Herpes Zoster and Exposure to the Varicella Zoster Virus in an Era of Varicella Vaccination

Abstract

Objectives. We performed a case–control study to determine if participants with herpes zoster had fewer contacts with persons with varicella or zoster, and with young children, to explore the hypothesis that exposure to persons with varicella zoster virus (VZV) results in “immune boosting.”

Methods. Participants were patients of the multispecialty Marshfield Clinic in Wisconsin. We identified patients aged 40 to 79 years with a new diagnosis of zoster from August 2000 to July 2005. We frequency matched control participants to case participants for age. We confirmed diagnoses by chart review and assessed exposures by interview.

Results. Interviews were completed by 633 of 902 eligible case participants (70.2%) and 655 of 1149 control participants (57.0%). The number of varicella contacts was not associated with zoster; there was no trend even at the highest exposure level (3 or more contacts). Similarly, there was no association with exposure to persons with zoster or to children, or with workplace exposures.

Conclusions. Although exposure to VZV in our study was relatively low, the absence of a relationship with zoster reflects the uncertain influence of varicella circulation on zoster epidemiology.

Herpes zoster, also known as shingles, is a condition caused by reactivation of the latent varicella zoster virus (VZV), which lies dormant in sensory ganglia after primary infection (i.e., chickenpox).¹ Clinically, the condition is usually characterized by a painful vesicular rash in a dermatomal pattern. A summary of several investigations estimates the annual number of zoster cases at 1 million after age adjustment to the 2000 US population.² Although the factors that allow virus reactivation in healthy individuals are not completely understood, it is believed that a decline in cell-mediated immunity plays a pivotal role.^3,4

Since licensure of the varicella vaccine in 1995, there has been a dramatic decline in the number of chickenpox cases in the United States.⁵ However, there is concern that the relative absence of VZV circulating in the population may result in reduced immune boosting and an increase in the incidence of zoster.^6,7 Evidence to support this hypothesis was provided by a case–control study in the United Kingdom that found that contacts in the previous 10 years with varicella and with children reduced the risk of zoster.⁸ We describe a case–control study to test the hypothesis that persons with zoster have fewer contacts with persons with varicella or zoster, and fewer contacts with young children, compared with persons without zoster. The latter exposure is included because varicella is infectious before onset of the rash,⁹ and because some varicella probably occurs among vaccinated children. These breakthrough cases are relatively mild and the diagnosis may be delayed or missed entirely.^10,11

METHODS

Participants in this study were from the mostly rural patient population of the Marshfield Clinic, a multispecialty clinic network in central Wisconsin. Participants were either residents of the Marshfield Epidemiologic Study Area or they were enrolled in Security Health Plan and had received care at Marshfield Clinic. The Marshfield Epidemiologic Study Area is a geographically defined region surrounding Marshfield and selected regional centers.^12,13 Security Health Plan is a managed care organization owned by Marshfield Clinic. A comprehensive electronic medical record documents nearly every aspect of care provided to Marshfield Clinic patients.

Case Participants

To identify presumptive cases of herpes zoster from the electronic medical record we used International Classification of Diseases, Ninth Revision (ICD-9)¹⁴ codes (053–053.99) assigned during ambulatory, urgent care, or emergency department visits, or hospitalizations. The case accrual period was August 2000 to July 2005. We considered a herpes zoster diagnosis to be presumptive if there was no other zoster diagnosis for at least 6 months before the first diagnosis in the accrual period. We used 6 months to avoid identifying follow-up visits as incident cases of zoster rather than to rule out recurrent cases of zoster, which are rare. Additional eligibility requirements included age between 40 and 79 years as of their reference date and at least 6 months' residency in the Marshfield Epidemiologic Study Area or enrollment in Security Health Plan prior to their reference date. For this analysis, we defined reference date as the date of zoster onset for case participants and a randomly assigned date during the study period for control participants.

We confirmed a presumptive case of zoster if there was a provider diagnosis of zoster and no other diagnosis considered to be as likely or more likely based on the medical record. Trained abstractors reviewed the medical record for at least 6 months before and 3 months after the reference date. The zoster onset date was the earliest of either the date of the initial clinical encounter or the date of onset as reported in the clinical narrative.

