Herpes Zoster and Long‐Term Risk of Cardiovascular Disease

We investigated the longitudinal association of herpes zoster (HZ), commonly known as “shingles,” and long‐term risk of stroke or coronary heart disease (CHD) among participants in 3 large US cohorts, the NHS (Nurses' Health Study), NHS II (Nurses' Health Study II), and HPFS (Health Professionals Follow‐Up Study).

Participants were 79 658 women in the NHS (2000–2016), 93 932 women in the NHS II (2001–2017), and 31 440 men in the HPFS (2004–2016), without prior stroke or CHD. Information on HZ, stroke, and CHD was collected on biennial questionnaires and confirmed by medical record review. Cox proportional hazards regression models were used to estimate multivariable‐adjusted hazard ratios for stroke and for CHD according to years since HZ compared with never HZ. During >2 million person‐years of follow‐up, 3603 incident stroke and 8620 incident CHD cases were documented. History of HZ was significantly and independently associated with higher long‐term risk of stroke and CHD. In pooled analyses, compared with individuals with no history of HZ, the multivariable‐adjusted hazard ratios (95% CIs) for stroke were 1.05 (0.88–1.25) among those with 1 to 4 years since HZ, 1.38 (1.10–1.74) for among those with 5 to 8 years since HZ, 1.28 (1.03–1.59) among those with for 9 to 12 years since HZ, and 1.19 (0.90–1.56) among those with ≥13 years since HZ. For CHD, the corresponding multivariable‐adjusted hazard ratios (95% CIs) were 1.13 (1.01–1.27) for 1 to 4 years, 1.16 (1.02–1.32) for 5 to 8 years, 1.25 (1.07–1.46) for 9 to 12 years, and 1.00 (0.83–1.21) for ≥13 years.

HZ is associated with higher long‐term risk of a major cardiovascular event. These findings suggest there are long‐term implications of HZ and underscore the importance of prevention.

The NHS was established in 1976 when 121 700 female registered nurses, aged 30 to 55 years, were enrolled by completing a baseline questionnaire. In 1989, the NHS II was established and enrolled 116 429 female registered nurses, aged 25 to 42 years. The HPFS began in 1986 and enrolled 51 529 male health professionals, aged 40 to 75 years. Participants completed questionnaires at baseline and every 2 years on a wide range of demographic, health, diet, and lifestyle factors, including detailed information on health history and medication use. The follow‐up rates in all 3 cohorts exceed 90% of eligible person‐time.^22, 23 For the present analyses, baseline was defined in each cohort based on the year when information on date of HZ event was first available through most recent follow‐up cycle (NHS: 2000–2016; NHS II: 2001–2017; and HPFS: 2004–2016). We excluded participants for whom the year of their HZ event was not provided and those with a history of stroke or CHD before study baseline. In total, 79 658 women in the NHS, 93 932 women in the NHS II, and 31 440 men in the HPFS were included in the analyses (total n=205 030). The Institutional Review Boards of Brigham and Women's Hospital and Harvard T.H. Chan School of Public Health approved the study protocol. The return of the questionnaires was considered to imply informed consent.

Information on history of “shingles” (HZ) and the date it occurred was collected on the 2000, 2004, 2008, and 2012 NHS questionnaires, the 2001, 2005, 2013, and 2017 NHS II questionnaires, and the 2004, 2006, and 2008 HPFS questionnaires. Participants were asked about clinician‐diagnosed “shingles” and the year of diagnosis. To confirm the validity of self‐report in our participants, we mailed a supplemental questionnaire to 99 NHS women that asked for permission to obtain medical records relating to the diagnosis of shingles, which were reviewed by a physician investigator (G.C.C.). The diagnosis of HZ was confirmed in 97%, demonstrating that self‐reported diagnosis of shingles (HZ) in this cohort of nurses is highly reliable. Notably, the participants in all 3 of our cohorts are health care professionals, and the reliability of health‐related information collected from our participants has been demonstrated for a large number of health outcomes.²² The primary exposure for this study was categorized according to time (in years) since the participant's HZ event. Those with no history of HZ were the referent group.

Stroke was classified as total stroke, encompassing ischemic stroke and hemorrhagic stroke, confirmed by physician review of medical records according to criteria specified in the National Survey of Stroke,²⁴ requiring evidence of a sudden or rapid onset of neurologic deficit that persisted for >24 hours or until death. Participants who reported a nonfatal stroke on a biennial questionnaire were asked for permission to review their medical records. Deaths were reported by next of kin or postal authorities or determined by systematic searches of the National Death Index, and permission for medical records was sought. Our group previously reported that 98% of the deaths in each cohort are ascertained.^25, 26 Cases of nonfatal stroke that were confirmed by letter/interview but lacked sufficient medical documentation were considered probable stroke. We considered both stroke confirmed by medical record review and probable stroke as cases in our analyses.

