See related commentary on pg 1007

ORIGINAL ARTICLE

Very Severe Psoriasis Is Associated with Increased Noncardiovascular Mortality but Not with Increased Cardiovascular Risk
Robert S. Stern1,2 and Annemieke Huibregtse1,3
It has been hypothesized that severe psoriasis is an independent risk factor for cardiovascular disease (CVD). We prospectively studied patients with severe psoriasis treated with psoralens and ultraviolet-A therapy (PUVA) who enrolled in a cohort study in 19751976. From 1977 to 2005, 617 of the 1,376 patients (45%) died. Compared with the general population, cohort death rates were significantly higher than expected (standard mortality ratio (SMR) 1.1, 95% confidence interval (CI) 1.021.20). The number of deaths due to CVD (SMR 1.02, 95% CI 0.91.6) was nearly identical to the expected number. Deaths due to liver disease were significantly elevated (SMR 4.04, 95% CI 2.765.70). Patients with exceptionally severe psoriasis at entry (442% body surface area (BSA)) had a significantly increased risk of death compared with less severely affected cohort members (all-cause hazard ratio (HR) 1.42, 95% CI 1.181.69) as well as for deaths because of causes other than cancer or CVD (multivariate HR 1.56, 95% CI 1.142.13). Only patients with exceptionally severe psoriasis had an increased mortality risk compared with both the general population and other cohort members with less extensive but still severe psoriasis. These increases were not significant for CVD. Our data do not support the hypothesis that severe psoriasis is an independent risk factor for CVD. However, exceptionally severe psoriasis is associated with increased all-cause mortality.
Journal of Investigative Dermatology (2011) 131, 11591166; doi:10.1038/jid.2010.399; published online 20 January 2011

INTRODUCTION Psoriasis is a common, chronic skin disease affecting B2% of white adults. Less than 5% of those affected report as much as 10% of body surface area (BSA) affected (Stern et al., 2004; Kurd and Gelfand, 2008). Studies in specialized populations have found an association of psoriasis, especially severe psoriasis, with an increased risk of cardiovascular disease (CVD) (Gelfand et al., 2006, 2007; Ludwig et al., 2007; Shapiro et al., 2007; Gerdes et al., 2008; Kimball et al., 2008a, b). These findings are largely based on studies of specialized populations or administrative databases with severity defined according to therapies utilized rather than an objective severity metric.
1

Department of Dermatology, Beth Israel Deaconess Medical Center, Boston, Massachusetts, USA; 2Department of Dermatology, Harvard Medical School, Boston, Massachusetts, USA and 3Department of Dermatology, Leiden University Medical Center, Leiden, The Netherlands Correspondence: Robert S. Stern or Annemieke Huibregtse, Department of Dermatology, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, GZ-522, Boston, Massachusetts 02215, USA. E-mail: rstern@bidmc.harvard.edu (Robert S. Stern) or ahuibregste@hotmail.com (Annemieke Huibregtse) Abbreviations: BSA, body surface area; CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio; ICD-9, International Classification of Disease, 9th Edition; NDI, National Death Index; NHANES-1, First National Health and Nutrition Examination Survey; PUVA, psoralens and ultraviolet-A therapy; SMR, standard mortality rate

These findings are the primary basis for the hypothesis that psoriasis is an independent risk factor for CVD, most likely as a result of inflammation (Sterry et al., 2007; Fridewald et al., 2008; Spah, 2008; Kimball et al., 2008a, b). A secondary hypothesis is that systemic anti-inflammatory immunologic therapies for cutaneous psoriasis might also reduce cardiovascular morbidity and mortality (Prodanovich et al., 2005; Gelfand et al., 2006, 2007; Sterry et al., 2007; Fridewald et al., 2008; Spah, 2008; Kimball et al., 2008a, b). However, recent studies in European populations suggest that psoriasis is unlikely to be a substantial independent risk factor for CVD (Brauchli et al., 2009; Wakkee et al., 2010). In a 30-year prospective study of patients with severe psoriasis, we determined cause-specific mortality, tested the hypothesis that excess in mortality among patients with severe psoriasis is greater for CVD than that due to other causes, and examined the relation of mortality risk with the extent of psoriasis. RESULTS Since 1 January 1977, the 1,376 cohort patients have been studied prospectively for a total of 30,817 person-years. Data collected from direct contact with patients, their physicians, and relatives span 26,285 (85%) person-years. Information from the National Death Index (NDI) comprises an additional 4,532 (15%) person-years of follow-up. Death certificates were sought for all reported deaths.
www.jidonline.org 1159

Received 23 December 2009; revised 11 November 2010; accepted 17 November 2010; published online 20 January 2011

& 2011 The Society for Investigative Dermatology

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

Table 1. Attributes and exposures of PUVA study cohort at entry to study in 19751976 for 1,376 of the 1,380 originally enrolled who survived to 1 January 1977
(N)
Gender Male Female Education Post high school/unknown High school or less Unknown Married No Yes Unknown Body mass index (kg m2) 1,2 p302 4302 Unknown BSA (quartiles)3 BSA 215% BSA 1626 BSA 2742 BSA 442 Unknown Blood pressure
1

