Herpes Zoster and the Risk for Stroke in Patients with Autoimmune Diseases

Abstract

Background

Herpes zoster (HZ) is an opportunistic infection caused by varicella zoster virus (VZV), and observed with increasing frequency in patients on immunosuppressive therapies. Prior literature has suggested that the risk of stroke may increase shortly after herpes zoster, but little is known about this association for patients with autoimmune (AI) diseases, who are at increased risk both for zoster and stroke.

Methods

Medicare data (2006–2013) was used to identify patients with autoimmune diseases. The outcome of interest was hospitalized stroke. The hypothesis tested was that the incidence of stroke immediately following HZ was increased compared to the incidence of stroke at later time points. Secondary analysis included assessment of the impact of antiviral therapy on subsequent stroke as well as the influence of VZV-related complications on stroke incidence.

Results

The crude incidence of stroke ranged from a high of 2.30 (0.96–5.52) per 100 patient years within 90 days of HZ in patients who had HZ-related cranial nerve complications and did not receive treatment to a low of 0.87 (0.75–1.02) per 100py (>1 year after uncomplicated HZ). After multivariable adjustment for multiple stroke-related factors, the incidence rate ratio (IRR) for stroke in the first 90 days after HZ was 1.36 (1.10–1.68) compared to stroke >1 year after HZ. The risk was greater for patients with zoster with cranial nerve complications (IRR 2.08, 95%CI 0.99–4.36). Prompt antiviral therapy was associated with lower incidence of subsequent stroke (IRR 0.83, 95% CI 0.70–0.98).

Conclusion

In patients with autoimmune diseases, incident HZ was associated with up to a two-fold increased risk for stroke in the subsequent few months. These data provide urgency for developing strategies to reduce the risk of VZV.

Introduction

Herpes zoster (HZ) infection, also known as shingles, is caused by reactivation of latent varicella-zoster virus infection generally acquired decades earlier. HZ has been variably considered an opportunistic infection: while it clearly occurs in “normal” hosts, it is readily understood that diminished cell mediated immunity leads to zoster [1]. While more than half of the individuals developing zoster are over age 60, patients with autoimmune disease and particularly those on immunosuppressive therapy are at 1.5–2× the risk for HZ versus healthy adult controls[2].

In general, HZ manifests as a painful vesicular rash that, even when uncomplicated, can be highly symptomatic. HZ is, however, often pernicious with frequent complications, the most common of which is post herpetic neuralgia (PHN) which can often lead to long term pain, depression and disability[3, 4]. In immunocompromised hosts, even more severe complications can be observed including vasculopathy, myelitis, motor nerve involvement, cranial nerve palsies and ocular involvement [5]. The annual health care costs related to HZ have been estimated in the hundreds of millions of dollars and this does not begin to account for the pain and reduced quality of life caused by the disease [6].

While most complications (aside from PHN) are rare, a series of recent epidemiologic studies suggest that HZ may be associated with a more common and serious complication in the form of increased risk of stroke. These studies - using varying methodology and performed on three continents - have demonstrated that in patients with HZ there is a statistical and clinically significant risk of stroke ranging from 30% to 430% peaking in the first few weeks after HZ and tapering off over the ensuing year[7–12].

Given that these studies examining the relationship between HZ and stroke have been performed in the general population, we designed our study to examine a patient population traditionally considered at increased risk for HZ (i.e. those with immune suppressive illnesses) to determine whether they too were at increased risk of stroke. Secondarily, we examined whether early antiviral therapy in patients with HZ had an ameliorative effect by reducing the incidence of stroke.

Methods

Study design and data sources

We conducted a retrospective cohort study using 2006–2013 Medicare claims data for 100% of patients with ankylosing spondylitis (AS), inflammatory bowel disease (IBD), psoriasis (PSO), psoriatic arthritis (PSA), and rheumatoid arthritis (RA). Medicare is a US federal benefit program that provides health insurance to approximately 93% of US adults 65 years of age and older; it also provides insurance coverage for approximately 7 million people younger than 65 with disabling conditions (e.g. rheumatoid arthritis) or who have end-stage renal disease (ESRD)[13]. The Institutional Review Board of the University of Alabama at Birmingham approved the study and the study data was governed by a Data Use Agreement from the Center for Medicare and Medicaid Services (CMS).

