Association between recombinant human erythropoietin and qualityof life and exercise capacity of patients receiving haemodialysis

Canadian Erythropoietin Study Group

AbstractObjective-To determine whether recombinant

human erythropoietin improves the quality of lifeand exercise capacity of anaemic patients receivinghaemodialysis.Design-A double blind, randomised, placebo

controlled study.Setting-Eight Canadian university haemodialysis

centres.Patients-118 Patients receiving haemodialysis

aged 18-75 with haemoglobin concentrations<90 g/l, no causes of anaemia other than erythro-poietin deficiency, and no other serious diseases.

Interventions-Patients were randomised to threegroups to receive placebo (n=40), erythropoietin toachieve a haemoglobin concentration of 95-110 g/l(n=40), or erythropoietin to achieve a haemoglobinconcentration of 115-130 g/l (n=38). Erythropoietinwas given intravenously thrice weekly, initially at100 units/kg/dose. The dose was subsequentlyadjusted to achieve the target haemoglobin con-centration. All patients with a serum ferritinconcentration less than 250 Ftg/l received oral orintravenous iron for one month before the study andas necessary throughout the trial.Main outcome measures-Scores obtained with

kidney disease questionnaire, sickness impactprofile, and time trade off technique; and results ofsix minute walk test and modified Naughton stresstest.Results-The mean (SD) haemoglobin con-

centration at six months was 74 (12) g/l in patientsgiven placebo, 102 (10) g/l in those in the lowerythropoietin group, and 117 (17) g/l in those inthe high erythropoietin group. Compared with theplacebo group, patients treated with erythropoietinhad a significant improvement in their scores forfatigue, physical symptoms, relationships, anddepression on the kidney disease questionnaire andin the global and physical scores on the sicknessimpact profile. The distance walked in the stress testincreased in the group treated with erythropoietin,but there was no improvement in the six minute walktest, psychosocial scores on the sickness impactprofile, or time trade off scores. There was nosignificant difference in the improvement in qualityof life or exercise capacity between the two groupstaking erythropoietin. Patients taking erythropoietinhad a significantly increased diastolic blood pressuredespite an increase in either the dose or number ofantihypertensive drugs used. Eleven of 78 patientstreated with erythropoietin had their sites of accessclotted compared with only one of 40 patients givenplacebo.

Conclusions-Patients receiving erythropoietinwere appreciably less fatigued, complained of lesssevere physical symptoms, and had moderateimprovements in exercise tolerance and depres-sion compared with patients not receivingerythropoietin. At the doses used in this trial therewas a higher incidence of hypertension and clottingof the vascular access in patients treated witherythropoietin.

IntroductionRecombinant human erythropoietin is given to

correct the anaemia of end stage renal disease' andthus decrease the likelihood of infections associatedwith transfusion and overload of iron. Uncontrolledstudies suggest that erythropoietin improves generalwellbeing, symptoms of fatigue, exercise tolerance,'and cognitive and sexual function.6 Adverse effectsthat are attributed to erythropoietin in patientsreceiving haemodialysis include hypertension, flu-likesymptoms, hyperkalaemia, seizures, and clottedvascular accesses."' These are common problems insuch patients, and a causal role for erythropoietin hasnot been established.We studied the effect of erythropoietin on the

quality of life, exercise capacity, and putative adverseeffects among anaemic patients receiving haemo-dialysis.

Patients and methodsWe studied 1 8 anaemic patients who were receiving

haemodialysis at 13 hospitals in eight Canadianuniversity centres. Patients were stratified by hospitaland randomised in blocks to receive placebo; erythro-poietin at a dose adjusted to maintain the haemoglobinconcentration at 95-110 g/l (low erythropoietin group);or erythropoietin at a dose adjusted to maintain thehaemoglobin concentration at 115-130 g/l (higherythropoietin group). The study was approved by theethics committees of all the hospitals.

Patients were included in the study if they were aged18 to 75, had been clinically stable for three months,had had an average haemoglobin concentration of<90 g/l during the preceding three months, and hadhad haemodialysis three times a week in a hospital orself care unit. Patients were excluded if their anaemiawas not caused by erythropoietin deficiency; theirquality of life or exercise capacity was affected byfactors other than renal failure; they were unable toperform a six minute walk test'; they would not be ableto understand the questionnaires we were going to usebecause of language or intellectual difficulties; or theywere unwilling or unable to give informed consent.

