HIV and human herpesvirus 8 co-infection across the globe: systematic review and meta-analysis

Abstract

HIV-infection is an important risk factor for developing Kaposi sarcoma (KS), but it is unclear whether HIV-positive persons are also at increased risk of co-infection with human herpesvirus 8 (HHV-8), the infectious cause of KS. We systematically searched literature up to 12/2012 and included studies reporting HHV-8 seroprevalence for HIV-positive and HIV-negative persons. We used random-effects meta-analysis to combine odds ratios (ORs) of the association between HIV and HHV-8 seropositivity and conducted random-effects meta-regression to identify sources of heterogeneity. We included 93 studies with 58,357 participants from 32 countries in sub-Saharan Africa, North and South America, Europe, Asia, and Australia. Overall, HIV-positive persons were more likely to be HHV-8 seropositive than HIV-negative persons (OR 1.99, 95% confidence interval [CI] 1.70 - 2.34) with considerable heterogeneity among studies (I² 84%).The association was strongest in men who have sex with men (MSM, OR 3.95, 95% CI 2.92 - 5.35), patients with hemophilia (OR 3.11, 95%CI 1.19 - 8.11) and children (OR 2.45, 95% CI 1.58 - 3.81), but weaker in heterosexuals who engage in low-risk (OR 1.42, 95% CI 1.16 - 1.74) or high-risk sexual behavior (OR 1.66, 95% CI 1.27 - 2.17), persons who inject drugs (OR 1.66, 95% CI 1.28-2.14), and pregnant women (OR 1.68, 95% CI 1.15 - 2.47), p-value for interaction <0.001. In conclusion, HIV-infection was associated with an increased HHV-8 seroprevalence in all population groups examined. A better understanding of HHV-8 transmission in different age and behavioral groups is needed to develop strategies to prevent HHV-8 transmission.

Introduction

HIV-infection is an important risk factor for developing Kaposi sarcoma (KS), but to date, it is unclear whether HIV-positive persons are also at increased risk of co-infection with human herpesvirus 8 (HHV-8), the infectious cause of KS. Seroepidemiological studies have produced conflicting results concerning the HIV/HHV-8 association, possibly due to the wide range of populations with diverse transmission patterns studied.

HIV is transmitted through unprotected anal or vaginal intercourse in sexually active adults, and vertically from mother-to-child in children. HHV-8 transmission routes are less clear, and both sexual and horizontal non-sexual HHV-8 transmission routes have been reported.¹ In sub-Saharan Africa, where HHV-8 prevalence reaches 50%,^2-4 HHV-8 is primarily transmitted during childhood from mother to child and between siblings.^5,6 In Europe and the United States the prevalence of HHV-8 is low in the general population (<4%)^7,8 and high in men who have sex with men (MSM).^9,10 In MSM HHV-8 seropositivity has been linked to sexual behavior, including the number of lifetime sexual partners, but the specific routes of transmission in these sexually active men are still unclear.^10,11 In heterosexual adults in HHV-8 endemic and non-endemic regions HHV-8 might be transmitted through sexual contacts, but results from different studies have been inconsistent.^7,12-14

We did a systematic review and meta-analysis to examine the association between HIV and HHV-8 seropositivity in different population groups and geographic regions.

Methods

Eligibility criteria

We included cross-sectional, cohort and case-control studies, as well as clinical trials that reported data on HHV-8 seropositivity on a total of at least 50 eligible HIV-positive and HIV-negative children or adults, published as full articles, short reports, letters or abstracts. We excluded items published before the first description of HHV-8 in 1994.¹⁵ HIV-positive and HIV-negative subjects had to be either recruited from the same source population, or matched for relevant confounding factors, and tested with the same HHV-8 assay. We excluded studies without HHV-8 positive participants in both the HIV-positive and HIV-negative groups.

