Incidence and determinants of mortality among HIV infected adult patients on antiretroviral therapy in Ethiopia: A systematic review and Meta-analysis

generate summary data and Articles were comprehensively searched on Pub Med, Google Scholar, Cochrane library, Scopus, and DOAJ databases using Boolean operators. A Dersimonian and Laird methods of random effect model was used to estimate incidence and determinants of mortality. Heterogeneity, publication bias and quality of each study were checked. Subgroup analysis was employed. Relevant data from each study were extracted. STATA software version 14 was used for all statistical analysis.


Conclusion
Incidence of mortality was high particularly early in the course of therapy. Advanced WHO clinical stage, CD4 cells count low body weight, low hemoglobin level presence of TB infection, bad medication adherence older age and non-working functional status were signi cant determinants of incidence of mortality. Comprehensive service and strict follow up should be given to avert this high rate of mortality. among adult populations due to HIV/AIDS was 9,294. Among these female's death due to AIDS accounted a total of 5019 (5). In Ethiopia, the incidences of mortality due to AIDS-related causes dropped by 50 percent in the two years following the introduction of free ART in 2005. Between 2001 and 2009, the mortality rate attributable to AIDS fell from 50-14% for women, and 40-11% for men (6,7). Every individual will be free of HIV, free from discrimination and with dignity, equality and healthy, and any people infected with HIV should receive the treatment, protection, care and support to survive and improve their quality of life (8).
HIV primarily affects the most productive age group, and it affects not only health of individuals, but also households, communities, and nations. Many of the countries hardest hit by HIV also face serious challenges due to other infectious diseases, food insecurity, and additional global health and development problems. HIV/AIDS has remained one of the top priority's health agenda for the Ethiopian government over last thirty-year period (5).
Clinical bene t of Highly Active Anti-Retroviral Therapy (HAART) for HIV infected patients, in terms of mortality reduction and improved quality of life, is well established but shows regional variations. Higher incidence of mortality within the rst years of HAART initiation especially during the rst 3-6 months of treatment and determinants contributing to this high mortality are poorly understood. There are several determinants of mortality for HIV infected patients started HAART which include, CD4 count, total lymphocytes count, body mass index, and medication adherence, WHO clinical stage, years lived with the virus, presence of opportunistic infections. Different studies conducted in Ethiopia indicated that there were variations among the determinants of mortality due to AIDS related illness. So, a better knowledge of determinant factors would allow closer follow-up and more targeted interventions for HIV infected patients, thus reduce excess mortality, alter morbidity and improve quality of life (9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19).
There is only one study tried to describe the determinants of mortality by systematic review, but he failed to do meta-analysis (20).Hence, studying the pooled incidence rate and determinants of mortality among HIV infected adult patients who started HAART will provide better evidence-based quality data that are important for the healthcare system of the country, particularly for ART program.

Method Protocol and Registration
This systematic review and meta-analysis were conducted in accordance with the Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) statement of 2009. The protocol for this systematic review and meta-analysis was registered in the international prospective register of systematic reviews (PROSPERO) which can be accessed using http://www.crd.york.ac.ku/PROSPERO with the registration number CRD42019123380.

Eligibility criteria
This systematic review and Meta-analysis considered all those published and unpublished studies conducted throughout Ethiopia (Table 1).

Literature Searching strategy
A comprehensive internet-based search was conducted to nd research articles on the electronic databases with keywords and MESH terms stringed out using "AND", "OR" and "NOT" Boolean operators. Search was used based on COCOPO ( Table 1). The ndings from the electronic freely access databases, hand search of the reference lists and institutional repositories were pooled for subsequent selection. We have searched articles from the reference list of the included studies directly from the internet.

Study selection
All retrieved studies were exported to Endnote X8 reference manager (Thomson Reuters) and duplicate studies were removed. The remaining studies were screened based on their titles and abstracts and then a full-text review was performed to determine their eligibility. Lastly, studies that ful lled the eligibility criteria were included in the nal review and synthesis.

