We appreciate the interest in our article by Jauhar and colleagues and welcome the opportunity to address a number of misleading points in their letter.
No argument they raise undermines our central points: antidepressants do not demonstrate a clinically important difference from placebo according to any suggested threshold, including that proposed by NICE, empirical correlation with clinician evaluation or other means. There is no conclusive evidence of an above-average response sub-group so far demonstrated. The evidence for relapse prevention properties is highly problematic because of confounding by withdrawal effects. Common adverse effects are well-established, a...
We appreciate the interest in our article by Jauhar and colleagues and welcome the opportunity to address a number of misleading points in their letter.
No argument they raise undermines our central points: antidepressants do not demonstrate a clinically important difference from placebo according to any suggested threshold, including that proposed by NICE, empirical correlation with clinician evaluation or other means. There is no conclusive evidence of an above-average response sub-group so far demonstrated. The evidence for relapse prevention properties is highly problematic because of confounding by withdrawal effects. Common adverse effects are well-established, and there are some worrying signals about rare but serious adverse effects, on which we need more data. Quibbling over whether antidepressants cause withdrawal in 30%, 40% or 50% of people misses the more important public health point that any of these proportions means that millions of people are at risk of either suffering in coming off their drugs or being forced to stay on them because of the aversive effects of coming off them. While more studies of this neglected area are always welcome to help define the problem more precisely, authorities have already sensibly acknowledged that the problem is vast enough to require decisive action.[1–3]
Efficacy
The principal argument put forward by Jauhar and colleagues is that we were wrong to suggest that the difference between antidepressants and placebo is clinically insignificant. This ignores the key point that the modest difference between antidepressants and placebo is likely inflated by significant biases including lack of blinding, the use of the placebo run-in design and short trial duration.[4] There is general agreement amongst meta-analyses that the mean difference between placebo and antidepressants is 2 points on the 52-point HDRS scale.[4,5] Any observer would likely judge this a small difference, especially in light of the fact that placebo recipients routinely improve by 7-13 points on the HDRS scale in meta-analyses.[6,7] Indeed, Professor Cipriani, one of the co-authors, has said: “No research evidence or consensus is available about what constitutes a clinically meaningful difference in Hamilton (HDRS) scores, but...a difference of less than 3 points seems unlikely to be regarded as such.”[8] NICE used a cut-off of 0.5 SMD (equivalent to 3.3 HDRS points)4 in its latest draft guidance on depression. [9]
Jauhar and colleagues claim that an analysis [10] of more than 7000 patients from 43 drug trials which aimed to correlate change on the HDRS scale into clinical global impression of change (rating of patients by clinicians from ‘very much worse’ to ‘very much improved’) which is ‘intuitively understood by clinicians’ is not informative both because the study examined people taking mirtazapine, which has differences in receptor targets to SSRIs and SNRIs, and because within-person changes cannot be extrapolated to between-group differences, citing a letter on urinary incontinence. While we accept the possibility that medications with different receptor targets may produce different effects (e.g. being more or less sedating) we do not accept the argument that this study has no relevance to interpreting other drugs because a measurement instrument like a ruler (or the HDRS) is useful whether it is used to measure the length of a single plank or the difference in length between two planks, and can be used whether the plank is made of wood or metal. Having different receptor targets to other drugs has limited relevance and the issue of whether it looks at within-group or between-group differences has no relevance: the point is to correlate change in one metric that has no a priori clinical significance (the 52-point HDRS scale) with one that is understandable to clinicians (degree of clinical improvement). Furthermore, it is not true to say that mirtazapine exerts ‘large effects on HDRS’ compared to other medications: it shows broadly similar effects to other modern antidepressants (2-3 points on the HDRS, according to Appendix in Cipriani et al. (2018).[11]
Indeed, contrary to their assertion that ‘experts’ suggest that a minimally important difference (MID) should only apply to within-groups change and not between-groups change, expert consensus is that MIDs can - and should - inform evaluation of treatment effects (which includes placebo-drug differences),[12–15] and these differences are routinely used for power calculations of studies.[16] As detailed by Revicki et al [17]:
“Although demonstrating responsiveness is a key component to establishing an instrument’s construct validity, it is also important to determine the MID to assist in interpreting statistically significant PRO (patient-reported outcome) results in clinical trials. In addition, the MID for a PRO instrument that is specified as a primary or important secondary endpoint is clearly useful for calculating statistical power and for determining sample sizes for clinical trials.”
On the topic of a suggested alternative, Jauhar and colleagues are notably silent. We therefore think the finding that a change in over 7 points on the HDRS is required for a clinician to measure even a difference of ‘minimally improved’ is a useful guide, also considered useful in a recent paper in World Psychiatry.[18] Other authors, using anchor-based and distribution-based approaches have estimated the MID for the HDRS to be 3 to 8 points, which also included trials with SSRI and SNRI drugs.[19]
The authors then suggest that examining ‘key mood items’ is a more ‘nuanced’ approach which has shown ‘clear benefits’.[20] However, there are several problems with focusing on one item out of the 17 HDRS items (‘depressed mood’) as the paper cited does - firstly, this single item lacks content validity as it fails to capture the many symptoms of depression, as defined, for example, by DSM and ICD, including effects on interest/pleasure, sleep, appetite, suicidality, energy levels, etc.[21,22] Secondly, the item, ‘depressed mood’, reproduced below, itself lacks face validity:
1 |__| These feeling states indicated only on questioning.
2 |__| These feeling states spontaneously reported verbally.
