Antidepressant discontinuation trial misleading as it likely mis-interprets withdrawal effects as relapse
Dear Editor
Lewis and colleagues randomly assigned patients with multiple depressive episodes, stable on antidepressants, to either continue antidepressants or to stop them over 4-8 weeks.[1] The authors concluded that continuing antidepressants reduces the chance of relapse even for those patients feeling well enough to stop as 39% of patients in the maintenance arm met criteria for relapse, compared to 56% in the discontinuation arm.
This conclusion is not warranted because the authors neglected to account for the possibility of antidepressant withdrawal effects being mis-classified as relapse, a fundamental problem in discontinuation trials.[2,3] Stopping antidepressants commonly causes withdrawal symptoms, with about half of patients experiencing them.[4] This is reflected in the current study where the average number of withdrawal symptoms was more than double in the discontinued group at 12 weeks (3.1 vs 1.3 out of 15 total symptoms).[1] Even this may have been an underestimate as the modified DESS scale used did not measure severity. ‘Dizziness’, for example, may mean different things to people in the maintenance group compared to the discontinuation group (such withdrawal symptoms can be so severe they have led to people being investigated for stroke).[5]
Although antidepressants were, by design, discontinued more slowly than in previous trials, we now know that 8 weeks is still a relatively short taper in patients who had been on the drugs for more than 2 years (half the dose for one month, then halve the dose every second day for one month, before stopping). Although this approach was consistent with recommendations at the time the trial was set up, since then guidance from the Royal College of Psychiatrists[6] and draft guidance from NICE[7] recommends stopping antidepressants over “months or more.”[7] Also the final doses before stopping (one quarter the highest dose), although seemingly small, have large effects on target receptors and therefore cause large shifts in effect when reduced to zero.[8] Hence withdrawal symptoms are still likely to occur.
Antidepressant withdrawal symptoms overlap with most domains of the depression scale (rCIS-R) used to detect relapse,[9] and are therefore likely to artificially inflate relapse rates in the discontinuation group.[2] The withdrawal symptoms measured by the authors include: ‘severe nervousness or anxiety’, ‘confusion or trouble concentrating’, ‘agitation’, ‘brain fog, forgetfulness or problems with memory’, ‘trouble sleeping, insomnia’, ‘fatigue, tiredness’, ‘sudden panic or anxiety.’[9] These symptoms would also register on the scales used to detect anxiety (GAD-7), depression (PHQ-9) and relapse (rCIS-R), which include the same domains.
There are several lines of evidence that withdrawal symptoms from antidepressants were indeed mis-classified as relapse in this study. There was high correlation between mean differences at the different time points on the withdrawal scale (DESS) and mean differences on the depression scale (PHQ-9) (R2=0.97) and anxiety scale (GAD-7) (R2=0.89). The differences between the average of the groups for the DESS at weeks 12, 26, 39 and 52 were 2.2, 0.7, 0.9 and 0.1, respectively; for the GAD-7 they were 2.2, 0.8, 0.6, 0.3 and for the PHQ-9 they were 2.2, 0.8, 0.6, 0.3. Together, with the overlap of withdrawal symptoms with measures of mood and relapse, this suggests that the withdrawal symptoms may account for the increase in symptom scores and relapse rate. The reverse direction of causation would be unlikely, since withdrawal symptoms include physical symptoms that are not intrinsically related to depression, including dizziness, electric shocks and headache. Whilst it is possible that relapse and withdrawal effects could co-incide, Occam’s razor would suggest one condition causes several symptoms rather than requiring several conditions.
Withdrawal confounding of relapse is also consistent with the finding that most relapses occurred when withdrawal effects are at their peak, within 6-12 weeks of when the drugs were stopped (at week 8). Indeed, 90% of the total difference in relapse rates between the two arms of the trial (at 52 weeks) was present at 12 weeks after the drugs were stopped (although this accounts for only 27% of the total follow-up time (12/(52-8)*100%). Furthermore, patients stopping fluoxetine, associated with fewer withdrawal effects than other antidepressants, because of its long elimination half-life, relapsed about 25% less than people stopping citalopram and sertraline again suggesting withdrawal effects.
