Prevalence of metabolic syndrome among HIV-positive and HIV-negative populations in sub-Saharan Africa—a systematic review and meta-analysis

Background Metabolic syndrome (MetS) is a constellation of conditions that increase the risk of cardiovascular diseases. It is an emerging concern in sub-Saharan African (SSA) countries, particularly because of an increasingly aging population and lifestyle changes. There is an increased risk of MetS and its components among people living with Human immune deficiency syndrome (HIV) individuals; however, the prevalence of metabolic syndrome in the SSA population and its differential contribution by HIV status is not yet established. This systematic review and meta-analysis were conducted to estimate the pooled prevalence of metabolic syndrome in people living with HIV and uninfected populations, its variation by sub-components. Methods We performed a comprehensive search on major databases—MEDLINE (PubMed), EBSCOhost, and Cochrane Database of Systematic Reviews and Web of sciences for original epidemiological research articles that compared proportions of the MetS and its subcomponents between people living with HIV and uninfected patients and published between January 1990–December 2017. The inclusion criteria were adults aged ≥ 18 years, with confirmed HIV status. We assessed the risk of bias using a prevalence studies tool, and random effect meta-analyses were used to compute the pooled overall prevalence. Results A total of four cross-sectional studies comprising 496 HIV uninfected and 731 infected participants were included in the meta-analysis. The overall prevalence of MetS among people living with HIV was 21.5% (95% CI 15.09–26.86) versus uninfected 12.0% (95% CI 5.00–21.00%), with substantial heterogeneity. The reported relative risk estimate for MetS among the two groups was twofold (RR 1.83, 95% CI 0.98–3.41), with an estimated predictive interval of 0.15 to 22.43 and P = 0.055 higher for the infected population. Hypertension was the most prevalent MetS sub-components, with diverse proportions of people living with HIV (5.2–50.0%) and uninfected (10.0–59.0%) populations. Conclusions The high range of MetS prevalence in the HIV-infected population compared to the uninfected population highlights the possible presence of HIV related drivers of MetS. Also, the reported high rate of MetS, irrespective of HIV status, indicates a major metabolic disorder epidemic that requires urgent prevention and management programs in SSA. Similarly, in the era of universal test and treat strategy among people living with HIV cohorts, routine check-up of MetS sub-components is required in HIV management as biomarkers. Systematic review registration PROSPERO CRD42016045727


Introduction
The problem of metabolic syndrome (MetS) has been the main scourge of high mortality and morbidity [1]. Globally, the prevalence of metabolic syndrome (MetS) is unknown [2], and country-specific prevalence varies with estimated prevalences in excess of 25.0% in developed countries [3][4][5]. In sub-Saharan Africa (SSA), the prevalence of metabolic syndrome is not well established. The growing burden of the global metabolic disorder is occurring at a time when SSA is experiencing an epidemiological transition, whereby the continent is affected by the dual burden of infectious and non-communicable diseases [6]. Further, SSA remains the worst affected region globally with over 25 million people living with HIV [7]. The coexistence of infectious diseases and non-communicable diseases is well documented in developed countries, and the intensity of this comorbidity is incomparable in SSA [8].
The global response to HIV has averted 30 million new infections and nearly 8 million AIDS-related deaths, as a result of antiretroviral therapy (ART) uptake [9]. This has resulted in an aging population of people living with HIV, living longer on ART and at greater risk of chronic diseases and metabolic disorders [10,11]. While the global focus is on preventing and managing HIV infections, less attention is on the metabolic impact of HIV infection and treatment on infected individuals. The global pooled prevalence of metabolic risk factors among people living with HIV range from 16.7 to 31.3% [12].
Increased risk of MetS and its subcomponents among people living with HIV individuals is well documented and attributed to HIV infection, antiretroviral therapy, and other related factors [13,14]. Of which is similar to the burden of MetS risk in the general population, as a result of associated modifiable risk factors [15,16]. However, little is known about the pooled prevalence of MetS and the prevalence difference among people living with HIV and uninfected population in SSA. Studies from developed countries predominantly report MetS among HIV-positive cohorts than in negative counterparts [15][16][17]. Studies suggest that MetS outcomes are much lower among people living with HIV compared to the general population, but ART-treated patients have a higher risk of metabolic complication [18]. Whether this risk is more or less among people infected with HIV compared to uninfected population remains controversial. This systematic review and meta-analysis was conducted to understand the burden of metabolic syndrome and its subcomponents among people living with HIV and uninfected population in SSA.