In 1977, the patients mean age was 46 (SD15). A total of 15% reported fair or poor health; 92% had at least 10% of BSA involved with psoriasis at entry. Nearly all had used systemic therapy or phototherapies other than psoralens and ultraviolet-A therapy (PUVA) or had been hospitalized for psoriasis.
Population and risk factors

891 (65%) 485 (35%)

738 (53%) 545 (40%) 93 (7%)

402 (30%) 971 (70%) 3 (o1%)

975 (71%) 204 (15%) 197 (14%)

Table 1 provides the distribution of patient characteristics, including risk factors for CVD determined from health histories, laboratory, and medical examinations performed at entry into the clinical trial (19751976) or at the first PUVA follow-up study interview (1977). Using a standardized form and method, dermatologists assessed BSA involved with psoriasis before first PUVA treatment (19751976). Table 2 provides the distribution of levels of exposures to smoking and psoriasis therapies that varied over the course of the study. For each year, the patients cumulative exposure was compared with that of all other cohort patients to determine that patients relative level of exposure compared with the mean level for surviving cohort members in that year. Each patients category for each year was used in the analysis for that year.
Deaths overall

326 (26%) 322 (23%) 347 (25%) 349 (25%) 32 (2%)

o135 and o95 X135 or X95 Unknown General health4 Fair or poor Good or excellent Unknown Alcohol use1 No Yes Unknown Complete blood count (thousands) o6.5 6.58.1 48.1 Unknown Uric acid (mg/dl) Normal5 High Unknown

983 (71%) 104 (8%) 289 (21%)

210 (15%) 1,153 (84%) 13 (o1%)

230 (17%) 956 (69%) 190 (14%)

423 (31%) 480 (35%) 467 (34%) 6

879 (64%) 462 (34%) 35 (3%)

Abbreviations: BSA, body surface area; PUVA, oral psoralen and UVA photochemotherapy. 1 One center (184 patients) did not collect these data. 2 Body mass index formula: weight (kg)/[(height (m))2; p30: normal weight, 430: obese. 3 Body surface area estimation by dermatologist. 4 Self-assessed general health at entry to clinical trial. 5 For males, 3.47.0 and for females, 2.411.0 (mg/dl).

We documented 617 deaths compared with 560 expected (standard mortality ratio (SMR) 1.1, 95% confidence interval (CI) 1.021.20). Table 3 presents the observed and expected deaths and 95% CIs for all-cause mortality and major causes of death. As detailed in Supplementary Table S1 online, except for liver-related death, the causes of death in our cohort closely paralleled those in the general population. Table 4 presents the results of the Cox proportional hazards model for eight patient attributes and exposures listed in Tables 1 and 2 that were available for most patients and were significantly associated with all-cause mortality, after adjustment for age and sex. Supplementary Table S2 online provides these estimates including three potential confounders for which data were unavailable for B15% of patients. In our cohort, the significant risk factors for death are similar to those for the general population including obesity, smoking, being unmarried, having less education, fair or poor health, hypertension, elevated uric acid, and white blood cell count (Table 4). Among the four quartiles of the extent of psoriasis at entry (215%, 1626%, 2742%, and 442% BSA), only patients in the highest quartile (442% BSA on entry) had a significantly higher risk of death (hazard ratio (HR) 1.55, 95% CI 1.231.93) than those with the least psoriasis at entry. The mortality risk of patients with intermediate extents of psoriasis was nearly identical to those with less psoriasis (Table 4). The results of the primary multivariate and all univariate and multivariate sensitivity analyses demonstrated significant associations between very severe psoriasis and all-cause mortality and with deaths not due to CVD or cancer. Although CVD was the most frequent cause of death in our cohort, no significant relationship between increasing extent of psoriasis and CVD mortality was seen in these analyses (Tables 4 and 5, and Supplementary Tables S2 and S3 online).

1160 Journal of Investigative Dermatology (2011), Volume 131

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

Table 2. Patient exposures that were documented at multiple interviews and may have changed over time (N=30,817 patient-years of follow-up)
No. of follow-up years
Smoking more than the mean Yes No Unknown
1

Table 3. Observed and expected deaths, SMR, and 95% CI by cause of death and extent of psoriasis among 1,376 cohort patients
Observed (N)
Underlying cause of death All causes Major cardiovascular diseases1 Neoplasms2 Other causes3 617 246 146 225 560.28 241.2 143.4 161.4 1.11 (1.021.20) 1.02 (0.901.16) 1.02 (0.861.20) 1.39 (1.221.59)

Expected

SMR (95% CI)

9,354 (30%) 19,916 (65%) 1,547 (5%)

PUVA dosage more than mean2 Yes No 11,699 (38%) 19,118 (62%)

Percent psoriasis at entry (quartiles) o16 1626 137 137 138 190 15 140.6 135.9 135.2 123.4 10.7 0.97 (0.821.15) 1.01 (0.851.19) 1.02 (0.861.21) 1.54 (1.331.77) 1.40 (0.782.31)

MTX months more than mean3 Yes No 7,610 (24%) 23,207 (76%)

2742 442 Missing

Abbreviations: MTX, methotrexate; PUVA, oral psoralen and UVA photochermotherapy. 1 We calculated the mean total pack-years of smoking for all smokers in the calendar year and compared it with the individual patients total that year. 2 We calculated mean total PUVA treatments for each year for the group and compared it with the individual patients total that year. 3 For MTX users, we calculated mean total MTX in use months for each year for the group and compared it with the individual patients total that year.

