Prevalence of HIV and hepatitis B coinfection in Ghana: a systematic review and meta-analysis

To identify relevant studies, comprehensive searches were conducted by RO in PubMed, Science direct, Google SCHOLAR, and Africa journals online (AJOL) databases. The key words used were Hepatitis B, Hepatitis B surface antigen (HBsAg), HBV, HBV-DNA, human immunodeficiency virus, HIV, AIDS, coinfection, Prevalence and Ghana. In many instances, a combination of these keywords were explored (See "Appendix" section). AA also searched the digital institutional repository of the two leading Ghanaian universities, University of Ghana Legon (http://​www.​ugspace.​ug.​edu.​gh/​) and Kwame Nkrumah University of Science and Technology (http://​www.​ir.​knust.​edu.​gh/​) to identify unpublished thesis related to the subject. The websites of the Ministry of Health (http://​www.​moh-ghana.​org/​) and the Ghana Health Service (http://​www.​ghanahealthservi​ce.​org/​) were also searched for related reports and reviews using selected phrases such as ‘Hepatitis B and HIV coinfection’. Online searches were conducted between 01/03/2016 and 12/03/2016. All references in selected articles were further screened for additional publications.

Studies were included only if they reported chronic HBV infection prevalence rate among HIV-infected persons in Ghana. As generally recommended, laboratory diagnosis of chronic HBV infection focuses on the detection of hepatitis B surface antigen (HBsAg) [15]. Hence, only studies reporting prevalence of HBV in HIV persons based on HBsAg seropositivity were included. Studies reporting HBV prevalence in general populations (HIV negative cohorts) or those presenting HIV/hepatitis C (HCV) coinfection prevalence were excluded. Only studies published in English were selected. For duplicate studies, the version published first or one with complete dataset were included.

Studies’ qualities were assessed using a 12-point scoring system based on the Downs and Black checklist as adopted in similar reviews [10, 16, 17]. These were: (objective of the study clearly described, study design clearly stated, participants representative of the population from which they were recruited, participants accrued during the same time period, modest sample size, management of missing data, age, gender and other characteristics explored/reported, e.g. were confounders reported, was detection method of HBV reported, were potential biases reported, was outcome clearly described?), the assessment also included other items known to be associated with study quality [10, 17]. Each study was issued with a unique number for identification purposes and the following descriptive information collected; author details, year of publication, region of Ghana, type of study population, mean age of subjects, number of subjects involved (sample size), setting (rural vs urban), gender of study participants and the HIV/HBV coinfection prevalence rate. Data were independently extracted by RO and AA and compared. Any disagreements or discrepancies were resolved by consensus-based discussions.

We analyzed the results by meta-analysis proportions performed with OpenMeta (analyst) software, an open-source, cross-platform software for advanced meta-analysis [18] and StatsDirect statistical software (Version 3.0.0, StatsDirect Ltd, Cheshire UK) [19]. Individual study proportions were assessed at 95 % confidence interval (CI) as well as the pooled effect. Between-study heterogeneity was assessed by the Quoran (Q) statistic test and the I ² statistic, which represents the percentage of total variation across studies, attributable to heterogeneity rather than to chance [20]. A p value of <0.1 was considered to be statistically significant for the Q-statistics test and an I² >50 % was deemed to represent meaningful heterogeneity in which case the random effect model (DerSimonian-Laird) was adopted over fixed effect model in the summary of pooled analysis [20]. To assess the publication bias and small-study bias, a funnel plot of the data was applied. In addition, Egger and Begg’s tests were used to detect publication bias [21, 22]. A leave-one-out sensitivity analysis was performed by iteratively removing one study at a time to confirm how each individual study affects the overall estimate of the rest of the studies [23]. For all computations except the between-study heterogeneity testing, statistical significance was set at p < 0.05.

An ethical approval was not required for this study as it was based on data/information retrieved from published studies already available in the public domain.

Figure 1 outlines the articles’ search and retrieval steps. A total of 1220 citations were identified through electronic search and other sources. After the exclusion of duplicates and irrelevant studies based on titles and abstracts, fourteen (14) articles were retrieved for detailed full-text analysis. Out of the 14 studies, twelve (12) met the inclusion criteria for addition to the review [24‐35]. The 12 studies (Table 1) reported HBV coinfection prevalence rate in a total HIV-positive patient’s population size of 8162. The sample size (number of HIV infected patients) across the 12 studies ranged from 12 to 3108. The studies were conducted across seven (7) of the ten (10) regions of Ghana. The regional breakdown of the studies were as follows; Ashanti (n = 3), Brong-Ahafo (1), Eastern (2), Western (1), Northern (1), Greater Accra (3) and one study that involved multiple regions (Greater Accra, Ashanti, Eastern and Volta). Fifty-eight percent (7/12) of studies were conducted among HIV patients visiting HIV clinics whereas 17 % (2/12) were conducted among blood donors with same proportion (17 %, 2/12) of studies being carried out in pregnant women attending antenatal clinics. The oldest identified study was published in 1999 [24] and the most recent study identified was published in 2016 [35]. Ninety-two percent (92 %) of studies were published within the last decade (2006–2016) and as much as seventy-five percent (75 %) of studies were published within the last five (5) years (2011–2016). In studies that presented gender distribution, female participants were of higher proportion. The overall quality grading identified 59, 33 and 8 % of studies included in the review to be of high, moderate and low quality respectively.

A total of 8162 HIV infected patients were involved in the 12 studies included in this review. The pooled HIV/HBV coinfection prevalence rate (Fig. 2) across the twelve (12) studies published between 1999 and 2016 in Ghana was 13.6 % (95 % CI 10.3–16.8 %; p < 0.001). Heterogeneity Chi squared (Q) was 161.7 (degree of freedom, d.f = 11), and I² was determined as 93.2 % for the degree of inconsistency. The estimate of between-study variance Tau-squared was 0.002.

A funnel plot of HBV/HIV coinfection prevalence rates did not reveal a completely symmetrical display of the prevalence rates reported by the various studies (Fig. 3). However, we did not deduce any strong evidence of publication bias as revealed by Egger’s (p = 0.1604) and Begg’s (p = 0.7373) tests.

To assess the robustness of the HIV/HBV coinfection prevalence results, we performed a leave-one-out sensitivity analysis by iteratively removing one study at a time while recalculating the coinfection prevalence rate. The weight of the individual summaries on the pooled effect showed that the prevalence estimate was dominated by Walana et al. [31], Archampong et al. [35] and Kye-Duodu [29] (Fig. 4).