We performed a literature search in PubMed, EMBASE, OVID Medline, Scopus, Clinicaltrials.gov, and Cochrane Library using HZ terms and immunocompromising conditions among studies published between January 1, 1993, and January 8, 2019 outlined in a supplemental table. References cited by the retrieved articles for were searched for additional references.

We included full-text, original research articles reporting data on HZ risk in at least one of five immunocompromised U.S. adult populations (hematopoietic stem cell transplant recipients [HCT], hematologic malignancies [HM], solid organ transplant [SOT], solid tumor malignancies [STM], or people living with HIV). Clinical trials of HZ vaccination were included even if they had study sites outside the US, and estimates of HZ risk among unvaccinated (placebo) groups. We excluded studies limited to children or the elderly (>65 years). We excluded abstracts, case reports, and animal or molecular studies. For studies in HIV patients, we excluded estimates from populations not receiving standard antiretroviral therapy (ART).

Reviewers assessed studies for eligibility and then abstracted information on setting, design, demographics, antiviral prophylaxis, and outcomes using a standardized abstraction worksheet.

We extracted outcome measures for HZ risk, and post-herpetic neuralgia (PHN), disseminated HZ, HZ ophthalmicus, and hospitalization. HZ risk was measured as either cumulative incidence (the number of cases reported during the specified follow-up period, as a percentage of all study participants) or incidence (cases per 1,000 person-years). We defined PHN as pain lasting >90 days after HZ and reported PHN as a proportion of total HZ cases. We relied on the authors’ definitions for disseminated disease, HZ ophthalmicus and when hospitalizations were due to HZ.

We evaluated study design, outcome definition, adequate adjustment of confounding factors, generalizability of the findings and risk of bias. Observational studies and open-label phase 1 and 2 clinical trials were evaluated for methodological issues specified by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) working group [15]. The risk of bias for clinical trials was assessed according to guidelines of the Cochrane Collaboration [16]. We rated a study low-risk of bias if it had 0 categories or 1 category with minor issues, moderate if 1 or 2 categories were selected, or high if 3 or more categories were selected or if 1 category had critical problems. Data abstraction and risk of bias assessments were performed by SM, AG, or KD; extracted data were assessed for accuracy by a second reviewer. All reviewers adjudicated any data discrepancies identified during the second review.

Of the 3,765 studies screened, 869 full-text articles and identified 57 studies were eligible for inclusion and data extraction (figure 1). Characteristics of these studies are summarized in the table 1[17–72]. Among the 57 articles abstracted, 19 were clinical trials and 38 observational. The median study size was 49 (interquartile range [IQR]=27–78) for clinical trials (excluding populations who received vaccine as an intervention) and 355 (IQR=122–1739) for observational studies. Five studies (9%) reported a median or average study population age ≥65 years old, while 25 studies (45%) reported a median or average study population age of ≤50 years.

We assessed each study for risk of bias and classified 23 studies as high, 19 moderate, and 15 low-risk. Subsequent description of outcomes and figures report estimates from the 34 studies we deemed to have low or moderate-risk of bias since these were more likely to report valid and generalizable estimates.

Most studies included (32/34) reported the total number of zoster cases that occurred during the study period (cumulative incidence). A total of 40 cumulative incidence estimates were reported: 16 for HCT; 11 for SOT; 5 for HM; 4 for STM; and 4 for HIV (figure 2)[18, 20, 22, 24–26, 29, 30, 32, 35–38, 40, 41, 49, 51, 52, 55, 58–60, 62, 64, 65, 67–73]. The cumulative incidence of HZ ranged from 0% to 41% (IQR=2–13%). Follow-up time reported as median, average, or maximum was categorized into <1 year, 1 to 2 years, or >2 years. The point estimates generally increased with increasing follow-up time: <1 year (median=3.6%, IQR=1–4%, n=7), 1 to 2 years (median=6%, IQR=2–13%, n=21), and >2 years (median=13.7%, IQR=8–19%, n=9). Among study populations with no reported HZ cases, 3/6 were from clinical trials with active follow-up [21, 55]. Interim cumulative incidence estimates when reported, showed that the risk of HZ is not constant over time, with the highest risk occurring in the first 2 years of follow-up[25, 41, 70].

The 6 highest estimates of cumulative incidence were reported in HCT recipients ([22, 25, 38, 41, 62, 71]. Among HCT recipients, the risk varied by transplant type; 12 articles reported cumulative incidence estimates for autologous HCT (median 18.6%; IQR=11–23%) and 5 estimates for allogeneic HCT (median 9%; IQR=4–25%)[22, 25, 30, 32, 38, 40, 41, 55, 58, 60, 62, 64, 65, 68, 71, 73]. All estimates for HM, STM, and HIV were below 10%, but estimates varied within the immunocompromising category. Among SOT recipients, the HZ risk ranged from 0 to 16% and varied depending on the organ transplanted (e.g. heart, kidney, liver, and other visceral transplants)[18, 37, 42, 49, 59, 67, 70]. The highest risk was reported for heart transplant recipients (16% and 11%). Most reported estimates were for kidney transplants (n=3, median 5%)[19, 38, 68].

