In this issue of Acta Psychiatrica Scandinavica, Isacsson and Adler (1) show that the Hamilton Depression Scale does not always act as a global depression scale, i.e. that the adding-up of the scores of the individual 17 symptoms (HAM-D17) to produce a total score is not valid. They analysed the patient-level HAM-D17 data (1) from the meta-analysis by Fournier et al. (2), who had concluded that the efficacy of antidepressants was non-existent to negligible in patients with mild to moderate depression.
Over the past three decades, Acta Psychiatrica Scandinavica has released papers in which item response theory analyses of the HAM-D17 have shown that the total score is not always a sufficient statistic, most recently within the patient-level data in the STAR*D (sequenced treatment alternatives to relieve depression) (3). However, it was shown here (3) that the six-item subscale (HAM-D6) did fulfil the item response theory model, as now also demonstrated by Isacsson and Adler (1). The specific items in HAM-D6 are depressed mood, guilt, work and interests, psychomotor retardation, psychic anxiety and tiredness.
The fact that the HAM-D17 is not a global scale, but includes a specific depression subscale, implies that subscale scores rather than a global score should be employed when using this scale in trials of antidepressants. In their criticism of the HAM-D17, Moncrieff and Cohen (4) claim that the scale contains too many unspecific depression symptoms, namely three sleep items, agitation, somatic anxiety, gastrointestinal symptoms and body weight, sexual problems and hypochondriasis. According to Moncrieff and Cohen (4), any drug with sedative effects would be likely to outperform placebo on many of these items. These nine non-specific items (HAM-D9) cover the physiological arousal symptoms. The remaining two items, suicidal thoughts and insight (HAM-D2), are very important to assess separately, as they are the most idiographic symptoms in depression.
The meta-analysis performed by Fournier et al. (2) was planned to include 23 placebo-controlled trials of antidepressants in mild to moderate depression, but 17 trials were excluded because the authors of these trials were not able to provide patient-level HAM-D17 data for this meta-analysis. Among the trials not to be found in the Fournier (2) analysis is the trial performed by Paykel (5) – still to be considered as the landmark study in this field. The Paykel study was carried out in a general practice setting; however, it was an experienced psychiatrist who performed the HAM-D17 assessments during the 6-week, placebo-controlled trial with amitriptyline. The original HAM-D17 version was used, supplemented by Paykel’s Clinical Interview for Depression (5). Using the HAM-D17 total score, amitriptyline was found superior to placebo (P < 0.05) in moving the patient scores down to remission after 6 weeks of therapy, provided that their baseline HAM-D17 was between 13 and 24. For HAM-D17 scores below 13 at baseline, placebo was as effective as amitriptyline. Furthermore, Paykel (5) showed that the superiority of amitriptyline over placebo was just as powerful for baseline scores between 13 and 15 (mild depression) as for scores between 16 and 24 (moderate depression). A closer look at the individual HAM-D items assessed by the Clinical Interview for Depression approach demonstrated that the HAM-D6 items in particular discriminated between placebo and amitriptyline at endpoint (5).
Isacsson and Adler conclude (1) that it is vital to develop better measurement techniques for depression severity than the total score of the HAM-D17. However, it seems that they are in the process of throwing out the baby with the bathwater. In point of fact, the HAM-D17 should be considered as a set of triplets, covering the whole spectrum of the family of depression symptoms, namely (A): the specific HAM-D6, (B): the physiological arousal symptoms HAM-D9 and (C): the idiographic HAM-D2. When used properly (5), the HAM-D17, thus, supports the major conclusion reached by Isacsson and Adler (1), namely that the effect of antidepressants is unrelated to depression severity, i.e. is in operation in both mild and moderate depression (a HAM-D17 score going from 13 to 24).