Approaches for discontinuation versus continuation of long-term antidepressant use for depressive and anxiety disorders in adults

In Bockting 2018, 19% (8/43) of participants in the discontinuation group reduced their dose by at least 50%.

“16 of 85 (split 8/43) (19%) were able to reduce the dose by a minimum of 50%.”

• cites Bockting 2018 sourceDisputed

  In the Results baseline table (page unavailable), 16/85 (19%) refers to participants receiving current psychological or psychotherapeutic treatment, not to antidepressant dose reduction. The Statistical analysis section defines a 50% dose reduction for a per-protocol analysis, but does not report 8/43 or 16/85 as achieving it. (LLM, conf=0.86)

  Checked by a reasoning model — may be wrong.

  Experimental — accuracy not yet measured

Randomised studies with tapering periods up to 14 days showed no or minimal reduction in withdrawal symptom severity compared with abrupt discontinuation.

“randomised studies with a tapering period up to 14 days showed either no or minimal reduction in withdrawal symptom severity compared with abrupt discontinuation (Baldwin 2006;Bogetto 2002;Montgomery 2004a).”

• cites Baldwin 2006 sourceDisputed

  The Tolerability assessments section describes a randomized comparison of stopping treatment immediately versus continuing the same dose for one more week before stopping—not tapering versus abrupt discontinuation. The DESS results section says the dose-reduction group had no randomized withdrawal comparison and omits its results from the figure. This paper therefore does not show that tapering reduced withdrawal symptoms minimally or not at all compared with abrupt discontinuation. (LLM, conf=0.86)

  Checked by a reasoning model — may be wrong.

  Experimental — accuracy not yet measured
• cites Bogetto 2002 sourceNot checked — no full text of the source
• cites Montgomery 2004aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The trials supporting continuation treatment did not measure withdrawal effects.

“although none of the trials on which this conclusion was based measured withdrawal effects (Cohen 2019;Hengartner 2020;Recalt 2019)”

• cites Cohen 2019 sourceNot checked — no full text of the source
• cites Hengartner 2020 sourcePartly supported

  The ‘Withdrawal confounding in relapse prevention trials’ section says that symptom scales used to assess relapse cannot distinguish withdrawal symptoms from depression relapse. The ‘Summary and conclusion’ likewise says withdrawal confounding makes the trials difficult to interpret. Neither passage establishes the stronger claim that none of the trials measured withdrawal effects. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites Recalt 2019 sourceNot checked — no full text of the source

Antidepressant treatment for up to one year results in lower relapse rates among responders compared with discontinuation in anxiety disorders, irrespective of anxiety disorder type.

“Antidepressant treatment duration of up to one year results in lower relapse rates among responders compared with treatment discontinuation in anxiety disorders, irrespective of the type of anxiety disorder (Batelaan 2017).”

• cites Batelaan 2017 sourceSupported

In Segal 2010, among stable remitters, there was no difference in relapse between discontinuation and continuation groups (50% vs 59%; P=0.49).

“For stable remitters (all HAM-D scores ≥ 7 during remission), there were no differences in relapse of absence of withdrawal symptoms measured a er antidepressants were tapered (relapse rates: 50% discontinuation group, 59% antidepressant group; HR not reported; P = 0.49).”

• cites Segal 2010 sourcePartly supported

  The Relapse section reports 50% relapse with placebo after antidepressant taper versus 59% with maintenance antidepressants among stable remitters, with P=.49 for that comparison. However, the Study Phases section defines stable remission as HRSD scores ≤7, not ≥7 as the citing sentence states. The cited passages also do not establish the sentence’s suggestion about withdrawal symptoms. (LLM, conf=0.90)

  Experimental — accuracy not yet measured

Wilson 2003 reported 2/57 (3.5%) discontinuation and 2/56 (3.6%) continuation participants experienced adverse events necessitating discontinuation.

“Wilson 2003 (n = 113) reported that 2 of 57 (3.5%) participants in remission of a first depression in the discontinuation group experienced adverse events that necessitated discontinuation compared to 2 of 56 (3.6%) participants treated with antidepressants.”

• cites Wilson 2003 sourcePartly supported

  The Study sample section reports 56 participants assigned to sertraline and 57 to placebo, matching the claim’s group sizes. However, the passages in our copy do not report two adverse-event-related discontinuations in either group; searches for adverse events and reasons for withdrawal did not establish those counts. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

In Segal 2010, among unstable remitters, the discontinuation group had a higher risk of relapse than the continuation group (71% vs 27%; HR 4.16, 95% CI 1.12 to 14.3; P=0.03).

“Among unstable remitters (one or more with HAM-D score ≥ 7 during remission), participants in the discontinuation group showed higher risk of relapse compared with those in the continuation group (relapse rates: discontinuation group 71%, 27% antidepressant; HR 4.16, 95% CI 1.12 to 14.3; P = 0.03).”

