Dr. McMahon Replies

To The Editor: We agree that very little is known about treatment-emergent suicidal ideation as a phenotype. Since it is an uncommon and transient event, treatment-emergent suicidal ideation cannot be assessed by the usual genetic epidemiologic methods of family and twin studies. Thus, as Dr. Levinson correctly observes, treatment-emergent suicidal ideation has not been shown to be a heritable phenotype. He states that the lack of data on heritability indicates that “the prior probability of [a genetic] association is very low.” This may be true, but the same could be said even for conditions in which heritability has been clearly established; for example, experience with disorders such as schizophrenia and autism demonstrates high heritability of a trait that does little to ensure that genetic associations will be valid (1) . In contrast, much less heritable conditions such as type 2 diabetes have produced several robust genetic associations (2) . For tests of association, it is not so much the general heritability (of the liability) but rather the heritability attributable to a specific marker that is relevant. Thus, the relationship between the heritability of the trait and valid genetic associations is, in our view, not clear.

We defined subjects with treatment-emergent suicidal ideation as those participants who initially denied suicidal ideation but then endorsed it during treatment, which is the definition commonly used by regulatory agencies and in the literature. (Participants who endorsed suicidal ideation both initially and during treatment were not considered to be subjects with treatment-emergent suicidal ideation and were indeed included—as Dr. Levinson correctly assumes—among the set of comparison subjects who endorsed suicidal ideation at the initial visit.) Dr. Levinson questions whether some of the participants who met our case definition might have been “embarrassed or afraid to admit suicidal ideation at a first visit with a new treatment team.” While this is a possibility, it seems to us unlikely. As shown in Table 2 of our article, treatment-emergent suicidal ideation subjects showed no general tendency to deny symptoms, since they had baseline symptom scores that were similar to those of the other participants. Moreover, they often endorsed other potentially embarrassing symptoms such as marital discord and sexual dysfunction (data not shown in the article). More detailed, longitudinal studies of suicidality during treatment may shed some light on this issue, but suicidal ideas, similar to most psychiatric symptoms, are fundamentally a subjective phenomenon. We are all limited by our patients’ ability to reveal to us the contents of their conscious minds (3) .

Dr. Levinson questions the decision of the NIH Office of Technology Transfer to license the markers reported in our study for commercial development. Such licensing gives the NIH some control over how the markers are used commercially. All data produced by laboratories within the NIH Intramural Research Program are the property of the people of the United States. The professionals in the Office of Technology Transfer have devoted their careers to protecting and managing this common property for the public good. We respect their decision.

However, we agree with Dr. Levinson that it is premature to introduce a test based on these results to the clinic until they are independently replicated. Independent replication serves two vital roles for genetic association findings: 1) verification of true positive associations and 2) better estimation of the true effect size. Experience and statistical theory show that highly significant p values alone are poor indicators of true associations and that the first study to detect an association will typically overestimate the effect size—the so-called winner’s curse (4) . Thus, independent replication is the essential next step.

But is independent replication sufficient to justify offering a genetic test in the clinic? What other criteria should be applied to research findings in judging their readiness for clinic use? Should we withhold from patients access to genetic information that could help prevent bad outcomes?

Questions such as these will arise with increasing frequency and urgency in the near future (5) . We submit that it is now time for the field of psychiatry to begin an active debate on the issue of clinical genetic testing. Criteria will probably differ for tests intended to predict severe adverse outcomes, tests intended to identify patients most likely to improve with treatment, and tests intended to support a clinical diagnosis. In any case, we as a profession need to develop some guidelines as to what clinical genetic tests should be used, when psychiatrists should offer them, and how they should be interpreted in the context of diagnosis and treatment. If we fail to act promptly, then the marketplace will fill the vacuum, which has already begun to occur in other fields of medicine, and psychiatrists may lose the initiative in a debate in which the outcome could have real consequences for our patients and their families.