European perspectives in psychiatry: antidepressants as an example.

While public mental health issues have traditionally been followed also by supra-national or international organizations like WHO and EC, specific professional aspects of mental health care and disease management were so far regarded as locally specific matters, almost exclusively in the competence of local governments. Health care has been often politically directed from lower than national level of government, i.e. by cantons, counties and municipalities. This is also why it is so difficult to introduce real medicine to European agenda and why international professional associations, like EPA, have had almost zero influence on the quality of mental health care including treatment (there are no European standards or guidelines as yet) and ethics (coercion measures, involuntary treatment). Nevertheless, as many times before, the reality outran policy. Recently, article of Irving Kirsch et al. (1) triggered worldwide emotional discussion by provocative conclusion that antidepressants are generally no more effective than placebo. This was immediately grasped by journalists and anti-psychiatrists, largely publicized in media, misused by mental health care payers and misunderstood by patients and their relatives, who felt in the long term terribly cheated. So beyond our intention a paneuropean medical cause was suddenly here. This was recently reflected in several editorials of psychiatric journals (2, 3) and in a hot debate at 16th European Congress of Psychiatry, held in April in Nice, France. The debate between one of the authors (B. J. Deacon) and an opponent (H.-J. Möller) was recorded and the webcast is available on the internet (4). Let us briefly summarize how it happened. Kirsch et al. (1) report a meta-analysis of clinical trials, in which all data were available, submitted to the U.S. Food and Drug Administration (FDA) for approval of four new-generation antidepressants. Authors assessed linear and quadratic effects of initial illness severity expressed as a total score of Hamilton Rating Scale for Depression (HAMD) on clinical improvement both in subjects on active compound and on placebo, and also the difference between the active compound and placebo. This difference increases as a function of initial severity. They found that antidepressants have a limited impact relative to placebo except in samples with highest levels of depression. Authors conclude that the difference between active compound and placebo in the trials involving fluoxetine, venlafaxine, nefazodone and paroxetine increases with the initial severity of the illness being clinically significant only in the most severe cases. In general, the difference does not reach the NICE criterion for clinically significant effect size. In addition, authors suggest that the initial severity-dependent increase of response to antidepressants is attributable rather to decreasing efficacy to placebo than to the effect of antidepressants. The discussion, which followed, impugns methodology of the meta-analysis and drastically limits its interpretation. The law of initial value says that the worse is the patient at the beginning of the study, the higher is the improvement (i.e. difference between initial and final measure). At the same time, however, we often see that the worse is the patient at the beginning the worse is also at the end of the study. In other words, the correlation between the initial value (of a scale) and the improvement is positive, but the correlation between the initial and final value is positive as well. This could be avoided by expressing this relationship as a ratio of final and initial value, which is unfortunately only exceptionally used. In the discussion in PLoS I wondered how the results of the meta-analysis would change if authors took into account not the difference, but only a final value of HAMD. Could it significantly change the conclusion? It would be clinically more relevant because we do not care as much how the patient improved in a study but rather how he ⁄ she is finally doing! P.J. Leonard responded to this challenge, recalculated the results of the metaanalysis and published it on his blog (5). And then, behold some of Kirsch evaluated antidepressants do reach NICE recommended effect size, which is clinically relevant (the difference in response to a drug and to placebo higher than 3 points of HAMD total score). After all, some other opponents (2) doubt the concept of clinically relevant Acta Psychiatr Scand 2008: 118: 89–90 All rights reserved DOI: 10.1111/j.1600-0447.2008.01218.x Copyright 2008 The Author Journal Compilation 2008 Blackwell Munksgaard