Abbreviations: CI, confidence interval; SMR, standard mortality ratio. 1 International Classification of Disease, 9th Edition (ICD-9)-codes: 390434, 436448. 2 ICD-9 codes: 140208. 3 Includes 19 patients for whom we could not establish the cause of death.

Cardiovascular death

A total of 246 deaths due to CVD occurred in our cohort, compared with the 241 expected (SMR 1.02, 95% CI 0.901.16). The SMRs for subsets of CVD, including myocardial infarction (ICD-9 (International Classification of Disease, 9th Edition) code 410; SMR 0.95, 95% CI 0.761.23) and all heart diseases (ICD-9 codes 390398, 402, 404, 410, and 429; SMR 1.03, 95% CI 0.921.21), did not suggest that severe psoriasis is associated with CVD mortality. The well-established cardiovascular risk factors were also significantly related to cardiovascular mortality in our cohort (Table 4). In both the univariate and multivariate analyses, patients in the highest quartile of extent of psoriasis (442%) were at slightly but not statistically significantly higher risk of dying from CVD than those in the lowest quartile. No apparent difference in the risk of death because of CVD was observed among the first three quartiles of psoriasis severity (Table 4). Multivariate models showed similar associations (Table 5 and Supplementary Table S2 online). The cohorts proportional mortality odds ratio for CVD standardized for age and gender equaled 0.40 versus 0.44 for the US white population (P 0.08).
Cancer mortality

(Table 4). In the univariate analysis, for those with greatest extent of psoriasis, the risk of cancer was B50% higher than those with least extensive disease, but did not reach statistical significance (Table 4). In the multivariate models, the HR for cancer death among cohort members with 442% BSA was lower, as expected, compared with those with o15% BSA and also not significant (HR 1.31, 95% CI 0.901.92) (Table 5 and Supplementary Table S2 online).
Other causes of death

Patients in our cohort were significantly more likely to die from causes other than CVD and cancer (SMR 1.39, 95% CI 1.221.59). The univariate, multivariate, and sensitivity analyses, all indicate that very severe psoriasis (442% BSA) was associated with an increased risk of death from causes other than CVD and cancer compared with those with less extensive psoriasis (Tables 4 and 5 and Supplementary Table S2 online).
CVD versus other causes of death

In both univariate and multivariate stratified analyses, the point estimate of the risk of death because of causes other than CVD and cancer in the patients with most extensive psoriasis compared with that observed for CVD was 41 but did not reach statistical significance (multivariate estimate HR 1.27, 95% CI 0.861.99). DISCUSSION Studies from specialized populations, including the General Practice Research Database of outpatient records of British
www.jidonline.org 1161

Observed cancer deaths were nearly identical to that expected (SMR 1.02, 95% CI 0.861.20). Among the potential confounders evaluated, only smoking and methotrexate use were significant predictors of death from cancer

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

Table 4. Cox proportional HRs with 95% CIs for PUVA cohort for death by cause and by patient attributes and exposuresadjusted for sex and age
Cause of death
All deaths (N=617) Attribute/exposure Percent psoriasis (quartiles) o15 1626 2742 43100 Body mass index (kg m2) p30 X30 Education Post high school High school or less Married No Yes Blood pressure o95 and o135 X95 or X135 General health Good or excellent Fair or poor Alcohol No Yes Smoking No or less than mean pack-years Yes and more than mean pack-years PUVA treatments Less than mean number Greater than mean number Methotrexate use, months None or less than mean Greater or equal to mean White blood cell count (thousands) o6.5 6.58.1 48.1 Uric acid (mg/dl) Normal High 1 1.43 (1.211.67) 1 1.41 (1.091.81) 1 1.55 (1.112.15) 1 1.37 (1.041.79) 1 1.27 (1.041.57) 1.72 (1.402.11) 1 1.00 (0.731.37) 1.29 (0.941.75) 1 1.89 (1.222.92) 1.90 (1.212.99) 1 1.32 (0.911.89) 2.25 (1.613.17) 1 1.14 (0.951.36) 1 0.97 (0.721.31) 1 1.41 (0.992.00) 1 1.17 (0.871.57) 1 0.81 (0.680.96) 1 0.74 (0.560.97) 1 0.98 (0.691.37) 1 0.80 (0.601.06) 1 1.43 (1.201.69) 1 1.35 (1.031.77) 1 1.67 (1.172.38) 1 1.39 (1.051.85) 1 0.81 (0.651.00) 1 0.79 (0.561.11) 1 0.81 (0.521.26) 1 0.82 (0.581.18) 1 1.96 (1.642.42) 1 2.46 (1.843.27) 1 1.08 (0.661.77) 1 2.08 (1.512.86) 1 1.45 (1.131.86) 1 1.69 (1.152.49) 1 1.27 (0.752.14) 1 1.39 (0.912.12) 1 0.70 (0.580.83) 1 0.70 (0.520.93) 1 0.83 (0.571.21) 1 0.62 (0.460.83) 1 1.38 (1.171.64) 1 1.41 (1.091.85) 1 1.40 (1.001.97) 1 1.34 (1.001.78) 1 1.27 (1.031.57) 1 1.27 (0.901.78) 1 1.43 (0.952.15) 1 1.18 (0.841.68) 1 1.04 (0.821.31) 1.03 (0.811.30) 1.55 (1.231.93) 1 1.05 (0.731.53) 1.00 (0.691.45) 1.37 (0.971.94) 1 0.99 (0.611.63) 1.09 (0.671.76) 1.49 (0.942.36) 1 1.07 (0.711.58) 1.02 (0.681.58) 1.79 (1.242.59) HR, 95% CI Cardiovascular deaths (N=246) HR, 95% CI Cancer (N=146) HR, 95% CI All other deaths (N=225) HR, 95% CI