Literature search

We searched MEDLINE and EMBASE up to December 21, 2012, without language restrictions. We combined keywords related to seropositivity and seroprevalence with keywords related to HHV-8 and Kaposi sarcoma associated herpes virus (see Appendix Box 1). We also screened reference lists of relevant papers and conference proceedings from the International Conference on Malignancies in AIDS and Other Acquired Immunodeficiencies (ICMAOI), the Conference on Retroviruses and Opportunistic Infections (CROI) and the annual meeting of the American Society of Clinical Oncology (ASCO). Where eligibility could not be determined from titles and abstracts, we retrieved full texts. Two out of three reviewers independently screened references for eligibility (ER, NW, ZH). When consensus could not be reached, a fourth reviewer (JB) was consulted.

Data extraction and quality assessment

We extracted study and participant characteristics, and assessed the study objective and primary outcome, the comparability of HIV-positive and HIV-negative persons, and the degree of adjustment for potential confounders. We extracted prevalence data, and unadjusted and confounder-adjusted odds ratios (ORs) for the association between HIV and HHV-8 seropositivity for each population group of interest: children, patients with hemophilia, pregnant women, persons who inject drugs (PWID), heterosexuals with low-risk or high-risk sexual behavior, and MSM. We extracted results as reported for different combined or single HHV-8 assays: latent or lytic antigens, enzyme immunoassay (EIA) or immunofluorescence assay (IFA). For study populations that were included in several publications, we chose the publication with the largest sample size. Data were extracted by one reviewer (ER, JB, NW, ZF, ZH) and checked for accuracy by a second reviewer (ER, JB, NW, ZF, ZH), using standardized, piloted data extraction forms. We then entered the data in duplicate into an electronic database created for this study.

Statistical analyses

We pooled the unadjusted and, where available, adjusted ORs of the association between HIV and HHV-8 seropositivity using random-effects meta-analysis. We measured between-study heterogeneity with the I-squared statistic,¹⁶ and conducted stratified analyses and random-effects meta-regression to identify sources of heterogeneity and potential effect modifiers.¹⁷ The following variables were considered: population group (children, patients with hemophilia, pregnant women, PWID, heterosexuals with low-risk or high-risk sexual behavior, and MSM), sex (men/boys versus women/girls), HHV-8 tests used (EIA versus IFA), HHV-8 antigens used (latent versus lytic), source of study data (cross-sectional study versus cohort study versus clinical trial versus case-control study), combined antiretroviral therapy (cART) period (pre-cART versus cART period), geographic region (sub-Saharan Africa versus outside Africa), and country income level (low- versus middle- versus high-income). We used the World Bank classification of Gross National Income per Capita 2013, dividing countries into low-income (per capita income of US$ 1045 or less per year), middle income (US$ 1046 to 12,745 per year), or high income (US$ 12,746 or more per year).¹⁸ We defined pre-cART period as sample collection occurring before 2004 in low- and middle-income countries and before 1996 in high-income countries. The group of low-risk heterosexuals included patients from general hospitals, blood donors, members of the general population, and factory workers. The group of heterosexuals with high-risk sexual behavior included patients attending sexually transmitted infections clinics, female sex workers, male clients of sex workers, and wives of HIV-positive men. For publications that reported results for several distinct population groups, we considered each population as a separate study. We selected the results from lytic over latent antigen tests, EIA over IFA, and from combined tests over single test if studies reported results for several HHV-8 testing strategies. We used meta-regression to examine differences in the strength of the association between HIV and HHV-8 seropositivity and funnel plots and a regression test for funnel plot asymmetry to assess potential publication and small study bias.¹⁹ Results of the meta-analysis are presented in forest plots. All data analyses were done in Stata version 12.1 (Stata Corporation, College Station, Texas, USA).

To assess the impact of confounder-adjustment on the effect estimate, we restricted the analyses to studies that reported confounder-adjusted estimates in a sensitivity analysis. In further sensitivity analyses we excluded studies where HIV-positive and HIV-negative persons had not been recruited from the same source population; we included studies that did not report results for a distinct population group; we selected results from latent over lytic antigen tests and IFA over EIA; and used a fixed-effects model to combine results.

Changes and amendments to the published protocol of this systematic review²⁰ are as follows: we excluded studies with less than 10 HIV-positive or HIV-negative persons, studies with no HHV-8 positive persons in both groups, and studies which did not present results for a distinct population group. We also included studies that recruited study participants from different sources and matched these participants for relevant confounding factors.