We excluded persons from the study if at the time of sampling they were deceased or institutionalized, if they had difficulties with communication or comprehension, or if during the 6 months before the reference date they were likely to have immunosuppression caused by disease (e.g., cancer) or drugs (e.g., chemotherapy).

Control Participants

We selected potential control participants on the basis of incidence density sampling as described elsewhere.^15,16 After we created a roster of potential control participants from the membership files of the Marshfield Epidemiologic Study Area and Security Health Plan, we frequency matched control participants to case participants by randomly selecting control participants from within case strata defined by 1-year age groups and year of zoster onset. Control participants were subjected to the same screening process (including medical record review) and eligibility criteria as were the case participants, but the criteria referred to a randomly assigned date (in lieu of an onset date for disease) during the case accrual period. By definition, control participants had no zoster diagnosis before but could have had a diagnosis after their reference date. We excluded control participants with a self-reported history of zoster before the reference date.

Variables

After being mailed an introductory letter, eligible case participants and control participants were contacted by telephone (10 attempts or fewer). The median interval between zoster onset and interview was about 2 years and ranged between 2 months and 5 years. Interviewers asked participants to estimate the number of persons with varicella (or zoster) with whom they had contact in the 10 years before their reference date and to categorize such exposures as household, family members (nonhousehold), friends or social, or occupational. In addition, because unrecognized varicella was likely to be an important source of VZV, we asked about contacts with children aged 1 to 10 years in the same time period. We specifically inquired about exposures in the household, with nonhousehold family members (e.g., grandchildren), and in social settings (e.g., friends). We asked participants for the number of children, and for each child we asked for their age, duration of contact, and the child's first name to facilitate the participant's recall of events. Exposures to children were restricted to those that were relatively regular (i.e., at least once per week, on average) during 1 or more intervals, which were specified by the participant. For each regular contact with a child, the specified starting and ending dates were recorded. If exposure to a child was discontinuous (e.g., lived with grandparents from June 2001 to January 2002, then again from June 2003 to September 2003) there would be multiple records with different starting and ending dates. These exposures were then aggregated in the analysis.

The interviewers asked similar questions about occupational exposures, which were grouped into teaching, health care, child care, and other occupations. For questions relating to social and occupational exposures, we defined contact as personal interactions at least once per week with children aged younger than 10 years with collection of starting and ending dates for each unique exposure.

For all exposures, interviewers sought to avoid collecting data in which the exposure was only very brief and superficial (i.e., casual contact). No explicit proximity or duration of contact was required, because we felt that respondents would not be able to recall exposures in such detail. However, the clear consensus from study staff was that the exposures were generally at least 30 minutes or longer, with participants in close contact with the children of interest or persons with varicella or zoster. During each interview, we used working tables that included names, dates, and types of relationships to track exposures and identify inconsistencies. To further facilitate recall of exposures, interviewers prompted participants with memory aid devices such as a calendar of significant events that occurred during the study period (e.g., presidential elections). We modeled our questions after those used in the UK investigation, which were generously provided by Sara Thomas.⁸ All interviews were recorded and a random sample of 10% was reviewed for quality control. The study hypothesis was unknown to interviewers and participants.

Analyses

We performed multivariable logistic regression to examine the association between potential risk factors and zoster. Quadratic age and dichotomous gender terms were forced into all models. Exposure to children in various settings (e.g., household) was represented in regression models by a dichotomous term indicating whether exposure had occurred and a linear term representing duration of exposure (i.e., person-time). Because the number of independent variables under consideration was relatively large, we screened variables for inclusion in the initial multivariable model; variables with P < .5 were retained. For person-time variables that met the screening criteria, we generated plots derived from generalized additive models¹⁷ to confirm that linear representation was adequate. Model reduction used a backward elimination procedure. A variable was retained if it achieved statistical significance (P ≤ .05) or if its exclusion altered the adjusted odds ratio for another variable by greater than 10%. We used SAS version 9.1 (SAS Institute Inc, Cary, NC) to perform analyses.

The primary exposure variables screened for inclusion in the multivariable models included the number of varicella and zoster contacts and contact (number and person-time) with children aged younger than 10 years. We categorized each exposure according to the setting in which it occurred: household members, nonhousehold family members, nonhousehold friends, and occupational.

