ACTG 398 Longitudinal Viral Load Data for Semiparametric Time-Varying Coefficients Regression Models (Sun and Wu, Scan. J.  Statist. 2005).


Multicenter, randomized, 4-arm trial, double-blind and placebo-controlled for second PI, conducted between October 1998 and April 2000, for which there was a 24-week primary analysis with extension to 48 weeks. Thirty-one participating AIDS (acquired immunodeficiency syndrome) Clinical Trials Units in the United States. A total of 481 human immunodeficiency virus (HIV)-infected persons with prior exposure to a maximum of 3 PIs and viral load above 1000 copies/mL. Selectively randomized assignment (per prior PI exposure) to saquinavir (n = 116); indinavir (n = 69); nelfinavir (n = 139); or placebo twice per day (n = 157); in combination with amprenavir, abacavir, efavirenz, and adefovir dipivoxil. The main findings from this study are reported in Hammer et al (2002).

Sun and Wu (2005) applied the semiparametric time-varying coefficients regression model to the viral load data of this study. More details can be found from the paper.



Sun, Y. and Wu, H. (2005), Semiparametric Time-Varying Coefficients Regression Model for Longitudinal Data, Scan. J.  Statist., 32, 21-47.

Hammer SM, Vaida F, Bennett KK et al. (2002), Dual vs single protease inhibitor therapy following antiretroviral treatment failure: a randomized trial. JAMA, 288, 169-180.

Definition of Variable Names:


patid: patient id.

nnrti: 1=NNRTI-experineced 0=NNRTI-naive.

trtarm: treatment arms. All patients received two NRTIs (abacavir and adefovir), one NNRTI (efavirenz), and at least one PI (amprenavir). In addition, arms A, B, and C received a second PI (saquinavir, indinavir, and nelfinavir, respectively) and arm D received a placebo.

calwk: the calculated scheduled visit week.

txday: number of days on treatment.

logrna: log10 HIV RNA.

cens:  censoring indicator, 0=actually observed, 1=censored below, 0=censored above.