Results

Number of eligible, included and excluded studies

The literature searches identified 1,957 references in MEDLINE and 1,820 in EMBASE. After removing duplicates, we screened 2,760 references for eligibility. We excluded 1,572 irrelevant records based on titles and abstracts, and assessed full text papers for the remaining 1,188 references. Of these, 1,124 study reports were excluded for the reasons described in Figure 1, including: no comparison group (334 references); review articles, book chapters or editorials (250 references); studies where HHV-8 antibodies were measured in tissue/body fluids other than blood (167 references). By searching for related publications of eligible studies and screening reference lists we identified two additional eligible publications.^21,22 The 66 publications eligible for inclusion contained a total of 160 studies in different population groups, of which we excluded 67 studies mainly because they had no comparison group (40 studies), were restricted to patients with KS (10 studies), or had a sample size below 10 in the HIV-positive or HIV-negative group (5 studies). We therefore included 66 publications that described 93 relevant studies in our systematic review. These 93 studies represent the denominator of our meta-analysis.

Figure 1. Inclusion of studies (flow diagram).

The flow diagram shows the process of selecting eligible studies. HHV-8, human herpesvirus 8; KS, Kaposi sarcoma; PCR, polymerase chain reaction.

Characteristics of included studies

The 93 studies included 58,357 participants; the median number of participants per study was 287 (interquartile range [IQR] 120-556); for details see Table 1 and Appendix Table 1. Thirty-eight studies (41%) were conducted in sub-Saharan Africa and 55 studies (59%) in Europe, North and South America, Asia, and Australia. Study subjects were drawn from 32 different countries (13 sub-Saharan African and 19 non-sub-Saharan African countries). Five countries contributed about half of the studies: the USA (16 studies), Italy (10 studies), South Africa (8 studies), Uganda (7 studies), and Brazil (5 studies). Studies in sub-Saharan Africa mainly examined heterosexuals with low-risk sexual behavior (19 studies), pregnant women (9 studies), and children (6 studies). Outside Africa, the most commonly studied population groups were MSM (16 studies), heterosexuals with high-risk sexual behavior (15 studies), and PWID (10 studies). Two of the 10 PWID studies included up to 20% of persons who did not inject drugs.^22,23 Half of the studies (47 studies) used only one HHV-8 assay; the other half combined different HHV-8 assays. Sera were specifically collected for HHV-8 analyses in 24 studies; 63 studies used stored blood samples that had been collected for other purposes. In six studies it was not clear if blood samples were specifically collected for HHV-8 analyses or not. All HHV-8 seroprevalence data included in our meta-analysis were cross-sectional, however, in several instances the data were taken from other study design types, for example, cohort studies (28 studies), clinical trials (3 studies), or case-control studies (3 studies). The association between HIV and HHV-8 seropositivity was the primary outcome in one study.²⁴ In the other studies, HHV-8 seroprevalence (68 studies), risk factors for HHV-8 infection in general (40 studies), and HHV-8 transmission modalities (11 studies) were primary outcomes. Confounder-adjusted ORs were available for 24 studies. Sociodemographic, behavioral, or medical characteristics were reported separately for HIV-positive and HIV-negative persons in 27 studies. Sixty-six studies did not compare HIV-positive and HIV-negative groups.

Table 1.

Characteristics of included studies.

