A systematic review and meta-analysis of the association between integrated tuberculosis and Human Immuno-deficiency Virus therapy and tuberculosis treatment outcomes among adults in sub-Saharan Africa

Background: Despite the scale-up of programmes integrating treatment of tuberculosis (TB) and Human Immuno-deficiency Virus (HIV) in sub-Saharan Africa (SSA), TB treatment outcomes in TB/HIV co-infected patients in the region remain sub-optimal. Objective: To summarize the available evidence on the association between integrated TB/HIV treatment and TB treatment outcomes specifically, successful treatment and all-cause mortality in TB/HIV co-infected adults in SSA. Method: A systematic review of studies published between March 2004 and 10 July 2019 was performed. Seven electronic databases including Medline, Embase and Cochrane were searched to identify interventional and observational quantitative studies reporting on integrated TB/HIV treatment in SSA. Two investigators independently screened the search output, reviewed the eligible studies, and rated the quality of eligible studies using quality assessment tools of the National Heart Lung and Blood Institute. Pooled odds ratios (ORs) were derived using random-effects meta-analysis. Heterogeneity across studies was assessed using the I 2 statistic. The confidence in the pooled ORs was rated using the GRADE. The final review was reported using the PRISMA. Results: Eleven studies including 4181 participants were retained. The studies were of moderate to good quality, with 10 being quasi-experimental and cohort. Pooling of relevant studies showed that the odds of treatment success with integration was 1.1 times (95% CI: 0.93-1.29; I 2 =74.5%; p-value for heterogeneity=0.001) the odds of treatment success without integration and the odds of mortality with integration was 1.27 (95% CI 1.02-1.59; I 2 =87.3%; p-value for heterogeneity<0.0001) times the odds of mortality without integration. On sensitivity analysis, the odds of mortality with integration decreased to 1.06 (95% CI: 0.83-1.34; I 2 =80.1%; p-value for heterogeneity<0.0001) times the odds of mortality without integration and there was strong evidence of an association between sample size variation and heterogeneity (p=0.01). Good quality studies (4/11) tended to support the effectiveness Conclusion: Evidence on the effect of and all-cause mortality in TB/HIV co-infected patients in SSA is inconclusive but the few available good quality studies tend to favour the effectiveness of these services. More robust primary studies are warranted.