• cites Segal 2010 sourcePartly supported

  The Relapse section reports adjusted relapse rates of 71% for placebo (antidepressant discontinuation) and 27% for maintenance antidepressants among unstable remitters, with P=.03. Its hazard ratio is expressed in the opposite direction: 0.24 (95% CI 0.07–0.89) for maintenance antidepressants versus placebo, approximately 4.16 (1.12–14.3) when inverted. The Study Phases section defines unstable remission by occasional elevated HRSD scores of 8–14, not by a score of ≥7, as the citing sentence states. Page numbers are unavailable in our copy. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

In Bockting 2018, 40% (17/43) of participants in the discontinuation with PCT group discontinued their antidepressant after six months.

“Bockting 2018 reported that 34 of 85 (split to allow multiple-arm comparisons: 17/43) (40%) participants in the discontinuation with PCT group followed the discontinuation advice and discontinued their antidepressant a er six months”

• cites Bockting 2018 sourcePartly supported

  The Results (page unavailable) identify 85 participants assigned to PCT while tapering antidepressants, supporting the group size. They report that 30 of 50 participants still in the study after six months adhered to the antidepressant protocol, but do not report that 34 of 85 discontinued antidepressants or establish the citing paper’s split figure of 17/43. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

Huijbers 2016 reported no differences between groups on six quality of life scales after 15 months of follow-up.

“Huijbers 2016 reported no differences between groups on six scales (World Health Organization (WHO) Quality of Life overall perception, Q1 and Q2, and WHO Quality of Life emotional, physical, psychological, and environmental domains) a er 15 months of follow-up (see Table 2).”

• cites Huijbers 2016 sourcePartly supported

  The Results (Secondary outcomes; page unavailable) report no between-group differences in quality of life at 15 months, supporting the main finding. Our copy does not identify the six named scales or report their results separately, so that level of specificity is not verified. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

In Kuyken 2015, 162 of 212 (76%) participants in the continuation group remained on the therapeutic dose and 50 (24%) did not.

“In all, 162 of 212 (76%) participants in the continuation group remained on the therapeutic dose and 50 (24%) did not remain on the therapeutic dose.”

• cites Kuyken 2015 sourceSupported

In Bockting 2018, 38 of 100 participants in the control group adhered to antidepressant treatment (maintaining at least 20 mg) after six months.

“In all, 38 of 100 participants in the control group adhered to antidepressant treatment (maintaining ≥ 20 mg antidepressant) a er six months.”

• cites Bockting 2018 sourcePartly supported

  The Results report that 38 of 66 participants still in the antidepressants-alone group after six months adhered to the antidepressant protocol; 100 were originally assigned to that group. The Procedures specify a maintenance dose of at least 20 mg fluoxetine equivalent, but the claim uses the original group size as the adherence denominator rather than the 66 assessed at six months. (LLM, conf=0.89)

  Experimental — accuracy not yet measured

There was no difference in risk of relapse between participants discontinuing antidepressants with support from MBCT or PCT compared to those continuing antidepressants (HR 0.89, 95% CI 0.66 to 1.19; 690 participants, 3 studies).

“There was no difference in risk of relapse (as defined by study authors) between participants discontinuing antidepressants with support from MBCT or PCT compared to those continuing antidepressants (HR 0.89, 95% CI 0.66 to 1.19; 690 participants, 3 studies).”

• cites Bockting 2018 sourcePartly supported

  The Results report no significant difference in recurrence between PCT with antidepressant tapering and continued antidepressants (HR 0.86, 95% CI 0.56–1.32). The Discussion also describes three earlier MBCT trials as finding no clinically meaningful difference from maintenance medication. The paper does not report the claimed pooled HR of 0.89 (95% CI 0.66–1.19) for 690 participants across three MBCT-or-PCT studies. (LLM, conf=0.88)

  Experimental — accuracy not yet measured
• cites Kuyken 2008 sourceNot checked — no full text of the source
• cites Kuyken 2015 sourcePartly supported

  The Summary and Results report no significant difference in relapse for the PREVENT trial’s 424 participants (HR 0.89, 95% CI 0.67–1.18). The Systematic review section also describes pooling three trials, but reports a different measure and result: a relative risk of 0.76 (95% CI 0.59–0.98). It does not support the claimed three-study HR, confidence interval, 690 participants, or inclusion of PCT. Page numbers were not provided for these passages. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

In Kuyken 2015, 13.7% (29/212) of participants in the discontinuation group reduced their dose by at least 50%.

“and 29 of 212 (13.7%) reduced the dose by a minimum of 50%.”