	Sub-Saharan Africa	Europe	North America	South America	Asia	Australia	All regions
All comparisons	38 (100%)	22 (100%)	20 (100%)	7 (100%)	5 (100%)	1 (100%)	93 (100%)
Population group
Children	6 (16%)	2 (9%)	0	3 (43%)	0	0	11 (12%)
Patients with hemophilia	0	3 (14%)	0	0	0	0	3 (3%)
Pregnant women	9 (24%)	0	1 (5%)	0	0	0	10 (11%)
Low-risk heterosexuals	19 (50%)	0	2 (10%)	1 (14%)	2 (40%)	0	24 (26%)
High-risk heterosexuals	4 (11%)	7 (32%)	5 (25%)	1 (14%)	2 (40%)	0	19 (20%)
Persons who inject drugs	0	6 (27%)	3 (15%)	0	1 (20%)	0	10 (11%)
Men who have sex with men	0	4 (18%)	9 (45%)	2 (29%)	0	1 (100%)	16 (17%)
Country income level*
Low	20 (53%)	0	0	0	1 (20%)	0	21 (23%)
Middle	18 (47%)	0	3 (15%)	7 (100%)	4 (80%)	0	32 (34%)
High	0	22 (100%)	17 (85%)	0	0	1 (100%)	40 (43%)
Sex
Men/boys	3 (8%)	8 (36%)	11 (55%)	3 (43%)	0	1 (100%)	26 (28%)
Women/girls	18 (47%)	4 (18%)	6 (30%)	0	1 (20%)	0	29 (31%)
Both sexes	14 (37%)	6 (27%)	2 (10%)	2 (29%)	4 (80%)	0	28 (30%)
Not reported	3 (8%)	4 (18%)	1 (5%)	2 (29%)	0	0	10 (11%)
HHV-8 tests used
EIA	16 (42%)	3 (14%)	8 (40%)	1 (14%)	3 (60%)	0	31 (33%)
IFA	14 (37%)	15 (68%)	9 (45%)	3 (43%)	1 (20%)	0	42 (45%)
EIA and IFA	8 (21%)	4(18%)	3 (15%)	3 (43%)	1 (20%)	1 (100%)	20 (22%)
HHV-8 antigens used for testing
Latent	5 (13%)	4 (18%)	4 (20%)	1 (14%)	0	0	14 (15%)
Lytic	21 (55%)	6 (27%)	9 (45%)	1 (14%)	1 (20%)	0	38 (41%)
Latent and lytic	12 (32%)	12 (55%)	7 (35%)	5 (71%)	4 (80%)	1 (100%)	41 (44%)
Source of study data
Cross-sectional study	23 (61%)	18 (82%)	2 (10%)	5 (71%)	4 (80%)	0	52 (56%)
Cohort study	6 (16%)	4 (18%)	15 (75%)	1 (14%)	1 (20%)	1 (100%)	28 (30%)
Clinical trial	3 (8%)	0	0	0	0	0	3 (3%)
Case -control study	3 (8%)	0	0	0	0	0	3 (3%)
Other/unclear	3 (8%)	0	3 (15%)	1 (14%)	0	0	7 (8%)
cART period**
Pre-cART period	25 (66%)	1 (5%)	6 (30%)	5 (71%)	1 (20%)	0	38 (41%)
cART period	6 (16%)	3 (14%)	6 (30%)	0	2 (40%)	1 (100%)	18 (19%)
Pre-cART and cART period	4 (11%)	12 (55%)	3 (15%)	0	0	0	19 (20%)
Unclear/unknown	3 (8%)	6 (27%)	5 (25%)	2 (29%)	2 (40%)	0	18 (19%)

^*as defined by the World Bank classification

^**Study samples collected in pre-cART period defined as before 1996 for high-income countries and before 2004 for low- and middle-income countries; cART period starts thereafter.

cART, combined antiretroviral therapy; EIA, enzyme immunoassay; HHV-8, human herpesvirus 8; IFA, immunofluorescence assay.

Main findings

HIV-positive persons were more likely to be HHV-8 seropositive than HIV-negative persons (OR 1.99, 95% confidence interval [CI] 1.70-2.34). Studies were highly heterogeneous (I² 84%, p<0.001), but there was no evidence for funnel plot asymmetry (Egger's test p=0.14), see Appendix Figure 1. The weighted median HHV-8 seroprevalence was 47% (IQR 26-68%) in HIV-positive populations and 24% (IQR 11-37%) in HIV-negative populations.