Erythropoietin was injected intravenously as a 10 mlbolus at the end of each session of dialysis (threetimes a week). The dose in both groups receivingerythropoietin was initially 100 units/kg. Whenpatients' haemoglobin concentration reached the lowerend of the target range the dose was usually decreasedby 25-50 units/kg at weekly intervals to maintain astable concentration within the range. If the targetconcentration was not achieved by eight weeks in thosein the low erythropoietin group or 12 weeks in those inthe high erythropoietin group the dose was increasedby 25-50 units/kg. To avoid iron depletion patientswith a serum ferritin concentration less than 250 [tg/lwere given iron orally or intravenously for one monthbefore randomisation and as necessary throughout thetrial.To ensure that the study was double blind we

established two teams of staff at each study centre. Theunblinded team consisted of a doctor, a pharmacist,

BMJ VOLUME 300 3 MARCH 1990

Members of the CanadianErythropoietin Study Groupare given at the end of thispaper.

Correspondence andrequests for reprints to:Dr Andreas Laupacis,Robarts Research Institute,PO Box 5015, London,Ontario N6A 5KB, Canada.

BrMed3 1990;300:573-8

573

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from

and a data clerk and was responsible for adjusting thedose of erythropoietin, prescribing iron supplementsor transfusions, and sending haematological data to thecoordinating centre. The blinded team consisted ofnurses in the dialysis unit and our study group and alldoctors in the dialysis unit other than those in theunblinded team; this team carried out routine clinicalcare and recorded adverse reactions and other clinicalevents but did not have access to the results ofhaematological tests or know the dose of erythropoietinor placebo that each patient was receiving. The nursesin the study group administered tests to assess qualityof life and exercise capacity.

All patients were followed up for six months, andoutcome measures were assessed before randomisationand at two, four, and six months. The two mainoutcomes assessed were quality of life and exercisecapacity. We thought that quality of life would bemost effectively assessed by a questionnaire aimedspecifically at patients receiving haemodialysis. To putthe results into perspective, however, we included twoglobal measures of quality of life.We thus developed a kidney disease questionnaire

using the methods suggested by Guyatt et al.' We alsoused the sickness impact profile, which assesses healthstate based on behaviour and has been used in thenational kidney dialysis and kidney transplant study,9and the time trade off technique described by Churchillet al," from which a "utility" is derived for eachpatient. These three measures of quality of life aredescribed in more detail in the appendix.To assess exercise capacity we used the six minute

walk test and an exercise stress test. The six minutewalk was conducted in enclosed corridors that werebetween 25 and 40 m long. Patients were instructed towalk from end to end, covering as much ground aspossible in the allotted time. We gave patients anexercise stress test using a modified Naughtonprotocol2 conducted in 11 two minute stages, with thespeed of the treadmill increasing from 1 -6 to 5 5 km/hand the incline increasing from 0 to 16%. If the patientwas able to exercise for 22 minutes the test wascontinued for a further eight minutes at a speed of5 5 km/h and a gradual increase in the incline from 16%to 25%. Standard encouragement was given duringboth exercise tests.We calculated the size of sample that would ensure

that a clinically important difference among the threegroups at six months would be detected with the timetrade off technique, sickness impact profile, and sixminute walk test (the only outcome measures for which

TABLE i-Characteristics ofpatients at beginning ofstudy into effects oflow and high doses oferythropoietin.Values are means (SD) unless stated otherwise

Placebo Low ervthropoietin High ervthropoietingroup group group(n=40) (n=40) n-38)

Age (years) 48 (16) 44 (16) 43 (15)No ofmen 25 19 26No who were anephric 6 13 10Nowho had had a transplant 14 16 12Duration of haemodialysis (years)* 25 (3 1) 46(47) 44 51)No of transfusions during previous year 7-3 (8 3) 6 6 (6 8) 5 6 (7 4)No dependent on transfusiont 19 19 11Systolic blood pressure (mm Hg)t 143 (19) 137 (22) 143 (21)Diastolic blood pressure (mm Hg)# 80 (12) 78 (14) 78 12))No receiving antihypertensive drugs 10 13 12Haemoglobin (g/l) 71 (9) 69 (10) 71 (12)Serum iron ([tmol/l) 19 (14) 18 (9) 22 (15)Serum ferritin (ug/l) 703 (866) 1044 (1690) 1273 (2303)Median serum ferritin ([g/l) 318 406 345Serum creatinine (4imolI1) 1057 (246) 1090 (260) 1076 290)Serum urea (mmol/l) 28-0 (6 0) 27-3 (7 0) 27 8 (7 2)Serum potassium (mmol/l) 5-0 (0 6) 5-1 (0 6) 51 (0-7)

*p=0 07 Among groups.tPatients who had received dialysis for one or more years were considered to be dependent on transfusion if theyreceived six or more units of blood in the year before randomisation. Patients who had received dialyzsis for less thanone year were considered to be dependent on transfusion if they had received more than two units of blood in thethree months before randomisation.tPlacebo group n= 37; low erythropoietin group n= 39; high ervthropoietin group n= 37.

