TABLE 3.

Effect of HAART^* versus no HAART on change in HIV-1^* RNA viral load under a series of models for the construction of inverse probability weights, Multicenter AIDS^* Cohort Study and Women's Interagency HIV^* Study, 1996–2005

Specification	Description	Estimated weights		Difference in viral load, log10 copies/ml
		Mean (SD*)	Minimum/maximum	Estimate	SE*†
1	Numerator includes linear terms for baseline CD4, RNA, and time. Denominator includes linear terms for baseline CD4, RNA, time, CD4−1, and RNA−1.	1.07 (1.47)	0.03/26.4	−1.94	0.170
2	Numerator and denominator are as in step 1 but replace linear terms for baseline and time-varying CD4 and RNA with step functions (i.e., categories‡).	1.05 (0.65)	0.11/16.6	−1.66	0.104
3	Numerator and denominator are as in step 1, with three-knot splines to all linear terms.	1.05 (1.17)	0.03/37.0	−1.91	0.139
4	Numerator and denominator are as in step 3, plus a product between CD4−1 and time in denominator.	1.04 (1.15)	0.03/46.8	−1.91	0.132
5	Numerator and denominator are as in step 4, plus three-knot splines for CD4−2 and RNA−2 in denominator.	1.04 (1.67)	0.03/83.5	−1.95	0.133
6	Numerator and denominator are as in step 4, plus indicators for AIDS-1 and presence of HIV symptoms−1 in denominator.	1.05 (1.37)	0.03/68.4	−1.91	0.130

^*HAART, highly active antiretroviral therapy; HIV-1, human immunodeficiency virus type 1; AIDS, acquired immunodeficiency syndrome; HIV, human immunodeficiency virus; SD, standard deviation; SE, standard error.

^†The standard deviation of 500 nonparametric bootstrap sample estimates; 500 always converged.

^‡The CD4 count categories were as follows: <200, 200–500, >500 cells/mm³; and viral load was detectable at 400 copies/ml or not.