Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Antidepressant Drug Adherence


One of those battles that seems to rage in the background is the question of whether SSRIs produce better patient compliance than other antidepressants. A major problem is how to measure compliance, and one measure which has been used is the number of people who withdraw from clinical trials. A new Cochrane review does more than most reviews (of which a number are to be found in the literature) to help us think about this.

It is important, because this question is at the heart of which treatment should come first in the treatment of depression. If efficacy and harm are balanced, cost rules. If efficacy and harm are not balanced, then the question of cost-effectiveness emerges, and cost-effectiveness in healthcare systems is not an exact science.


The review sought randomised trials comparing SSRIs with other antidepressants in patients suffering from depression diagnosed by any criteria. The outcomes extracted were the number of people withdrawing during the study in total, and those withdrawing due to inefficacy and those withdrawing due to adverse events. Several electronic databases were searched, including a specialist depression register, with hand searching of specialist journals plus attempts to obtain unpublished material. There was no language restriction. Studies were grouped by older tricyclics (amitriptyline, imipramine), newer tricyclics (clomipramine, desipramine, for example) and heterocyclics (maprotiline, mianserin, for example).


There were 136 trials that met the inclusion criteria. Their duration was four weeks to one year, but most were longer than six weeks, and over half were conducted in an outpatient setting. Most of the studies enrolled patients suffering from major depression, using standard diagnostic criteria.

There were fewer total withdrawals with SSRIs. On average total withdrawals were 27% with SSRIs and 30% for comparator drugs. More SSRI withdrawals due to lack of efficacy were counterbalanced by fewer withdrawals due to adverse events.

There was considerable variability in adverse event withdrawal rates in individual studies, best exemplified by the large data sets for amitriptyline (Figure 1) and imipramine (Figure 2). Group sizes in these comparisons varied between 10 and 380 patients. Of the 122 separate groups, 80 had 50 or fewer patients, 37 between 51 and 100, 12 between 101 and 200 and three over 200 patients.

Figure 1: Adverse event withdrawals in trials comparing SSRI with amitriptyline

Figure 2: Adverse event withdrawals in trials comparing SSRI with imipramine

Adverse event withdrawals for particular drug comparisons with SSRIs are shown in Table 1 and Figure 3 for those where there were at least three trials. SSRIs had significantly fewer adverse event withdrawals than amitriptyline, imipramine and clomipramine, with numbers needed to treat to prevent one adverse event withdrawal of about 15 to 20. Other comparisons either had no significant difference or there were fewer than 200 patients in the comparator group. Figure 3 shows the 95% confidence interval for the absolute adverse event withdrawal rate. Where the numbers of patients was small, the confidence interval was wide.

Figure 3: Absolute adverse event withdrawal rates (95% CI). Total in parenthesis

Table 1: Adverse event withdrawals in comparisons of particular antidepressants and SSRIs

    Antidepressant SSRI    
Comparison SSRI and - Number of trials Number/ total Percent (95%CI) Number/ total Percent (95%CI) Relative benefit (95%CI) NNT (95% CI)
Amitriptyline 32 231/1563 15 (13-17) 184/1670 11(10-13) 1.4 (1.1 to 1.8) 27 (16 to 69)
Imipramine 29 301/1508 20 (18-22) 233/1844 13 (11-14) 1.8 (1.4 to 2.4) 14 (10 to 22)
Clomipramine 11 117/1105 11 (9-12) 74/1021 7 (6-9) 1.6 (1.2 to 2.1) 23 (14 to 55)
Maprotiline 7 40/451 9 (6-12) 27/452 6 (4-8) 1.5 (0.9 to 2.3) not calculated
Dothiepin 5 11/171 6 (3-10) 29/173 17 (11-22) 0.4 (0.2 to 0.8) -10 (-6 to -28)
Mianserin 4 15/113 13 (7-20) 20/110 18 (11-25) 0.7 (0.4 to 1.3) not calculated
Doxepin 3 29/197 15 (10-20) 38/200 19 (14-24) 0.8 (0.5 to 1.2) not calculated
Desipramine 3 10/70 15 (6-22) 2/71 3 (0-7) 3.7 (1.1 to 12) 9 (5 to 58)


Firstly, this was a fine and thorough review, short and readable, and with a useful discussion about the likely importance of dose. SSRIs were grouped together, so no comments can be made about withdrawals for particular SSRIs.

Secondly, it shows the importance of size in determining small differences between groups. With the information we have, we can be sure that adverse event withdrawal rates are lower with SSRIs than amitriptyline and imipramine. For others there is doubt. For clomipramine, for instance, there are significantly more adverse event withdrawals than for SSRIs (Table 1), but the percentage of withdrawals is no different than the overall rate for SSRIs (Figure 3).

Why the uncertainty? This comes from three main areas - dose, collecting adverse event data, and size.

  • Higher doses of medicines are likely to elicit greater adverse event rates. The results that we are presented with make no allowance for dose, so this could be a contributory factor.
  • Collecting adverse event information using diaries elicits higher rates of adverse events than spontaneous reporting [2]. Different collection methods may also influence withdrawal rates, and definitions of withdrawals may not always be the same.
  • Size is crucial, because smaller amounts of information are more susceptible to the random play of chance. Figure 4 shows the SSRI adverse event withdrawal rates in 100 trials according to the size of the SSRI group. Below 200 patients the rates are highly variable. For many comparisons (Table 1) we have fewer than 200 patients, reflecting the variable SSRI withdrawal rates and affecting the chances of showing any true difference, let alone the size of that difference.

Figure 4: SSRI adverse event withdrawal rate and size of trial. Vertical line is overall mean

So our conclusion is that while there are lower adverse event withdrawal rates with SSRIs, extrapolating from those to real-life situations or whole healthcare systems will not be easy.


  1. C Barbui et al. Selective serotonin reuptake inhibitors versus tricyclic and heterocyclic antidepressants: comparison of drug adherence. In: The Cochrane Library, Issue 4, 2000. Oxford: Update Software.
  2. JE Edwards et al. Reporting of adverse effects in clinical trials should be improved. Lessons from acute postoperative pain. Journal of Pain and Symptom Management 1999; 18:427-37.
previous or next story in this issue