Abbreviations: CI, confidence interval; HR, hazard ratio; PUVA, oral psoralen and UVA photochermotherapy.

1162 Journal of Investigative Dermatology (2011), Volume 131

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

Table 5. Multivariate HR with 95% CI estimates for PUVA cohort by cause of death1 and percent psoriasis
Cause of death
All deaths (N=577) HR, 95% CI Percent psoriasis (quartiles) o15 1626 2742 43100 1 1.04 (0.821.32) 1.04 (0.821.32) 1.45 (1.161.82) 1 1.01 (0.691.49) 1.00 (0.681.48) 1.23 (0.851.80) 1 1.01 (0.621.67) 1.00 (0.611.63) 1.32 (0.822.14) 1 0.93 (0.601.43) 0.97 (0.631.49) 1.57 (1.012.26)
2

Cardiovascular deaths (N=209)2 HR, 95% CI

Cancer deaths (N=133)2 HR, 95% CI

Other (N=188)2 HR, 95% CI

Highest quartile versus all less severe

1.42 (1.181.69)

1.22 (0.891.66)

1.31 (0.901.92)

1.56 (1.142.13)

Abbreviations: CI, confidence interval; HR, hazard ratio; PUVA, oral psoralen and UVA photochermotherapy. 1 Adjusted for all significant potential confounders except obesity, alcohol, and blood pressure, which in bivariate analysis altered HR by o10% (see Materials and Methods). Potential confounders included in model by cause of death: all deaths: uric acid; cardiovascular deaths: education, general health, PUVA, and uric acid; cancer deaths: education and methotrexate; other causes: education, general health, smoking, and uric acid. 2 Number of deaths included in the multivariate model (see Table 4 for total deaths by cause).

general practitioners, have suggested that patients with severe psoriasis have increased risks of CVD that are independent of usual cardiovascular risk factors and attributable to their psoriasis (Rocha-Pereira et al., 2001; Reynoso-von Drateln et al., 2003; Gelfand et al., 2006, 2007; Melerba et al., 2006; Neimann et al., 2006; Akhyani et al., 2007; Huerta et al., 2007; Ludwig et al., 2007; Shapiro et al., 2007; Wakkee et al., 2007; Gerdes et al., 2008; Kaye et al., 2008; Kimball et al., 2008a, b; Ohtsuka, 2008; Balci et al., 2009). Some have suggested that the increased risk of CVD observed in patients with severe psoriasis may be a result of cutaneous inflammation and immunologic alterations common to psoriasis and CVD (Sterry et al., 2007; Fridewald et al., 2008; Spah, 2008; Federman et al., 2009; Kimball et al., 2008a, b). Proponents of this view had stated that in particular it is important to determine the impact of clinical markers of psoriasis, such as body surface area, on the risks of MI in patients with psoriasis (Gelfand et al., 2006). An association between psoriasis severity and C-reactive protein levels, a modest risk factor for CVD, has been observed (Strober et al., 2008). A clinical trial to test the hypothesis that tumor necrosis factor-a blocker will decrease the risk of heart disease is underway (ClinicalTrials.gov). In our 30-year prospective study of a cohort of US patients with severe psoriasis, mortality rates were slightly and significantly higher than those for the comparable US population of whom substantially B0.1% have psoriasis of comparable severity (Stern et al., 2004). When deaths due to liver disease are excluded, mortality in the cohort was nearly identical to that expected (SMR 1.06, 95% CI 0.971.15). In our cohort, patients with exceptionally severe psoriasis (442% BSA) have significantly higher overall mortality rates than cohort patients with less extensive but still severe psoriasis as well as that observed in the general population. However, our data do not support the hypothesis that either severe psoriasis or greater extent of psoriasis is a unique or significant independent risk factor for death due to CVD.

