EXTENDED REPORT

Risk of mortality in patients with psoriatic arthritis,rheumatoid arthritis and psoriasis: a longitudinalcohort studyAlexis Ogdie,1 Kevin Haynes,2 Andrea B Troxel,2 Thorvardur Jon Love,3

Sean Hennessy,2 Hyon Choi,4 Joel M Gelfand5

Handling editor Tore K Kvien

▸ Additional material ispublished online only. To viewplease visit the journal online(http://dx.doi.org/10.1136/annrheumdis-2012-202424).

For numbered affiliations seeend of article

Correspondence toDr Alexis Ogdie, Division ofRheumatology, Center forClinical Epidemiology andBiostatistics, Center forPharmacoepidemiologyResearch and Training,Perelman School of Medicineat the University ofPennsylvania, 8 Penn Tower,1 Convention Ave,Philadelphia, PA 19104, USA;alexis.ogdie@uphs.upenn.edu

Received 28 July 2012Revised 12 November 2012Accepted 2 December 2012Published Online First21 December 2012

To cite: Ogdie A, Haynes K,Troxel AB, et al. Ann RheumDis 2014;73:149–153.

ABSTRACTBackground There are conflicting reports in theliterature of the mortality risk among patients withpsoriatic arthritis (PsA). The objective of this study wasto examine the risk of mortality in patients with PsAcompared with matched controls, patients with psoriasisand those with rheumatoid arthritis (RA).Methods A longitudinal cohort study was performed ina large UK medical record database, The HealthImprovement Network, among patients with PsA,rheumatoid arthritis (RA) or psoriasis with data from1994 to 2010. Unexposed controls were matched onpractice and start date within the practice for eachpatient with PsA. Cox proportional hazards models wereused to calculate the relative hazards for death.Results Patients with PsA (N=8706), RA (N=41 752),psoriasis (N=138 424) and unexposed controls(N=82 258) were identified; 1 442 357 person-yearswere observed during which 21 825 deaths occurred.Compared with population controls, patients with PsAdid not have an increased risk of mortality after adjustingfor age and sex (disease-modifying antirheumatic drug(DMARD) users: HR 0.94, 95% CI 0.80 to 1.10;DMARD non-users: HR 1.06, 95% CI 0.94 to 1.19)whereas patients with RA had increased mortality(DMARD users: HR 1.59, 95% CI 1.52 to 1.66; DMARDnon-users: HR 1.54, 95% CI 1.47 to 1.60). Patientswith psoriasis who had not been prescribed a DMARDhad a small increased risk of mortality (HR 1.08, 95% CI1.04 to 1.12) while those who had been prescribed aDMARD, indicating severe psoriasis, were at increasedrisk (HR 1.75, 95% CI 1.56 to 1.95).Conclusions Patients with RA and psoriasis haveincreased mortality compared with the generalpopulation but patients with PsA do not have asignificantly increased risk of mortality.

INTRODUCTIONPsoriatic arthritis (PsA) is a chronic progressiveinflammatory disease that affects over 500 000Americans and can cause permanent joint damageand severe disability.1 However, it is unclearwhether PsA confers an increased risk of mortalityas, to date, observational studies have yielded con-flicting evidence.2 Such studies have been limitedby small sample size with few events and the poten-tial for selection bias in clinic-based studies.Furthermore, very little is known about the risk ofmortality in PsA from a population-basedperspective.

Given that previous studies have demonstratedincreased mortality among patients with psoriasisand rheumatoid arthritis (RA),3 4 we hypothesisedthat patients with PsA would similarly haveincreased rates of mortality compared with thegeneral population. The objective of this study wasto measure the risk of all-cause mortality amongpatients with PsA compared with the general popu-lation and with patients with psoriasis only(without a diagnosis of PsA) and RA.

