Mortality and its predictors among human immunodeficiency virus-infected children under 15 years of age receiving antiretroviral therapy in Ethiopia: a systematic review and meta-analysis | BMC Infectious Diseases

Study selection

A comprehensive search of the medical database identified 8,283 records. After duplicates were removed, 190 remained and were screened based on their titles and abstracts, and 123 were removed as not related to the study domain. Sixty-seven full-text articles were evaluated against the inclusion criteria, of which 45 were excluded. Finally, 22 articles were included in the meta-analysis (Fig. 1).

Figure 1
Figure 1

Flowchart of the article selection and screening process.

Study characteristics

Twenty-two articles (7, 8, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34) involving Se They included 8,731 children under 15 years of age. 51% (north= 4406) of the study participants were men. All studies employed a retrospective cohort epidemiological study design and reported data were extracted from ART databases, patient registries, and hospital and health center follow-up forms. The publication period of the original articles ranged from 2011 (22) to 2022 (7, 24, 27, 31). According to the administrative regions of Ethiopia, two studies were conducted in Oromia (16, 19), nine in Amhara (7, 8, 18, 21, 26, 30,31,32, 35), four in Addis Ababa, the capital of the country. (15, 22, 23, 28), four in the Southern Nations, Nationalities and Peoples’ Region (SNNPR) (20, 27, 29, 33), two in Tigray (24, 26) and one in the Harari regions (17) . The sample ranges between 228 (29) and 757 (29). The minimum and maximum person-time of observation was 329 child-years (22) and 4112 child-years (15), respectively. Mortality estimates from individual studies range from 1.08 per 100 child-years (7) to 12.94 per 100 child-years (8). The overall evaluation scores range from 9 to 11, indicating that the quality of the articles included in the meta-analysis was high (Table 1).

Table 1 Descriptive characteristics of the studies included in the systematic review and meta-analysis


We analyzed the statistical heterogeneity of the 22 retrospective follow-up studies (Yo2= 91.1% and P< 0.001). Weighted random-effects meta-analysis revealed that the pooled incidence density of mortality among HIV-infected children after ART initiation was 3.08 (95% CI, 25.2 to 36.4) per 100 years. child in Ethiopia (Fig. 2).

Figure 2
Figure 2

Forest plots of pooled mortality proportion among HIV-infected children after ART initiation in Ethiopia

Subgroup analysis and meta-regression.

A subgroup meta-analysis was performed using sample size, publication period, and administrative region of Ethiopia. Statistical heterogeneity ranged from 0.0 to 94.8%. Subgroup analysis revealed that mortality was 4.60 (95% CI, 2.93 to 6.26) per 100 child-years in Harari region, 3.26 (95% CI, 2.57 to 3.95) per 100 child-years when the sample size was less than 500, and 3.08 (95% CI, 2.52 to 3.64) per 100 child-years in articles published in 2016 and later (Table 2). Additionally, a meta-regression was performed and the results revealed that although there was a significant difference in the year of publication (t = -2.37, P= 0.029), the sample size was insignificant (t = -1.71, P= 0.104).

Table 2 A subgroup analysis of mortality among HIV-infected children after ART initiation

Sensitivity analysis

A leave-one-out sensitivity analysis was performed to assess the impact of each study on the pooled mortality rate and gradually excluding each study. The results showed that the combined effects did not change significantly (Fig. 3).

Fig. 3
figure 3

A sensitivity analysis leaving one out

Publication bias

in a P-value of < 0.05, Egger and Begg tests for small study effects plots revealed publication bias. Funnel plots of pooled incidence density of mortality among HIV-infected children after ART initiation in Ethiopia showed significant asymmetry, on visual inspection (Fig. 4) and pooling of test results. Egger regression-based (t = 8.52, P< 0.001), and Begg's non-parametric rank correlation test (Z = 3.52, P< 0.001) showed the presence of evidence of small study effects. When evaluated with the Egger© regression test, the P value was <0.001. Therefore, the test provides strong evidence for the presence of a small study effect.

Figure 4
Figure 4

Funnel plots of pooled mortality proportion among HIV-infected children after ART initiation in Ethiopia

Furthermore, as illustrated in Figure 5, although only four estimates barely touched the regression line, the remaining data points were scattered far from the regression line.

figure 5
Figure 5

Regression plot of mortality incidence density among HIV-infected children after ART initiation in Ethiopia

Additionally, Duval and Tweedie’s non-parametric trim-and-fill funnel plot asymmetry tests were performed, which provides a way to assess the impact of missing studies due to publication bias in the meta-analysis. Therefore, meta-trim analysis demonstrated the presence of 10 unpublished studies (Fig. 6) and full meta-analysis revealed that the pooled incidence density of mortality was 1.83 (95% CI, 1.22). to 2.43) per 100 childhood years.

Figure 6
figure 6

Trim and Fill Analysis to Determine Mortality Incidence Density Among HIV-Infected Children After ART Initiation in Ethiopia

Predictors of mortality after initiation of antiretroviral therapy

In this systematic review and meta-analysis, being a rural resident, having poor adherence to ART, not initiating CPT, having low hemoglobin levels, the presence of OI, malnutrition (underweight, wasting and growth retardation), based regimens in NVP and advanced levels of WHO classified clinical stage of HIV were predictors of mortality. On the other hand, initial CD4+ T cell count, IPT, and history of treatment failure were not significant.

To begin with, the mortality risks were 2.18 times (HR, 2.18 (95% CI, 1.20 to 3.98); Yo2= 88.3%) higher among children living in rural environments than among urban residents. The risk of mortality was 2.85 times (HR, 2.85 (95% CI, 1.39 to 5.88); Yo2= 93.8%) higher among children with poor adherence to ART. Children who did not take CPT had 2.16 times (HR, 2.16 (95% CI, 1.52 to 3.07); Yo2= 81.1%) higher mortality risks than those who received prophylactic chemotherapy.

Children with anemia (hemoglobin < 10 g/dl) had 2.28 times (HR, 2.28 (95% CI, 1.51 to 3.45); Yo2= 94.4%) increased risks of mortality compared to those who had normal baseline hemoglobin levels. Mortality risks were 52% higher (HR, 1.52 (95% CI, 1.15 to 2.00); Yo2= 92.9%) among children diagnosed with any of the OIs than those who did not have the condition. Children who were underweight, wasted, and stunted had a 74% higher risk (HR, 1.74 (95% CI, 1.26 to 2.41); Yo2= 91.6%), 2.54 times higher (HR, 2.54 (95% CI, 1.56 to 4.16); Yo2= 82.6%), and 2.02 times higher (HR, 2.02 (95% CI, 1.63 to 2.51); Yo2= 0.0%) mortality risks compared to well-nourished children, respectively. Children with advanced WHO clinical stages at presentation (III and IV) had a 71% higher risk (HR, 1.71 (95% CI, 1.42 to 2.05); Yo2= 83.5%) mortality risk than those with mild or asymptomatic stages. Finally, children receiving NVP-based regimens had 3.91 times (HR, 3.91 (95% CI, 3.09 to 4.95); Yo2= 89.9%) (Table 3).

Table 3 Meta-analysis of mortality predictor among HIV-infected children after ART initiation