Our prospective study included severely affected patients enrolled at 16 university centers nationwide in the first large clinical trial of a widely heralded new therapy for psoriasis (Parrish et al., 1974). The only health-related exclusions to entry were pregnancy and inability to stand for 20 minutes (Melski et al., 1977; Stern et al., 1984). However, the robustness of our findings that compare cause-specific mortality in our cohort with that expected for a comparable US population depend on our study having limited selection bias. A well cohort effect (i.e., selection bias) could account for our failure to find an increased risk of CVD deaths in the cohort overall, but our data suggest that this is unlikely. At enrollment, 15% of our cohort assessed their health as fair or poor, significantly higher than the 11% reporting fair or poor health in a comparable population (Hendershot, 1988, 2006). At enrollment, 29% of the cohort had hypertension and 15% had body mass index 430 kg m2; age-adjusted prevalence consistent with those for the general population (NCHSa). A comparison of the population-based First National Health and Nutrition Examination Survey (NHANES-1) cohort and ours also argues against our observation reflecting a well cohort effect. The average age of whites in NHANES-1 was slightly older than our population (50 vs. 46) (McGee et al., 1996). The proportion of patients in the PUVA cohort with white blood cell counts in the highest tercile for NHANES was similar to that observed in the NHANES cohort (Gillum et al., 2005). Abnormal uric acid levels were more frequent in our cohort than in the NHANES cohort (Fang and Alderman, 2000). The increases in risk for both all-cause mortality and cardiovascular mortality associated with elevated white blood cell count and uric acid that we observed were similar to those seen in NHANES-1 (Fang and Alderman, 2000; Gillum et al., 2005). Including these biological variables in our multivariate models further reduced the risk associated with very severe psoriasis. Our cohort enrolled residents of many states. However, death rates for the states that accounted for most patients are
www.jidonline.org 1163

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

B4% lower than national averages. Therefore, our estimates of SMRs may be slight underestimates. However, the differences between the national death rates we utilized and state-specific rates would not impact our findings concerning the relation of increasingly severe psoriasis and mortality. The direction and magnitude of risks associated with potential confounders we observed are generally similar to those observed in the general population (Fang and Alderman, 2000; Gillum et al., 2005; Fiscella and Tancredi, 2008). If present, the effect of selection bias with respect to CVD would be expected to decrease over time, which would be reflected in a higher SMR for CVD in the later years of the study that was not seen (P 0.52). All members of our cohort had severe psoriasis based on clinical or General Practice Research Database criteria (Stern et al., 2004; Gelfand et al., 2006; Leonardi et al., 2008). More than half of our cohort has used methotrexate. The cohort has had 4450,000 UVB and 200,000 PUVA treatments. For each end point, univariate and multivariate statistical models, including those of our sensitivity analyses, yielded highly consistent results. Cohort members with the most extensive psoriasis (top quartile) are significantly more likely to die than those with less psoriasis overall and from causes other than CVD and cancer. Our study has a number of strengths and limitations. It was prospective and achieved a high rate of follow-up. It enrolled a geographically dispersed cohort with severe psoriasis and did not exclude patients because of ill health or medication use. Based on death certificates and other medical information, we used standard methods to assess the cause of death. Hence, our classification of the cause of death should be comparable with that used in calculating the population incidence data. Death records are likely to be more specific than outpatient records used in other studies of CVD in psoriasis (Gelfand et al., 2006; Hammad et al., 2008). With 430,000 person-years of prospective, we only have sufficient power to assess frequent causes of death. Our findings demonstrate that exceptionally extensive psoriasis and/or its treatment are associated with increased mortality from noncardiovascular causes. Our findings strongly argue against a substantial or greater association of severe and extremely severe psoriasis with CVD than with other causes of death. The high prevalence of risk factors for CVD in patients with severe psoriasis and the nearly significant increased risk of CVD mortality in people with exceptionally severe psoriasis suggest that clinicians screen patients with severe psoriasis for CVD risk factors and be sure that psoriasis patients with risk factors receive appropriate counseling and treatment. MATERIALS AND METHODS
Study population
In 19751976, 16 dermatology departments at university centers established the Clinical Cooperative Trial of Oral Methoxsalen Therapy (PUVA) to evaluate a new and highly promising therapy for psoriasis (Melski et al., 1977). The eligibility criteria were severe psoriasis and the ability to stand for 20 minutes and not be pregnant. At enrollment, a detailed questionnaire and laboratory and physical
1164 Journal of Investigative Dermatology (2011), Volume 131

examinations established medical history and physical attributes including some risk factors for CVD. In the late 1976, a long-term follow-up study was organized enrolling 1,380 of the 1,450 participants in the clinical trial. They form the PUVA Follow-Up Study cohort. Its methods have been detailed previously (Stern et al., 1984, 1997). Follow-up information gathered on the cohort included that from 22 cycles of follow-up interviews, 9 cycles of dermatologic examinations, as well as medical records and death certificates. To decrease lead time bias, this report is restricted to 1,376 of the 1,380 cohort patients who survived to 1 January 1977 and were prospectively followed until 2005. Our study conformed to the Declaration of Helsinki Principles. The study was approved by the BIDMC Committee on Clinical Investigation (IRB). Written informed consent was obtained at entry to the original study (19751976) and at enrollment in the follow-up study (1976).

