University of Groningen

Population based glaucoma screeningStoutenbeek, Remco

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2010

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Stoutenbeek, R. (2010). Population based glaucoma screening Groningen: s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 21-05-2018

RIJKSUNIVERSITEIT GRONINGEN

Population based glaucoma screening

Proefschrift

ter verkrijging van het doctoraat in de Medische Wetenschappen

aan de Rijksuniversiteit Groningen op gezag van de

Rector Magnificus, dr. F. Zwarts, in het openbaar te verdedigen op

woensdag 6 januari 2010 om 16:15 uur

door

Remco Stoutenbeek

geboren op 28 maart 1978 te Deventer

Promotores : Prof. dr. J.M.M. Hooymans Prof. dr. P.T.V.M. de Jong Copromotor : Dr. N.M. Jansonius Beoordelingscommissie : Prof. dr. J.E.E. Keunen Prof. dr. J.B. Jonas Prof. dr. E. Buskens

Paranimfen : Dr. R.P.H.M. Müskens Drs. Y. Stoutenbeek

ISBN : 978-90-367-4123-1 Copyright © 2009 R. Stoutenbeek All rights reserved. No parts of this thesis may be reproduced, stored, or transmitted in any form by any means without prior written permission of the author. Publication of this thesis was financially supported by the Professor Mulder Stichting and the Rijksuniversiteit Groningen. All studies contained in this thesis were (partially) funded by the University Medical Center Groningen.

Stellingen

behorende bij het proefschrift

Population based glaucoma screening

1. Het adagium "voorkomen is beter dan genezen" staat bij de invoer van een screeningsprogramma juist ter discussie. (dit proefschrift)

2. Bevolkingsonderzoek naar glaucoom is op dit moment economisch niet

haalbaar in Nederland. (dit proefschrift) 3. Glaucoomscreening via opticiens lijkt een veelbelovend alternatief. (dit

proefschrift) 4. De participatiegraad van glaucoomscreening via de opticien is

potentieel hoger dan bij een bevolkingsonderzoek. (dit proefschrift) 5. De wet van de verminderde meeropbrengst is een economisch principe

met brede toepasbaarheid en geldt zeker voor glaucoomscreening. (dit proefschrift)

6. Lengthbias heeft een negatieve impact op de kosten-effectiviteit van

glaucoomscreening. (dit proefschrift) 7. Het aantal fout-positieven bij een screeningstest voor glaucoom is bijna

twee keer zo hoog per individu als per oog. (dit proefschrift) 8. Stilstand is vooruitgang bij de behandeling van glaucoom. 9. Bij invoering van nieuwe ICT projecten zoals het elektronisch voorschrijf

systeem en het elektronisch patiëntendossier kun je beter als laatste aan boord stappen.

10. Marktwerking in de door schaarste gekenmerkte gezondheidszorg

heeft niet geleid tot de beoogde kostenbeheersing, maar dreigt wel de zorg te verschralen.

11. You cannot depend on your eyes when your imagination is out of focus.

(Mark Twain / Samuel Langhorne Clemens; 1835 - 1910)

Table of contents 1. Introduction p. 09 2. Review of glaucoma literature relevant in respect to the

Wilson & Jungner criteria for appraising the validity of screening programmes:

p. 13

2.1 The condition being screened for should be an important health problem

p. 15

p. 22 2.2 The natural history of the condition should be well understood

2.3 There should be a detectable early stage p. 24 p. 26 2.4 Treatment at an early stage should be of more benefit than

at a later stage 2.5 A suitable test should be devised for the early stage p. 28

p. 33 2.6 The test should be acceptable 2.7 Intervals for repeating the test should be determined p. 34 2.8 Adequate health service provision should be made for the

extra clinical workload resulting from screening p. 36

2.9 The risks, both physical and psychological, should be less

than the benefits p. 39

p. 40 2.10 The costs should be balanced against the benefits

2.11 References p. 44

3. Frequency doubling perimetry screening mode compared to

the full-threshold mode. p. 55

p. 69 4. Strategies for improving the diagnostic specificity of the frequency doubling perimeter.

p. 83 5. Glaucoma screening during regular optician visits: can the population at risk of developing glaucoma be reached?

p. 95 6. The additional yield of a periodic screening programme for open-angle glaucoma - a population-based comparison of incident glaucoma cases detected in regular ophthalmic care with cases detected during screening.

p. 111 7. Supra-threshold perimetry compared to standard automated perimetry in glaucoma.

p. 125 8. General discussion

p. 126 8.1 Cost-effectiveness p. 129 8.2 Screening bias p. 130 8.3 Verdict on glaucoma screening p. 132 8.4 Alternatives p. 135 8.5 Suggestions for further research p. 137 8.6 References

p. 141 9. Summary

p. 145 10. Samenvatting (Summary in Dutch)

p. 151 11. Dankwoord (Acknowledgements)

p. 154 12. Curriculum Vitae

PUBLICATIONS Chapter 3 Stoutenbeek R, Heeg GP, Jansonius NM.

Frequency doubling perimetry screening mode compared to the full-threshold mode. Ophthalmic Physiol Opt. 2004;24:493-7.

Chapter 4 Heeg GP, Stoutenbeek R, Jansonius NM.

Strategies for improving the diagnostic specificity of the frequency doubling perimeter.

Acta Ophthalmol Scand. 2005 Feb;83(1):53-6. Chapter 5 Stoutenbeek R, Jansonius NM.

Glaucoma screening during regular optician visits: can the population at risk of developing glaucoma be reached? Br J Ophthalmol. 2006;90:1242-4. Epub 2006 Jul 19.

Chapter 6 Stoutenbeek R, de Voogd S, Wolfs RCW, Hofman A, de Jong

PTVM, Jansonius NM. The additional yield of a periodic screening programme for open-angle glaucoma: a population-based comparison of incident glaucoma cases detected in regular ophthalmic care with cases detected during screening. Br J Ophthalmol. 2008 Sep;92(9):1222-6.

Chapter 7 Stoutenbeek R, Hooymans JMM, Jansonius NM.

Supra-threshold perimetry compared to standard automated perimetry in glaucoma. [submitted]

9

Chapter 1 Introduction

Chapter 1

Screening is the systematic application of a test or inquiry, to identify individuals at sufficient risk of a specific disorder to benefit from further investigation or direct preventive action, among persons who have not sought medical attention on account of symptoms of that disorder.1 There are several ongoing population based screening programmes in the Netherlands, focused on early detection of breast cancer (mammography), cervical carcinoma (Papanicolaou test, “Pap smear”), and congenital metabolic diseases (heel prick or Guthrie test). Early detection and intervention is intuitively attractive, because it may avoid potentially serious consequences of a disease. However, population based screening requires a large amount of resources to be allocated to a specific health problem at the expense of alternative potentially beneficial uses. Therefore, it is important to investigate whether a health problem is suitable for screening prior to introducing a screening programme for it. This thesis explores whether population based glaucoma screening should be introduced. Glaucoma is an eye disease that causes glaucomatous optic neuropathy with subsequent visual field loss, which can eventually lead to irreversible blindness. It is the second leading cause of blindness, both worldwide and in Western Europe.2 Diagnosis is based on perimetry and evaluation of the optic disc and surrounding retinal nerve fibre layer. Intraocular pressure (IOP) is the most important risk factor for onset and progression of glaucoma; treatment consists of lowering of IOP.3-6 Glaucoma is an insidious disease: early stages do not cause any symptoms, and visual field loss often goes unnoticed initially. Patients become aware of their eye disease only after extensive damage has occurred. Therefore, screening is often noted as an option to reduce the glaucoma burden. At present, a glaucoma screening programme does not exist in the Netherlands. However, opportunistic case finding by ophthalmologists, optometrists, and opticians has become common practice. Nevertheless, only about half of all people with glaucoma are known (i.e. detected as having glaucoma, and visiting an ophthalmologists regularly).7-9 Mass screening would be required to find the remaining half. The most common form of glaucoma in the Netherlands is primary open-angle glaucoma (POAG). This type of glaucoma is characterized by an open angle on gonioscopy combined with the absence of any identifiable ocular or systemic abnormalities that would lead to increased IOP. For easier readability, glaucoma will be used as a synonym for POAG throughout this thesis, unless otherwise specified. Outline of the thesis Chapter two (literature review) provides an overview of available relevant literature regarding different aspects of screening. The five studies described in

10

Introduction

11

chapters three through seven are aimed at lacunae in the existing literature, as identified in chapter two. In chapters three and four, the efficacy of frequency doubling perimetry (FDT) as a screening test for glaucoma is evaluated. Chapter five assesses whether the majority of the population at risk for glaucoma could be reached by optician based screening, which would attain a substantial cost reduction compared to a normal screening programme. Chapter six evaluates the additional yield of population based screening compared to opportunistic case finding. Chapter seven evaluates supra-threshold perimetry as a screening test for glaucoma, but also serves to interpret the results of chapter six more accurately. Whether or not a population based screening programme should be introduced in the Netherlands is discussed in chapter eight (general discussion).

Chapter 2 Literature review

13

Chapter 2

Whether glaucoma is suitable for population based screening depends on many factors. In 1968, at the World Health Organization conference in Geneva, Wilson and Jungner proposed a set of criteria for assessing whether screening for a particular disease is worthwhile.10 In the course of years, some of the criteria were modified, and - predominantly because of advancements in genetics - additional criteria were appended. Genetic screening for glaucoma is not yet at issue, because genetic mutations known to cause primary open angle glaucoma account for only a small minority of glaucoma cases.11 Therefore, the original criteria are still valid. They will be applied to glaucoma screening, and will serve as a scaffold for the literature review of this thesis. The Wilson & Jungner criteria for appraising the validity of a screening programme are:

1. The condition being screened for should be an important health problem

2. The natural history of the condition should be well understood 3. There should be a detectable early stage 4. Treatment at an early stage should be of more benefit than at a later

stage 5. A suitable test should be devised for the early stage 6. The test should be acceptable 7. Intervals for repeating the test should be determined 8. Adequate health service provision should be made for the extra clinical

workload resulting from screening 9. The risks, both physical and psychological, should be less than the

benefits 10. The costs should be balanced against the benefits

14

Literature review

2.1 The condition being screened for should be an important health problem

Glaucoma is the second leading cause of blindness worldwide; second to cataract in developing countries and second to Macular Degeneration in developed countries. World Health Organization (WHO) definitions for blindness and low vision are shown in table 2.1.1. Visual impairment includes both blindness and low vision. Table 2.1.1 Blindness, low vision, and visual impairment definitions.

BCVA in better eye VF in better eye

max min max min

vis. impaired* <6/18 lp – and/or <20° 0°

low vision* <6/18 1/20 and/or <20° 10°

blindness* <1/20 lp – and/or <10° 0°

blindness# ≤2/20 lp – and/or ≤20° 0° BCVA = Best Corrected Visual Acuity; VF = Visual Field (widest radius with Goldmann III-4 stimulus, measuring from the fixation point); lp – = no light perception * = as defined in the 10th Revision of the WHO International Statistical Classification of Diseases, Injuries and Causes of Death (ICD-10) # = as defined by United States federal legislation Table 2.1.2 shows an estimate of global blindness and visual impairment in 2002 per WHO region. Data were derived from a WHO paper by Resnikoff.2 The number of blind people worldwide was estimated to be 36,9 million. This corresponds to a prevalence of 0.6%. In Western Europe, the prevalence of blindness is 0.2%.

15

Chapter 2

Table 2.1.2 Magnitude of visual impairment worldwide.2

Tota

l

6214

161.

2

124.

3

36.9

100%

100%

Wes

tern

Pa

cific

Re

gion

1718

41.8

32.5

9.3

28%

25%

Sout

h-Ea

st

Asi

a Re

gion

1591

45.1

33.5

11.6

26%

32%

East

ern

Med

iterr

Re

gion

503

16.5

12.4

4.0

8%

11%

Afr

ican

Re

gion

672

26.8

20.0

6.8

11%

18%

Regi

on o

f th

e A

mer

icas

853

15.5

13.1

2.4

14%

7%

East

Eu

rope

an

Regi

on

463

9.1

7.4

1.8

7%

5%

Wes

t Eu

rope

an

Regi

on

415

6.4

5.4

0.9

7%

2%

popu

latio

n (x

mill

ion)

vis.

impa

irmen

t (x

mill

ion)

low

vis

ion

(x m

illio

n)

blin

d (x

mill

ion)

% o

f tot

al

popu

latio

n

% o

f tot

al

blin

dnes

s

16

Literature review

Causes of blindness are shown in figure 2.1.1a (worldwide) en figure 2.1.1b (Western Europe). Blindness due to glaucoma affects 4.5 million people worldwide (12.3% of 36.9 mln) and 170,000 Western Europeans (18% of 0.9 mln).2

glauc 12.3% others

#

*

drp

cataractarmd

Figure 2.1.1a Causes of blindness in 2002 worldwide.2 # = trachoma; * = corneal opacity; drp = diabetic retinopathy; armd = age related macular degeneration; glauc = glaucoma

17

Chapter 2

others glauc

18% *

drp cat

armd

Figure 2.1.1b Causes of blindness in 2002 for Western Europe.2 * = corneal opacity; drp = diabetic retinopathy; armd = age related macular degeneration; cat = cataract; glauc = glaucoma Quigley estimated12 that there would be 60.5 million people with primary glaucoma worldwide in 2010: 44.7 million with open-angle glaucoma (OAG) and 15.7 million with angle-closure glaucoma (ACG). Glaucoma in the European derived region (defined as Europe including the Russian Federation and Ukraine, United States, Bermuda, Canada, Greenland, Australia, New Zealand, and Israel) was estimated to affect 12.1 million people: 10.7 million with OAG and 1.4 million with ACG. The number of people blind from glaucoma worldwide in 2010 was estimated to be 8.4 million: 4.5 from OAG and 3.9 from ACG. This estimate by Quigley, although of somewhat comparable magnitude, is higher than the WHO estimate (4.5 million people blind from glaucoma in 2002, 12.3% of 36.9 million, see above). The two estimates differ, partly because of the increasing world population from 2002 to 2010, but mainly because of methodological issues. Blindness prevalence surveys often assign the most “treatable” disease as the primary cause of blindness. Cataract is considered more treatable than glaucoma, which leads to an underestimation of glaucoma blindness in developing countries in the WHO survey. Projections for 2020 from the same article expect an increase in glaucoma burden of 19.1 million people in one decade, to a total of 79.6 million people with glaucoma worldwide. Glaucoma related blindness will by then affect

18

Literature review

approximately 11.1 million people. The increase in glaucoma burden is the result of predicted population growth and longer life expectancy. Both Resnikoffs and Quigleys estimates are based on glaucoma prevalence and incidence studies. Key characteristics of important epidemiological studies on POAG prevalence and incidence are summarized in table 2.1.3 and 2.1.4 respectively (presented below figure 2.1.2). Reported prevalence rates vary between 0.8% and 8.8%. The pooled prevalence rate from these and several additional minor studies is 2.1%.8 Reported 5-year incidence rates vary between 1.1% and 3.1%. Discrepancies between studies in reported prevalence and incidence rates are predominantly due to differences in ethnicity, age, and glaucoma definition used. The latter is illustrated in figure 2.1.2 (see below). The graph shows the variation in prevalence of glaucoma by age, when glaucoma definitions from other population based studies are applied to data of the Rotterdam study.7

Figure 2.1.2 Variation in prevalence of open angle glaucoma when different criteria for glaucoma are applied to data of the Rotterdam study.7 Reproduction, courtesy of Dr. R.C.W. Wolfs.

19

Chapter 2

Table 2.1.3 POAG prevalence.

study period ethnicity age (years)

N prevalence

USA, Framingham13 1973-1975 Caucasian 52-85 2433 1.4%

Italy, Ponza14 1986 Caucasian ≥40 1034 2.5%

USA, Wisconsin, Beaver Dam15

1988-1990 Caucasian 43-84 4926 2.1%

Australia, Blue Mountains16 1992-1994 Caucasian ≥49 3654 3.0%

Italy, Egna-Neumarkt17 1995 Caucasian ≥40 4297 1.4%

The Netherlands, Rotterdam7

1990-1993 Caucasian ≥55 6281 0.8%

USA, Baltimore18 1985-1988 Caucasian ≥40 2913 1.1%

USA, Baltimore18 1985-1988 African ≥40 2395 4.2%

Caribbean, West Indies19

<1989 African-

Caribbean ≥30 1679 8.8%

Barbados20 1988-1992 African-

Caribbean 40-84 4631 6.7%

Japan, Yokohama21 2000 Asian 6-98 64394 1.2%

20

Literature review

Table 2.1.4 POAG incidence.

study period ethnicity age (years)

N incidence (5 year risk)

Sweden, Dalby22 1977-1986 Caucasian ≥55 1295 1.2%

Sweden, Tierp23 1984/86-1988/91 Caucasian 65-74 413 3.1%

Barbados24 1988/92-1992/97

African ≥40 3427 2.8%

Australia, Melbourne25

1992/94-1997/99 Caucasian ≥40 3257 1.1%

The Netherlands, Rotterdam26

1990/93-1997/99 Caucasian ≥55 3842 1.8%

21

Chapter 2

2.2 The natural history of the condition should be well understood The term “glaucoma” comes from “glaukosis”, which was first described by Hippocrates in his book Concerning Vision of the Corpus Hippocraticum around 400 BC. Glaukosis may have been derived from the Greek word “glaukos”, which means greenish glaze. It probably represented cataract instead of glaucoma.27 POAG is currently defined by the European Glaucoma Society as a chronic progressive optic neuropathy with characteristic morphological changes at the optic nerve head and retinal nerve fibre layer, in the absence of other ocular disease or congenital anomalies. Progressive retinal ganglion cell death and VF loss are associated with these changes.28 The relative risk for POAG rises continuously with the level of IOP. Elevated IOP results from increased resistance to aqueous outflow at the level of the trabecular meshwork.29 Individuals with a statistically normal IOP can also develop POAG, which is then termed Normal Tension Glaucoma, and is not uncommon.30 The pathogenesis of glaucoma remains unclear. Apoptosis of retinal ganglion cells with concurrent demise of their axons is the established final common pathway of glaucomatous optic neuropathy.31 The most likely anatomical location where damage to the optical pathway must take place, can be deduced from knowledge of the progression patterns of VF loss. Scotomas are unlikely to cross the horizontal meridian when their size increases due to glaucoma progression.32 Only at the level of the optic disk can this finding be accounted for, so this is where damage must occur. 33 Morphologic disk changes such as cupping and bowing of the lamina cribrosa support this finding. The underlying mechanisms that take place at the optic disc and lead to apoptosis of ganglion cells, are subject of debate however, and probably multi-factorial. Two popular theories are: mechanical compression of axons at the level of the lamina cribrosa, hampering axonal transport;34;35 and ischemia of the optic nerve head as a result of deficient blood flow autoregulation.31;36 Even though the pathogenesis of glaucoma has not been elucidated, there is a good understanding of its natural history. Important studies that contributed most to the understanding of untreated glaucoma are the Early Manifest Glaucoma Trial (EMGT)3;4 and a study by Wilson et al.37 The EMGT is a clinical trial that evaluated the effectiveness of IOP reduction by randomising newly detected, previously untreated glaucoma patients to either treatment with trabeculoplasty and betaxolol eye drops (n=129) or no treatment (n=126).3;4 An average IOP reduction of 5.1 mmHg was achieved, which constituted a 25% reduction from baseline. The median follow-up period was 6 years (range 51-102 months). Glaucoma progression was considered present when either the visual field worsened, or optic disc cupping increased. Progression was less frequent in the treatment group (58/129=45%) than in the

22

Literature review

non-treatment group (78/126=62%; P=0.002). Interestingly, they also calculated the Mean Deviation (MD) change in dB per month, which was found to be -0.05 dB ±0.07 dB (± Standard Deviation, SD) in the untreated group. These figures can be used to estimate the long term outcome of untreated glaucoma, when we assume that the deterioration is linear. Over the course of 10 years, the MD will progress on average -6 dB. In the very worst case scenario however, reflected by adding twice the SD, the MD progression becomes -22.8 dB in ten years. Since baseline MD in the non-treatment group was -4.4 dB ±3.3 SD, these worst case scenario patients develop end-stage glaucoma in ten years. Moreover, the exclusion of high tension cases (mean IOP >30 mmHg or any IOP >35 mmHg), might have led to an underestimation of the actual progression rate. In 2002, the same year that the EMGT results were published, a study by Wilson et al37 appeared that presented a more direct view of the natural course of glaucoma. Their study took place in St. Lucia, West Indies, where a national survey had been conducted from 1986 to 1987 regarding the prevalence of primary open-angle glaucoma among 1,679 Afro-Caribbean individuals. 147 subjects (8.8%) had been diagnosed with glaucoma, and 217 subjects were diagnosed as glaucoma suspects. Those with severe glaucoma received surgery and those requiring glaucoma medications received them at no charge. However, in 1988 the infrastructure for subsidized glaucoma care changed, so unfortunately only a few cases were actually treated. This was discovered ten years later when Wilson conducted a follow-up study in 1997. It became a unique opportunity to study the natural course of untreated glaucoma. Of the 364 subjects at baseline, 159 were lost to follow up. Of the remaining 205 patients, 59 right and 64 left eyes were excluded. Based on the Advanced Glaucoma Intervention Study (AGIS)38 criteria, 55% of 146 right eyes and 52% of 141 left eyes showed progression of visual field loss. The cumulative probability of reaching end-stage glaucoma (AGIS score ≥18) in ten years in at least one eye was 16%; for bilateral end-stage glaucoma the cumulative probability was 7%. Various glaucoma risk factors have been identified, mainly through population based studies. Important risk factors are: intraocular pressure, 7;26;39 age,7 family history of glaucoma,40 black ethnicity,18 central corneal thickness41, and myopia42.

23

Chapter 2

2.3 There should be a detectable early stage Screening is aimed at detecting a disease before symptoms arise. The purpose is to treat the disease in an earlier stage, which is often beneficial: easier treatment, better prognosis, less health damage. Depicted in figure 2.3.1 is a schematic representation of an individual who becomes ill at a certain point in life.

screening

P Q R

birth

Figure 2.3.1 Disease phases. Adapted from Vandenbroucke JP et al.43 (N)D-PCP = (non)detectable preclinical phase Point P marks the biological onset of disease, but since the disorder is initially very small, detection is not yet possible. From point Q onwards, screening tests are able to detect abnormalities. From point R onwards, the patient becomes symptomatic (manifest disease). Between biological onset and manifest disease lies the preclinical phase. This preclinical phase can be subdivided in a non-detectable preclinical phase (ND-PCP) and a detectable preclinical phase (D-PCP). Cases detected through a screening programme are on average halfway between point Q and point R. Therefore, the gain in time that results from detection through screening is defined as half of the D-PCP, and is known as the “lead time”. As a consequence, the longer the D-PCP is, the longer the lead time, and the more suitable a disease becomes for screening. Glaucoma is characterised by its insidious onset and long D-PCP. Visual field loss goes unnoticed by most individuals, because scotomas are compensated for by the fellow eye, or are being filled in by the brain. Visual acuity typically remains intact until late in the disease process, when extensive optic nerve damage has occurred. In the Netherlands, many cases are presently discovered through case finding by opticians, optometrists, and ophthalmologists, and not as a result of glaucoma related symptoms. Even with case finding, only about half of all people

death

preclinical phase (PCP) manifest disease

ND-PCP D-PCP

24

Literature review

with glaucoma are known (i.e. detected as having glaucoma, and visiting an ophthalmologists regularly).7-9 Quigley et al performed a questionnaire study in 2003 among 308 glaucoma patients (362 eligible, response rate 85%) to determine how glaucoma patients are identified.44 Only 25% of the study subjects reported having eye symptoms at their initial ophthalmologist visit. Moreover, in at least three-quarters of these cases, the reported eye symptoms seemed unrelated to glaucoma.

25

Chapter 2

2.4 Treatment at an early stage should be of more benefit than at a later stage

Glaucomatous damage to the optic nerve is irreversible. Treatment (which consists of lowering of IOP) can slow down the progressive loss of nerve fibres, but can not stop it. In advanced glaucoma, much effort needs to be made to keep the rate of progression to an absolute minimum. A lower target IOP becomes necessary45, which requires stricter monitoring (more frequent visits), more medications, and more often glaucoma surgery. It is not always possible to prevent visual impairment and blindness. Early glaucoma on the other hand permits a more lenient approach regarding both follow-up and treatment, since patients are unlikely to become visually impaired. In the 1990s, two prospective randomised controlled trials were conducted: the Early Manifest Glaucoma Trial (EMGT)4 and the Ocular Hypertension Treatment Study (OHTS)5. Both studies prove that early treatment is beneficial (see below). The EMGT evaluated whether lowering of IOP decreased the progression of glaucomatous damage in early glaucoma patients. The EMGT is discussed in paragraph 2.2. In summary, progression was less frequent in the treatment group (45%) than in the non-treatment group (62%; P=0.002).3 The OHTS evaluated whether lowering of IOP decreased the conversion rate from ocular hypertension (OHT) to glaucoma in subjects with OHT. Participants with no evidence of glaucomatous damage, and with an IOP between 24 mm Hg and 32 mm Hg in one eye and between 21 mm Hg and 32 mm Hg in the other eye, were randomized to either treatment with topical ocular hypotensive medication (n=702) or observation (n=706). IOP reduction was 22.5% in the medication group versus 4.0% in the observation group. Conversion to glaucoma was considered present when either reproducible visual field abnormalities developed or reproducible optic disc deterioration was found. At 60 months follow-up, the cumulative probability of developing POAG was 4.4% in the medication group and 9.5% in the observation group (hazard ratio 0.40; 95% confidence interval 0.27-0.59; P<.0001).6 Several other influential prospective studies also evaluated the effect of IOP reduction on glaucoma progression, but these are not specifically aimed at early glaucoma, the focus of this paragraph. The Collaborative Normal-Tension Glaucoma Study (CNTGS) evaluated 140 patients with normal tension glaucoma. Patients were randomized to either treatment (n=61) or observation (n=79). Glaucoma progression was less frequent in the treatment group (12%) than in the non-treatment group (35%; P=0.002).46 The Advanced Glaucoma Intervention Study (AGIS)38 was designed to compare two surgical treatment strategies for patients with open-angle glaucoma that could no longer be adequately controlled by medications alone. A large

26

Literature review

proportion of the 738 participants had moderate or advanced glaucoma. Data was not only used to decide which surgical treatment strategy was more favourable, but also to examine the relationship between intraocular pressure and progression of visual field damage. For the latter purpose, participants were stratified into subgroups based on their IOP. After seven years of follow-up, eyes with lower IOP showed less glaucomatous visual field progression than eyes with higher IOP (P<0.001).45 In a meta-analysis by Maier et al.,39 data from five OHT treatment studies and two glaucoma treatment studies (the EMGT and the CNTGS) were pooled. Treatment had a significant protective effect on conversion from OHT to glaucoma, as well as on progression of established glaucoma. The hazard ratio for OHT treatment was 0.56 (95% confidence interval 0.39 to 0.81, P=0.01) and for glaucoma treatment 0.65 (95% confidence interval 0.49 to 0.87, P=0.003). Hattenhauer et al. evaluated the probability of blindness from treated OAG and treated OHT in a retrospective, community-based, longitudinal study in Minnesota.47 The United States federal legislation definition of blindness was used (see table 2.1.1). Newly diagnosed OAG and OHT cases identified between 1965 and 1980 were included in the study. The mean follow-up time was 15 years. Kaplan–Meier cumulative probability of bilateral blindness at 20 years was estimated to be 4% for treated OHT patients (n=177), and 22% for treated OAG patients (n=118). Obviously, these rates do not apply to modern glaucoma care, since in 1965 the standards for glaucoma care were different, and options for lowering IOP with medication were limited. Also, the definition of blindness as established by US legislation is fairly liberal, particularly with respect to the visual field criterion (see last row to table 2.1.1), which also contributes to the high rate of blindness. Nevertheless, the large difference in blindness rates between OAG and OHT patients is an indication that treatment at an early stage is preferable to treatment at a later stage. Chen conducted a similar study48, using the same definition for blindness. Patients undergoing glaucoma treatment in 2000 at the Eye Clinic of the University of Washington Medical Center that were newly diagnosed between 1975 and 1998 were eligible. 186 patients were included, their mean follow-up time was 10.2 years. The Kaplan-Meier cumulative probability of bilateral blindness at 15 years was estimated to be 6.4%. The amount of baseline visual field (VF) loss was identified as the most important risk factor for blindness

�(relative risk 1.20/dB deterioration in baseline MD (P<0.001).