CHD was defined as nonfatal or fatal MI, fatal CHD, or coronary revascularization procedure (coronary artery bypass graft or percutaneous transluminal coronary angioplasty). Participants who reported an incident event on a biennial questionnaire were asked for permission to obtain medical records, which were reviewed by physicians who were blinded to the exposure status and the specifc research question under study. Nonfatal MI was confirmed according to the World Health Organization criteria, which required typical symptoms plus diagnostic electrocardiographic findings or elevated enzyme levels.²⁷ Fatal CHD was identified by medical records or autopsy reports or if CHD was listed as the cause of death on the death certificate along with prior evidence of CHD. The study designated as probable fatal CHD deaths when no medical records surrounding the death were available, but CHD was the underlying cause on the death certificate or National Death Index search, or a family member provided supporting information on the diagnosis. Deaths were identified by report by the next of kin or the postal system or by searching the National Death Index using the methods described for stroke above.^25, 26 We considered both CHD confirmed by medical record review and probable CHD as cases in our analyses, as well as self‐reported revascularization procedures.^28, 29

In our multivariable‐adjusted analyses, we adjusted for several factors that could potentially be related to HZ and stroke or CHD, including age, race, smoking history, body mass index, waist circumference, physical activity, medical conditions strongly related to cardiovascular risk (eg, diabetes, hypertension, and elevated cholesterol), use of aspirin, thiazide or loop diuretics, statins, and antihypertensive medication, diet quality (Alternative Healthy Eating Index 2010) score, menopausal status (women), postmenopausal hormone therapy use (women), oral contraceptive use (NHS II), family history of heart disease (defined as maternal history of MI before 65 years of age or paternal history of MI before 55 years of age), and self‐reported medical conditions that potentially compromise immunity because of disease or treatment (eg, cancer other than nonmelanoma skin cancer, rheumatoid arthritis, Crohn's disease, ulcerative colitis, systemic lupus erythematosus, asthma, diabetes, chronic obstructive pulmonary disease, and oral steroid use). Covariate information, including age, weight, physical activity, smoking status, medication use, and physician diagnosis of chronic diseases, was assessed and updated every 2 to 4 years by biennial questionnaires administered throughout follow‐up. Diet information was collected every 4 years using a validated semiquantitative food frequency questionnaire.^{30, 31, 32, 33} Among the women in the NHS and NHS II, information on menopausal status, postmenopausal hormone therapy, and oral contraceptive use (NHS II only) was also obtained. Self‐reported information on height and weight was used to calculate body mass index. If covariate information was missing, we carried the information forward from up to 2 previous questionnaire cycles. The validity of covariate information collected in these cohorts has been demonstrated and is described in previous publications.^{22, 28, 34, 35}

The short‐term (up to 1 year) relation of HZ and cardiovascular disease (CVD) outcomes has been demonstrated previously.¹³ This study focuses on long‐term relations. The analyses were conducted using a prospective design, with information on HZ collected before the stroke or CHD event. We had information on the precise date of the CVD event, but the reporting of the date of HZ was less precise. To reduce the possibility that the CVD event preceded the episode of HZ, we censored participants who reported their HZ and had a stroke or CHD event within ±12 months. Thus, our study did not evaluate the short‐term relation of HZ and CVD. Person‐time was calculated for each participant based on the return date of the questionnaire to the diagnosis of stroke, CHD, death, or the end of follow‐up, whichever occurred first. Time (in years) since the HZ event was calculated based on the duration of the time interval from the reported diagnosis of HZ until the beginning of each 2‐year time period. HZ and covariate status were updated with each time period. We categorized time since HZ as never, 1 to 4 years since HZ, 5 to 8 years since HZ, 9 to 12 years since HZ, and ≥13 years since HZ. We used Cox proportional hazards models with time‐varying covariates and age as the underlying time scale, with stratification by calendar time (in 2‐year intervals), to estimate hazard ratios (MVHRs [95% CIs]) and assess the association between time since HZ and the subsequent risk of stroke or CHD event. In each cohort, separate analyses were conducted for risk of stroke, risk of CHD, and risk of a composite CVD outcome (a composite of stroke or CHD, whichever came first). We also performed pooled analyses for each outcome with the use of fixed‐effects meta‐analysis with inverse‐variance weighting. Heterogeneity was assessed with the I² statistic, and low to moderate heterogeneity was observed (I²<40%). We conducted sensitivity analyses that restricted the CHD outcome to fatal and nonfatal MI and fatal CHD events and censored those who reported coronary revascularization procedures (coronary artery bypass graft or percutaneous transluminal coronary angioplasty) during follow‐up. We also conducted stratified analyses among those with and without potentially immunocompromising health conditions, including cancer, rheumatoid arthritis, Crohn's disease, ulcerative colitis, systemic lupus erythematosus, asthma, diabetes, chronic obstructive pulmonary disease, or oral steroid use.

