Our main limitation is thus confounding by indication, where higher risk of suicide deaths during time periods with SSRI may depend on fluctuations in severity of the underlying L-AP4 depression that is related to SSRI-use rather than to the SSRI-use in itself. It is not possible to establish to what extent confounding by indication affects our overall risk estimates. In an attempt to explore this confound we performed the same study with TCAs instead that may be less likely to produce an ��activation syndrome��. Since we obtained an OR of 2.00 for TCA these results could be Leelamine hydrochloride interpreted as slightly stronger evidence in favor of a serotonergic mechanism involved in at least a small subset of suicides. However, the descriptive question on whether there is an increased risk of suicide in the early phase of treatment, which is of great clinical relevance, is not confounded by indication. The bias only interferes with our understanding of the causation of this increased risk. However, the peak in the risk estimate in the second and third week after initiation in part could speak against confounding by indication, as the suicide risk rather would be anticipated to decrease with time. Also, the suicide peak in day 8 to 11 and in day 12-15 argue both in favor of the activation syndrome theory and of the biological mechanism through alteration in serotonin levels. We avoided seasonal effects by choosing the control period 364 days prior to the suicide. This approach also ensures that the index date in the control period occurs on the same week-day as the suicide. Despite this, we cannot rule out the effect of potential differences in the severity of the underlying psychiatric disease in the compared time intervals from a potential triggering effect of SSRI initiation, as discussed above. Although SSRIs are the first-line treatment for major depression, 30�C40% of the patients do not show a significant response. If so we might overestimate the risk since some of the suicides might be due to lack of response rather than due to a biological mechanism from SSRI or activation syndrome. However, a possible lack of response would most probably not change over time, which leads us to believe it is not a major problem in our study. To avoid changes and trends in classification we included deaths coded as undetermined intent which is common when studying suicide in Europe. Our subdivided analysis showed however higher ORs among certain suicides compared to undetermined intent. The most common suicide method among deaths coded as undetermined intent is poisoning, i.e. non-violent, whereas among certain suicide around 80% are violent. We regarded a dispensed prescription of SSRI within 28 days of index date and no dispensed prescription of SSRI in the preceding 4 months as being exposed to SSRI initiation. However, there was no information on when or whether the patients had actually taken their medication, which is a limitation.