27

Chapter 2

2.5 A suitable test should be devised for the early stage The ideal mass screening test should be safe, cheap, fast, easy to perform by both examiner and examinee, and have a high specificity and reasonable sensitivity. Specificity is more important than sensitivity in the setting of mass screening, since it is important to minimise the number of false positive test results. Some sensitivity can be sacrificed in order to maintain high specificity, provided that advanced cases are not missed and can still be treated adequately. Diagnostic testing for glaucoma consists of two modalities: perimetry and imaging. In perimetry, the visual field (VF) is evaluated. Glaucoma is the leading cause of VF loss, but there are many non-glaucomatous causes.49 Perimetry is a subjective test of function, and thus requires cooperation and concentration from the test subject. In glaucoma imaging, the structure of the optic nerve head or its surrounding nerve fibre layer is evaluated. Imaging is an objective test of anatomic proportions, which requires less cooperation, but it yields an indirect measure of function. Anatomical variations of the optic disc, as seen in tilted or myopic discs, present problems for imaging devices, as well as the presence of peripapillary atrophy in otherwise normal discs. Perimetry and imaging are discussed below in paragraph 2.5.1 and 2.5.2. A third category of diagnostic tests is measurement of the IOP. The gold standard for measuring IOP is Goldmann applanation tonometry; most opticians perform non-contact tonometry. Elevated IOP is a risk factor for glaucoma, but does not provide direct information about the presence or absence of glaucoma. A characteristic of glaucoma is its increased fluctuation in IOP throughout the day, hence glaucoma patients with on average elevated IOP may sometimes have normal readings. Moreover, a not insignificant part of glaucoma patients has statistically normal IOP continuously (so called normal tension glaucoma – NTG). As a consequence of these problems, only about half of the glaucoma patients have elevated IOP at a single visit.50 On the whole, tonometry is unsuitable for population based screening, although it most definitely remains useful for case finding, as demonstrated by Peeters et al.51 2.5.1 Perimetry Standard Automated Perimetry (SAP) has become the most commonly used form of perimetry for glaucoma diagnosis and follow-up. Commonly used SAP devices are the Humphrey Field Analyzer (HFA, Carl Zeiss Meditec Inc., Dublin, CA, USA) and the Octopus perimeter (Haag-Streit AG, Koeniz, Switzerland), among others. In SAP, the threshold luminance of every test location in the visual field is determined by a staircase procedure: the stimulus luminance is either increased

28

Literature review

or decreased stepwise until a change occurs in the response of the subject.52;53 This detailed VF information allows for careful monitoring of glaucoma patients during follow-up. For screening purposes however, it is only necessary to differentiate a normal VF from a glaucomatous VF; follow-up is not at issue. Fewer requirements allow for a simpler test. The frequency doubling perimeter (FDT - Carl Zeiss Meditec, Dublin, California, USA; Welch-Allyn, Skaneateles, New York, USA) is particularly suitable for screening. It is an automated static perimeter that uses only 17 test locations in a 24° VF, compared to 52 and 76 test locations for the HFA24-2 and 30-2 respectively. In C20-1 screening mode, the FDT uses a supra-threshold instead of a full threshold algorithm, designed for high specificity. Testing times are typically 40 seconds per eye in case of no VF loss, which can increase to 120 seconds in case of extensive VF loss. Frequency doubling perimetry uses a special stimulus type, specifically a sinusoidal spatial waveform composed of alternative white-and-black stripes that counter-phase flicker at 25 Hz. This creates an optical illusion: the target is perceived to have twice its actual spatial frequency.54-56 Several studies reported on the diagnostic performance of the FDT C20-1. Some of them are shown in table 2.5.1. Classification (glaucoma or no glaucoma) was done on a by-patient basis in the study by Stoutenbeek et al.57 (see chapter three) and on a by-eye basis in the other studies. This difference in methodology explains the higher sensitivity and somewhat lower specificity reported by Stoutenbeek et al., since by-patient testing gives each individual two chances (two eyes) to be classified as abnormal. Table 2.5.1 Diagnostic performance of the Frequency Doubling perimeter in C20-1 screening mode.

study period N sensitivity specif.

glauc. normal early moderat severe

Trible et al.58 2000 125 95 39% 86% 100% 95%

Stouten- beek et al.57 2004 100 108 74% 100% 100% 88%

Ferreras et al.59 2007 110 92 32% 65% 100% 100%

Iwase et al.60 2007 171 5295 41% 88% 93%

early = early glaucoma, defined as HFA mean deviation (MD) > -6 dB; moderate = -6 dB ≥ MD > -12 dB; severe = MD ≤ -12 dB

29

Chapter 2

As can be seen in table 2.5.1, FDT C20-1 has a high specificity. Sensitivity for early glaucoma is limited, however no cases with severe glaucoma (defined as HFA mean deviation ≤ -12 dB) were missed. FDT C20-1 perimetry is therefore appropriate as a diagnostic test for glaucoma screening. Learning effects for FDT in perimetric novices are very small in comparison to test-retest variability and compared to standard automated perimetry (i.e. HFA perimetry).61 2.5.2 Imaging Currently, there are three principal imaging modalities used for analysing the optic nerve head (ONH) and the surrounding retinal nerve fibre layer (RNFL). [1]. The confocal scanning laser ophthalmoscopy (CSLO) uses a laser beam to scan the three dimensional shape of the ONH. [2]. Optical coherence tomography (OCT) measures the thickness of the RNFL based on the coherence of reflected laser light, similar to the physical principals of ultrasound. [3]. Scanning laser polarimetry (SLP) measures the thickness of the RNFL by assessing the retardation that results from the birefringent properties of ganglion cell axons. See table 2.5.2 for an overview. Table 2.5.2 Glaucoma imaging devices.

device modality target dilation

Heidelberg Retina Tomograph (HRT) CSLO ONH no

Stratus Optical Coherence Tomography (OCT) OCT RNFL yes

GDx Nerve Fiber Analyzer (GDx) SLP RNFL no

dilation = pupil dilation required; CSLO = confocal scanning laser ophthalmoscopy; SLP = scanning laser polarimetry; ONH = optic nerve head; RNFL = retinal nerve fibre layer All three devices have similar diagnostic performance.62-65 OCT requires pupil dilation, and is for that reason less interesting for screening purposes (though the latest generation OCT may be able to operate through pupil sizes as small as 3mm). Diagnostic performance of the HRT compared to the GDx is shown in table 2.5.3. Each study determined sensitivity and specificity rates for both the HRT and the GDx within a single group of subjects.

30

Literature review

Table 2.5.3 Diagnostic performance of the Heidelberg Retina Tomograph (HRT) compared to the GDx Nerve Fiber Analyzer (GDx).

study period N sensitivity specif.

glauc. normal HRT GDx both

Medeiros et al.62 2004 107 76 59% 61% 95%

Pueyo et al.63 2007 73 66 85% 48% 95%

Badala et al.65 2007 46 46 70% 78% 95%

Reus et al.66 2007 48 40 85% 92% 95%

Medeiros et al.67 2008 40 42 35% 50% 95%

At 95% specificity, reported sensitivity rates for HRT and GDx vary considerably from 35% to 85% and from 48% to 92% respectively. This is primarily due to differences in average glaucoma severity of enrolled patients. Since structural damage to the optic nerve generally precedes VF loss,6;68-70 most imaging studies focussed on early glaucoma and glaucoma suspects, yielding lower sensitivity. Only a few studies contain stratified data for moderate and severe glaucoma cases, see table 2.5.4. The lower specificity as found by Heeg et al. results from classification on a by-patient basis, analogous to the FDT perimetry study by Stoutenbeek et al. discussed earlier in this paragraph (see paragraph 2.5.1).

31

Chapter 2

Table 2.5.4 Sensitivity to moderate and severe glaucoma for the Heidelberg Retina Tomograph (HRT) and the GDx Nerve Fiber Analyzer (GDx).

study period device N sensitivity specif.

moderat severe moderat severe

Ferreras et al.71 2007 HRT 32 21 94% * 95% * 86%

Reus et al.72 2004 GDx 28 81 93% * 100% # 96%

Heeg et al.73 2005 GDx 329 95% # 78%

* severity of glaucoma rated according to Hodapp–Parrish–Anderson score74 # severity of glaucoma rated conform legend to table 2.5.1 Grading of glaucoma severity based on the Hodapp–Parrish–Anderson score74 (see table 2.5.4) is more complex and more strict than grading based on HFA mean deviation as described in the legend to table 2.5.1. This needs to be taken into account when comparing table 2.5.4 (imaging) to 2.5.1 (perimetry). Both the HRT and GDx seem appropriate diagnostic tests for glaucoma screening, and are on par with perimetry (FDT). The upper limit in test specificity for all glaucoma screening tests (perimetry and imaging) is probably at 95%; higher specificities are currently not feasible since sensitivity for severe glaucoma would fall. These conclusions are in concordance with a recently published meta-analysis on glaucoma screening tests by the United Kingdom Health Technology Assessment programme.75 It is important to keep in mind that the stated specificity of 95% is applicable to the testing of one eye only; specificity will be lower if both eyes are examined. See paragraph 8.1 regarding by-eye versus by-patient specificity for further discussion.

32

Literature review

2.6 The test should be acceptable Glaucoma tests mentioned in paragraph 2.5 (FDT, HRT, GDx) are all non-invasive. No ionizing radiation is used for imaging, and pupil dilation is not necessary (excluding OCT). Testing is not painful, awkward, or time-consuming. Subjects are required to concentrate on their visual task (for FDT), or keep their eyes steady on a point of focus (for imaging). Testing takes less than three minutes per eye. Measurement of intra ocular pressure (IOP) can be performed by non-contact tonometry, which uses a puff of air to applanate the cornea. Topical anaesthesia is not required (as opposed to Goldmann applanation tonometry), and since corneal touch does not occur, chances of microbiological and prion transmission are minimal. Testing takes less than a minute per eye. These glaucoma screening tests place far less of a burden on screened subjects than existing population based screening programmes in the Netherlands (mammography, heel prick test, and Pap smear – see introduction to chapter two).

33

Chapter 2

2.7 Intervals for repeating the test should be determined The optimal time period between two consecutive screening tests is determined by a balance of two opposing main interests. To minimize the chances that an individual reaches end-stage glaucoma between screening visits, the interval should be kept short. However, the length of the inter-test interval has considerable impact on the costs of a screening programme, since it directly determines how many people need to be screened each year. A larger screening interval is more likely to be economically feasible. An acceptable compromise between these conflicting interests depends to some extent on the society’s willingness to pay: how much money is made available for avoiding glaucoma induced blindness. Nevertheless, data on rate of progression of glaucoma can be used to estimate a reasonable screening interval. In the EMGT study, the progression rate of untreated glaucoma was analysed (see paragraph 2.2).3;4 Extrapolating from their results and assuming linear progression, the worst case scenario (97.5th percentile) patient progresses -22.8 dB in ten years, which is equivalent to end-stage glaucoma. A ten year screening interval thus seems unfavourable, even if none of the participants would have any baseline loss. In reality, baseline loss will not be uncommon, because in every screening round quite a few early glaucoma cases will have been missed, the so called false negatives (early glaucoma is defined as HFA MD better than -6 dB; see legend to table 2.5.1). This results from the need for high test specificity in a screening setting (see paragraph 2.8), which implies that some test sensitivity for early glaucoma must be sacrificed to attain high overall specificity. Missed cases in the prior screening round start their successive screening interval with baseline loss. Taking baseline loss into account and choosing a five year screening interval (arbitrarily chosen), the worst case scenario patient would start at -6 dB and progress -11.4 dB to end up at -17.4 dB at the successive screening test. This seems just acceptable. Other studies also provide empirical data concerning glaucoma progression over a period of approximately five years. In the Groningen Longitudinal Glaucoma Study (GLGS), a cohort of patients at risk for developing glaucoma (patients with ocular hypertension or a family history of glaucoma, but without visual field abnormalities at baseline) was followed prospectively in a clinical setting for 4 years with SAP, FDT, and GDx.69;73;76 Of 174 glaucoma suspect patients, 26 had developed reproducible glaucomatous visual field loss on SAP. At the end of follow-up, the HFA MD of the worse eye of the 26 incident glaucoma cases ranged from -7.6 to -0.8 dB, with an average value of -3.6 dB. The Rotterdam Study is a prospective, population-based cohort study of residents aged 55 and older living in a district of Rotterdam.7;26 Participants were examined at baseline and follow-up, 87 of 3842 participants developed incident glaucoma during a

34

Literature review

mean follow-up time of 6.5 years. 78 of 87 incident glaucoma cases were available for analyses, and 45 of these 78 cases had visual field loss on supra-threshold visual field testing (see chapter six; in the remaining 33 incident glaucoma cases, diagnosis was based on progressive optic disc cupping).77 At the 75th percentile, glaucoma progression over 6.5 years was –2.8 dB in the better eye and –9.3 dB in the worse eye. Missed points on Supra-Threshold Perimetry (STP) have been converted into HFA MD scores according to the method described in chapter seven. Both the EMGT, the GLGS, and the Rotterdam Study yield comparable results: during a five year interval, considerable glaucoma progression can take place in unfavourable cases, but even when baseline loss was also present, the resulting amount of glaucomatous damage will fall short of end-stage glaucoma and is amenable to treatment. Recent cost-effectiveness studies also use a five year screening interval for modelling of a screening programme78;79 (see paragraph 2.10).

35

Chapter 2

2.8 Adequate health service provision should be made for the extra clinical workload resulting from screening

Implementation and upkeep of a population based screening programme requires a great logistical effort and consumes considerable amounts of resources. However, screening is only the first step towards prevention of morbidity. All subjects with a positive screening test will need further evaluation. Some of them will prove to have glaucoma (the true positives) and require follow-up and treatment. Borderline cases will also need follow-up. Unconfirmed cases (the false positives) pose a distinct problem: they may either return to periodic screening or quite the reverse, see paragraph 8.5 (general discussion). For the estimates given below, false positive cases were assumed to return to periodic screening or even opportunistic case finding and to not require ongoing follow-up, as this approach seems most likely to be economically feasible. Effects of accumulation of normal subjects that repeatedly fail their screening test every screening round are ignored since estimates are performed for a single 5 year screening round only. The clinical workload that results from true and false positive screening tests, depends on the prevalence of undetected glaucoma, and on the sensitivity and specificity of the screening test. Prevalence of glaucoma in the Dutch 40+ population is about 2% (see table 2.1.3 – POAG prevalence). At least half of all glaucoma patients are undetected,7-9 so the prevalence of undetected glaucoma is approximately 1%. Since this is a relatively low percentage, the specificity of the screening test will be of far greater importance with respect to ensuing clinical workload than the sensitivity. Based on data from table 2.5.1, 2.5.3, and 2.5.4, sensitivity and specificity for a glaucoma screening test are arbitrarily determined as 75% and 95% respectively. Table 2.8.1 shows the corresponding estimated number of true and false positives.

36

Literature review

Table 2.8.1 Population of the Netherlands for different age groups; rough estimates of clinical workload that results from glaucoma screening.

[ estimated figures ]

age inhabitants (x million)

known glau- coma patients

true positives

false positives

all 16.4

40+ 8.2 66 000 49 000 404 000

45+ 6.9 63 000 47 000 339 000

50+ 5.6 56 000 42 000 277 000

inhabitants = number of Dutch inhabitants on 01-01-2008 according to Statistics Netherlands;80 known glaucoma patients = number of inhabitants with known glaucoma, assuming a prevalence of

1% for age over 50 & 0.5% for age 45 to 50 & 0.25% for age 40 to 45 years; true positives = number of inhabitants with unknown glaucoma that would be discovered by

screening, assuming a test sensitivity of 75% and a prevalence of unknown glaucoma of 1% for age over 50 & 0.5% for age 45 to 50 & 0.25% for age 40 to 45 years;

false positives = number of inhabitants without glaucoma that would have a false positive screening test result, assuming a test specificity of 95% As can be seen in table 2.8.1, the lower age limit of the population to be screened has a profound influence on the number of false positives. Moreover, glaucoma prevalence is very low for subjects aged 40 to 50 years, so screening would yield few cases in this stratum. This is illustrated in figure 2.1.2, where the prevalence of glaucoma in the Rotterdam Study7 drops below 0.1% for subjects under age 55 for most glaucoma definitions. On the other hand, young undetected glaucoma patients run the highest risk of becoming blind, and will on average spend more years being blind, because they have the longest life expectancy. A reasonable compromise is to start glaucoma screening at age 50. The estimated number of true and false positive test subjects would then be 319,000, if everyone would attend for screening. Participation rates for breast cancer screening vary considerably, but 70% is regarded to be achievable.81 When we apply this rate to glaucoma screening, we can expect 223,000 subjects per screening round referred for further evaluation. For a test interval of five years (see paragraph 2.7), this corresponds to a clinical workload of 45,000 subjects with a true or false positive test result each year. In the clinical setting, each subject requires once again perimetry and/or imaging, and at least one or two ophthalmologist visits. About 10,000 glaucoma and borderline cases will need ongoing follow-up. This places a heavy burden on ophthalmologists and technical personnel. Currently there are 557 ophthalmologists registered in the Netherlands, a considerable

37

Chapter 2

proportion of whom work part time (based on data from the Dutch Ophthalmologic Society, yearbook 2008). At least an additional 15 full time ophthalmologists and 30 technical ophthalmologic assistants are needed initially, and requirements may increase in the course of years.

38

Literature review

2.9 The risks, both physical and psychological, should be less than the benefits

As discussed in paragraph 2.6, diagnostics tests for glaucoma are non-invasive and do not use ionizing radiation. Consequently, screening participants are not exposed to any risks whatsoever, only burdened by having to attend the screening. There are no reports regarding the psychological consequences of false positive test results, but inconvenience from necessary follow-up and anxiety to be diagnosed with glaucoma seem likely. Medical treatment for glaucoma initially consisted of IOP lowering eye drops in the form of β-blockers and miotics. Recently, an increasing number of topical agents have become available, including carbonic anhydrase inhibitors, α2-agonists, and prostaglandin analogues. As a result of the broader therapeutic armamentarium at the ophthalmologists’ disposal, operation rates have declined in favour of medical management.82 Conscientiously prescribed medication will have few topical or systemic side effects. Occasionally it takes some trial and error to find the most suitable choice. An increasing number of preparations are also available without preservative, so that patients with allergic reactions to certain preservatives can also be treated. Only a small minority of patients can not be maintained on medication and need surgery. In contrast to medical treatment, glaucoma surgery caries a significant risk of serious ocular complications.83-86 The risk of surgical complications is weighed against the risk of glaucomatous progression due to failure to medically control IOP. In summary, treatment for glaucoma poses little risk to patients, except for difficult cases. Glaucoma has a negative influence on quality of life (QoL) in several ways. The diagnosis in itself causes depression, anxiety, changes in health perception, and worry about blindness. Furthermore, patients dislike having to use medication every day, as it reminds them of their disorder. Instillation of eye drops takes effort, it stings, and drops may have local and systemic side effects. Finally, there is a direct relationship between various QoL measures and both visual acuity and the visual field.87-92 Glaucoma reduces patients’ health-related QoL mainly in advanced stages of the disease, when severe visual field damage has occurred in both eyes.93;94 A considerable proportion of glaucoma patients would not have become blind when left untreated, before deceasing (see chapter six).77 These patients profit little from being detected by screening, but they do bear the disadvantages as mentioned above. However, the small proportion of glaucoma patients for whom screening and earlier treatment prevents visual impairment gain a substantial amount of QoL, since visual impairment has a major impact on QoL.89;90

39

Chapter 2

2.10 The costs should be balanced against the benefits Breast cancer screening programmes were initially considered successful at lowering mortality rates associated with breast cancer,95 but later on, initiated by a Swedish epidemiological study in 1999,96 they became controversial.97;98 Once a screening programme has become operational, it is extremely difficult to confirm or invalidate it’s effectiveness. Observational studies that investigate ongoing screening programmes are prone to several types of bias related to screening (see paragraph 8.2 – general discussion). It is therefore crucial to investigate whether a health problem is suitable for screening prior to the introduction of a screening programme. Unambiguous evidence for cost-effectiveness requires a randomized controlled trial (RCT), that compares glaucoma screening to the current practice of opportunistic case finding. A literature review did not identify any RCT’s on glaucoma screening, which is not surprising. Such a trial would need a very large number of participants as well as long-term follow-up. Another approach is to estimate the cost-effectiveness of glaucoma screening by developing an economic model of all major costs and effects involved (see table 2.10.1). A systemic review of all cost-effectiveness studies on glaucoma screening published until December 2005 was performed by Hernandez et al.99 They found four studies that met their inclusion criteria.100-103 These were conducted between 1983 and 1997, the latest of which was based on data from 1995. They reported that all four studies suffered from numerous methodological weaknesses, that the screening tests and treatment protocols have become outdated, that the data was not sufficiently detailed to allow for reinterpretation, and that the outcomes ranged widely between studies. Therefore, they conclude that there is insufficient economic evidence on which to base recommendations regarding screening for OAG. Since then, two additional cost-effectiveness studies have been published which will be discussed below, one conducted by Vaahtoranta-Lehtonen et al. in Finland,78 the other conducted by Hernandez et al. in the United Kingdom.79 Both studies compared population based screening to opportunistic case finding, and both studies used a Markov model for this purpose. A Markov model is suitable for modelling a chronic progressive diseases like glaucoma.104 It uses Markov states to represent all relevant outcomes in the screening and treatment process, in which individuals stay for a period of time called a ‘cycle’. Initial distribution of individuals over the separate Markov states is based on prevalence figures. The flow through the model is determined by using transition probabilities, which reflect the chance of an individual moving from one Markov state to another. Costs and benefits are calculated according to time spent per individual per Markov state.

40

Literature review

Table 2.10.1 Costs and benefits of a glaucoma screening programme.

costs

screening personnel (organisation, technicians, clinicians, overhead)

logistics (invitation of participants)

test equipment

premises

clinical diagnostic evaluation (of participants with abnormal test results)

follow-up (of borderline and glaucoma cases)

treatment (drugs; glaucoma surgery; secondary cataract surgery)

societal QoL loss (anxiety in participants with false positive test result; inconvenience)

productivity loss (due to having to attend the screening programme, and possibly needing to visit an ophthalmologist for further evaluation)

transport (travel of participants to the examination centres and back)

benefits

screening [none]

clinical earlier treatment (which means easier treatment, see paragraph 2.4)

societal reduced loss of QoL (in individuals that were successfully identified and treated due to the screening programme)

reduced blindness incurred costs (social services / rehabilitation costs; aids for reading, writing, mobility; disability homes)

reduced productivity loss from blindness (minimal effect, because progression to blindness will almost exclusively occur after retirement)

QoL = Quality of Life

41

Chapter 2

In the Finnish study,78 an organized screening programme in a population aged 50-79 years at 5 year intervals was simulated. The examinations carried out in the screening arm were IOP measurement, autorefraction, digital imaging of the fundus, and VF examination. In positive cases the same tests were repeated once. They found that the incremental cost of one Quality Adjusted Life Year (QALY) gained by screening compared to opportunistic case finding was €9023, which makes screening an appealing alternative. Costs per QALY were much higher in the younger age groups, whereas screening became dominant (i.e. both cheaper and more effective) in the eldest cohorts (>70 years). The effects of assumptions and uncertainties in the input data of the model were tested with sensitivity analyses. Diagnostic test specificity and cost per screening were revealed to have the most significant influence on the final cost per QALY in their model. Unfortunately, both of these parameters were chosen optimistically. A test specificity of 98% was assumed, based on combining different diagnostic tests and repeating abnormal test results. 98% is a very high specificity, especially since it applies to by-patient classification. This issue is further discussed in paragraph 8.2. Cost per screening was estimated to be €33, which seems rather inexpensive for a battery of four tests. Moreover, productivity loss due to having to attend screening was not taken into account, and costs for invitation of participants were also missing. Finally, although screening was found to be the most cost-effective in the eldest cohorts, no data indicating correction for the reduced life expectancy of elderly subjects was mentioned,105 and no Markov state “deceased” seemed to be present in the model. Despite these points of criticism, the majority of the study design was thorough, and the authors are lauded for their vast work in this matter. In the UK study,79 multiple simulations were carried out using various combinations of different settings in the input data of their model. These include: cohorts starting at 40, 60 or 75 years of age; screening interval of 3, 5 or 10 years; prevalence of glaucoma from 1 to 10%; test specificity from 80 to 100%; society’s willingness to pay of £10,000, £20,000, £30,000, and £50,000; and several other parameters. Two screening strategies were explored, screening by a technician and screening by a glaucoma-trained optometrist, and both were compared to opportunistic case finding. They concluded that general population screening for glaucoma is unlikely to be cost-effective. Screening at any age was expected to cost more than £30,000 per gained QALY at realistic age adjusted glaucoma prevalence rates (probably about £60,000 per QALY; rough estimate, no exact figures given). Only in certain subgroups of the 40 years old age cohort, specifically participants with a family history of glaucoma and participants of black ethnicity, screening might become cost-effective (defined as cost per QALY <£30,000; see paragraph 8.1 – general discussion) as a result of the increased glaucoma prevalence.

42

Literature review

Both the Finnish and the UK study used Markov models, and in both studies these models were populated with data based on systematic literature reviews when possible. Although the outcome measures of both studies in terms of costs per QALY were in the same order of magnitude, those costs still differed considerably, and in consequence, the authors ended up with opposing conclusions. Sensitivity analyses of both studies confirmed the common sense notion that modelling of so many variables each with its own confidence limits creates a very wide margin of uncertainty in the final output. In any case, the conflicting evidence precludes a definite statement on whether or not population based glaucoma screening should be introduced. This discussion will continue in chapter eight, appended by information from the publications contained in this thesis (chapters three to seven).

