Premium
A method for deciding early stopping of inconclusive case‐control studies in settings where data are stratified
Author(s) -
Strömberg Ulf
Publication year - 1997
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/(sici)1097-0258(19971030)16:20<2327::aid-sim663>3.0.co;2-y
Subject(s) - continuation , early stopping , computer science , contrast (vision) , econometrics , confounding , statistics , control (management) , mathematics , artificial intelligence , artificial neural network , programming language
Abstract Statistical methodology for early stopping of clinical trials is well developed. However, in epidemiologic studies, methods for deciding early stopping have rarely been considered. In contrast to clinical trials, termination due to an early significant result is indeed seldom relevant, because further expenditure is then likely to be worthwhile in order to obtain a more precise effect estimate and to more precisely account for confounders. On the other hand, if data obtained early indicate an inconclusive final result and if continuation of the epidemiologic study is expensive, then it might be desirable to stop the study early. The present paper proposes a method for evaluating the possibility that an ongoing case‐control study will produce an inconclusive final result. The method is developed for stratified case‐control data. In principle, probable final results, given the available data, are predicted by repeatedly simulating the remaining data using plausible assumptions on the true exposure effect. These assumptions involve the current effect estimate as well as some alternatives, chosen with respect to the investigator's prior hypothesis concerning a harmful exposure effect, which are in reasonable agreement with the available data. © 1997 John Wiley & Sons, Ltd.