Dr. Lagakos's current research involves a variety of issues concerning AIDS, with particular emphasis on statistical methods and analyses relating to therapies for persons infected with HIV.

One area of research interest involves statistical methods for assessing noncompliance in clinical trials and the implications of noncompliance for the interpretation of study results. The traditional "gold standard" for assessing noncompliance to study medications in clinical trials is a pharmacological assay based on blood drawn from study participants. However, because the half-life of chronically-administered drugs such as AZT is very short, these assays cannot provide a reliable assessment of the longer-term compliance of subjects. We have explored methods of assessing compliance based on routinely collected laboratory data, such as uric acid and mean corpuscular volume (MCV) of erythrocytes. Individual estimates of the time of noncompliance are obtained by extending statistical methods for change-point problems, and population estimates of the distribution of time to onset of noncompliance are based on extensions of failure-time methods to allow for measurement error. We also have investigated the consequences of noncompliance for comparing treatments in a clinical trial.

Another area of interest is the development and investigation of stochastic models to help guide treatment strategies for persons infected with HIV. By the end of 1996, there will be a variety of different types of anti-HIV drugs available to clinicians, yet how to best combine and sequence these is not clear. Because each drug's effectiveness appears to be transient, rules for when to begin and end the use of a particular drug regimen need to be established. We have evaluated the use of HIV viral load as a means of guiding treatment strategies in such an environment.

We are also working on statistical methods for comparing two or more groups with respect to their effects on repeated observations of a laboratory marker, such as the CD4 cell count, or on a recurrent clinical event such as an AIDS-defining opportunistic infection. Of particular interest are methods that make minimal assumptions and retain validity in the presence of missing data.

Selected Publications

Li Q.H. and Lagakos S.W. Use of the Wei-Lin-Weissfeld method for the analysis of a recurring and a terminating event. Statistics in Medicine 1996;16:925-9407.

Elashoff M. and Lagakos S.W. HIV treatment strategies utilizing virologic and immunologic markers as criteria for changing treatments. Statistics in Medicine 1996;15:2425-2443.

Kim H.M. and Lagakos S.W. Nonparametric inference of a failure time distribution when the failure times are estimated. Statistics in Medicine 1996;15: 2475-2490.

Kim H.M. and Lagakos S.W. Estimation of individual and population noncompliance using longitudinal marker data, with application to AIDS. Drug Information Journal 1996;29:1601S-1614S.

Volberding P, Lagakos S, et al. A comparison of immediate with deferred zidovudine therapy for asymptomatic HIV-infected adults with CD4 cell counts 500 or more per cubic millimeter. N Engl J Med 1995;333:401-07.

Gomez G and Lagakos S. Estimation of the infection time and latency distribution of AIDS with doubly censored data. Biometrics 1994;50:204-12.

Kim M and Lagakos S. Assessing drug compliance using longitudinal marker data, with application to AIDS. Statistics in Medicine 1994;13:2141-53.

Dawson J and Lagakos S. Size and power of two-sample tests based on repeated measures data. Biometrics 1993;49:1022-32.

Lagakos S, Barraj L, and De Gruttola V. Nonparametric analysis of truncated survival data, with application to AIDS. Biometrics 1988;515-23.