METHODSStudy design and data sourceA cohort study was undertaken using data fromThe Health Improvement Network (THIN) in theUK between 1994 and 2010. The UK is an idealhealth system for capturing medical record data asthe general practitioner (GP) is the primary contactfor all aspects of the patient’s care.5 In the UK,95% of patients are registered with a GP.6

Participating GPs record data as part of routinepatient care (eg, demographics, medical diagnoses,laboratory data and prescriptions) in the electronicmedical record including recommendations madeby specialists in secondary and tertiary care.(However, some treatments may initially be pre-scribed by a specialist and often the GP will there-after assume prescribing the medication. Tumournecrosis factor α prescriptions are not generallyrecorded in THIN as these are exclusively pre-scribed by the specialist.) The data are anonymisedand collected by THIN, assessed for quality andmade available for research use.7 THIN drawspatients from 514 general practices and is represen-tative of the UK population in terms of age, sex,geography and medical conditions.8–11

Study populationAll patients with PsA, psoriasis or RA between theages of 18 and 89 at the start date were included ifthey had observation time in THIN after Visionsoftware implementation. Patients were excluded ifthey died or transferred out of the practice prior tothe implementation of Vision software. Up to 10unexposed controls from the general populationwere randomly selected for each patient with PsAand were matched on practice and start date withinthe practice. Unexposed controls were assigned a‘diagnosis’ date within 6 months of the diagnosisdate of the patient with PsA. This algorithm wasdesigned to minimise bias by ensuring that patients

Ogdie A, et al. Ann Rheum Dis 2014;73:149–153. doi:10.1136/annrheumdis-2012-202424 149

Clinical and epidemiological research

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from

with PsA and unexposed controls are followed by similardoctors during similar time periods. Patients were not eligible toserve as unexposed controls if they had PsA, psoriasis or RA orwere using disease-modifying antirheumatic drugs (DMARDs).

Person-time calculationCohort entry was defined as the latest of diagnosis, 6 monthsafter initial registration with the practice, DMARD initiation,implementation of Vision software in the patient’s practice or apractice acceptable mortality reporting (the first three arepatient-level factors and last two are practice-level factors).7 12

The rationale for choosing these elements is further described inthe online supplementary methods section. Censoring occurredwhen the patient died, left the practice, the practice stoppedparticipating in THIN or the study ended in September 2010.

Exposure definitionsPsA, psoriasis and RA were defined by the presence of at leastone READ code consistent with these diseases. READ codes area comprehensive hierarchical alphanumerical clinical languagedeveloped in the UK to record diagnoses, symptoms and tests,similar to International Classification of Diseases codes.13

READ codes for psoriasis and RA have been previously vali-dated within the same or analogous large medical record data-bases.14–16 READ codes for PsA have a positive predictive valueof 85% as determined by a survey of 100 randomly selectedGPs caring for patients with PsA.17 We have also used this defin-ition of PsA in other studies.18

Outcome definitionThe primary outcome, death, was defined by specific codesnoting death and/or codes indicating the patient was transferredout of the practice because of his/her death.19 An algorithmrecommended by Cegedim, the administrators of THIN, wasused and identifies death codes from within the patient, medicaland administration files. This algorithm has been used in otherstudies and has a sensitivity of 99.7%.19

Covariates of interestAll covariates of interest were measured prior to cohort entry.The following potential confounders were measured: Charlsoncomorbidity score,20 smoking, body mass index, blood pressureat baseline, depression, prior hospitalisation in the baselinefollow-up period, year of cohort entry, socioeconomic status(via Townsend deprivation score), urban versus rural livingenvironment, chronic kidney disease, heart disease, atrial fibril-lation, diabetes, hypertension, history of cancer, asthma, chronicobstructive pulmonary disease and liver disease. Furthermore, apriori we hypothesised a statistical interaction between diseasestatus and DMARD use. DMARDs included methotrexate,sulfasalazine, azathioprine, leflunamide, cyclosporine, mycophe-nolate, hydroxychloroquine and biologic disease-modifyingagents including adalimumab, etanercept and infliximab. In theUK, these medications can be prescribed by consultants (specia-lists) but should be captured by GP records with the exceptionof the biologic medications which are rarely recorded.17

Statistical analysisPower calculations prior to the start of the study revealed that,with 7000 patients with PsA and 35 000 unexposed patients, wewould have 96% power to detect a HR as small as 1.05 forpatients with PsA with an average of 5 years of follow-up perpatient. Descriptive statistics were used to examine age, sex,person-time and covariate distribution between the four groups.

