Two issues were investigated in the present study: 1) whether there is a statistical inverse relationship between the sales of SSRIs and the suicide rates in four Nordic countries in the time period 1990-98, and 2) whether the switch from TCAs to SSRIs has had a significant impact on the suicide rate.
We found no negative association between the increasing sales figures of SSRIs and the declining suicide rates in the period using our study design. We did not find any association between the rapid decline in the sales figures of TCAs and the simultaneous decline in the suicide rates either.
The sales figures of antidepressant were slowly increasing in all four countries before 1990 (Figure 2). In contrast, the corresponding suicide rates differed between the nations, increasing in Norway and Finland, and slowly decreasing in Sweden (Figure 1). After 1999, sales figures of SSRIs are still increasing in Norway, Sweden and Denmark, but there is no significant decline in the corresponding suicide rates. Thus, there is no consistency in the associations between the trends in figures 1 and 2 before 1990. Consistency is one of the Hill's criteria of causation .
We wanted to see if a strong increase in sales figures of SSRI in one year was associated with a strong decline in suicide rates in the same year; i.e. if there was a dose-response relationship. Our analysis is inspired by the dose-response causal criteria . The analyses in Tables 2-3 are simplified versions of the Mantel-Haenszel method , which is commonly used to study dose-response relationships. Stratifying data into quartiles instead of two groups also gave non-significant results.
The associations between the curves in figures 1 and 2 can also be studied using alternative statistical methods. Correlation coefficients estimate the strength of a linear association between two variables. Regression models allow for adjustment of trends and confounding. Our method does not assume linear relationships and we do not have to model the dose-response relationship either.
Because intoxication only accounts for about 25 percent of all suicides and because intoxications are caused by many other substances than TCAs, a simple calculation shows that a 50 percent decline in the absolute sales figures of TCAs can only have a marginal effect on the overall suicide rates, if there is any at all. For example, if 20 percent of all fatal intoxications are caused by TCAs in Norway, then the predicted reduction in the suicide rate is (20-10)/4% = 2.5%. This means that the observed reduction in TCA sales figures can only explain a small part of the observed reduction in suicide rates.
The two hypotheses above are related because many people have switched from using TCAs to using SSRIs. Because the sales figures of TCAs and SSRIs are negatively correlated, one may argue that the statistical analyses are mathematically dependent. However, the correlation coefficient is only -0.36, and of fourteen TCA differences below the TCA median difference, only six of the corresponding SSRI differences were above the median SSRI differences.
During the period 1990-98 the sales figures of SSRIs increased more than the sales figures of TCAs declined, allowing more people to be treated; however, there is no association between increased sales of SSRIs and decrease in the suicide rates. In contrast, Bramness et al.  reported a small time-dependent inverse association in Norway, but when including data from the other Nordic countries, this association disappears. If increased use of antidepressants (and more patients being treated) prevents suicide, this will cause an underlying declining trend in suicide rates.
Our study of the relationship between sales of SSRIs and suicide rates is using the same data as Reseland et al. ; however, their conclusion is based on when the suicide rates started to decline and not on the direct association between sales figures and suicide rates per se. The decline in the suicide rates in Denmark and Sweden pre-dated the introduction of SSRIs by ten years or more, and the rates continued to decline thereafter . In Norway, the association was only present in the first three years after the introduction of SSRIs, and during the period when the major increase in sales of SSRIs occurred, there were no major changes in the suicide rates [2, 3]. The estimated association reported in Norway  is only valid for a small subset (less than 10 percent) of the Norwegian data. In Finland, an association between increased sales of antidepressants and reduction of suicide rates has been demonstrated, but the effect was not as strong for females as for males . In Iceland, the sales figures are the highest in the Nordic Countries, but suicide rates have been relatively constant for 50 years . Thus, there is mixed evidence that increased use of antidepressants has coincided with a reduction in suicides .
The association between the decline in suicide rates in Norway and the increase in sales figures of SSRIs is different from the other Nordic countries. Initially, the association is very strong but after three years (post 1994) there is no association. Thus, the conclusion by Bramness and colleagues , that "the fall in suicide rates in Norway and its counties was related to the increased sales of non-TCAs" seems not to be warranted. The authors claim that the effect was mostly a result of a sales increase in the lower sales segment (page 1), due to "a change from the more toxic TCAs, or heightened awareness of depression and its treatment " (page 1). However, in Denmark there was no inverse association between sales of SSRIs and suicide rates in the same period, while there was such an association after 1994. The same pattern was found in Sweden and Finland, while there was no inverse association in Norway after 1994. Further, in all Nordic countries, except Norway, associations are also present in the higher sales segment. Thus, in the present study, we have shown that the claimed association in Norway  is contradicted when using an alternative analysis and including data from all Nordic countries. In the US, an inverse association between the suicide rate and the use of SSRIs has been demonstrated for the time period 1990-2000, but more detailed analyses have revealed that the association is only valid for older and male adolescents residing in low-income regions. Furthermore, the analyses did not include factors like psycho-social interventions and substance abuse . In Australia, Hall et al.  claimed a link between increased use of SSRIs and reduced rates of suicide. Their study attracted a number of reactions, pointing at a simple shift in prescribing full-dose SSRIs instead of low-dose TCAs  and a marked pre-existing decline in elderly suicide rates , essentially related to the strong reduction in overdosing on barbiturates as a suicide method . In the words of Sakinofsky , "Hall's conclusion [...] overstates the evidence" (page 71).
The shift from TCAs to SSRIs in Denmark was similar to that in Norway . The proportions of TCAs in the total sales figures of antidepressants in Norway and Denmark in 2001 were 9.5% and 11%, respectively. The proportion of TCAs in the total sales in Norway in 1991 was 69%. We do not know the proportion in Denmark; however, before 1983, proportions were similar between Norway and Denmark . The inconsistencies between the patterns in different countries are also shown by the different trends in Norway and Sweden during the period 1974-1988. The sales figures for antidepressants were rather similar in the two countries, and slightly increasing in both cases. During the same years, suicide rates in Norway increased by approximately 70%, whereas they decreased slightly in Sweden. The reduction in the incidence of suicides caused by intoxication in Norway can explain about half of the drop in suicides in the 1990s.