Outcome of interest
The primary outcome of this study was to compare the pooled prevalence of metabolic syndrome among people living with HIV and uninfected populations in SSA. The secondary aim was to compare metabolic syndrome subcomponents (namely visceral obesity, hypertension, diabetes, triglycerides, HDL cholesterol) among people living with HIV and uninfected populations in SSA.

Protocol and registration
A study protocol (published) was developed prior to the conduct of this review [19]. The protocol was registered in the PROSPERO international prospective register of systematic reviews (CRD42016045727). The protocol was amended by removing the aspect of co-morbid diabetes and hypertension components. The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines were followed [20,21].

Eligibility criteria
All studies (randomized control trials, cross-sectional, case-control, and cohort studies) among adults (18+) published from January 1990-December 2017 reporting the prevalence of metabolic syndrome and its subcomponents in people living with HIV and/or uninfected populations were considered for inclusion. We excluded studies that presented estimates from study participants with unknown HIV status. The full inclusion and exclusion criteria are presented in the aforementioned published protocol [19].
Search strategy and selection process MEDLINE (PubMed), EBSCOhost, and Cochrane Database of Systematic Reviews and Web of sciences databases were searched for papers published between January 1990 and December 2017. The search terms used a combination of relevant medical subject headings (MeSH) and database specific terms with an African search filter. The search strategy and the number of returned items are presented in Appendix 1 and 2. The reference lists of identified articles were traced through a web of science, and conference proceedings checked using the International AIDS society abstract archives. The titles of retrieved articles were examined to exclude ineligible articles. Given a large number of Francophile countries in SSA, this review included eligible studies published in French and were reviewed by a French-speaking reviewer (SM). The selection of studies was a multistep process with two reviewers (OO, SM) independently screening the abstract and full text for potential eligibility using the inclusion criteria, and discrepancies were resolved through arbitration with a third reviewer (BS). The flow diagram of the study selection and exclusion process is presented in Fig. 1. The interrater agreement for abstract and full-text screening was high at 90% and 100% respectively.

Assessment of the methodological quality and risk of bias of included studies
The methodological quality and risk of bias of selected articles were assessed using the Effective Public Health Practice Project/McMaster Evidence Review and Synthesis Centre Tool: Quality Assessment Tool for Quantitative Studies [22], and the risk of bias tool for prevalence studies [23]. A multistep process approach was employed by two reviewers (OO, SM), and the interrater agreement on quality and risk of bias were 80% and 90% respectively. A summary of the areas considered in the assessment of each domain is included in Appendix 1 and 2-the risk of bias and quality assessment of included studies.

Data item and collection process
Abstraction of meta-data from the included articles was performed using a specifically designed Google form, by two reviewers (OO and SM) independently. Information extracted included publication details,  population sampled and sample size, metabolic prevalence estimates, participant's characteristics, and HIV status. Prevalence figures and 95% confidence intervals (CIs) were extracted or calculated from the available data using the Clopper-Pearson method [24]. Certain authors were contacted for clarifications and/or further data requests, and if contacted three times and no response, the article concerned was removed from the included studies.