We performed sample size calculations assuming that 30% of participants would be exposed to VZV. The exposure rate was based on Centers for Disease Control and Prevention estimates of exposures because of wild-type varicella and breakthrough disease. We found a sample size of 600 case participants and 600 control participants to have 80% power to detect protective odds ratios (ORs) of 0.72 or lower (α = 0.05; 1 sided).

RESULTS

Of the 1099 case participants with presumptive zoster identified, we eliminated 197 (17.9%) for various reasons (Table 1). Of the remaining 902 eligible case participants, 633 (70.2%) were interviewed, 187 (20.7%) refused, and 82 (9.1%) were unreachable. Of the 1404 presumptive control participants identified, we eliminated 255 (18.2%) for various reasons. Of the remaining 1149 eligible control participants, 655 (57.0%) were interviewed, 340 (29.6%) refused, and 154 (13.4%) were unreachable. There were no significant differences between case participants and control participants in demographic or socioeconomic characteristics (Table 2). Cases who agreed to participate in the study were slightly older (aged 58 versus 56 years; P = .04) and more likely to be female (59.2% versus 54.0%; P = .16) than were nonparticipants. Control participants and non-control participants showed similar characteristics.

TABLE 1.

Description of Herpes Zoster Case and Control Participants: Marshfield Clinic, Wisconsin, 2000–2005

	Case Participants (n = 1099), No. (%)	Control Participants (n = 1404), No. (%)	Total (n = 2503), No. (%)
Screened
Immunosuppressive condition	54 (4.9)	14 (1.0)	68 (2.7)
Questionable herpes zoster diagnosisa	35 (3.2)	NA	35 (1.4)
Prior herpes zoster diagnosis	27 (2.5)	8 (0.6)	35 (1.4)
Insufficient information in record	19 (1.7)	120 (8.5)	139 (5.6)
Cognitive or psychiatric condition	12 (1.1)	32 (2.3)	44 (1.8)
Speech or hearing impairment	5 (0.5)	12 (0.8)	17 (0.7)
Institutionalized	2 (0.2)	2 (0.1)	4 (0.2)
Deceased	6 (0.5)	2 (0.1)	8 (0.3)
Multiple exclusions	6 (0.5)	20 (1.4)	26 (1.0)
Interview not attemptedb	26 (2.4)	36 (2.6)	62 (2.5)
Other exclusions	5 (0.5)	9 (0.6)	14 (0.6)
Interviews attempted	902 (82.1)	1149 (81.8)	2051 (81.9)
Refusedc	187 (20.7)	340 (29.6)	527 (25.7)
Unreachablec	82 (9.1)	154 (13.4)	236 (11.5)
Interviews completedc	633 (70.2)	655 (57.0)	1288 (62.8)

Note. NA = not applicable.

^aThese cases of herpes zoster were questionable after chart review and these participants were determined not to have zoster after adjudication.

^bThese eligible participants were screened but not interviewed because recruitment goals had been achieved.

^cPercentages were based on participants with an interview attempted.

TABLE 2.

Characteristics of Interviewed Case Participants and Control Participants: Marshfield Clinic, Wisconsin, 2000–2005

	Case Participants (n = 633), No. (%) or Mean (SD)	Control Participants (n = 655), No. (%) or Mean (SD)	P
Age, y	61.1 (10.7)	61.2 (10.6)	.84
Women	375 (59.2)	361 (55.1)	.15
US born	625 (98.7)	649 (99.1)	.74
Live overseas ≥ 6 mo	90 (14.2)	103 (15.7)	.50
Non-Hispanic White	619 (97.8)	638 (97.4)	.79
≥ High school education	542 (85.9)	565 (86.4)	.86
Retired	211 (33.3)	219 (33.5)	.98
Income ≤ $40 000	306 (58.2)	307 (56.8)	.68
Cigarette smoker	79 (12.5)	73 (11.2)	.52

Note. Percentages based on varying denominators because of missing values.

