Table 3:

Results of separate Cox’s models predicting relative hazard of early or late lost to follow-up and death after HAART initiation in the PART initiative, Yaoundé and Douala, Cameroon, (March 2003– January 2005)

	Death	First 6 months		Death	After 6 months
		Loss to follow-up	Death or lost to follow-up a		Loss to follow-up	Death or lost to follow-up a
Variable
CD4 counts
≥50	1.0	1.0	1.0	1.0	1.0	1.0
< 50	1.8 (0.3–10.7)	2.3 (1.4–3.7) b	2.3 (1.5–3.7) b	1.4 (0.1–16.3)	1.0 (0.4–2.0)	1.0 (0.5–2.1)
Active TB
No	1.0	1.0	1.0	1.0	1.0	1.0
Yes	....	1.8 (1–3.6) b	1.6 (0.8–3.2)	.....	2.2 (0.8–6.5)	1.9 (0.7–5.6)

Note. Data are adjusted hazard ratio (95% CI). Hazard ratios were estimated for the time-to-event outcomes of death, lost to follow-up and death or lost to follow-up using an extended Cox model allowing for a time-dependent variable (SAS PHREG Procedure). We defined for each variable in the model, a product term with a heavy side function of time (before and after 6 month) based on the pattern of the survival curves. Analyses were adjusted on site, WHO stage, previous HAART. TB, tuberculosis.

^a We performed sensitivity analysis to estimate to hazard of death in a worst-case scenario where we considered all the losses to follow-up as deaths.

^b Significant hazard ratio at 5% level of significance.