Testing For Center E ects In Multicenter Survival Studies: A Monte Carlo Comparison Of Fixed And Random E ects Tests

SUMMARY The problem of testing for a center eeect following a proportional hazards regression is considered. Two approaches to the problem can beused. One approach ts a proportional hazards model with a xed covariate included for each center. The need for a center speciic adjustment is evaluated using either a score, Wald or likelihood ratio test of the hypothesis that all the center speciic covariates are equal to zero. An alternative approach is to introduce a random eeect or frailty for each center into the model. Recently, Commenges and Andersen 11, have proposed a score test for this random eeects model. By a Monte Carlo study we compare the performance of these two approaches when either the xed or random eeects model holds true. The study shows that for moderate samples the xed eeects tests have nominal levels much higher than speciied, but the random eeect test performs as expected under the null hypothesis. Under the alternative hypothesis the random eeect test has goodpower to detect relatively small xed or random center eeects. Also if the center eeect is ignored the estimator of the main eeect may be quite biased and the estimator is inconsistent. The tests are illustrated on a retrospective multicenter study of the recovery from bone marrow transplantation.