3 |__| Communicates feeling states non-verbally, i.e. through facial expression, posture, voice and tendency to weep.
4 |__| Patient reports virtually only these feeling states in his/her spontaneous verbal and non-verbal communication.
In the analysis performed by Hieronymus and colleagues, the mean of this score at baseline and endpoint were between 1 and 2.[20] It is not self-evident that having to be asked if one feels sad (score = 1) indicates that someone is less depressed than if they report this spontaneously (score = 2). Thirdly, studies looking at this single item are likely subject to reporting bias as the study cohorts consisted entirely of industry-conducted trials that made individual patient data available to them.[20] In any case, the effect size for the depressed mood is still small (SMD=0.40) and below the NICE criterion (0.5 SMD) for a clinically relevant difference.[9] Furthermore, all the same methodological biases that apply to studies employing the full HDRS, also apply to the single item of ‘depressed mood’, likely meaning the true effect is even smaller.[4]
Elsewhere the authors have claimed larger effects for a sub-set of the HDRS-17, the HDRS-6 (Bech-subscale), but in a recent analysis antidepressants only showed an effect size of 0.32 in severe depression (as compared to 0.39 in non-severe depression),[7] hardly different from the finding on the full HDRS-17 scale (0.30), and well below all estimates of MID.19 These subsets of symptoms suffer from the same lack of content validity as a single mood item.[21,22]
Jauhar and colleagues then argue that even if the average effect of antidepressants is small there may be a sub-group with a much larger response. However, analyses that have examined heterogeneity in antidepressant response have so far failed to find evidence for it.[23–26] Unfortunately, Jauhar and colleagues cite a study to support their notion of a strongly responding sub-group [27] that was retracted [28] because the authors used an incorrect analysis. When the proper analysis was conducted the heterogeneity in the antidepressant arms of the trials was not found to be greater than that in the placebo arms of the trials,[25] consistent with all other analyses. Although of course there is variability in the response to both placebo and antidepressant, as the authors state, the cited studies find that the variation is the same for placebo and antidepressant. The simplest explanation for the equal variances is that there are constant treatment effects and there is no treatment heterogeneity.[29]
A final argument made by the authors is that the small size of effect for antidepressants is similar to other medical treatments, citing an analysis by Leucht. [30] This is misleading for two reasons – one, because outcomes for other treatments were mostly based on hard clinical outcomes such as survival or cardiovascular events. If antidepressants were held to the similarly hard clinical outcomes (such as suicide attempts or hospitalisations) rather than subjective symptom rating scales, there would be no effect for antidepressants. Secondly, the duration of treatment in many of the studies in non-psychiatric drugs are months or years (e.g. 5.8 years for aspirin in stroke and heart disease, 10.7 years for metformin in diabetes, 15 years for chemotherapy for breast cancer), highly relevant to the long-term use of these medications, whereas for antidepressants the duration of treatment is generally 6 to 12 weeks.[11] This duration of treatment has limited relevance to the treatment of depression in practice which is often for months, years or even decades.
Lastly, it is also worth considering just how large expectation effects are, which may overshadow any chemical effect of the drug, and account for much or all of the placebo-drug differences in trials. For example, in a negative study of hypericum, sertraline and placebo, people who guessed they were taking one of the active drugs improved by 4.6 to 5.7 points more than those who guessed they were taking placebo, regardless of actual allocation.[31] People allocated to placebo who guessed they were taking sertraline improved by 5.4 points more than people allocated to sertraline who guessed they were taking the placebo.31 This is consistent with another study in which patients were all given escitalopram, but half told what it was (overt) and the other half deceived into thinking it was an active placebo (covert), in which the overt group showed a reduction in MADRS of 9.66 points, 5 points greater than the 4.64 point reduction in the covert group.[32]
In relation to the STAR*D study, although Jauhar and colleagues put forward the unusual position that the original study authors should have the exclusive right to perform the definitive analysis of data, this is not the guiding principle of open science. For example, a re-analysis of paroxetine trials in adolescents found that the data was negative for all eight pre-specified primary and secondary outcomes, although the study was reported as positive by the original authors.[33] The re-analysis of the STAR*D data is important because the original authors decided to switch out the intended primary outcome (HDRS) for the Quick Inventory of Depressive Symptomatology (QIDS) and elected not to use intent-to-treat analysis to derive an exaggerated figure for remission.[34] Pigott and colleagues employed intent-to-treat analysis and found that 108 patients out of 4041 had remitted and not dropped out at 12 months after four cycles of antidepressant treatment according to gold standard treatment.34 Methods used to inflate remission rates in the original publication in a wide variety of ways has been previously discussed,[35] and the overall effect on the HDRS was only 6.55 points,[36] less than the effect of placebo in most studies.[7]
Adverse effects
With regards to adverse effects, one limitation of the field is the lack of long-term studies addressing this issue. As recently pointed out, studies are designed to identify positive effects by their manufacturers, often employing multiple symptom measures to do so.[37] These measures are themselves composites of several symptoms, each interrogated for severity (e.g. the HDRS).[37] In contrast, adverse effects are considered in isolation, with no consideration of their severity, such that their overall impact is impossible to judge, and most studies restrict the adverse effects that are queried. For example, manufacturer studies do not often ask patients about potentially important symptoms such as ‘fogginess’, or ‘emotional numbing’ so a survey with a self-selected group is the only evidence we have of these potentially important adverse effects.[38]
The naturalistic study we cited, used by Jauhar to suggest no relationship to time period, was not set up well to analyse such a relationship as the study authors themselves say: “associations between SEs and…duration may have been weakened by routine clinical measures like lowering of dosages, switching and discontinuation of the antidepressant” and point to the issue of tolerance to adverse effects if patients had used similar drugs before the start of the study.[39] Nevertheless, this study demonstrates that antidepressants cause multiple common adverse effects in the short term as well as the long term.
