Per Bech and colleagues present an independent reanalysis of data from the Genome-based Therapeutic Drugs for Depression (GENDEP) study, which suggests that escitalopram may be superior to nortriptyline in its effect on core depressive symptoms. This differs from our conclusion that the two antidepressants do not differ in overall antidepressant efficacy [1]. We welcome this opportunity to explain why the two analyses may give different results.

GENDEP included 811 treatment-seeking adults recruited in nine European centres and diagnosed with major depressive disorder of at least moderate severity. Whereas 468 were randomly allocated to receive either nortriptyline or escitalopram, the other 343 individuals were given the antidepressant that was judged preferable based on relative and absolute contraindications [1]. We measured depression severity weekly for up to 12 weeks with three depression rating scales: Montgomery-Åsberg Depression Rating Scale (MADRS), Hamilton Rating Scale for Depression (HRSD) and Beck Depression Inventory (BDI). None was declared the primary outcome measure a priori. Psychometric analysis indicated that MADRS measures depression severity most accurately and therefore we used MADRS as the primary outcome measure. We analyzed the outcomes with linear mixed-effect models for repeated measures which use available information at all weeks whilst controlling for baseline severity of depression and accounting for clustering of individuals within centres. We found no significant differences in antidepressant efficacy on any of the three outcome scales [1].

Bech and colleagues' analysis differed from our own in several important respects. First, they decided to focus exclusively on the HRSD, although it was the scale with the least satisfactory psychometric properties [2]. Second, they excluded individuals with incomplete baseline data on any scale, leading to a reduced sample of 765. Third, they defined remission as a score of five or less on a six-item core subscale of HRSD. This corresponds to a total HRSD score of 9 or 10 [2], which is higher than the conventional cut-off score for remission of 7. Fourth, they analyzed the data with last observation carried forward, a method which systematically biases the results in favour of a treatment that has smaller drop-out rates. They also used mixed-effect models, but it remains unclear whether centre effects were taken into account. Bech and colleagues report results stratified by baseline HRSD and omit 38 individuals who they classified as ‘doubtful depression’ (although they received a diagnosis of at least moderately severe depressive episode in a diagnostic interview).

We think it is likely that these differences in analytic approach fully explain why the results of Bech and colleagues are at variance with our own.

A slight advantage of escitalopram for core depressive symptoms is in keeping with our results for the ‘observed mood’ symptom dimension [1]. Failure to take into account non-random allocation also favours escitalopram as nortriptyline was non-randomly given to more severe cases. The last observation carried forward approach coupled with larger drop-out rate among nortriptyline-treated individuals also favours escitalopram, as illustrated in the supplementary material to the original report [1]. It appears that these biases have joint forces to give an apparent advantage to escitalopram over nortriptyline in the analyses presented by Bech and colleagues.

We therefore hold firm in maintaining that the original GENDEP analysis was preferable. The relative merits of total scale or ‘core’ subscales may depend on antidepressant used [3]. Whereas escitalopram may have a slightly better effect on core mood symptoms, nortriptyline helps more with sleep [1, 3], an effect that many patients value highly. However, there is a twist to the story. A recent reanalysis of GENDEP that considers informative drop-out also suggests an efficacy advantage for escitalopram [4]. This demonstrates the complexity of comparing the efficacy of two drugs that differ in tolerability. However, in practice we will continue using serotonin reuptake inhibitors, including escitalopram, as first-line treatment options in most cases of depression, not because of efficacy, but because of better tolerability and safety in overdose compared with tricyclic antidepressants.


  1. Top of page
  2. References
  • 1
    Uher R, Maier W, Hauser J et al. Differential efficacy of escitalopram and nortriptyline on dimensional measures of depression. Br J Psychiatry 2009; 194: 252259.
  • 2
    Uher R, Farmer A, Maier W et al. Measuring depression: comparison and integration of three scales in the GENDEP study. Psychol Med 2008;38:289300.
  • 3
    Boessen R, Groenwold RH, Knol MJ, Grobbee DE, ROES KC. Comparing HAMD(17) and HAMD subscales on their ability to differentiate active treatment from placebo in randomized controlled trials. J Affect Disord 2013;145:363369.
  • 4
    Power RA, Muthen B, Henigsberg N et al. Non-random dropout and the relative efficacy of escitalopram and nortriptyline in treating major depressive disorder. J Psychiatr Res 2012;46:13331338.