Cohort

Our study cohort consisted of patients experiencing HZ as identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis code for zoster (Appendix Table 1) from either an inpatient or outpatient health care provider encounter. Follow up started at the date zoster was diagnosed (the ‘index date’). To be included in the analysis, we also required all patients to have at least 12 consecutive observable months (enrolled in Medicare Part A, Part B and Part D, and not enrolled in Medicare part C) prior to the index date and have at least one of the five included inflammatory diseases defined by two ICD-9-CM diagnosis codes separated by 7 days and within one year using all available data prior to index date. Patients with any inpatient or outpatient health care provider visit claim with any diagnosis code for stroke (using all available data prior to the index date) or anti-viral treatment for zoster during base line were excluded. Follow up ended at the earliest date of: a hospitalized ischemic or hemorrhagic stroke; 24-months after index date; loss of Medicare coverage; or the end of study (Dec 31, 2013).

Appendix Table 1.

ICD9 codes for HZ categories and stroke

Disease	Disease subgroup	ICD9 diagnosis code	Description
Zoster		053
	With cranial nerve complication	05311	GENICULATE HERPES ZOSTER
		05312	POSTHERPETIC TRIGEMINAL NEURALGIA
		05320	HERPES ZOSTER DERMATITIS OF EYELID
		05321	HERPES ZOSTER KERATOCONJUNCTIVITIS
		05322	HERPES ZOSTER IRIDOCYCLITIS
		05329	HERPES ZOSTER WITH OTHER OPHTHALMIC COMPLICATIONS
		05371	OTITIS EXTERNA DUE TO HERPES ZOSTER
	With other complication	0530	HERPES ZOSTER WITH MENINGITIS
		05310	HERPES ZOSTER WITH UNSPECIFIED NERVOUS SYSTEM COMPLICATION
		05313	POSTHERPETIC POLYNEUROPATHY
		05314	HERPES ZOSTER MYELITIS
		05319	HERPES ZOSTER WITH OTHER NERVOUS SYSTEM COMPLICATIONS
		05379	HERPES ZOSTER WITH OTHER SPECIFIED COMPLICATIONS
		0538	HERPES ZOSTER WITH UNSPECIFIED COMPLICATION
	Without complication	0539	HERPES ZOSTER WITHOUT COMPLICATION
Stroke	Hemorrhagic	430	SUBARACHNOID HEMORRHAGE
		431	INTRACEREBRAL HEMORRHAGE
	Ischemic	43301	OCCLUSION AND STENOSIS OF BASILAR ARTERY WITH CEREBRAL INFARCTION
		43311	OCCLUSION AND STENOSIS OF CAROTID ARTERY WITH CEREBRAL INFARCTION
		43321	OCCLUSION AND STENOSIS OF VERTEBRAL ARTERY WITH CEREBRAL INFARCTION
		43331	OCCLUSION AND STENOSIS OF MULTIPLE AND BILATERAL PRECEREBRAL ARTERIES WITH CEREBRAL INFARCTION
		43381	OCCLUSION AND STENOSIS OF OTHER SPECIFIED PRECEREBRAL ARTERY WITH CEREBRAL INFARCTION
		43391	OCCLUSION AND STENOSIS OF UNSPECIFIED PRECEREBRAL ARTERY WITH CEREBRAL INFARCTION
		43401	CEREBRAL THROMBOSIS WITH CEREBRAL INFARCTION
		43411	CEREBRAL EMBOLISM WITH CEREBRAL INFARCTION
		43491	CEREBRAL ARTERY OCCLUSION UNSPECIFIED WITH CEREBRAL INFARCTION
	Other	436	ACUTE BUT ILL-DEFINED CEREBROVASCULAR DISEASE