We fit a Cox proportional hazards regression model, adjustingfor age and sex, to determine the overall HR for each groupcompared with the unexposed group. We then tested thehypothesised statistical interaction and fit models with thehypothesised confounders using a purposeful selection model-ling approach.21 More detail with regard to the modellingapproach can be found in the online supplementary methods.Log-log survival plots were generated to assess the assumptionof proportionality. All statistical analysis was performed usingSTATA 12.0 (College Station, Texas, USA).

Sensitivity analysesThe following sensitivity analyses were conducted: (1) multipleimputation was performed for body mass index and smokingstatus where missing and models containing these variables wereretested; (2) DMARDs were included as a time-varying covariaterather than a stratification variable; (3) we restricted patients toonly those followed for at least 1 year prior to the start date toensure capture of comorbidities; (4) to examine whether missedDMARD prescriptions would have an impact on the results, weimputed additional DMARD users by first creating a propensityscore for DMARD use and then assigning 25% of thenon-DMARD users with the highest propensity scores toDMARD use; (5) we restricted the cohort to only those withincident disease defined as patients with at least 1 year offollow-up prior to the first diagnosis code; and (6) we con-ducted an unmeasured confounder analysis to test the assump-tion that there is a confounder that we are unable to measurethat may skew the results had we been able to measure this con-founder. More detail with regard to the methods used in thesensitivity analyses can be found in the online supplementarymethods.

RESULTSBetween 1994 and 2010, 8706 patients with PsA, 41 752 withRA and 138 424 patients with psoriasis met the inclusion cri-teria and 82 258 unexposed patients were randomly selectedusing the described criteria. Baseline characteristics are shown intable 1. Patients with RA were significantly older and were pre-dominantly female, while the male: female ratio was closer to1:1 in the other three groups. Mean person-time contributedwas 5.3 years. Median year of cohort entry was 1 year earlierfor patients with RA compared with the other three groups.DMARDs were prescribed to 48%, 52% and 3% of patientswith PsA, RA and psoriasis, respectively. The prevalence ofcomorbidities was highest among patients with RA in nearly allcategories.

Among the 271 140 patients, 21 825 deaths were observedover 1 442 238 person-years (table 2). The unadjusted incidenceof mortality was highest in patients with RA (table 3). The inter-action between disease status and history of DMARD use wasstatistically significant and therefore results stratified byDMARD use are presented. The addition of hypothesised con-founders to the model did not change the model substantially(all of the covariates listed in the Methods section were testedbut individual p values and HRs are not provided due to spacerestrictions). Therefore, the final model is the age- and sex-adjusted model demonstrating increased mortality risk amongpatients with RA (DMARD users: HR 1.59, 95% CI 1.52 to1.66; DMARD non-users: HR 1.54, 95% CI 1.47 to 1.60) andpsoriasis (DMARD users: HR 1.75, 95% CI 1.56 to 1.95;DMARD non-users: HR 1.08, 95% CI 1.04 to 1.12) but noincreased mortality risk among patients with PsA (DMARD

150 Ogdie A, et al. Ann Rheum Dis 2014;73:149–153. doi:10.1136/annrheumdis-2012-202424

Clinical and epidemiological research

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from

users: HR 0.94, 95% CI 0.80 to 1.10; DMARD non-users: HR1.06, 95% CI 0.94 to 1.19).

Reasons for censoring (patient leaving the practice, practicestopped contributing to THIN and end of study in September2010) were not significantly different among the groups (seeonline supplementary table 1). The five sensitivity analyses con-ducted did not change the results of the final model (see onlinesupplementary table 2). An unmeasured confounder analysisshowed that even if the unmeasured confounder has a HR of10.0 and a prevalence of 60% among patients with PsA, themortality risk for patients with PsA would not significantlychange (see online supplementary table 3).

DISCUSSIONIn this large population-based study we found that patients withPsA did not have a statistically significant increase in mortalitycompared with the general population. Furthermore, patientswith RA and those with severe psoriasis (DMARD users) hadsignificantly higher mortality than the general population. Theincreased all-cause mortality risk in patients with psoriasis not

using DMARDs (8% increase over the general population) issmall and statistically similar to the PsA groups based on the95% CIs. These results are consistent with the findings of otherpopulation-based studies3 4 22–29 and were robust to multiplesensitivity analyses. This is the first study to compare directlythe risk of all-cause mortality in PsA with an unexposed popula-tion rather than using standardised mortality ratios (SMR) tocompare with census statistics. Internal controls are generallyfelt to provide a better approximation of the true effect thanSMRs.30

