Herpes zoster, the reactivation of latent varicella-zoster virus infection, occurs commonly in HIV-infected patients and sometimes is the initial clue to the underlying immunodeficiency.1,2 Unlike most opportunistic infections, zoster occurs at a wide range of CD4 cell counts in HIV-infected patients.3,4 Several studies conducted prior to the availability of highly active antiretroviral therapy (HAART) suggested that herpes zoster incidence does not vary by level of HIV-induced immunosuppression,3,5,6 while other studies have found increased zoster incidence in patients with low CD4 cell counts.7,8
The impact of HAART on the epidemiology of herpes zoster is not well understood. Recent reports suggest that zoster incidence increases shortly after initiation of HAART and that it may be a consequence of immune reconstitution.9-11 These studies, however, did not examine incidence of zoster after longer periods of HAART administration nor did they compare this incidence to that in persons without HIV infection.
To describe the long-term epidemiology of herpes zoster in the current era of treating HIV with HAART, we evaluated self-reported zoster in HIV-infected and uninfected women participating in the prospective, multicenter Women's Interagency HIV Cohort Study (WIHS). We hypothesized that the risk of zoster in HIV-infected women, even with longer-term use of HAART, would remain higher than the risk in HIV-uninfected women.
The WIHS enrolled HIV-infected and at-risk HIV-uninfected women at 6 sites (Manhattan/Bronx, NY; Brooklyn, NY; Washington, DC; northern California; Los Angeles/southern California; and Chicago, IL) from October 1994 through November 1995. Details of the cohort design and baseline characteristics of the subjects are published elsewhere.12 Women are seen at 6-month intervals (semesters) and undergo extensive interviews to ascertain clinical information, physical examinations, and collection of blood and other specimens. The present study included data through visit 16 (conducted April through September 2002).
At each study visit, women were asked: "Since your previous study visit, has a health care provider, either doctor, dentist, nurse practitioner, nurse, or physician's assistant, told you that you had shingles (herpes zoster)?" Those responding affirmatively were then asked: "How many different times in the past 6 months did you have this?" At the initial study visit, women were asked in a similar fashion about whether they were ever diagnosed with herpes zoster and whether they were diagnosed with it in the preceding 6 months.
Trained interviewers ascertained self-reported antiretroviral use in the period since the previous visit aided by photomedication cards. HAART was defined as one of the following combinations13: ≥2 nucleoside reverse transcriptase inhibitors (NRTIs), in combination with at least one protease inhibitor (PI) or one nonnucleoside reverse transcriptase inhibitor (NNRTI); one NRTI in combination with at least one PI and at least one NNRTI; one NRTI in combination with ritonavir plus saquinavir; or an abacavir-containing regimen of ≥3 NRTIs in the absence of both PIs and NNRTIs. Due to their antagonism, the combination of stavudine and zidovudine with either an NNRTI or PI was not considered to be a HAART regimen.
HIV-1 RNA in plasma was originally quantified for all participants using the isothermal nucleic acid sequence based amplification (NASBA) method (Organon Teknika Corp., Durham, NC) in laboratories that were certified by the National Institutes of Health, Virology Quality Assurance Laboratory proficiency testing program, with a lower limit of quantification 4000 copies/mL, and starting with visit 7 (October 1997) with Nuclisens, the more sensitive method of NASBA quantification (lower limit of detection 80 copies/mL). For all women who initiated HAART, HIV-1 RNA determinations were repeated using the more sensitive assay, if needed, at the visit just prior to that at which HAART use was first reported, and for all visits after HAART initiation. CD4 and CD8 lymphocyte subsets were quantified using standard flow cytometric methods in laboratories participating in the NIH/NIAID Flow Cytometry Quality Assessment Program.
