This review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [13] (S1 Checklist). The protocol for this study was registered on the PROSPERO registry for systematic reviews (CRD42016045194) [14]. Searches on the PubMed^® (includes MEDLINE^® content) and Scopus^® (includes Embase^® content) electronic databases were performed in December 2016 to identify relevant articles. Text headings and medical subject heading (MeSH) terms used in the searches included herpes zoster, shingles, herpes zoster ophthalmicus, herpes zoster oticus, zoster sine herpete, herpes zoster oticus, stroke, transient ischemic attack, myocardial infarction, angina, cardiac death, coronary heart disease, coronary revascularization, and vascular diseases. This search strategy emphasized terms for the most prevalent cerebrovascular and cardiovascular events [15] (S1 File). We reviewed the reference lists of eligible articles and performed searches on Google Scholar and OAIster (database of the Open Archives Initiative) to identify additional research not indexed by the searched databases.

This review included studies that: (1) investigated the association between herpes zoster, or a subtype of herpes zoster, with cerebrovascular and cardiovascular events as defined in the primary studies and included stroke, transient ischemic attack, myocardial infarction, angina, coronary heart disease, cardiac death, coronary revascularization, and vascular diseases; (2) used a trial, cohort, case-control, case-crossover, or cross-sectional (with a temporal relationship established between exposure and outcome) design; (3) were written in English; (4) were published in a peer-review journal or as a conference paper; and (5) were published between January 1, 1960 and December 28, 2016. Excluded studies did not meet the inclusion criteria or (1) were a case-study/series, abstract (with no accompanying full text), review article, or cross-sectional study without the temporal relationship established between the exposure and outcome; or (2) examined primary VZV infection (i.e., chickenpox) as the exposure; or (3) restricted to patients less than 18 years of age; or (4) studied immunocompromised populations (e.g., HIV, cancer).

One author (N.E.) reviewed all publication titles to eliminate articles that covered a subject not relevant to this review, described a primary exposure that was not herpes zoster or its subtypes, or used an immunocompromised, pediatric, or non-human sample. Two authors (N.E. and H.T.) reviewed abstracts of the remaining articles to determine inclusion for assessment of the full text. Discrepancies were resolved by group consensus and independent adjudication (R.G.) if consensus could not be reached. One author (N.E.) extracted data onto a standardized data collection form. Extracted information included study title and authors, year of publication, source of participants and sample size, mean age and gender composition of the study sample, type of herpes zoster studied (e.g., herpes zoster, herpes zoster ophthalmicus), cerebrovascular and cardiovascular event outcomes (e.g., stroke, transient ischemic attack, myocardial infarction, angina, coronary heart disease, cardiac death, coronary revascularization, and vascular diseases), and effect estimates with accompanying 95% confidence intervals. The extracted forms were reviewed (H.T.) for accuracy with the source paper.

To assess the quality of abstracted information, including the clarity of reporting study methods, error, and bias, we used questions adapted from the 32-item questionnaire developed by Downs and Black to rate patient intervention studies [16]. Following the approach of other systematic reviews [17–19], we removed items irrelevant to observational studies not studying an intervention as the primary exposure. Due to a lack of validation, we did not calculate a composite quality score but report scores in the appendix. Since the number of studies we included was low, we did not report the results of the Eggers test.

We performed qualitative and quantitative syntheses. We organized our findings according to type of herpes zoster, outcome (cerebrovascular or coronary event), and time examined between the onset of herpes related illness and our pre-defined disease outcomes. We considered meta-analysis when 2 or more studies could be pooled after evaluating the studies for clinical and statistical heterogeneity. Clinical heterogeneity was examined by evaluating for differences in study population, intervention or comparators. Statistical heterogeneity was estimated using the I² statistic to evaluate both the random and fixed effects models. I² values of 30–60% represented a moderate level of heterogeneity [20].

We conducted both random and fixed effects meta-analysis. We conducted random effects meta-analysis using the inverse variance method for pooled odds ratios (ORs). We used the fixed effects model when the number of studies was low. We assumed similarity between the Ors and other relative measures, such as relative risk, rate ratios, or hazard ratios because the cardiac events examined were relatively infrequent (< 1/10) events [21]. Where possible, we pooled adjusted Ors from the primary studies; otherwise, we used the unadjusted Ors with no correction for baseline differences or confounding. We assessed for publication bias using the Eggers test. All meta-analyses were conducted in Stats Direct.

Since prior studies have reviewed the association between herpes zoster and cerebrovascular events, but not cardiac events, and estimated pooled ORs, we summarized the findings of these studies forest plots to facilitate comparisons with our results.

Table 1 presents an overview of the eight retrospective cohort studies [22–30] and three self-case controlled case series [31–33] included in this review. We included self-controlled case series since these studies have a design akin to that of an efficient cohort study that minimizes confounding and produces measures of association, unlike a traditional case series [34]. The studies sampled populations from six countries including enrollees of the national health insurance systems in Taiwan [23, 25, 28, 29] and South Korea [24], Danish hospital and prescription drug users [26], British primary care patients [22, 31], American Medicare enrollees 65 years and older [32], residents 50 years and older in Olmsted County (MN, US) [30], enrollees of four private insurers in Germany [33] and residents of Västra Götaland County, Sweden [27]. All included studies performed retrospective reviews of electronic medical record databases; no study prospectively followed patients.