Most previous studies of mortality in PsA have been per-formed within specialty clinics. These studies have had mixedresults ranging from no difference in mortality compared withlocal census statistics to an SMR of 1.62 (95% CI 1.21 to2.12).31 (Of note, the SMR was 1.36 (95% CI 1.12 to 1.64) inthe same cohort 10 years later32). Clinic- and hospital-basedstudies demonstrating higher mortality could be capturing alarger proportion of patients with severe disease reflecting selec-tion bias. In the general population a larger proportion ofpatients are likely to have mild disease. It is possible that severityof disease is a driver of mortality risk. We were unable to testthis hypothesis in THIN, however a sensitivity analysis demon-strated that the effect size of this unmeasured confounder mustbe substantial (HR=10) in order to change the results. Finally, itis possible that mortality has declined over time in the PsApopulation.32 However, adjusting the regression model for startyear in the cohort did not change the results of the final model.

Population-based studies using large medical record databaseshave tremendous advantages in examining mortality.Consideration must be given to the choice of the unexposed orcontrol population in such studies as control patients may nothave regular contact with their physicians and new diagnoses orevents may not be captured. It was for this reason that we choseto include only patients who had contact with their GP aroundthe time the matched patient with PsA was diagnosed. Selectingpatients in this way may increase the ‘illness level’ of patients inthe unexposed group and reduce the HR for mortality inpatients in the exposed groups. It is therefore possible that thereis a minimally increased risk of mortality in patients with PsA.On the other hand, we included patients with prevalent diagno-ses in this study. Studies including patients with prevalent diag-noses are generally thought to have higher death rates thanthose with incident disease.33 However, a sensitivity analysis

Table 1 Demographic data of study subjects

Control PsA RA Psoriasis(N=82 258) (N=8706) (N=41 752) (N=138 424)

Male sex (%) 45% 51% 29% 49%Mean age (years) 49.9 50.7 61.4 47.6Cohort time (years) 5.25 5.21 5.29 5.38Start year (median) 2003 2003 2002 2003Townsend (median)* 2.57 2.53 2.68 2.66DMARDs, N (%) 0 4213 (48%) 21910 (52%) 3577 (3%)Hospitalised inbaseline period

11558 (14%) 1383 (16%) 8094 (19%) 16861 (12%)

Disabled 2106 (2.6%) 447 (5.1%) 3053 (7.3%) 3611 (2.6%)Smoking, N (%)Never 39529 (55%) 4117 (52%) 18854 (51%) 54437 (45%)Past 13823 (19%) 1878 (24%) 8865 (24%) 26375 (22%)Current 18025 (25%) 1857 (24%) 9193 (25%) 40820 (34%)Missing 10881 (13%) 854 (10%) 4840 (12%) 16792 (12%)

BMI, N (%)Normal 21639 (43%) 2274 (34%) 13391 (40%) 44819 (41%)Underweight 1472 (3%) 83 (1%) 1008 (3%) 2879 (3%)Overweight 16854 (34%) 2439 (36%) 11442 (34%) 37219 (34%)Obese 9924 (20%) 1993 (29%) 7433 (22%) 23409 (22%)Missing 32369 (39%) 1917 (22%) 8478 (20%) 30098 (22%)

Comorbidities, N (%)CharlsonComorbidity Index(mean)

0.44 0.41 0.65 0.39

Diabetes 4619 (6%) 608 (7%) 3307 (8%) 7087 (5%)Heart disease 6169 (7%) 576 (7%) 5767 (14%) 9761 (7%)Hypertension 15349 (19%) 1901 (22%) 11895 (28%) 22899 (17%)Dyslipidaemia 6919 (8%) 849 (10%) 4432 (11%) 10879 (8%)Stroke 2586 (3%) 219 (3%) 2372 (6%) 3987 (3%)Kidney disease 1730 (2%) 196 (2%) 1608 (4%) 2275 (2%)Cancer 6982 (8%) 601 (7%) 4465 (11%) 10006 (7%)Liver disease 1039 (1%) 208 (2%) 750 (2%) 2075 (1%)Depression 14447 (18%) 1908 (22%) 8874 (21%) 26180 (19%)Substance abuse 2583 (3%) 355 (4%) 1090 (3%) 6216 (4%)

*Socioeconomic status was assessed by Townsend deprivation scores using quintiles1–5, with higher quintiles representing more deprivation.BMI, body mass index; DMARD, disease modifying anti-rheumatic drug; PsA, psoriaticarthritis; RA, rheumatoid arthritis.

