AEGiS-13IAC: Increased power from inclusion of prevalent individuals in natural history studies.

13th International AIDS Conference

Durban, South Africa - July 9-July 14, 2000

Increased power from inclusion of prevalent individuals in natural history studies.

Int Conf AIDS 2000 Jul 9-14; 13:(abstract no. TuOrC370)

Geskus R;;; R. Geskus, Municipal Health Service, Nieuwe Achtergracht 100, 1018 WT Amsterdam, Netherlands, Tel.: +31 20 555 5524, Fax: +31 20 555 5533, E-mail: rgeskus@gggd.amsterdam.nl

BACKGROUND: In many HIV/AIDS cohort studies, seroconverters do not provide enough power to study natural history in case of small size effects. Often, prevalent cases constitute a substantial part of the study population, but including them requires an unbiased estimate of their date of seroconversion (sconv).

METHODS: A date of sconv is imputed by conditional mean imputation and by multiple imputation, using an estimate of the sconv distribution obtained from retrospectively identified seroconverters. Both methods are compared in a simulation study and the best performing one is applied to data from the Amsterdam cohort study among homosexual men (HOM-study). Most prevalent individuals seroconverted during the start of the epidemic, hence the effect of early infection can be studied. The effect of age at sconv and calendar period are studied in a Cox proportional hazards model. In order to study the AIDS risk of long term survivors, we looked at the shape of the hazard function.

RESULTS: In the simulation study, inclusion of prevalent individuals and using conditional mean imputation does not increase bias in survival estimates. In the HOM-study, the number of individuals used in the analysis increases from 133 (67 AIDS diagnoses until Jan. 1, 1997) to 333 individuals (208 AIDS diagnoses) after inclusion of prevalent individuals. The effect of age at sconv switches and becomes significant (RR per ten year increase: seroconverters: 0.84 (95% CI 0.60-1.17), prevalent included: 1.24 (95% CI 1.002-1.54)). Moreover, sconv before April 1985 delays progression to AIDS (RR = 0.71; 95% CI 0.50-0.99). No calendar time effect was present (p-value 0.45). After an initial increase, the hazard function flattens, but no decreasing trend was found.

CONCLUSIONS: In the Amsterdam HIV/AIDS cohort study among homosexual men, information from 200 seroprevalent individuals can safely be included in analyses of AIDS-free time. Older age and earlier date of sconv increase risk of AIDS.

Keywords: AEGIS, HIV Seropositivity, Acquired Immunodeficiency Syndrome, Homosexuality, HIV, Cohort Studies, HIV Infections, HIV Seroprevalence, Proportional Hazards Models, Power (Psychology), Natural History, Anti-HIV Agents, Bias (Epidemiology), Human, Male
000709
TuOrC370