Biometrics 62, 893–900 September 2006 DOI: 10.1111/j.1541-0420.2005.00516.x A Comparison of Eight Methods for the Dual-Endpoint Evaluation of Efficacy in a Proof-of-Concept HIV Vaccine Trial Devan V. Mehrotra, 1, ∗ Xiaoming Li, 1 and Peter B. Gilbert 2 1 Merck Research Laboratories, UN-A102, 785 Jolly Road, Blue Bell, Pennsylvania 19422, U.S.A. 2 Fred Hutchinson Cancer Research Center and Department of Biostatistics, University of Washington, Seattle, Washington 98109, U.S.A. ∗ email: devan mehrotra@merck.com Summary. To support the design of the world’s first proof-of-concept (POC) efficacy trial of a cell-mediated immunity-based HIV vaccine, we evaluate eight methods for testing the composite null hypothesis of no- vaccine effect on either the incidence of HIV infection or the viral load set point among those infected, relative to placebo. The first two methods use a single test applied to the actual values or ranks of a burden-of-illness (BOI) outcome that combines the infection and viral load endpoints. The other six methods combine separate tests for the two endpoints using unweighted or weighted versions of the two-part z, Simes’, and Fisher’s methods. Based on extensive simulations that were used to design the landmark POC trial, the BOI methods are shown to have generally low power for rejecting the composite null hypothesis (and hence advancing the vaccine to a subsequent large-scale efficacy trial). The unweighted Simes’ and Fisher’s combination methods perform best overall. Importantly, this conclusion holds even after the test for the viral load component is adjusted for bias that can be introduced by conditioning on a postrandomization event (HIV infection). The adjustment is derived using a selection bias model based on the principal stratification framework of causal inference. Key words: Burden of illness; Causal inference; Cell-mediated immunity; Fisher’s test; HIV vaccine; Multiple endpoints; Principal stratification; Selection bias; Simes’ test. 1. Introduction More than 20 million people worldwide have died of AIDS since the first cases were identified in 1981, including 3 million deaths in 2004 alone. An estimated 40 million people are cur- rently living with HIV/AIDS, and approximately 15,000 new HIV infections are being added each day (UNAIDS, 2004). An efficacious prophylactic HIV vaccine (administered to HIV uninfected persons) is urgently needed. The first-generation candidate HIV vaccines, developed in the 1980s and early 1990s, were designed to prevent HIV acquisition by stimulating anti-HIV antibodies. However, antibody-based vaccines failed to lower the rate of HIV in- fection compared to placebo in the first two large-scale HIV vaccine efficacy trials (The rgp120 HIV Vaccine Study Group, 2005). The absence of protection has been explained, in part, by the inability of the tested vaccines to elicit antibodies that neutralize HIV particles freshly sampled from popula- tions (Burton et al., 2004). Due to HIV’s expansive genetic diversity and its many mechanisms of evading neutralization, development of an effective antibody-based HIV vaccine has proven to be an extremely difficult task. Second-generation HIV vaccine candidates have been de- signed not to elicit humoral immune responses (antibodies), but rather to elicit cell-mediated immune (CMI) responses (Graham, 2002). These candidates are motivated by increas- ing evidence that CMI responses, mediated primarily by CD8+ cytotoxic T lymphocytes, play a key role in the con- trol of acute and chronic HIV infection (Borrow et al., 1994; Shiver et al., 2002). To establish the efficacy of an antibody-based HIV vaccine in a randomized, placebo-controlled clinical trial, it would suf- fice to demonstrate a statistical difference in the HIV infec- tion rates between vaccine and placebo recipients. But how does one establish the efficacy of a CMI-based HIV vaccine? Vaccine-induced CMI responses (unlike antibody responses) are not expected to impact the initial entry of host cells by HIV. However, they could abort an infection before it be- comes fully established (implying a negative HIV diagnostic test), or contain the viral load at a low “set point” in people who become infected despite vaccination. As noted in Gilbert et al. (2003b), the latter outcome would likely provide sub- stantial clinical benefit by preventing or delaying the onset of AIDS, and would decrease the rate of secondary transmission of HIV. These considerations support the use of HIV infection and viral load set point as co-primary endpoints in an efficacy trial of a CMI-based HIV vaccine. The first proof-of-concept (POC) efficacy trial of a CMI- based HIV vaccine began enrolling volunteers in December 2004. This groundbreaking trial is being conducted by Merck Research Laboratories, in collaboration with the HIV Vaccine Trials Network and the Division of AIDS in the U.S. Na- tional Institutes of Health. The candidate vaccine, developed C  2006, The International Biometric Society 893