We excluded from all analyses women who seroconverted to HIV after enrollment in WIHS (n = 12) and women who were recruited during a 2nd wave of recruitment at visits 15 and 16 whose follow-up was therefore limited. Data from visits that occurred ≤3 months or >12 months from the preceding visit were excluded. In exploratory analyses, we noted a dramatically higher rate of reporting of zoster at the initial study visit compared with subsequent study visits (ie, occurrence of zoster within 6 months of the initial study visit was reported by 126 [4.8%] of 2613 women). We attributed this presumed overreporting to the lack of a readily identifiable anchor in calendar time for the preceding 6-month interval at the initial visit, as opposed to a dramatically declining zoster incidence following study entry. Consequently, we began our analyses of the incidence of zoster at the 2nd study visit and excluded women who did not attend any visits beyond visit 1. Overall, our study population consisted of 1832 of the 2058 HIV-infected women and 489 of the 568 HIV-uninfected women enrolled in the original cohort.
We calculated the probability of reporting zoster during each (between visit) study semester for HIV-infected and uninfected women. Statistical analysis was done using SAS version 8.2 (SAS Institute, Carey, NC). We assessed the effects of covariates on zoster occurrence during the period immediately prior to the visit using generalized estimating equations (GEE)14 with a logit link function, fit using all visits of each person (the cluster) as separate observations. In these models, each visit at which zoster was reported was counted as a separate event for a given subject.
Predictor variables included age, log CD8 cell count, CD4 cell count, and log HIV RNA level from the visit. Although the distribution of CD4 and CD8 cell counts were both skewed, log transformation of CD4 cell counts worsened the fit of this covariate as assessed by χ2 value, whereas log transformation of CD8 cell counts improved the fit. Consequently, only CD8 cell counts were log transformed. Use of HAART was modeled in an intent-to-treat fashion such that once a woman initiated HAART she was assumed to have continued on HAART. The models also included a covariate for time interval between the prior and current visits to account for the expected increased or decreased probability of reporting zoster to the degree this time was extended or shortened over the target 6 months.
Separate models were constructed for the visits of all women, all HIV-infected women, and HIV-infected women following initiation of HAART. Final models were selected to maximize explanatory value and to include covariates for which there was biologic plausibility for an association with herpes zoster or potential as confounders. To determine if there was a change in risk of reporting zoster over calendar time, we constructed an additional GEE model with self-reported zoster as the dependent variable and a 3-level, categorical, indicator variable denoting whether the woman's visit was her visits 2nd-6th (reference category), 7th-11th, and 12th-16th scheduled WIHS visit.
The study population consisted of 2321 women who contributed 22,657 study visits. Of these, 1832 women (18,265 visits) were HIV-infected and 489 (4392 visits) were HIV-uninfected women. Table 1 summarizes the baseline demographic and clinical characteristics of the women. Of the HIV-infected women, 565 (30.9%) had reported a clinical AIDS-defining event at or prior to their initial WIHS study visit and an additional 482 (26.3%) women reported an AIDS-defining event during follow-up. Sixty-eight percent of the HIV-infected women initiated HAART during follow-up. The median CD4 and CD8 cell counts at the visit prior to initiating HAART were 270 cells/μ L (interquartile range 142-430; n = 992) and 771 cells/μL (interquartile range 524-1113; n = 992), respectively. The median HIV RNA level at the visit prior to initiating HAART was 4.04 log10 copies/mL (interquartile range 3.25-4.79; n = 960).
At the initial study visit, 109 (6.0%) HIV-infected women reported a history of herpes zoster compared with 3 (0.61%) HIV-uninfected women. During follow-up, 337 (18.4%) HIV-infected women reported at least 1 episode of herpes zoster compared with 7 (1.4%) HIV-uninfected women. Of women who reported zoster during follow-up, both HIV-infected and HIV-uninfected women reported a median number of episodes of 1, though 32.9% of the HIV-infected women reported ≥2 episodes compared with none of the HIV-uninfected women. Of the HIV-infected women, 109 who had not yet initiated HAART reported an initial on-study episode of zoster; of these, 75 initiated HAART at a subsequent visit. An additional 295 women reported an initial on-study zoster episode after having initiated HAART.
