Table 4.

Sensitivity analyses results

All LMIC	Male (n)	Female (n)	≤12a	df	I 2 (%)	13–35a	df	I 2 (%)	36–59a	df	I 2 (%)	60–108a	df	I 2 (%)	Overall	df	I 2 (%)
Model 1	186,452	283,811	1.42 (1.25–1.63) p=0.002	15	57.5	1.39 (1.23–1.57) p=0.000	13	70.1	1.36 (1.22–1.52) p=0.000	14	66.7	1.62 (1.46–1.80) p=0.000	21	74.8	1.46 (1.38–1.56) p=0.000	66	75.2
Model 2	175,554	266,621	1.42 (1.25–1.63) p=0.002	15	57.5	1.40 (1.38–1.56) p=0.000	12	72.3	1.37 (1.22–1.53) p=0.000	13	69.1	1.62 (1.46–1.80) p=0.000	21	74.8	1.47 (1.38–1.56) p=0.000	64	75.9
Model 3	119,470	168,465	1.42 (1.25–1.63) p=0.002	15	57.5	1.43 (1.24–1.64) p=0.000	11	70.8	1.46 (1.21–1.75) p=0.000	10	72.3	1.49 (1.35–1.54) p=0.006	14	54.2	1.44 (1.35–1.54) p=0.000	53	62.7
Model 4	86,233	117,719	1.42 (1.21–1.67) p=0.002	12	60.7	1.48 (1.23–1.78) p=0.027	5	60.4	1.50 (1.18–1.91) p=0.000	8	77.8	1.49 (1.29–1.71) p=0.005	13	56.8	1.46 (1.35–1.59) p=0.000	41	63.4
Model 5	85,552	116,994	1.44 (1.22–1.71) p=0.000	11	63.0	1.48 (1.23–1.78) p=0.000	5	60.4	1.54 (1.18–2.02) p=0.002	7	80.6	1.49 (1.29–1.71) p=0.000	13	56.8	1.47 (1.35–1.60) p=0.000	39	64.8
Model 6	85,431	116,756	1.37 (1.20–1.56) p=0.000	10	43.9	1.48 (1.23–1.78) p=0.000	5	60.4	1.54 (1.18–2.02) p=0.002	7	80.6	1.43 (1.27–1.61) p=0.000	12	42.2	1.43 (1.33–1.55) p=0.000	37	58.9

df, degrees of freedom; LTFU, lost to follow-up; HR, hazard ratio.

Model 1: all HRs.

Model 2: high-quality studies only (7–9 stars).

Model 3: HRs with LTFU <18%.

Model 4: HRs with LTFU <15% (all also scored high on the NOS).

Model 5: HRs with LTFU <15% and adjusted for age (excludes n=2 HRs).

Model 6: HRs with LTFU <15%, adjusted for age, and accounts for potential publication extremes on the funnel plot (excludes two additional HRs beyond Model 5).

^aMonths since initiation of ART