Table 2 Incidence of death

Deaths Person-years (PY) Incidence (deaths/1000 PY)Control 5330 431730 12.35

PsAAll 470 45334 10.37DMARD 159 20377 7.80No DMARD 311 24958 12.46

RAAll 7004 220855 31.71DMARD 2903 113859 24.50No DMARD 4101 106996 38.33

PsoriasisAll 9021 744318 12.12DMARD 328 14781 22.19No DMARD 8693 729537 11.92

DMARD, disease-modifying antirheumatic drug; PsA, psoriatic arthritis; RA,rheumatoid arthritis.

Ogdie A, et al. Ann Rheum Dis 2014;73:149–153. doi:10.1136/annrheumdis-2012-202424 151

Clinical and epidemiological research

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from

including only cases with at least 1 year prior to the first diagno-sis code (a commonly used definition of incident disease) didnot substantially change the results.

Potential limitations of this study include misclassification ofdiagnoses and missing information regarding DMARD use. Ourprevious validation study showed limited misclassification ofpatients with a PsA diagnosis (ie, a high positive predictive valuefor the diagnostic code).17 In fact, there was an even higher posi-tive predictive value (93%) for patients with a diagnostic codefor PsA patients with a DMARD prescription, further decreasingthe likelihood that misclassification could be the reason for thenull result (unpublished data). While we are confident in thediagnostic code for PsA, there may be patients with PsA cate-gorised as psoriasis only (eg, undiagnosed cases of PsA amongpatients with psoriasis).34 It is not possible to identify the fre-quency of this phenomenon in this population-based study.Population-based studies such as those performed in THIN oftenlack information on disease activity measures and disease charac-teristics. However, the goal of this study was to examine patientswith already diagnosed PsA, a more specific cohort than allpatients with possible PsA. Misclassification of the outcome isnot a major concern as mortality is unlikely to be subject to sur-veillance bias because it has near complete ascertainment.19 Thispopulation-based study draws from a larger cohort, THIN,which is felt to be representative of the UK but may over-represent more affluent areas.11 However, adjusting for socio-economic status (via Townsend deprivation score) did not changethe results. Finally, use of DMARDs is difficult to model in anobservational study and can be performed in numerous ways,though all are limited by confounding by indication.35 36 Wechose to classify patients as ‘ever’ or ‘never’ users because wewere using DMARDs as a stratification variable, and we hypothe-sised that patients prescribed treatment may be different fromthose who had not been prescribed treatment. There may beunder-reporting of DMARD use in cases where the consultant isthe primary prescriber, despite the fact that this may be recordedby the GP in the medical record, and this could result in mis-classification. However, increasing the number of DMARD usersby 25% did not significantly change the results.

In conclusion, we present the results of a large population-based study demonstrating increased mortality in patients with

severe psoriasis or RA but no statistically significant increase inmortality among patients with PsA. The lack of higher mortalityin patients with PsA has been previously demonstrated insmaller population-based cohorts. Despite a lack of higher mor-tality, there is still significant morbidity in patients with PsAincluding concomitant illnesses and impaired quality of life.Future research efforts are needed to identify mechanisms bywhich we can improve comorbidities and quality of life forpatients with PsA.

Author affiliations1Division of Rheumatology, Center for Clinical Epidemiology and Biostatistics, Centerfor Pharmacoepidemiology Research and Training, Perelman School of Medicine atthe University of Pennsylvania, Philadelphia, Pennsylvania, USA2Department of Biostatistics and Epidemiology, Center for Clinical Epidemiology andBiostatistics, Center for Pharmacoepidemiology Research and Training, PerelmanSchool of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania,USA3Division of Rheumatology, Landspitali University Hospital, Reykjavik, Iceland4Section of Rheumatology and the Clinical Epidemiology Unit, Boston UniversitySchool of Medicine, Boston, Massachusetts, USA5Department of Dermatology, Center for Clinical Epidemiology and Biostatistics,Center for Pharmacoepidemiology Research and Training, Perelman School ofMedicine at the University of Pennsylvania, Philadelphia, Pennsylvania, USA

Acknowledgements We thank Dr Peter Merkel for helpful discussions regardingthis paper.