Exposure and risk windows

The zoster phenotype was classified into one of three mutually exclusive categories based on diagnosis codes (Appendix Table 1) within 7 days after first diagnosis code (hierarchically) as follows: zoster with cranial nerve complication, zoster with other complication and zoster without complication. If patients met criteria for more than one phenotype, they were assigned to the highest (i.e. with complications). This classification was used to allow the hypothesis to be tested in a subgroup analysis that zoster with cranial nerve complications would have a stronger association with stroke than zoster without complications. The risk windows used for stroke ascertainment were defined as the time since the index date (0–90 days, 91–365 days months, and 366–733 days) and were selected based on prior literature that showed HZ stroke associations in a non-inflammatory disease population [7–12]. Antiviral treatment for HZ (i.e. acyclovir, valacyclovir, famciclovir, foscavir, ganciclovir) was identified within 7 days after the zoster diagnosis.

Outcome

Hospitalized stroke was defined by patients having a diagnosis code of stroke (Appendix Table 1) in the primary or non-primary position on a hospital claim. A subgroup analysis was conducted that was limited to hospitalized ischemic stroke. The positive predictive value of zoster diagnosis alone to identify incident zoster exceeds 85% and is even higher when requiring anti-viral drug use [12,14, 29].

Covariates

We evaluated demographic characteristics including age, gender and race. In addition to the inflammatory conditions of interest, we also studied comorbidities (Appendix Table 2) including diabetes, hypertension, atrial fibrillation, hyperlipidemia, obesity, transient ischemic attack, tobacco use, and carotid artery disease; these were assessed using ICD 9 diagnosis codes and/or procedure codes using a 12-month baseline period. Medications of interest included biologics disease modifying antirheumatic drugs (DMARD) (yes/no), non-biologic DMARDs (yes/no), and mean glucocorticoid dose averaged over the 183 days prior to the start of follow-up. Covariates were selected based upon content knowledge of risk factors for stroke or zoster that would be relevant for RA patients.

Appendix Table 2.

ICD 9 diagnosis code for covariates

Co-morbidities	ICD9 code	Description
Atrial fibrillation	42731	Atrial fibrillation
Carotid or other artery disease	41401	Coronary atherosclerosis of native coronary artery
	43310	Occlusion and stenosis of carotid artery without cerebral infarction
	4479	Unspecified disorders of arteries and arterioles
Diabetes	250	Diabetes with/without complication
	3572	Polyneuropathy in diabetes
	3620	Diabetic retinopathy
	36641	Diabetic cataract
Hyperlipidemia	2720	Pure hypercholesterolemia
	2721	Pure hyperglyceridemia
	2722	Mixed hyperlipidemia
	2724	Other and unspecified hyperlipidemia
Hypertension	4010	Malignant essential hypertension
	4011	Benign essential hypertension
	4019	Unspecified essential hypertension
Obesity	V852	Body mass index 25.0–29.9
	V853	Body mass index 30.0–39.9
	V854	Body mass index 40.0 and above
	27800	Obesity unspecified
	27801	Morbid obesity
	27802	Over weight
	79391	Image test inconclusive due to excess body fat
Transient ischemic attack	4359	Unspecified transient cerebral ischemia
Ankylosing spondylitis	7200	Ankylosing spondylitis
Inflammable bowel disease	555	Regional enteritis
	556	Ulcerative colitis
Psoriasis	6961	Other psoriasis and similar disorders
Psoriatic arthritis	6960	Psoriatic arthropathy
Rheumatoid arthritis	7140	Rheumatoid arthritis
	7142	Other rheumatoid arthritis with visceral or systemic involvement
	71481	Rheumatoid lung

Statistical methods

We generated descriptive statistics for demographic characteristics, co-morbidities, and medications use, stratified by each of the three zoster phenotypes. We calculated incidence rates (IRs) for subsequent hospitalized stroke by dividing the number of events by person years and generated 95% confidence intervals (CI) using Poisson regression for each exposure group and risk window. Generalized linear models for repeated measures were used to estimate incidence rate ratios (IRRs) and 95% confidence intervals. We conducted a pooled analysis that included patients regardless of whether or not they were treated with anti-viral medications, and included this medication use as a term in the multivariable model to estimate the effect of anti-viral medication for zoster on subsequent stroke. Finally, as part of a sensitivity analysis, we used Cox proportional hazards models to compute hazard ratios that adjusted for the competing risk for death. All analyses were performed using SAS 9.4 software (SAS Institute).