Introduction
Tuberculosis (TB) is the most common opportunistic infection in persons living with HIV/AIDS (PLWHA) [1]. In 2017, it was estimated that there were 10 million incident TB disease cases worldwide, with PLWHA accounting for 900, 000 (9%) of these [2]. Sub-Saharan Africa (SSA) bears the greatest burden of HIV and TB co-infection in the world; in 2017, up to 72% of all TB patients living with HIV/AIDS resided in the region [2]. Currently, TB accounts for as high as one in three AIDS-related deaths on the global scale [3] and in low-income settings like SSA, it is the most common cause of death in PLWHA [4]. TB treatment outcomes in co-infected patients remain suboptimal in several lowincome settings like SSA [7]: for instance, among the 900,000 co-infected patients recorded worldwide in 2017, up to 300,000 (33.3%) died because of TB, with the majority of cases being registered in SSA [7].
There is substantial evidence indicating that patients who are co-infected with TB and HIV are more likely to experience unsuccessful TB treatment outcomes [6][7][8]. Studies have demonstrated that a key intervention to address this issue is to initiate and maintain HIV-infected TB patients on antiretroviral drugs [9][10][11][12]. However, in SSA, the delivery of therapy to patients who are co-infected with TB and HIV has traditionally been through separate (vertical) TB and HIV programmes [13]. This approach has been associated with diagnostic and therapeutic delays which lead to unnecessary mortality [13]. In order to foster coordination of TB/HIV treatment and improve outcomes, the integration of HIV and TB treatment services was proposed. Integrated TB/HIV treatment aims to merge treatment commodities for both TB and HIV, reduce fragmentation of TB/HIV treatment services, improve access to and quality of concurrent TB/HIV treatment and enhance the effectiveness of concurrent TB/HIV treatment [14].
In 2004, the World Health Organization (WHO) issued guidelines on collaboration between TB and HIV programmes. These guidelines sought to promote integration of TB/HIV treatment and care [15]. The guidelines were based on incomplete evidence and were thus termed provisional guidelines [16]. In 2012, WHO issued a review of these guidelines [16]. Overall, the updated guidelines present a similar framework as the interim policy but lay emphasis on establishing mechanisms for delivering integrated treatment for TB and HIV [16]. A gradual scale-up in various integration strategies has been reported following the new guidelines [17].
While acknowledging the rationale for integrating TB/HIV treatment and scaling-up integration strategies in a high burden setting like SSA, few successful strategies have been reported [18] and evidence on the downstream effects of TB/HIV treatment integration on therapeutic outcomes of TB in co-infected patients in routine clinical settings in the region remain unclear. Studies from diverse locations such as those by Gandhi et al in South Africa [19], Jack and et al in South Africa [20], Pathmanathan et al in Swaziland [21] and Musaazi et al in Uganda [22] describe high rates of successful TB treatment outcomes following treatment integration. But in all these studies, the evidence is derived from study populations with no comparison groups [19,21,23] and is therefore inadequate as evidence of the effectiveness of treatment integration in improving TB outcomes.
Meanwhile, some observational studies reported findings that contradict the widely upheld beneficial Pioneer systematic reviews by Legido-Quigley et al in 2012 [13] and Uyei et al in 2011 [26], provide evidence on models of delivering integrated care for TB/HIV in low-income settings but do not investigate the estimated effect of integrated TB/HIV treatment on TB treatment outcomes. Also, the evidence included in these reviews were essentially based on the 2004 WHO guidelines and older recommendations [13,26]. Consequently, there is lack of recent comprehensive reviews estimating the effect of TB/HIV treatment integration on TB treatment outcomes in SSA. Objectives 1.
To assess the association between integrated TB/HIV treatment and For studies reporting other measures of effect such as risk ratio, these measures were converted to proportions and the proportions converted to odds ratios as explained above. The quantitative data entered in the excel spreadsheet was then exported to STATA version 15 software. Presuming the variation in the true effect of integrated treatment on TB outcomes between studies, a random effect meta-analysis was fitted to estimate the summary odds ratios and their 95% CI. The analysis was conducted for treatment outcomes for which we obtained at least five effect sizes. The results of the meta-analysis were presented as forest plots and funnel plots (assessing risk of publication bias). The percentage of variation in effect estimates that was due to heterogeneity across studies was assessed using the I 2 statistic. In this study, the degree of heterogeneity was interpreted as none (I 2 <25%), low (25≤ I 2 ≤49%), moderate (50≤ I 2 ≤74%) or high (I 2 ≥75%). For each treatment outcome, the p-value for heterogeneity was used to determine whether heterogeneity was associated with variations in the odds ratio across studies. When methodological limitations could explain the size of the odds ratio in a study, sensitivity analysis was conducted by omitting the study and repeating the meta-analysis.
Meta-regression was used to assess the effect of sample size variations on summary estimates. All reported p-values were two-sided with a significance level of 0.05.

Reporting method:
The review procedure was developed and reported as per the Preferred Reporting for Systematic Reviews and Meta-analysis (PRISMA) guidelines (supplementary file 1) Ethical considerations: Ethical approval was not needed for the study as it is a systematic review and does not involve the collection of data from human participants.

Systematic review registration: A study protocol was submitted and registered with the
International Prospective Register of Systematic Reviews (PROSPERO), under the registration number CRD42020173701. Figure 1 shows the steps that were followed to arrive at the final articles included in the review. Twohundred and sixteen studies were retained for full text review after removal of duplicates, review of the reference list of relevant studies, and screening of titles and abstracts. Eleven studies assessing the association between TB/HIV treatment integration and TB treatment outcomes were retained. The included studies had a total of 4181 participants. Seven of these were of moderate quality and the rest of good quality (supplementary file 2). With regards to setting, 5/11 studies were conducted in East Africa and 5/11 in Southern Africa. Five studies were exclusively from urban settings, four from rural settings and two studies were from both rural and urban settings. Regarding study designs, 6/11 studies were quasi-experimental (before-and-after) studies. Table 2 is a comparative data table on the studies included in the review.