Differences in strength of association

Consistent with the overall pattern, HIV-infection was associated with an increased risk of HHV-8 seropositivity in all examined population groups (children, patients with hemophilia, pregnant women, heterosexuals with low-risk or high-risk sexual behavior, PWID, and MSM), see Figure 2 and Appendix Figure 2. However, the strength of the association varied between different population groups (p-value for interaction <0.001), see Table 2. The strongest association was found in MSM (OR 3.95, 95%CI 2.92-5.35), patients with hemophilia (OR 3.11, 95%CI 1.19-8.11), and children (OR 2.45, 95%CI 1.58-3.81). The association was weaker in heterosexuals with low-risk (OR 1.42, 95%CI 1.16-1.74) or high-risk sexual behavior (OR 1.66, 95% CI 1.27-2.17), PWID (OR 1.66, 95% CI 1.28-2.14), and pregnant women (OR 1.68, 95% CI 1.15-2.47). The association between HIV and HHV-8 seropositivity was similar in pregnant women (OR 1.68, 95% CI 1.15-2.47) and other women from low-risk groups (OR 1.55, 95% CI 1.10-2.18). There was some evidence for interaction with sex (p-value for interaction=0.001) but this disappeared when we excluded MSM (p-value for interaction=0.57). The association between HIV and HHV-8 seropositivity was less pronounced in studies from sub-Saharan Africa (OR 1.59, 95% CI 1.34-1.90) than in studies from outside Africa (OR 2.50, 95% CI 1.99-3.14; p-value for interaction=0.001). Similarly, the association was also weaker in low-income countries than in middle- and high-income countries (p-value for interaction=0.007). Heterogeneity between studies was not explained by the type of HHV-8 test(s) used (EIA or IFA; p-value for interaction=0.79), the HHV-8 antigens used for testing (latent or lytic; p-value for interaction=0.99), or source of study data (cross-sectional study, cohort study, clinical trial, or case-control study; p-value for interaction=0.36). The strength of the HIV/HHV-8 association was also not modified by the sample collection period (pre-cART or cART period; p-value for interaction=0.89).

Figure 2. Association between HIV and HHV-8 seropositivity in different population groups.

The center of the diamonds is the pooled point estimate of the studies included in each population group, and the width of the diamonds represents the 95% confidence interval (CI) of the pooled odds ratio.

Table 2.

Association between HIV and HHV-8 seropositivity, stratified by different explanatory variables.

		Studies (N)	Odds ratio (95% CI)	p-value for interaction
Overall		93	1.99 (1.70-2.34)
Population group	Children	11	2.45 (1.58-3.81)	<0.001
	Patients with hemophilia	3	3.11 (1.19-8.11)
	Pregnant women	10	1.68 (1.15-2.47)
	Low-risk heterosexual	24	1.42 (1.16-1.74)
	High-risk heterosexuals	19	1.66 (1.27-2.17)
	Persons who inject drugs	10	1.66 (1.28-2.14)
	Men who have sex with men	16	3.95 (2.92-5.35)
Region	Sub-Saharan Africa	38	1.59 (1.34-1.90)	<0.001
	Europe	22	2.06 (1.60-2.66)
	North America	20	2.70 (1.91-3.83)
	South America	7	4.33 (3.13-5.99)
	Asia	5	1.26 (0.75-2.14)
	Australia	1	8.60 (3.55-20.86)
Country income level*	Low	21	1.43 (1.12-1.82)	0.007
	Middle	32	1.98 (1.59-2.46)
	High	40	2.47 (1.90-3.21)
Sex	Men/boys	26	2.93 (2.12-4.06)	<0.001
	Women/girls	29	1.61 (1.32-1.96)
	Both sexes†	28	1.54 (1.20-1.98)
	Not reported†	10	3.62 (2.78-4.73)
HHV-8 tests used	EIA	31	2.09 (1.62-2.71)	0.79
	IFA	42	2.03 (1.65-2.50)
	EIA and IFA†	20	1.69 (1.16-2.47)
HHV-8 antigens used for testing	Latent	14	1.78 (1.21-2.64)	0.99
	Lytic	38	1.78 (1.49-2.13)
	Latent and lytic†	41	2.32 (1.80-2.98)
Source of study data	Cross-sectional study	52	1.93 (1.59-2.33)	0.36
	Cohort study	28	2.33 (1.72-3.17)
	Clinical trial	3	1.35 (0.77-2.38)
	Case-control study	3	1.94 (0.93-4.05)
	Other/unclear†	7	1.45 (0.68-3.10)
cART period**	Pre-cART period	38	1.87 (1.55-2.24)	0.89
	cART period	18	1.81 (1.29-2.53)
	Pre-cART and cART period†	19	2.33 (1.51-3.60)
	Unclear/unknown†	18	2.46 (1.64-3.69)

^*as defined by the World Bank classification

^**Study samples collected in pre-cART period defined as before 1996 for high-income countries and before 2004 for low- and middle-income countries; cART period starts thereafter.