Moreover anxiety, and depression scores were the same for both groups at the end of the study (week 52). Although 44% of the discontinued group returned to their medication by this point, even with twice as many people on antidepressants in the maintenance group (89% were still on medication) there was no difference in symptom scores. This suggests that discontinuation of antidepressants did not worsen mood after the period in which withdrawal symptoms had settled (there were only small differences in DESS scores by 52 weeks).
Lastly, 71% of patients in the discontinuation group correctly guessed their allocation to placebo, possibly because of withdrawal symptoms, and the expectation that they would get worse might have contributed to their worsening.[10]
As there was no effort made to manage the potential confounding of relapse by withdrawal the current study suffers the same flaws as previous discontinuation studies and cannot provide evidence of the benefits of long-term treatment, only the difficulties of stopping it. The authors could resolve some of these concerns by analysing the correlation of withdrawal symptoms with mood scores and relapse amongst individual patients to verify if withdrawal symptoms might account for relapse. They could also re-analyse their data by excluding patients who experienced significant withdrawal symptoms (e.g. modified DESS ≥ 2) from qualifying for a diagnosis of relapse. This would provide a more robust measure of relapse, reducing the potential for the misclassification of relapse as withdrawal. They could also test whether unblinding was associated with relapse.
Uncritical interpretation of this study may lead to the erroneous conclusion that antidepressants should be continued to prevent relapse, when in reality all they may be doing is preventing withdrawal symptoms. The more accurate conclusion would be that such symptoms are temporary withdrawal symptoms that can be minimised by stopping the drug more gradually, as recognised by the authors in media interviews, although not in the published paper.
References
1 Lewis G, Marston L, Duffy L, et al. Maintenance or Discontinuation of Antidepressants in Primary Care. N Engl J Med 2021;385:1257–67. doi:10.1056/NEJMoa2106356
2 Hengartner MP, Plöderl M. Prophylactic effects or withdrawal reactions? An analysis of time-to-event data from antidepressant relapse prevention trials submitted to the FDA. Ther Adv Psychopharmacol 2021;11:20451253211032052. doi:10.1177/20451253211032051
3 Hengartner MP. How effective are antidepressants for depression over the long term? A critical review of relapse prevention trials and the issue of withdrawal confounding. Ther Adv Psychopharmacol 2020;10:2045125320921694. doi:10.1177/2045125320921694
4 Davies J, Read J. A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based? Addict Behav 2019;97:111–21. doi:10.1016/j.addbeh.2018.08.027
5 Haddad P, Devarajan S, Dursun S. Antidepressant discontinuation (withdrawal) symptoms presenting as ‘stroke’. J Psychopharmacol 2001;15:139–41. doi:10.1177/026988110101500210
6 Burn W, Horowitz M, Roycroft G, et al. Stopping antidepressants. Stopping Antidepressants. 2020;:1–9.https://www.rcpsych.ac.uk/mental-health/treatments-and-wellbeing/stoppin...
7 NICE. Safe prescribing and withdrawal management of prescribed drugs associated with dependence and withdrawal. 2019.https://www.nice.org.uk/guidance/indevelopment/gid-ng10141 (accessed 6 Nov 2019).
8 Horowitz MA, Taylor D. Tapering of SSRI treatment to mitigate withdrawal symptoms. The Lancet Psychiatry 2019;6:538–46. doi:10.1016/S2215-0366(19)30032-X
9 Rosenbaum JF, Fava M, Hoog SL, et al. Selective serotonin reuptake inhibitor discontinuation syndrome: A randomized clinical trial. Biol Psychiatry 1998;44:77–87. doi:10.1016/S0006-3223(98)00126-7
10 Moncrieff J. Are antidepressants overrated? A review of methodological problems in antidepressant trials. J Nerv Ment Dis 2001;189:288–95.
Competing interests:
MH and BP have no conflicts of interest. JM reports grants from the National Institute of Health Research, that she is co-chairperson of the Critical Psychiatry Network (an informal group of psychiatrists) and a board member of the unfunded organisation, the Council for Evidence-based Psychiatry. Both are unpaid positions.