The authors point to an umbrella review of observational studies [40] but this study suffers from all the same limitation as the observational studies Jauhar and colleagues criticised, and was restricted to a few serious but rare adverse events. These are notoriously difficult to establish reliably, and it is appropriate to have lower standards of evidence for serious harms. Nevertheless, there was “convincing” or “highly suggestive evidence” (of moderate to high quality according to A Measurement Tool to Asses Systematic Reviews-2 (AMSTAR-2) for maternal complications, foetal malformations, osteoporotic fractures, upper GI bleeding and suicidal behaviour among young people among other effects which remained statistically significant in the sensitivity analysis of studies that adjusted for multiple confounders and confounding by indication.[40]
Convincing evidence on common adverse effects comes from a meta-analysis of randomised placebo-controlled antidepressant trials,[5] which therefore does not suffer from confounding by indication, but which may still under-report adverse effects.[41] This meta-analysis found 28 adverse effects to be significantly more common in antidepressants than placebo: including abnormal ejaculation (5.4-fold), anorexia (2.8 fold), nausea (2.5 fold), somnolence (2.3 fold), sweating (2.2 fold), asthenia (1.7-fold), insomnia (1.5 fold), libido decreased (3.5-fold), fatigue (1.7-fold), and nervousness (1.4-fold).[5]
Withdrawal effects
Jauhar and colleagues mis-represent the existing studies when they dismiss evidence for withdrawal effects being common, effectively minimising the important task of helping patients to safely stop these medications. Their view is less widely held these days given that severe and long-lasting withdrawal symptoms are recognised by NICE,[42] the Royal College of Psychiatrists [43] and Public Health England.[2] While a systematic review the authors have been highly critical of did include surveys, it is misleading to suggest that this represents the whole review.[44] This review identified 6 double-blind randomised controlled trials that examined withdrawal effects from antidepressants and a weighted average of these studies found that withdrawal effects occurred in 53.8% of patients (albeit slightly less than the 56.4% found when observational studies and surveys were also included).
Instead the authors cite their own paper which included only studies conducted by drug companies which looked at withdrawal effects after an average of 12 weeks of antidepressant treatment.[45] This review is misleading as it is known that longer duration of treatment gives rise to a greater likelihood of withdrawal effects [3] and so estimates based on these short-term studies will underestimate the chance of withdrawal for half of the patients in England who have been on antidepressants for more than 2 years.[46] Even if we accept the figure derived from this review of 31% (after subtraction of the largest nocebo response rate from one of the studies), this means that more than 2 million of the 7 million people in England who are prescribed antidepressants each month are at risk of having trouble stopping their medication due to withdrawal effects.
It is also false to state that there are no staggered double-blind RCTs conducted on this topic – there are at least two such studies.[47,48] In Rosenbaum et al. (1998) patients were allocated to placebo substitution in a staggered design (with two possible periods for substitution) with both raters and patients blind to this procedure.[47] The proportion of patients who experienced more than 4 withdrawal symptoms in this study were 66% for people taking paroxetine, 60% for sertraline, and 14% for fluoxetine (likely because the 5-8 day interruption period was too short for withdrawal symptoms from this long half-life drug to emerge). The other staggered double-blind study found that people on paroxetine experience, on average, 4 new withdrawal symptoms (when subtracting out active treatment) in 4 days of placebo substitution, though the period was likely too short for withdrawal symptoms for sertraline and fluoxetine to reach significance.[48]
Lewis and colleagues have conducted the first ever double-blind randomised controlled study that examines withdrawal effects from antidepressant over periods longer than a few weeks and recorded a significant increase in withdrawal effects that lasted for at least 9 and perhaps 12 months, demonstrating the existence of long-lasting withdrawal effects.[49] Notably, calculation of the standardised mean difference at 3 months for withdrawal symptoms between the maintenance and discontinuation groups from data provided in the original paper yields an SMD of 0.60 (95% CI 0.41 to 0.78), twice the size of the short-term antidepressant effect.
It is possible that withdrawal effects in the maintenance group in this study could reflect the fact that some patients in this group also stopped treatment (15% of patients in the maintenance arm did so in the study, p.34 of Appendix) and/or the fact that there is some overlap of withdrawal symptoms with adverse effects of antidepressants and with depression and anxiety. The presence of some withdrawal symptoms in the maintenance group does not therefore invalidate the important finding that there was a significant increase in withdrawal effects for many months in the discontinuation group, as reported by the study authors in a rapid response.[50]
Jauhar and colleagues correctly find fault with our statement that relapse occurred almost exclusively within 12 weeks of stopping medication. We should have been more precise: the difference in relapse rates between maintenance and discontinuation arms evident at the end of the study (interpreted as relapse prevention properties) were already present at week 20 (which is 12 weeks after week 8, the point at which medication was stopped for most patients). The difference in relapse rate at the end of the study was 17% (56% -39%). At week 20 the difference in relapse rates between the two groups was already 22% (slightly higher than the final difference between the groups). This is consistent with a withdrawal effect because a true prophylactic drug effect should be evenly distributed over time, but in this study the difference between the two groups was evident within 12 weeks of stopping the drugs (just 27% (12/44) of the length of the follow up period of the study after drugs were stopped). As there should be no temporal correlation between the time of relapse for patients with relapsing-remitting conditions with their own individual periodicity, the clustering of relapse in the weeks following cessation is most plausibly explained by withdrawal effects being mis-diagnosed as relapse.[51,52]
Consistent with the less than impressive evidence of efficacy of antidepressants the recent NICE guidance, now in draft form, albeit with some evidence of bias towards present practice, and a methodology that underestimates the harms of treatments, has moved away from medication as the focus for treatment and now offer 9 treatments ‘equally effective’ for less severe depression and 8 treatments that are ‘equally effective’ for more severe depression than either antidepressants alone or antidepressant in combination with CBT.[9] It is notable that many of these treatments including various forms of psychotherapy, exercise, problem solving therapy, and mindfulness have a more favourable side effect profile than medication.
