Anestis et al. Respond

We read the letter by Stuart et al. eagerly, because we admire their work and appreciate useful feedback. The authors’ primary point appeared to be that regression is an imperfect analytic approach, a point with which we agree. As the authors themselves noted, however, we explained this openly within the article, much as the authors discussed the shortcomings of their analytic decisions in the work they promoted in their letter. The authors noted, “although Anestis et al. include language about the need to take their results with caution, it is all too easy for readers to set those warnings aside.” We agree that readers, at times, overlook limitations and arrive at conclusions that exceed what is appropriate or are even in conflict with the data; however, we do not believe that the limitations of our study were more extensive than those of other studies on this topic and in this Journal. We did not find any reason to infer that readers are more likely to make this error based on our work than in the work of Stuart et al. or others.

If the authors’ point was that they have used more “sophisticated” analyses in their own work, here too we would agree. We also have used more sophisticated analyses when we believed that the question at hand called for such an approach. If the authors’ point was that more sophisticated analyses are superior to regression, however, then, here we disagree. The appropriateness of the approach depends on the question being asked, with some questions offering several equally useful and imperfect options. We believe that our approach represented an entirely appropriate follow-up to Anestis and Anestis,¹ extending that work through the inclusion of an additional year of data and several additional meaningful covariates (e.g., gun ownership, past-year rates of depressive episodes) while using the same analytic approach as in the original study. The authors wish we had taken a different approach, and we can certainly respect that. This disagreement in approach, however, in no way alters our findings or changes our interpretation of the results.

We agree with the authors that it would be highly inappropriate to conclude, based on our findings, that the laws in question explain the entirety of the differences in suicide rate changes across states between 2013 and 2014. Without the ability to experimentally manipulate law status, determining the specific effect of laws on suicide (and variability of that effect from state to state) remains impossible. However, we noted these points in our article. Furthermore, this is true not only with regression analyses but also with the analytic approach promoted by the authors. Despite this, we wholeheartedly believe that such work, which provides evidence supporting (or disputing) the potential utility of legislation in suicide prevention efforts, is vital.

We believe in the importance of good science and are glad that Stuart et al. share our opinion on this front. We also believe that suicide prevention moves forward as a field when multiple independent groups vigorously test hypotheses on similar topics, submit their work to peer review, and adjust their vision as the data dictate that they do so. This has and will continue to be our approach to what we believe to be a sorely underresearched issue: the role of firearms in suicide. We look forward to pursuing our shared research goals with Stuart et al. moving forward and are hopeful that future discussions of this sort will prove more productive.