Results

We identified 267,906 patients with inflammatory diseases (any one diagnosis code) who developed HZ. The following were excluded: those without at least 12 enrolled months before first zoster (n=169,135); patients who used antiviral medication within 12 months before the index date (n=14,633); patients who did not meet our study criteria requirement for autoimmune inflammatory diseases (n=33,272); and patients with a diagnosis of stroke prior to the index date (n=7,339). The final cohort contained 43,527 patients. Categorizing these individuals, 3,080 had zoster with cranial nerve complications; 4,494 had zoster with other complications, and 35,953 had zoster without complication (Figure 1). Demographics, co-morbidities, and medication use were similar among the three zoster groups (Table 1).

Figure 1.

Flow-chart for cohort selection

1. Observable: Enrolled in Medicare Part A, Part B, part D, and not in Part C.

2. Two ICD9 diagnosis code separated by 7 days and within 365 days for inflammatory diseases (ankylosing spondylitis, inflammatory bowel disease, psoriasis, psoriatic arthritis, or rheumatoid arthritis).

Table 1.

Baseline^* characteristics by herpes zoster-related complications

	Zoster with cranial nerve complication (N=3080)	Zoster with other complication (N=4494)	Zoster without complication (N=35953)
Demographic
Age in years, mean (STD)	72.2 (11.2)	72.0 (11.5)	71.0 (11.6)
Female, %	74.8	77.9	76.6
Race, %
Black	6.6	7.7	6.7
Other	6.5	8.2	6.4
White	86.9	84.0	86.9
Co-morbidities, %
Diabetes	22.4	23.8	22.4
Hypertension	55.3	57.7	55.3
Atrial fibrillation	9.0	9.5	8.8
Hyperlipidemia	33.8	34.8	33.0
Obesity	7.5	10.0	9.0
Transient ischemic attack	1.7	1.5	1.5
Carotid artery disease	16.6	17.6	15.4
Ankylosing spondylitis	1.8	1.6	1.7
Inflammatory bowel disease	15.7	15.2	16.6
Psoriatic arthritis	5.1	5.0	4.8
Psoriasis	20.2	18.3	19.2
Rheumatoid arthritis	65.4	67.9	65.2
Medication use, %
Biologic DMARD	20.2	19.5	20.4
Non-biologic DMARD	52.1	53.5	54.7
Steroid use/dose*
none	61.1	57.3	59.9
<7.5 mg	29.4	30.6	29.3
≥7.5 mg	9.6	12.2	10.8
Anti-viral drug***	70.3	55.9	79.6
Zoster vaccination†	5.1	5.5	5.0

DMARD: Disease modifying anti-rheumatic drug

^*measured in the 12 months prior to the start of follow-up, unless otherwise indicated

^**daily prednisone-equivalent dose calculated in the 6 months prior to the start of follow-up

^***Antiviral treatment for HZ (i.e. acyclovir, valacyclovir, famciclovir,foscavir, ganciclovir) was identified within 7 days after zoster diagnosis

^†Using all available data prior to the start of follow-up

In the overall cohort (Table 2, left column), the crude IR of stroke declined from 1.24 (95%CI: 1.05–1.49) per 100 patient years within 0–90 days of zoster to 0.94 (95%CI: 0.83–1.06) in days 366–730 post-zoster. In the antiviral treated zoster sub-group, the IRs of stroke declined from 1.13 (95%CI: 0.92–1.39) in the 0–90 day risk window to 0.89 (95%CI: 0.77–1.03) in the 366–730 day risk window. In the untreated zoster subgroup, the IRs of stroke were numerically higher than in the group treated for zoster with antivirals and followed a similar pattern. They declined from 1.60 (95%CI: 1.17–2.20) in 0–90 days to 1.10 (95%CI: 0.86–1.39) in the 366–730 day interval. The stroke incidence rates in the 91–365 day risk window were intermediate between these two other risk windows.

