Response to Tricyclic Antidepressants: Independent of Gender?

The database incorporated 32 studies submitted between 1979 and 1991 to the Medicines Evaluation Board as part of a registration dossier for different drugs under development. The Medicines Evaluation Board is the regulatory authority in the Netherlands that is responsible for granting marketing authorizations to medical compounds.

The studies consisted of three treatment arms: an experimental compound, an active control (tricyclic antidepressant), and placebo. The data for the placebo and the tricyclic antidepressant arms was made available to the Medicines Evaluation Board by the pharmaceutical companies.

All patients had received a DSM-III-R diagnosis of major depression. The Hamilton Depression Rating Scale was used as the outcome measure for efficacy.

Statistical analyses were based on the intent-to-treat population. This population included all patients who were randomly assigned to receive double-blind treatment, had taken at least one dose of study medication, and had at least one efficacy evaluation. The last observation was carried forward for patients who had dropped out before the planned end of the study.

Several outcome variables were examined: change in Hamilton depression scale scores from baseline to end of study, response rate (response defined as a 50% reduction from baseline in Hamilton depression scale score), remission rate (remission defined as a Hamilton depression scale score ≤7 [6]), and patient dropout rate. Gender differences in response to treatment on each of these outcomes were examined in each study by testing the interaction term between gender and treatment in a regression model (linear regression in the case of the Hamilton depression scale change score and logistic regression for response, remission, and dropout rates).

In addition, efficacy of tricyclic antidepressant compared with placebo in terms of Hamilton depression scale change scores was estimated separately among men and among women older and younger than 50 in order to evaluate the effect of hormonal status on response. This was done over all studies by using a random effect model meta-analysis.

Of the 32 submitted studies, two had no information about gender of responders and hence had to be excluded from further analysis. The remaining 30 studies included a total of 3,886 patients, of whom 2,331 (60%) were women, and 1,555 (40%) were men. The majority of women (1,809 [78%]) were younger than 50 years. Imipramine was the active comparator in 27 (90%) of the 30 studies; amitriptyline was the active comparator in the remaining three (10%).

Two studies (out of the 30 that were examined) showed a significant interaction (at the 0.05 level) between gender and treatment when change in Hamilton depression scale score was the dependent variable. With response as the dependent variable, two different studies showed a significant interaction; with remission as the defining variable, two yet different studies showed a significant interaction. With dropouts as the dependent variable, none of the studies showed a significant interaction. In some of the studies with significant interaction, women showed a more favorable response, whereas in others men responded better.

The random effect estimates (from the meta-analyses) of the difference in Hamilton depression scale score improvement between tricyclic antidepressants and placebo were essentially similar for men (–2.38, 95% confidence interval [CI]=–3.21 to –1.55), women older than 50 (–3.04, 95% CI=–5.08 to –1.01), and women younger than 50 (–2.44, 95% CI=–3.30 to –1.58). The heterogeneity test came out as nonsignificant for men (χ²=31.43, df=29, p<0.35) and for women younger than 50 (χ²=34.03, df=28, p=0.20). For women older than 50, heterogeneity was significant (χ²=39.27, df=23, p<0.02), but this is probably due to the small sample sizes in this stratum (more than half of the studies included less than 15 women older than 50 years).

The results presented in this study show no gender differences in response to tricyclic antidepressants in this large database consisting of 3,886 patients from 30 randomized clinical trials. Only few significant differences were found, and these are most likely the result of chance, given the number of comparisons that were performed. The fact that, depending on the measure of efficacy (i.e., change in Hamilton depression scale scores or rates of remission, response, or dropout), different sets of two studies showed significant treatment-by-gender interactions supports the contention that these interactions are chance results. With a type I error level of 5%, having two studies with significant results is only slightly higher than what could be expected by chance in a series of 30 studies. Furthermore, examination of the significant studies could not reveal any special characteristic that could distinguish these studies from the others. Finally, efficacy estimates based on a meta-analysis of these studies indicate that efficacy estimates were essentially similar among men and women older and younger than age 50.

These results corroborate those of Quitkin et al. (4), this time in a larger and more representative data set of depressed patients and using a variety of possible definitions of treatment efficacy, thereby providing more certainty regarding the generalizability of these results. The results suggest that previously found gender differences in response to tricyclics could have appeared by chance the same way as some of the studies presented in this study resulted in significant gender differences for some of the possible treatment outcome measures.

A limitation of this study is that only tricyclic antidepressants were examined. The issue of gender differences in response to SSRIs, which are much more widely used, remains unaddressed. It should also be noted that although no differential response of pre- and postmenopausal women was found, the power to detect a difference in this dataset was relatively small because of the few women older than 50. Other factors that may differentially influence efficacy were not examined in this paper. These include age, concomitant medication, and smoking status.

Received April 12, 2003; revision received July 3, 2003; accepted July 5, 2003. From the Medicines Evaluation Board of the Netherlands, Den Haag; and the Psychiatric Department of the Academic Medical Center, Amsterdam. Address reprint requests to Dr. Wohlfarth, Medicines Evaluation Board of the Netherlands, Kalvermarkt 53, P.O. Box 16229, 2500 BE Den Haag, the Netherlands; [email protected] (e-mail). The authors thank the following pharmaceutical companies for providing the data used in this study: Bristol-Meyers Squibb Company; Eli Lilly and Company; GlaxoSmithKline; H. Lundbeck A/S; Organon Inc.; Pfizer Inc.; Solvay Pharmaceuticals, Inc.; and Wyeth BV. The authors also thank Nefarma for making this study possible.