Data collection process
Quantitative data were extracted from papers included in the review using the standardized data extraction tool. Data was collected from April 30,2019 to March 30,2019. Two reviewers MPH in Epidemiology have extracted the data and cross-checked to ensure consistency from all included studies. The following Information's author and year of publication, study area, study subjects, study design, the total sample size, the number of participants and events in each groups, median follow-up period, incidence of mortality due to AIDS related causes, and Hazard Ratio among exposed and non-exposed groups were extracted. There was no discrepancy happening during data extraction between the reviewers. Standardized data extraction tool from Joanna Briggs Institute Meta-Analysis of Statistics Assessment and Review Instruments were used

Data quality control measures
Quality of all studies included in the review and analysis was assessed and evaluated using Joanna Briggs Institute (JBI 2017) quality appraisal tools adapted for studies reporting observational studies.
Based on the quality assessment tools all the included studies scored above 82%. Search beyond published literature was done to reduce the risk of publication bias.

Synthesis of results
A DerSimonian and Laird method of random-effect model was employed to compute the overall pooled incidence of mortality with its corresponding 95% con dence interval. The pooled hazard ratios of the determinants of mortality were estimated from each study. Heterogeneity across studies was assessed using the Cochrane Q test and quanti ed by Higgins and Thompson's I 2 statistic. Subgroup and sensitivity analysis were performed to explore the possible causes of heterogeneity. Publication bias across studies was checked using funnel plot and Egger's regression test. STATA software version 14 was used for all statistical analysis. The results of the analysis were summarized and presented graphically using Forest Plot.

Operational de nitions
Incidence of mortality: Incidence rate was de ned as the rate of number of participants died during the follow up period to the total person year of observation. (events/person years of observation).
WHO clinical stage: Study participants are classi ed into two groups based on WHO categories. Those patients with WHO clinical stage 1 and 2 are categorized as Non-exposed group and those patients who satis ed WHO clinical stage 3 and 4 are as exposed group.
CD4 + cells count: Patients with baseline CD4 cells count less than 200 cells/µl was considered as exposed group while those patients with CD4 cells count greater than 200 cells/µl was considered as unexposed group.
Adherence: Patients who had good medication adherence was classi ed as non-exposed group and those patients with poor and fair adherence were classi ed as exposed group.
Functional status: Patients who were classi ed as working functional status at enrollment to chronic care are grouped into non-exposed group and those ambulatory and bedridden patients are exposed group. Anemia: Patients with hemoglobin level below 10 g/dl had anemia and classi ed as exposed grouped whereas patients with hemoglobin level above 10 g/dl had no anemia and classi ed as non-exposed grouped.
Body weight: patients who had baseline body weight lower than 45 kg were assigned to exposed group and above 45 kg as non-exposed group.
Age: patients whose age was above 40 years were categorized as exposed group and below 40 years as non -exposed group.

Results
Description of studies A total of 423 articles were retrieved. After removing duplicated retrievals, 423 articles remained, of which 392 were removed during initial assessment due to exclusion criteria's set at the review. From the remaining 31 articles 10 were removed based full text articles. Finally, 21 articles were included for the nal synthesis (Fig. 1). All articles included in the nal reviews and analyses were retrospective cohort studies which were conducted throughout the country from September 01, 2006 up to October 25, 2018.

Study characteristics
The studies comprised a total sample size of 23904 participants ranging from 152 (ID 08) to 5664 (ID 11) that contributed to a total of 63, 107.7 person years of observations during their follow up. Of these, 13390 (56% of the participants) were females. There were 2, 804 events occurred in the study period. The study contributed a total of 1154.75 months of follow up period, a median follows up period of 60 months ranged from a minimum of 55 weeks to a maximum of 96 months.

Incidence of mortality
Incidence of mortality analyzed from eighteen studies. The random pooled estimate of the incidence of mortality among adult HIV infected patients after HAART initiation was 5 deaths per 100 person years of observation (95% CI: 4-5/100pyo, p-value < 0.001; heterogeneity I 2 = 97.39%) (Fig. 2).