data on variance were available before the studv). " Werequired 30 patients in each group (two tailed (C of 0 05and a power of 90o). Assuming a withdrawal rate of25% due to renal transplantation, adverse effects, andconcomitant events, this meant that we needed asample of 120 patients.The outcome variables were analysed only for

patients who had completed all four evaluations(before randomisation and at two, four, and sixmonths). Analysis of variance for repeated measureswas used. The main hypothesis tested was that themean response profiles of v . rariables evaluated at thefour time points were the same among the tl-ireegroups.4 In addition to the overall comparison amongthe three treatment groups, two orthogonal contrastswere specified to compare the patients receivingplacebo with the patients treated with erythropoietin(both groups) and the patients in the low erythropoietingroup with those in the high erythropoietin group.Pearson's correlation coefficient was computedto examine the relation between the change inhaemoglobin concentration and the change in qualitvof life and exercise capacity. Kendall's T statisticand the corresponding p values were computed todetect any significant associations between theincidence of side effects and treatment witherythropoietin.

ResultsPATIENT CHARACTERISTICS

Forty patients were randomised to receive placebo,40 low dose erythropoietin, and 38 high dose erythro-poietin. Renal failure was due to glomerulonephritis(50 patients), pyelonephritis or interstitial nephritis(24), polycystic kidney disease (14), renal vasculardisease (seven), chronic renal failure of unknown cause(three), nephritis induced by drugs (two), and othercauses (18). Table I shows the characteristics ofpatients in the three groups at the beginning of thestudy; the three groups were comparable. Patientsreceiving placebo had been treated by dialysis for ashorter time than the two other groups (p=0 07).When age, duration of dialysis, whether patients hadreceived a transplant or were anephric, dependence ontransfusion, and antihypertensive treatment at thetime of randomisation were examined with effect oftreatment in the analysis the results of the unadjustedanalysis were hardly affected.

WITHDRAWALS FROM STUI)Y

Nineteen patients were withdrawn during the study:eight in the placebo group (because of transplarntation(five), non-compliance (one), reaction to transfusion(one), seizure and death (one)); six in the low erythro-poietin group (transplantation (two), hypertension(one), hypertension and seizure (one), subarachnoidhaemorrhage and seizure (one), pregnancy (one)); andfive in the high erythropoietin group (transplantation(three), hypertension (two)). Six patients werewithdrawn before the follow up at two months, and the13 others were withdrawn before the follow up at fourmonths. The patient who became pregnant continuedto receive erythropoietin but had a spontaneousmiscarriage at 11- 1 2 weeks' gestation.

CHANGES IN HAEMOGLOBIN CONCENTRATION

Figure 1 shows the changes in haemoglobinconcentration in the three groups. The mean (SD)haemoglobin concentration at six months was 74(12) g/l in the placebo group, 102 (10) g/l in the lowerythropoietirl group, and 117 (14) g/l in the higherythropoietin group. The mean haemoglobinconcentration reached the target range by six weeks inpatients in the low erythropoietin group and by

BMJ VOLUME 300 3 MARCH 1990574

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from

IABLE I- Scores for qualitv of liJe and exercise capacity obtained for patienits wzith end stage renal disease before and two and six months aflter start oftreatment with placebo or low orhigh doses of ervthropoietin

Placebo group(n=32

Low ervthropoietin groupn =34)

High erythropoietin group(n = 33)

At At At At At AtBefore 2 months 6 months Before 2 months 6 months Before 2 months 6 months

Significance ofdifference

Among the Ervthropoietin High dose vthree z} low dosegroups placebo ervthropoletin

Kidney disease questionnaire:Phvsical*FatigueRelationshipsDepressionFrustration

Sickness impact profilet:GlobalPhysicalPsychosocial

Time trade off techniqueExercise stress test: (minutes walkedlSix minute walk test' distance walked m

*Placebo group n= 31.tLowest score represents best quality of life.

tPlacebo group n= 29; low ervthropoietin group n= 31; high ers thropoietin group n= 27.,Placebo group n= 31; low erythropoietin group n 32; high erv thropoietin group n= 30.