Analysis
Data were analyzed using Stata 13.0 (StataCorp. 2013. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP). Heterogeneity between study estimates was assessed using the I 2 statistic, i.e., the percentage of variation not due to sampling error across studies. An I 2 value above 50% indicates high heterogeneity. The meta-weighted prevalence (95% confidence interval (CI)) of metabolic syndrome among HIV-positive and HIV-negative populations was undertaken using a random effects model (to account for heterogeneity). For the studies with both HIV-positive and control HIV-negative population, we estimated a meta-weighted relative risk using a random effects approach. The influence of the estimates for all included studies was assessed by the level of risk of bias, quality of the study, study settings, and data collection methods. The risk of bias and study quality were classified as either low, moderate, or high, and we performed the Egger test (in addition to funnel plot for the HIV-positive only meta-analysis) to assess for potential publication/small-study bias. The association between HIV infection and MetS was estimated directly using the "metan" function in STATA, where the default is RR, i.e., pooled risk ratio. We also check for influential outlying studies using a random effects variance shift outlier model (RVSOM) for detecting and accommodating outliers in a meta-analysis [25]. In a meta-analysis of prevalence, when the estimate for a given study tends towards either 0.0% or 100.0%, the variance for that study moves towards zero, and as a result, its relative weight may be overestimated [26]. Thus, we transformed the prevalence estimates using the double arcsine method to correct for this potential discontinuity [26,27]. For data analysis, we merged the estimates of people living with HIV population by ATP III, IDF, JIS, and WHO definitions criteria (average prevalence across those with multiple definitions). The final pooled meta-estimates and 95% CIs were back-transformed for ease of interpretation. We were unable to stratify prevalence estimates by age, sex, and location with sufficient power, due to the limited number of studies (four) that compared estimates between HIV-negative and HIV-positive and resultant sample sizes.

Characteristics of included studies Study participation
The characteristics of the included studies are summarized in Table 1. Of the 18 studies included in this review, 4 studies [33,35,38,40] compared the prevalence of MetS among people living with HIV and uninfected populations while the other 14 presented MetS prevalence estimates among people living with HIV subjects only [28-32, 34, 36, 37, 39, 41-45].
In the "Methods" section of the included studies, the projected sample sizes were bigger than the actual number of participants analyzed and reported. The number of participants presented in the studies analysis ranged between 50 to 300 HIV-negative participants and 79 to 755 participants in people living with HIV. About 25% of the total study participant's samples were men, and the age of participants ranged from 18 to 70 years. In the 12 studies with data on the duration of ART, the duration ranged from 3 to 92 months.
The included studies applied various international criteria to diagnose MetS. Six studies defined metabolic syndrome using the International Diabetes Foundation (IDF 2005) criteria, four studies used the Third Report  Table 2.

Risk of bias
A summary of the risk of bias of the included articles is shown in Appendix 1 and 2. Sixteen studies (88%) were considered to be at low risk of bias while the remaining two studies were classified as medium risk of bias. None of the studies was classified as high risk of bias. The major risks of bias were the lack of presentation of the representativeness of the study sample in relation to the broader target population (n = 8 studies, 44%) and the lack of specification of a random selection of subjects (n = 11 studies, 61%).

Quality assessment
A summary of the quality assessment of the included studies shows that half of the studies (n = 9, 50%) were considered to be of low quality, six studies were considered to be of medium quality, and three studies were considered to be of high quality. Most of the studies were cross-sectional in design, and this was deemed the weakest identified quality domain during our assessment due to the associated temporality bias.
Similarly, the estimated prevalence using IDF definition among males ranged from (1/38; 3.1%) [28] in Benin to ( Notable, MetS were shown to be consistently more prevalent in female than male across the two criteria. There is a relative estimate of 12.7% was among the female against 3.6% using IDF criteria, and ATP III definition estimated MetS prevalence at 19.7% among females and 15.7% in males respectively. The sex variation in MetS prevalence among HIV-negative cohort could not be ascertained because it was not reported.   [39] study was a prominent and influential outlier in our study. Thus, further meta-regression analysis to identify the source of heterogeneity was performed and also suggests that risk of bias score and year were potential sources of heterogeneity. However, a meta-regression model including year and risk of bias score and excluding the aforementioned influential outlier study only reduced the adjusted I 2 to~65%. Hence, there was still a large residual heterogeneity which is a potential limitation of the underlying data.