Table 2.

Incidence rate of hospitalized stroke^* by zoster phenotype

		Overall			zoster with cranial nerve complication			zoster with other complication			zoster without complication
		Event	Person- years	IR (95% CI) per 100py	Event	Person -years	IR (95% CI) per 100py	Event	Person- years	IR (95% CI) per 100py	Event	Person- years	IR (95% CI) per 100py
Overall	0–90 days	130	10463.2	1.24 (1.05–1.48)	17	740.9	2.29 (1.43–3.69)	18	1074.5	1.68 (1.06–2.66)	95	8647.9	1.10 (0.90–1.34)
	91–365 days	295	26908.6	1.10 (0.98–1.23)	25	1924.1	1.30 (0.88–1.92)	34	2711.5	1.25 (0.90–1.75)	236	22273.1	1.06 (0.93–1.20)
	366–730 days	255	27156.4	0.94 (0.83–1.06)	27	1968.0	1.37 (0.94–2.00)	26	2713.2	0.96 (0.65–1.41)	202	22475.2	0.90 (0.78–1.03)
With anti-viral treatment	0–90 days	91	8032.4	1.13 (0.92–1.39)	12	523.0	2.29 (1.30–4.04)	12	605.5	1.98 (1.13–3.49)	67	6903.8	0.97 (0.76–1.23)
	91–365 days	214	20817.0	1.03 (0.90–1.18)	16	1367.7	1.17 (0.72–1.91)	20	1553.0	1.29 (0.83–2.00)	178	17896.3	0.99 (0.86–1.15)
	366–730 days	188	21050.1	0.89 (0.77–1.03)	16	1400.1	1.14 (0.70–1.87)	14	1555.2	0.90 (0.53–1.52)	158	18094.8	0.87 (0.75–1.02)
Without anti-viral treatment	0–90 days	39	2430.8	1.60 (1.17–2.20)	5	217.8	2.30 (0.96–5.52)	6	468.9	1.28 (0.57–2.85)	28	1744.0	1.61 (1.11–2.33)
	91–365 days	81	6091.6	1.33 (1.07–1.65)	9	556.4	1.62 (0.84–3.11)	14	1158.4	1.21 (0.72–2.04)	58	4376.8	1.33 (1.02–1.71)
	366–730 days	67	6106.3	1.10 (0.86–1.39)	11	567.9	1.94 (1.07–3.50)	12	1158.0	1.04 (0.59–1.82)	44	4380.4	1.00 (0.75–1.35)

IR = incidence rate

^*includes both ischemic and hemorrhagic stroke

In the cohort of patients with zoster who had cranial nerve complication, the IRs of stroke were numerically greater and showed large changes in risk over time. The IRs were highest in the 0–90 day risk window: 2.29 (95%CI: 1.43–3.69) per 100 patient years within 0–90 days to 1.37 (95%CI: 0.94–2.00) in the 366–730 day window (Table 2). In the antiviral treated subgroup, IRs of stroke decline from 2.29 (95%CI: 1.30–4.04) within 0–90 days to 1.17 (95%CI: 0.72–1.91) in the 91–365 day risk window. The corresponding IRs in the same three risk windows among untreated patients were 2.30 (95%CI: 0.96–5.52), 1.62 (95%CI: 0.84–3.11) and 1.94 (1.07–3.50).