43

Chapter 2

References

1. Wald NJ. Guidance on terminology. J Med Screen 2001;8:56.

2. Resnikoff S, Pascolini D, Etya'ale D, et al. Global data on visual impairment in the year 2002. Bull World Health Organ 2004;82:844-51.

3. Heijl A, Leske MC, Bengtsson B, et al. Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial. Arch Ophthalmol 2002;120:1268-79.

4. Leske MC, Heijl A, Hyman L, Bengtsson B. Early Manifest Glaucoma Trial: design and baseline data. Ophthalmology 1999;106:2144-53.

5. Gordon MO, Kass MA. The Ocular Hypertension Treatment Study: design and baseline description of the participants. Arch Ophthalmol 1999;117:573-83.

6. Kass MA, Heuer DK, Higginbotham EJ, et al. The Ocular Hypertension Treatment Study: a randomized trial determines that topical ocular hypotensive medication delays or prevents the onset of primary open-angle glaucoma. Arch Ophthalmol 2002;120:701-13.

7. Wolfs RC, Borger PH, Ramrattan RS, et al. Changing views on open-angle glaucoma: definitions and prevalences--The Rotterdam Study. Invest Ophthalmol Vis Sci 2000;41:3309-21.

8. Burr JM, Mowatt G, Hernandez R, et al. The clinical effectiveness and cost-effectiveness of screening for open angle glaucoma: a systematic review and economic evaluation. Health Technol Assess 2007;11:iii-x, 1.

9. Hollows FC, Graham PA. Intra-ocular pressure, glaucoma, and glaucoma suspects in a defined population. Br J Ophthalmol 1966;50:570-86.

10. Wilson JM, Jungner YG. [Principles and practice of mass screening for disease]. Bol Oficina Sanit Panam 1968;65:281-393.

11. Hulsman CA, de Jong PT, Lettink M, et al. Myocilin mutations in a population-based sample of cases with open-angle glaucoma: the Rotterdam Study. Graefes Arch Clin Exp Ophthalmol 2002;240:468-74.

44

Literature review

12. Quigley HA, Broman AT. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthalmol 2006;90:262-7.

13. Kahn HA, Milton RC. Alternative definitions of open-angle glaucoma. Effect on prevalence and associations in the Framingham eye study. Arch Ophthalmol 1980;98:2172-7.

14. Cedrone C, Culasso F, Cesareo M, et al. Prevalence of glaucoma in Ponza, Italy: a comparison with other studies. Ophthalmic Epidemiol 1997;4:59-72.

15. Klein BE, Klein R, Sponsel WE, et al. Prevalence of glaucoma. The Beaver Dam Eye Study. Ophthalmology 1992;99:1499-504.

16. Mitchell P, Smith W, Attebo K, Healey PR. Prevalence of open-angle glaucoma in Australia. The Blue Mountains Eye Study. Ophthalmology 1996;103:1661-9.

17. Bonomi L, Marchini G, Marraffa M, et al. Prevalence of glaucoma and intraocular pressure distribution in a defined population. The Egna-Neumarkt Study. Ophthalmology 1998;105:209-15.

18. Tielsch JM, Sommer A, Katz J, et al. Racial variations in the prevalence of primary open-angle glaucoma. The Baltimore Eye Survey. JAMA 1991;266:369-74.

19. Mason RP, Kosoko O, Wilson MR, et al. National survey of the prevalence and risk factors of glaucoma in St. Lucia, West Indies. Part I. Prevalence findings. Ophthalmology 1989;96:1363-8.

20. Leske MC, Connell AM, Schachat AP, Hyman L. The Barbados Eye Study. Prevalence of open angle glaucoma. Arch Ophthalmol 1994;112:821-9.

21. Yoshida M, Okada E, Mizuki N, et al. Age-specific prevalence of open-angle glaucoma and its relationship to refraction among more than 60,000 asymptomatic Japanese subjects. J Clin Epidemio 2001;54:1151-8.

22. Bengtsson BO. Incidence of manifest glaucoma. Br J Ophthalmol 1989;73:483-7.

23. Ekstrom C. Elevated intraocular pressure and pseudoexfoliation of the lens capsule as risk factors for chronic open-angle glaucoma. A population-based five-year follow-up study. Acta Ophthalmol (Copenh) 1993;71:189-95.

45

Chapter 2

24. Leske MC, Connell AM, Wu SY, et al. Incidence of open-angle glaucoma: the Barbados Eye Studies. The Barbados Eye Studies Group. Arch Ophthalmol 2001;119:89-95.

25. Mukesh BN, McCarty CA, Rait JL, Taylor HR. Five-year incidence of open-angle glaucoma: the visual impairment project. Ophthalmology 2002;109:1047-51.

26. de Voogd S, Ikram MK, Wolfs RC, et al. Incidence of open-angle glaucoma in a general elderly population: the Rotterdam Study. Ophthalmology 2005;112:1487-93.

27. Fronimopoulos J, Lascaratos J. The terms glaucoma and cataract in the ancient Greek and Byzantine writers. Doc Ophthalmol 1991;77:369-75.

28. European Glaucoma Society. Terminology and Guidelines for Glaucoma. 2003. 20-2-2009. Ref Type: Report

29. Alvarado J, Murphy C, Juster R. Trabecular meshwork cellularity in primary open-angle glaucoma and nonglaucomatous normals. Ophthalmology 1984;91:564-79.

30. Hendrickx KH, van den Enden A, Rasker MT, Hoyng PF. Cumulative incidence of patients with disc hemorrhages in glaucoma and the effect of therapy. Ophthalmology 1994;101:1165-72.

31. Agarwal R, Gupta SK, Agarwal P, et al. Current concepts in the pathophysiology of glaucoma. Indian J Ophthalmol 2009;57:257-66.

32. Boden C, Sample PA, Boehm AG, et al. The structure-function relationship in eyes with glaucomatous visual field loss that crosses the horizontal meridian. Arch Ophthalmol 2002;120:907-12.

33. Levin LA. Pathophysiology of the progressive optic neuropathy of glaucoma. Ophthalmol Clin North Am 2005;18:355-64, v.

34. Agapova OA, Kaufman PL, Lucarelli MJ, et al. Differential expression of matrix metalloproteinases in monkey eyes with experimental glaucoma or optic nerve transection. Brain Res 2003;967:132-43.

35. Morrison JC, Dorman-Pease ME, Dunkelberger GR, Quigley HA. Optic nerve head extracellular matrix in primary optic atrophy and experimental glaucoma. Arch Ophthalmol 1990;108:1020-4.

46

Literature review

36. Grieshaber MC, Flammer J. Blood flow in glaucoma. Curr Opin Ophthalmol 2005;16:79-83.

37. Wilson MR, Kosoko O, Cowan CL, Jr., et al. Progression of visual field loss in untreated glaucoma patients and glaucoma suspects in St. Lucia, West Indies. Am J Ophthalmol 2002;134:399-405.

38. The Advanced Glaucoma Intervention Study (AGIS): 1. Study design and methods and baseline characteristics of study patients. Control Clin Trials 1994;15:299-325.

39. Maier PC, Funk J, Schwarzer G, et al. Treatment of ocular hypertension and open angle glaucoma: meta-analysis of randomised controlled trials. BMJ 2005;331:134.

40. Wolfs RC, Klaver CC, Ramrattan RS, et al. Genetic risk of primary open-angle glaucoma. Population-based familial aggregation study. Arch Ophthalmol 1998;116:1640-5.

41. Gordon MO, Beiser JA, Brandt JD, et al. The Ocular Hypertension Treatment Study: baseline factors that predict the onset of primary open-angle glaucoma. Arch Ophthalmol 2002;120:714-20.

42. Grodum K, Heijl A, Bengtsson B. Refractive error and glaucoma. Acta Ophthalmol Scand 2001;79:560-6.

43. Vandenbroucke JP, Hofman A. Vroege opsporing van ziekte. In: Vandenbroucke JP, Hofman A, eds. Grondslagen der epidemiologie. Elsevier; 2004:70-3.

44. Quigley HA, Jampel HD. How are glaucoma patients identified? J Glaucoma 2003;12:451-5.

45. The Advanced Glaucoma Intervention Study (AGIS): 7. The relationship between control of intraocular pressure and visual field deterioration.The AGIS Investigators. Am J Ophthalmol 2000;130:429-40.

46. Comparison of glaucomatous progression between untreated patients with normal-tension glaucoma and patients with therapeutically reduced intraocular pressures. Collaborative Normal-Tension Glaucoma Study Group. Am J Ophthalmol 1998;126:487-97.

47. Hattenhauer MG, Johnson DH, Ing HH, et al. The probability of blindness from open-angle glaucoma. Ophthalmology 1998;105:2099-104.

47

Chapter 2

48. Chen PP. Blindness in patients with treated open-angle glaucoma. Ophthalmology 2003;110:726-33.

49. Ramrattan RS, Wolfs RC, Panda-Jonas S, et al. Prevalence and causes of visual field loss in the elderly and associations with impairment in daily functioning: the Rotterdam Study. Arch Ophthalmol 2001;119:1788-94.

50. Katz J, Tielsch JM, Quigley HA, et al. Automated suprathreshold screening for glaucoma: the Baltimore Eye Survey. Invest Ophthalmol Vis Sci 1993;34:3271-7.

51. Peeters A, Schouten JS, Webers CA, et al. Cost-effectiveness of early detection and treatment of ocular hypertension and primary open-angle glaucoma by the ophthalmologist. Eye 2008;22:354-62.

52. Bebie H, Fankhauser F, Spahr J. Static perimetry: strategies. Acta Ophthalmol (Copenh) 1976;54:325-38.

53. Johnson CA, Chauhan BC, Shapiro LR. Properties of staircase procedures for estimating thresholds in automated perimetry. Invest Ophthalmol Vis Sci 1992;33:2966-74.

54. Kelly DH. Frequency Doubling in Visual Responses. J Opt Soc Am 1966;56:1628-33.

55. Maddess T, Goldberg I, Dobinson J, et al. Testing for glaucoma with the spatial frequency doubling illusion. Vision Res 1999;39:4258-73.

56. Cello KE, Nelson-Quigg JM, Johnson CA. Frequency doubling technology perimetry for detection of glaucomatous visual field loss. Am J Ophthalmol 2000;129:314-22.

57. Stoutenbeek R, Heeg GP, Jansonius NM. Frequency doubling perimetry screening mode compared to the full-threshold mode. Ophthalmic Physiol Opt 2004;24:493-7.

58. Trible JR, Schultz RO, Robinson JC, Rothe TL. Accuracy of glaucoma detection with frequency-doubling perimetry. Am J Ophthalmol 2000;129:740-5.

59. Ferreras A, Larrosa JM, Polo V, et al. Frequency-doubling technology: searching for the optimum diagnostic criteria for glaucoma. Acta Ophthalmol Scand 2007;85:73-9.

48

Literature review

60. Iwase A, Tomidokoro A, Araie M, et al. Performance of frequency-doubling technology perimetry in a population-based prevalence survey of glaucoma: the Tajimi study. Ophthalmology 2007;114:27-32.

61. Heeg GP, Ponsioen TL, Jansonius NM. Learning effect, normal range, and test-retest variability of Frequency Doubling Perimetry as a function of age, perimetric experience, and the presence or absence of glaucoma. Ophthalmic Physiol Opt 2003;23:535-40.

62. Medeiros FA, Zangwill LM, Bowd C, Weinreb RN. Comparison of the GDx VCC scanning laser polarimeter, HRT II confocal scanning laser ophthalmoscope, and stratus OCT optical coherence tomograph for the detection of glaucoma. Arch Ophthalmol 2004;122:827-37.

63. Pueyo V, Polo V, Larrosa JM, et al. Diagnostic ability of the Heidelberg retina tomograph, optical coherence tomograph, and scanning laser polarimeter in open-angle glaucoma. J Glaucoma 2007;16:173-7.

64. Leung CK, Cheung CY, Lin DS, et al. Longitudinal Variability of Optic Disc and Retinal Nerve Fiber Layer Measurements. Invest Ophthalmol Vis Sci 2008.

65. Badala F, Nouri-Mahdavi K, Raoof DA, et al. Optic disk and nerve fiber layer imaging to detect glaucoma. Am J Ophthalmol 2007;144:724-32.

66. Reus NJ, de Graaf M, Lemij HG. Accuracy of GDx VCC, HRT I, and clinical assessment of stereoscopic optic nerve head photographs for diagnosing glaucoma. Br J Ophthalmol 2007;91:313-8.

67. Medeiros FA, Vizzeri G, Zangwill LM, et al. Comparison of retinal nerve fiber layer and optic disc imaging for diagnosing glaucoma in patients suspected of having the disease. Ophthalmology 2008;115:1340-6.

68. Quigley HA, Addicks EM, Green WR. Optic nerve damage in human glaucoma. III. Quantitative correlation of nerve fiber loss and visual field defect in glaucoma, ischemic neuropathy, papilledema, and toxic neuropathy. Arch Ophthalmol 1982;100:135-46.

69. Jansonius NM, Heeg GP. The Groningen Longitudinal Glaucoma Study. II. A prospective comparison of frequency doubling perimetry, the GDx nerve fibre analyser and standard automated perimetry in glaucoma suspect patients. Acta Ophthalmol 2008.

49

Chapter 2

70. Hood DC, Kardon RH. A framework for comparing structural and functional measures of glaucomatous damage. Prog Retin Eye Res 2007;26:688-710.

71. Ferreras A, Pajarin AB, Polo V, et al. Diagnostic ability of Heidelberg Retina Tomograph 3 classifications: glaucoma probability score versus Moorfields regression analysis. Ophthalmology 2007;114:1981-7.

72. Reus NJ, Lemij HG. Diagnostic accuracy of the GDx VCC for glaucoma. Ophthalmology 2004;111:1860-5.

73. Heeg GP, Blanksma LJ, Hardus PL, Jansonius NM. The Groningen Longitudinal Glaucoma Study. I. Baseline sensitivity and specificity of the frequency doubling perimeter and the GDx nerve fibre analyser. Acta Ophthalmol Scand 2005;83:46-52.

74. Hodapp E, Parrish RK, Anderson DR. Clinical Decisions in Glaucoma. St. Louis: Mosby; 1993:52-61.

75. Mowatt G, Burr JM, Cook JA, et al. Screening tests for detecting open-angle glaucoma: systematic review and meta-analysis. Invest Ophthalmol Vis Sci 2008;49:5373-85.

76. Heeg GP, Jansonius NM. The groningen longitudinal glaucoma study III. The predictive value of frequency-doubling perimetry and GDx nerve fibre analyser test results for the development of glaucomatous visual field loss. Eye 2008.

77. Stoutenbeek R, de Voogd S, Wolfs RC, et al. The additional yield of a periodic screening programme for open-angle glaucoma: a population-based comparison of incident glaucoma cases detected in regular ophthalmic care with cases detected during screening. Br J Ophthalmol 2008;92:1222-6.

78. Vaahtoranta-Lehtonen H, Tuulonen A, Aronen P, et al. Cost effectiveness and cost utility of an organized screening programme for glaucoma. Acta Ophthalmol Scand 2007;85:508-18.

79. Hernandez RA, Burr JM, Vale LD. Economic evaluation of screening for open-angle glaucoma. Int J Technol Assess Health Care 2008;24:203-11.

80. Population of the Netherlands on 01-01-2008. Available at: www.cbs.nl. Accessed August 21, 2008.

50

Literature review

81. Bonfill X, Marzo M, Pladevall M, et al. Strategies for increasing women participation in community breast cancer screening. Cochrane Database Syst Rev 2001;CD002943.

82. Bateman DN, Clark R, Azuara-Blanco A, et al. The effects of new topical treatments on management of glaucoma in Scotland: an examination of ophthalmological health care. Br J Ophthalmol 2002;86:551-4.

83. Jampel HD, Musch DC, Gillespie BW, et al. Perioperative complications of trabeculectomy in the collaborative initial glaucoma treatment study (CIGTS). Am J Ophthalmol 2005;140:16-22.

84. Feiner L, Piltz-Seymour JR. Collaborative Initial Glaucoma Treatment Study: a summary of results to date. Curr Opin Ophthalmol 2003;14:106-11.

85. Stein JD, Ruiz D, Jr., Belsky D, et al. Longitudinal rates of postoperative adverse outcomes after glaucoma surgery among medicare beneficiaries 1994 to 2005. Ophthalmology 2008;115:1109-16.

86. Mills KB. Trabeculectomy: a retrospective long-term follow-up of 444 cases. Br J Ophthalmol 1981;65:790-5.

87. Janz NK, Wren PA, Guire KE, et al. Fear of blindness in the Collaborative Initial Glaucoma Treatment Study: patterns and correlates over time. Ophthalmology 2007;114:2213-20.

88. Janz NK, Wren PA, Lichter PR, et al. Quality of life in newly diagnosed glaucoma patients : The Collaborative Initial Glaucoma Treatment Study. Ophthalmology 2001;108:887-97.

89. Bechetoille A, Arnould B, Bron A, et al. Measurement of health-related quality of life with glaucoma: validation of the Glau-QoL 36-item questionnaire. Acta Ophthalmol 2008;86:71-80.

90. Broman AT, Munoz B, Rodriguez J, et al. The impact of visual impairment and eye disease on vision-related quality of life in a Mexican-American population: proyecto VER. Invest Ophthalmol Vis Sci 2002;43:3393-8.

91. Freeman EE, Munoz B, West SK, et al. Glaucoma and quality of life: the Salisbury Eye Evaluation. Ophthalmology 2008;115:233-8.

51

Chapter 2

92. Rahi JS, Cumberland PM, Peckham CS. Visual impairment and vision-related quality of life in working-age adults: findings in the 1958 British birth cohort. Ophthalmology 2009;116:270-4.

93. Tuulonen A, Airaksinen PJ, Erola E, et al. The Finnish evidence-based guideline for open-angle glaucoma. Acta Ophthalmol Scand 2003;81:3-18.

94. Burr JM, Kilonzo M, Vale L, Ryan M. Developing a preference-based Glaucoma Utility Index using a discrete choice experiment. Optom Vis Sci 2007;84:797-808.

95. Nystrom L, Rutqvist LE, Wall S, et al. Breast cancer screening with mammography: overview of Swedish randomised trials. Lancet 1993;341:973-8.

96. Sjonell G, Stahle L. [Mammographic screening does not reduce breast cancer mortality]. Lakartidningen 1999;96:904-13.

97. Gotzsche PC, Olsen O. Is screening for breast cancer with mammography justifiable? Lancet 2000;355:129-34.

98. Baum M. Breast cancer screening comes full circle. J Natl Cancer Inst 2004;96:1490-1.

99. Hernandez R, Rabindranath K, Fraser C, et al. Screening for open angle glaucoma: systematic review of cost-effectiveness studies. J Glaucoma 2008;17:159-68.

100. Boivin JF, McGregor M, Archer C. Cost effectiveness of screening for primary open angle glaucoma. J Med Screen 1996;3:154-63.

101. Gottlieb LK, Schwartz B, Pauker SG. Glaucoma screening. A cost-effectiveness analysis. Surv Ophthalmol 1983;28:206-26.

102. Tuck MW, Crick RP. The cost-effectiveness of various modes of screening for primary open angle glaucoma. Ophthalmic Epidemiol 1997;4:3-17.

103. Gooder P. Screening for Glaucoma. Development an Evaluation Committee (DEC). 38. 1995. Bristol. Ref Type: Report

52

Literature review

53

104. Drummond MF, Sculpher MJ, Torrance GW, et al. Methods for the Economic Evaluation of Health Care Programmes. 3rd ed. Oxford University Press; 2005.

105. Wesselink C, Stoutenbeek R, Jansonius NM. Until what age should glaucoma be monitored and/or treated? A novel index that facilitates clinical decision making. 200. Ref Type: Unpublished Work

Chapter 3 Frequency doubling perimetry screening mode compared to the full threshold mode

55

Chapter 3

FREQUENCY DOUBLING PERIMETRY SCREENING MODE COMPARED TO THE FULL THRESHOLD MODE Remco Stoutenbeek, Govert P. Heeg and Nomdo M. Jansonius Department of Ophthalmology, University Hospital Groningen, P.O. Box 30.001, 9700 RB Groningen, The Netherlands Abstract The diagnostic performance of the frequency doubling perimetry (FDT) C20-1 screening mode was compared to that of the C20 full-threshold mode. For the number of defects p < 1% in the total deviation plot, both modes appeared to perform similarly in terms of sensitivity, specificity, and area under the receiver–operating characteristic (ROC) curve. Different cut-off points should be applied for both modes to obtain equal sensitivity and specificity values, and – related to that – for most subjects more defects were found in full-threshold mode than in screening mode. For the screening mode, we found a sensitivity of 0.91 and a specificity of 0.88 at a cut-off point of >0 defects, and an area under the ROC curve of 0.93.

56

Frequency doubling perimetry screening mode compared to the full threshold mode

3.1 Introduction Glaucoma is a chronic disease that may cause irreversible blindness, but the early stages of glaucoma do not cause any symptoms. Treatment of glaucoma may arrest or slow down its progress (e.g. Heijl et al., 2002). Therefore, screening for glaucoma might be advisable. Various studies have suggested that the frequency doubling perimeter (FDT) may be suitable for this purpose (Johnson and Samuels, 1997; Quigley, 1998; Cello et al., 2000). FDT can be applied in two different modes, a full-threshold mode and a screening mode. In full-threshold mode, a contrast threshold is determined for each of the test locations by means of a staircase procedure. In screening mode, on the contrary, a suprathreshold strategy is applied. The full-threshold mode produces two global indices (mean deviation and pattern standard deviation) and total and pattern deviation plots. The output of the screening mode is limited to a single plot that can be considered the equivalent of the total deviation plot in full-threshold mode. Therefore, the full-threshold mode might be the more informative one. The screening mode, however, has a much shorter testing time. Besides logistical advantages, the diagnostic performance of the screening mode could be better because a shorter concentration time span is required. In an accompanying study (Müskens et al., 2004), we compared all previously published algorithms for the interpretation of FDT test results in full-threshold mode. We found that none of the algorithms performed substantially better than simply counting the number of defects p < 1% in the total deviation plot. Interestingly, this algorithm can also be applied in screening mode. The question is, however, whether the algorithm performs equally well in both modes. Trible et al. (2000) found for the C20-1 screening mode a sensitivity of 0.39 (n = 51 eyes), 0.86 (n = 42 eyes), and 1.00 (n = 32 eyes), for early, moderate, and severe glaucoma, respectively, and a specificity of 0.95 (n = 95 normal eyes). In full threshold mode they found sensitivities of 0.59, 0.93, and 1.00, respectively, and a specificity of 0.82. Thomas et al. (2002) found for the C20-1 screening mode a sensitivity of 0.81 and a specificity of 0.95. According to their methods, they also performed an FDT N30 full-threshold mode test on all subjects. Unfortunately, these results were not reported in their article. Finally, a third comparison of both modes by Burnstein et al. (2000) was based on only 20 glaucoma patients and nine suspects. The aim of the present study is to investigate whether the algorithm `counting the number of defects p < 1% �in the total deviation plot ’ performs at least as well in screening mode as compared to the full-threshold mode. To this end we compared sensitivity, specificity, and the area under the receiver–operating characteristic (ROC) curve in a large group of glaucoma patients and normal subjects.

57

Chapter 3

3.2 Methods Patient data and study protocol The glaucoma patients incorporated in the present study were recruited from a large cohort that is currently followed prospectively with FDT and GDx (Nerve Fibre Analyser; Laser Diagnostic Technologies, San Diego, CA, USA) in the glaucoma service of our outpatient department. In this cohort, 452 patients out of 1051 consecutive visitors to the glaucoma service were classified as glaucoma patients. Classification was based on conventional perimetry. Classification of patients can be done both on by-eye basis and on by-patient basis. We decided to perform by-patient analysis because results of this type of analysis are easier to extrapolate to clinical practice: any patient with an abnormality in any eye needs further attention whereas only patients with normal test results in both eyes can be reassured. Thus, in our study a glaucoma patient was defined as a patient with a reproducible visual field defect on conventional perimetry in at least one eye. Defects had to be compatible with glaucoma and without any other explanation. For further details see Heeg et al. (2003). The worse eye of the glaucoma patients had an average mean deviation (MD) Humphrey Field Analyser (HFA) of -11.5 dB (SD 8.6 dB). In 100 consecutive patients out of the 452 glaucoma patients we added an FDT measurement in screening mode to the usual measurement in full-threshold mode. The mean age of these patients was 68 years (SD 13 years; range 28–89). In addition to these 100 glaucoma patients from the glaucoma service, 108 normal subjects were recruited outside the hospital by advertising in local senior departments, among volunteers from the local blood bank, and in other public places. Subjects regularly visiting an ophthalmologist for glaucoma-related reasons were excluded. Both in the glaucoma patients and in the normal subjects no (further) exclusion criteria were applied in order to get samples as representative as possible of the populations studied. The mean age of the normal subjects was 61 years (SD 11 years; range 36–83). We alternated the sequence of the tests; subgroup analyses did not reveal any influence of the testing sequence. Frequency doubling perimetry Testing was performed with the frequency doubling perimeter (Carl Zeiss Meditec AG, Jena, Germany) using the C20 full-threshold mode and the C20-1 screening mode. FDT was performed in both the eyes, and a patient was considered positive if there was an abnormal test result or an unfeasible test in at least one eye. The test result was based on the number of defects p < 1% (called `mild’ in C20-1 screening mode) in the total deviation plot. In case of only one functional eye, only that eye was used for classifying the patient and for determining the test result. For details on frequency doubling perimetry see e.g. Maddess and Henry (1992); Johnson and Samuels (1997), and Maddess et al. (1999).

58

Frequency doubling perimetry screening mode compared to the full threshold mode

Analysis Sensitivities, specificities, and Areas Under the ROC Curve (AUC) were calculated for both modes. Confidence limits for sensitivity and specificity were calculated with PEPI 4.0 (Abramson and Gahlinger, 2001); AUCs were calculated using SPSS 10.0.07. Possible differences in sensitivity or specificity values between both modes were analysed with McNemar’s test (paired proportions); differences in AUC values were compared with the technique described by Hanley and McNeil (1983). Analyses were performed both for the group as a whole and after stratifying the glaucoma patients into two groups (early and moderate/severe) according to the level of damage. Early glaucoma was defined as MD (HFA) -6 dB or better in the worse eye; moderate/ severe glaucoma as MD(HFA) <-6 dB. 3.3 Results Table 3.3.1 shows sensitivity and specificity values of both the full-threshold mode and the screening mode at various cut-off points. Figure 3.3.1 presents the corresponding ROC curves. To establish a sensitivity of 0.90, the full-threshold mode and the screening mode appeared to need different cut-off points, >1 and >0, respectively. At these cut-off points, specificities were 0.83 (95% confidence limits 0.75–0.89) and 0.87 (0.80–0.91) for the full-threshold mode and the screening mode respectively. This difference is not significant (p = 0.30). The area under the ROC curve was 0.93 (95% confidence limits 0.89–0.96) for the full-threshold mode and also 0.93 (0.90–0.97) for the screening mode (no significant difference). A useless test has an area under the ROC curve of 0.50; a faultless test has an area of 1.00.

