In Reply.—We appreciate Dr Brown's interest in our recent report. We are aware of the problems raised by him, problems that occur in any retrospective clinical study and for which there are no completely adequate solutions. The problem of transfer bias does not seem to us to be as severe a problem as he describes since the transfer was based on status of a patient's respiratory disease and was not related to the severity of the patient's eye disease. In fact, transfer occurred before eye disease status was known. In addition, transfer decisions were made in a reasonably uniform manner for all patients.
The question of the appropriate number of df for the t test raises the question of how to handle an imperfect data set. Our system of obtaining weighted average values for each of the 92 patients was an attempt to make the most appropriate use we