^†excluded from test for interaction

cART, combined antiretroviral therapy; CI, confidence interval; EIA, enzyme immunoassay; HHV-8, human herpesvirus 8; IFA, immunofluorescence assay.

The interaction between population group and the HIV/HHV-8 association remained statistically significant (p-value for interaction <0.001) when we included geographic region (sub-Saharan Africa versus outside sub-Saharan Africa), or country income level (high versus middle versus low) in the regression model. In contrast, after adjustment for population group there was no longer evidence for a statistically significant interaction of geographic location or country income level with the HIV/HHV-8 association (p-values for interaction=0.44 and 0.52, respectively). The meta-regression coefficients and the modeled effect estimates of these analyses are shown in the Appendix Tables 2+3.

Sensitivity analyses

The overall association between HIV and HHV-8 seropositivity and the interaction with population group remained similar when we selected results from IFA over EIA, latent antigens over lytic antigens, or used fixed-effects instead of random-effects models for analyses (Appendix Table 4). Exclusion of four studies^25,26 that recruited HIV-positive and HIV-negative participants from different sources (but matched them for confounders), and inclusion of small and mixed study populations did also not affect the overall result or the interaction between population group and the HIV/HHV-8 association (Appendix Table 4). The pooled estimate in the 24 studies that reported confounder-adjusted ORs (OR 1.80, 95% CI 1.42-2.28) was slightly weaker than the pooled estimate in the same set of studies when we used unadjusted ORs (OR 1.88, 95% CI 1.48-2.39). Both pooled estimates were similar to the result from the main meta-analysis for which we pooled both confounder-adjusted and unadjusted ORs (OR 1.99, 95% CI 1.70-2.34). The number of studies with adjusted ORs per population group was too small to allow meaningful comparisons (Appendix Table 4). In the 24 studies that reported both confounder-adjusted and unadjusted odds ratios, the two estimates were mostly similar (Appendix Table 5). The variables used for adjustment in the individual studies are listed in the Appendix Table 1.

Discussion

Our meta-analysis shows that HIV-infection is associated with an increased HHV-8 seroprevalence. The association was strongest in MSM, patients with hemophilia, and children, but weaker in heterosexual adults with low-risk or high-risk sexual behavior, in pregnant women, and in PWID. There was some evidence for a geographic variation of the HIV/HHV-8 association in univariable analysis; however, this was mainly explained by the different population groups studied in these regions. When we adjusted for population group the effect of geographic region disappeared.

Our findings extend those of a systematic review of the HIV/HHV-8 association done for China,²⁷ which reported a pooled OR of 2.97 (95% CI 2.22-3.97). Our study was based on an extensive literature search that covered all regions and countries. We developed a study protocol²⁰ with rigorous inclusion criteria and assessed the quality of studies carefully. To reduce potential confounding and selection biases, we only included studies that recruited HIV-positive and HIV-negative participants from the same population and studies that recruited participants from different sources, but matched them for relevant confounding factors. Because we identified a large number of eligible studies, we could assess the association between HIV and HHV-8 seropositivity within different age and behavioral subgroups.

Our results are based on cross-sectional data, so we cannot determine the temporal sequence of infection with the two viruses. Also, most of the studies we included were not designed to assess the association between HIV and HHV-8 seropositivity. Their sampling methods and the comparability of HIV-positive and HIV-negative participants with respect to relevant confounders were often reported incompletely. This limited our ability to assess potential confounding and selection bias. However, in the studies that reported both adjusted and unadjusted effect estimates, adjustment for confounders had little effect on the association between HIV and HHV-8 seropositivity. We observed substantial heterogeneity between studies and explored sources of heterogeneity using meta-regression analysis. Findings from meta-regressions of aggregate data, however, have to be interpreted cautiously. For example, grouping studies into low-risk or high-risk sexual behavior categories will not be a perfect reflection of the risk status of the individual persons studied. High-risk behaviors may overlap across groups; for example injection drug use may be common among commercial sex workers. More detailed data on risk factors, such as number of sexual partners or histories of sexually transmitted infections stratified by HIV and HHV-8 infection status were available only for few studies. Ideally, analyses of the HIV/HHV-8 association should therefore be based on individual participant data, but such data were not available for the current meta-analysis. Only three studies in patients with hemophilia and no studies in MSM or PWID from sub-Saharan Africa were included in the analysis, limiting the precision of estimates for these groups. Because only one of the included studies reported whether HIV-positive participants received cART or not,²⁶ we could not assess whether cART modified the HIV/HHV-8 association. Finally, we will have missed studies that measured HHV-8 seropositivity in HIV-positive and HIV-negative persons but failed to report the results.

