Application of Case–Crossover and Case–Time–Control Study Designs in Analyses of Time-varying Predictors of T-cell...IntroductionMaterials and methodsStudy PopulationIdentification of T-cell Homeostasis Failure (TCHF)Case–Crossover Design and AnalysisCase–Time–Control Design and AnalysisResultsCase–Crossover AnalysisCase–Time–Control AnalysisDiscussionREFERENCESApplication of Case–Crossover and Case–Time–Control Study Designs inAnalyses of Time-varying Predictors of T-cell Homeostasis FailureMICHAEL F. SCHNEIDER, MS, STEPHEN J. GANGE, PHD, JOSEPH B. MARGOLICK, MD, PHD,ROGER DETELS,MD, JOAN S. CHMIEL, PHD, CHARLES RINALDO, PHD, ANDHAROUTUNE K. ARMENIAN,MD, DRPHPURPOSE: To evaluate the association of sexual behavior and recreational drug exposures with T-cellhomeostasis failure (TCHF), which corresponds to the onset of a rapid decline in an individual’s Tlymphocyte count, which occurs on average approximately 1.75 years prior to an initial diagnosis ofacquired immunodeficiency syndrome (AIDS).METHODS: A case–crossover design and a case–time–control design, both nested within the MulticenterAIDS Cohort Study of 4954 homosexual and bisexual men initiated in 1983.RESULTS: In the case–crossover analysis, use of both recreational drugs and hashish were found to beprotective against TCHF (odds ratios % 0.41), based on comparisons with four earlier control periods.However, a significant decreasing trend in the prevalence of these exposures was observed over time, thusmotivating the implementation of the case–time–control design. Using the latter approach, theassociations of drug use (odds ratio Z 0.53; 95% confidence interval (CI): 0.22, 1.28) and hashish use(odds ratio Z 0.46; 95% CI: 0.20, 1.05) with TCHF were no longer statistically significant.CONCLUSIONS: The difference in inferences between these approaches demonstrates the importanceof evaluating temporal trends in exposures when using a case–crossover design.Ann Epidemiol 2005;15:137–144. Ó 2004 Elsevier Inc. All rights reserved.KEY WORDS: HIV/AIDS, Epidemiological Methods, Biological Markers, Crossover Studies.INTRODUCTIONSince the beginning of the human immunodeficiency virus(HIV) epidemic, there have been intense efforts toinvestigate factors that influence HIV disease progression,in particular those that affect the occurrence or timing ofdeath and clinical manifestations of acquired immunodefi-ciency syndrome (AIDS) (1). Despite these efforts,relatively few epidemiological and behavioral factors havebeen associated with disease progression or intermediateevents. Age at time of seroconversion is well established (2),while co-infection with hepatitis C virus or GB virus C,ethnicity, and sex are suggestive but less clear (3–7).Compared with factors associated with the developmentof clinical disease, relatively little investigation has goneinto identifying factors associated with intermediate events.The modeling of CD4Clymphocyte counts may be anexception, where the approximate linear decline of thismarker over time has facilitated the identification ofpredictors of fast or slow CD4Cdeclines over time. Thepaucity of studies using markers is partly due to the relativelyrecent understanding of intermediate pathogenesis (5, 8, 9)as well as the introduction of highly active antiretroviraltherapy (HAART), which complicates evaluation ofnatural history and marker trends. An advantage of usingintermediate biomarkers is that they may provide moreinsight than studies of clinical disease into mechanisms ofpathogenesis and may eliminate some of the variabilityinherent in clinical outcomes (e.g., opportunistic infectionprophylaxis) (10). However, substantial efforts includingstandardized, prospective follow-up, are required to obtainhigh-quality intermediate biomarker data.From the Department of Epidemiology, Johns Hopkins BloombergSchool of Public Health, Baltimore, MD (M.F.S., S.J.G., H.K.A.);Department of Molecular Microbiology and Immunology, Johns HopkinsBloomberg School of Public Health, Baltimore, MD (J.B.M.); Departmentof Epidemiology, University of California Los Angeles School of PublicHealth, Los Angeles, CA (R.D.); Department of Preventive Medicine,Feinberg School of Medicine, Northwestern University, Chicago, IL(J.S.C.); and Department of Infectious Diseases and Microbiology,Graduate School of Public Health, University of Pittsburgh, Pittsburgh,PA (C.R.).Address correspondence to: Michael F. Schneider, M.S., Johns HopkinsBloomberg School of Public Health, Department of Epidemiology, 615 N.Wolfe Street, Room E7012, Baltimore, MD 21205. Tel.: (410) 955-4320;Fax: (410) 955-7587. E-mail: [email protected] in this manuscript were collected by the Multicenter AIDS CohortStudy (MACS) with centers (Principal Investigators) at The JohnsHopkins Bloomberg School of Public Health (Joseph B. Margolick, AlvaroMun˜oz), Howard Brown Health Center and Northwestern UniversityMedical School (John Phair), University of California, Los Angeles (RogerDetels, Beth Jamieson), and University of Pittsburgh (Charles Rinaldo).The MACS is funded by the National Institute of Allergy and InfectiousDiseases, with additional supplemental funding from the National CancerInstitute. UO1-AI-35042, 5-MO1-RR-00722 (GCRC), UO1-AI-35043,UO1-AI-37984, UO1-AI-35039, UO1-AI-35040, UO1-AI-37613,UO1-AI-35041. Website located at http://statepi.jhsph.edu/macs/macs.html.Received January 20, 2004; accepted May 17, 2004.Ó 2004 Elsevier Inc. All rights reserved. 1047-2797/05/$–see front matter360 Park Avenue South, New York, NY 10010 doi:10.1016/j.annepidem.2004.05.002Beginning in the mid 1990s, a series of studies identifiedthe occurrence of a rapid decline in total T lymphocytecounts, closely corresponding to the sum of the CD4CandCD8Ccell counts, prior to the onset of clinically definedAIDS (11–14). This rapid decline, the onset of which istermed T-cell homeostasis failure (TCHF) (11), begins atvariable times after HIV infection but is more well-definedin relation to the onset of AIDS, beginning an average of1.75 years before the initial diagnosis of AIDS. A variety ofinvestigations have evaluated immunologic and virologiccharacteristics around TCHF, but no studies have in-vestigated whether modifiable epidemiologic factors are alsoassociated with TCHF (15, 16).Therefore, the goal of the present study was to un-derstand whether continued exposure to factors associatedwith the acquisition of HIV infection (e.g., sexual