Similar proportions of case participants and control participants reported contact with at least 1 person with varicella in the 10 years before the reference date (Table 3). Varicella exposures in specific settings (e.g., household) were also similar between case participants and control participants. Compared with varicella exposures, past exposures to persons with zoster were somewhat more frequent overall, but were also similar between case participants and control participants. As with varicella, case participants and control participants reported similar proportions of zoster exposures in specific settings. More than a third of both case participants and control participants reported at least some personal contact with children in the 10 years prior to the reference date, and the proportions were similar in every exposure category. Note that the numbers reporting the various types of contacts are less than the actual numbers of case participants and control participants because some participants could not recall. However, the proportion with missing data is similar for case participants and control participants within each type of contact.

TABLE 3.

Case and Control Participants Who Reported Contact With 1 or More Persons With Varicella or Herpes Zoster, or With Children Aged Younger Than 10 Years in the 10 Years Prior to the Reference Date: Marshfield Clinic, Wisconsin, 2000–2005

	Case Participants, No. or No. (%)	Control Participants, No. or No. (%)	P
Type of varicella contactsa	578	613
Any contact	127 (22.0)	119 (19.4)	.28
Household members	30 (5.2)	33 (5.4)	.89
Family members, nonhousehold	68 (11.8)	65 (10.7)	.54
Friends or social settings	13 (2.3)	19 (3.1)	.36
Workplace	37 (6.4)	33 (5.4)	.47
Type of herpes zoster contactsa	597	628
Any contact	166 (27.8)	193 (30.7)	.26
Household members	26 (4.4)	36 (5.7)	.27
Family members, nonhousehold	74 (12.4)	86 (13.8)	.48
Friends or social settings	37 (6.2)	46 (7.3)	.43
Workplace	46 (7.8)	48 (7.7)	.98
Type of contact with childrena	628	650
Any contact	226 (36.0)	250 (38.5)	.36
Household members	77 (12.2)	80 (12.3)	.95
Family members, nonhousehold	158 (25.1)	181 (27.9)	.25
Friends or social settings	41 (6.5)	48 (7.4)	.53
Workplace	105 (16.6)	111 (17.0)	.86

Note. Reference date is date of herpes zoster onset for case participants, and a randomly selected date for control participants.

^aThe sample sizes shown for case participants and control participants reflect the number of participants who knew about “any contact” with varicella or herpes zoster, or with children. For the initial herpes zoster exposure question, 36 (5.7%) case participants and 27 (4.1%) control participants (P < .24) were unable to provide an answer. For the initial varicella exposure question, 55 (8.7%) case participants and 42 (6.4%) control participants (P < .15) were unable to provide an answer. For the initial contact with children questions, 5 (0.8%) case participants and 5 (0.8%) control participants were unable to provide an answer. The sample sizes for the other specific types of contacts may vary if participants failed to recall occurrence of that type of contact.

Roughly similar numbers of case participants and control participants were exposed to 1, 2, and 3 or more persons with varicella (Table 4). Compared with participants with no contact and after adjustment for age and gender, the number of varicella contacts was not associated with zoster, even at a relatively high level of exposure (3 or more contacts; OR = 1.4; 95% confidence interval [CI] = 0.8, 2.3). Neither was there any trend in the risk of zoster related to the number of varicella contacts. When varicella exposure was restricted to household members, the OR of zoster associated with 3 or more contacts was 0.5 but had a wide CI (95% CI = 0.1, 2.8); only 6 participants reported that level of exposure. The ORs for 1 and 2 household varicella contacts were approximately 1.0.

TABLE 4.

Odds Ratios for Prior Exposure to Persons With Varicella or Herpes Zoster Over a 10-Year Period: Marshfield Clinic, Wisconsin, 2000–2005

	No. of Case Participantsa	No. of Control Participantsa	ORb (95% CI)
No. of varicella contacts in previous 10 y
0 (Ref)	451	494	1.00
1	45	39	1.25 (0.79, 1.98)
2	24	30	0.86 (0.49, 1.50)
3 or more	38	30	1.37 (0.82, 2.27)
No. of herpes zoster contacts in previous 10 y
0 (Ref)	431	435	1.00
1	112	137	0.82 (0.62, 1.09)
2	31	29	1.07 (0.63, 1.82)
3 or more	10	14	0.69 (0.30, 1.58)

Notes. CI = confidence interval; OR = odds ratio. Exposures included household, nonhousehold family, friends or social settings, and workplace.