59

Chapter 3

Table 3.3.1 Sensitivity and specificity of FDT full-threshold mode and screening mode at various cut-off points.

full-threshold mode screening mode cut-off

sensitivity specificity sensitivity specificity

>0 0.96 0.72 0.91 0.88

>1 0.91 0.83 0.84 0.94

>2 0.87 0.85 0.78 0.95

>3 0.82 0.90 0.68 0.97

>4 0.75 0.92 0.65 0.98

Figure 3.3.1 Receiver–operating characteristic (ROC) curves of the FDT full-threshold mode (FT) and screening mode (SC). Diamonds correspond to cut-off point >1 for full-threshold mode and >0 for screening mode.

60

Frequency doubling perimetry screening mode compared to the full threshold mode

Table 3.3.2 shows 2 x 2 matrix form in which the test results of the full-threshold mode (cut-off > 1) and the screening mode (cut-off > 0) are compared in glaucoma patients and normal subjects. As can be seen in this table, four of 100 glaucoma patients and 15 of 108 normal subjects were classified differently by the two modes. Figure 3.3.2 presents a scatter plot of the screening mode and the full-threshold mode results of all 208 subjects. This plot illustrates that for most subjects more defects were found in full-threshold mode than in screening mode. Table 3.3.2 Test results of the full-threshold mode and the screening mode compared in glaucoma patients and normal subjects.

glaucoma patients normal subjects

SC = 0 SC > 0 SC = 0 SC > 0

FT ≤ 1 7 2 85 5

FT > 1 2 89 10 8

FT, full-threshold mode; SC, screening mode.

61

Chapter 3

Figure 3.3.2 Scatter plot showing the screening mode and the full-threshold mode test results (number of defects p < 1% in the total deviation plot) of both the glaucoma patients (◊) and the normal subjects (+). Noise with a maximum amplitude of 0.3 was added in order to avoid overlapping dots. Table 3.3.3 presents the sensitivity of both modes after stratifying the sample into two groups according to the level of damage. The stratifying process yielded 34 patients with early glaucoma and 66 patients with moderate/severe disease. At the cut-off points as mentioned above (>1 for full-threshold; >0 for screening mode), both modes achieved a sensitivity of 1.00 for moderate/severe disease, i.e. they missed none of the moderate and severe glaucoma cases. The sensitivity for early glaucoma did not differ between both modes.

62

Frequency doubling perimetry screening mode compared to the full threshold mode

Table 3.3.3 Sensitivity of both modes after stratifying the sample into early and moderate/severe glaucoma.

early glaucoma moderate/severe glaucoma cut-off

FT SC FT SC

> 0 0.88 0.74 1.00 1.00

> 1 0.74 0.59 1.00 0.97

> 2 0.62 0.47 1.00 0.94

> 3 0.56 0.29 0.95 0.88

> 4 0.41 0.26 0.92 0.85

FT, full-threshold mode; SC, screening mode. 3.4 Discussion In this study, we compared the diagnostic performance of the FDT screening mode to that of the full-threshold mode. For the number of defects p < 1% in the total deviation plot, both modes appear to perform similarly in terms of sensitivity, specificity, and area under the ROC curve. However, different cut-off points should be applied for both modes and – related to that – for most subjects more defects were found in full-threshold mode than in screening mode. In the present study we found, like Trible et al. (2000), that the diagnostic performance of the screening mode is at least equal to that of the full-threshold mode. Moreover, we confirmed their interesting finding that both modes need different criteria to obtain equal sensitivity and specificity values. Trible et al. used >0 defects p < 0.5% in the total deviation plot to define an abnormal test result in the full-threshold mode and >0 defects p < 1% in the screening mode. We found >1 defects p < 1% for the full-threshold mode and >0 defects p < 1% for the screening mode. These findings are important because they are essential for a proper comparison of FDT test results obtained with different modes in the same subject.

63

Chapter 3

Figure 3.3.2 suggests a rather poor correlation between the full-threshold mode and the screening mode. This might be caused by the fact that the two modes are actually measuring different things. Part of the apparently poor correlation, however, could also be caused by a large test–retest variability of FDT, independent of the mode used (see Heeg et al., 2003). Figure 3.3.2 also shows that most points are below the diagonal, i.e. there is apparently more damage in full-threshold mode. Differences in testing strategy (staircase vs supra-threshold) could be a possible explanation. Fatigue in the longer lasting full-threshold mode could be another factor. In our data, the mean duration of a test in full-threshold mode was 272s (SD 24s) in normal subjects and 282s (SD 37s) in glaucoma patients; the mean duration in screening mode was 45s (SD 7s) in normal subjects and 79s (SD 36s) in glaucoma patients. The duration of a test in screening mode depends on the number of defects because only missed stimuli are repeated at a higher contrast value. For that reason, Patel et al. (2000) suggested that any test in screening mode lasting longer than 90s could be considered abnormal irrespective of the test result. Alternatively, a new screening mode could be designed that skips the repeated testing at p < 0.5% and at full contrast. Information regarding the depth of a defect can be skipped without loss of diagnostic performance for screening (Müskens et al., 2004). For other purposes (e.g. disease staging or follow-up), however, information on defect depth might be very useful. In conclusion, the FDT C20-1 screening mode saves time and performs for screening purposes at least as well as the FDT C20 full-threshold mode. A criterion for an abnormal result in screening mode with a reasonable trade-off between false-positive and false-negative test results is: one or more defects p < 1% in the total deviation plot. Acknowledgements The authors wish to thank Johan Drost, José Eisses, and Wietse Wieringa, technicians in our outpatient department, for performing the FDT measurements. This research was supported by the Dutch Health Care Insurance Council (CVZ) through the Department of Medical Technology Assessment (MTA) of the University Hospital Groningen, the Netherlands.

64

Frequency doubling perimetry screening mode compared to the full threshold mode

References for chapter three

1. Abramson JH, Gahlinger PM. Computer Programs for Epidemiologists PEPI Version 4.0. Sagebrush Press 2001; Salt Lake City, UT.

2. Burnstein Y, Ellish NJ, Magbalon M, Higginbotham EJ. Comparison of

frequency doubling perimetry with humphrey visual field analysis in a glaucoma practice. Am J Ophthalmol 2000;129:328–333.

3. Cello KE, Nelson-Quigg JM, Johnson CA. Frequency doubling technology

perimetry for detection of glaucomatous visual field loss. Am J Ophthalmol 2000;129:314–322.

4. Hanley JA, McNeil BJ. A method of comparing the areas under receiver–

operating characteristic curves derived from the same cases. Radiology 1983;148:839–843.

5. Heeg GP, Ponsioen TL, Jansonius NM. Learning effect, normal range, and

test–retest variability of frequency doubling perimetry as a function of age, perimetric experience, and the presence or absence of glaucoma. Ophthal Physiol Opt 2003;23:535–540.

6. Heijl A, Leske MC, Bengtsson B, Hyman L, Bengtsson B, Hussein M.

Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial. Arch Ophthalmol 2002;120: 1268–1279.

7. Johnson CA, Samuels SJ. Screening for glaucomatous visual field loss

with frequency-doubling perimetry. Invest. Ophthalmol. Vis. Sci. 1997;38:413–425.

8. Müskens RPHM, Heeg GP, Jansonius NM. An evaluation of algorithms

designed to classify the results from frequency doubling perimetry. Ophthal Physiol Opt 2004;24:498–503.

9. Maddess T, Henry GH. Performance of a nonlinear pathway in subjects

with ocular hypertension and glaucoma. Clin Vis Sci 1992;7:371–382.

10. Maddess T, Goldberg I, Dobinson J, Wine S, Welsh AH, James AC. Testing for glaucoma with the spatial frequency doubling illusion. Vision Res 1999;39:4258– 4273.

65

Chapter 3

11. Patel SC, Friedman DS, Varadkar P, Robin AL. Algorithm for interpreting

the results of frequency doubling perimetry. Am J Ophthalmol 2000; 129:323–327.

12. Quigley HA. Identification of glaucoma-related visual field abnormality

with the screening protocol of frequency doubling technology. Am J Ophthalmol 1998;125: 819–829.

13. Thomas R, Bhat S, Muliyil JP, Parikh R, George R. Frequency doubling

perimetry in glaucoma. J Glaucoma 2002;11:46–50.

14. Trible JR, Schultz RO, Robinson JC, Rothe TL. Accuracy of glaucoma detection with frequency doubling perimetry. Am J Ophthalmol 2000; 129:740–745.

66

Frequency doubling perimetry screening mode compared to the full threshold mode

67

Chapter 4 Strategies for improving the diagnostic specificity of the frequency doubling perimeter

69

Chapter 4

STRATEGIES FOR IMPROVING THE DIAGNOSTIC SPECIFICITY OF THE FREQUENCY DOUBLING PERIMETER Govert P. Heeg, Remco Stoutenbeek and Nomdo M. Jansonius Department of Ophthalmology, University Hospital Groningen, Groningen, The Netherlands Abstract Purpose: To evaluate various strategies designed to improve the specificity of the interpretation of results obtained with the frequency doubling technology perimeter (FDT) used in the full-threshold mode. Methods: Three different strategies were compared using data from 452 glaucoma patients and 237 healthy subjects: combining several FDT parameters from a single test, combining the FDT test with a GDx test, and confirming an abnormal FDT test result with a repeat test. Results: Confirming an abnormal FDT test result with a repeat test yielded a specificity increase of 0.10, from 0.80 to 0.90, at the expense of some loss of sensitivity for early but not for moderate or severe glaucoma. Combining several FDT parameters from a single test and combining FDT with GDx did not yield any noticeable increase in diagnostic performance. Conclusions: A modest increase in FDT diagnostic performance can be obtained by the confirmation of an abnormal test result with a repeat test.

70

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

4.1 Introduction Glaucoma is a chronic disease that may cause irreversible blindness. Early stages of glaucoma do not result in any symptoms. Treatment of glaucoma may arrest or slow down its progress (Heijl et al. 2002). Screening for glaucoma may, therefore, be advisable and various studies have suggested that the frequency doubling technique (FDT; Carl Zeiss Meditec Inc., Dublin, California, USA) may be suitable for this purpose (Johnson & Samuels 1997; Quigley 1998; Cello et al. 2000). Sensitivities and specificities reported in previous data vary. The FDT parameter with the best diagnostic performance was found to be the number of depressed test-points p<0.01 in the total deviation probability plot, with a specificity of 0.81 at a sensitivity of 0.90 (cut-off point >1; Heeg et al. 2005). False-positive test results will be much more common than true-positive test results with this specificity because the prevalence of glaucoma is only about 1% in the general population (Wolfs et al. 2000). Several investigators have tried to improve the diagnostic performance of FDT. Khong et al. (2001) investigated whether the specificity of FDT could be improved by confirming an abnormal test result. They found that this increased the specificity from 0.62 to 0.69. Horn et al. (2003) evaluated the diagnostic performance of the combined use of FDT (C20-5 screening mode) and the nerve fibre analyser (GDx; Laser Diagnostic Technologies, San Diego, California, USA) and found that the combined use of both techniques was superior to testing with only one method. The aim of the present study was to investigate the extent to which the strategies mentioned above, and others, are able to decrease the number of false-positive FDT test results in a large group of subjects. Three strategies were explored in order to do this. Firstly, we tried to combine the FDT result (number of depressed test-points p<0.01 in the total deviation probability plot) with another FDT parameter. Secondly, we added information from a GDx measurement. Thirdly, we added a second FDT measurement. 4.2 Material and Methods Patient data and study protocol The data and other details have been previously described (Heeg et al. 2005). To summarize, 875 regular visitors to the glaucoma service at our outpatients department underwent frequency doubling perimetry and a GDx measurement as a baseline measurement for a prospective follow-up. In this cohort, 452 patients were classified as glaucoma patients. A glaucoma patient was defined as a patient with a reproducible visual field defect in at least one eye using

71

Chapter 4

conventional perimetry. Defects had to be compatible with glaucoma and without any other explanation. In addition to these 452 glaucoma patients from the glaucoma service, 237 healthy subjects were recruited from outside the hospital. Very few exclusion criteria were applied so that the samples were as representative as possible of the populations studied. All 452 glaucoma patients underwent at least one FDT and one GDx test. A second FDT test was performed on 120 of the glaucoma patients. All 237 healthy subjects completed at least one FDT and 108 of them completed a GDx test. A second FDT measurement was performed on 129 healthy subjects. Thus, the various strategies were explored in different groups and subgroups. Frequency doubling technique Testing was performed with the frequency doubling perimeter (FDT; Version 2.60; Carl Zeiss Meditec Inc., Dublin, California, USA) using the C-20 full threshold mode. An FDT measurement was considered positive if there was an abnormal test result, or an unfinished test, in at least one eye. An abnormal test result was defined as >1 depressed test-point p<0.01 in the total deviation probability plot (TD>1; Heeg et al. 2005). Three additional parameters were studied: mean deviation in dB (MD), pattern standard deviation in dB (PSD), and the number of depressed test-points p<0.01 in the pattern deviation probability plot (PD). For details of FDT, see Maddess & Henry (1992), Johnson & Samuels (1997) and Maddess et al. (1999). GDx - nerve fibre analyser Testing was performed with the nerve fibre analyser (GDx, Version 2.0.10; Laser Diagnostic Technologies, San Diego, California, USA). We confined the analyses within this study to the GDx parameter the Number. The Number is a global parameter that gives an indication of glaucoma probability. Its value ranges from 0 (no glaucoma) to 100 (glaucoma). A GDx measurement was considered positive if there was an abnormal test result or an unfinished test in at least one eye. An abnormal test result was defined as the Number >29 (Heeg et al. 2005). Six images were recorded for each eye. Another six images were made if the first series did not contain an image with sufficiently high image quality. High image quality required a well centred optic nerve head, an in-focus image, equal illumination in all quadrants and an absence of motion artefacts. A mean image was created if at least two images with high image quality were available. For scanning laser polarimetry details, see Weinreb et al. (1990) and Dreher & Reiter (1992). Strategies Three different strategies to improve the diagnostic performance of FDT were explored:

72

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

1. combining TD with another parameter of FDT: MD, PSD or PD; 2. combining TD with GDx parameter the Number, and 3. repeating the FDT measurement after a positive test result

where TD is the number of test-point locations depressed at the p<0.01 level in the total deviation probability plot, MD the mean deviation in dB, PSD the pattern standard deviation in dB, and PD the number of test-point locations depressed at the p<0.01 level in the pattern deviation probability plot. Analyses For strategies 1 and 2, the parameters MD, PSD, PD and the Number were combined with TD and the resulting sensitivities and specificities were subsequently calculated. Combining parameters can be carried out in many ways. We used the AND/OR operators. If the OR operator is used for defining an abnormal (i.e. positive) combination, then there must be at least one abnormal parameter. An abnormal combination in the case of the AND operator requires both parameters to be abnormal. For strategy 3, only the AND operator was applied (see Discussion), so that the overall test result was considered positive if both the first and second tests were abnormal. In addition to exploring the effects of the strategies on the group as a whole, the strategies were also tested after the exclusion of patients with early glaucoma. Early glaucoma was defined as MD(HFA) ≥- 6 dB in the worse eye. 4.3 Results Tables 4.3.1 A and B show the sensitivity and specificity for combinations of TD with FDT parameters MD, PSD and PD (Table 4.3.1 A) and with GDx parameter the Number (Table 4.3.1B). The original cut-off points (Heeg et al. 2005) were used for both TD and the second parameter. At this cut-off point, TD by itself had a sensitivity of 0.90 in the 452 glaucoma patients, a specificity of 0.81 in the healthy subjects as presented in Table 4.3.1 A, and a specificity of 0.86 in the subgroup of healthy subjects as presented in Table 4.3.1 B. As would be expected, the application of the AND operator resulted in an increase in specificity and a decrease in sensitivity, whereas the opposite occurred in case of implementing the OR operator. As can be seen in Table 4.3.1 A, no substantial increase in diagnostic performance could be obtained by combining FDT parameters. The combination of TD AND the Number resulted in a specificity increase of 0.08 (from 0.86 to 0.94) at the expense of a sensitivity decrease of 0.07 (from 0.90 to 0.83). This is shown in Table 4.3.1 B. After the exclusion of patients with early glaucoma, the sensitivity of TD AND the Number increased to 0.99. This value should be

73

Chapter 4

compared to the sensitivity of TD alone, which is 1.00 after the exclusion of patients with early glaucoma. Table 4.3.1. A Sensitivity and specificity after combining TD with a second parameter of FDT. Based on 452 glaucoma patients and 237 healthy subjects. Original cut-off points from Heeg et al. (2005) were used for all parameters.

criterion cut off sensitivity specificity

TD > 1 0.90 0.81

MD < - 1.8 0.90 0.72

PSD > + 4.8 0.91 0.68

PD > 0 0.90 0.59

TD and MD 0.87 0.83

TD or MD 0.92 0.70

TD and PSD 0.85 0.84

TD or PSD 0.96 0.65

TD and PD 0.83 0.84

TD or PD 0.97 0.56

TD = the number of depressed test-points p<0.01 in the total deviation probability plot; MD = mean deviation in dB; PSD = pattern standard deviation in dB; PD = number of depressed test-points p<0.01 in pattern deviation probability plot.

74

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

Table 4.3.1. B Sensitivity and specificity after combining TD with a second parameter of GDx. Based on 452 glaucoma patients and 108 healthy subjects. Original cut-off points from Heeg et al. (2005) were used for all parameters.

criterion cut off sensitivity specificity

TD > 1 0.90 0.86

Number > 29 0.90 0.78

TD and Number 0.83 0.94

TD or Number 0.96 0.69

TD = the number of depressed test-points p<0.01 in the total deviation probability plot; Number = GDx parameter the Number. Table 4.3.2 shows the results for strategy 3, which involved the repetition of a FDT measurement after a positive test result. The first test yielded an abnormal result in 105 of 120 glaucoma patients and in 26 of 129 healthy subjects. The resulting sensitivity (0.88) and specificity (0.80) in this subgroup of 249 subjects do not differ from those of the original 452 glaucoma patients and 237 healthy subjects (0.90 and 0.81, respectively). Applying the AND operator yielded a specificity increase of 0.08 (0.80 to 0.88) at the expense of some sensitivity (0.83 instead of 0.88). The requirement that at least two depressed test locations have the same location on retest resulted in a specificity of 0.90 at a sensitivity of 0.82. Repeating a test with five or more depressed test locations in any eye appeared to be superfluous. If these tests had been considered as positive without performing a retest, this would have prevented 87 of 105 retests carried out in the glaucoma patients, and nine of 26 retests carried out in the healthy subjects, without any loss of diagnostic performance. This approach is depicted as the last strategy in Table 4.3.2. The sensitivity in this strategy returned to 0.99 after the exclusion of patients with early glaucoma. After the exclusion, 77 glaucoma patients remained.

75

Chapter 4

Table 4.3.2. Sensitivity and specificity after repeating an abnormal first FDT test (FDT1) with a second FDT test (FDT2) for a subgroup of 249 patients (120 glaucoma patients and 129 healthy subjects). An abnormal FDT test was defined as more than one depressed test-point p<0.01 in the total deviation probability plot. For same location, at least two depressed test-points had to have the same location on retest. For the last strategy, a second test had only to be performed if the worse eye on FDT1 had 2–4 depressed test locations (<2 in both eyes on FDT1: negative final test result, no retest required; >4 in any eye on FDT1: positive final test result, no retest required).

patient group criterion sensitivity specificity

FDT1 > 1 0.88 0.80

FDT1 > 1 and FDT2 > 1 0.83 0.88

FDT1 > 1 and FDT2 > 1, same location 0.82 0.90

all patients

FDT1 > 5 or (FDT1 > 1and FDT2 > 1), same location 0.84 0.90

FDT1 > 1 0.99 0.80

FDT1 > 1 and FDT2 > 1 0.99 0.88

FDT1 > 1 and FDT2 > 1, same location 0.97 0.90

early glaucoma excluded

FDT1 > 5 or (FDT1 > 1and FDT2 > 1), same location 0.99 0.90

FDT = Frequency Doubling Technoloy perimetry. Only the third strategy of those tested yielded some gain in diagnostic performance. However, even when employing that strategy a substantial increase in specificity appeared to be impossible without a small decrease in sensitivity for early glaucoma cases. Because this might also be achieved by simply raising the cut-off point of TD, we explored that as well. Table 4.3.3 presents the results. Increasing the cut-off point of TD as a single parameter is seen to be less effective than the third strategy in increasing the diagnostic performance.

76

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

Table 4.3.3 Sensitivity and specificity of TD as a function of its cut-off point based on 452 glaucoma patients (329 after exclusion of early glaucoma) and 237 healthy subjects.

cut off TD patient group sensitivity specificity

all patients 0.90 0.81 > 1

early glaucoma excluded 1.00 0.81

all patients 0.86 0.85 > 2

early glaucoma excluded 0.99 0.85

all patients 0.82 0.88 >3

early glaucoma excluded 0.97 0.88

TD = the number of depressed test-points p<0.01 in the total deviation probability plot 4.4 Discussion In this study, we evaluated three strategies intended to improve the specificity of FDT. The most successful strategy appears to be confirming an abnormal FDT test result with a repeat test. This strategy yields a substantial increase in specificity at the expense of some loss of sensitivity for early but not for moderate and severe glaucoma. Neither combining FDT with another FDT parameter nor combining FDT with GDx seems to be a sensible approach. Khong et al. (2001) found a specificity of 0.62 using the FDT C20-5 screening mode, which improved to 0.69 after a second measurement in patients with abnormal first test results. It is difficult to compare their results with our approach because they excluded almost 50% of the original 223 subjects, whereas we excluded very few patients. Nevertheless, the specificity increase they found (0.07) seems to be surprisingly similar to that in our results. Joson et al. (2002) measured the effect of repeated testing in healthy subjects using the C20-5 screening mode. They found a specificity increase from 0.85 to 0.96. Neither Khong et al. (2001) nor Joson et al. (2002) acknowledged the fact that repeated testing might compromise sensitivity. Khong et al. (2001) did mention a sensitivity, but the sensitivity of 1.00 they found was based on results from only two patients. The corresponding 95%

77

Chapter 4

confidence intervals for n = 2 would range from 0.22 to 1.00 (Abramson & Gahlinger 2001). Horn et al. (2003) obtained sensitivities of 0.85 and 0.64, for FDT and GDx, respectively, at a predefined specificity of 0.95 in 252 patients. When FDT was combined with GDx the sensitivity increased by 0.07 to 0.92. The increase in diagnostic performance found by Horn et al. (2003) when combining both devices is difficult to compare with our results because there are several methodological differences between the two studies. As mentioned in Methods, there are many ways to combine parameters. Horn et al. (2003), for example, used a two-dimensional discriminant analysis to combine their FDT score with the GDx Number. In our study, we selected the straightforward AND/OR operators. The AND operator is especially interesting from a logistical point of view: a second test need only be performed if the first test yields an abnormal test result. All other methods for combining parameters require both tests to be performed on all patients. The present study was conducted using the original version of the FDT perimeter with 17 test locations. A 24–2 version (Matrix) has recently been developed with 54 test locations. The original version, however, is still more common in clinical practice, is better documented than the newer instrument, and is probably more attractive as a screening instrument. A new GDx version with a variable corneal compensator (GDx-VCC) was launched last year. The GDx-VCC seems to display a slightly better diagnostic performance when compared to the original GDx (Zhou & Weinreb 2002; Bagga et al. 2003; Reus et al. 2003). In conclusion, a modest increase in FDT diagnostic performance could be obtained by confirmation of an abnormal test result with a repeat test. This strategy resulted in a specificity of 0.90, which is still rather a low specificity for a screening test. Combining TD with another FDT parameter or combining FDT with GDx did not seem to be useful. The combination of FDT with GDx also requires the purchase of a second device, whereas confirming an abnormal test result with a repeat test is more in keeping with common clinical practice. The latter strategy would, therefore, seem to be the most sensible approach. Acknowledgements This research was supported by the Dutch Health Care Insurance Council (CVZ) through the Department of Medical Technology Assessment (MTA) of the University Hospital Groningen, the Netherlands.

78

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

References for chapter four

1. Abramson JH, Gahlinger PM. Computer Programs for Epidemiologists PEPI Version 4.0. Sagebrush Press 2001; Salt Lake City, UT.

2. Bagga H, Greenfield DS, Feuer W, Knighton RW. Scanning laser

polarimetry with variable corneal compensation and optical coherence tomography in normal and glaucomatous eyes. Am J Ophthalmol 2003;135:521–529.

3. Cello KE, Nelson-Quigg JM, Johnson CA. Frequency doubling technology

perimetry for detection of glaucomatous visual field loss. Am J Ophthalmol 2000;129:314–322.

4. Dreher AW, Reiter K. Retinal laser ellipsometry: a new method for

measuring the retinal nerve fibre layer thickness distribution? Clin Vis Sci 1992;7:481–488.

5. Heeg GP, Blanksma LJ, Hardus PLLJ, Jansonius NM. The Groningen

Longitudinal Glaucoma Study. I. Baseline sensitivity, specificity of the frequency doubling perimeter and the GDx nerve fibre analyser. Acta Ophthalmol Scand 2005;83:46–52.

6. Heijl A, Leske MC, Bengtsson B, Hyman L, Bengtsson B, Hussein M, The

Early Manifest Glaucoma Trial Group Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial. Arch Ophthalmol 2002;120:1268–1279.

7. Horn FK, Nguyen NX, Mardin CY, Junemann AG. Combined use of

frequency doubling perimetry and polarimetric measurements of retinal nerve fibre layer in glaucoma detection. Am J Ophthalmol 2003;135: 160–168.

8. Johnson CA, Samuels SJ. Screening for glaucomatous visual field loss

with frequency doubling perimetry. Invest Ophthalmol Vis Sci 1997;38: 413–425.

9. Joson PJ, Kamantigue ME, Chen PP. Learning effects among perimetric

novices in frequency doubling technology perimetry. Ophthalmology 2002;109:757–760.

79

Chapter 4

10. Khong JJ, Dimitrov PN, Rait J, McCarty CA. Can the specificity of the FDT for glaucoma be improved by confirming abnormal results? J Glaucoma 2001;10:199–202.

11. Maddess T, Goldberg I, Dobinson J, Wine S, Welsh AH, James AC. Testing

for glaucoma with the spatial frequency doubling illusion. Vision Res 1999;39:4258–4273.

12. Maddess T, Henry GH. Performance of a non-linear pathway in subjects

with ocular hypertension and glaucoma. Clin Vis Sci 1992;7:371–382.

13. Quigley HA. Identification of glaucoma- related visual field abnormality with the screening protocol of frequency doubling technology. Am J Ophthalmol 1998;125:819–829.

14. Reus NJ, Colen TP, Lemij HG. Visualization of localized retinal nerve fibre

layer defects with the GDx with individualized and with fixed compensation of anterior segment birefringence. Ophthalmology 2003;110:1512– 1516.

15. Weinreb RN, Dreher AW, Coleman A, Quigley H, Shaw B, Reiter K.

Histopathologic validation of Fourierellipsometry measurements of retinal nerve fibre layer thickness. Arch Ophthalmol 1990;108:557–560.

16. Wolfs RC, Borger PH, Ramrattan RS et al. Changing views on open-angle

glaucoma: definitions and prevalences. The Rotterdam Study. Invest Ophthalmol Vis Sci 2000;41:3309–3321.