27 October 2021
Mark A. Horowitz
Clinical Research Fellow in Psychiatry
Professor Joanna Moncrieff, Dr Beth Parkin
UCL
Maple House, Department of Psychiatry, 149 Tottenham Court Rd, UCL
Rapid Response:
Antidepressant discontinuation trial misleading as it likely mis-interprets withdrawal effects as relapse
Dear Editor
Lewis and colleagues randomly assigned patients with multiple depressive episodes, stable on antidepressants, to either continue antidepressants or to stop them over 4-8 weeks.[1] The authors concluded that continuing antidepressants reduces the chance of relapse even for those patients feeling well enough to stop as 39% of patients in the maintenance arm met criteria for relapse, compared to 56% in the discontinuation arm.
This conclusion is not warranted because the authors neglected to account for the possibility of antidepressant withdrawal effects being mis-classified as relapse, a fundamental problem in discontinuation trials.[2,3] Stopping antidepressants commonly causes withdrawal symptoms, with about half of patients experiencing them.[4] This is reflected in the current study where the average number of withdrawal symptoms was more than double in the discontinued group at 12 weeks (3.1 vs 1.3 out of 15 total symptoms).[1] Even this may have been an underestimate as the modified DESS scale used did not measure severity. ‘Dizziness’, for example, may mean different things to people in the maintenance group compared to the discontinuation group (such withdrawal symptoms can be so severe they have led to people being investigated for stroke).[5]
Although antidepressants were, by design, discontinued more slowly than in previous trials, we now know that 8 weeks is still a relatively short taper in patients who had been on the drugs for more than 2 years (half the dose for one month, then halve the dose every second day for one month, before stopping). Although this approach was consistent with recommendations at the time the trial was set up, since then guidance from the Royal College of Psychiatrists[6] and draft guidance from NICE[7] recommends stopping antidepressants over “months or more.”[7] Also the final doses before stopping (one quarter the highest dose), although seemingly small, have large effects on target receptors and therefore cause large shifts in effect when reduced to zero.[8] Hence withdrawal symptoms are still likely to occur.
Antidepressant withdrawal symptoms overlap with most domains of the depression scale (rCIS-R) used to detect relapse,[9] and are therefore likely to artificially inflate relapse rates in the discontinuation group.[2] The withdrawal symptoms measured by the authors include: ‘severe nervousness or anxiety’, ‘confusion or trouble concentrating’, ‘agitation’, ‘brain fog, forgetfulness or problems with memory’, ‘trouble sleeping, insomnia’, ‘fatigue, tiredness’, ‘sudden panic or anxiety.’[9] These symptoms would also register on the scales used to detect anxiety (GAD-7), depression (PHQ-9) and relapse (rCIS-R), which include the same domains.
There are several lines of evidence that withdrawal symptoms from antidepressants were indeed mis-classified as relapse in this study. There was high correlation between mean differences at the different time points on the withdrawal scale (DESS) and mean differences on the depression scale (PHQ-9) (R2=0.97) and anxiety scale (GAD-7) (R2=0.89). The differences between the average of the groups for the DESS at weeks 12, 26, 39 and 52 were 2.2, 0.7, 0.9 and 0.1, respectively; for the GAD-7 they were 2.2, 0.8, 0.6, 0.3 and for the PHQ-9 they were 2.2, 0.8, 0.6, 0.3. Together, with the overlap of withdrawal symptoms with measures of mood and relapse, this suggests that the withdrawal symptoms may account for the increase in symptom scores and relapse rate. The reverse direction of causation would be unlikely, since withdrawal symptoms include physical symptoms that are not intrinsically related to depression, including dizziness, electric shocks and headache. Whilst it is possible that relapse and withdrawal effects could co-incide, Occam’s razor would suggest one condition causes several symptoms rather than requiring several conditions.
Withdrawal confounding of relapse is also consistent with the finding that most relapses occurred when withdrawal effects are at their peak, within 6-12 weeks of when the drugs were stopped (at week 8). Indeed, 90% of the total difference in relapse rates between the two arms of the trial (at 52 weeks) was present at 12 weeks after the drugs were stopped (although this accounts for only 27% of the total follow-up time (12/(52-8)*100%). Furthermore, patients stopping fluoxetine, associated with fewer withdrawal effects than other antidepressants, because of its long elimination half-life, relapsed about 25% less than people stopping citalopram and sertraline again suggesting withdrawal effects.
