To the Editor: Mario Maj, M.D., Ph.D., raises two caveats regarding our report—that the statistical power was small and that bias may have resulted from the inclusion of some patients who had discontinued lithium previous to study entry.

To illustrate the first concern, Dr. Maj noted that the Tondo et al. group elsewhere described an extended study group in which patients experienced a significantly higher proportion of time ill during the retreatment period. Because this study group was mentioned as "unpublished data" in a review chapter published in the same year as the Tondo et al. article, the methods used to acquire and describe the additional subjects are difficult to critique. Dr. Maj’s point, though, does raise a methodological issue to which our report only alluded.

Study groups drawn from treatment-seeking populations are biased toward relative illness severity and persistence. By the same token, patients who are attending a clinic more frequently are more likely to be in a problematic phase of their illness than those with less frequent visits. If the group described by Tondo et al. was drawn from all of those "attending" a clinic in a given period of time, those who were experiencing a relatively difficult period in their illness would be overrepresented, and this would produce the impression that current therapies are less effective than past therapies. The prospective ascertainment of illness course provided with our data avoids this problem in that patients were tracked regardless of whether, or how often, they continued to seek treatment.

The limited statistical power in our data would have been notable had trends existed toward longer times for recovery in the second treatment phase. A trend in the opposite direction was apparent, however; nearly 50% of the patients had recurrences within 2 years in the first well period, but just over 30% had recurrences within that time in the second well period.

It may be, as Dr. Maj’s second point implies, that the first lithium discontinuation—although not subsequent discontinuations—increases the likelihood of subsequent therapeutic resistance. We are not aware that this has been asserted in the pertinent literature, however.

We do not wish to prematurely dismiss the possibility that lithium discontinuation has long-term effects on lithium responsiveness. Given the biases likely to be operating in the reports of this effect, though, its application in clinical management would be premature at this point. Other prospectively observed study groups of patients with bipolar disorder exist, and the data from those studies could help settle this question.