Contributors AO and JMG conceptualised and designed the study with input fromKH, SH, ABT, TJL and HC, and these authors were integral in interpretation of theresults. AO performed the programming, statistical analysis, preparation of the dataand the first draft of the manuscript. KH performed data abstraction from The HealthImprovement Network and assisted in programming. All authors were involved incritical review of the data as well as drafting and revision of the manuscript, and allhave approved the final version of the paper to be published.

Funding This project was funded by the American College of Rheumatology. Datafrom The Health Improvement Network is supported by the Clinical and TranslationalScience Award at the University of Pennsylvania (8UL1TR000003from the NationalCenter for Research Resources). AO was supported by NIH T32 GM075766-05 andthe American College of Rheumatology Research and Education Foundation; SH wassupported by R01AG025152; JMG was supported by R01HL111293; and TJL wassupported by The Icelandic Research Fund #120433021.

Competing interests JMG serves as a consultant to Amgen, Abbott, Centocor,Celgene, Novartis and Pfizer and has received honoraria. He has received grants fromAmgen, Abbott, Pfizer, Novartis and Genentech. The remaining authors have nocompeting interests. Cegedim Strategic Data (CSD) Medical Research UK is an expert inUK anonymous patient data for the healthcare industry. CSD is a commercial

Table 3 Cox proportional hazards models

HR (unadjusted) HR (age- and sex-adjusted) HR (unadjusted stratified)* HR (age- and sex-adjusted, stratified)*Control Ref Ref Ref Ref

PsAAll 0.84 (0.77 to 0.92) 1.02 (0.92 to 1.12)No DMARD 1.00 (0.90 to 1.13) 1.06 (0.95 to 1.19)DMARD 0.64 (0.55 to 0.75) 0.94 (0.80 to 1.10)

RAAll 2.57 (2.48 to 2.67) 1.56 (1.51 to 1.62)No DMARD 3.09 (2.97 to 3.22) 1.54 (1.48 to 1.61)DMARD 2.08 (1.99 to 2.18) 1.59 (1.52 to 1.66)

PsoriasisAll 0.98 (0.95 to 1.01) 1.10 (1.06 to 1.13)No DMARD 0.96 (0.93 to 1.00) 1.08 (1.04 to 1.12)DMARD 1.85 (1.65 to 2.07) 1.75 (1.56 to 1.95)

Age 1.10 (1.10 to 1.10) 1.10 (1.10 to 1.10)Sex 0.71 (0.69 to 0.73) 0.71 (0.69 to 0.73)

*Interaction p<0.001 (likelihood ratio/Wald test).DMARD, disease-modifying antirheumatic drug; PsA, psoriatic arthritis; RA, rheumatoid arthritis.

152 Ogdie A, et al. Ann Rheum Dis 2014;73:149–153. doi:10.1136/annrheumdis-2012-202424

Clinical and epidemiological research

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from

organisation that supplies data and trains and supports researchers in the use of primarycare patient data. Data are available for use in medical research in the academic settingas well as in industry for a fee which varies depending on the type of data requested.Aside from undergoing ethical review by the Scientific Review Committee at Cegedim,independent academic groups who voluntarily act as an ethical review body, thisprotocol was not in any way discussed with Cegedim nor were any changes made bythe company. No financial support or other forms of computational or analytical supportwere received from Cegedim/THIN. The data were collected by Cegedim and the generalpractitioners without knowledge of the study objectives and hypotheses.

Ethical approval This study was approved by the University of PennsylvaniaInstitutional Review Board and Cegedim’s Scientific Review Committee. All data inthe study were anonymous to the investigators.

Provenance and peer review Not commissioned; externally peer reviewed.

Data sharing statement Data from this study will be shared upon request. Pleaseemail the corresponding author for such inquires.