MetS prevalence overview
To our knowledge, this is the first systematic review and meta-analysis of the MetS prevalence in the sub-Saharan African population by HIV status. Notably, the total prevalence of MetS among people living with HIV population was significantly higher at 21.5% (95% CI 15.09 to 26.86) with estimated predictive interval of (0.02 to 0.52) irrespective of the definition criteria compared with their HIV-negative counterparts at 12.0% (95% CI 5-21%). However, the wide prevalence range indicates substantial heterogeneities and this is as a result of influential outlying estimates from one of the included study.   As mentioned above this review suggests a twofold higher risk of MetS (RR 1.83, 95% CI 0.98-3.41), with an estimated predictive interval of (0.15 to 22.43) P value = 0.055 among people living with HIV subjects compared to their HIV-negative counterparts, and this ratio was not statistically significant. This finding suggests HIV infection and ART appear to contribute to a significant excess burden of MetS over and above the contribution of traditional lifestyle-related risk factors. The findings of this review are somewhat related to ones discussed in other studies [46][47][48]. The meta-prevalence of MetS among people living with HIV populations in SSA countries has shown to be higher when compared to reported estimates from developed countries [18,49]. Similarly, the estimate of MetS among the uninfected population in this review is similar to the AGEhIV cohort study in the Netherlands, underpinning the significance of excess MetS risk among people living with HIV compared to HIV-negative [50]. Irrespective of the risk differences of MetS among people living with HIV and uninfected patients in this review not being significant, it is notable that the prevalence of MetS in both cohorts are high and ranged within the earmarked global burden. This implies that the burden of MetS is growing vehemently in SSA with or without HIV; hence, other related factors such as lifestyles, diets, aging, and other interlinked factors require crucial prevention and management beyond HIV. Across the major criteria (ATP 2001, IDF 2005, JIS) used by most studies included in this review, the estimated MetS prevalence was highest based on the IDF 2005 definition (25.7%). This was different from a similar review that found higher estimates based on the ATP definition criteria [12]. This implies a large waist circumference band among the infected cohorts included in this review, especially females, as this is a compulsory requirement using IDF definition. This agrees with the finding underlining high adiposity (based on body mass index) and waist circumference among people living with HIV subjects [18,51]. Further research is needed to understand the difference between waist circumferences by HIV status, as this was not established in this review.
The analysis of the meta-prevalence of MetS individual subcomponents was limited given the lack of adequate reporting in the included studies. Diabetes and hypertension were the most reported sub-components, but among people living with HIV cohorts. Nonetheless, studies have described the outcome of the high prevalence of hypertension and diabetes among people living with HIV populations [52,53]. Further research is required to understand which of the sub-component is the most prevalent and contributes to the development of MetS by HIV status.
With the widespread ART uptake and the introduction of treat, all strategy irrespective of viral load and CD4 count, we anticipated an increase in the burden of MetS irrespective of age and sex in SSA. Despite an unbalance representation of people living with HIV studies against uninfected groups in this review, the prevalence of MetS was essentially higher among the infected population. The absence of study conducted specifically on populations with confirmed negative HIV status might have accounted to these findings. Otherwise, it may be the true reflection of the ubiquitous MetS burden in sub-Saharan Africa irrespective of HIV status, as a result of improved life expectancy, globalization, and lifestyles. Thus, a scale-up in the awareness, prevention, and management of metabolic disorder is directly needed in this continent, to curb the emerging epidemic.