Ascertainment and classification of mortality

From enrollment to 2004, trained interviewers performed 22 cycles of telephone interviews that documented patients health status. If patients could not be reached, clinicians and patient-designated relatives and friends were contacted. We attempted to obtain death certificate information for all those reported dead or we could not contact. We also utilized the NDI, a nationwide database that ascertains cause of death for 99% of US residents (NCHSb; NDI, 2005). We used the same methods to determine the cause of death as are used to calculate US mortality rates (NCHSc, d). After 1998, death certificates utilized the tenth revision of the ICD codes. Therefore, we converted these codes to ICD-9 codes using the ICD code translator (TDRdata.com) to determine the cause of death. Two physicians evaluated all deaths with any ambiguity concerning the proper application of coding rules to determine the cause of death. For 41 of 96 such cases, full agreement about correct code was reached. For 47 cases, ambiguity between two diagnoses persisted. For 19 of these 47 cases, we coded CVD as the underlying cause of death. For 17 of these 19 cases, diabetes mellitus was an alternative. For only 6 of the 28 remaining cases, CVD could be considered a likely cause of death. To test for consistency between our coding of underlying cause of death and the official NDI coding, which is used to calculate causespecific death rates, we compared the causes of death in nearly 60% sample of deaths available as a complete file of NDI codes. In this sample, our coding agreed with the NDI in 490% of cases and the total number of deaths coded as CVD differed by o2%. We used diagnosis, gender, age, and calendar-year-specific death rates for whites to calculate the expected number of deaths (CDCa; Doody et al., 2001; NCHSe, f). We did this for deaths overall, for all diagnostic (three-digit) categories with X4 observed deaths, and NCHS (National Center for Health Statistics)-defined groupings for CVD and cancer (CDCb).

Statistical analysis
To evaluate our cohorts overall and cause-specific mortality experience, we compared the observed and expected numbers of deaths. We calculated the expected number of deaths for all-cause mortality and for specific diseases based on the United States. We applied age (5-year intervals), sex, and calendar-year-

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

specific mortality rates for whites from the NCHS Multiple-Cause Mortality Files of the Centers for Disease Control and Prevention (CDCa, b). For 19771978, we lacked detailed data, and used 1979 death rates for these years. Rates for 1979 differed by o2% from 19771978 rates. Based on Poisson distribution, we calculated the SMR and its exact 95% CI for each cause of death in our cohort with at least four deaths. All analyses were performed using statistical software Stata Version 10 (StataCorp, College Station, TX). Using percent BSA at entry to the study, as assessed by specially trained dermatologists who categorized the percentage of BSA covered with psoriasis for each of 33 body areas, we stratified our populations extent of psoriasis into quartiles. At entry, the lowest quartile of patients had BSA 215%, BSA in the second and third quartiles were 1626% and 2742%, respectively. Those in the highest quartile had at least 43% BSA involvement (Table 1). To assess the association of increasingly severe psoriasis with the risk of death, we used Cox proportional hazard models (Cleaves et al., 2008). We graphed the hazard function over time for deaths because of all causes as well as cardiovascular deaths and deaths due to other causes separately. Visual inspection of the log(-log) survival plots against time confirmed that the proportional hazard assumption were met. We developed Cox regression models that related psoriasis severity to mortality overall and separately for CVD, cancer, and other causes (see Supplementary Table S1 online for classification of ICD-9 codes included in each category). We estimated the HR for psoriasis severity and potential confounders (see Tables 1 and 2) to each end point. Following recent STROBE (strengthening the reporting of observational studies in epidemiology) recommendations, we tested all potential confounders that were significant at Po0.1 in the univariate analysis and did bivariate analyses of each of these variables with psoriasis severity (Von Elm et al., 2007). In our primary multivariate models, any potential confounder that altered the point estimate of the HR for the most severe quartile of psoriasis by X7% (e.g., from 1.5 to p1.45 or X1.55) were included. All models included age and sex. For 8 of the 11 potential confounding variables, data were available for 495% of patients (see Tables 1 and 2). One center with 184 patients did not provide blood pressure, alcohol use, and body mass index data. Therefore, these variables were excluded from the primary multivariate models. The results of multivariate analyses including these three variables are presented in Supplementary Table S2 online. We used stratified Cox regression to estimate differences in HRs for competing risk (CVD versus other causes of death) (Lunn and McNeil, 1995). We performed a number of sensitivity analyses. Using Cox regression analysis, we developed a multivariate model that included all potential confounders significantly related to each outcome and with Po0.20 in the multivariate model that related extent of psoriasis to risk of death by cause. We also used binomial regression to determine the association of psoriasis severity and death by cause. Our measures of risk in this model were agegender, cause, and year-specific death rates for whites in the United States. These binomial regression models also included all significant potential confounders with Po0.20 in the univariate analyses.
CONFLICT OF INTEREST
RSS has served as a consultant to Johnson and Johnson on issues not related to psoriasis and for divisions that do not market drugs for psoriasis.

ACKNOWLEDGMENTS
We thank more than 30 investigators who participated in the PUVA follow-up study data collection and the 1,380 patients, who shared their medical information and freely gave their cooperation and time over nearly 30 years. We thank Professor Rebecca Betensky for her helpful statistical advice. This study was funded by the NIH contract NO1-AR-02446 and the Beth Israel Deaconess Department of Dermatology. The funding organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.