It is unfortunate that Jauhar and colleagues quibble about marginal differences between mood items and the full HDRS, both consistent with small effects of questionable clinical significance and minimise withdrawal effects, when an important project should be reducing the unnecessary use of these drugs and helping patients to safely stop these medications, which in many cases may cause more harm than benefit.
Acknowledgements: We would like to thank Dr Klaus Munkholm (Cochrane Denmark and Centre for Evidence-Based Medicine Odense) for his useful input into this letter.
References
[ 1] Wendy Burn: Medical community must ensure that those needing support to come off antidepressants can get it. The BMJ. 2020; published online Sept 25. https://blogs.bmj.com/bmj/2020/09/25/wendy-burn-medical-community-must-ensure-that-those-needing-support-to-come-off-anti-depressants-can-get-it/ (accessed March 2, 2022).
[2] Public Health England. Dependence and withdrawal associated with some prescribed medicines. An evidence review. 2019. https://www.gov.uk/government/publications/prescribed-medicines-review-report.
[3] NICE. Medicines associated with dependence or withdrawal symptoms: safe prescribing and withdrawal management for adults. Medicines associated with dependence or withdrawal symptoms: safe prescribing and withdrawal management for adults - draft guideline. 2021; published online Oct. https://www.nice.org.uk/guidance/GID-NG10141/documents/draft-guideline (accessed Nov 9, 2021).
[4] Munkholm K, Paludan-Müller AS, Boesen K. Considering the methodological limitations in the evidence base of antidepressants for depression: a reanalysis of a network meta-analysis. BMJ Open 2019; 9: e024886.
[5] Jakobsen JC, Katakam KK, Schou A, et al. Selective serotonin reuptake inhibitors versus placebo in patients with major depressive disorder. A systematic review with meta-analysis and Trial Sequential Analysis. BMC Psychiatry 2017; 17: 58.
[6] Hieronymus F, Nilsson S, Eriksson E. A mega-Analysis of fixed-dose trials reveals dose-dependency and a rapid onset of action for the antidepressant effect of three selective serotonin reuptake inhibitors. Transl Psychiatry 2016; 6: e834-7.
[7] Hieronymus F, Lisinski A, Nilsson S, Eriksson E. Influence of baseline severity on the effects of SSRIs in depression: an item-based, patient-level post-hoc analysis. The Lancet Psychiatry 2019; 6: 745–52.
[8] Barbui C, Cipriani A. Novel melatonin-based treatments for major depression. Lancet. 2012; 379: 215; author reply 217-9.
[9] National Institute for Health and Care Excellence. Depression in adults:Guideline (Draft for consultation, November 2021). https://www.nice.org.uk/guidance/indevelopment/gid-cgwave0725/documents (accessed Dec 2, 2021).
[10] Leucht S, Fennema H, Engel R, Kaspers-Janssen M, Lepping P, Szegedi A. What does the HAMD mean? J Affect Disord 2013; 148: 243–8.
[11] Cipriani A, Furukawa TA, Salanti G, et al. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet 2018; 391: 1357–66.
[12] Man-Son-Hing M, Laupacis A, O’Rourke K, et al. Determination of the clinical importance of study results. J Gen Intern Med 2002; 17: 469–76.
[13] Leichsenring F, Jaeger U, Masuhr O, et al. To be or not to be improved: Patients’ perception of symptom improvement - linking the SCL-90-R to patient-rated global improvement in a large real-world treatment sample. Psychother Psychosom 2020; 89: 357–62.
[14] Barrett B, Brown D, Mundt M, Brown R. Sufficiently important difference: expanding the framework of clinical significance. Med Decis Making 2005; 25: 250–61.
[15] Bauer-Staeb C, Kounali D-Z, Welton NJ, et al. Effective dose 50 method as the minimal clinically important difference: Evidence from depression trials. J Clin Epidemiol 2021; 137: 200–8.
[16] Salaminios G, Duffy L, Ades A, et al. A randomised controlled trial assessing the severity and duration of depressive symptoms associated with a clinically significant response to sertraline versus placebo, in people presenting to primary care with depression (PANDA trial): study protocol for a randomised controlled trial. Trials 2017; 18: 496.
[17] Revicki D, Hays RD, Cella D, Sloan J. Recommended methods for determining responsiveness and minimally important differences for patient-reported outcomes. J Clin Epidemiol 2008; 61: 102–9.
[18] Leichsenring F, Steinert C, Rabung S, Ioannidis JPA. The efficacy of psychotherapies and pharmacotherapies for mental disorders in adults: an umbrella review and meta-analytic evaluation of recent meta-analyses. World Psychiatry 2022; 21: 133–45.
