1) In the context of a study, the quality of a measurement (e.g. the mean estimate of a treatment effect) that is correct or that reflects the actual effectiveness of the treatment.
Note: Not to be confused with preciseness: an estimate can be accurate, yet not be precise, if it is based on an unbiased method that shows great random variations (severity of the disease, coexisting conditions).
2) In the context of a diagnostic test, the proportion in which the results correspond to those of the chosen refererence test, i.e. the sum of the true positives and the true negatives, divided by the size of the sample of the population studied.
(Related concepts: sensitivity and specificity).