Efficient use of biological banks for biochemical epidemiology: exploratory hypothesis testing by means of a sequential t-test.

In view of recent advances in molecular and biochemical epidemiology, there is growing interest in the creation of biological banks of blood, urine, tissue, or other biological specimens collected from participants in prospective cohort studies. The existence of biological banks may make it possible to study a multitude of etiologic hypotheses, by comparing biochemical parameters measured in the biological specimens of subjects who will eventually develop the disease of interest ("cases") and of control subjects, using a nested case-control or a case-cohort design. In practice, however, the amount of biological material available per subject (in particular, that of cases) will limit the number of hypotheses that can be tested. The present paper discusses the use of a sequential t-test which, compared with an analogous fixed sample procedure, will on average require fewer biological specimens before a given study hypothesis can be accepted or rejected. The sequential test should thus facilitate an early decision on whether a new hypothesis is worth further investigation, while avoiding wasting too much biological material on testing hypotheses that may eventually prove unfruitful. If the test reveals an exposure difference of interest, the study may be extended so that relevant epidemiologic effect measures can be estimated more accurately.