As exposures, six of the included studies examined any form of herpes zoster (herpes zoster type unspecified, which includes herpes zoster ophthalmicus) [24, 26–30], five examined herpes zoster type unspecified with sub-group analyses for herpes zoster ophthalmicus [22, 23, 31–33], and one study only examined patients with herpes zoster ophthalmicus [25].

Included studies examined cerebrovascular and cardiovascular events over time periods ranging from 1 week to 24 years after onset of herpes zoster. The studies used a heterogeneous definition of cerebrovascular and cardiac events. Eleven of the included studies examined associations between herpes zoster type unspecified and the following: non-specified stroke [22, 23, 27, 30, 31, 33], ischemic stroke [32, 33], hemorrhagic stroke [33] [33], TIA [22], a composite of stroke and TIA [24, 26], myocardial infarction [22, 30, 32], acute coronary syndromes (e.g., acute myocardial infarction and unstable angina) [28], and incident coronary artery disease, including angina and myocardial infarction [29]. Among studies examining patients with herpes zoster ophthalmicus, five studies examined non-specified stroke [22, 23, 25, 31, 33], two studies examined ischemic stroke [32], one examined hemorrhagic stroke [33], and one examined myocardial infarction [32] as endpoints.

In examining the relationships between herpes zoster, type unspecified, and cerebrovascular events, five studies found significantly positive associations with non-specific stroke [22, 23, 27, 30, 31, 33], two studies found associations with ischemic stroke [32, 33], one study with hemorrhagic stroke [33], one study with TIA [22], and two studies with a composite endpoint of stroke and TIA [24, 26].

The results of the meta-analyses of the association of herpes zoster, type unspecified, and cerebrovascular events stratified by the length of follow-up are shown in Fig 2. Meta-analyses found that patients with herpes zoster were at significantly increased odds of a cerebrovascular event within 3 months of onset using the random effects and fixed effects models (pooled Ors for fixed and random effects models 1.34, 95% CI: 1.22 to 1.46). There was no evidence for statistical heterogeneity (I² = 0.0%) for the pooled results for cerebrovascular events. Similarly, the odds of a cerebrovascular event were significantly increased up to 1 year after herpes zoster onset (pooled Ors for fixed and random-effects models 1.22, 95% CI: 1.15 to 1.29), but with severe statistical heterogeneity (I² = 99.7%). The odds of experiencing a cerebrovascular event over periods of time greater than 1 year after herpes zoster onset was significant in the fixed effects model (pooled OR 1.40, 95% CI: 1.38 to 1.43), but not the random effects model (pooled OR 1.20, 95% CI: 0.82 to 1.75), with severe statistical heterogeneity (I² = 99.7%).

Overall, three studies found significantly positive associations between exposure to herpes zoster, type unspecified, with myocardial infarction within one week [32], three months [30], or twenty-four years after zoster onset [22], with acute coronary syndromes within 12 years after zoster onset [28], and with incident coronary artery disease within two and 10 years after zoster onset (Table 2) [29].

Meta-analyses of the association of herpes zoster, type unspecified, and various cardiac events stratified by length of follow-up is shown in Fig 3. Patients with herpes zoster were at significantly increased odds of a cardiac event at 3 months after onset using the fixed effects model (pooled OR 1.31, 95% CI: 1.02 to 1.70), but not the random effects model (pooled OR 1.34. 95% CI 0.98 to 1.82). We found a low level of statistical heterogeneity (I² = 23.9%) for the pooled results for myocardial infarction. The odds of a cardiac event up to 1 year after onset were significantly higher using the fixed and random effects model (both pooled Ors 1.19, 95% CI: 1.01 to 1.41). We found no evidence of statistical heterogeneity (I² = 0.0%). Likewise, the odds of a cardiac event over periods of greater than 1 year after zoster onset were significantly higher using both the fixed and random effects models (both pooled Ors 1.12, 95% CI: 1.08 to 1.16). We found no evidence of statistical heterogeneity (I² = 0.0%).

Exposure to herpes zoster ophthalmicus was statistically significantly associated with development of non-specific stroke over time periods of up to one year in four studies [23, 25, 31, 33], with the development of ischemic stroke over periods up to three months in two studies [32, 33], but not with hemorrhagic stroke [33] (Table 3). Meta-analyses of the association of herpes zoster ophthalmicus and cerebrovascular events stratified by length of follow up are shown in Fig 4. Compared with unexposed individuals, patients with herpes zoster ophthalmicus were at significantly increased odds of cerebrovascular events within 3 months of illness onset in meta-analyses using fixed-effects (pooled OR 1.39, 95% CI: 1.25 to 1.54) and random effects (pooled OR 1.85, 95% confidence interval 1.09 to 3.12) models. We found a high level of statistical heterogeneity (I² = 88.9%) for the pooled results for cerebrovascular events. Similarly, patients exposed to herpes zoster ophthalmicus had an increased odds of cerebrovascular events within 1 year of onset in meta-analyses using both the fixed and random effects models (both pooled ORs 4.42, 95% confidence interval 2.75 to 7.11), without any evidence of statistical heterogeneity (I² = 0%).

The one study to examine the association between herpes zoster ophthalmicus and acute cardiac events showed that the relative frequency of myocardial infarction was significantly higher in the first week after disease onset (Table 4) [32].

The Forest plots from previous meta-analysis examining the association between herpes zoster, type unspecified and herpes zoster ophthalmicus with stroke are shown in Fig 5. All meta-analyses show a consistently positive relationship between herpes zoster and herpes zoster ophthalmicus with stroke.