Sensitivity analysis of incidence of mortality
The leave -one -out sensitivity analysis has been performed for incidence of mortality. The sensitivity analysis con rmed that the nding of this study was independent of a single study. The pooled estimated incidence of mortality varied between 4.0 deaths per 100 person year of observation (95% CI: 3.0-5.0) and 5.0 deaths per 100 person year of observation (95% CI: 4.0-6.0) after removing one study at a time from the analysis ( Table 2).    (Fig. 7). Sensitivity analysis was done but no change on Hazard ratio.
Sensitivity analysis was done but no change on Hazard ratio.  (Fig. 9). Sensitivity analysis was done but no change on Hazard ratio.

Low body weight
The overall pooled hazard ratio of patients with low body weight at the initiation of antiretroviral treatment was 1.59 higher than those patients with normal body weight (95% CI: 1.15-2.19, p -value < 0.001; heterogeneity I 2 = 88.6%), (Fig. 11). Sensitivity analysis was done but no change on Hazard ratio.

Publication Bias
The studies were scattered symmetrically in funnel plot which indicated that there was no publication bias among the studies (Fig. 14).

Egger Test
The estimated publication bias coe cient of egger test was 0.5 with a standard error of 1.36 and pvalue of 0.716 which indicated that there is no evidence for the presence of small study effect among the included studies (Table 3).

Discussion
In this review, the summary incidence of mortality among adult HIV infected patients after taking HAART Patients who had low CD4 cells counts at initiation of HAART was almost twice more likely to experience hazard of death than patients with higher CD4. This is a similar nding to the systematic review and meta-analysis conducted by Anglemyer et al (39), study at Iran (40) Patients presented with non -working functional status at the initiation of HAART in Ethiopia were four times more likely to experience death than those patients with working functional status. A study done in Maryland, USA to examine the association between physical performance and mortality on HIV infected patients indicated that reduced physical performance among HIV infected individuals had been associated with increased mortality (hazard ratio 2.52, 95% CI: 1.59-4.00) (42). HIV infected individuals frequently exposed to physical performance impairment because of loss of muscle mass, altered body composition, decrease physical function, frailty, and disability. These predisposing causes may restrict the day to day activities of the patients especial during advanced HIV diseases and increased the likelihood of mortality (39).
Different studies reported that patients with non -adherence to treatment had been associated with increased risk of mortality than those who adhere to their treatment. HAART will work best for patients who adhere to their medication and these patients have increased chance of survival. Poor adherence is associated with the risk of early treatment failure and rapid development of drug resistance (43,44). The hazard of death among adult HIV infected patients with poor medication adherence at initial enrollment to chronic care had nearly ve times higher than those patients with good medication adherence.
In this analysis, patients diagnosed with TB at initiation of HAART were three times more likely to die than those patients without TB. This nding is in line with study done in assessing determinants of mortality in South Africa (45), Iran (40)  In this review, HIV infected patients with severe anemia at initiation of HAART had two times risk of dying than those patients without anemia. This nding is similar with the nding from Tanzania showed that severe anemia had been associated with increased mortality with odds ratio of 6.6 (95% CI, 3.4-12.9) (37). Another study at LMIC indicated that anemic patients with hemoglobin level < 8 gm/dl have experienced the greatest risk of mortality. Anemia among HIV infected individuals are easily exposed to chronic diseases like chronic heart failure that may be fatal to the patients could be the best explanation for the nding (35,36).
HIV infected patients who did not take a cotrimoxazole preventive therapy before or at the initiation of HAART were 53% more risk to die than patients who took cotrimoxazole preventive therapy. The nding is similar with systematic review and meta-analysis conducted at LMIC, patients who had taken CPT in their course of therapy had a 58% reduced incidence of mortality among HIV infected adults (46). This might be CPT has been recommended for all HIV infected patients developing advanced AIDS stage III & IV and patients with WHO stage I & II whose CD4 counts < 350cells/µl. Cotrimoxazole is safe, well tolerated, widely available and inexpensive which is used to prevent opportunistic infections which can be developed when the immune system of the body is compromised due to HIV infection (40).
Patients with low body weight at the initiation of HAART had 59% higher risk of death than those patients with normal range of body weight. A systematic review and meta-analysis conducted in LMIC among adult patients initiating HAART showed that low Body Mass Index/ body weight was independently associated with early mortality (36). In the trend mortality at Asia -Paci c region, lower BMI was a signi cant determinant associated with increased mortality in which underweight patients with BMI value < 18.5 had three times more hazard of death than patients who were in the normal BMI range of 18.5-24.9 (HR = 3.33, 95% CI: 2.31-4.81) (36). This may be due to the underlying effects of poor nutrition which may result in low body weight and hence associated with early mortality. Loss of appetite, inability to eat/ di culty to swallow foods and stress associated with HIV/AIDS make the immune system to weaken and easily vulnerable to advanced AIDS stage. Body weight was used as a proxy indicator of nutritional status of a person.
In this analysis, being male had 41% higher risk of death due to HIV/AIDS related cause than female after the initiation of HAART in Ethiopia. This nding is consistent with a national ART study conducted at Botswana that predicted women had a lower risk of mortality (about 46% of reduction) due to AIDS related causes when compared with men (odds ratio = 0.64; 95% CI 0.60-0.69) (35). In Tanzania one study indicated that being male had almost two times higher odds of death than females (odds ratio = 1.8, 95% CI: 1.1-3.0) (37) and a systematic review and meta-analysis in LMIC revealed that male sex was an independent risk factor associated with early mortality in about 30 (60%) of the included studies (35). The reasons why males were in general more likely to die early in the course of therapy than females may be due to the difference in health seeking behavior, biological differences in ART response resulting in poorer adherence in males, difference in HIV treatment outcomes.
In this study older ages above 40 years had 21% higher hazard of dying from HIV/AIDS when compared to patients below the age of 40 years. This result is in line with a study done at LMIC, Botswana and China indicated that older age above 40 years had been strongly associated with increased odds of AIDS related mortality (34,35,47). This is probably due to the fact that older people would experience to have poor adherence to ARV drugs because of low level of education and their desire to reduce the burden and stigma on their family associated with HIV/AIDS, older people may be underserved by public health care facility and has been associated with late presentation, diagnostic delays, poorer CD4 immune reconstitution may also explain the nding as is was mentioned in these studies.