In the cohort of patients with zoster with complication other than cranial nerve involvement, patterns were similar but smaller in magnitude compared to those with cranial nerve complications (Table 2). In the antiviral treated subgroup, the IRs declined quickly from 1.98 (95%CI: 1.13–3.49) in the 0–90 day window to 1.29 (95%CI: 0.83–2.00) in the 91–365 days window and further declined to 0.90 (95%CI: 0.53–1.52) in the 366–730 days window. Rates of stroke among patients with zoster but without any zoster-related complications were smaller in magnitude and showed smaller trends over time (Table 2, right-most column). Overall, the majority of strokes in the overall analysis were ischemic (87%) and a sensitivity analysis with hospitalized ischemic stroke as the outcome revealed the patterns were similar to the main analysis that included all strokes (not shown).

In the multivariable adjusted results in the overall cohort (Table 3) stratified by HZ-associated complications, there was a 36% (HR = 1.36, 95% CI 1.10 – 1.68) increased risk for stroke within 90 days of HZ. This effect was larger in magnitude among those with HZ-related cranial nerve complications where the risk was increased by 76%. The trend was similar among those with zoster without complications and was not significant (HR = 1.26, 95% CI 0.99–1.60). In the overall analysis, use of anti-viral therapy was associated with a lower risk for subsequent stroke (multivariable-adjusted IRR 0.83, 95% CI 0.70–0.98). Results from the sensitivity analysis that adjusted for the competing risk for death were consistent with the primary results (not shown).

Table 3.

Incidence rate ratio for hospitalized stroke by zoster phenotype

		Overall		zoster with cranial nerve complication		zoster with other complication		zoster without complication
		Crude	Adjusted*	Crude	Adjusted*	Crude	Adjusted*	Crude	Adjusted*
Overall	0–90 days	1.33 (1.07–1.64)	1.36 (1.10–1.68)	1.68 (0.91–3.08)	1.76 (0.96–3.23)	1.75 (0.96–3.20)	1.77 (0.97–3.23)	1.23 (0.96–1.56)	1.26 (0.99–1.60)
	91–365 days	1.17 (0.99–1.38)	1.18 (1.00–1.40)	0.95 (0.55–1.63)	1.00 (0.58–1.74)	1.31 (0.79–2.18)	1.30 (0.78–2.18)	1.18 (0.98–1.42)	1.19 (0.99–1.43)
	366–730 days	reference	reference	reference	reference	reference	reference	reference	reference
With anti-viral treatment	0–90 days	1.27 (0.99–1.63)	1.30 (1.02–1.67)	2.01 (0.95–4.25)	2.08 (0.99–4.36)	2.21 (1.02–4.78)	2.21 (1.01–4.84)	1.11 (0.84–1.48)	1.13 (0.85–1.51)
	91–365 days	1.15 (0.95–1.40)	1.15 (0.94–1.39)	1.02 (0.51–2.05)	1.04 (0.51–2.14)	1.43 (0.72–2.84)	1.40 (0.70–2.81)	1.14 (0.92–1.41)	1.12 (0.91–1.39)
	366–730 days	reference	reference	reference	reference	reference	reference	reference	reference
Without anti-viral treatment	0–90 days	1.47 (0.99–2.18)	1.51 (1.02–2.24)	1.19 (0.41–3.42)	1.21 (0.42–3.50)	1.24 (0.47–3.29)	1.26 (0.48–3.31)	1.60 (1.00–2.57)	1.66 (1.04–2.67)
	91–365 days	1.21 (0.88–1.68)	1.24 (0.90–1.72)	0.84 (0.35–2.02)	0.86 (0.36–2.07)	1.17 (0.54–2.52)	1.18 (0.54–2.55)	1.32 (0.89–1.96)	1.37 (0.93–2.02)
	366–730 days	reference	reference	reference	reference	reference	reference	reference	reference

^*Adjusted for age, gender, race, diabetes, hypertension, atrial fibrillation, transient ischemic attack, glucocorticoids use.

Other variables not included in final model but tested included: smoking, obesity, carotid artery disease, biologic disease-modifying anti-rheumatic drug use, non-biologic disease-modifying anti-rheumatic drug use, zoster vaccination.