Of the 337 HIV-infected women who reported zoster during follow-up, demographic and lymphocyte subset data were available on 334 women at 1 visit or 2 visits prior to reporting zoster. Their median age was 40 years (interquartile range 34-45 years), 61.4% were white, and 38.6% were African American. The median CD4 cell count at the closest visit preceding the report of zoster was 262 cells/μL, (interquartile range 146-446 cells/μL) and the median CD8 cell count was 818 cells/μL (intequartile range 539-1132 cells/μL). Of the 7 HIV-uninfected women who reported zoster during follow-up, demographic data were available on all 7 and lymphocyte subset data were available on 6. Their median age was 37 years (interquartile range 33-41 years), 42.9% were white, and 57.1% were African American. The median CD4 cell count at the closest visit preceding the report of zoster was 763 cells/μL (interquartile range 614-987 cells/μL) and the median CD8 cell count was 448 cells/μL (intequartile range 268-474 cells/μL).
Among the HIV-infected women, the incidence of herpes zoster per study semester (ie, between semiannual study visits) varied by current CD4 cell count with the highest incidence at the lowest CD4 stratum. Specifically, the probability of reporting zoster since the prior study visit was 1.2% for women with CD4 cell count >750 cells/μL, 1.7% for CD4 500-749 cells/μL, 2.5% for CD4 350-499 cells/μL, 3.2% for CD4 200-349 cells/μL, and 4.2% for CD4 <200 cells/μL. At all CD4 cell count ranges, the probability of reporting zoster since the prior study visit among HIV-infected women was greater than in the HIV-uninfected women, for whom the probability was 0.14%. Figure 1 illustrates the odds ratios for reporting herpes zoster since the prior visit by CD4 cell stratum and HIV status, adjusted for semester on study. Using HIV-infected women with CD4 >750 cells/μL as the reference category, the odds ratios for HIV-infected women to report herpes zoster since the prior visit by CD4 cell count stratum were: 1.43 (95% CI 0.86-2.37) for CD4 500-749 cells/μL, 2.07 (95% CI 1.27-3.38) for CD4 350-499 cells/μL, 2.72 (95% CI 1.66-4.46) for CD4 200-349 cells/μL, and 3.16 (95% CI 1.92-5.18) for CD4 <200 cells/μL, compared with 0.11 (95% CI 0.046-0.26) for HIV-uninfected women.
In univariate analyses of all HIV-infected women adjusted for time on study (Table 2), higher CD4 (OR 0.87 per 100 cells/μL, 95% CI 0.83-0.91) and CD8 cell counts (OR 0.67 per log10 cells/μL, 95% CI 0.41-0.90) were each associated with decreased odds of reporting herpes zoster since the prior study visit. Higher HIV RNA level (OR 1.23 per log10 copies/mL, 95% CI 1.12-1.35) and increased time since previous study visit (OR 1.09 per month, 95% CI 1.03-1.15) were also associated with increased odds of reporting zoster. In multivariate analysis (Table 2), CD4 cell count (adjusted OR 0.90 per 100 cells/μL, 95% CI 0.84-0.94) and time since previous visit (adjusted OR 1.08 per month, 95% CI 1.02-1.14) remained independently associated with zoster. In multivariate analyses that adjusted for the effects of HAART on CD4 cell count and HIV RNA level, HAART use in and of itself was not associated with zoster (adjusted OR 1.05, 95% CI 0.79-1.39; P = 0.72). Since low CD4 count and high HIV RNA are indications for HAART use, it was impractical to obtain an "unadjusted" effect of HAART that would not be influenced by these variables.
