Table 3.

Proportional hazards modelling of possible risk factors for infant HIV infection by 48 weeks

	Unadjusted*		Adjusted†
	Hazard ratio (95% CI)	p value	Hazard ratio (95% CI)	p value
Infant nevirapine vs control	0·47 (0·28–0·77)	0·0030‡	0·25 (0·13–0·49)§	<0·0001
Maternal antiretroviral vs control	0·58 (0·36–0·94)	0·0271	0·46 (0·27–0·81)§	0·0072
Maternal baseline CD4 count (<350 vs ≥350 cells per μL)	2·40 (1·60–3·61)	<0·0001	2·54 (1·68–3·83)	<0·0001
Maternal age (per increase of 1 year)	0·96(0·92–1·00)	0·0628	0·95 (0·91–0·99)	0·0198
Reported breastfeeding vs no breastfeeding¶	4·55 (2·02–10·3)	0·0003	9·06 (3·83–21·4)	<0·0001

^*Proportional hazards model with each risk factor as the only covariate, and stratified by nutritional group. Infant nevirapine, maternal antiretroviral, maternal baseline CD4, and maternal age included as time-independent covariates.

^†Proportional hazards model adjusted for all variables listed, stratified by nutritional group.

^‡Proportional hazards assumption violated; p<0·0001.

^§Hazard ratio estimates for infant-nevirapine and maternal-antiretroviral groups are based on interaction terms with reported breastfeeding (ie, the effect of antiretroviral treatment while breastfeeding); the hazard of infection did not differ significantly between antiretroviral and control groups when the mother reported that she did not breastfeed.

^¶Reported breastfeeding included as a time varying covariate. Breastfeeding based on mothers' reports at study visits describing infant feeding since previous visit. Reported breastfeeding as a time-dependent covariate is sometimes missing at the time of events (HIV infections), so corresponding events (HIV infections) are not counted in the proportional hazards model analysis. To avoid loss of information about events, we assumed that the infant was not weaned at the time of an event when it occurred within 1 month of the mother reporting she was still breastfeeding and no additional information was available.