Thirteen studies reported 15 incidence rates of HZ: 4 estimates for HCT; 2 for HM; 4 for SOT; 3 for STM; and 2 for HIV (figure 3)[18, 24, 29, 35, 36, 41, 42, 51, 59, 62, 69, 71–73]. The incidence rate of HZ ranged between 9 and 95/1,000 person-years (PY). Incidence rates exceeding 40/1,000 PY were reported for both HCT and HM populations with the highest estimates from placebo recipients in phase III clinical trial studies who underwent active follow-up [35, 71]. Other populations of immunocompromised patients had median HZ incidence estimates <30/1,000 PY. Two studies reported HZ incidence for more than one immunocompromising population. Chen et al. reported estimates for multiple immunocompromised populations and found that the incidence of HZ in HCT populations (43/1,000 PY) was nearly 3 times that of SOT or HIV populations (each 17/1,000 PY)[29]. Habel et al. reported an incidence of HZ in HM patients (31/1,000 PY) over two times higher than that for STM patients (12.3/1,000 PY)[36].

Only one study in our review, reported data that allowed for comparison of the risk for HZ by age group across immunocompromising groups. Here, the incidence (per 1,000 PY) for those aged 18–49 years versus 60–64 years was: HCT (40 vs 51), SOT (13 vs 20) HIV (18 vs 16) cancer (8 vs 13)[29].

Twelve studies reported 14 estimates of the proportion of HZ cases that developed PHN according to our definition of pain lasting >90 days after HZ[26, 29, 36, 40–42, 52, 59, 62, 68, 71, 73]. The risk of developing PHN ranged between 6% and 45% across the immunocompromising conditions. Within each of the immunocompromising categories there was heterogeneity between PHN estimates; HCT (range = 6–41%, n=6), SOT (range = 7–45%, n=3), HM (range = 6–40%, n=3), STM (9%, n=1) and HIV (6%, n=1). Although PHN estimates varied between studies, estimates within a single study were similar to one another; Chen et al. reported PHN estimates for HCT (10%), SOT (7%), and HIV (6%)[29], and Habel et al. reported PHN for STM (9%) and HM (6%) populations[36].

Eighteen studies reported 25 estimates for occurrence of disseminated HZ[18, 22, 24–26, 30, 32, 36, 40–42, 52, 60, 62, 64, 68, 71, 73]. The majority of estimates (n=18) were among HCT recipients and ranged from 0% to 32%, with a median of 3% (IQR=0.2–3.8%)[22, 25, 30, 32, 40, 41, 60, 62, 64, 68, 71, 73]. Two recent large randomized clinical vaccine trials in autologous HCT patients reported 2% and 4% disseminated HZ, respectively[71, 73]. Truong et al. retrospectively analyzed rates of HZ following autologous HCT and found that patients on the least stringent prophylaxis regimen (until neutrophil recovery ≥500/μL) had the highest rate of disseminated disease (4%) compared with none with prophylaxis of 6 months or longer[68, 73].

Four studies reported 7 estimates for ocular HZ complications which were ≤1%[32, 42, 62, 71]. Erard et al. retrospectively reviewed medical records of three HCT cohorts receiving prophylaxis regimens of increasing duration and reported no ocular complications (n=2,635)[32]. HZ-associated hospitalization was reported by two studies. Arness et al. reported that among 37 cases of HZ in kidney transplant patients, 7 (19%) individuals were hospitalized[18]. In a vaccine clinical trial, Winston et al. reported that among 113 individuals who developed HZ after HCT, 16 (14%) were hospitalized[71].

Seventeen studies conducted on HCT patients provided 34 estimates of cumulative incidence of HZ post-transplant[22, 25, 30, 32, 38, 40, 41, 55, 58, 60, 62, 64, 65, 68, 71–73]. The HZ risk estimates are plotted by follow-up time, prophylaxis duration, and population size (figure 4). Overall, the cumulative incidence of HZ increased with increased follow-up time. The duration of antiviral prophylaxis appears to modify risk of HZ; HCT patients who received short-term prophylaxis had the highest cumulative incidence of zoster at a given follow-up time. The effect of long-term prophylaxis against HZ persists at one year after prophylaxis is discontinued[25, 32, 40], however Boeckh et al. found that by 5 years post-transplant, the HZ risk of patients who received 12 months of prophylaxis was similar to that of patients who received no prophylaxis.