^aSome participants could not provide an exact number of contacts (varicella, n = 40; herpes zoster, n = 26). These participants were excluded from this analysis.

^bAdjusted for age and gender.

To determine if more recent varicella exposures were protective, we reduced the exposure window from 10 years to the 5 years immediately before the reference date. The odds of zoster associated with 3 or more varicella contacts relative to those with no contacts was not significant (OR = 0.9; 95% CI = 0.4, 2.1). Ten case participants and 12 control participants had 3 or more contacts in the preceding 5 years. Contacts with zoster were also not significantly associated with case status regardless of the number (Table 4).

The number of children and the total person-time of exposure to children were not associated with zoster regardless of the setting or category (i.e., household, nonhousehold family members, and friends or social settings). For example, 45 case participants (7.2%) and 42 control participants (6.5%) reported at least some personal interaction with 2 or more children in the household during the 10 years before the reference date (P = .76). Relative to those with no such exposures, the ORs of zoster were 1.1 (95% CI = 0.6, 1.7). Incorporating a dichotomous variable indicating exposure and a continuous modifier variable representing the duration of exposure did not change this null association. When restricted to those aged 40 to 60 years, the OR ranged from 0.8 (95% CI = 0.5, 1.4) for exposure to 1 child in the household to 1.4 (95% CI = 0.6, 3.2) for exposure to 3 or more children. Similarly, there was no effect among participants aged older than 60 years.

A total of 216 participants (105 case participants, 111 control participants) reported at least some contact with children while at work. The median number of person-months of exposure to children was similar between case participants and control participants (118 vs 117). Approximately 32% of the occupations were in teaching, 22% were in health care, and 9% were in child care; the remaining occupations were unlikely to have significant contact with children. The distribution of job types between case participants and control participants was similar (P = .52). When occupation was entered into the model as a dichotomous variable indicating exposure and a continuous modifier variable representing duration of exposure, there was no association with zoster (P = .95).

DISCUSSION

We found no evidence of a relationship between herpes zoster and exposure to VZV in the previous 10 years. We also observed that exposure to VZV in our population was relatively low, a likely consequence of an effective varicella vaccination program that began in the United States more than 10 years ago.⁵ A national survey of persons aged 65 years and older similarly reported a low exposure rate and no protection against self-reported zoster.¹⁸ However, our results contrast with a study conducted in the United Kingdom⁸ that found significantly protective trends with both increasing numbers of varicella contacts and social contacts with children in the 10 years before zoster onset.

There are several possible explanations for the differences between the UK study and ours. The source populations for these 2 investigations differ markedly in the incidence of varicella. In 1996, varicella vaccine was recommended in the United States for children aged 12 to 18 months and older susceptible individuals.¹⁹ The vaccine was highly effective; as coverage rates increased to 76% by 2001,²⁰ the incidence of varicella decreased by more than 70%.^5,21 In contrast, the United Kingdom has no varicella vaccination program and its varicella incidence rate is consistent with that expected for temperate climates.^9,22

Although a direct comparison of the Marshfield study population with the UK study population⁸ is problematic, our results are consistent with a lower level of varicella exposure in the United States. A smaller proportion of both case participants and control participants in Marshfield reported 1 or more varicella contacts over 10 years compared with the UK study (case participants: 22.0% versus 26.6%; control participants: 19.4% versus 41.6%). The differences were even greater when one considered 3 or more varicella contacts, with 17% of UK control participants reporting such exposures compared with 4.6% of Marshfield control participants. In contrast, the proportion with 1 or more zoster contacts, which should not have been affected by the vaccination program, was similar for both case participants (27.8% for Marshfield versus 22.5% for United Kingdom) and control participants (30.7% for Marshfield versus 30.3% for United Kingdom). The UK study described a strong protective effect with 3 or more varicella contacts, but the effect of fewer contacts was minimally protective and nonsignificant.⁸ Our study showed no trend with the level of varicella exposure regardless of exposure category. Exposure to zoster, which has less potential for transmission than varicella, conferred protection in neither study.