REFERENCES1 Gladman D, Antoni C, Mease P, et al. Psoriatic arthritis: epidemiology, clinical

features, course, and outcome. Ann Rheum Dis 2005;64(Suppl II):ii14–17.2 Gladman DD. Mortality in psoriatic arthritis. Clin Exp Rheumatol 2008;26(Suppl

51):S62–5.3 Gelfand JM, Troxel AB, Lewis JD, et al. Risk of mortality in patients with psoriasis:

results from a population-based study. Arch Dermatol 2007;1493–9.4 Sokka T, Abelson B, Pincus T. Mortality in rheumatoid arthritis: 2008 Update. Clin

Exp Rheumatol 2008;26(S51):S35–61.5 deLusignan S, Stephens P, Naeema N, et al. Does feedback improve the quality of

computerized medical records in primary care? J Am Med Inform Assoc2002;9:395–403.

6 Ogdie A, Langan S, Parkinson J, et al. Medical record databases. In: Strom B,Kimmel S, Hennessy S. eds. Pharmacoepidemiology. 5th edn. Oxford:Wiley-Blackwell, 2012:224–43.

7 Cegedim Strategic Data. THIN Data Quality Assurance. http://csdmruk.cegedim.com/our-data/data-quality.shtml (accessed 8 Nov 2012).

8 Cegedim Strategic Data. The Health Improvement Network: Our Data. http://csdmruk.cegedim.com/our-data/our-data.shtml (accessed 8 Nov 2012).

9 Lewis JD, Schinnar R, Bilker WB, et al. Validation studies of the health improvementnetwork (THIN) database for pharmacoepidemiology research. PharmacoepidemiolDrug Saf 2007;16:393–401.

10 Haynes K, Bilker WB, Tenhave TR, et al. Temporal and within practice variability inthe health improvement network. Pharmacoepidemiol Drug Saf 2011;20:948–55.

11 Blak B, Thompson M, Dattani H, et al. Generalisability of The Health ImprovementNetwork (THIN) database: demographics, chronic disease prevalence and mortalityrates. Inform Prim Care 2011;19:251–5.

12 Maguire A, Blak B, Thompson M. The importance of defining periods of completemortality reporting for research using automated data from primary care.Pharmacoepidemiol Drug Saf 2009;18:76–83.

13 Chishom J. The READ clinical classification. BMJ 1990;300:1092.14 Watson DJ, Rhodes T, Bing C, et al. Lower risk of thromboembolic cardiovascular

events with Naproxen among patients with rheumatoid arthritis. Arch Intern Med2002;162:1105–10.

15 Rodriguez LA, Tolosa LB, Ruigómez A, et al. Rheumatoid arthritis in UK primarycare: incidence and prior morbidity. Scand J Rheumatol 2009;38:173–7.

16 Seminara NM, Abuabara K, Shin DB, et al. Validity of The Health ImprovementNetwork (THIN) for the study of psoriasis. Br J Dermatol 2011;164:602–9.

17 Ogdie A, Langan S, Love T, et al. Prevalence and treatment patterns of psoriaticarthritis in the United Kingdom. Rheumatology Published Onlone First: 7 Dec 2012.doi: 10.1093/rheumatology/kes324

18 Love T, Zhu Y, Zhang Y, et al. Obesity and the risk of psoriatic arthritis: apopulation-based study. Ann Rheum Dis 2012;71:1273–7.

19 Hall GC. Validation of death and suicide recording on THIN UK primary caredatabase. Pharmacoepidemiol Drug Saf 2009;18:120–31.

20 Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognosticcomorbidity in longitudinal studies: development and validation. J Chron Dis1987;40:373–83.

21 Bursac Z, Gauss CH, Williams DK, et al. Purposeful selection of variables in logisticregression. Source Code Biol Med 2008;3:17.

22 Shbeeb M, Uramoto KM, Gibson LE, et al. The epidemiology of psoriatic arthritis inOlmsted County, Minnesota, USA 1982–1991. J Rheumatol 2000;27:1247–50.

23 Watson DJ, Rhodes T, Guess HA. All-cause mortality and vascular events amongpatients with rheumatoid arthritis, osteoarthritis, or no arthritis in the UK GeneralPractice Research Database. J Rheumatol 2003;30:1196–202.

24 Bjornadal L, Baecklund E, Yin L, et al. Decreasing mortality in patients withrheumatoid arthritis: results from a large population based cohort in Sweden,1964–95. J Rheumatol 2002;29:906–12.