All adjusted IRRs were re-calculated using the time window of 1–2 years after zoster without complication as the referent (Figure 2). The IRRs were greatest for zoster with cranial nerve complications, particularly in the 0–90 day risk window after zoster. For these individuals, the adjusted IRRs exceeded two, and later time windows were not associated with significantly increased risk. Trends were similar for patients who had zoster with other complications, although the magnitude of the IRRs in the 0–90 day window was somewhat lower, albeit still statistically significant. No results for any risk window were significant among patients who had zoster without complications.

Figure 2.

Adjusted* incidence rate ratios of hospitalized stroke by time since herpes zoster and zoster phenotype

Risk window 366–730 days for zoster without complication as the reference.

*adjusted for age, gender, race, diabetes, hypertension, atrial fibrillation, transient ischemic attack, glucocorticoids use.

Discussion

In this large cohort of older patients with autoimmune and inflammatory diseases, we found a significantly elevated and time-dependent risk for hospitalized stroke following HZ, particularly among those who had HZ-related complications including cranial nerve involvement. Based upon the risk windows we examined (0–90 days, 90–365 days, and beyond 365 days), risk was greatest early and declined over time. Patients treated for HZ had a lower risk for stroke than those not treated, and this effect was more pronounced among those with HZ-related complications.

Our data documenting HZ as a risk factor for stroke are consistent with several previous studies performed on the general population[7–12] but not specifically addressing this vulnerable population. Kang et al[8] demonstrated, in a study utilizing the Taiwan National Health Research Institute database with matched controls, an adjusted hazard rate of stroke of 1.31 for the 12 months following infection. In a subsequent study utilizing a follow-on cohort from the same database with age and gender matched controls and adjusted for cardiovascular risks, Lin et al[15] found that patients with herpes zoster ophthalmicus had a 4.52-fold increase in stroke over a 1 year follow up but, importantly, patients with rheumatoid arthritis were excluded. A Danish study [10] analyzing 117,926 patients demonstrated an increased risk of stroke of 126% in the first 2 weeks from the onset of the infection and 17% from 2 weeks to one year. Langan et al [9] examined more than 6000 patients over the age of 18 with a history of HZ and stroke and compared incidence of stroke after HZ to other time periods. They found stroke increased after HZ with an age adjusted incidence rate of 1.63 over weeks 1–4, gradually decreasing over weeks 5–26. Similar to our study, a stronger effect was observed with herpes zoster ophthalmicus, rising to a >3 fold rate 5–12 weeks following infection. Utilizing an alternative approach, Breuer et al [7] performed a retrospective study of 106,601 HZ cases and 213,202 controls matched for age, sex, and general practice in the UK. Cox proportional hazard models were used[16] to adjust for multiple cardiovascular risk factors including smoking. Overall risk factors for vascular disease were increased in HZ patients but hazard ratios were also increased in the overall group for TIA and MI but not stroke. In a subset of subjects less than 40 years of age stroke was increased over the study period with an adjusted hazard ratio of 2.42. Three other recent studies of alternative design including one single health system analysis [12, 17] and two registry based cohort studies [11, 18] produced similar findings.

Our study has added to these recent investigations in several ways. First, ours is the first study to examine specifically an immunosuppressed patient population. This is particularly relevant given that such patients have been demonstrated to have an increased incidence of HZ by 1.5–2.0 over non immunosuppressed patient populations of similar age [2]. Secondly and consistent with several other studies [7–9, 15, 18] we have demonstrated that more serious and complex forms of HZ are attended by the highest rates of stroke. We performed this analysis by separating HZ by ICD code into ‘zoster without complications’ as well as ‘zoster with complications’. We further divided ‘zoster with complications’ into ‘zoster with cranial nerve complications’ (identified by specific ICD-9 codes) and ‘zoster with other complications’ (identified by any of the remaining 17 ICD-9 codes (Appendix Table 1). This latter group included mostly post herpetic neuralgia and other non-specific designations. The rationale for this segregation relates to the known capacity for varicella to establish latency in cranial nerve ganglia, especially cranial nerve V, and spread retrograde via sensory nerve fibers to the cerebral circulation [19].

