We are pleased, but not surprised, that the reanalysis of the National Cancer Institute study by Walker, et al,' has yielded the same findings that we reported2 3. Our preliminary analysis showed no evidence of any association between bladder cancer risk and past consumption of artificial sweeteners (AS) in the total study population. However, we noted a slight tendency toward increased risk with increased intensity of use (amount of AS used daily); this trend was much more prominent in the two subgroups chosen before the start of the study for particular attention, i.e., nonsmoking females and heavy-smoking males. Walker and his colleagues, using a slightly different statistical methodology, has faithfully reproduced these results. Our interpretation of these results was that "past AS use has had a minimal effect, if any, on bladder cancer rates." We reported that "inconsistencies in the data suggest that the positive associations may be due to chance, but that it is noteworthy that the subgroups were chosen, a priori, to test hypotheses derived from laboratory experiments."3 The Walker interpretation resembles ours, although some differences exist. We feel that either a causal association or chance may be responsible for the positive findings in the subgroups, while the Walker research team favors chance as the explanation. We do not agree with their argument against a causal interpretation based on a "risk score" analysis of our data. Indeed, the findings from two other studies4,5 suggest that a causal explanation cannot yet be dismissed. Walker, et al, claim that "control for a variety offactors through multivariate techniques diminished the plausibility of earlier interpretations" of the subgroup findings. This apparently refers to the lack of any evidence of a relationship between risk and intensity of AS use in either the lowest or highest "risk category," with the groups defined on the basis of a multivariate "risk score." Although this is apparently at odds with the associations seen in our defined high-risk and low-risk groups, the observations are actually compatible with each other. We chose non-smoking, non-occupationally exposed females a priori because their baseline risk of bladder cancer is low and there were enough cases to evaluate the effect of AS use. Similarly, we chose heavysmoking males as a high-risk group a priori. Evaluation of the bladder cancer risks of these two groups after the data were collected verified these designations. The baseline risk