17. Zhou Q, Weinreb RN. Individualized compensation of anterior segment

birefringence during scanning laser polarimetry. Invest Ophthalmol Vis Sci 2002;43:2221–2228.

80

Strategies for improving the diagnostic specificity of the frequency doubling perimeter

81

Chapter 5 Glaucoma screening during regular optician visits: can the population at risk of developing glaucoma be reached?

83

Chapter 5

GLAUCOMA SCREENING DURING REGULAR OPTICIAN VISITS: CAN THE POPULATION AT RISK OF DEVELOPING GLAUCOMA BE REACHED? Remco Stoutenbeek and Nomdo M. Jansonius Abstract Aim: To determine the percentage of the population at risk of developing glaucoma, which can potentially be reached by conducting glaucoma screening during regular optician visits. Methods: 1200 inhabitants aged >40 years were randomly selected from Dutch community population databases. A questionnaire was mailed to these inhabitants with questions on their latest optician visit and risk factors for glaucoma. A second questionnaire was sent to their opticians, who were asked about their willingness to conduct an additional glaucoma screening programme in the future. Results: The questionnaire was returned by 959 of 1200 inhabitants and 37 of 50 opticians. The percentage of inhabitants who visited an optician during a 5-year period was 83% (95% confidence interval (CI) 80% to 85%). This percentage was adjusted for the presence of risk factors for glaucoma to obtain the percentage of the population at risk of developing glaucoma. The percentage of opticians willing to cooperate in a glaucoma screening programme extended beyond a non-contact tonometry measurement alone was 91% (95% CI 77% to 98%). Conclusion: By conducting glaucoma screening during regular optician visits, a large section of the population at risk of developing glaucoma can be reached.

84

Glaucoma screening during regular optician visits

5.1 Introduction Glaucoma is a chronic disease that may cause irreversible blindness. Owing to its insidious nature, the early stages of glaucoma do not cause any symptoms, treatment of glaucoma may arrest or slow down its progress.1 Therefore, screening might be advisable. The essential conditions required for a successful implementation of any screening programme were listed by Wilson and Jungner2 almost 40 years ago. With regard to glaucoma screening, several of these conditions have been met, whereas others are still under debate. One of the Wilson and Jungner criteria deals with economic feasibility, which largely depends on how a screening programme is organised. Currently, ophthalmologists carry out glaucoma screening on all patients. Although many patients with glaucoma are diagnosed in this way, only a limited proportion of the general population is being reached and, as a result, only about half of the people with glaucoma are detected.3-5 At the other end of the spectrum lies straightforward implementation of a large-scale screening programme. This presumably would allow for the largest proportion of the population to be screened, but it is expensive and requires substantial and ongoing logistic efforts. Opticians are a third possibility for glaucoma screening. In fact, they already carry out some screening by means of non-contact tonometry during regular patient visits. However, only about half of the patients with glaucoma have raised intraocular pressure at a single visit.6 Screening could be improved by adding a second test. For example, frequency doubling perimetry (Carl Zeiss Meditec, Dublin, California, USA) in screening mode offers a short and simple test. With an appropriate cut-off point, it combines high specificity (typically 95%) with acceptable sensitivity for early glaucoma (about 50%) and high sensitivity for moderate (about 90%) and severe (100%) glaucoma.7-8 A prerequisite for effective glaucoma screening is that a sufficient proportion of those at risk of developing glaucoma regularly visit an optician. Therefore, the first aim of the present study was to determine how often people at risk of developing glaucoma visit an optician. This was achieved by sending a questionnaire to a large randomly selected group of inhabitants of the city of Groningen and its surrounding rural area. As some people are more likely than others to develop glaucoma, questions on known risk factors for glaucoma were included in the questionnaire and used as weights in the analyses. A second prerequisite is the willingness of the participating opticians to extend the screening programme beyond non-contact tonometry. Thus, a second questionnaire was sent to the relevant opticians.

85

Chapter 5

5.2 Materials and methods Frequency of visiting an optician The ethics board of the University Medical Center Groningen, Groningen, The Netherlands, approved the study. In all, 1200 inhabitants aged >40 years were randomly selected from the community population databases of Groningen, Veendam, Roden and Siddeburen. Groningen is a city of 180 000 inhabitants. Veendam and Roden are towns with 28 000 and 15 000 inhabitants, respectively, both of which have several opticians. Siddeburen is a village situated in a rural area with 2000 inhabitants; there are several small supermarkets but no local optician. All inhabitants received a questionnaire in the post. Non-responders received a reminder after 6 weeks. Data collection was discontinued after 12 weeks. The questionnaire consisted of 17 questions, covering gender, age, time since latest optician and ophthalmologist visit, use of spectacles or contact lens, cataract surgery, and educational level. As we were interested in the proportion of potential patients with glaucoma that can be detected via the optician, we also asked about established and possible risk factors for glaucoma (myopia, family history of glaucoma, diabetes, hypertension, migraine and ethnicity).9-16 If a risk factor has a higher prevalence in the group that regularly visits an optician (visitors) than in the group that does not regularly visit an optician (non-visitors), then the actual proportion of patients with glaucoma that can be detected during regular optician visits exceeds the proportion as estimated from the raw visiting frequency data. Arbitrarily, we considered at least one visit in 5 years as regular. A shorter screening interval is unlikely to be efficient.4 Cooperation of participating opticians A second questionnaire was sent to all opticians visited by at least two inhabitants. They were asked about their willingness to conduct specific glaucoma screening during regular patient visits in the future, in addition to the non-contact tonometry carried out at present. A reasonable fee for this additional service was offered (ie, to cover costs but not to provide a profit). Names and locations of the opticians were obtained from the first questionnaire. Non-responders received a reminder after 4 weeks. Data collection was discontinued after 10 weeks. Analysis For the univariate analyses we used the x2 test for proportions, Yates corrected when appropriate, and the Mann–Whitney U test for continuous variables. For the multivariate analysis, logistic regression was carried out using SPSS V.12.0. Fisher’s exact 95% confidence intervals for proportions were calculated using PEPI V.4.1.17 p<0.05 was considered to be significant.

86

Glaucoma screening during regular optician visits

For logistic regression analysis, the parameter age was stratified into six groups comprising one decade per group. The parameter educational level was ranked on an ordinal scale from 1 (primary school) to 8 (university graduate). The parameter address was analysed by using two dummies ‘town’ and ‘village’. Likewise, refraction was analysed as ‘myopia’ and ‘hypermetropia’. With regard to the parameter ethnicity, we discriminated only between African origin and non-African origin. The dependent variable was visitor: optician visit yes or no during the 5 years before our study. 5.3 Results Frequency of visiting an optician The questionnaire was returned by 959 of 1200 inhabitants (80%). Table 5.3.1 shows a comparison of responders and non-responders. On average, responders were older and more often women. The response percentages were similar in all three subregions (city, town and village). Table 5.3.1 Demographics of responders and non-responders.

parameter responders

(n = 959) non-responders

(n = 241) p value

age, years* 58 (48 – 69) 55 (46 – 66) 0.011

female gender, % 53 45 0.019

city,% 78 22

town,% 82 18 0.452

village,% 80 20

*median (25th and 75th centiles). Figure 5.3.1 shows the frequencies of visits to opticians and ophthalmologists. Of the 959 inhabitants who returned the questionnaire, 764 (80%, 95% CI 77% to 82%) visited an optician during the 5 years before our study; 359 (37%, 95% CI 34% to 41%) visited an ophthalmologist during this period; and 804 (84%, 95% CI 81% to 86%) visited one or the other.

87

Chapter 5

Figure 5.3.1 Distributions of time since most recent optician visit (A) and most recent visit to an ophthalmologist (B). Table 5.3.2 shows univariate comparisons of all investigated parameters (demographics and self-reported prevalences of risk factors) between visitors (defined as latest optician visit ≤5 years) and non-visitors. Table 5.3.3 shows the corresponding multivariate analysis. A significant difference between visitors and non-visitors was found for the parameters myopia, hypermetropia, age and previous cataract surgery. Of these, only myopia and age are presumed risk factors for open-angle glaucoma.

88

Glaucoma screening during regular optician visits

Table 5.3.2 Demographics and glaucoma risk factors for responders, stratified for time since most recent optician visit, univariate analyses.

parameter visitors

(≤5 years; n = 764) non-visitors

(>5 years; n = 195) p value

myopia, % 31 6 <0.001

hypermetropia, % 34 7 <0.001

age, years* 60 (51 – 70) 51 (44 – 64) <0.001

male gender, % 44 55 0.007

cataract surgery, % 8 4 0.077

hypertension, % 29 23 0.114

village, % 33 38 0.199

town, % 34 30 0.279

city, % 33 32 0.888

migraine, % 28 25 0.410

family history, % 9 10 0.699

diabetes, % 8 8 0.888

educational level* 3 (2 – 4) 3 (2 – 5) 0.969

ethnicity, % 1 1 1.000

*Median (between 25th and 75th centiles).

89

Chapter 5

Table 5.3.3 Demographics and glaucoma risk factors for responders, multivariate analysis with latest optician visit ≤5 years as the dependent variable

parameter odds ratio p value 95% CI

myopia 17.5 <0.001 9.3 to 33.0

hypermetropia 12.7 <0.001 7.0 to 22.9

cataract surgery 4.7 0.001 2.0 to 11.1

age 1.2 0.043 1.0 to 1.4

family history 0.6 0.174 0.3 to 1.2

gender 0.8 0.176 0.5 to 1.1

village 1.2 0.399 0.8 to 1.9

ethnicity 1.9 0.469 0.3 to 10.2

hypertension 1.2 0.477 0.7 to 1.9

diabetes 0.8 0.574 0.4 to 1.6

town 1.1 0.719 0.7 to 1.7

migraine 1.1 0.774 0.7 to 1.6

educational level 1.0 0.836 0.9 to 1.1

*Median (between 25th and 75th centiles). On the basis of an estimated relative risk of 2 for people with myopia of developing glaucoma,9-11 we calculated that the visitors (80% of total) represent 83% (80–85%) of the population at risk for developing glaucoma. Inhabitants who visited either an optician or an ophthalmologist (84% of total) represent 86% (84–88%) of the population at risk. We did not adjust for age in our analysis as

90

Glaucoma screening during regular optician visits

although the prevalence and incidence of glaucoma increases with age,4,5 younger people presumably benefit more from screening. Cooperation of participating opticians Of the 959 inhabitants who returned our questionnaire, 826 had visited an optician at least once; 770 remembered the name of the optician last visited. Of the 50 opticians, 37 (74%) returned the questionnaire. Of these 37 opticians, 32 (91%, 95% CI 77% to 98%) expressed willingness to participate in an extended glaucoma screening programme. 5.4 Discussion We investigated whether glaucoma screening during regular optician visits could be a suitable way to reach the population at risk for glaucoma. We found that 83% (95% CI 80% to 85%) of the population at risk for glaucoma visited an optician during a 5-year period. Most opticians expressed willingness to cooperate in some kind of extended screening programme for patients with glaucoma. Most of the criteria for appraising the validity of a screening programme as stated by Wilson and Jungner2 have been met. The fact that screening during a regular optician visit seems to be feasible, as shown by our study, is a step forward with regard to Wilson and Jungner’s criterion of economic feasibility. However, the extra clinical workload for ophthalmologists resulting from any screening programme must also be taken into account. Shifting the cut-off criterion for any screening test towards a high specificity is an important first measure and, hence, it should be accepted that some cases will be missed. Finally, it is the community that determines the amount of money that may be spent to prevent a single case of visual impairment or blindness. In the case of limited resources, this should be weighed against other beneficial measures that could be adopted for the same cost. In conclusion, by conducting glaucoma screening during regular optician visits, a large section of the population at risk of developing glaucoma can be reached.

91

Chapter 5

References for chapter five

1. Heijl A, Leske MC, Bengtsson B, et al. Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial Arch Ophthalmol 2002;120:1268–79.

2. Wilson JM, Jungner G. Principles and practice of screening for disease.

Geneva; World Health Organization 1968:1–163.

3. Hollows FC, Graham PA. Intra-ocular pressure, glaucoma, and glaucoma suspects in a defined population. Br J Ophthalmol 1966;50:570–86.

4. De Voogd S, Ikram MK, Wolfs RC, et al. Incidence of open-angle glaucoma

in a general elderly population: the Rotterdam Study. Ophthalmology 2005;112:1487–93.

5. Wolfs RC, Borger PH, Ramrattan RS, et al. Changing views on open-angle

glaucoma: definitions and prevalences—the Rotterdam Study. Invest. Ophthalmol Vis Sci 2000;41:3309–21.

6. Katz J, Tielsch JM, Quigley HA, et al. Automated suprathreshold screening

for glaucoma: the Baltimore Eye Survey. Invest Ophthalmol Vis Sci 1993;34:3271–7.

7. Trible JR, Schultz RO, Robinson JC, et al. Accuracy of glaucoma detection

with frequency-doubling perimetry. Am J Ophthalmol 2000;129:740–5.

8. Stoutenbeek R, Heeg GP, Jansonius NM. Frequency doubling perimetry screening mode compared to the full-threshold mode. Ophthalmic Physiol Opt 2004;24:493–7.

9. Mitchell P, Hourihan F, Sandbach J, et al. The relationship between

glaucoma and myopia: the Blue Mountains Eye Study. Ophthalmology 1999;106:2010–15.

10. Wong TY, Klein BE, Klein R, et al. Refractive errors, intraocular pressure,

and glaucoma in a white population. Ophthalmology 2003;110:211–17.

11. Grodum K, Heijl A, Bengtsson B. Refractive error and glaucoma. Acta Ophthalmol Scand 2001;79:560–6.

92

Glaucoma screening during regular optician visits

93

12. Wolfs RC, Klaver CC, Ramrattan RS, et al. Genetic risk of primary open-angle glaucoma. Population-based familial aggregation study. Arch Ophthalmol 1998;116:1640–5.

13. Bonovas S, Peponis V, Filioussi K. Diabetes mellitus as a risk factor for

primary open-angle glaucoma: a meta-analysis. Diabet Med 2004;21: 609–14.

14. Mitchell P, Lee AJ, Rochtchina E, et al. Open-angle glaucoma and

systemic hypertension: the Blue Mountains Eye Study. J Glaucoma 2004; 13:319–26.

15. Wang JJ, Mitchell P, Smith W. Is there an association between migraine

headache and open-angle glaucoma? Findings from the Blue Mountains Eye Study. Ophthalmology 1997;104:1714–19.

16. Racette L, Wilson MR, Zangwill LM, et al. Primary open-angle glaucoma in

blacks: a review. Surv Ophthalmol 2003;48:295–313.

17. Abramson JH, Gahlinger PM. Computer programs for epidemiologists: PEPI version 4.1. Salt Lake City, UT: Sagebrush Press, 2001:43–9.

Chapter 6 The additional yield of a periodic screening programme for open-angle glaucoma: a population-based comparison of incident glaucoma cases detected in regular ophthalmic care with cases detected during screening

95

Chapter 6

THE ADDITIONAL YIELD OF A PERIODIC SCREENING PROGRAMME FOR OPEN-ANGLE GLAUCOMA: A POPULATION-BASED COMPARISON OF INCI-DENT GLAUCOMA CASES DETECTED IN REGULAR OPHTHALMIC CARE WITH CASES DETECTED DURING SCREENING R Stoutenbeek,1,2 S de Voogd,2 R C W Wolfs,2,3 A Hofman,2 P T V M de Jong,2,4,5 N M Jansonius1,2 1University Medical Center Groningen, University of Groningen, Groningen, The Netherlands; 2 Department of Epidemiology & Biostatistics, Erasmus Medical Center, Rotterdam, The Netherlands; 3 Department of Ophthalmology, Erasmus Medical Center, Rotterdam, The Netherlands; 4 The Netherlands Institute for Neuroscience, RAAS, Amsterdam, The Netherlands; 5 Department of Ophthalmology, Academic Medical Center, Amsterdam, The Netherlands Abstract Aim: To study the additional yield of a periodic screening programme for open-angle glaucoma (OAG) by comparing, in a population-based setting, incident OAG (iOAG) cases detected in regular ophthalmic care with those detected during screening. Methods: Participants aged 55 and over from the population-based Rotterdam Study underwent the same ophthalmic examination at baseline (1991–3) and follow-up (1997–9), including visual field testing and simultaneous stereo optic disc photography. Of 3842 participants, 87 (2.3%) developed iOAG during a mean follow-up time of 6.5 years. Of these 87 iOAG cases, 78 (90%) were included in this study. Results: Of the 78 iOAG cases detected at follow-up, 23 (29%) had already been detected before during regular ophthalmic care. The remaining 55 (71%) undetected iOAG cases more often showed glaucomatous optic neuropathy without glaucomatous visual field loss (29 of 55 (53%)) as compared with the detected cases (four of 23 (17%); p = 0.009). Of the undetected iOAG cases, only four had developed significant visual field loss in their better eye. Conclusion: The additional yield of a periodic OAG screening programme is lower than expected from published prevalence data. In the discussion, the authors estimate that—in a white population with a low prevalence of pseudoexfoliation—about one in 1000 screened persons could be saved from bilateral end-stage OAG.

96

The additional yield of a periodic screening programme for open-angle glaucoma

6.1 Introduction Open-angle glaucoma (OAG) is a chronic disease leading to glaucomatous optic neuropathy (GON) and causing irreversible blindness. Due to its insidious nature, the early stages of OAG do not cause any symptoms. Treatment of OAG can arrest or slow its progress.1,2 Therefore, screening is often noted as an option to reduce the OAG burden. Patients detected by a periodic screening programme can differ from those detected in another way. One reason for this is that cases with a severe and progressive variant of a disease are more likely to be detected before their next periodic screening test takes place. This phenomenon is described as ‘length bias’ and is a major source of ongoing debate on the benefit of screening programmes.3-5 Many OAG cases are currently detected by case finding during a visit to an optician or ophthalmologist (regular ophthalmic care). However, only about half of all OAG cases are detected in this way.6,7 This discovery is often used as an argument in favour of starting a periodic OAG screening programme. Whether such a programme in addition to the current practice of case finding is indeed an efficient approach, however, depends— inter alia—on length bias. If the currently undetected cases are mainly elderly patients with a slowly progressive form of the disease at its early stages, a periodic screening programme is unlikely to be efficient for blindness prevention. The aim of the present study was to investigate the additional yield of a periodic OAG screening programme. For this purpose, we used data from the population-based Rotterdam Study.8 In the ophthalmic part of this study, participants were examined on two different occasions in order to determine the incidence of OAG.9 Since the average interval between the baseline and follow-up examinations was 6.5 years (range 5.0–9.4 years), our study applies to a hypothetical situation in which a population is screened periodically at 6.5- year intervals. We compared iOAG cases (cases with OAG at follow-up but not at baseline) that had already been detected in regular ophthalmic care before the follow-up examination with those who remained undetected until the follow-up examination. From these data, we estimated the number of persons that could be saved from bilateral end-stage OAG by screening. 6.2 Methods Study population In the ophthalmic section of the Rotterdam Study— a prospective, population-based cohort study of all residents aged 55 and older living in a district of

97

Chapter 6

Rotterdam—home interviews and ophthalmic examinations at an examination centre were conducted after approval by the Medical Ethics Committee of the Erasmus University Rotterdam.7,10 All participants gave written informed consent. After the baseline examination between 1991 and 1993, the first follow-up examination for OAG was performed from 1997 to 1999. The average time to follow-up was 6.5 years. The ophthalmic part of the Rotterdam Study included 6773 participants (response rate 78%). Baseline examination identified 221 cases (3.2%) with OAG in at least one eye.9 Of the 6552 (6773 – 221) participants at risk of developing OAG during follow-up, 3842 (59%) completed the follow-up examination. Of the 2710 (6552 – 3842; 41%) non-participants, 1244 (19%) had died, and 1466 (22%) were unable or unwilling to participate (see Discussion). There were 87 cases (2.3%) with iOAG in at least one eye at follow-up (for definitions of OAG and iOAG, see below). 9 Of these 87 cases, six had already been referred to an ophthalmologist at baseline because of an elevated IOP without further signs of OAG at that moment. These six cases were excluded from this analysis (see Discussion), as were three cases with missing visual-field (VF) data (see Discussion), leaving 78 iOAG cases. Data collection The ophthalmic examination, identical at baseline and follow-up, included Goldmann applanation tonometry, direct and indirect ophthalmoscopy, and simultaneous stereoscopic fundus photography in pharmacological mydriasis. The VF of each eye was screened using a 52-point supra-threshold test that covered the central field with a radius of 24° (Humphrey Field Analyzer 640; Carl Zeiss Meditec, Dublin, CA). When the participant did not respond to the light stimulus (6 dB above a threshold-related estimate of the hill of vision) in at least three contiguous test points (or four including the blind spot), visual field loss (VFL) was considered as present. If the first VF test was unreliable (>33% false positive or false-negative catch trials) or a reliable test showed VFL in at least one eye, a second supra-threshold test was performed on that eye. When VFL remained present on the second supra-threshold test, or if the test was unreliable again, Goldmann kinetic perimetry (Haag Streit, Bern, Switzerland) was performed on both eyes by a skilled perimetrist. VFL on Goldmann perimetry was considered to be glaucomatous visual-field loss (GVFL) only after excluding all other possible causes.10 Simultaneous stereo colour transparencies (20°) of the optic disc were digitised and analysed with a semiautomatic image analyser (ImageNet; Topcon Optical Company, Tokyo) to obtain vertical cup-to-disc ratio (VCDR) and minimal neural rim widths. Cut-off values for GON were based on the 99.5 percentile (probable GON) and 97.5 percentile (possible GON) of the distribution in this population, preferentially on ImageNet data.7 Probable GON was defined as a VCDR ≥0.8, or asymmetry between both eyes ≥0.3, or minimal neural rim width <0.05 with

98

The additional yield of a periodic screening programme for open-angle glaucoma

ImageNet. Where no ImageNet data were available (six of 78), ophthalmoscopically determined probable GON was defined as a VCDR ≥0.9, or asymmetry between both eyes ≥0.3. Minimal rim widths were not quantified by ophthalmoscopy. Likewise, possible GON was defined as VCDR ≥0.7, or asymmetry ≥0.2 for both ImageNet and ophthalmoscopy, or minimal neural rim width <0.10 (ImageNet criterion only). OAG cases were classified into four subgroups—(1) definite OAG (both GVFL and any GON), (2) probable OAG based on GVFL only, (3) probable OAG based on probable GON only, and (4) possible OAG based on possible GON only.7 iOAG was defined as no or possible OAG at baseline and probable or definite OAG at follow-up. Excluded from this incidence definition were participants with possible GON without GVFL at baseline and probable GON without GVFL at follow-up because they were considered not to have changed sufficiently to count as iOAG cases.9 At follow-up, iOAG cases were classified as detected cases if they were using intraocular pressure (IOP) lowering medication and/or stated that they were visiting an ophthalmologist regularly for glaucoma-related reasons. Medication use was verified through an automated computer database of all medical prescriptions from the seven pharmacies located in the district where the study was conducted, from 1991 onwards. Patient records were reviewed for verification of glaucoma-related visits. The remaining iOAG cases were classified as undetected cases. Analysis We used the chi-squared test for proportions, with Yates correction where appropriate. The Student t test and Mann– Whitney U test were used for continuous variables (SPSS 12.0.2 for Windows, SPSS, Chicago). 6.3 Results Of the 78 iOAG cases, 23 (29%) had already been detected during regular ophthalmic care before their follow-up examination (detected cases), whereas 55 (71%) had remained undetected (undetected cases). Table 6.3.1 shows the characteristics of the detected and undetected iOAG cases, with univariate comparisons. Of the iOAG cases detected prior to the follow-up examination, 12 of 23 (52%) had IOP-lowering treatment at follow-up. This is reflected in the tendency towards a lower mean IOP at follow-up as compared with baseline in this group (table 6.1; 15.8 vs 14.4 mm Hg; p=0.083).

99

Chapter 6

Tabel 6.3.1 Characteristics (at baseline, unless otherwise indicated) of detected and undetected cases with incident open-angle glaucoma.

characteristics detected cases

(n = 23) undetected cases

(n = 55) p value

age at follow-up, years* 77.0 (6.0) 74.0 (7.4) 0.091

male gender† 43.5 49.1 0.841

IOP,‡ mm HG* 15.8 (3.1) 15.6 (3.4) 0.854

IOP‡ at follow-up, mmHg* 14.4 (3.0) 15.5 (3.9) 0.235

emmetropia†,§ 26.1 25.5 0.823

family history of OAG† 39.1 1.8 <0.001

diabetes mellitus† 4.3 7.5 1.000

educational level*,¶ 3.0 (2.0) 3.8 (2.0) 0.100

vertical cup-disc ratio *, ** 0.66 (0.09) 0.56 (0.14) 0.003

possible GON 8/23 10/55 0.345

*Mean (SD) †Percentage ‡Mean of both eyes §Between -0.99 and +0.99 D spherical equivalent ¶Ordinal scale from 1 (primary school) to 8 (university) **Larger value of both eyes IOP = intraocular pressure; OAG = open-angle glaucoma; GON = glaucomatous optic neuropathy. Table 6.3.2 presents a comparison of detected and undetected cases with respect to the severity of iOAG. The distributions are significantly different (p=0.001). Of the 55 undetected cases, only 26 (47%) had GVFL (definite OAG or probable OAG based on GVFL only); the remaining 29 undetected cases had GON only. Of the 23 detected cases, 19 (83%) had GVFL (p=0.009). Of all cases with GVFL, 19 of 45 (42%) were detected in regular ophthalmic care before their follow-up examination.

100

The additional yield of a periodic screening programme for open-angle glaucoma

Tabel 6.3.2 Distribution of detected and undetected cases with incident open-angle glaucoma according to severity of disease

severity of glaucoma detected

cases (n = 23)

undetected cases

(n = 55)

P value

definite OAG 13 13

probable OAG based on GVFL only 6 13

based on GON only 4 29

0.001

GON = glaucomatous optic neuropathy; GVFL = glaucomatous visual-field loss (for definitions, see Methods); OAG = open-angle glaucoma. We performed a more detailed comparison within the subgroup of iOAG cases with GVFL (n=45) between detected (n=19) and undetected (n=26) cases, based on their second supra-threshold VF test score (table 6.3.3; number of missed points). Calculations were performed both for the better eye (the eye with the lower number of missed points) and for the worse eye. There was no significant difference between detected and undetected cases in the number of missed points for either the better (p=0.65) or the worse eye (p=0.98).

101

Chapter 6

Tabel 6.3.3 Number of missed points on supra-threshold visual field test for detected and undetected cases with incident glaucomatous visual-field loss (n=45).

eye percentile detected

cases (n = 19)

undetected cases

(n = 26) p value

p25 1 1

p50 4 3 better eye

p75 7 8

0.65

p25 8 9

p50 14 14 worse eye

p75 20 19

0.98

Better eye = eye with the lower number of missed points on VF test; p25 = 25th percentile (25% of cases have a number of missed points lower than or equal to this value); p50 = median; p75 = 75th percentile; worse eye, = eye with the higher number of missed points on VF test. From the perspective of blindness prevention, the severity of GVFL in the undetected cases is the most important measure of the additional screening yield. Figure 6.3.1 shows the distributions of VF scores for the undetected cases with GVFL (n=26) for the better and the worse eyes. As can be seen in this figure, only four of these cases had >10 missed points in their better eyes.