Association between TB/HIV treatment integration and TB treatment success
Six of the studies (two of good quality) had data that could be used to estimate the odds of TB treatment success with implementation of integrated treatment compared to the odds of successful treatment outcomes in the absence integrated treatment. As shown in figure 2, between-study heterogeneity (variability in the estimated crude odds ratios) was moderate (I 2 =74.5%) and the pooled odds ratio was 1.1 (blue diamond. In TB/HIV co-infected patients, the odds of experiencing successful outcomes under treatment integration was similar to the odds of experiencing successful outcomes when no integration strategy was employed. The 95% confidence interval (95% CI) for the pooled odds ratio was 0.93-1.29. We can be reasonably (95%) confident that the pooled odds of treatment success following the implementation of TB/HIV treatment integration is between 0.93 to 1.29 times the odds of treatment success in the absence of a strategy to integrate therapy. The 95% CI overlaps 1; it is possible that the observed pooled effect is due to chance. There was strong evidence of an association between heterogeneity and variations in the odds ratios across studies (p= 0.001). On meta-regression, there was borderline evidence of an association between sample size variation and the observed heterogeneity (p=0.064). Supplementary file 3 is a funnel plot to assess the possibility of publication bias. The data points (representing the six studies) are symmetrically distributed; there is minimal risk of publication bias.

Association between TB/HIV treatment integration and all-cause mortality
Seven studies (four of good quality) estimated the odds of mortality with integrated treatment services compared to the odds of mortality in the absence of treatment integration. As shown in figure 3, the variability in the estimated crude odds ratios (between-study heterogeneity) was high (I 2 =87.3%) and the pooled odds ratio was 1.27 (blue diamond). In TB/HIV co-infected patients, the odds of death following the implementation of a treatment integration was about 1.3 times the odds of mortality in the absence of integrated treatment services. The 95% confidence interval (95% CI) for the pooled estimate was 1.02-1.59. We can be reasonably (95%) confident that the pooled odds of death following the implementation of TB/HIV treatment integration is between 1.02 to 1.59 times the odds of mortality in the absence of integrated therapy. The 95% CI excludes 1; it is unlikely that the observed pooled OR is due to chance. Meta-regression showed borderline evidence of an association between sample size variation and heterogeneity (p= 0.057). There was very strong evidence of an association between heterogeneity and variations in the odds ratios across studies (p<0.0001). obtained. Contrary to previous systematic reviews that limited their scope to hospital-based studies and did not describe quality assessments of the studies or the confidence in the cumulative evidence [13,26], the current review includes evidence from community-based studies and describes how the quality of included studies as well as the confidence in the measured outcomes were rated.
Importantly, meta-analyses showed that there was minimal risk of publication bias, i.e, it is reasonable to consider that the studies with effect estimates on successful TB treatment and mortality included in the meta-analyses, constitute a representative sample of the available evidence on TB treatment success and all-cause mortality in integrated TB/HIV treatment programmes. As this review mainly included routine programmatic data, its findings reflect real life situations, and this is extremely useful in informing TB/HIV integration programmes on their effectiveness in reducing mortality or increasing treatment success.
Evidence from community-based studies have tended to demonstrate that integrated TB/HIV therapy leads to better clinical outcomes and use limited resources more efficiently [19,42]. However, patients enrolled for community-based treatment are generally carefully selected and in concurrent TB/HIV therapy these patients are required to fulfil essential criteria like clinical stability, undetectable viral load and optimal CD4 counts before being down-referred to community-based clinics [19,42,43].
As a result, community-based programmes integrating TB/HIV treatment may tend to overestimate treatment success or underestimate all-cause mortality and so the effects of such programmes on treatment outcomes should be interpreted with caution. that the conduct of further randomised trials in this regard could be considered unethical [9,10,44].