TABLE III-Symptoms defined by patients as being a problem associated with end stage renal disease before and two and six months after start of treatment with placebo or low or highdoses oferythropozetin

Placebo group

FatigueDecreased strengthSleeping abnormalitiesAching legs and bonesShortness of breathSexualitV

Noof

patients

674545352018

At AtBefore 2 months 6 months Before

4 14 14 03 63-64-6

4 64.74.34.44.74-6

4 14-24.54.34.45 .2

Low ervthropoietin group High erythropoietin group

At At At At2 months 6 months Before 2 months 6 months

3-12-83 83 04.33 0

4 34-04-04 25.44 2

5 45 34 04.45.94 2

37 4.9 504-0 5 3 5 34.3 5 3 564-0 5-0 5.74-2 5 8 5-83-0 3.9 5.0

Significance of diffeience

Among the Erythropoietin High dose z'three Z) low dosegroups placebo ervthropoietin

<0.001<0 001NSNSNSNS

<0 001<0 001NSNSNSNS

100

90

30

E 80 / T

70. 'IIII

*I

60 IJ50

40

30

0 2 4 6 8 10 12 14 16 18 20 22 24 26

WeekFIG 1-Mean (SD) changes in haemoglobin concentration in three treatment groups during

11 weeks in patients in the high erythrop4After eight weeks 23 patients in the placebin the low erythropoietin group, and oneervthropoietin group had had blood transpatient in the low erythropoietin grotransfusion because of gastrointestinal hand the patient in the high erythropoietintransfusion during a hysterectomy.The mean (SD) dose of erythropoietin o

was 204 (167) units/kg/week in the low er

Higherythropoietingroup

.Lowerythropoietingroup

0-0520-023NSNSNSNS

group and 248 (146) units/kg/week in the higherythropoietin group. At six months iron supplementshad been given to 11 patients in the placebo group,

12 in the low erythropoietin group, and 19 in the higherythropoietin group (p=0- 10 among groups); theywere given intravenously to four, five, and 11 of thepatients, respectively. At six months the mean changein serum ferritin concentration from the value beforetreatment was significantly different among the groups(+ 197 (662) [tg/l in the placebo group, - 341 (769) [tg/lin the low erythropoietin group, and -241 (953) Fg/l inthe high erythropoietin group; p=0 02).

QUALITY OF LIFE AND EXERCISE CAPACITY

The scores for quality of life before treatment were

comparable among the three groups except for the time

Placebo walked during the stress test (p-002, table II) and atrend for the section on physical symptoms in thekidney disease questionnaire (p=0 06). When thesetwo variables were reanalysed with the value beforetreatment being included as a covariate factor theresult for the stress test was more significant and no

substantial change was found for the physical effects.Data from the four month visits on quality of life andexercise capacity were used to calculate significancebut are not included in tables II or III as the improve-ments that occurred were generally apparent by two

28 months, and none of the scores at four months were

significantly better than those at six months.study Patients treated with erythropoietin showed a

significant and clinically important improvement inoietin group. their response to questions in the sections on fatigueo group, one and physical symptoms in the kidney disease question-e in the high naire. There were also smaller improvements in theirfusions. The response to questions in the sections on relationshipsup required and depression, but no significant improvement withLaemorrhage, regard to the section on frustration.group had a Among the physical symptoms mentioned by the

patient as most affecting his or her life there was a

at six months highly significant improvement in fatigue and strengthrythropoietin in patients treated with erythropoietin (table III).

BMJ VOLUME 300 3 MARCH 1990

424.44.95 04.9

10 34.99.10 42

11 4421

4.74.75 15 24.9

8 13 25 90-4613-0

437

4-64.55 05.14.97.44-24-80-4213 2

440

3 64 14.94.74.912 06 410.90-49

11 2418

4.74.95 35 05 0

8-54 08 10 5114 2

456

5 25 05 55 14.9

6 72 66-00 5114 8

451

3.94 24.94 84 5

12 26 3

11 80-5216 1

470

5 05 25 45 34-8

6-03.35 20 5819 8

509

5 35 35 55 54.9

4.42 43 00 5819 7

521

<0.001<0 0010 0090 045NS

0 0290 039NSNS0-104NS

<0.001<0.0010.00100 18NS

0-0240-005NSNS0-018NS

NSNSNSNSNS

NSNSNSNSNSNS

575

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from

TABLE Iv-Mean blood pressure (mm Hg) in the three treatment groups before (time 0) and two, four, and six months after start of treatment

Placebo group (n = 29) Low ervthropoietin group n =31) High erythropoietin group (n=32) Significance of difference

Among the Erythropoietin High dose V lOw dose0 2 4 6 0 2 4 6 0 2 4 6 three groups v placebo erythropoietin

Sy'stolic 147 144 142 143 137 138 140 137 144 146 141 144 NS NS NSDiastolic 80 78 77 79 76 79 80 78 78 84 83 85 0-023 <0-001 0 063

TABLE v-Number of patients reporting vanrous side effects while taking placebo or low or high doses ofervthropoietin

Placebo group Low ervthropoietin group High ervthropoietin group(n=40) (n =40) (n=38)

Seizure 1 2 0Clotting of vascular access* 1 4 7Clotting of tubing in dialvsis machine 4 4 4Pain in chest 6 7 6Epistaxis or haemorrhage 7 5 5Abnormal sense of taste 6 7 4Headache 19 13 13Redness of eyes 0 2 3Flu-like symptoms 12 9 9Aches in bone or muscle 9 8 12

*Thrombectomy with or without revision of fistula was required in one patient in the placebo group, three in the lowervthropoietin group, and six in the high ervthropoietin group.