102

The additional yield of a periodic screening programme for open-angle glaucoma

Figure 6.3.1 Number of missed points on supra-threshold visual field (VF) screening test for undetected cases with incident glaucomatous visual-field loss (n=26) for the better and the worse eye. For definitions of better and worse eye, see legend to table 6.3.3. 6.4 Discussion Incident OAG cases that remained undetected until the follow-up examination more often had iOAG based on GON only as compared with cases already detected by regular ophthalmic care (table 6.3.2). This suggests that the detected iOAG cases had a faster progressing disease, indicating the presence of length bias in OAG screening. With a more conservative definition of OAG, however, requiring the presence of GVFL, no differences between the two groups were found (table 6.3.3). More important, as can be seen in table 6.3.1, a small albeit significant difference in vertical cup–disc ratio between detected and undetected iOAG cases was already visible at baseline. This difference precludes a final statement on the presence or absence of length bias in glaucoma screening. The average vertical cup–disc ratio at baseline in the Rotterdam Study was 0.52, with a standard deviation of 0.15 (n=6552; larger value of both eyes). As mentioned in the Methods, 1466 of 6552 participants (22%) were unable or unwilling to participate in the follow-up examination. Non-participants were older, more often female and had more often a history of stroke or dementia.9 Most important for this study, the mean VCDR at baseline was not larger in the non-participants (0.51) when compared with the participants (0.53).

103

Chapter 6

The most obvious difference between the detected and undetected iOAG cases is the presence of a positive family history of glaucoma (table 6.3.1). A possible explanation of this difference is an elevated visit rate by relatives of OAG patients to opticians or ophthalmologists, as is widely recommended. Of all participants at baseline, 8.5% had a positive family history of glaucoma. Of the 3842 participants at risk of OAG at baseline, 87 developed iOAG during a follow-up period of 6.5 years. Of the 78 iOAG cases included in this study, 23 had already been diagnosed by an ophthalmologist before the follow-up examination took place. Only four of the remaining 55 undetected cases had >10 missed points (arbitrarily chosen) in their better eye on VF testing (fig 6.3.1). If we tentatively assume that these four iOAG cases would become blind without treatment while the other undetected cases would retain useful vision (see also next paragraph), it can be calculated that about 1000 OAG screening tests would have to be performed in order to prevent one OAG case from becoming severely visually impaired or blind (about 200 tests if we would start this discussion with >10 missed points in the worse eye, aiming to prevent unilateral loss). Assuming a test specificity of 95% for a suitable OAG screening test, 1000 screening tests will produce 50 false-positive test results, all needing further investigation. At first glance, incorporating prior selection criteria like myopia or family history might improve the feasibility of screening.11–17 However, only a minority of the undetected cases (18 of 55) had an OAG risk factor (17 of 55 myopia; 1 of 55 positive family history; see also table 6.3.1). The use of supra-threshold testing, as performed in our study, might have resulted in missing cases with early glaucoma. It is unlikely, however, that cases with moderate or severe glaucoma, the most important ones for the estimate as presented here, have been overlooked.18 We re-estimated the number of undetected iOAG cases that would develop end-stage OAG before death, now using data from the study by Wilson et al.19 They analysed progression of VFL in untreated black OAG cases in the West Indies. Following the AGIS scoring system,20 7% of their cases had developed end-stage OAG in both eyes after a follow-up of 10 years. The average life expectancy at follow-up of our 55 undetected cases was 11.4 years.21 From this it can tentatively be estimated that, given the higher prevalence of OAG in blacks, 22 four at most (7% of 55) of 3842, that is 0.1% of the white participants, might have become blind before dying if they had remained undetected. Obviously, this is only a rough estimate. For example, we assumed that the disease severity of our 55 undetected iOAG cases at follow-up could be compared with the baseline findings of Wilson et al,1,9 and we based our estimate on average life expectancy, assuming that the lower incidence of glaucoma blindness in those dying earlier than average balances the higher incidence in those living longer than average. Both our estimates suggest, however, that the real yield of a periodic OAG screening programme in terms of preventing severe visual impairment or

104

The additional yield of a periodic screening programme for open-angle glaucoma

blindness (estimated to be 0.1%) is much lower than the yield as estimated from the prevalence of undetected OAG (typically 1%).6,7

As mentioned in the Methods section, six iOAG cases were excluded because at baseline, they were referred to an ophthalmologist because of an elevated IOP without further signs of OAG at that stage. From a methodological point of view, it might have been better not to report any baseline abnormality to these participants. However, ethically, and in accordance with the guidelines for unexpected findings in the Rotterdam Study, this would have been unacceptable. Of inhabitants aged 40 years and older and living in The Netherlands, 84% visit an optician or an ophthalmologist at least once every 5 years,23 and most opticians perform noncontact tonometry in clients in this age group. Hence, at least some if not most of these six iOAG cases with elevated IOP at baseline would have been detected before the follow-up measurement. Likewise (see Methods), three iOAG cases had to be excluded because of missing VF data. These three cases were all undetected at follow-up. Based on disc data, all three cases presumably had unilateral disease. Within the remaining 78 cases, only one had unreliable perimetric results according to the criteria as listed in the Methods section. This case was not excluded from the analyses. Grodum et al. compared 402 OAG cases identified through a large population screening among elderly citizens of Malmo, Sweden, with 354 cases identified through retrospective patient record analysis at the Eye Department of Malmo University Hospital.24 They found that the latter group had considerably more VFL than the former group. This cross-sectional finding provides information on the screening yield if a population is screened for the first time. Our data are complementary to their findings—based on incident cases only, our data provide information on the ongoing yield of screening if a population is screened periodically. Most of the participants of the Rotterdam Study were white. As the prevalence of OAG depends on ethnicity,25 our findings should not automatically be applied to non-white populations. We did not exclude cases with pseudoexfoliation (hence OAG rather than POAG), but none of the iOAG cases found at follow-up suffered from pseudoexfoliation. The latter finding suggests that pseudoexfoliation is of little importance to the OAG burden in The Netherlands. This might be different elsewhere.26,27 Due to the design of the Rotterdam Study, we were unable to address the additional yield of screening in a younger population or in a rural area directly. A recent study from Finland suggested that screening is unlikely to be efficient under 55 years of age.28 In an earlier study, we did not find any difference in optician/ophthalmologist visit frequency between urban and rural area.23 That study reported that of inhabitants aged 40 years and older and living in The Netherlands, 84% visit an optician or an ophthalmologist at least once every five

105

Chapter 6

years. This apparently high percentage may have contributed to the poor additional yield as found in this study. In summary, the additional yield of periodic OAG screening is less than expected from published prevalence data because (1) many cases had already been detected at early disease stages in regular ophthalmic care and (2) only a minority of the undetected cases had severe enough disease to be seriously at risk of reaching end-stage OAG in both eyes during life—were they to remain undetected. Funding The Netherlands Organization for Health Research and Development (ZonMw) grant 2200.0035, The Hague. Foundations: Stichting Nederlands Oogheelkundig Onderzoek, Nijmegen/Rotterdam; Optimix, Amsterdam; Netherlands Organisation for Scientific Research (NWO), The Hague; Physico Therapeutic Institute, Rotterdam; Blindenpenning, Amsterdam; Sint Laurens Institute, Rotterdam; Bevordering van Volkskracht, Rotterdam; Blindenhulp, The Hague; Algemene Nederlandse Vereniging ter Voorkoming van Blindheid, Doorn; Rotterdamse Blindenbelangen Association, Rotterdam; OOG, The Hague; kfHein, Utrecht; Prins Bernhard Cultuurfonds, Amsterdam; Van Leeuwen Van Lignac, Rotterdam. All in The Netherlands. Unrestricted grants were obtained from Topcon Europe BV, Capelle aan de IJssel, The Netherlands, and from Heidelberg Engineering, Dosselheim, Germany.

106

The additional yield of a periodic screening programme for open-angle glaucoma

References for chapter six

1. Heijl A, Leske MC, Bengtsson B, et al. Early Manifest Glaucoma Trial Group. Reduction of intraocular pressure and glaucoma progression: results from the Early Manifest Glaucoma Trial. Arch Ophthalmol 2002;120:1268–79.

2. Maier PC, Funk J, Schwarzer G, et al. Treatment of ocular hypertension

and open angle glaucoma: meta-analysis of randomised controlled trials. BMJ 2005;331:134–9.

3. Wilson JM, Jungner G. Principles and practice of screening for disease.

34th edn. Geneva: World Health Organization, 1968:1–163.

4. Johnson GJ, Minassian DC, Weale RA, et al. The epidemiology of eye disease. Second edn. London: Hodder Arnold, 2003.

5. Baum M. Breast cancer screening comes full circle. J Natl Cancer Inst

2004;96:1490–1.

6. Hollows FC, Graham PA. Intra-ocular pressure, glaucoma, and glaucoma suspects in a defined population. Br J Ophthalmol 1966;50:570–86.

7. Wolfs RC, Borger PH, Ramrattan RS, et al. Changing views on open-angle

glaucoma: definitions and prevalences—The Rotterdam Study. Invest Ophthalmol Vis Sci 2000;41:3309–21.

8. Hofman A, Grobbee DE, de Jong PT, et al. Determinants of disease and

disability in the elderly: the Rotterdam Elderly Study. Eur J Epidemiol 1991;7:403–22.

9. De Voogd S, Ikram MK, Wolfs RC, et al. Incidence of open-angle glaucoma

in a general elderly population: the Rotterdam Study. Ophthalmology 2005;112:1487–93.

10. Ramrattan RS, Wolfs RC, Panda-Jonas S, et al. Prevalence and causes of

visual field loss in the elderly and associations with impairment in daily functioning: the Rotterdam Study. Arch Ophthalmol 2001;119:1788–94. [Erratum in: Arch Ophthalmol 2002;120:525]

107

Chapter 6

11. Mitchell P, Hourihan F, Sandbach J, et al. The relationship between glaucoma and myopia: the Blue Mountains Eye Study. Ophthalmology 1999;106:2010–15.

12. Grodum K, Heijl A, Bengtsson B. Refractive error and glaucoma. Acta

Ophthalmol Scand 2001;79:560–6.

13. Wong TY, Klein BE, Klein R, et al. Refractive errors, intraocular pressure, and glaucoma in a white population. Ophthalmology 2003;110:211–17.

14. Tielsch JM, Katz J, Sommer A, et al. Family history and risk of primary

open angle glaucoma. The Baltimore Eye Survey. Arch Ophthalmol 1994;112:69–73.

15. Wolfs RC, Klaver CC, Ramrattan RS, et al. Genetic risk of primary open-

angle glaucoma. Population-based familial aggregation study. Arch Ophthalmol 1998;116:1640–5.

16. Mitchell P, Rochtchina E, Lee AJ, et al. Bias in self-reported family history

and relationship to glaucoma: the Blue Mountains Eye Study. Ophthalmic Epidemiol 2002;9:333–45.

17. Le A, Mukesh BN, McCarty CA, et al. Risk factors associated with the

incidence of open-angle glaucoma: the visual impairment project. Invest Ophthalmol Vis Sci 2003;44:3783–9.

18. Topouzis F, Coleman AL, Yu F, et al. Sensitivity and specificity of the 76-

suprathreshold visual field test to detect eyes with visual field defect by Humphrey threshold testing in a population-based setting: the Thessaloniki eye study. Am J Ophthalmol 2004;137:420–5.

19. Wilson MR, Kosoko O, Cowan CL Jr, et al. Progression of visual field loss in

untreated glaucoma patients and glaucoma suspects in St. Lucia, West Indies. Am J Ophthalmol 2002;134:399–405.

20. AGIS Investigators. Advanced Glaucoma Intervention Study. 2. Visual

field test scoring and reliability. Ophthalmology 1994;101:1445–55.

21. Statistics Netherlands. Life expectancy 1996–2000. http://www.cbs.nl (accessed 14 Nov 2006).

22. Leske MC, Connell AM, Schachat AP, et al. The Barbados Eye Study.

Prevalence of open angle glaucoma. Arch Ophthalmol 1994;112:821–9.

108

The additional yield of a periodic screening programme for open-angle glaucoma

109

23. Stoutenbeek R, Jansonius NM. Glaucoma screening during regular optician visits: can the population at risk of developing glaucoma be reached? Br J Ophthalmol 2006;90:1242–4.

24. Grodum K, Heijl A, Bengtsson B. A comparison of glaucoma patients

identified through mass screening and in routine clinical practice. Acta Ophthalmol Scand 2002;80:627–31.

25. Tielsch JM, Sommer A, Katz J, et al. Racial variations in the prevalence of

primary open-angle glaucoma. The Baltimore Eye Survey. JAMA 1991;266:369–74.

26. Ball SF. Exfoliation syndrome prevalence in the glaucoma population of

South Louisiana. Acta Ophthalmol Suppl 1988;184:93–8.

27. Lindblom B, Thorburn W. Observed incidence of glaucoma in Halsingland, Sweden. Acta Ophthalmol (Copenh) 1984;62:217–22.

28. Vaahtoranta-Lehtonen H, Tuulonen A, Aronen P, et al. Cost effectiveness

and cost utility of an organized screening programme for glaucoma. Acta Ophthalmol Scand 2007;85:508–18.

Chapter 7 Supra-threshold perimetry compared to standard automated perimetry in glaucoma

111

Chapter 7

SUPRA-THRESHOLD PERIMETRY COMPARED TO STANDARD AUTOMATED PERIMETRY IN GLAUCOMA Remco Stoutenbeek,1,2 Johanna M.M. Hooymans,1 Nomdo M. Jansonius.1,2 1University Medical Center Groningen, University of Groningen, Groningen, The Netherlands; 2 Department of Epidemiology & Biostatistics, Erasmus Medical Center, Rotterdam, The Netherlands; Abstract Purpose: To compare the severity of glaucomatous damage measured with supra-threshold perimetry (STP) to the severity measured with standard automated perimetry (SAP). Design: Prospective, cross-sectional study. Methods: One hundred thirty-one (131) glaucoma patients and glaucoma suspect patients scheduled for SAP (Humphrey Field Analyzer 24-2 SITA Standard) were invited to undergo two additional 52-points STP tests in one eye. Ninety-two (92) patients were included. Eyes with non-glaucomatous visual field loss were not eligible, as well as eyes with best corrected visual acuity <0.5 (20/40) except when caused by glaucoma. If both eyes were eligible, one eye was randomly selected. Test sequence was also randomized. The Mean Deviation (MD) of the SAP test was compared to the number of missed points on STP. Results: There was a linear relationship between the number of missed points on STP and the SAP MD, with a correlation coefficient of -0.92. The MD can be estimated from the STP score by multiplying the number of missed points by -0.75. STP was nearly twice as fast as SAP. Conclusion: STP appears to be a fast and reliable method for estimating the severity of glaucomatous damage.

112

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

7.1 Introduction Standard Automated Perimetry (SAP) has become the most commonly used form of perimetry for glaucoma diagnosis and follow-up. Commonly used SAP devices are the Humphrey Field Analyzer (HFA, Carl Zeiss Meditec Inc., Dublin, CA, USA) and the Octopus perimeter (Haag-Streit AG, Koeniz, Switzerland), among others. In SAP, the threshold luminance of every test location in the visual field is determined by a staircase procedure: the stimulus luminance is either increased or decreased stepwise until a change occurs in the response of the subject.1,2 Initially, the HFA employed a full-threshold strategy, wherein the staircase step size is 4dB until the threshold is crossed. After the first crossing, the step direction reverses and a second crossing is performed with a step size of 2dB. The second crossing occurs in either an ascending or descending direction, and threshold is designated as the last-seen stimulus luminance. Full-threshold testing accurately determines threshold luminance, but at the expense of a long examination duration (typically 15 minutes per eye).3 In an effort to reduce test duration, FASTPAC was introduced around 1993, using 3dB steps and no reversals.4,5 A more sophisticated algorithm was introduced in 1997, the Swedish Interactive Threshold Algorithm (SITA). It reduces test time by taking into account information from nearby test locations to estimate threshold luminance, as well as by optimized test pacing and several other optimizations.6,7 Despite these improvements, SAP remains a difficult and time-consuming task that can be too demanding for some patients. Alternatives are desirable. Goldmann kinetic perimetry is easier for the patient, but equally time-consuming and it relies, more than SAP does, on the perimetric skills of the examiner. SAP with a larger stimulus (size V) reduces test-retest variability, which may be an indication that the test is easier to perform,8 but, since SITA is not available for unconventional stimulus sizes, testing time is inevitably long. Modern techniques such as Frequency Doubling Technology perimetry (FDT, Carl Zeiss Meditec Inc., Dublin, CA, USA) are also of interest. However, there are not enough longitudinal data available to decide whether FDT is suitable for glaucoma follow-up. Furthermore, FDT requires the purchase of an additional device. Yet another option is supra-threshold perimetry (STP), which will be explored in this study. In STP, stimuli are presented at an intensity several decibels above the peer-average threshold luminance. If a stimulus is seen, then it is assumed that there is no significant defect at that particular location. If a stimulus is not seen, then there is either a relative or an absolute scotoma. By accepting the loss of information about scotoma depth, STP achieves a substantial reduction in test time. To what extent this loss of information compromises the applicability of STP in glaucoma care is largely unknown.

113

Chapter 7

Besides being an interesting alternative to SAP from a clinical point of view, there is another reason why a more detailed knowledge of the performance of STP is mandatory. Several of the major population-based studies that reported on glaucoma incidence, the Rotterdam Study,9-11 the Blue Mountains Eye Study,12 the Baltimore Eye Survey,13 and the Beaver Dam Eye Study,14 used STP for visual field screening because of examination time constraints, especially since SITA was not yet available at that time. The interpretation of the results from these studies would also benefit from a detailed comparison of STP and SAP. The aim of this study was to compare glaucomatous damage on STP and SAP. This was performed cross-sectionally, in a clinical setting. In STP, we used the number of missed test points as our outcome measure; in SAP, we used the Mean Deviation (MD). There are a few studies that have compared STP to SAP,15-18 but these were aimed at the diagnostic performance of STP (presence or absence of abnormalities) and their data do not allow for a quantitative comparison of glaucomatous damage. 7.2 Methods Study population Glaucoma patients and glaucoma-suspect patients, who were scheduled for SAP at our outpatient department in 2008, received a letter of invitation. We invited only one randomly selected patient per day in order to avoid long waiting times in our tightly booked department. The selected patients were requested to perform two additional STP tests in one eye. Glaucoma-suspect patients were subjects with ocular hypertension and/or a positive family history of glaucoma and/or glaucomatous optic neuropathy but no glaucomatous visual field loss at their latest visit. Ocular hypertension was defined as an intra-ocular pressure (IOP) >20 mmHg on at least two separate visits. Positive family history of glaucoma was defined as one or more first-degree relatives with glaucoma. Glaucomatous optic neuropathy was defined as a vertical cup-disc ratio of at least 0.7. Glaucoma patients were patients with glaucomatous visual field loss in their enrolled study eye. Glaucomatous visual field loss was defined using SAP (HFA 24-2 SITA standard, see next section) as either: 1. Glaucoma Hemifield Test outside normal limits, and/or 2. Pattern Standard Deviation P<0.05, and/or 3. a scotoma of three adjacent points P<0.05 in the pattern deviation probability plot in which at least one point P<0.01, with all points being on the same side of the horizontal meridian. The field loss had to be reproducible, that is, it had to be present on at least two consecutive fields (not including the first field ever made) in the same hemifield and with at least one depressed test point having exactly the same location. Fields had to be reliable and the field loss had to be compatible with glaucoma

114

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

and without any other explanation. A test result was considered unreliable if false positives exceeded 10% or if both false negatives and fixation losses exceeded 10% and 20% respectively. All participants had perimetric experience. Patients with concurrent disease that might compromise the visual field were excluded. Eyes with best corrected visual acuity <0.5 (20/40) were also excluded, except when caused by glaucoma. Perimetry Visual fields (VF) were obtained with the Humphrey Field Analyzer (HFA 640; Carl Zeiss Meditec, Dublin, CA, USA). For STP, we used a test identical to the screening test used in the Rotterdam Study.9 It has a 52-point grid covering the central field with a radius of 24º, and a threshold-related test strategy (see Appendix). In this strategy, a threshold sensitivity is determined in four “seed” locations, located at 10º eccentricity, one in each quadrant. The age-corrected average height of the hill of vision is then adjusted for each test subject based on the sensitivity of these four seed locations. This adjustment attempts to compensate for differences in media clarity, pupil size, and general responsiveness. The maximum amount of adjustment is restricted by the lower limit of the seed locations (26dB). The actual test is performed with stimuli that are 6dB brighter than the predicted sensitivity of each particular test point. Stimuli that are not seen will be presented for a second time, and only if missed twice are they reported as missed points on the test printout. We used the total number of missed points for quantifying the amount of visual field loss. For SAP, the HFA 24-2 SITA Standard program was used. It has the same 52-point grid as our custom STP test field, but has two additional points nasally, extending its radius to 30º at that location. The global index Mean Deviation (MD) was used as a measure of glaucomatous damage on SAP. If both eyes were eligible for the study, one eye was randomly selected. The selected eye underwent three consecutive VFs: two STP tests and one HFA 24-2, with five-minute breaks in between. Testing always started with STP, but the order of the second and third test was determined at random to avoid any systematic influence of diminishing concentration. Thus, the test sequence was either STP > STP > SAP, or STP > SAP > STP. Data from the second STP test were used for comparison with the SAP MD score; data from the first STP test were used only for calculating STP test-retest variability. Analysis We used the Student’s t-test for continuous variables, and the chi-squared test for proportions (with Yates’ correction where appropriate). Spearman’s rho was used to determine correlation coefficients, since the data were not normally distributed. Test-retest variability was expressed as the coefficient of repeatability, which equals twice the standard deviation of the differences between the first and second STP tests, as described by Bland and Altman.19 Calculations were performed in SPSS 16.0.2 for Windows, SPSS Inc. Chicago, IL, USA.

115

Chapter 7

7.3 Results Ninety-two out of 131 invited patients (70%) participated in the study. Participants and non-participants were similar with respect to age and gender. Table 7.3.1 presents the characteristics of the 92 participants, consisting of 20 glaucoma suspects and 72 glaucoma patients. In the group of glaucoma suspects, 9 had OHT, 9 had a positive family history for glaucoma, and 6 had glaucomatous optic neuropathy (several participants were classified as glaucoma suspect for more than one reason). Table 7.3.1 Characteristics of participants

glaucoma suspects (n=20)

glaucoma patients (n=72)

all participants (n=92)

age (yrs; mean (SD)) 67 (11) 59 (12) 65 (12)

gender (% male) 50 49 49

right eye selected (%) 55 47 49

test sequence STP > SAP > STP (%) 60 47 50

SD = Standard Deviation; STP = Supra Threshold Perimetry custom test; SAP = Standard Automated Perimetry Table 7.3.2 presents the amount of visual field loss and testing time of STP and SAP. The STP test was almost twice as short as the SAP test (P<0.001).

116

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

Table 7.3.2 Visual field loss and test time of STP and SAP

glaucoma suspects (n=20)

glaucoma patients (n=72)

all participants (n=92)

HFA MD (dB; mean (SD)) -1.5 (1.7) -11.8 (7.6) -9.5 (8.0)

STP1 (missed points; mean (SD)) 1.5 (1.1) 13.2 (10.5) 10.7 (10.5)

STP2 (missed points; mean (SD)) 1.5 (1.0) 14.7 (11.8) 11.9 (11.8)

STP test time (s; mean (SD)) 184 (15) 233 (42) 223 (43)

HFA test time (s; mean (SD)) 305 (37) 407 (63) 385 (72)

HFA = Humphrey Field Analyzer 24-2 SITA Standard programme; MD = Mean Deviation; SD = Standard Deviation; STP = Supra Threshold Perimetry custom test; 1= first test, 2= second test Figure 7.3.1 shows the relationship between SAP MD and the number of missed points on STP. The coefficient of determination (R2) of the linear regression line was 0.89. The regression line is described as MD = -0.72*STP – 0.68; when forcing the line to go through the origin, we found MD = -0.75*STP. The residuals of the fit followed a normal distribution. The Spearman’s rho correlation coefficient was -0.92 (P<0.001). Figure 7.3.2 presents a scatter plot of the first versus the second STP test. The regression line was almost equal to the line of equality (y=x). Slightly more points were missed on the second STP test (11.9, SD 11.8 points) than on the first test (10.7, SD 10.5; P=0.001). The coefficient of repeatability was 7.0 missed points. Subgroup analysis revealed that patients with test sequence STP > STP > SAP had a trend towards a better coefficient of repeatability (5.9 missed points) than patients with test sequence STP > SAP > STP (7.9 missed points), but this difference was not significant (P=0.118).

117

Chapter 7

Figure 7.3.1 Scatter plot showing the (minus of the) HFA 24-2 MD plotted against the number of missed points on STP. Lines show the regression line and corresponding 95% confidence interval.

Figure 7.3.2 Scatter plot of the first versus the second STP test. Lines show the regression line and corresponding 95% confidence interval.

