use of mean or median follow-up. However, the information is available in Tables 2-4 of our article. ' As judged from the letter, we did not point out clearly that the expansion of the data base by using "1 minus mortality in prostate cancer" as a basis for cancer-specific survival was done only to increase precision, at a cost of a possible decrease in validity. In addition, we failed to emphasize that the precision was not increased for external radiation therapy. We obtained information from 2567 irradiated patients: a weighted mean of a 74% 10-year cancer-specific survival could be based on information from 1035 patients, but "1 minus mortality" could be based solely on three reports with 132 patients. The latter were the only reports where the number of evaluable patients, mean or median follow-up time, and number of patients who died of prostate cancer during follow-up were provided. However, we suggest that the more precise and valid figure of 74% is used before the "1 minus mortality." In general, "1 minus mortality" can only be used when the mortality is low (see the Methods section),' but the transformation does not, as suggested by Levitt and Aeppli, depend on the distribution of the hazard rate of prostate cancer.
Recently, a more valid and precise method than ours has been used by Chodak et aL3 for deferred treatment. They expanded the base to 762 men, including the data we used, and enrolled original records for a combined Kaplan-Meyer estimate. The 10-year disease-specific survival rate was 87.5% for Grade 1 tumors, 85.8% for Grade 2 tumors, and 27.2% for Grade 3 tumors. If our figure of 83% is corrected for the 7.2% Grade 3 tumors in the series, an estimate almost identical to that of Chodak et al. is obtained.
Concerning the final words by Levitt and Aeppli, we stress that no "statistical exercise" in a strict sense ( P values, confidence intervals for the rate differences) was done in our article. Validity, not precision, was the critical issue if the groups were to be compared, so we abstained from statistics and used only arithmetic. Relative to the "effective sample size," this again is only of interest for the variance of a comparative figure, if such was estimated.
We share Levitt and Aeppli's concerns considering the documentation of the patient series with external radiation therapy. Again, we were obviously not clear enough in our writing, because Levitt and Aeppli apparently believe that we drew some unjustified conclusions (which?). Specifically, we did not make an overall judgment on the treatment efficacy, if any, for external radiation therapy. Moreover, rather than supporting current practice in the management of localized prostate cancer, our article implies that current opinions are influenced by potentially biased studies. Our intention was not only to draw the attention to this fact but also to point out some of the sources of bias (by compilation of available data on prognostic factors and modes of follow-up) if the treatment groups were to be compared.