Table 2:

Unadjusted and adjusted relative risks of virological non-suppression and virological failure among patients on first-line antiretroviral therapy (ART)

	Virological non-suppression				Virological failure
	Unadjusted analysis		Adjusted analysis		Unadjusted analysis		Adjusted analysis
	RR (95% CI)	p value	RR (95% CI)	p value	RR (95% CI)	p value	RR (95% CI)	p value
Sex
Male	1·02 (0·99–1·06)	0·20	1·07 (1·02–1·13)	0·0080	1·07 (0·96–1·19)	0·20	1·08 (0·97–1·19)	0·17
Female	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)
Age at ART initiation, years
18–19	1·22 (1·07–1·40)	0·0032	1·28 (1·14–1·45)	<0·0001	1·71 (1·23–2·37)	0·0015	1·70 (1·22–2·37)	0·0016
20–29	1·12 (1·04–1·21)	0·0031	1·16 (1·07–1·25)	0·0002	1·26 (1·06–1·51)	0·0093	1·31 (1·10–1·22)	0·0021
30–39	1·11 (1·04–1·17)	0·0008	1·12 (1·06–1·18)	<0·0001	1·20 (1·03–1·41)	0·022	1·22 (1·04–1·42)	0·012
40–49	1·03 (0·98–1·10)	0·18	1·04 (0·99–1·10)	0·16	1·06 (0·91–1·22)	0·45	1·05 (0·91–1·22)	0·49
≥50	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)
Time on ART, years	1·21 (1·17–1·25)	<0·0001	1·16 (1·13–1·20)	<0·0001	1·11 (1·05–1·18)	0·0007	1·23 (1·15–1·32)	<0·0001
First viral load ≥12 months after ART initiation	0·97 (0·91–1·03)	0·32	··	··	0·78 (0·71–0·87)	<0·0001	0·66 (0·59–0·75)	<0·0001
Documented dispense ratio ≥51%	0·82 (0·78–0·87)	<0·0001	0·91 (0·86–0·97)	0·0009	0·78 (0·69–0·88)	0·0001	0·83 (0·73–0·95)	0·0055
INSTI (eg, dolutegravir) exposure (cumulative)*	0·79 (0·74–0·85)	<0·0001	··	··	0·40 (0·34–0·46)	<0·0001	··	··
INSTI exposure, by ART naive or experienced status*
Did not receive an INSTI	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)
ART naive	0·60 (0·54–0·67)	<0·0001	0·67 (0·61–0·74)	<0·0001	0·37 (0·31–0·44)	<0·0001	0·45 (0·37–0·54)	<0·0001
ART experienced	0·94 (0·87–1·00)	0·052	0·85 (0·80–0·92)	<0·0001	0·41 (0·35–0·48)	<0·0001	0·40 (0·34–0·46)	<0·0001
Range of first on-treatment viral load
Virological suppression (≤50 copies per mL)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)	1 (ref)
LLV (51–999 copies per mL)†	1·65 (1·58–1·71)	<0·0001	··	··	2·27 (2·05–2·51)	<0·0001	··	··
LLV range (copies per mL)†
51–199	1·49 (1·43–1·55)	<0·0001	1·48 (1·42–1·54)	<0·0001	1·85 (1·64–2·09)	<0·0001	1·80 (1·60–2·03)	<0·0001
200–399	1·78 (1·66–1·91)	<0·0001	1·73 (1·61–1·85)	<0·0001	2·39 (2·01–2·84)	<0·0001	2·31 (1·93–2·76)	<0·0001
400–999	2·09 (1·96–2·24)	<0·0001	2·03 (1·91–2·17)	<0·0001	3·68 (3·13–4·32)	<0·0001	3·54 (3·01–4·15)	<0·0001

Log-binomial analysis, unadjusted, and adjusted relative risks (RRs) using first on-treatment viral load after at least 24 weeks of ART. 95% CIs were calculated using cluster-robust standard errors to account for clustering at the facility level. Only patients with three or more viral load results (n=112 316) were included in analysis of virological failure, defined as two consecutive viral load results of at least 1000 copies per mL (figure 1). INSTI=integrase strand transfer inhibitor. LLV=low-level viraemia.

^*Patients who received INSTI as their initial ART are denoted ART naive, and patients who received non-INSTI regimens but later received INSTI are denoted ART experienced. INSTI exposure was treated as both a dichotomous variable and as a categorical variable in the univariable model, and as a categorical variable in the multivariable model. A version of the multivariable analysis was conducted using INSTI exposure as a dichotomous variable (appendix p 8).

^†LLV range was treated as both a dichotomous variable and as a categorical variable in the univariable model, and as a categorical variable in the multivariable model. A version of the multivariable analysis was run using LLV as a dichotomous variable (appendix p 8).