In univariate analyses of women who had initiated HAART (Table 3), adjusted for time on study, higher CD4 cell count was associated with decreased odds of reporting herpes zoster since the prior study visit (OR 0.87 per 100 cells/μL, 95% CI 0.81-0.92). Higher HIV RNA level was associated with increased odds of reporting zoster (OR 1.23 per log10 copies/mL, 95% CI 1.09-1.39). There were trends for increased odds of reporting zoster in association with African American race (OR 1.34, 95% CI 0.98-1.84) and increased time since previous study visit (OR 1.08 per month, 95% CI 1.00-1.17). In multivariate analysis of women who had initiated HAART (Table 3), CD4 cell count was the only variable that remained statistically significantly associated with reporting zoster (adjusted OR 0.88 per 100 cells/μL, 95% CI 0.82-0.94, P = 0.0006). In a multivariate model that adjusted for time since previous study visit, the risk of zoster was not statistically different at the first visit after reporting HAART initiation compared with the last pre-HAART visit (adjusted OR 1.21, 95% CI 0.71-2.05; P = 0.48).
There was a trend toward decreased odds of reporting zoster at more recent study visits. The odds ratios for reporting zoster at visits 7-11 and 12-16, compared with visits 2-6, were 0.91 (95% CI 0.74-1.12; P = 0.36) and 0.81 (95% CI 0.65-1.01; P = 0.058), respectively.
We found a strong relationship between herpes zoster incidence and degree of immunosuppression, as indicated by CD4 cell count, among HIV-infected women in the WIHS cohort. Notably, even HIV-infected women at the highest CD4 cell count stratum (>750 cells/μL) were at a nearly 9-fold higher risk of developing zoster compared with HIV-uninfected women. HIV-infected women with the lowest CD4 cell counts (<200 cells/μL) were nearly 25-fold more likely than HIV-uninfected women to develop zoster. In multivariate analyses, both lower CD4 cell count and higher HIV RNA level were associated with zoster incidence when all HIV-infected women were evaluated, whereas only lower CD4 cell count remained associated with zoster in women who had initiated HAART, presumably due to the dramatic effect of HAART on reducing HIV RNA levels.
There are limited data on the incidence of herpes zoster in HIV-infected patients receiving longer-term HAART. Rastogi et al15 presented a retrospective analysis of herpes zoster at the Johns Hopkins HIV clinic. They found that the incidence of zoster was identical in the HAART era (1997-2001) compared with prior published data from their clinic from the pre-HAART era16 (3.2 per 100 person-years in each time period), although there was a trend for a decreased incidence of complications of zoster in the latter years of the study period. In our study, there was a trend for decreased reporting of zoster at the more recent study visits that was not statistically significant but could be attributable to a rising CD4 cell count in the HIV-infected cohort as a whole due to use of HAART. No data were collected on complications of herpes zoster in our cohort.
Our data are notable for the robust independent association of CD4 cell count with zoster incidence in HIV-infected women, including the subset who initiated HAART. Studies done prior to the availability of antiretroviral therapy or HAART yielded conflicting results as to whether the incidence of zoster varies by CD4 cell count. In a cohort of homosexual men in San Francisco, Buchbinder et al5 reported a significant increase in the incidence of herpes zoster associated with HIV seropositivity. However, there was no effect of the duration of HIV infection, suggesting a lack of association between zoster and degree of immunosuppression, at least in the pre-HAART era. It should be noted, though, that CD4 cell counts were not explicitly reported, so this parameter was not specifically addressed. Similarly, data from an urban Ugandan cohort showed no relationship between CD4 cell count and zoster incidence.6
In contrast, data from the Amsterdam Cohort showed an increasing relative risk of zoster with decreasing CD4 cell count.8 A combined analysis of data from the Multicenter Hemophilia Cohort Study and District of Columbia Gay Cohort Study found a relatively constant incidence of zoster with CD4 cell counts above >200 cells/μL but a sharp increase in risk with CD4 counts less than <200 cells/μL.7 The reason for these discrepant findings reported among these studies, including our own, is not clear. Since most of the prior studies have included only men, it is possible that an interaction between sex and CD4 cell count with regard to risk of zoster may be a factor; however, there is no obvious biologic explanation for such an interaction.