[19] Hengartner MP, Plöderl M. Estimates of the minimal important difference to evaluate the clinical significance of antidepressants in the acute treatment of moderate-to-severe depression. BMJ evidence-based medicine 2021; published online Feb. DOI:10.1136/bmjebm-2020-111600.
[20] Hieronymus F, Lisinski A, Nilsson S, Eriksson E. Efficacy of selective serotonin reuptake inhibitors in the absence of side effects: a mega-analysis of citalopram and paroxetine in adult depression. Mol Psychiatry 2018; 23: 1731–6.
[21] Bagby RM, Ryder AG, Schuller DR, Marshall MB. The Hamilton Depression Rating Scale: has the gold standard become a lead weight? Am J Psychiatry 2004; 161: 2163–77.
[22] Fried EI. Are more responsive depression scales really superior depression scales? J. Clin. Epidemiol. 2016; 77: 4–6.
[23] Volkmann C, Volkmann A, Müller CA. On the treatment effect heterogeneity of antidepressants in major depression: A Bayesian meta-analysis and simulation study. PLoS One 2020; 15: e0241497.
[24] Munkholm K, Winkelbeiner S, Homan P. Individual response to antidepressants for depression in adults-a meta-analysis and simulation study. PLoS One 2020; 15: e0237950.
[25] Maslej MM, Furukawa TA, Cipriani A, et al. Individual Differences in Response to Antidepressants: A Meta-analysis of Placebo-Controlled Randomized Clinical Trials. JAMA Psychiatry 2021; 78: 490–7.
[26] Plöderl M, Hengartner MP. What are the chances for personalised treatment with antidepressants? Detection of patient-by-treatment interaction with a variance ratio meta-analysis. BMJ Open 2019; 9: e034816.
[27] Maslej MM, Furukawa TA, Cipriani A, Andrews PW, Mulsant BH. Individual differences in response to antidepressants: A meta-analysis of placebo-controlled randomized clinical trials. JAMA Psychiatry 2020; 77: 607–17.
[28] Öngür D, Bauchner H. Notice of retraction: Maslej et al. Individual differences in response to antidepressants: A meta-analysis of placebo-controlled randomized clinical trials. JAMA psychiatry. 2020;77(6):607-617. JAMA Psychiatry 2020; 77: 786.
[29] Cortés J, González JA, Medina MN, et al. Does evidence support the high expectations placed in precision medicine? A bibliographic review. F1000Res 2019; 7: 30.
[30] Leucht S, Hierl S, Kissling W, Dold M, Davis JM. Putting the efficacy of psychiatric and general medicine medication into perspective: review of meta-analyses. Br J Psychiatry 2012; 200: 97–106.
[31] Chen JA, Papakostas GI, Youn SJ, et al. Association between patient beliefs regarding assigned treatment and clinical response: reanalysis of data from the Hypericum Depression Trial Study Group. J Clin Psychiatry 2011; 72: 1669–76.
[32] Faria V, Gingnell M, Hoppe JM, et al. Do You Believe It? Verbal Suggestions Influence the Clinical and Neural Effects of Escitalopram in Social Anxiety Disorder: A Randomized Trial. EBioMedicine 2017; 24: 179–88.
[33] Le Noury J, Nardo JM, Healy D, et al. Restoring Study 329: efficacy and harms of paroxetine and imipramine in treatment of major depression in adolescence. BMJ 2015; 351.
[34] Pigott HE, Leventhal AM, Alter GS, Boren JJ. Efficacy and effectiveness of antidepressants: Current status of research. Psychother Psychosom 2010; 79: 267–79.
[35] Pigott HE. The STAR*D Trial: It Is Time to Reexamine the Clinical Beliefs That Guide the Treatment of Major Depression. Can J Psychiatry 2015; 60: 9–13.
[36] Kirsch I, Huedo-Medina TB, Pigott HE, Johnson BT. Do outcomes of clinical trials resemble those of “real world” patients? A reanalysis of the STAR*D antidepressant data set. Psychology of Consciousness: Theory, Research, and Practice 2018; 5: 339–45.
[37] Warren JB. The trouble with antidepressants : why the evidence overplays benefits and underplays risks — an essay by John B Warren. 2020. DOI:10.1136/bmj.m3200.
[38] Read J, Williams J. Adverse Effects of Antidepressants Reported by a Large International Cohort: Emotional Blunting, Suicidality, and Withdrawal Effects. Curr Drug Saf 2018; 13: 176–86.
[39] Bet PM, Hugtenburg JG, Penninx BWJH, Hoogendijk WJG. Side effects of antidepressants during long-term use in a naturalistic setting. Eur Neuropsychopharmacol 2013; 23: 1443–51.
[40] Dragioti E, Solmi M, Favaro A, et al. Association of antidepressant use with adverse health outcomes: A systematic umbrella review. JAMA Psychiatry 2019; 76: 1241–55.
[41] Golder S, Loke YK, Wright K, Norman G. Reporting of Adverse Events in Published and Unpublished Studies of Health Care Interventions: A Systematic Review. PLoS Med 2016; 13: e1002127.
[42] Iacobucci G. NICE updates antidepressant guidelines to reflect severity and length of withdrawal symptoms. BMJ 2019; 367: l6103.
[43] Royal College of Psychiatrists. Position statement on antidepressants and depression. 2019. https://www.rcpsych.ac.uk/docs/default-source/improving-care/better-mh-policy/position-statements/ps04_19---antidepressants-and-depression.pdf?sfvrsn=ddea9473_5.