It is possible that some of the varicella contacts reported by participants in our study were less likely to elicit a VZV-specific immune response. As the vaccination program has matured, the proportion of disease occurring in vaccinated persons (i.e., breakthrough varicella) has increased from 1% in 1996 to 60% in 2004.²³ Compared with varicella in unvaccinated persons, breakthrough varicella is milder and has a lower transmission rate.^10,11 It is also possible that some of the reported varicella was another rash illness (e.g., herpes simplex); as the prevalence of varicella decreases, the likelihood for a misdiagnosis increases.²⁴ In our study, a participant was considered a case if medical record review confirmed a physician's diagnosis. When we required case participants to have unilateral rash and pain noted in the record, the ORs changed only slightly, suggesting that case misclassification was minimal.

The minimum age of 40 years was selected for this study because the burden of zoster is largely shouldered by older adults.^3,25 However, younger adults are more likely to have contact with children and mount a more vigorous immune response.^26,27 When we repeated our analysis in participants aged 40 to 54 years, there was no evidence of a protective effect; the risk of zoster for participants with 3 or more varicella contacts was 1.0 (95% CI = 0.5, 2.1). We also restricted our analysis to varicella exposures occurring in the first 5 years of our exposure window, but found no evidence of protection. Nor was there any trend when we examined consecutive biyearly estimates. The UK study included persons aged as young as 16 years, but did not report age-specific effect estimates except to state that there was a protective effect among persons aged older than 60 years.⁸

On the basis of mathematical models, others have estimated the duration of immune boosting to be up to 20 years.⁶ The actual duration is unknown, but can be expected to wane over time; more recent exposures may have a stronger effect. When we restricted our analysis of varicella exposures to those occurring in the 5 years immediately preceding the reference date, there was no protective effect.

Our results were likely affected by limitations of participant recall. The interval between zoster onset and interview was much longer than in the UK study.⁸ We would expect less exposure misclassification if the interview were closer to disease onset, but this effect is attenuated when the period of interest is long.²⁸ Recall of events that occurred 10 years before zoster onset is another potential source of misclassification. However, it seems unlikely that it would entirely explain the differences with the UK study, which also asked participants about exposures in the prior 10 years.⁸ The fact that our rates of zoster exposure were similar to those in the UK study is reassuring. Recall for exposures in certain settings, such as household, would be expected to be better than for other settings, but the effect estimates were fairly uniform across the different settings. Recall may have varied by case–control status, but neither group was informed of the specific hypothesis being tested, and the sometimes lengthy period between disease onset and interview may have attenuated the differences in recall between case participants and control participants.

Selection bias may have occurred as the proportion of eligible control participants who agreed to participate was 57% compared with 70% for case participants. The mean age of control participants was slightly older than that of nonparticipants (58 versus 56 years; P = .001) and participants were more likely to be female (55.1% verus 47.4%; P = .01). However, case participants and non-case participants were similar by age and gender, and case participants and control participants were similar in all sociodemographic characteristics (Table 2). Thus, we do not believe our findings were influenced by selection bias.

There is widespread acceptance of the hypothesis that an adequate level of cell-mediated immune response is critical to preventing zoster,¹ but there is debate over the mechanism by which that immunity is maintained. One hypothesis holds that zoster is held in check by an immune response boosted following exposures to persons with varicella—also termed exogenous boosting.^6,8,29 The concern is that universal immunization would reduce exposure to varicella and lead to a zoster epidemic. A second hypothesis is that immunity against VZV is maintained by a clinically silent reactivation of latent virus—termed endogenous boosting.^30–32 In this case, zoster incidence should decrease because the incidence rate in those who have had varicella will not be affected by a declining varicella prevalence and because those vaccinated for varicella have a smaller risk of zoster.^33,34 The long latent period of zoster means that the question of whether it has been affected by the varicella vaccination program may not be known for years.

We conducted a case–control study 10 years after the start of the varicella vaccination program and observed no relationship between exposure to VZV and zoster. If exogenous boosting influences the incidence of zoster, perhaps it does so only after relatively frequent or intense exposure to VZV.^8,35,36 Given the significant morbidity attributable to zoster, ongoing monitoring is needed to assess the impact, if any, of varicella immunization on the epidemiology of zoster. The recently approved zoster vaccine offers the promise of reduced incidence and disease severity,³⁷ but its influence on the epidemiology of zoster remains to be seen.