Two recent studies reported a short-term increased incidence of herpes zoster immediately after the initiation of HAART. Domingo et al11 described the occurrence of zoster in 24 cases from a cohort of 316 patients who initiated antiretroviral therapy in Spain during the period of 1997-1999. All cases occurred within 17 weeks of beginning HAART. Similarly, Martinez et al10 reported a high incidence of zoster among 193 patients with AIDS over short time periods following addition of a PI to a baseline regimen of ≥1 nucleoside analogues. Twelve of 14 patients who developed zoster during a median follow-up of 64 weeks had their episodes between 4 and 16 weeks after starting the PI. In contrast to these studies, we did not observe a short-term increase in zoster incidence at the first visit following initiation of HAART. Our ability to define clearly the incidence of zoster relative to initiation of HAART was limited, however, by the lack of knowledge of exact dates of HAART initiation and zoster occurrence. For example, a woman could have reported both an episode of zoster and initiation of HAART since her prior study visit, and we would not have known which occurred first.
Although the mechanism by which zoster may occur at an increased rate in association with initiation of HAART is not known, the reports cited above10,11 found relationships to CD8 cell counts. In the Domingo et al11 study, cases of zoster were compared with 96 controls without zoster from the same cohort matched on age, sex, HIV viral load, CD4 cell count, and duration of follow-up. In multivariate analysis, only increase in CD8 cell count 1 month after initiation of antiretroviral therapy was associated with the development of zoster (adjusted OR 1.3 per percentage increase in CD8 count; 95% CI 1.1-1.5). Similarly, in the multivariate analysis done in the Martinez et al study, the baseline percentage of CD8+ lymphocytes and the increase in percentage of CD8+ lymphocytes at 1 month were associated with zoster incidence, whereas baseline values of or changes in HIV RNA level and CD4 cell percentage or absolute counts were not.10 We found no relationship between CD8 cell count and zoster among HAART users, but there was a protective effect of CD8 count in the univariate analysis of all HIV-infected women. The latter association may have been confounded by HIV RNA levels and CD4 cell counts, since the relationship to CD8 cell count disappeared in the multivariate model after adjustment for HIV RNA and other covariates.
Our study has several important limitations. We relied on self-reported herpes zoster that was not confirmed by review of medical records, and thus misclassification of zoster events may have occurred. The fact that approximately one-third of the HIV-infected women in this study reported ≥2 episodes of zoster raises the question of whether episodes of herpes simplex virus reactivation could have been reported as herpes zoster. Our finding of an approximately 10-fold higher incidence among HIV-infected compared with HIV-uninfected women, however, is in the range reported by others in cohort studies of men5,8 and suggests that there was not differential reporting of zoster by HIV serostatus. Furthermore, a study in the elderly has validated self-reported zoster as accurate,17 though the incidence of herpes simplex was presumably low in this population. We were not able to assess the potential impact of use of antiherpes drugs on zoster incidence in this study, since lack of data on exact start dates of medications precluded us from distinguishing whether these drugs were prescribed to treat zoster or were being used for other purposes prior to an episode of zoster, such as herpes simplex virus treatment or prophylaxis. Since HAART use was modeled using an intent-to-treat approach to reduce bias, we were unable to assess the potential impact of stopping and restarting HAART on zoster incidence. We also had no data on the clinical manifestations and complications of zoster in the cohort and can make no inferences about the severity of zoster in relation to CD4 cell count and use of HAART. This limitation is important in light of other reports of greater severity of zoster and increased risk of dissemination in HIV-infected patients with low CD4 cell counts.16,18 Lastly, we were not able to examine which, if any, demographic or clinical factors were associated with herpes zoster among the HIV-uninfected controls due to the low incidence of zoster in this group.
Taken together, the available data suggest that herpes zoster occurs at all stages of HIV disease but at higher rates with more advanced immunosuppression. Unlike many opportunistic infections, herpes zoster is likely to remain a common complication of HIV disease even in those who are treated with HAART. Further study of the clinical manifestations of zoster in HIV-infected patients is indicated to elucidate whether the complication rate, which was reported to be high prior to the advent of HAART,16,18 is truly reduced in the era of more effective antiretroviral therapy.