[44] 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.
[45] Jauhar S, Hayes J, Goodwin GM, Baldwin DS, Cowen PJ, Nutt DJ. Antidepressants, withdrawal, and addiction; where are we now? J Psychopharmacol 2019; 33: 655–9.
[46] Johnson CF, Macdonald HJ, Atkinson P, Buchanan AI, Downes N, Dougall N. Reviewing long-term antidepressants can reduce drug burden: a prospective observational cohort study. Br J Gen Pract 2012; 62: e773–9.
[48] Michelson D, Fava M, Amsterdam J, et al. Interruption of selective serotonin reuptake inhibitor treatment. Double-blind, placebo-controlled trial. Br J Psychiatry 2000; 176: 363–8.
[49] Lewis G, Marston L, Duffy L, et al. Maintenance or Discontinuation of Antidepressants in Primary Care. N Engl J Med 2021; 385: 1257–67.
[50] Lewis GLG. Half of people who stopped long term antidepressants relapsed within a year, study finds: Rapid Response. BMJ. 2021; published online Oct 20. https://www.bmj.com/content/374/bmj.n2403/rr-1 (accessed March 2, 2021).
[51] 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. Therapeutic advances in psychopharmacology 2021; 11: 20451253211032052.
[52] Récalt AM, Cohen D. Withdrawal Confounding in Randomized Controlled Trials of Antipsychotic, Antidepressant, and Stimulant Drugs, 2000–2017. Psychother Psychosom 2019; 88: 105–13.
I find it concerning that the authors pay no attention to the use of Shared Decision Making regarding the use of anti-depressants. Many patients have and continue to benefit from being prescribed an anti-depressant. It is important within prescribing that patients' views, experience and preferences are given due regard.
We read with interest the opinion piece by Horowitz and Wilcock [1], which highlights ‘considerable uncertainty about the benefit of antidepressant use in the short- and long-term’ and ‘the uncertain balance of benefits and harms’ with antidepressants, within its key learning points. We wish to counter these and address other assertions within their narrative review.
In the section ‘Questions relating to efficacy’, the authors mention meta-analyses with newer generation antidepressants (principally the selective serotonin reuptake inhibitors [SSRIs] and serotonin-noradrenaline reuptake inhibitors [SNRIs]) which show a difference from placebo of around 2 points on the Hamilton Depression Rating Scale (HDRS), noting this does not meet thresholds of 3-, 5- or even 6-point differences, regarded by some as indicating a minimally clinically important difference.
This is taken from data on the antidepressant mirtazapine (neither an SSRI nor SNRI) plotting change in HDRS against clinically important difference [2]. This does not have any relevance for SSRIs or SNRIs, as mirtazapine, by its effects on sleep and somatic symptoms, will exert large effects on HDRS without necessarily affecting low mood. The selection of a point on this graph was not made by authors of the original paper [3], and within-group changes from that selected study cannot be extrapolated to placebo-drug differences: a point made explicit by experts in this area [4], and by us in response to Dr H...
We read with interest the opinion piece by Horowitz and Wilcock [1], which highlights ‘considerable uncertainty about the benefit of antidepressant use in the short- and long-term’ and ‘the uncertain balance of benefits and harms’ with antidepressants, within its key learning points. We wish to counter these and address other assertions within their narrative review.
In the section ‘Questions relating to efficacy’, the authors mention meta-analyses with newer generation antidepressants (principally the selective serotonin reuptake inhibitors [SSRIs] and serotonin-noradrenaline reuptake inhibitors [SNRIs]) which show a difference from placebo of around 2 points on the Hamilton Depression Rating Scale (HDRS), noting this does not meet thresholds of 3-, 5- or even 6-point differences, regarded by some as indicating a minimally clinically important difference.
This is taken from data on the antidepressant mirtazapine (neither an SSRI nor SNRI) plotting change in HDRS against clinically important difference [2]. This does not have any relevance for SSRIs or SNRIs, as mirtazapine, by its effects on sleep and somatic symptoms, will exert large effects on HDRS without necessarily affecting low mood. The selection of a point on this graph was not made by authors of the original paper [3], and within-group changes from that selected study cannot be extrapolated to placebo-drug differences: a point made explicit by experts in this area [4], and by us in response to Dr Horowitz in the past [5].
The HDRS was conceived over 60 years ago, and more nuanced analyses, examining the key mood items, have shown clear benefits for SSRIs, using primary data from randomised double-blind, placebo-controlled trials (RCTs) [6]. The authors cite a review addressing these very issues [7], though do not acknowledge any of these points. That review also addressed practical aspects from trials which suggest the need for more care when evaluating effect sizes, illustrating the points above regarding the HDRS, study level data and practicalities in RCTs that make the expectation of high effect sizes implausible. In critiquing this review, Horowitz and Wilcock argue these inferences are made erroneously on the assumption of “a subgroup with greater than average response.” to SSRIs, and cite a study-level analysis of heterogeneity8 that they claim supports the notion that response to antidepressants is constant across patients – i.e. if the mean antidepressant-placebo difference in a trial is 2 HDRS points, then all patients receiving antidepressants in that study will have a 2-point improvement over placebo. The cited paper correctly acknowledges their analysis cannot disprove a non-constant effect [8]. However, most would agree that a non-constant effect occurs, that data from RCTs are normally distributed, and that there is individual variability in antidepressant and placebo response [9]. Critically, there are patient-level analyses supporting a non-constant effect of antidepressants in depression (Maslej et al., 2021) [10]. Accounting for problems with the HDRS, the ‘number needed to treat’ (NNT) for antidepressants, taken from short-term efficacy trials, is around 7-10, a range which is not dissimilar to many other medical treatments [11].
