Table 3. Studies in systematic review of cotrimoxazole’s effect on mortality and morbidity in adults with human immunodeficiency virus (HIV) infection receiving antiretroviral therapy (ART) .

First author, year	Study design	Follow-up duration	Cotrimoxazole		Timing of cotrimoxazole initiation relative to ART initiation	Analytical method used	Variables included in the statistical model	Losses to follow-up	How losses to follow-up were addressed	Risk of death or morbidity in individuals taking cotrimoxazole compared with those not taking cotrimoxazole
			Received (No.)	Not received (No.)
Walker, 201028	Prospective cohort study	Median duration: 4.9 years (IQR: 4.5–5.3)	1959	1220	62% of participants initiated cotrimoxazole before ART initiation, 28% initiated cotrimoxazole after ART initiation, and 10% never initiated cotrimoxazole	Marginal structural model	Baseline variables: CD4 cell count, haemoglobin level, body mass index, history of a WHO stage 3 or 4 HIV disease event, age, sex, randomization year, WHO disease stage, monitoring group and study site. Time-dependent variables: CD4 cell count, haemoglobin level, time on ART and body mass index	198 of 3 179 participants (6%) were last seen alive more than 4 months before the end of follow-up	Data from participants lost to follow-up was censored, although losses were infrequent and the effect of this censoring by additional weighting was small	Adjusted OR for death on cotrimoxazole: 0.65 (95% CI: 0.50–0.85). There were 83 deaths during 8 128 person–years of follow-up on cotrimoxazole and 105 deaths during 6 086 person–years of follow-up off cotrimoxazole. Adjusted OR for malaria on cotrimoxazole: 0.74 (95% CI: 0.63–0.88). Adjusted OR for new or recurrent WHO clinical stage 3 or 4 disease events on cotrimoxazole: 0.85 (95% CI: 0.74–0.98)
Hoffmann, 201029	Prospective cohort study	Mean duration: 0.77 years	7508	6589	Cotrimoxazole was initiated between 30 days before and 7 days after the initiation of ART	Cox proportional hazards model	Baseline variables: CD4 cell count, history of tuberculosis, WHO disease stage, sex and age. Time-updated variables: HIV RNA suppression	1186 participants (18%) not on cotrimoxazole and 901 on cotrimoxazole (12%) discontinued ART, left care or were lost to follow-up	Data from participants who discontinued ART, left care or were lost to follow-up were censored	Adjusted HR for death on cotrimoxazole: 0.64 (95% CI: 0.57–0.72). There were 558 deaths during 6 133 person–years of follow-up on cotrimoxazole and 732 deaths during 4 617 person–years of follow-up off cotrimoxazole
Lowrance, 200730	Retrospective cohort study	0.5 years	574	478	Cotrimoxazole was initiated after ART initiation	Cox proportional hazards model	Baseline variables: CD4 cell count, WHO disease stage, tuberculosis history, gender and age	22 of 574 participants who received cotrimoxazole, were lost to follow-up and 17 transferred out of the study. 51 of 478 patients not on cotrimoxazole, were lost to follow-up and 21 transferred out	Data from participants lost to follow-up were excluded from the analysis	Adjusted RR of death on cotrimoxazole: 0.59 (95% CI: 0.43–0.82). There were 57 deaths in 535 participants on cotrimoxazole and 73 deaths in 406 participants off cotrimoxazole
Madec, 200731	Prospective cohort study	Median duration: 1.08 years (IQR: 5–21 months)	Not reported	Not reported	Not reported	Cox proportional hazards model	Age, sex, WHO disease stage, CD4 cell count, haemoglobin level, body mass index, year of ART initiation, ART regimen and previous exposure to ART	37 participants (2.13%) were lost to follow-up	Participants lost to follow-up were regarded as having died	Adjusted HR for death on cotrimoxazole: 0.15 (95% CI: 0.11–0.21)
Fairall, 200832	Prospective cohort study	Median duration: 0.33 years (IQR: 1–9 months)	Not reported	Not reported	Not reported	Marginal structural model	Baseline variables: age, sex, district of residence, clinic, weight, CD4 cell count, tuberculosis status and ART status. Time-dependent variables: CD4 cell count, weight and ART status	Not reported	Not reported	Adjusted HR for death on cotrimoxazole: 0.37 (95% CI: 0.32–0.42)
van Oosterhout, 201033	Retrospective cohort study	0.5 years	319	274	Cotrimoxazole was initiated before ART initiation	Logistic regression model	Age, sex, WHO disease stage, body mass index, haemoglobin level and baseline CD4 cell count	7 participants were lost to follow-up and 13 stopped ART after 14 weeks. 16 were lost to follow-up and 15 stopped ART after 26 weeks	Not reported	After 14 weeks, the adjusted OR for death on cotrimoxazole was 0.61 (95% CI: 0.38–0.96). After 26 weeks, it was 0.71 (95% CI: 0.46–1.11)
Alemu, 201034	Retrospective cohort study	Mean duration: 2.01 years	240	31	Cotrimoxazole was initiated at or before ART initiation in 88.6% of participants	Cox proportional hazards model	Baseline variables: WHO disease stage and haemoglobin level	48 participants (18%) were lost to follow-up and 19 (7%) transferred to another facility	Data from participants who were lost to follow-up or transferred to another facility were censored	Adjusted HR for death on cotrimoxazole: 0.14 (95% CI: 0.05–0.37)
Miiro, 200935	Retrospective cohort study	1 year	179	32	Cotrimoxazole was initiated before ART initiation	Logistic regression model	ART, age, sex, baseline WHO disease stage, haemoglobin level and baseline CD4 cell count	Data from participants followed up for less than 1 year were censored	Data from 157 of the 219 participants (71.7%) were censored at the completion of follow-up	RR for all-cause morbidity on cotrimoxazole: 0.66 (95% CI: 0.41–1.06), although all-cause morbidity was undefined
Campbell, 200936	Randomized trial	0.32 years	452	384	Not reported	Not reported	Not reported	Not reported	Not reported	RR for malaria on cotrimoxazole 0.04 (95% CI: 0.01–0.17). RR for diarrhoea on cotrimoxazole 0.56 (95% CI: 0.43–0.77)

CD4 cell, CD4+ T lymphocyte; CI, confidence interval; HR, hazard ratio; IQR, interquartile range; OR, odds ratio, RNA, ribonucleic acid; RR, relative risk; WHO, World Health Organization.