4-

C 0

3 0

0 0 -0

2 0 0

0 0

1 C0 t $ 0m~ 0

U X ,D,'DO~*

-30 -20 -10 0 10 20 30 40 50 60 70 80 90

Change in haemoglobin (g /1)FIG 2 -Relation between changes in score forfatigue on kidney diseasequestionnaire and changes in haemoglobin concentration in threetreatment groups during study

Although erythropoietin seemed to cause an im-provement in most of the other physical symptomsmentioned, this was not significant, perhaps because ofthe fairly small number of patients complaining of eachsymptom.The global and physical scores on the sickness

impact profile improved in the patients treated witherythropoietin compared with those given placebo(table II); the average improvement in the globalscore achieved in six months was 5 3 in the lowerythropoietin group and 7 8 in the high erythropoietingroup but only 2 9 in the placebo group. The sicknessimpact profile detected significant improvements(p<005) in the response of patients to questionson body care and movement, home maintenance,ambulation, communication, and work amongthose treated with erythropoietin. There was noimprovement in the aggregate psychosocial score orin scores for sleep and rest, emotional behaviour,mobility, social interaction, alertness behaviour,recreation and pastimes, or eating.The time trade off technique showed no improve-

ment in overall quality of life. The difference in thetime walked in the exercise stress test between patientsgiven erythropoietin and those given placebo was

significant, but no significant improvement was foundin results of the six minute walk test.No significant difference between patients in the low

and high erythropoietin groups was found in thechange in any of the outcome measures assessed exceptdecreased strength (p=0-02, table III).

CORRELATION BETWEEN CHANGES IN HAEMOGLOBINCONCENTRATION AND CHANGES IN OUTCOME MEASURES

The mean haemoglobin concentrations beforerandomisation varied considerably among the patients,and some patients did not achieve their target con-centration. We therefore assessed the correlationbetween the change in haemoglobin concentration andthe change in quality of life and exercise capacity at sixmonths. A significant correlation was found betweenthe change in haemoglobin concentration and thechange in the global, physical, and psychosocial scoreson the sickness impact profile; all sections of the kidneydisease questionnaire except that on frustration; andthe time walked during the exercise test. The Pearson'scorrelation coefficients, however, were fairly low, thehighest being observed with the scores for the sectionson fatigue and physical symptoms in the kidney diseasequestionnaire (r=0 32, p=0 001 and r=0 31, p=0-002respectively; fig 2).

HYPERTENSION

Table IV shows the mean systolic and diastolic bloodpressures of the patients throughout the study. Therewas no significant difference in the change in systolicpressure throughout the study among the threegroups. Diastolic pressure, however, was significantlyincreased in the patients treated with erythropoietincompared with those given placebo (p<0001).The most striking increase was seen in the higherythropoietin group (p=0 006, high erythropoietin vplacebo; and p=0-063, high v low erythropoietin).Among patients who completed the trial antihyper-tensive treatment was either started or increasedin three patients given placebo compared witheight patients in the low erythropoietin group and10 patients in the high erythropoietin group. The doseof antihypertensive drugs was decreased in one patientin the placebo group, none of the patients in thelow erythropoietin group, and four patients in thehigh erythropoietin group. The association betweenerythropoietin treatment and an increase in the dose ofantihypertensive drugs approached conventionalsignificance (p= 0 06).Four patients (two in the low erythropoietin group

and two in the high erythropoietin group) hadhypertension that was sufficiently severe for theirdoctor to withdraw them from the study (at one, two,12, and 13 weeks after randomisation). None of thepatients had been taking antihypertensive drugs beforethe study. There were no permanent sequelae fromthese hypertensive episodes.

OTHER SIDE EFFECTS

Table V shows the incidence of other side effects.There was a positive association between treatmentwith erythropoietin and the incidence of clotting of thevascular access (p=001) and eye redness (p=004).There were no significant differences in the meanchanges in serum concentrations of potassium,

BMJ VOLUME 300 3 MARCH 1990576

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from

phosphorus, calcium, urea, and creatinine or whitecell count among the groups throughout the study. Atsix months the platelet count had fallen by a mean (SD)of 5 (39) x 109/1 in the placebo group and increased by25 (64)x109/l in the low erythropoietin group and23 (47) x 109/1 in the high erythropoietin group(p=0005, placebo v both erythropoietin groupsthroughout study).