Our adjusted analysis examined the effects of several important cardiovascular risk factors including age, sex, race, diabetes, HTN, atrial fibrillation, TIA and glucocorticoid use, which revealed minimal confounding effects consistent with previous studies [7–12]. We believe that the consistent findings of our and other recent studies identifying risk for stroke from HZ is even further strengthened by biologic plausibility given the capacity of the virus to replicate in arteries and produce disease [20]. Varicella has also long been associated with the rare occurrence of stroke in children in the time period following primary infection [21, 22]. In addition a rare syndrome of focal arteritis has been described in individuals with HZ opthalmicus in the weeks and months following infection; it has also been linked to a spectrum of vascular disorders involving the central nervous system found predominantly, but not exclusively, in immunosuppressed individuals [5, 23]. Finally, zoster has recently been identified in the temporal arteries of patients biopsied for suspected giant cell arteritis [24]. While these observations support a likely role for HZ in the development of stroke we have no pathologic data to help speculate on the immunopathogenic mechanisms involved, though a virally driven inflammatory and or pro-coagulant effect are possible. Further studies exploring pathogenic mechanisms are urgently needed given the implications of the large number of cases described and the implicit burden of morbidity and mortality linked to stroke.

Our study has several important limitations. We were not able to confirm the diagnosis of zoster from examination of medical records, although our criteria for the primary analysis (combining ICD-9 diagnosis plus antiviral therapy) has been shown to have a high positive predictive value (>85%). Another limitation in our study was that while we attempted to define the anatomic distribution and type of strokes in our population to determine if anatomic or pathologic trends could be identified, less than half had sufficient details reported for such analysis. We recognize the possibility that declining trends in stroke over time might be expected simply due to depletion of susceptibles, where the most ill patients are more likely to have serious adverse events (e.g. stroke) early. However, we would not expect that to be differential by zoster phenotype, nor by receipt of anti-viral treatment. Indeed, in the large subgroup of patients who had zoster without complications that received anti-viral therapy (n=403), there was essentially no time trend in the rate of stroke. Finally, while it would be desirable to examine the effects of HZ on stroke risk in younger patients (which has been demonstrated to be particularly elevated in at least one study [7] our analysis was limited to the age of Medicare beneficiaries (mean age 71 years). We therefore believe our findings are conservative given age-related risk factors for stroke in this cohort that would attenuate the associations of interest. In younger patients, we hypothesize that the HZ-related relative effect on stroke risk is greater than observed in this analysis.

Given the public health implications of these observations new urgency should be directed at increasing the administration rate of vaccination for zoster in vulnerable populations. The proportion of the older patients in our study who received vaccination was only 5% minimally different than in years past [25]. It is conceivable that an additional downstream benefit of preventing HZ may now include a lower risk of stroke, though this hypothesis remains speculative [25]. Recent studies have demonstrated that the risk of zoster is 1.5 to 2.0 times elevated in patients with autoimmune diseases versus older populations [2] affirming the vulnerability of patients with autoimmune inflammatory diseases. Additional analysis from this same study [2] identified clinically meaningful elevations of risk for HZ beginning at age 40 that were comparable or higher than the risk for HZ in healthy older people age 60+, implying that serious consideration should be given to recommending vaccination at a lower age than 60 which is the current ACIP-recommended minimum age for vaccination [26] or the FDA approved age of 50, as was recently recommended in the updated 2015 RA treatment guidelines [27]. In addition, many patients with autoimmune inflammatory diseases may have additional obstacles to vaccination including the use of biologic agents [27]. Thus the results of ongoing studies on the safety of the live zoster vaccine in the setting of TNF inhibitor therapy [28] and the further study of the safety and effectiveness of zoster vaccine(s) in high risk populations are clearly needed.