Data in this manuscript were collected by the WIHS Collaborative Study Group with centers (principal investigators) at New York City/Bronx Consortium (Kathryn Anastos); Brooklyn, NY (Howard Minkoff); Washington, DC, Metropolitan Consortium (Mary Young); The Connie Wofsy Study Consortium of Northern California (Ruth Greenblatt); Los Angeles County/Southern California Consortium (Alexandra Levine); Chicago Consortium (Mardge Cohen); Data Coordinating Center (Alvaro Munoz). The authors thank the study participants for their dedication.
1. Tyndall MW, Nasio J, Agoki E, et al. Herpes zoster
as the initial presentation of human immunodeficiency virus type 1 infection in Kenya. Clin Infect Dis.
2. Colebunders R, Mann JM, Francis H, et al. Herpes zoster
in African patients: a clinical predictor of human immunodeficiency virus infection. J Infect Dis.
3. Rogues A-M, Dupon M, Ladner J, et al. Herpes zoster
and human immunodeficiency virus infection: a cohort study of 101 coinfected patients. [letter] J Infect Dis.
4. Glesby MJ, Moore RD, Chaisson RE. Zidovudine Epidemiology
Study Group. Herpes zoster
in patients with advanced human immunodeficiency virus infection treated with zidovudine. J Infect Dis.
5. Buchbinder SP, Katz MH, Hessol NA, et al. Herpes zoster
and human immunodeficiency virus infection. J Infect Dis.
6. Morgan D, Mahe C, Malamba S, et al. Herpes zoster
and HIV-1 infection in a rural Ugandan cohort. AIDS.
7. Engels EA, Rosenberg PS, Biggar RJ. Zoster incidence in human immunodeficiency virus-infected hemophiliacs and homosexual men, 1984-1997. District of Columbia Gay Cohort Study. Multicenter Hemophilia Cohort Study. J Infect Dis.
8. Veenstra J, Krol A, van Praag RME, et al. Herpes zoster
, immunological deterioration and disease progression in HIV-1 infection. AIDS.
9. Aldeen T, Hay P, Davidson F, et al. Herpes zoster
infection in HIV-seropositive patients associated with highly active antiretroviral therapy. [letter] AIDS.
10. Martinez E, Gatell J, Moran Y, et al. High incidence of herpes zoster
in patients with AIDS soon after therapy with protease inhibitors. Clin Infect Dis.
11. Domingo P, Torres OH, Ris J, et al. Herpes zoster
as an immune reconstitution disease after initiation of combination antiretroviral therapy in patients with human immunodeficiency virus type-1 infection. Am J Med.
12. Barkan SE, Melnick SL, Preston-Martin S, et al. The Women
' s Interagency HIV Study. WIHS Collaborative Study Group. Epidemiology.
13. Dybul M, Fauci AS, Bartlett JG, et al. Panel on Clinical Practices for Treatment of HIV Infection
. Guidelines for using antiretroviral agents among HIV-infected adults and adolescents. Ann Intern Med.
14. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics.
15. Rastogi R, Polydefkis MJ, Moore RD, et al. The incidence of herpes zoster
among HIV patients in the post-HAART era [abstract 690]. In: Program and Abstracts of the 10th Conference on Retroviruses and Opportunistic Infections; February 10-14, 2003; Boston, MA.
16. Glesby MJ, Moore RD, Chaisson RE. Clinical spectrum of herpes zoster
in adults infected with human immunodeficiency virus. Clin Infect Dis.
17. Schmader K, George LK, Newton R, et al. The accuracy of self-report of herpes zoster
. J Clin Epidemiol.
18. Veenstra J, van Praag RME, Krol A, et al. Complications of varicella zoster virus reactivation in HIV-infected homosexual men. AIDS.