AUTHOR CONTRIBUTIONS
RSS and AH: substantial contribution to the conception and design or acquisition of data, or analysis or interpretation of data; drafting the article or revising it critically; final approval of the version to be published. RSS had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. SUPPLEMENTARY MATERIAL Supplementary material is linked to the online version of the paper at http:// www.nature.com/jid

REFERENCES
Akhyani M, Ehsani AH, Robati RM et al. (2007) The lipid profile in psoriasis: a controlled study. J Eur Acad Dermatol Venereol 21:13302 Balci DD, Balci A, Karazincir S et al. (2009) Increased carotid artery intimamedia thickness and impaired endothelial function in psoriasis. J Eur Acad Dermatol Venereol 23:16 Brauchli YB, Jick SS, Miret M et al. (2009) Psoriasis and risk of incident cancer: an inception cohort study with a nested case-control analysis. J Invest Dermatol 129:260412 CDCa (Centers for Disease Control and Prevention) CDC Wonder: Compressed Mortality File Archives. http://wonder.cdc.gov/mortarchives. html (accessed Spring 2008) CDCb (Centers for Disease Control and Prevention) CDC Wonder: About Compressed Mortality, 19791998. http://wonder.cdc.gov/cmf-icd9. html (accessed 20 February, 2009) Cleaves MA, Gould WM, Gutierrez RG et al. (2008) An Introduction to Survival Analysis Using Stata. 2nd edn. Stata Press: Texas, 372 ClinicalTrials.gov. Effects of TNF-alpha antagonism (etanercept) in patients with the metabolic syndrome and psoriasis. http://clinicaltrials.gov/ct2/ show/NCT00477191?term=metabolic+syndrome+psoriasis&rank=2 (accessed 5 May 2009) Doody MM, Hayes HM, Bilgrad R (2001) Comparability of national death index plus and standard procedures for determining causes of death in epidemiologic studies. Ann Epidemiol 11:4650 Fang J, Alderman MH (2000) Serum uric acid and cardiovascular mortality. The NHANES epidemiologic follow-up study, 19711992. JAMA 283:240410 Federman DG, Shelling M, Prodanovich S et al. (2009) Psoriasis: an opportunity to identify cardiovascular risk. Br J Dermatol 160:17 Fiscella K, Tancredi D (2008) Socioeconomic status and coronary heart disease risk prediction. JAMA 300:26668 Fridewald VE, Cather JC, Gelfand JM et al. (2008) AJC editors consensus: psoriasis and coronary artery disease. Am J Cardiol 102:163143 Gelfand JM, Neimann AL, Shin DB et al. (2006) Risk of myocardial infarction in patients with psoriasis. JAMA 296:173541 Gelfand JM, Troxel AB, Lewis JD et al. (2007) The risk of mortality in patients with psoriasis: results from a population-based study. Arch Dermatol 143:12939 Gerdes S, Zahl VA, Knopf H et al. (2008) Comedication related to comorbities: a study in 1203 hospitalized patients with severe psoriasis. Br J Dermatol 159:111623 Gillum RF, Mussolino ME, Madans JH (2005) Counts of neutrophils, lymphocytes and monocytes, cause-specific mortality and coronary heart disease: the NHANES-I epidemiologic follow-up study. Ann Epidemiol 15:26671

www.jidonline.org 1165

RS Stern and A Huibregtse

Psoriasis, Mortality, and Cardiovascular Disease

Hammad TA, McAdams MA, Feight A et al. (2008) Determining the predictive value of Read/OXMIS codes to identify acute myocardial infarction in the general practice research database. Pharmacoepidemiol Drug Saf 17:1197201 Hendershot GE (1988) Health of the foreign-born population: United States, 19851986. Adv Data 157:16 Huerta C, Rivero E, Garcia Rodriguez LA (2007) Incidence and risk factors for psoriasis in the general population. Arch Dermatol 143:155965 Kaye JA, Li L, Jick SS (2008) Incidence of risk factors for myocardial infarction and other vascular diseases in patients with psoriasis. Br J Dermatol 159:895902 Kimball AB, Gladman D, Gelfand JM et al. (2008a) National psoriasis foundation clinical consensus on psoriasis comorbidities and recommendations for screening. J Am Acad Dermatol 58:103142 Kimball AB, Robinson D, Wu Y et al. (2008b) Cardiovascular disease and risk factors among psoriasis patients in two US healthcare databases, 20012002. Dermatology 217:2737 Kurd SK, Gelfand JM (2008) The prevalence of previously diagnosed and undiagnosed psoriasis in US adults: results from NHANES 20032004. J Am Acad Dermatol 60:21824 Leonardi CL, Kimball AB, Papp KA et al. (2008) Efficacy and safety ustebinumab, a human interleukin- 12/23 monoclonal antibody, in patients with psoriasis: 76-week results from a randomized, doubleblind, placebo-controlled trial. Lancet 371:166574 Ludwig RJ, Herzog C, Rostock A et al. (2007) Psoriasis: a possible risk factor for development of coronary artery calcification. Br J Dermatol 156:2716 Lunn M, McNeil D (1995) Applying Cox regression to competing risks. Biometrics 51:52432 McGee D, Cooper R, Liao Y et al. (1996) Patterns of comorbidity and mortality risk in blacks and whites. Ann Epidemiol 6:3815 Melerba M, Gisondi P, Radaeli A et al. (2006) Plasma homocysteine and folate levels in patients with chronic plaque type psoriasis. Br J Dermatol 155:11659 Melski JW, Tanenbaum L, Parrish JA et al. (1977) Oral methoxsalen photochemotherapy for the treatment of psoriasis: a cooperative clinical trial. J Invest Dermatol 68:32835 NCHSa (National Center for Health Statistics) National Vital Statistics System. Prevalence of BMI 25+ by age, sex, and survey. http://www.cdc.gov/ nchs/data/nhanes/overweight.pdf (accessed 29 January 2008) NCHSb (National Center for Health Statistics) National Death Index Users Guide. http://www.cdc.gov/nchs/data_access/ndi/ndi_user_guide.htm (accessed 2 February 2009) NCHSc (National Center for Health Statistics) National Death Index Facts and Figures, 2008. http://www.cdc.gov/nchs/data/dvs/2a2008Final.pdf (accessed January and February 2008) NCHSd (National Center for Health Statistics) ICD-9-CM Official Guidelines for Coding and Reporting. http://www.cdc.gov/nchs/data/icd9/ icdguide09.pdf (accessed 22 February 2009) NCHSe (National Center for Health Statistics) Death rates by 10-year age groups and age-adjusted death rates for 113 selected causes, race, and sex: United States, 197998. http://www.cdc.gov/nchs/data/statab/ hist001r.pdf (accessed December 2007, January 2008, and February 2008)