During a total of 2 471 975 person‐years of follow‐up, 2461 cases of incident stroke among women in the NHS, 537 incident strokes among women in the NHS II, and 605 incident strokes among men in the HPFS were documented and included in the analysis. History of HZ was significantly and independently associated with higher long‐term risk of incident stroke (Table 2). In pooled multivariable analyses accounting for multiple potential confounding factors, compared with individuals with no history of HZ, the MVHRs (95% CIs) for incident stroke were 1.05 (0.88–1.25) among those with 1 to 4 years since HZ, 1.38 (1.10–1.74) among those with 5 to 8 years since HZ, 1.28 (1.03–1.59) among those with 9 to 12 years since HZ, and 1.19 (0.90–1.56) among those with ≥13 years since HZ. The findings for the individual cohorts are also shown in Table 2.

During a total of 2 350 066 person‐years of follow‐up, 4910 cases of incident CHD (nonfatal/fatal MI, fatal CHD, or coronary revascularization procedure) among women in the NHS, 1183 cases among women in the NHS II, and 2527 cases among men in the HPFS were documented and included in the analysis. History of HZ was significantly and independently associated with higher long‐term risk of CHD (Table 3). In pooled multivariable analyses accounting for multiple potential confounding factors, compared with individuals with no history of HZ, the MVHRs (95% CIs) for incident CHD were 1.13 (1.01–1.27) among those with 1 to 4 years since HZ, 1.16 (1.02–1.32) among those with 5 to 8 years since HZ, 1.25 (1.07–1.46) among those with 9 to 12 years since HZ, and 1.00 (0.83–1.21) among those with ≥13 years since HZ. The findings for the individual cohorts are also shown in Table 3.

The results from sensitivity analyses that restricted the CHD outcome to fatal and nonfatal MI and fatal CHD and censored those who reported coronary revascularization procedures are shown in Table S1. In the female cohorts, the findings were similar. However, in men, compared with those with no history of HZ, we observed a statistically significant higher risk of CHD among those with a history of HZ using the more restrictive outcome definition; compared with men with no history of HZ, the MVHR (95% CI) for incident MI was 1.36 (1.02–1.81) among those with ≥5 years since HZ.

In analyses examining HZ and risk of a composite CVD outcome (stroke or CHD, whichever came first), previous HZ was significantly and independently associated with risk of a major cardiovascular event (Table 4). In the pooled multivariable analyses, compared with individuals with no history of HZ, the MVHRs (95% CIs) for a cardiovascular event were 1.11 (1.01–1.23) among those with 1 to 4 years since HZ, 1.26 (1.13–1.41) among those with 5 to 8 years since HZ, 1.27 (1.11–1.46) among those with 9 to 12 years since HZ, and 1.08 (0.92–1.28) among those with ≥13 years since HZ. Table 4 also shows the findings for the individual cohorts.

We performed stratified analyses that compared the association of HZ and long‐term risk of stroke and CHD among those with and without potentially immunocompromising conditions. There was a suggestion that the magnitude of the elevated risk for stroke after ≥5 years since HZ was greater among women with potentially immunocompromising conditions; however, the P value for interaction was statistically significant only in the NHS II (P‐interaction=0.05) (Table S2). In men, we did not observe that the risk varied by immune status (P‐interaction=0.68), but the number of cases among men with a history of HZ was small. There was a suggestion that the magnitude of the risk for CHD after ≥5 years since HZ was greater among women with potentially immunocompromising conditions in the NHS II (P‐interaction=0.03) (Table S3), but there was no evidence that the risk differed by immune status in the other 2 cohorts (P‐interaction ≥0.2). Similarly, there was a suggestion that the magnitude of the risk for CVD (composite) after ≥5 years since HZ was greater among women with potentially immunocompromising conditions in the NHS II (P‐interaction=0.02) but not in the other 2 cohorts (P‐interaction ≥0.28) (Table S4).

In the male cohort, we did not have information on HZ that occurred after 2008; thus, men who had HZ after the return of the 2008 questionnaire were misclassified in the later time periods; this could have biased the results for the 1 to 4 and 5 to 8 years since HZ toward the null. When we conducted sensitivity analyses that ended follow‐up in 2010, the findings were similar.

In additional analyses that included intakes of alcohol and specific dietary factors as individual covariates, the results were not materially different (data not shown). To consider the possibility that the true effect size may differ among the cohorts, we also conducted meta‐analyses using the random effects approach. The findings were essentially unchanged.