BLINDING

At the end of the study all patients were askedwhether they thought they had received erythropoietinor placebo. Nineteen patients in the placebo groupand 52 patients receiving erythropoietin correctlyidentified their treatment. Six of the placebo groupthought that they had received erythropoietin whiletwo of the patients receiving erythropoietin believedthat they had been given placebo. The other patientscould not say which they had received.

DiscussionTreatment with erythropoietin corrected the

anaemia associated with end stage renal disease inpatients being treated by haemodialysis. Generally,patients receiving erythropoietin were much lessfatigued, complained of less severe physical symptoms,and had increased exercise tolerance compared withpatients given placebo. There was no significantdifference in quality of life or exercise tolerancebetween patients receiving low and high doses oferythropoietin. A significant correlation was foundbetween changes in haemoglobin concentration andchanges in quality of life and exercise capacity, butthe correlation coefficient was modest. Althougherythropoietin caused a significant improvement inquality of life and exercise capacity, many otherfactors, such as uraemia, bone disease, and thepsychological burdens of haemodialysis, also adverselyaffect the wellbeing of such patients, and these factorsmay not be improved by erythropoietin.The effect of erythropoietin on psychosocial

function was less impressive than that on fatigue,physical symptoms, and exercise tolerance. Somesignificant (but moderate) improvements were seen inresponses to the kidney disease questionnaire but notin the sickness impact profile. As the kidney diseasequestionnaire was developed for patients with renalfailure it is not surprising that it was more responsive inour patients.The two measures of overall quality of life yielded

different results, an improved quality of life beingdetected in patients treated with erythropoietin by thesickness impact profile but not by the time tradeoff technique. The time trade off technique allowspatients to give their own weighting to the physical,social, and emotional factors that affect their quality oflife, while the sickness impact profile applies fixedweights that may not precisely have reflected theconcerns of our patients. The time trade off technique,which asks patients how many years of their currenthealth state they would be willing to give up to receiveerythropoietin is a rigorous test; the fact that the timetrade off score did not change in patients takingerythropoietin probably indicates that despite the drugmany of the patients receiving dialysis still had severaladverse effects from their disease. The time trade offtechnique has been successfully used to detect changesin quality of life,'6 and complete insensitivity of the testis therefore unlikely to be the reason for the fairlyconstant scores in our study.We examined several outcome measures, and thus

multiple testing may possibly have caused some resultsto appear significant by chance. We expected the effectoferythropoietin on fatigue, however, to be the largest,

and the high levels of significance we obtained make itextremely unlikely that the results occurred by chancealone.

It is difficult to recommend confidently a targethaemoglobin concentration for patients treated witherythropoietin. In general, the patients' quality of lifeand exercise capacity rose with increasing haemoglobinconcentration, but patients in the high erythropoietingroup had a greater incidence of hypertension (andperhaps clotting of the vascular access) than patients inthe other groups. Decisions on each patient's targethaemoglobin concentration should be made clinicallytaking into account not only increased quality of lifeand exercise capacity but also the patient's likelihoodof developing hypertension or thrombosis.The ultimate clinical role of erythropoietin in the

treatment of anaemic patients receiving haemodialysiswill depend on the patient's need for transfusion,the ability of erythropoietin to prevent anti-HLAsensitisation, the cost of the drug, and the drug'sefficacy and side effects in other populations receivinghaemodialysis.

AppendixKIDNEY DISEASE QUESTIONNAIRE

The kidney disease questionnaire is specific forpatients with end stage renal disease and was developedfrom the methods described by Guyatt et al.8 Weinterviewed patients receiving haemodialysis andhealth care staff and reviewed existing quality of lifevariables. This generated a list of 130 items that couldpotentially have an adverse affect on the quality of lifeof the patients. Fifty patients receiving haemodialysiswere then asked to rank these items in terms of theirimportance.The items that were ranked most frequently and

judged to be most important by the patients wereincluded in the questionnaire. The final questionnaireconsisted of 26 questions divided into five sections:physical symptoms, fatigue, depression, relationships,and frustration. In the section on physical symptomseach patient was asked to identify those physicalsymtoms or problems that most affected his or her life,and these were used for that patient throughout thestudy.