NCHSf (National Center for Health Statistics) Vital Statistics Data Available Online. http://www.cdc.gov/nchs/data_access/Vitalstatsonline. htm#Mortality_Multiple (accessed December 2007, January 2008, and February 2008) NDI. National Death Index Plus: Coded Causes of Death Supplement to the National Death Index Users Manual. National Center for Health Statistics: Hyattsville, MD, 2005 Neimann AL, Shin DB, Wang X et al. (2006) Prevalence of cardiovascular risk factors in patients with psoriasis. J Am Acad Dermatol 55:82935 Ohtsuka T (2008) The relation between high-sensitivity C-reactive protein and maximum body mass index in patients with psoriasis. Br J Dermatol 158:113473 Parrish JA, Fitzpatrick TB, Tanenbaum L et al. (1974) Photochemotherapy of psoriasis with oral methoxsalen and longwave ultraviolet light. N Engl J Med 291:120711 Prodanovich S, Ma F, Taylor JR et al. (2005) Methotrexate reduces the incidence of vascular diseases in veterans with psoriasis or rheumatoid arthritis. J Am Acad Dermatol 52:2627 Reynoso-von Drateln C, Martinez-Abdunis E, Balcazar-Munoz BR et al. (2003) Lipid, profile, insulin secretion, and insulin sensitivity in psoriasis. J Am Acad Dermatol 48:8225 Rocha-Pereira P, Santos-Silva A, Rebelo I et al. (2001) Dislipidemia and oxidative stress in mild and in severe psoriasis as a risk for cardiovascular disease. Clin Chim Acta 303:339 Shapiro J, Cohen AD, David M et al. (2007) The association between psoriasis, diabetes mellitus and atherosclerosis in Israel: a case-control study. J Am Acad Dermatol 56:62934 Spah F (2008) Inflammation in atherosclerosis and psoriasis: common pathogenic mechanisms and the potential for an integrated treatment approach. Br J Dermatol 159:107 Stern RS, Laird N, Melski J et al. (1984) Cutaneous squamous-cell carcinoma in patients treated with PUVA. N Engl J Med 310:115661 Stern RS, Nichols KT, Vakeva LH (1997) Malignant melanoma in patients treated for psoriasis with methoxsalen (psoralen) and ultraviolet-a radiation (PUVA). The PUVA follow-up study. N Engl J Med 336:10415 Stern RS, Nijsten T, Feldman SR et al. (2004) Psoriasis is common, carries a substantial burden even when not extensive, and is associated with widespread treatment dissatisfaction. J Invest Dermatol 9:1369 Sterry W, Strober BE, Menter A (2007) Obesity in psoriasis: the metabolic, clinical and therapeutic implications. Report of an interdisciplinary conference and review. Br J Dermatol 157:64955 Strober B, Teller C, Yamauchi P et al. (2008) Effects of etanercept on Creactive protein levels in psoriasis and psoriatic arthritis. Br J Dermatol 159:32230 TDRdata.com. ICD-9 and -10 Translation Tools. http://www.tdrdata.com/IPD/ ipd_ICDTranslator.aspx (accessed 26 May 2009) Von Elm E, Altman DG, Egger M et al. (2007) The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 370:14537 Wakkee M, Herings RMC, Nijsten T (2010) Psoriasis may not be an independent risk factor for acute ischemic heart disease hospitalizations: results of a large population-based Dutch cohort. J Invest Dermatol 130:9627 Wakkee M, Thio HB, Prens EP et al. (2007) Unfavorable cardiovascular risk profiles in untreated and treated psoriasis patients. Artherosclerosis 190:19

1166 Journal of Investigative Dermatology (2011), Volume 131