25 Doran MF, Pond GR, Crowson CS, et al. Trends in incidence and mortality inrheumatoid arthritis in Rochester, Minnesota, over a forty-year period. ArthritisRheum 2002;46:625–31.

26 Myllykangas-Luosujarvi R, Aho K, Kautianien H, et al. Shortening of life span andcauses of excess mortality in a population-based series of subjects with rheumatoidarthritis. Clin Exp Rheumatol 1995;13:149–53.

27 Mikuls TR, Saag KG, Criswell LA, et al. Mortality risk associated with rheumatoidarthritis in a prospective cohort of older women: results from the Iowa Women’sHealth Study. Ann Rheum Dis 2002;61:994–9.

28 Thomas E, Symmons DP, Brewster DH, et al. National study of cause-specificmortality in rheumatoid arthritis, juvenile chronic arthritis, and other rheumaticconditions: a 20 year follow up study. J Rheumatol 2003;30:958–65.

29 Sihvonen S, Korpela M, Mustonen J, et al. Renal disease as a predictor of increasedmortality among patients with rheumatoid arthritis. Nephron Clin Pract 2004;96:c107–14.

30 Card TR, Solaymani-Dodaran M, Hubbard R, et al. Is an internal comparison betterthan using national data when estimating mortality in longitudinal studies? JEpidemiol Community Health 2006;60:819–21.

31 Wong K, Gladman DD, Husted J, et al. Mortality studies in psoriatic arthritis: resultsfrom a single outpatient clinic; causes and risk of death. Arthritis Rheum1997;40:1868–72.

32 Ali Y, Tom BDM, Schentag CT, et al. Improved survival in psoriatic arthritis withcalendar time. Arthritis Rheum 2007;56:2708–14.

33 Gordis L. Measuring the occurrence of disease: mortality. Epidemiology. 4th edn.Philadelphia, PA: Saunders Elsevier, 2004:59–84.

34 Ibrahim G, Waxman R, Helliwell PS. The prevalence of psoriatic arthritis amongpatients with psoriasis. Arthritis Care Res 2009;61:1373–8.

35 Strom B, Melmon KL. The use of pharmacoepidemiology to study beneficial drugeffects. In: Strom B, Kimmel S, Hennessy S. eds. Pharmacoepidemiology. 5th edn.Oxford: Wiley-Blackwell, 2012:655–77.

36 Dixon WG, Bansback N. Understanding the side effects of glucocorticoid therapy:shining a light on a drug everyone thinks they know. Ann Rheum Dis2012;71:1761–4.

Ogdie A, et al. Ann Rheum Dis 2014;73:149–153. doi:10.1136/annrheumdis-2012-202424 153

Clinical and epidemiological research

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from

doi: 10.1136/annrheumdis-2012-20242421, 2012

2014 73: 149-153 originally published online DecemberAnn Rheum Dis Alexis Ogdie, Kevin Haynes, Andrea B Troxel, et al. a longitudinal cohort studyarthritis, rheumatoid arthritis and psoriasis: Risk of mortality in patients with psoriatic

http://ard.bmj.com/content/73/1/149.full.htmlUpdated information and services can be found at:

These include:

Data Supplement http://ard.bmj.com/content/suppl/2012/12/21/annrheumdis-2012-202424.DC1.html

"Supplementary Data"

References http://ard.bmj.com/content/73/1/149.full.html#ref-list-1

This article cites 30 articles, 10 of which can be accessed free at:

serviceEmail alerting

the box at the top right corner of the online article.Receive free email alerts when new articles cite this article. Sign up in

CollectionsTopic

(2785 articles)Rheumatoid arthritis   � (4252 articles)Musculoskeletal syndromes   �

(4340 articles)Immunology (including allergy)   � (3977 articles)Degenerative joint disease   � (3650 articles)Connective tissue disease   �

(582 articles)Drugs: musculoskeletal and joint diseases   � (446 articles)Biological agents   �

(1189 articles)Epidemiology   � Articles on similar topics can be found in the following collections

Notes

http://group.bmj.com/group/rights-licensing/permissionsTo request permissions go to:

http://journals.bmj.com/cgi/reprintformTo order reprints go to:

http://group.bmj.com/subscribe/To subscribe to BMJ go to:

group.bmj.com on September 26, 2014 - Published by ard.bmj.comDownloaded from