118

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

As mentioned in the introduction, several population-based studies relied on STP for visual field screening. In the Rotterdam Study, for example, STP visual field loss was considered present when there was a scotoma that consisted of three or more contiguous missed points in two consecutive screening tests.10 Fifty-five out of our 72 glaucoma patients (76%) would have been classified as having visual field loss according to the Rotterdam criterion. All 20 glaucoma suspects would be classified correctly as not having visual field loss. The average MD of the 17 missed glaucoma patients was -4.4dB (SD 2.5dB, range -1.1 to -9.6dB). Sensitivity of the Rotterdam criterion (using glaucomatous visual field loss on SAP as the gold standard) was 28% for early glaucoma (MD better than -6dB; 5 out of 18 patients), 83% for moderate glaucoma (MD between -6dB and -12dB; 20 out of 24 patients), and 100% for severe glaucoma (MD worse than -12dB; 30 out of 30 patients). 7.4 Discussion In this study, we quantitatively compared glaucomatous damage on STP and SAP. We found a sound association between both types of perimetry. Severity of glaucomatous damage on SAP expressed in MD can be estimated from STP test results by multiplying the number of missed points by -0.75. Most patients with glaucomatous visual field loss on SAP also had STP abnormalities, except for patients with early glaucoma. A study by Mills et al.18 concerning 87 early glaucoma patients (defined as MD better than -7dB) used an STP test identical to ours. Their criterion for visual field loss was three or more missed points. They reported a sensitivity of 23%, which is similar to the sensitivity of 28% for early glaucoma patients as found in our study using the Rotterdam criterion. Based on these figures, STP does not seem suitable for identifying early glaucoma. Topouzis et al.15 compared a 76-point STP test to HFA 30-2 full-threshold SAP in 29 glaucoma patients of unspecified severity. They reported a sensitivity of 74%, which is similar to the overall sensitivity of 76% as found in our study. Several studies report on test-retest variability of SAP. Many of those compare the threshold sensitivity of individual test locations between consecutive tests, which is incompatible with STP output. A direct comparison can only be made with three studies. First, a study by Bjerre et al.20 concerning 74 glaucoma patients that underwent two HFA 24-2 SITA Standard tests. They counted the number of depressed points at P<5% as a global index of visual field loss, and then compared the first with the second test. The resulting coefficient of repeatability was 9.9 points. Second, a study by Smith et al.21 concerning 192 glaucoma patients that underwent several full threshold HFA tests. They used the MD as a measure of visual field loss and expressed test-retest variability as a variance. They

119

Chapter 7

found a variance of 4.6 dB2, which equals a coefficient of repeatability of 4.3 dB. The latter value can be converted to STP format by dividing it by 0.75, yielding 7.0 missed points. Finally, we re-analyzed data collected in the Groningen Longitudinal Glaucoma Study.22,23 Two consecutive HFA 30-2 SITA fast VF’s were selected to determine the MD test-retest variability of 221 glaucoma patients. The coefficient of repeatability was 4.0dB, that is, 6.6 missed points. All three of the above-mentioned figures are comparable to the coefficient of repeatability of 7.0 missed points as found for STP in this study, suggesting that the test-retest variability of SAP and STP are similar. There are some advantages of STP over SAP. Subjects find STP easier because test duration is a lot shorter, but also because – unlike in SAP – the majority of the stimuli are not close to threshold luminance. Furthermore, because we used a threshold-related STP test strategy (see methods), the number of missed points on STP was less affected by cataract and other media opacities than the global index MD. However, unlike the Pattern Standard Deviation (PSD) global index in SAP, the number of missed points on STP does not return to normal in end-stage disease, since the amount of adjustment for overall sensitivity loss is limited. Existing progression detection algorithms like the Glaucoma Progression Analysis (GPA; Carl Zeiss Meditec Inc., Dublin, CA, USA) and the AGIS criterion24 cannot be used with STP, but the recently introduced and evaluated Nonparametric Progression Analysis (NPA) algorithm25,23 can be applied to any continuous visual field loss measure, including the number of missed points on STP. The major drawback of STP is its lack of information about scotoma depth. This might be relevant in progression detection, since glaucoma progression has been shown to be a mixture of increase in both scotoma size and depth.26,27 Given the obvious linear relationship between the number of missed points on STP and SAP MD, however, STP appears to be a viable alternative to SAP for obtaining a general impression of glaucomatous damage. Since SAP can be too demanding, especially for elderly patients, further longitudinal studies to evaluate the role of STP in progression detection would seem worthwhile. Appendix For the test used in this study, access “System Setup” and then “Additional Setup” on your HFA device. Select “Custom Test”, then select “Create Screening Test”. Coordinates of each of the 52 test points can now be entered consecutively. The test is identical to the HFA 24-2 grid without the two nasal points. Spacing between the test points is 6º. Upon completion, the default test parameters are shown. “Test Mode” needs to be set to “Threshold Related”.

120

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

References for chapter seven

1. Bebie H, Fankhauser F, Spahr J. Static perimetry: strategies. Acta Ophthalmol (Copenh) 1976;54:325-38.

2. Johnson CA, Chauhan BC, Shapiro LR. Properties of staircase procedures

for estimating thresholds in automated perimetry. Invest Ophthalmol Vis Sci 1992;33:2966-74.

3. Wild JM, Pacey IE, O'Neill EC, Cunliffe IA. The SITA perimetric threshold

algorithms in glaucoma. Invest Ophthalmol Vis Sci 1999;40:1998-2009.

4. Flanagan JG, Wild JM, Trope GE. Evaluation of FASTPAC, a new strategy for threshold estimation with the Humphrey Field Analyzer, in a glaucomatous population. Ophthalmology 1993;100:949-54.

5. Flanagan JG, Moss ID, Wild JM, Hudson C, Prokopich L, Whitaker D,

O'Neill EC. Evaluation of FASTPAC: a new strategy for threshold estimation with the Humphrey Field Analyser. Graefes Arch Clin Exp Ophthalmol 1993;231:465-9.

6. Bengtsson B, Olsson J, Heijl A, Rootzén H. A new generation of algorithms

for computerized threshold perimetry, SITA. Acta Ophthalmol Scand 1997;75:368-75.

7. Bengtsson B, Heijl A. Evaluation of a new perimetric threshold strategy,

SITA, in patients with manifest and suspect glaucoma. Acta Ophthalmol Scand 1998;76:268-72.

8. Wall M, Brito CF, Woodward KR, Doyle CK, Kardon RH, Johnson CA. Total

deviation probability plots for stimulus size v perimetry: a comparison with size III stimuli. Arch Ophthalmol 2008;126:473-9.

9. Dielemans I, Vingerling JR, Wolfs RC, Hofman A, Grobbee DE, de Jong PT.

The prevalence of primary open-angle glaucoma in a population-based study in The Netherlands. The Rotterdam Study. Ophthalmology 1994;101:1851-5.

10. Wolfs RC, Borger PH, Ramrattan RS, Klaver CC, Hulsman CA, Hofman A,

Vingerling JR, Hitchings RA, de Jong PT. Changing views on open-angle

121

Chapter 7

glaucoma: definitions and prevalences--The Rotterdam Study. Invest Ophthalmol Vis Sci 2000;41:3309-21.

11. de Voogd S, Ikram MK, Wolfs RC, Jansonius NM, Hofman A, de Jong PT.

Incidence of open-angle glaucoma in a general elderly population: the Rotterdam Study. Ophthalmology 2005;112:1487-93.

12. Mitchell P, Smith W, Attebo K, Healey PR. Prevalence of open-angle

glaucoma in Australia. The Blue Mountains Eye Study. Ophthalmology 1996;103:1661-9.

13. Katz J, Tielsch JM, Quigley HA, Javitt J, Witt K, Sommer A. Automated

suprathreshold screening for glaucoma: the Baltimore Eye Survey. Invest Ophthalmol Vis Sci 1993;34:3271-7.

14. Klein BE, Klein R, Sponsel WE, Franke T, Cantor LB, Martone J, Menage MJ.

Prevalence of glaucoma. The Beaver Dam Eye Study. Ophthalmology 1992;99:1499-504.

15. Topouzis F, Coleman AL, Yu F, Mavroudis L, Anastasopoulos E, Koskosas

A, Pappas T, Dimitrakos S, Wilson MR. Sensitivity and specificity of the 76-suprathreshold visual field test to detect eyes with visual field defect by Humphrey threshold testing in a population-based setting: the Thessaloniki eye study. Am J Ophthalmol 2004;137:420-5.

16. Artes PH, Henson DB, Harper R, McLeod D. Multisampling suprathreshold

perimetry: a comparison with conventional suprathreshold and full-threshold strategies by computer simulation. Invest Ophthalmol Vis Sci 2003;44:2582-7.

17. Siatkowski RM, Lam BL, Anderson DR, Feuer WJ, Halikman AM.

Automated suprathreshold static perimetry screening for detecting neuro-ophthalmologic disease. Ophthalmology 1996;103:907-17.

18. Mills RP, Barnebey HS, Migliazzo CV, Li Y. Does saving time using

FASTPAC or suprathreshold testing reduce quality of visual fields? Ophthalmology 1994;101:1596-603.

19. Bland JM, Altman DG. Statistical methods for assessing agreement

between two methods of clinical measurement. Lancet 1986;1:307-10.

20. Bjerre A, Grigg JR, Parry NR, Henson DB. Test-retest variability of multifocal visual evoked potential and SITA standard perimetry in glaucoma. Invest Ophthalmol Vis Sci 2004;45:4035-40.

122

Supra-threshold perimetry compared to standard automated perimetry in glaucoma

123

21. Smith SD, Katz J, Quigley HA. Analysis of progressive change in

automated visual fields in glaucoma. Invest Ophthalmol Vis Sci 1996;37:1419-28.

22. GP, Blanksma LJ, Hardus PL, Jansonius NM. The Groningen Longitudinal

Glaucoma Study. I. Baseline sensitivity and specificity of the frequency doubling perimeter and the GDx nerve fibre analyser. Heeg Acta Ophthalmol Scand 2005;83:46-52.

23. Wesselink C, Heeg GP, Jansonius NM. Glaucoma monitoring in a clinical

setting: glaucoma progression analysis vs nonparametric progression analysis in the Groningen Longitudinal Glaucoma Study. Arch Ophthalmol 2009;127:270-4.

24. The AGIS Investigators. The Advanced Glaucoma Intervention Study

(AGIS): 7. The relationship between control of intraocular pressure and visual field deterioration. Am J Ophthalmol 2000;130:429-40.

25. Jansonius NM. Bayes' theorem applied to perimetric progression

detection in glaucoma: from specificity to positive predictive value. Graefes Arch Clin Exp Ophthalmol 2005;243:433-7. Epub 2004 Dec 1.

26. Mikelberg FS, Drance SM. The mode of progression of visual field defects

in glaucoma. Am J Ophthalmol 1984;98:443-5.

27. Boden C, Blumenthal EZ, Pascual J, McEwan G, Weinreb RN, Medeiros F, Sample PA. Patterns of glaucomatous visual field progression identified by three progression criteria. Am J Ophthalmol 2004;138:1029-36.

125

Chapter 8 General discussion

Chapter 8

The main issue that this thesis tries to address, is whether or not population based glaucoma screening should be introduced in the Netherlands. Chapter two (Literature review) gives an overview of current literature regarding different aspects of screening. Although most of the Wilson and Jungner criteria for population based screening are satisfied, two criteria in particular remain problematic. First and foremost, cost-effectiveness remains dubious (see paragraph 2.10). Second, although diagnostic tests for glaucoma have improved substantially in recent years, screening will still generate a lot of false positive subjects. Lack of a screening test with near-perfect specificity contributes to an unfavourable cost-effectiveness ratio of the screening programme (see paragraph 2.5). The five studies described in this thesis (chapter three to seven) were aimed at these two key matters. 8.1 Cost-effectiveness Cost-effectiveness is usually expressed as costs per Quality Adjusted Life Year (QALY). It is a measure that indicates whether a particular intervention is good value for money. An intervention is said to be cost-effective when the costs per QALY gained is within the limits of society’s Willingness to Pay (WTP), the exact amount varies per country. In the Netherlands, the cost-effectiveness threshold is set at €20,000/QALY for preventive interventions.1;2 In the United Kingdom, WTP is determined by the National Institute for Clinical Excellence to be about £30,000 (or €35,000).3 In the United States, WTP is $50,000 (also equalling €35,000).4;5 Economic implications of the introduction of population based screening are difficult to foresee in sufficient detail. As described in paragraph 2.10, the two major studies on cost-effectiveness of glaucoma screening6;7 disagree on the economic feasibility thereof, because their respective models yielded considerably different final costs per QALY: €9,000 as found by the Finnish model6 is well within the WTP limits, while around €70,000 (rough approximation) as estimated by the UK model7 clearly is not. So which one of them is right, or at least closest to the actual costs? Several points of criticism were already made in paragraph 2.10 regarding the Finnish study. The various elements that make up the overall cost of a screening programme are summarized in table 2.10.1 (Costs and benefits of a glaucoma screening programme). Several modifiers exert influence on the total costs of glaucoma screening. The most important parameters in this respect are: test specificity, glaucoma prevalence, age at which screening starts, and screening interval. The latter two are sufficiently covered in chapter two (Literature review). However, additional comments are necessary regarding test specificity and glaucoma prevalence, related to the cost-effectiveness studies mentioned above.

126

General discussion

Test specificity In paragraph 2.8, the number of false positive test subjects produced by population based screening in the Netherlands was estimated, assuming a screening test specificity of 95% based on data presented in paragraph 2.5. However, most of the studies cited in paragraph 2.5, concerning both perimetry and imaging tests, determined specificity on a by-eye basis (i.e. one eye is selected as the study eye; the fellow eye is excluded). In real life, most individuals have two eyes, hence two chances to be classified as abnormal. Data from the FDT perimetry measurements collected for the third chapter of this thesis8 was re-analysed to calculate the by-eye specificity and compare it to the original by-patient specificity. Out of 108 healthy elderly subjects, 11 had VF loss in one of their eyes (five right eyes and six left eyes), and only two had VF loss in both eyes. This corresponds to a by-patient specificity of 88%, and a mean by-eye specificity of 93%. The same subjects also underwent GDx imaging as part of the Groningen Longitudinal Glaucoma Study.9;10 These data were also re-analysed. By-eye specificity was set at 94% by choosing the appropriate cut-off point for the GDx test. The corresponding by-patient specificity was 90%. If we adjust for the discrepancies of by-eye versus by-patient specificity in regard to the specificity rate of 95% as applied in table 2.8.1, then the actual maximum feasible specificity of glaucoma screening for either a single perimetry test or a single imaging test would be closer to 91%. In that case, the clinical workload would nearly double, approaching 500,000 false positives instead of 277,000 every five years! The impact on cost-effectiveness of so many false positives is unacceptable, and needs to be negated. In chapter four, several strategies for increasing the test specificity were explored. The Finnish cost-effectiveness study6 assumed a test specificity of 98%, based on combining different diagnostic tests, and repeating abnormal test results. The highest specificity we were able to obtain was 94% by combining FDT and GDx results (classification on by-patient basis), see chapter four.11 Repeating abnormal tests in addition to combining different types of tests will probably result in a modest additional increase in specificity.11 However, a specificity of 98% remains unlikely to be achieved without a significant drop in sensitivity, including loss of sensitivity to moderate and severe glaucoma, which is especially undesirable. Glaucoma prevalence The prevalence of glaucoma in the Netherlands is about 2% in the general population aged >40 years (see paragraph 2.1).12 At least half of all glaucoma patients are undetected,12-14 so the prevalence of undetected glaucoma is approximately 1%. Glaucoma screening would naturally become much more cost-effective if prevalence were higher. This gain is clearly shown in the various sensitivity analyses carried out in the UK cost-effectiveness study discussed earlier.7 Therefore, the authors of that study suggest screening targeted at certain subpopulations as a potential alternative to screening of the general population. The aim is to attain a subpopulation in which the prevalence of glaucoma is

127

Chapter 8

increased, by limiting the screening programme to subjects with one or more risk factors for developing glaucoma. Important risk factors for glaucoma are summarized in the last section of paragraph 2.2. Positive family history of glaucoma,15 black ethnicity,16 and myopia17 are relevant in this regard. Screening in risk factor defined subpopulations can be effective if two conditions are met. Obviously, the Relative Risk (RR; also termed Risk Ratio) of a specific risk factor must be significantly increased. But if the risk factor itself is rare, then the screening programme targets only a fraction of the population, with little impact on the overall burden of a particular disease. The Population Attributable Risk (PAR) takes this into account. Since the proportion of the Dutch 50+ population with black ethnicity is less than 1.3%,17;18 the PAR of black ethnicity is insignificant even though it carries a considerably increased RR of 3.8.16 Positive family history of glaucoma might be a better candidate for targeted screening. Out of 6773 subjects that participated in the Rotterdam Study, 8.5% had a positive family history of glaucoma (see chapter six, discussion section).19 The RR may be as high as 9.2 according to Wolfs et al.,15 partially based on data from the Rotterdam Study. They reported a PAR of 16.4%, which could be considered worthwhile. However, as a result of current opportunistic case finding, the proportion of undetected glaucoma patients that have a positive family history of glaucoma is very small, whereas the proportion of known glaucoma patients with positive family history is far greater. Specifically 2% among undetected glaucoma patients and 39% among detected patients have a positive family history of glaucoma, P<0.001, see chapter six (percentages presented in table 6.1).19 Since the majority of patients who have a positive family history of glaucoma are apparently already detected quite efficiently through current practice of opportunistic case finding, screening targeted at this risk factor is redundant. That leaves myopia as the remaining option for targeted screening. Myopia is a common condition with an estimated prevalence of 26% in the Netherlands (see chapter five, table 5.2).20 However, the RR of myopia is only 1.7 (based on pooled data from three population based glaucoma studies).17;21;22 This modest RR is probably not worth the additional effort of inviting only myopic subjects for screening, improvement in cost-effectiveness will be minimal. Also, it is hard to justify from an ethical point of view that the non-myopic population is not eligible to glaucoma screening, based on such a minor difference in risk of developing glaucoma. A more practical approach would be to intensify opportunistic case finding taking place at opticians. Myopics need to visit an optician anyway because their spectacles will need adjustment or replacement periodically, so most of them will be accessible for evaluation. This alternative will be discussed further below. In summary, strategies to improve the cost-effectiveness of glaucoma screening by targeting subpopulations with higher glaucoma prevalence as a result of risk factors are not viable in the Netherlands.

128

General discussion

8.2 Screening bias Both observational studies that investigate ongoing screening programmes as well as modelling studies that try to predict the yield of a screening programme are susceptible to bias. There are at least three types of screening bias: length bias, lead time bias, and class bias. See paragraph 2.3 for background information about terminology used in this context. Length bias The progression rate of diseases in general and of glaucoma in particular can vary considerably. For some individuals glaucoma progresses rapidly, whereas others have a slowly progressing variant. Those with a rapidly progressive variant remain for only a relatively short period of time in the detectable preclinical phase (D-PCP), a stadium in which the disease is detectable by a screening test but does not yet cause any symptoms. In contrast, individuals with a slowly progressive variant will remain in the D-PCP for many years. Among cases identified by a periodic screening programme, the slowly progressive ones will be over-represented. After all, rapidly progressive cases will more often become symptomatic (manifest disease; clinical phase) during the screening interval (the time period in between two consecutive screening rounds), before their next screening round was due. Consequently, the benefits of screening fall short: the decrease in morbidity and/or mortality will turn out to be smaller than expected. This phenomenon is called length bias, and is essentially a kind of selection bias towards mild disease cases. Its effect in glaucoma screening is probably significant, because glaucoma is characterised by a long D-PCP on average, but at the same time there is considerable individual variation in the rate of progression. The negative impact of lengthbias on screening yields is difficult to quantify. The study described in chapter six was originally designed and intended to explore the presence of lengthbias in glaucoma, and give a rough estimate of the effect magnitude. Unfortunately, baseline differences came to light between opportunistically detected and undetected incident glaucoma patients, which precluded a final statement on the presence or absence of length bias in glaucoma screening. Still, glaucoma patients that were discovered through opportunistic case finding outside the Rotterdam Study had worse glaucoma than those that had remained undetected19 (see discussion of chapter six for more details). Lead time bias Screening allows for early detection of diseases compared to normal diagnosis which is initiated in a later stadium, when complaints or symptoms start to occur. Time gained by early detection is called lead time, and is equivalent to half the D-PCP (see paragraph 2.3). Survival is defined as the time span between biological onset of disease (in this case glaucoma) and end stage thereof (blindness). If a disease is detected earlier due to screening, then survival will increase

129

Chapter 8

automatically by an amount equal to the lead time. This is not a beneficial effect caused by screening, but purely the arithmetical result of starting to count from an earlier disease phase. Lead time bias occurs when this deviation is not taken into account. Evaluation of ongoing screening programmes is prone to lead time bias, whereas studies conducted prior to introduction of screening as well as modelling studies are not. Therefore, the impact of lead time bias is irrelevant at present. Class bias Individuals from a higher socio-economic class have a healthier lifestyle than individuals from a lower socio-economic class. They are also more likely to participate in health promoting projects such as a screening programme. If screening is attended predominantly by healthy people, then less disease will be detected. This problem is called class bias. Whether a healthy lifestyle has any influence on glaucoma onset or progression is questionable. Black ethnicity is a risk factor for glaucoma, and is also associated with a lower socio-economic class, which may cause some class bias. However, as mentioned earlier in this chapter, the proportion of Dutch 50+ inhabitants with black ethnicity is less than 1.3%.18 Any effects of class bias must therefore be minimal. In summary, among the three existing types of screening bias, only length bias is relevant for glaucoma screening in the context of this thesis. The modelling studies6;7 (discussed in paragraph 2.10 and also in this chapter) do not account for length bias, hence their cost-effectiveness estimates regarding glaucoma screening are too optimistic. 8.3 Verdict on glaucoma screening Most aspects relevant to glaucoma screening were reviewed in chapter two. An important argument against screening is the problem that screening is in all probability not cost-effective. This complicated issue is discussed in paragraph 2.10, as well as in paragraph 8.1 and 8.2, where the negative impact on cost-effectiveness of suboptimal by-patient specificity and lengthbias are assessed. Several additional aspects against or in favour of the introduction of a screening programme in the Netherlands will be discussed in this section. Part of the incident glaucoma cases identified by the Rotterdam Study23 were found to have also been detected by regular ophthalmic care outside the study, whereas other cases had remained undetected during the entire follow-up interval (see chapter six).19 Twenty-three cases (29%) had already been detected, 55 cases (71%) remained undetected. The severity of glaucoma was worse in detected cases compared to undetected cases (P=0.009). The additional yield of screening is therefore lower than would be expected from prevalence data: we

130

General discussion

estimated that only about one in 1000 screened individuals could be saved from bilateral end-stage glaucoma. This estimate might be somewhat conservative, since it is based on the assumption that only patients with >10 missed points on STP in their better eye will reach bilateral end-stage glaucoma. In chapter seven we found that >10 missed points on STP corresponds to a HFA MD < -7.5 dB (moderate glaucoma). A lower cut-off point is arguably more sensible. For >5 missed points on STP (equivalent to HFA MD < -3.8 dB, early glaucoma) estimated yields will be better: about one in 550 screened individuals might be saved from blindness. Nevertheless, yields are limited, and for every 1000 screened individuals, 50 cases will still require further investigation as a result of false positive test results (see paragraph 2.5 and 2.8). In the last section of paragraph 2.4, studies by Hattenhauer19;24 and Chen25 were discussed that reported on the cumulative probability of bilateral blindness from glaucoma, defined as either VF constriction to ≤20° or visual acuity ≤2/20 in the better eye. Most eyes were classified as blind due to VF constriction; only a small fraction had a visual acuity of ≤2/20. Although severe constriction of the VF is unquestionably debilitating, the impact on quality of life is limited as long as central visual acuity is preserved. Much et al. investigated the long-term survival of central visual field in end-stage glaucoma.26 Eighty-four eyes of 64 patients were included. Only fourteen eyes (17%) lost more than three lines of visual acuity during a mean follow-up period of 8.3 years. Thus, despite being considered legally blind based on VF criteria (see table 2.1.1 for definitions), central visual acuity of end-stage glaucoma patients may be preserved for many years. Definitions of blindness, low vision, and visual impairment are based on the remaining amount of visual function in the better seeing eye (see table 2.1.1). Glaucoma reduces health-related quality of life (QoL) mainly in advanced stages of the disease, when severe visual field damage has also occurred in both eyes.27;28 Therefore, glaucoma severity of the better eye is emphasized throughout this thesis, while the worse eye is essentially being ignored. An alternative point of view is that screening should be aimed at retaining useful vision in both eyes, and thus at preventing end-stage glaucoma in the worse eye. QoL in individuals with good vision in both eyes is higher than QoL in functionally monocular individuals.29 But the difference is only marginal, and consequently cost per QALY gained by a glaucoma screening programme designed to save the worse eye would be astronomical. Still, the introduction of a screening programme set to prevent bilateral blindness would as an added benefit also lower incidence of unilateral blindness with a concurrent modest increase in QoL. Verdict The cost-effectiveness studies6;7 discussed in paragraph 2.10 and this chapter are a good starting point for determining whether or not a periodic population based glaucoma screening programme should be introduced in the Netherlands. The

131

Chapter 8

UK study7 concluded that screening was not cost-effective by a wide margin. The Finnish study6 concluded that screening was indeed cost-effective, and in certain age cohorts even dominant. However, their results seem over-optimistic due to several apparent weak points in study design and chosen parameter values; see criticism expressed in relation to the discussion of this study in paragraph 2.10. In this thesis, several factors are identified that have a further negative impact on the already dubious cost-effectiveness of glaucoma screening. These factors are: • lengthbias (chapter 6 and paragraph 8.2) • suboptimal by-patient test specificity (paragraph 2.5 and 8.1; chapter 3 and 4) • large clinical workload from false positives (paragraph 2.8 and 8.5) This leads to the conclusion that at present periodic population based glaucoma screening is not feasible in the Netherlands. 8.4 Alternatives There are at least three commonsense alternatives to periodic population based screening. The most straightforward option is to maintain the current practice of opportunistic case finding. Another approach is once in a lifetime screening instead of periodic screening. This second option seems counter-intuitive since glaucoma prevalence slowly increases with age, and there are no specific opportunities during the course of glaucoma that warrant screening at any specific moment. This strategy is not explored in this thesis. The third and most interesting option is to intensify opportunistic case finding taking place at opticians, and is discussed in the following section. In chapter five it is reported that 80% of Dutch inhabitants aged >40 years visit an optician at least once every five years.20 As mentioned in paragraph 2.8, participation rates for breast cancer screening vary considerably, and 70% is considered to be achievable.30 This means that optician based case finding will potentially reach at least as many individuals as a nation wide screening programme. Optician based screening might therefore be a more valid designation than optician based case finding. Opticians are comfortable with providing ophthalmic care, and most of them already perform non-contact tonometry as a manner of glaucoma case finding voluntarily. 97% of optician shops is equipped with a non-contact tonometer (unpublished data, obtained from the questionnaires described in chapter five). There are approximately 3300 optician shops in the Netherlands.31 In a small pilot study among 50 opticians, 37 of whom responded to the questionnaire, 91% expressed willingness to participate in an extended glaucoma screening programme (i.e. more extensive than tonometry).20 Financial compensation was not ascertained, and may not be necessary. After all, opticians have acquired non-contact tonometers for competitive reasons of their own accord. The same trend is now discernible with

132

General discussion

respect to the FDT perimeter and the GDx nerve fibre analyser. In paragraph 2.8, it is established that a lower age limit of 50 years is suitable for a glaucoma screening programme in the Netherlands. For optician based screening, it may be more appropriate to start at age 45, so as not to miss individuals that visit an optician because of presbyopia. Optician based screening would reduce costs considerably as compared to a normal screening programme. See table 2.10.1 for an overview of the different types of screening related costs: all direct screening costs as well as the societal costs related to productivity loss and transport are practically eliminated in optician based screening. Still, substantial clinical costs remain because of the large clinical workload that results from screening, principally due to false positive test results (discussed in paragraph 2.8). Optician based screening is far more likely to be cost-effective than a normal glaucoma screening programme, but a definite statement requires additional research (see paragraph 8.5). If optician based screening were to be instituted, ophthalmologists should be made aware that around 90% of individuals that are referred by opticians because of an abnormal test result, are expected to turn out to be false positive cases. Left unaware, ophthalmologists might at least initially assume that the referring opticians are incompetent. A positive attitude and readiness to cooperate are important. Encouragement of opticians to participate in screening efforts may be a suitable task for ophthalmologists’ departments. Referral should not require an appointment with the general practitioner. The screening test strategy must be kept simple so that any employee in the optician shop can carry out the test, and understands what to do with the results. Cut-off points for screening tests should be chosen aiming for high specificity. An example of a simple decision tree for screening with the FDT perimeter is shown below in figure 8.4.1. The FDT can be substituted by another glaucoma screening device such as the GDx or HRT without the need to alter the rest of the decision tree. Repeat testing is probably necessary for all of current diagnostic devices in order to attain a high specificity. Regardless of the format of glaucoma screening, both opticians and ophthalmologists still need to perform tonometry on a routine basis. The reason is that individuals with normal screening tests but a (very) high IOP can progress so rapidly that a five year screening interval is inadequate.