All questions were scored on a seven point Likertscale (7=no problem, 1= a severe problem). Forexample, a question from the fatigue section, "Howoften in the past two weeks have you felt low inenergy?" was accompanied by a choice of answers: (1)All the time; (2) most of the time; (3) a good bit of thetime; (4) some of the time; (5) a little of the time;(6) hardly any of the time; and (7) none of the time.A change in mean score of 0 5 in each section

represented a minimal, clinically important difference,and a mean change of 1 0 represented a large clinicalchange. 17

SICKNESS IMPACT PROFILE

The sickness impact profile is a questionnaire basedon behaviour that contains 136 statements aboutdysfunction related to health in 12 types of activity:sleep and rest, eating, work, home management,recreation and pastimes, ambulation, mobility, bodycare and movement, social interaction, alertness,emotional behaviour, and communication.9 Some ofthese can be aggregated to give a physical score (bodycare and movement, mobility, and ambulation) anda psychosocial score (emotional behaviour, socialinteraction, ambulation, and communication). Anoverall score combining all questions can also bederived.When answering the questionnaire the patients were

asked to think about themselves that day and to

BMJ VOLUME 300 3 MARCH 1990 577

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from

determine which statements described them and wererelated to their health. The patient could answer onlyyes or no to each question. An example of a questionabout sleep and rest is, "I spend much of the day lyingdown in order to rest."

TIME TRADE OFF TECHNIQUE

The time trade off technique is used to derivea "utility," a score between 0 and 1, in which1 represents perfect health and 0 a state in which thepatient is indifferent between life and death." With thehelp of visual aids patients were asked how many yearsof their current health they would be willing to forgo toachieve perfect health. For example, if a 36 year oldpatient stated that she was unable to choose betweenfour years of perfect health and 40 years of her currenthealth the utility of her current health state was4/40=0 10 (she would be willing to give up 36 years ofher current health to achieve four years of perfecthealth).

The Canadian Erythropoietin Study Group comprises thefollowing groups. Executive committee: David Churchill,Paul Keown (chairman), Andreas Laupacis, NormanMuirhead, Dalice Sim, David Slaughter. Managementcommittee: Paul Keown, Andreas Laupacis (chairman),Norman Muirhead, Beth Sarazin, Denise Short, Dalice Sim,David Slaughter, Cindy Wong. Study centres: VictoriaGeneral Hospital, Halifax, Nova Scotia (Allan Cohen,* DavidHirsch, Kailash Jindal); Royal Victoria Hospital, Montreal,Quebec (Paul Barre,* Andrew Gonda, Tom Hutchinson,Sarah Prichard, Denis Roy); Ottawa Civic Hospital andOttawa General Hospital, Ottawa, Ontario (AndrewLazarovits,* Steven Nadler,* Gerald Posen, Eli Rabin); StJoseph's Hospital, Hamilton, Ontario (Colin Barnes, DavidChurchill,* Sandra Donnelly, David Ludwin, KinseySmith); University Hospital and Victoria Hospital, London,Ontario, and Laurentian Hospital, Sudbury, Ontario(William Fay, David Hollomby, Tony Jevnikar, PaulKeown, Norman Muirhead*); Health Sciences Centre and StBoniface Hospital, Winnipeg, Manitoba (Keevin Bernstein,Adrian Fine, John McKenzie,* Brian Penner, BrentSchacler, Ashley Thomson); Foothills Hospital, Calgary,Alberta (Ellen Burgess, Allan Jones, Henry Mandin*); StPaul's and Vancouver General Hospitals, Vancouver,

British Columbia (Anthony Chiu, Gershon Growe, DavidLandsberg, John Price, John Shepherd, Roger Sutton,*Linda Vickers, Ronald Werb). *=Principal investigators.We thank the study nurses, data coordinators, and

pharmacists at each of the centres; the support staff of RobartsResearch Institute; Ulrike Kuprath, Patricia Laplante,Angela Mongul, David Slaughter, and other staff at OrthoPharmaceutical (Canada) Ltd; and the patients. We alsothank Janice Coffey for secretarial help.

1 Eschbach JW, Egrie JC, Downing MR, Browne JK, Adamson JWCorrection of thc anicmia of end-stage renal disease with recombinant humancrythropoictin. Engl.ViMed 1987;316:73-8.

2 Winearls C(G, P'ippard AIj, t)owning MR, Oliver DO, Reid C, Cotes PMN.Ef'fect of htuman erythropoietin derived from recombinant DNA on theanaemia of patients tnaintaitted by chronic haemodialysis. Latncet1986;ii: 1 175-7.

3 Canadian Ersthropoietin Study Group. A prospective, randomized,douLble-blind stuLds of recombittant erythropoietin (r-Hu-EPO)0 in chronichemodialv sis [Abstract]. Kidsne Iti 1988;33:218A.

4 Eschbach JVI. 'rhc ancmia ot' chronic renal failure: pathophvsiologv and theeffect of recombinant crsthropoietin. Kidnicesv Int 1989;35:134-48.

5 Ltindin AP. Qualits ot' life: subjective and objectise improvements withrecombinant htuman crs thropoictin therapy. Semin Nephrol 1989;9:22-9.

6 Schacfer RAM, Kokot F, Wlcrnze H, Geiger H, Heidland A. Impros-ed sexualfunictioni in hcmodialssis patients on rccombinant erythropoietiit: a possiblerole for prolactin. Clin .\' phrol 1989;31:1-5.