Limitation of the Study
Patients who died at home without registered were considered as alive or loss to follow up which result in underestimation of incidence of mortality. Similarity, patients who died at health facility and recorded in the ART registry were considered as death due to HIV related causes as the exact cause of death was not differentiated routinely and this may overestimate the incidence of mortality.
This review and analysis may not include all the eligible studies because of lack of access to the databases. Only free databases (Pubmed, Google Scholar, Scopus, Cochrane and DOAJ) have been accessed for searching studies.

Conclusions And Recommendations
In summary, there was high incidence of mortality in this review and meta-analysis early in the course of therapy. Most of the death (67%) occurred during the rst year of follow up after initiation of HAART.
Advanced WHO clinical stage III&IV, low CD4 cells count below 200 µl/ml, low body weight below 40 kg, low hemoglobin level, presence of TB infection, non -working functional status, bad medication adherence, lack of cotrimoxazole preventive therapy, male sex and older age at the initiation of HAART were the signi cant determinants found to be associated with increased mortality in this review and meta analysis.
Routine intensi ed HIV testing and counseling services and early diagnosis of HIV infected patients and then early initiation of HAART before the deterioration of the functional status of the patients, before the reduction of CD4 cells counts and hemoglobin levels and before the development of advanced WHO clinical stage with continue monitoring of medication uptake and health education on adherence to ART regimens in these patients by all health facilities providing ART service will maximize the effectiveness of HAART and help to save their lives.
Approaches to reduce TB infection should be implemented and scaling up of TB prevention strategy including routine INH provision to all HIV infected adults, rapid TB detection and immediate treatment should be strengthened by health care facility.
Increase the utilization of cotrimoxazole preventive therapy to all eligible patients with adequate health education on its important in preventing opportunistic infections.
Easily accessible and available ART services for male and elderly people should be emphasized by service providers.

Multisectoral response o ce and stakeholders working on HIV prevention and control program should
give emphasis on ART services updates and refreshing health care workers and supportive staffs routinely. This is important to provide comprehensive health package to HIV infected patients at all level of service provision.  Figure 1