When the authors consider regulatory issues, they continue, “indeed, longer studies of antidepressants show poor outcomes: in the STAR-D trial only 108 (2.7%), out of 4041 participants who originally enrolled in the study, achieved remission and did not relapse or drop out at the endpoint of the study at 12 months”. However, the original publication in the New England Journal of Medicine [12] shows that STAR-D randomised people with depression to staged treatment with symptom remission as outcome. The reference cited by Horowitz and Wilcock is to a secondary analysis by non-trial authors [13], which is difficult to interpret as 36.7% of patient participants dropped out of the study one month after remission: cumulative remission in the STAR-D study was 67%, though this may be an over-estimate as drop-outs were not included in this figure (and there was no placebo group).
The section on ‘harms of long-term use of antidepressants’ cites a naturalistic study [14] and a self-selected survey [15] in which the survey response rate is not mentioned. Horowitz and Wilcock state that rates of adverse effects are ‘often greater in naturalistic studies of long-term antidepressant users’ (our italics) than those mentioned in 6-12 week efficacy studies’. The authors of the cited naturalistic study state, "We find slightly lower side effects (SE) reporting with prolonged antidepressant use, while no associations are detected between specific SEs and prolonged use. During long-term antidepressant use, all specific SEs seem to persist.“ Horowitz and Wilcock overlooked a recent substantial ‘umbrella review’ of meta-analyses [16], which found a low rate of adverse effects attributable to antidepressants. This is particularly important as this umbrella review included cohorts of people not receiving antidepressants, therefore adjusting for a common problem with observational studies: confounding by indication.
A subsequent paragraph on withdrawal effects with antidepressants notes a recent (2019) Public Health Review acknowledging the very low to moderate quality of the evidence. Horowitz and Wilcock cite a review of the incidence, duration and severity of antidepressant withdrawal reactions [17] to assert that “about half” of patients experience withdrawal symptoms on stopping antidepressants, abruptly or with taper; however, not all evidence is aligned with this contention (e.g. the sample of “half” included self-selected surveys with no mention of response rate; and RCTs indicate substantially lower rates, varying by compound) [18-19]. Randomised placebo-controlled studies with a double-blind staggered design probably represent the best approach to ascertain the relative burden of symptoms which emerge during or after stopping placebo or antidepressant treatment.
Horowitz and Wilcock appear to favour use of the Discontinuation Emergent Signs and Symptoms (DESS) scale15 for assessing antidepressant withdrawal symptoms, although there are potential flaws in the utility of this check-list instrument in research and clinical practice [20]. They cite use of a modified version of the DESS (a 15-item checklist) [21] in a recent RCT of antidepressant treatment discontinuation in UK primary care [22] stating that withdrawal effects were ‘almost triple’ in the discontinuation group. In the original publication, the mean score for withdrawal symptoms was reported as 1.3+2.4 in the maintenance group and 3.1+3.5 in the discontinuation group, with an estimated difference between groups of 1.9 points [95% CI, 1.5-2.3]. Many participants experienced an increase in withdrawal symptoms, regardless of allocation to continued treatment or treatment discontinuation. The further problem of the DESS picking up symptoms of relapse can only be clarified by a fuller examination of the patient level data.
Elucidating the clinical validity of these changes in measures based on rating scales would seem reasonable before any pronouncement about withdrawal effects can be made definitively. Further work, ideally using the double-blind placebo-controlled staggered design, is needed to fully characterise the overall burden of symptoms following the withdrawal of antidepressant treatment (with consideration of incidence, severity, persistence, and impairment). Horowitz and Wilcock make a factual error when reporting findings in this UK study: they say that new depressive episodes (relapses) in the group who had discontinued antidepressants “occurred almost exclusively within 12 weeks of stopping medication, the time during which withdrawal effects were most pronounced”, thus implying these events were not depressive relapses but rather withdrawal symptoms. However, the Kaplan–Meier analysis from the paper clearly indicates that less than half of all relapses occurred within 12 weeks.
It is difficult to draw reliable conclusions from the selective reporting of Horowitz and Wilcock’s viewpoint. The authors appear to have come to their own conclusions, and then selectively cited only the evidence that supports them.
References
[1] Horowitz M, Wilcock M. Newer generation antidepressants and withdrawal effects: reconsidering the role of antidepressants and helping patients to stop. DTB 2022; 60: 7–12.
[2] Leucht S, Fennema H, Engel R, Kaspers-Janssen M, Lepping P, Szegedi A. What does the HAMD mean? J Affect Disord 2013; 148: 243–8.
[3] Moncrieff J, Kirsch I. Empirically derived criteria cast doubt on the clinical significance of antidepressant-placebo differences. Contemp Clin Trials 2015; 43: 60–2.
[4] Furukawa TA. Measuring Clinical Importance in a Trial of Interventions for Mixed Urinary Incontinence. JAMA 2020; 323: 479.
[5] Jauhar S, Marwaha S, Morrison PD, Upthegrove R. Weighing up scientific evidence requires balance, not opinion. The British Journal of Psychiatry 2021; 219: 619–619.