Moreover anxiety, and depression scores were the same for both groups at the end of the study (week 52). Although 44% of the discontinued group returned to their medication by this point, even with twice as many people on antidepressants in the maintenance group (89% were still on medication) there was no difference in symptom scores. This suggests that discontinuation of antidepressants did not worsen mood after the period in which withdrawal symptoms had settled (there were only small differences in DESS scores by 52 weeks).
Lastly, 71% of patients in the discontinuation group correctly guessed their allocation to placebo, possibly because of withdrawal symptoms, and the expectation that they would get worse might have contributed to their worsening.[10]
As there was no effort made to manage the potential confounding of relapse by withdrawal the current study suffers the same flaws as previous discontinuation studies and cannot provide evidence of the benefits of long-term treatment, only the difficulties of stopping it. The authors could resolve some of these concerns by analysing the correlation of withdrawal symptoms with mood scores and relapse amongst individual patients to verify if withdrawal symptoms might account for relapse. They could also re-analyse their data by excluding patients who experienced significant withdrawal symptoms (e.g. modified DESS ≥ 2) from qualifying for a diagnosis of relapse. This would provide a more robust measure of relapse, reducing the potential for the misclassification of relapse as withdrawal. They could also test whether unblinding was associated with relapse.
Uncritical interpretation of this study may lead to the erroneous conclusion that antidepressants should be continued to prevent relapse, when in reality all they may be doing is preventing withdrawal symptoms. The more accurate conclusion would be that such symptoms are temporary withdrawal symptoms that can be minimised by stopping the drug more gradually, as recognised by the authors in media interviews, although not in the published paper.
References
1 Lewis G, Marston L, Duffy L, et al. Maintenance or Discontinuation of Antidepressants in Primary Care. N Engl J Med 2021;385:1257–67. doi:10.1056/NEJMoa2106356
2 Hengartner MP, Plöderl M. Prophylactic effects or withdrawal reactions? An analysis of time-to-event data from antidepressant relapse prevention trials submitted to the FDA. Ther Adv Psychopharmacol 2021;11:20451253211032052. doi:10.1177/20451253211032051
3 Hengartner MP. How effective are antidepressants for depression over the long term? A critical review of relapse prevention trials and the issue of withdrawal confounding. Ther Adv Psychopharmacol 2020;10:2045125320921694. doi:10.1177/2045125320921694
4 Davies J, Read J. A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based? Addict Behav 2019;97:111–21. doi:10.1016/j.addbeh.2018.08.027
5 Haddad P, Devarajan S, Dursun S. Antidepressant discontinuation (withdrawal) symptoms presenting as ‘stroke’. J Psychopharmacol 2001;15:139–41. doi:10.1177/026988110101500210
6 Burn W, Horowitz M, Roycroft G, et al. Stopping antidepressants. Stopping Antidepressants. 2020;:1–9.https://www.rcpsych.ac.uk/mental-health/treatments-and-wellbeing/stoppin...
7 NICE. Safe prescribing and withdrawal management of prescribed drugs associated with dependence and withdrawal. 2019.https://www.nice.org.uk/guidance/indevelopment/gid-ng10141 (accessed 6 Nov 2019).
8 Horowitz MA, Taylor D. Tapering of SSRI treatment to mitigate withdrawal symptoms. The Lancet Psychiatry 2019;6:538–46. doi:10.1016/S2215-0366(19)30032-X
9 Rosenbaum JF, Fava M, Hoog SL, et al. Selective serotonin reuptake inhibitor discontinuation syndrome: A randomized clinical trial. Biol Psychiatry 1998;44:77–87. doi:10.1016/S0006-3223(98)00126-7
10 Moncrieff J. Are antidepressants overrated? A review of methodological problems in antidepressant trials. J Nerv Ment Dis 2001;189:288–95.
Competing interests: MH and BP have no conflicts of interest. JM reports grants from the National Institute of Health Research, that she is co-chairperson of the Critical Psychiatry Network (an informal group of psychiatrists) and a board member of the unfunded organisation, the Council for Evidence-based Psychiatry. Both are unpaid positions.