Meta-analyses including large randomised trials (in which TB
Even the SAPiT trial, a pioneer clinical trial that was crucial in establishing the effectiveness of early initiation of antiretrovirals (during TB treatment) on improving therapeutic outcomes was interrupted prematurely because early initiation clearly had more benefits over delayed initiation and it was deemed unethical to continue enrolling patients in the delayed initiation arm [9,45]. Even though it is expected that compared to vertical treatment of TB and HIV, integrated TB/HIV treatment should lead to higher or much earlier initiation of ART in co-infected patients and presumably better treatment outcomes, there is considerable evidence indicating that in routine clinical settings, there are serious shortfalls in the uptake of ART and its timeliness relative to the start of TB treatment in integrated TB/HIV clinics [21]. In routine clinical settings, there are factors such as psychosocial, economic and The current systematic review has some limitations. It was not possible to obtain odds ratios adjusted for confounders in most of the quantitative studies, so pooled odds ratios were obtained from crude estimates which are unadjusted for confounders. Moreover, the data assessed in this review was not thick and the included studies did not consistently report quantitative indicators of treatment progress such as clinical staging of patients, viral load, CD4 counts. As such, even though randomeffects meta-analysis was appropriately used, given the high heterogeneity in effect estimates, there was insufficient data to perform meta-regression to assess the effect of many variables on heterogeneity. Nonetheless, on sensitivity analysis, it was observed that there was strong evidence of an association between sample size variation and heterogeneity when assessing the association between integrated treatment and all-cause mortality. It is also worth noting that TB

Conclusion
Based on the evidence included in this review, the role of TB/HIV treatment integration in increasing successful outcomes and decreasing mortality in TB/HIV co-infected patients was inconclusive. But these results should be interpreted with caution for a number of reasons. The available good quality studies tended to support the effectiveness of TB/HIV treatment integration in increasing successful outcomes and decreasing mortality, but these studies were few, and smaller in number than the moderate quality studies. Additionally, much of the evidence was from retrospective analyses of routine programmatic data which restrict investigators' ability to reduce confounding and other sources of bias. The cumulative evidence in this review does not indicate that TB/HIV treatment integration improved TB treatment outcome in SSA but there is low confidence in this cumulative evidence. More robust primary studies are therefore required to arrive at more conclusive evidence on the association between TB/HIV treatment integration and TB treatment outcomes. Such evidence will be crucial in updating policies and clinical guidelines on integrated TB/HIV treatment. Some important considerations that could improve on the validity of future research include: prospective data collection and follow-up of study participants over several months in order to properly report outcomes; use of impact evaluation methods that are more robust than before-and-after studies; optimal adjustments for confounders and missing/poorly reported data; and comparing integrated and non-integrated treatment services using other proxies of treatment outcomes (beyond routine WHO standards). Such proxies could include timeliness of ART uptake relative to TB treatment and TB   Ndagijimana 2015, Rwanda [38] To evaluate one-stop TB-HIV services in Rwanda by comparing the TB treatment outcomes before and after their implementation Historical (before-and-after) study in 12 public facilities in rural and urban areas of Rwanda involving 888 patients: 413 postintervention cases and 475 pre-intervention cases ------------One-stop services w operation facilities. treatmen after the interv comparab the interv Owiti 2015, Kenya [39] To assess the uptake and timing of cotrimoxazole preventive therapy and antiretroviral treatment as well as anti-tuberculosis treatment outcomes among HIV-infected TB patients before and after the introduction of integrated TB-HIV care.
Before-and-after study of routine data on 797 patients: 347 preintegration (March-October 2010) and 450 postintegration (March-October 2012) cases in 17 [20] To determine the feasibility and effectiveness of integrating antiretroviral therapy into existing tuberculosis directly observed therapy programs Prospective study of 20 coinfected patients in an urban tuberculosis clinic