Strengths and limitations
Strengths This review strictly adhered to the PRISMA guidelines to maximize the robustness and rigor of the employed methodology. We also conducted a very rigorous quality and risk of bias assessment. To our knowledge, this is the first systematic review to attempt to compare the burden of MetS by HIV status. Furthermore, we did not exclude French-based studies given the region of focus in this review and a large number of Francophile countries with HIV burden.

Limitations
Our findings may not be generalizable to all people living with HIV and uninfected individuals given the small number of included studies and potential non-representativeness. Given the varied methodological designs of the studies included in the review, the calculation of the pooled prevalence estimate may have been affected by this heterogeneity, as suggested by the very high I 2 statistic of 96.05% based on the final eligible pool of studies included in this meta-analysis. Some of the included study's authors were contacted for raw data and further clarifications; however, some estimates had to be calculated from data provided in the article. One of the included studies among people living with HIV cohort was a randomized control trial study, but we do not suspect that the inclusion and exclusion criteria of the trial participants would introduce substantial bias to our study, as these are not related to the metabolic risk factors/conditions we are attempting to assess among adults living with HIV. The pooled risk ratio of the burden of metabolic syndrome among HIV-infected population compared to their negative counterparts should be interpreted with caution given that only 4 out of 18 studies compared MetS by HIV status. Furthermore, limited comparison of pooled prevalence of MetS combining all 18 studies versus meta-estimate among combined HIV-negative subjects in the 4 studies (and relatively small pooled sample size of HIV-negative subjects) can be done as the underlying epidemic profile varies tremendously across the region. A further limitation was the lack of standardized MetS definition employed across the various study settings.
Another limitation of this review was the limited subgroup stratification of MetS burden by sex and ART regimen among the eligible studies. However, the association of antiretroviral therapy with MetS has been previously documented [12,54]. Also, the relationship between the use of ART and naïve with MetS prevalence could not be ascertained in this review, as 16 out of 18 studies on HIV-infected subjects reported that the participants were ART-experienced. This may explain the high prevalence of MetS quantified in this review. It is thus important to include routine and regular metabolic disorder check in the routine follow-ups of people living with HIV to optimize prevention and management, especially in the era of treat-all.
Thus, further research is required in estimating the extent and association of HIV status and metabolic syndrome and/or its subcomponents within SSA. Moreover, a more standardized approach of classifying MetS in SSA should be adopted to allow better comparability across countries in the region and whether conventional waist circumference cutoffs are appropriate in the Africa context.

Conclusion and implications
MetS prevalence in people living with HIV and uninfected individuals is high in sub-Saharan Africa; however, based on our findings, this appears to be a nonsignificant high prevalence comparing HIV-positive and HIV-negative group. However, this review needs to be interpreted with caution given the weaknesses alluded to above. More primary research is required in SSA to give a better understanding of the difference in the burden of metabolic syndrome in the context of high HIV burden. However, the findings of this review have implications for public health practice and policymakers within SSA as HIV-positive individual's life expectancy increases in the post ART rollout era and also in the context of an unfolding epidemiological transition where an increasing burden of non-communicable in the context of a high dual and persistent burden of infectious disease. Implementing an inclusive/integrated care plan for people living with HIV populations in the region is essential. This implies the increased presence of other healthcare needs beyond HIV and other communicable infection that might overburden our already overstrained healthcare systems. Early screening of metabolic syndrome subcomponents irrespective of HIV status to reduce future metabolic syndrome epidemic cost is important in the era of increased population aging and obesity, and this has been observed in SSA. OR Zambia OR Zimbabwe OR "Central Africa" OR "Central African" OR "West Africa" OR "West African" OR "Western Africa" OR "Western African" OR "East Africa" OR "East African" OR "Eastern Africa" OR "Eastern African" OR "South African" OR "Southern Africa" OR "Southern African" OR "sub Saharan Africa" OR "sub Saharan African" OR "sub Saharan Africa" OR "sub Saharan African" NOT "guinea pig" NOT "guinea pigs" NOT "aspergillus niger" ))))