Professor Isacsson raises an important issue concerning the interpretation of national suicide data before and after the introduction (in 1969) of a new classification of deaths, ‘injury undetermined whether accidentally or purposely inflicted’ (ICD–8). The points he raises do not, however, invalidate our conclusions.

The exclusion of pre-1969 or even pre-1979 (the period when the use of suicide and undetermined categories had stabilised in Sweden) data from our analyses does not alter our main finding that suicide reductions in three of the four Nordic countries preceded the widespread use of SSRIs in the early 1990s. With the exception of Sweden, suicide rates continued to increase, rather than decrease, in the period 1969–79 in the Nordic countries, indicating that the changed classification had a minor impact on apparent trends in these countries.

There are well-recognised problems with interpreting ecological data to infer causal effects. Isacsson cites data from a number of countries where declines in suicide rates have coincided with increased antidepressant prescribing. However, data from other countries, such as England and Wales, Ireland and Italy, demonstrate the opposite pattern (Reference Gunnell and AshbyGunnell & Ashby, 2004). Professor Isacsson suggests that the reduction in suicide rate in Sweden in 1979–89, prior to the use of SSRIs, may be a result of the increased use of tricyclic antidepressants. This is possible, but data from Norway suggest that increased use of non-SSRI antidepressants in the 1970s and 1980s was associated with increases in suicide rates.

Isacsson suggests that the stabilisation in the decline in suicide rates is expected because not all people with depression consult doctors and conditions other than depression contribute to overall suicide numbers. We agree with this analysis. Nevertheless, the continued year-on-year rise in antidepressant use in the study period does indicate a wider population of individuals, presumably some of whom are at risk of suicide, being treated by these drugs.

Our assessment of suicide and antidepressant prescribing in the Nordic countries was more comprehensive than Isacsson's original analysis and in our view provides weaker evidence than that originally presented (Reference IsacssonIsacsson, 2000). Nevertheless the most comprehensive assessment of the ecological data to date (Reference Ludwig and MarcotteLudwig & Marcotte, 2005) does support Isacsson's view. In an area where the influence of the pharmaceutical industry is widespread we favour a more cautious interpretation of the ecological data.