7 Guvatt GH, Sullisan M1J, Fhompson P1, ct al. The 6-minute walk: a newmeasure of excrcise capacits in patients with chronic hcart failure. Can MedAssocj 1985;132:919-23.

8 Guyatt GH, Bombardier C, I'ugwell P. Measuring disease-specific quality oflife in clinical trials. Can Med .Assoc7 1986;134:889-95.

9 Bergner M, Bobbitt RA, Carter WB, Gilson BS. The sickniess impact profile:deselopment anid final res-ision of a health status measure. Med Care1981;19:787-805.

10 Hart LG, Evans RW. The functional status of ESRD patients as measured bythe sickness impact profile. ] Chronic Dis 1987;40(suppl 1): I 17-30S.

11 Churchill DN, Torrance GW, Taylor DX', et al. ieasurement of quality of'lifcin end-stage renal disease: the time trade-off approach. Clin Invest Med1987;10: 14-20.

12 Naughton J, Sevelius G, Balke B. Physiological responses otf' normal andpathological subjects to a modified work capacity test. J Sports Med P/s'sFitness 1963;3:201-7.

13 Schwartz D, Flamant R, Lellouch J. Clinical trio/ls. New York: AcademicPress, 1980:137.

14 Winer BS. Statistitcal prnciples in experimental design. New York: MSc(Graw-Hill, 1971:518-39.

15 Gibbons J[). NnparantietrIc statistical inference. Ncw Ytork: MIcGraw-Hill,1971:2209-26.

16 Mohiide EA, Torrance GW, Streitter D)t, ltrittgle l)Ni, Gilbert R. MNeasuringthe wcllbeing of family caregis-crs usitag thc timc trade-otff technique.Cl/inical Epidemiots/s 1988;41:475-82.

17 Jacschkc R, Singer J, Guyatt G, Health status mcasuremcnt: ascertaining theminimal clinically important diff'ercncc. CAontrolled C/in Irials 1989;10:407-15.

(Accepte7d 19 IJecenther /989

Short term increase in risk ofbreast cancer associated withfull term pregnancy

E M I Williams, L Jones, M P Vessey,K McPherson

Women who have their first full term pregnancy afterage 35 are at higher risk of breast cancer thannulliparous women,' and the proportion of youngwomen with breast cancer who have had children ishigher than expected.7 These and other findingssuggest that women who have full term pregnancieshave a transiently increased risk of breast cancer that isfollowed by long term protection.) We applied theanalysis described by Bruzzi et al3 to data obtainedfrom an investigation which explored the relationbetween lifestyle factors and risk of breast cancer.4

Patients, methods, and resultsBriefly, we recruited 1996 married women aged 25-

59 between 1980 and 1984 from seven hospitals. Theycomprised 998 patients with newly diagnosed breastcancer that had been confirmed histologically and 998controls, who had been admitted electively withconditions not originally related to breast cancer.

Patients and controls were matched for admittinghospital and within five year age groups but not forparity. The present analysis aimed at detecting anyincrease in the risk of breast cancer shortly after a fullterm pregnancy. The interval between the date ofdiagnosis of breast cancer and the last term birth wasstudied. As this is related to age, age at first term birth,and parity these variables were adjusted for in theanalysis.Only women under 50 with two or more children

were included to accord with the analysis by Bruzzi etal,' and this resulted in the study becoming unmatched(422 cases and 447 controls). The generalised inter-active modelling (GLIM) package was used to estimatethe maximum likelihood of effects. Graduated levels ofexposure were assessed by linear trend tests.The two groups were comparable in terms of centre

of recruitment, which was ignored in subsequentanalyses. An increased risk of breast cancer wasassociated with decreasing interval since last term birth(p=0-021), increasing age at first term birth (p=0 006),and decreasing parity (p=0-002) (table). When womenaged under 40 and 40-49 were considered separatelythe trends were broadly similar, although not all weresignificant.

CommentOur results suggest that a transient increase in the

risk of breast cancer occurs after full term pregnancy.

Department of CommunityMedicine and GeneralPractice, RadcliffeInfirmary, OxfordOX2 6HEE M I Williams, MB, seniorregistrarL Jones, BA, computer scientistM P Vessey, FFCM, professorK McPherson, PHD, lecturer

Correspondence to: DrE M I Williams, OxfordRegional Health Authority,Old Road, Headington,Oxford OX3 7LF.

Br,1edj 1990;300:578-9

578 BMJ VOLUME 300 3 MARCH 1990

on 5 January 2021 by guest. Protected by copyright.

http://ww

w.bm

j.com/

BM

J: first published as 10.1136/bmj.300.6724.573 on 3 M

arch 1990. Dow

nloaded from