133

Chapter 8

age 45 years or more

Figure 8.4.1 Decision tree optician based glaucoma screening by FDT perimetry

yes no

screening indicated no screening

intra-ocular pressure:25 or higher

yes no

FDT screening test:two or more abnormal squares in one eye

yes no

repeat FDT screening test:at least two abnormal squares

at the same location as the abnormal squares in the first test

yes no

refer to ophthalmologist no referral necessary

134

General discussion

8.5 Suggestions for further research During the process of writing this thesis, several issues were identified that would benefit from further research. False positives The majority of glaucoma screening participants with a positive (abnormal) test result will be classified as normal upon further evaluation. These individuals are called ‘false positives’ (see paragraph 2.8). There are several options regarding how to deal with false positives in the course of the screening programme. Viable approaches are: return false positives to periodic screening; retain false positives under clinical follow-up; or, exclude false positives from screening indefinitely and rely on opportunistic case finding only. If false positives are allowed to return to periodic screening, accumulation of normal subjects that repeatedly fail their screening test every screening round will likely ensue, which will negatively affect the cost-effectiveness of the screening programme. It would be interesting to know what proportion of false positives falls into this category of ‘reoffenders’. If the proportion is significant, then it becomes important to determine the reason why false positives occur. This knowledge can help to choose the best course of action. For example, a false positive subject who has cataract may return to periodic screening after phacoemulsification is performed. Some of the false positives will prove to be unsuitable for either imaging or perimetry (see first section of paragraph 2.5 for overview of causes thereof). They can return to periodic screening if an alternate screening test modality is used (i.e. switch from perimetry to imaging or vice versa); if the screening programme by default relies on a combination of two screening test modalities, then the inappropriate test is left out and cut-off points for the remaining test may be adjusted. False positives who are unsuitable for both perimetry and imaging should be excluded from further screening rounds, case finding only seems the most cost-effective for this group. Borderline cases should be followed clinically. Increasing the specificity High specificity is important in a screening setting, as stressed in paragraphs 2.5 and 8.1. The Finnish cost-effectiveness study6 discussed in paragraphs 2.10 and 8.1 assumed a very high by-patient test specificity of 98% for their Markov model, attained theoretically by combining different diagnostic tests and repeating abnormal test results. A study is needed to explore what effects such an extensive screening test strategy will have on sensitivity in general and sensitivity to moderate and severe glaucoma in particular (with cut-off points set to achieve a 98% specificity). Blindness from glaucoma Why do glaucoma patients in the Netherlands go blind? This question is of fundamental importance with respect to the yield of glaucoma screening. If the

135

Chapter 8

predominant cause of glaucoma related blindness is late detection, then screening will have the greatest impact on preventing blindness (see paragraph 2.4 for advantages of early detection). In contrast, if suboptimal treatment of glaucoma progression during follow-up also plays a significant role, then the beneficiary effects of screening will be limited. To clarify this issue, a study is required that determines the proportion of blind glaucoma patients in the Netherlands that was treated suboptimally. Continuation of the current line of research Further research in the field of glaucoma screening in our department is aimed at optician based glaucoma screening. Preparations for a pilot project investigating this matter were recently started in collaboration with opticians in the North of the Netherlands who are willing to participate in glaucoma screening. Severity of glaucoma at diagnosis will be determined prior to and after introduction of the pilot project. This will provide more definite data on the yield of glaucoma screening. Cost-effectiveness of optician based glaucoma screening should be explored before it is introduced nationwide.

136

General discussion

References

1. de Kok IM, van Ballegooijen M, Habbema JD. Cost-effectiveness analysis of human papillomavirus vaccination in the Netherlands. J Natl Cancer Inst 2009;101:1083-92.

2. Coupe VM, van Ginkel J, de Melker HE, et al. HPV16/18 vaccination to prevent cervical cancer in The Netherlands: model-based cost-effectiveness. Int J Cancer 2009;124:970-8.

3. Devlin N, Parkin D. Does NICE have a cost-effectiveness threshold and what other factors influence its decisions? A binary choice analysis. Health Econ 2004;13:437-52.

4. Hirth RA, Chernew ME, Miller E, et al. Willingness to pay for a quality-adjusted life year: in search of a standard. Med Decis Making 2000;20:332-42.

5. Ubel PA, Hirth RA, Chernew ME, Fendrick AM. What is the price of life and why doesn't it increase at the rate of inflation? Arch Intern Med 2003;163:1637-41.

6. Vaahtoranta-Lehtonen H, Tuulonen A, Aronen P, et al. Cost effectiveness and cost utility of an organized screening programme for glaucoma. Acta Ophthalmol Scand 2007;85:508-18.

7. Hernandez RA, Burr JM, Vale LD. Economic evaluation of screening for open-angle glaucoma. Int J Technol Assess Health Care 2008;24:203-11.

8. Stoutenbeek R, Heeg GP, Jansonius NM. Frequency doubling perimetry screening mode compared to the full-threshold mode. Ophthalmic Physiol Opt 2004;24:493-7.

9. Heeg GP, Blanksma LJ, Hardus PL, Jansonius NM. The Groningen Longitudinal Glaucoma Study. I. Baseline sensitivity and specificity of the frequency doubling perimeter and the GDx nerve fibre analyser. Acta Ophthalmol Scand 2005;83:46-52.

10. Jansonius NM, Heeg GP. The Groningen Longitudinal Glaucoma Study. II. A prospective comparison of frequency doubling perimetry, the GDx

137

Chapter 8

nerve fibre analyser and standard automated perimetry in glaucoma suspect patients. Acta Ophthalmol 2008.

11. Heeg GP, Stoutenbeek R, Jansonius NM. Strategies for improving the diagnostic specificity of the frequency doubling perimeter. Acta Ophthalmol Scand 2005;83:53-6.

12. Burr JM, Mowatt G, Hernandez R, et al. The clinical effectiveness and cost-effectiveness of screening for open angle glaucoma: a systematic review and economic evaluation. Health Technol Assess 2007;11:iii-x, 1.

13. Hollows FC, Graham PA. Intra-ocular pressure, glaucoma, and glaucoma suspects in a defined population. Br J Ophthalmol 1966;50:570-86.

14. Wolfs RC, Borger PH, Ramrattan RS, et al. Changing views on open-angle glaucoma: definitions and prevalences--The Rotterdam Study. Invest Ophthalmol Vis Sci 2000;41:3309-21.

15. Wolfs RC, Klaver CC, Ramrattan RS, et al. Genetic risk of primary open-angle glaucoma. Population-based familial aggregation study. Arch Ophthalmol 1998;116:1640-5.

16. Tielsch JM, Sommer A, Katz J, et al. Racial variations in the prevalence of primary open-angle glaucoma. The Baltimore Eye Survey. JAMA 1991;266:369-74.

17. Grodum K, Heijl A, Bengtsson B. Refractive error and glaucoma. Acta Ophthalmol Scand 2001;79:560-6.

18. Population of the Netherlands on 01-01-2009 by origin.Available at: www.cbs.nl. Accessed August 30, 2009.

19. Stoutenbeek R, de Voogd S, Wolfs RC, et al. The additional yield of a periodic screening programme for open-angle glaucoma: a population-based comparison of incident glaucoma cases detected in regular ophthalmic care with cases detected during screening. Br J Ophthalmol 2008;92:1222-6.

20. Stoutenbeek R, Jansonius NM. Glaucoma screening during regular optician visits: can the population at risk of developing glaucoma be reached? Br J Ophthalmol 2006;90:1242-4.

138

General discussion

139

21. Mitchell P, Hourihan F, Sandbach J, Wang JJ. The relationship between glaucoma and myopia: the Blue Mountains Eye Study. Ophthalmology 1999;106:2010-5.

22. Wong TY, Klein BE, Klein R, et al. Refractive errors, intraocular pressure, and glaucoma in a white population. Ophthalmology 2003;110:211-7.

23. de Voogd S, Ikram MK, Wolfs RC, et al. Incidence of open-angle glaucoma in a general elderly population: the Rotterdam Study. Ophthalmology 2005;112:1487-93.

24. Hattenhauer MG, Johnson DH, Ing HH, et al. The probability of blindness from open-angle glaucoma. Ophthalmology 1998;105:2099-104.

25. Chen PP. Blindness in patients with treated open-angle glaucoma. Ophthalmology 2003;110:726-33.

26. Much JW, Liu C, Piltz-Seymour JR. Long-term survival of central visual field in end-stage glaucoma. Ophthalmology 2008;115:1162-6.

27. Tuulonen A, Airaksinen PJ, Erola E, et al. The Finnish evidence-based guideline for open-angle glaucoma. Acta Ophthalmol Scand 2003;81:3-18.

28. Burr JM, Kilonzo M, Vale L, Ryan M. Developing a preference-based Glaucoma Utility Index using a discrete choice experiment. Optom Vis Sci 2007;84:797-808.

29. Sach TH, Foss AJ, Gregson RM, et al. Second-eye cataract surgery in elderly women: a cost-utility analysis conducted alongside a randomized controlled trial. Eye 2009.

30. Bonfill X, Marzo M, Pladevall M, et al. Strategies for increasing women participation in community breast cancer screening. Cochrane Database Syst Rev 2001;CD002943.

31. Key figures of the optician market.Available at: www.nuvo.nl. Accessed September 20, 2009.

141

Chapter 9 Summary

Chapter 9

Glaucoma is a common eye disease that causes glaucomatous optic neuropathy with subsequent visual field loss, and can eventually lead to irreversible blindness. Early stages do not cause any symptoms, and visual field loss often goes unnoticed initially. Patients become aware of their eye disease only after extensive damage has occurred. Therefore, screening seems a logical approach to reduce the glaucoma burden. At present, a glaucoma screening programme does not exist in the Netherlands, but opportunistic case finding by ophthalmologists, optometrists, and opticians has become common practice. Despite case finding, about half of all glaucoma cases are currently undetected and thus do not receive any treatment or monitoring. Population based screening could help identify these individuals. Wilson and Jungner proposed a set of criteria for appraising the validity of a screening programme. Chapter two (literature review) provides an overview of relevant literature regarding different aspects of glaucoma screening based on these criteria. Most of the criteria are satisfied; characteristics of glaucoma which make it suitable for screening are: glaucoma is a common disease and a leading cause of blindness, the detectable preclinical phase is long in glaucoma; early treatment is advantageous; diagnostic tests are non-invasive and do not use ionizing radiation; and finally, preventing blindness has a large beneficial effect on quality of life. Problems identified in the literature review are mainly related to the cost-effectiveness of glaucoma screening, which is dubious at best. Chapters three and four assess the diagnostic performance of the Frequency Doubling Technology perimeter (FDT), a device suitable for glaucoma screening. In chapter three, FDT screening mode is compared to full-threshold mode. Both modes perform similar in terms of by-patient sensitivity and specificity, but they require a different cut-off point: one or more missed points in screening mode is equivalent to two or more missed points in full-threshold mode. For the screening mode, we found a sensitivity of 91% and a specificity of 88%. For the full-threshold mode, we found a sensitivity of 91% and a specificity of 83%. Both modes had an area under the receiver operating characteristic curve of 93%. Chapter four explores several strategies to improve the diagnostic specificity of the FDT. Confirming an abnormal FDT test result with a repeat test yielded a specificity increase from 80% to 90%, at the expense of some loss of sensitivity to early but not to moderate or severe glaucoma. Combining several FDT parameters from a single test or combining FDT parameters with GDx Nerve Fiber Analyzer parameters did not yield any noticeable increase in diagnostic performance. Glaucoma screening by opticians is a potential alternative to population based screening. An important prerequisite is that the population at risk for glaucoma must visit an optician sufficiently frequent. In chapter five, the optician visiting frequency of Dutch inhabitants over age 40 is determined by questionnaire.

142

Summary

143

Response rate was 80% (959 of 1200 inhabitants) and 80% of responders had visited an optician during a five year period. Another prerequisite is willingness of opticians to participate in a glaucoma screening programme. A second questionnaire was sent to 50 opticians, 37 of whom responded, and 91% of responders expressed willingness to participate in an extended glaucoma screening programme (i.e. more extensive than tonometry). In chapter six, the additional yield of a periodic glaucoma screening programme compared to current opportunistic case finding is estimated. Part of the incident glaucoma cases identified by the Rotterdam Study were found to have also been detected by regular ophthalmic care outside the study, whereas other cases had remained undetected during the entire follow-up interval. Twenty-three cases (29%) had already been detected, 55 cases (71%) remained undetected. The severity of glaucoma was worse in detected cases compared to undetected cases (P=0.009). The additional yield of screening is therefore lower than would be expected from prevalence data: we estimated that only about one in 1000 screened individuals could be saved from bilateral end-stage glaucoma. In chapter seven, supra-threshold perimetry (STP) is compared to full-threshold standard automated perimetry (SAP). Because of extensive clinical experience with SAP, glaucoma severity can be assessed adequately by quantitatively relating the number missed points on STP to the SAP mean deviation (MD). Results show a linear relationship between the number of missed points on STP and the SAP MD with a correlation coefficient of -0.92. The MD can be estimated from the STP score by multiplying the number of missed points by -0.75. STP was nearly twice as fast as SAP. STP appears to be a fast and reliable method for estimating the severity of glaucomatous damage. Whether or not a population based screening programme should be introduced in the Netherlands is discussed in chapter eight (general discussion). Cost-effectiveness studies regarding glaucoma screening indicate that economic viability is dubious. In this thesis, several factors are identified that have a further negative impact on the cost-effectiveness of glaucoma screening. These factors are: lengthbias, suboptimal by-patient test specificity, and large clinical workload from false positives. This leads to the conclusion that at present population based glaucoma screening is not feasible in the Netherlands. Optician based screening is a promising alternative, and will be subject of further research projects.

145

Chapter 10 Samenvatting

Chapter 10

Glaucoom is een veel voorkomende oogziekte die glaucomateuze opticopathie veroorzaakt met gezichtsveld uitval als gevolg, en kan uiteindelijk leiden tot onomkeerbare blindheid. Aanvankelijk zijn er geen symptomen en beginnende gezichtsveld uitval blijft vaak onopgemerkt. Patiënten ontdekken hun oogziekte pas nadat er al uitgebreide schade is opgetreden. Daarom lijkt screening een logische aanpak ter vermindering van de glaucoom problematiek. Op dit moment vindt er geen systematische glaucoom screening plaats in Nederland. Wel is het gebruikelijk dat mensen bij een bezoek aan de oogarts, optometrist of opticien gecontroleerd worden op de aanwezigheid van verhoogde oogdruk en/of glaucomateuze opticopathie; dit wordt ‘case finding’ of ‘actieve geneeskunde’ genoemd. Ondanks actieve geneeskunde blijft op enig moment ongeveer de helft van de glaucoom patiënten onopgemerkt, en staat dus niet onder behandeling of controle bij een oogarts. Met bevolkingsonderzoek zouden deze individuen opgespoord kunnen worden. Wilson en Jungner hebben een aantal criteria geformuleerd waaraan de validiteit van een screeningsprogramma kan worden getoetst. Hoofdstuk 2 (literature review) geeft een overzicht van relevante literatuur met betrekking tot de verschillende aspecten van glaucoom screening gebaseerd op deze criteria. Aan de meeste van deze criteria wordt voldaan. Kenmerken van glaucoom die het tot een voor screening geschikte oogziekte maken, zijn: glaucoom is een belangrijke oorzaak van blindheid; de detecteerbare pre-klinische fase duurt lang; vroegtijdige behandeling is gunstig; diagnostische onderzoeken zijn niet invasief en maken geen gebruik van ioniserende straling; en tenslotte, het voorkómen van blindheid heeft een groot positief effect op de kwaliteit van het leven. Problemen die bij het literatuur onderzoek geïdentificeerd werden zijn met name gerelateerd aan de kosten-effectiviteit van glaucoom screening, welke op zijn best twijfelachtig is. Hoofdstuk 3 en 4 beoordelen de diagnostische prestatie van de Frequency Doubling Technology perimeter (FDT), een voor glaucoom screening potentieel geschikt apparaat. In hoofdstuk 3 wordt de FDT screening mode vergeleken met de full-threshold mode. Beide FDT modi presteren gelijkwaardig wat betreft sensitiviteit en specificiteit, maar vereisen een verschillend afsnijdpunt: één of meer gemiste punten in screening mode is equivalent aan twee of meer gemiste punten in full-threshold mode. Voor de screening mode vonden we een sensitiviteit van 91% en een specificiteit van 88%. Voor de full-threshold mode vonden we een sensitiviteit van 91% en een specificiteit van 83%. Beide modi hadden een oppervlakte onder de receiver operating characteristic curve van 93%. Hoofdstuk 4 exploreert verschillende strategieën om de diagnostische specificiteit van de FDT te verbeteren. Het bevestigen van een afwijkende test uitslag door de test te herhalen levert een verbetering op van de specificiteit van 80% naar 90%. Dit gaat wel gepaard met een bescheiden afname in sensitiviteit voor vroeg glaucoom, maar gaat niet ten koste van de sensitiviteit voor matig of

146

Samenvatting

ernstig glaucoom. Het combineren van verschillende FDT parameters uit dezelfde test, of het combineren van FDT parameters met de GDx Nerve Fiber Analyzer test parameters leidde niet tot een verbetering van diagnostische prestaties. Glaucoom screening door opticiens is een mogelijk alternatief voor bevolkingsonderzoek. Een belangrijk vereiste is dat het deel van de populatie dat risico loopt op het ontwikkelen van glaucoom voldoende vaak een opticien bezoekt. In hoofdstuk 5 is de frequentie van opticien bezoek van Nederlandse inwoners ouder dan 40 jaar onderzocht met behulp van een vragenlijst. De respons was 80% (959 van de 1200 personen stuurden de enquête ingevuld retour) en 80% van de respondenten had de afgelopen 5 jaar een opticien bezocht. Een ander vereiste is bereidwilligheid van opticiens om te participeren in een glaucoomscreening programma. Een tweede vragenlijst werd verstuurd naar 50 opticiens, waarvan er 37 reageerden. 91% van de respondenten gaf aan te willen participeren in een uitgebreidere vorm van glaucoom screening (d.w.z. uitgebreider dan alleen oogdruk meting). In hoofdstuk 6 wordt de additionele opbrengst van invoering van een periodiek glaucoomscreening programma geschat ten opzichte van de huidige opportunistische situatie bestaande uit case finding. Een deel van de incidente glaucoom casus die opgespoord zijn door de ERGO studie (Erasmus Rotterdam Gezondheid Onderzoek – the Rotterdam Study) bleken ondertussen ook al te zijn ontdekt via de reguliere gezondheidszorg, terwijl andere gevallen onopgemerkt waren gebleven gedurende het gehele follow-up interval van de ERGO studie. Drieëntwintig gevallen (29%) waren ondertussen ontdekt, 55 gevallen (71%) bleven onopgemerkt. De mate van glaucomateuze schade was ernstiger in de wel ontdekte gevallen vergeleken met de niet ontdekte gevallen (P=0.009). Het additionele voordeel van bevolkingsscreening is daarom geringer dan verwacht zou mogen worden op grond van prevalentie data: naar schatting zal slechts 1 op de 1000 gescreende individuen gered kunnen worden van bilateraal eindstadium glaucoom. In hoofdstuk 7 wordt supra-threshold perimetrie (STP; boven-drempelig) vergeleken met full-threshold standaard automatische perimetrie (SAP). Vanwege de uitgebreide klinische ervaring met SAP kan de ernst van glaucomateuze uitval nauwkeurig bepaald worden door het aantal gemiste punten op de STP testuitslag kwantitatief te relateren aan de Mean Deviation (MD) van de SAP testuitslag. Uit de resultaten blijkt dat er een lineair verband is tussen het aantal gemiste punten bij STP en de SAP MD, met een correlatie coëfficiënt van -0.92. De MD kan geschat worden op basis van een STP test door het aantal gemiste punten te vermenigvuldigen met -0.75. STP was bijna twee keer zo snel als SAP. STP lijkt een snelle en betrouwbare methode te zijn om de ernst van glaucomateuze schade in te schatten.

147

Chapter 10

Of er wel of niet een bevolkingsonderzoek naar glaucoom zou moeten worden geïntroduceerd in Nederland is onderwerp van discussie in hoofdstuk 8. Uit kosten-effectiviteit studies betreffende glaucoomscreening blijkt dat de economische haalbaarheid twijfelachtig is. In dit proefschrift zijn verschillende factoren geïdentificeerd die een verdere negatieve impact hebben op de kosten-effectiviteit. Deze factoren betreffen: lengthbias, suboptimale specificiteit van potentiële screeningstests, en daaraan gerelateerd de hoge klinische werkdruk die zal resulteren van fout positieve test uitslagen. Dit leidt tot de conclusie dat bevolkingsonderzoek naar glaucoom op dit moment niet haalbaar is in Nederland. Glaucoomscreening via het bestaande netwerk van opticiens is een veelbelovend alternatief, en is onderwerp van een nieuw onderzoeksproject dat recent van start is gegaan.

148

Samenvatting

149

Dankwoord Dit proefschrift is tot stand gekomen in de periode 2003 tot 2009 in combinatie met mijn opleiding tot oogarts. Tijdens de verschillende onderzoeksprojecten heb ik met allerlei mensen samengewerkt, die ik graag wil bedanken voor hun bijdrage. Prof. dr. J.M.M. Hooymans, mijn promotor, wil ik graag bedanken voor het vertrouwen dat ze in mij stelde door mij een opleidingsplaats toe te kennen bij de oogheelkunde in het UMCG. Beste Anneke, ik waardeer je informele prettige manier van samenwerken en overleggen, en je hulp met de afronding van dit proefschrift. Daarnaast vind ik het erg leuk om nu deel uit te maken van de staf. Dr. N.M. Jansonius, mijn copromotor, was voor mij de ideale begeleider. Beste Nomdo, je doet veel moeite om een goed onderzoeksklimaat te creëren voor de verschillende promovendi onder jouw hoede, en dat lukt uitstekend. Je hebt op alle vlakken bijgedragen aan dit proefschrift. Door je scherpe analytisch vermogen wist je steeds inventieve oplossingen te bedenken voor de diverse epidemiologische en statistische vraagstukken die we zijn tegengekomen. Ik heb door de jaren heen veel van je geleerd, en ben je erg dankbaar voor alle hulp. Prof. dr. P.T.V.M. de Jong ben ik zeer erkentelijk voor zijn bijdrage aan hoofdstuk 6 en zijn bereidheid om als promotor op te treden. Prof. dr. J.E.E. Keunen, prof. dr. J.B. Jonas en prof. dr. E. Buskens dank ik voor hun bereidwilligheid zitting te nemen in de leescommissie. Mijn grote broer en tevens paranimf Ysbrand Stoutenbeek. Ys, je bent een geweldig fijne broer, en ik ben trots dat je bij mijn promotie aanwezig zult zijn als paranimf. Collega en paranimf Rogier Müskens. Beste Rogier, vanaf dat ik als keuze-coassistent bij de oogheelkunde kwam werken heb je mij op sleeptouw genomen. Als collega in zowel de patiëntenzorg als het glaucoom onderzoek hebben we veel samengewerkt; je enorm vriendelijke en collegiale opstelling

151

waardeer ik zeer. Lunchtijd is altijd erg gezellig, werktijd trouwens ook want je beheerst de kunst om een praatje te maken tijdens mijn spreekuur als je de kans krijgt. Alleen je missie mij aan de koffie te krijgen is nog niet gelukt, maar blijf het vooral proberen ;-). Dank dat je mijn paranimf wilt zijn. Beste Govert Heeg, bedankt voor de prettige samenwerking bij de tot stand koming van de eerste twee artikelen. Aangezien je destijds praktisch in de assistenten kamer woonde was je altijd bereikbaar maar ook altijd bereidwillig voor overleg en een helpende hand, wat ik zeer op prijs heb gesteld. Ik hoop dat je master plan met de Jaguar zijn vruchten af zal werpen, lijkt mij waterdicht, maar pas wel op voor paaltjes op parkeerplaatsen. Beste Berry Middel, dank voor je bijdrage aan hoofdstuk 5. Het maken van een goede enquête bleek een hele andere tak van sport dan klinische patiënten metingen, en jouw ervaring op dit gebied was essentieel. De arts assistenten oogheelkunde wil ik graag bedanken voor alle gezelligheid en collegialiteit: oud assistenten Carolien, Tanja, Angela, Kochoi, Lonneke, Govert, Rogier, Nicole, Karen en Esther, alsmede de huidige assistenten Dyonne, Suzan, Dirk, Theo, Danna, Annemiek, Eylem, Marieke, The Anh, Rosemarie, Stijn en Mathijs. Het is altijd een hechte groep geweest, waar ik vijf jaar lang met veel plezier deel van heb uitgemaakt (en nu met ijzeren hand supervisie aan geef ;-)). Daarnaast wil ik de stafleden bedanken voor hun aandeel in mijn opleiding en de prettige samenwerking. Nu ik de assistenten groep moet missen is het goed te merken dat de staf etentjes net zo gezellig zijn. Ook veel dank aan alle andere medewerkers op de poli (TOA’s, optometristen, orthoptistes, verpleegkundigen, planners en de administratie) voor de prettige sfeer op de werkvloer en voor alle hulp met de gezichtsveld metingen. Ella, Fenna, Stella en Albert wil ik bedanken voor de goede ondersteuning. De onderzoekers van de ERGO Studie in Rotterdam dank ik voor de plezierige vergaderingen en het beschikbaar stellen van onderzoeksdata ten bate van hoofdstuk 6. Lieve familie, Peter, Roos en Ysbrand, dank voor jullie onvoorwaardelijke liefde en steun. Met jullie achter me staand heb ik het zo ver kunnen brengen. Bedankt voor jullie grote interesse in al mijn activiteiten, en de stimulans om overal het beste uit te halen. De laatste loodjes waren zwaar, het is fijn om dan op jullie terug te kunnen vallen. Lieve Marlous, bedankt voor de fijne tijd die we samen gehad hebben.

152

Tot slot wil ik Pepijn, Jan Willem, Michiel, Jeroen, Rouben, Krischan, Sander, Raphael, Ferdie, Willem, Cor en Alex bedanken voor hun vriendschap en voor de nodige afleiding tussen de bedrijven door. Remco Stoutenbeek, november 2009

153

154

Curriculum Vitae Remco Stoutenbeek was born in Deventer, the Netherlands, on the 28th of March 1978. After completing secondary school at the Geert Groote College in Deventer in 1996, he started his medical education in Groningen. He obtained a Masters degree in medicine and graduated as a Medical Doctor in August 2003. After a few months of working at the emergency department of Delfzicht hospital in Delfzijl, he started his ophthalmologist training in February 2004 at the University Medical Center Groningen (UMCG), department of Ophthalmology. Work on this thesis had started in 2003 as a scientific program as part of the medical curriculum, and was now resumed under ongoing supervision by dr. Jansonius. In February 2009, he completed his training as ophthalmologist, and currently works in the UMCG as cornea fellow under supervision of drs. Wijdh. After completion of the fellowship, he will continue his practice at the UMCG as a staff member. Correspondence: r.stoutenbeek@ohk.umcg.nl