[6] Hieronymus F, Lisinski A, Nilsson S, Eriksson E. Efficacy of selective serotonin reuptake inhibitors in the absence of side effects: a mega-analysis of citalopram and paroxetine in adult depression. Mol Psychiatry 2017; published online July 25. DOI:10.1038/mp.2017.147.
[7] Hieronymus F, Jauhar S, Østergaard SD, Young AH. One (effect) size does not fit at all: Interpreting clinical significance and effect sizes in depression treatment trials. J Psychopharmacol 2020; : 0269881120922950.
[8] Volkmann C, Volkmann A, Müller CA. On the treatment effect heterogeneity of antidepressants in major depression: A Bayesian meta-analysis and simulation study. PLOS ONE 2020; 15: e0241497.
[9] Maslej MM, Furukawa TA, Cipriani A, Andrews PW, Mulsant BH. Individual Differences in Response to Antidepressants: A Meta-analysis of Placebo-Controlled Randomized Clinical Trials. JAMA Psychiatry 2020; 77: 607–17.
[10] Hieronymus F, Lisinski A, Hieronymus M, Näslund J, Eriksson E, Østergaard SD. Determining maximal achievable effect sizes of antidepressant therapies in placebo-controlled trials. Acta Psychiatrica Scandinavica 2021; 144: 300–9.
[11] Leucht S, Hierl S, Kissling W, Dold M, Davis JM. Putting the efficacy of psychiatric and general medicine medication into perspective: review of meta-analyses. The British Journal of Psychiatry 2012; 200: 97–106.
[12] Rush AJ, Trivedi MH, Wisniewski SR, et al. Bupropion-SR, Sertraline, or Venlafaxine-XR after Failure of SSRIs for Depression. New England Journal of Medicine 2006; 354: 1231–42.
[13] Pigott HE, Leventhal AM, Alter GS, Boren JJ. Efficacy and effectiveness of antidepressants: current status of research. Psychother Psychosom 2010; 79: 267–79.
[14] Bet PM, Hugtenburg JG, Penninx BWJH, Hoogendijk WJG. Side effects of antidepressants during long-term use in a naturalistic setting. Eur Neuropsychopharmacol 2013; 23: 1443–51.
[15] Read J, Williams J. Adverse Effects of Antidepressants Reported by a Large International Cohort: Emotional Blunting, Suicidality, and Withdrawal Effects. Curr Drug Saf 2018; 13: 176–86.
[16] Dragioti E, Solmi M, Favaro A, et al. Association of Antidepressant Use With Adverse Health Outcomes: A Systematic Umbrella Review. JAMA Psychiatry 2019; 76: 1241–55.
[17] Davies J, Read J. A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based? Addict Behav 2018; published online Sept 4. DOI:10.1016/j.addbeh.2018.08.027.
[18] Jauhar S, Hayes J. The war on antidepressants: What we can, and can’t conclude, from the systematic review of antidepressant withdrawal effects by Davies and Read. Addict Behav 2019; published online Jan 23. DOI:10.1016/j.addbeh.2019.01.025.
[19] Jauhar S, Hayes J, Goodwin GM, Baldwin DS, Cowen PJ, Nutt DJ. Antidepressants, withdrawal, and addiction; where are we now? J Psychopharmacol (Oxford) 2019; : 269881119845799.
[20] Baldwin DS. A personal account of reducing and stopping antidepressant treatment. Int Clin Psychopharmacol 2020; 35: 194–200.
[22] Lewis G, Marston L, Duffy L, et al. Maintenance or Discontinuation of Antidepressants in Primary Care. New England Journal of Medicine 2021; 385: 1257–67.
We appreciate the interest in our article by Jauhar and colleagues and welcome the opportunity to address a number of misleading points in their letter.
No argument they raise undermines our central points: antidepressants do not demonstrate a clinically important difference from placebo according to any suggested threshold, including that proposed by NICE, empirical correlation with clinician evaluation or other means. There is no conclusive evidence of an above-average response sub-group so far demonstrated. The evidence for relapse prevention properties is highly problematic because of confounding by withdrawal effects. Common adverse effects are well-established, a...
Show MoreI find it concerning that the authors pay no attention to the use of Shared Decision Making regarding the use of anti-depressants. Many patients have and continue to benefit from being prescribed an anti-depressant. It is important within prescribing that patients' views, experience and preferences are given due regard.
We read with interest the opinion piece by Horowitz and Wilcock [1], which highlights ‘considerable uncertainty about the benefit of antidepressant use in the short- and long-term’ and ‘the uncertain balance of benefits and harms’ with antidepressants, within its key learning points. We wish to counter these and address other assertions within their narrative review.
In the section ‘Questions relating to efficacy’, the authors mention meta-analyses with newer generation antidepressants (principally the selective serotonin reuptake inhibitors [SSRIs] and serotonin-noradrenaline reuptake inhibitors [SNRIs]) which show a difference from placebo of around 2 points on the Hamilton Depression Rating Scale (HDRS), noting this does not meet thresholds of 3-, 5- or even 6-point differences, regarded by some as indicating a minimally clinically important difference.
This is taken from data on the antidepressant mirtazapine (neither an SSRI nor SNRI) plotting change in HDRS against clinically important difference [2]. This does not have any relevance for SSRIs or SNRIs, as mirtazapine, by its effects on sleep and somatic symptoms, will exert large effects on HDRS without necessarily affecting low mood. The selection of a point on this graph was not made by authors of the original paper [3], and within-group changes from that selected study cannot be extrapolated to placebo-drug differences: a point made explicit by experts in this area [4], and by us in response to Dr H...
Show More