Because there is no gold standard for HHV-8 serological testing, we were confronted with different tests of varying sensitivity and specificity.^28,29 However, within each study the same HHV-8 test was used in HIV-positive and HIV-negative groups and the odds ratios will therefore be more comparable across studies than absolute measures, such as seroprevalence, would have been. Indeed, we found no statistical evidence for differences in odds ratios by type of HHV-8 test (IFA versus EIA) or antigens (latent versus lytic) used. Only two of the included studies reported data on HHV-8 DNA in peripheral blood mononuclear cells³⁰ or saliva³¹ for HIV-positive and HIV-negative persons. In both studies HHV-8 DNA was detected slightly more often in the HIV-positive group compared to the HIV-negative group, but these findings were not statistically significant.

Biological mechanisms may account for the positive association between HIV and HHV-8 seropositivity. These mechanisms include reactivation of pre-existing HHV-8 infections in HIV-positive persons,³² increased susceptibility for HHV-8 infection in the presence of HIV co-infection^9,33,34 or shared routes of transmission for HIV and HHV-8. Combined ART might influence the HIV/HHV-8 association by stimulating the immune response against HHV-8 and reducing HHV-8 viremia.^35,36 However, we found no evidence that the strength of the HIV/HHV-8 association varied between the pre-cART and the cART period. Reactivation and increased susceptibility are expected to lead to similar magnitudes of the association between HIV and HHV-8 seropositivity in different population groups. Conversely, shared transmission routes, for example through anal intercourse, would lead to stronger associations in MSM than in heterosexual populations, as observed in our study. Interestingly, associations were similar in high-risk and low-risk heterosexual populations, indicating that vaginal intercourse may be less important in this context. The finding of a strong association between HIV and HHV-8 in children could be explained by transmission through saliva: HHV-8 DNA is detected more frequently and at a higher viral load in saliva than in semen, or vaginal and prostatic secretions.^37,38 Of note, HIV- infection may increase HHV-8 shedding in different body fluids, including saliva^38-40 and semen.⁴¹ Further studies are needed to clarify what exposures may explain the strong association between HIV and HHV-8 seropositivity in MSM and children.^42,43

In conclusion, this large-scale systematic review and meta-analysis found that HIV-infection was associated with an increased HHV-8 seroprevalence in all population groups examined. The association was strongest in MSM and children. Shared routes of sexual transmission of HIV and HHV-8 in MSM, and shedding of HHV-8 in HIV-positive mothers could explain these associations. Interventions that prevent the spread of HHV-8 are likely to reduce the burden of KS, however, a better understanding of HHV-8 transmission in different age and behavioral groups will be required before such preventive strategies can be developed and evaluated.

What's new?

Human herpesvirus 8 (HHV-8) is the infectious cause of Kaposi sarcoma (KS) and HIV-infection is an important risk factor for developing KS. To date, it is unclear whether HIV-positive persons are at increased risk of co-infection with HHV-8. This large-scale systematic review and meta-analysis found that HIV-infection was associated with an increased HHV-8 seroprevalence in all population groups examined. The association was strongest in men who have sex with men and children.

Abbreviations used

CI

confidence interval

EIA

enzyme immunoassay

HHV-8

human herpesvirus 8

HIV

human immunodeficiency virus

IFA

immunofluorescence assay

IQR

interquartile range

KS

Kaposi sarcoma

MSM

men who have sex with men

OR

odds ratio

PWID

persons who inject drugs