Reply

We appreciate the thoughtful comments of Vinceti et al.

Vinceti et al. claim that our data show an attenuation of the reduction in risk of prostate, colorectal, and lung cancer resulting from 25 months of additional follow-up. They point to excesses of some cancer sites associated with selenium treatment (breast cancer, melanoma, head and neck cancer, lymphoma and leukemia).

They also point to an excess of cancer risk in selenium-treated patients with the highest plasma selenium levels at baseline. The effects, associations, or modifications to which Vinceti et al. draw attention are not in every case statistically significant. Thus, there does appear to be an attenuation of the reduction in incidence of prostate, colorectal, and lung cancer. Whether this is a chance finding is not clear. Our analyses did not show a statistically significant change over time in the effect of treatment. Similarly, the “excesses” that Vinceti et al. note are not statistically significant; as nearly as we can tell, these “excesses” are well within what one would expect to see given no association of selenium treatment and increased risk. Vinceti et al. refer to a reported excess risk in selenium-treated patients with the highest baseline plasma selenium levels. Clearly, as they observe, the point estimate of relative risk associated with selenium treatment for those with the highest baseline levels is above 1.0. The 95% confidence intervals for that relative risk, however, extend well beyond 1.0; the relative risk observed could well occur even with no association of plasma selenium treatment in the risk of cancer.

On the other hand, the observation of Vinceti et al. that the reduction in risk of cancer is confined to males is true; we find no alteration of risk among women, and the difference in relative risk between men and women is statistically significant. Similarly, the interaction between baseline plasma selenium levels and the impact of selenium treatment is statistically significant. Those with the lowest selenium status at baseline profited from selenium treatment; the others did not.

Vinceti et al. suggest that further follow-up from this experimental cohort beyond the end of active intervention would be of great interest. We have in fact followed this cohort from February of 1996 through the fall of 2002 and are evaluating the impact of treatment assignment after the period of blinded ascertainment ended. Those results will be forthcoming.

Vinceti et al. suggest that this trial offers opportunities to evaluate the possible occurrence of toxicity in people. We are in the process of analyzing the toxicity data. Some individuals were on selenium treatment (200 μg/day) for over 10 years; we are analyzing the data now and will have a report on it in the very near future. Vinceti et al. suggest that toxicity may depend on the form of selenium exposure. The selenium used in this experiment was largely selenomethionine; this organic form is probably less likely than some others to induce toxicity. We have blood specimens that were collected yearly from the beginning of selenium treatment to the end of the study, so we should have the ability to examine the toxicity that Vinceti et al. think is a possibility in a great deal of detail.

We appreciate these comments and will be doing our best to provide answers to the questions raised by Vinceti et al.