• cites Kuyken 2015 sourcePartly supported

  Table 2 (Results; page not available) reports that 29 participants reduced their antidepressant dose, but this was among 176 participants who attended at least four MBCT-TS sessions (17%), not a reported 29 of 212 (13.7%). The table does not specify that their reductions were at least 50%. (LLM, conf=0.89)

  Experimental — accuracy not yet measured

Bockting 2018 reported mean QALYs over 24 months of 1.59 (range 0.64 to 1.94) for the discontinuation with PCT group (n=42) and 1.62 (range 0.95 to 1.95) for the continuation and PCT group (n=104).

“Mean QALYs over 24 months were 1.59 (range 0.64 to 1.94) for the discontinuation with PCT group (n = 42) and 1.62 (range 0.95 to 1.95) for the continuation and PCT group (n = 104).”

• cites Bockting 2018 sourceCould not verify

  The Outcomes section (page unavailable) says cost-effectiveness results would be reported elsewhere; our copy does not report QALYs. The Results identify a 24-month follow-up and 104 participants allocated to PCT plus maintenance antidepressants, but do not give the claimed QALY means, ranges, or discontinuation-group sample of 42 (searched: QALYs, quality-adjusted life years, and the stated values). (LLM, conf=0.78)

  Experimental — accuracy not yet measured

There were no differences between discontinuation and continuation in the physical domain of quality of life (MD -0.22, 95% CI -2.16 to 1.73; 455 participants, 2 studies; I2 = 65%; random-effects model).

“Data show no differences between discontinuation and continuation in the physical domain (MD -0.22, 95% CI -2.16 to 1.73; 455 participants, 2 studies; I = 65%; random-effects model)”

• cites Kuyken 2008 sourceNot checked — no full text of the source
• cites Kuyken 2015 sourcePartly supported

  The Results report no evidence that mindfulness-based cognitive therapy with support to taper or discontinue antidepressants was superior to maintenance antidepressants on secondary outcomes, and include the WHO-QoL physical-health domain. But this is a single trial comparing treatment strategies, not a two-study random-effects analysis of discontinuation versus continuation; it does not report the quoted pooled mean difference, confidence interval, participant count, or I². (LLM, conf=0.88)

  Experimental — accuracy not yet measured

There were no differences between discontinuation and continuation in the social relations domain of quality of life (MD 0.05, 95% CI -0.56 to 0.66; 455 participants, 2 studies; I2 = 9%; random-effects model).

“or the social relations domain (MD 0.05, 95% CI -0.56 to 0.66; 455 participants, 2 studies; I = 9%; random-effects model) a er longterm follow-up.”

• cites Kuyken 2008 sourceNot checked — no full text of the source
• cites Kuyken 2015 sourcePartly supported

  The quality-of-life table (page unavailable in our copy) reports similar social-relationships scores for mindfulness-based cognitive therapy with support to taper or discontinue antidepressants and maintenance antidepressant treatment over follow-up. This single trial does not report the claimed pooled mean difference, confidence interval, 455 participants across two studies, or I² statistic; its intervention was not discontinuation alone. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

There was no difference in depressive symptom severity on the HAM-D scale between antidepressant discontinuation with MBCT and continuation of antidepressant (MD -0.42, 95% CI -1.82 to 0.98; 484 participants, 2 studies; I2 = 28%).

“Data show no difference in depressive symptom severity on the HAM-D scale between antidepressant discontinuation with MBCT and continuation of antidepressant (MD -0.42, 95% CI -1.82 to 0.98; 484 participants, 2 studies; I = 28%).”

• cites Kuyken 2008 sourceNot checked — no full text of the source
• cites Kuyken 2015 sourcePartly supported

  The paper’s Table 4 reports GRID-HAMD symptom scores for MBCT with tapering support and maintenance antidepressants, consistent with no clear difference between the groups. But the Results describe one trial with 424 participants; this paper does not report the claimed pooled mean difference, confidence interval, 484 participants across two studies, or I² statistic. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

In Segal 2010, there was no difference between discontinuation with MBCT and continuation of antidepressants in stable remitters (62% vs 59%; P = 0.77).

“However, there was no difference between discontinuation with MBCT and continuation of antidepressants in stable remitters (62% for MBCT and discontinuation, 59% for continuation of antidepressant; P = 0.77)”

• cites Segal 2010 sourceSupported

In Segal 2010, there was no difference between discontinuation with MBCT and continuation of antidepressants in unstable remitters (28% vs 27%; HR 1.07, 95% CI 0.25 to 4.49; P = 0.93).

“and unstable remitters (28% for MBCT and discontinuation, 27% for continuation of antidepressant; HR 1.07, 95% CI 0.25 to 4.49; P = 0.93).”

• cites Segal 2010 sourceSupported

There were no differences between discontinuation and continuation in the psychological domain of quality of life (MD 0.37, 95% CI -0.75 to 1.49; 455 participants, 2 studies; I2 = 46%; random-effects model).

“the psychological domain (MD 0.37, 95% CI -0.75 to 1.49; 455 participants, 2 studies; I = 46%; random-effects model)”

• cites Kuyken 2008 sourceNot checked — no full text of the source
• cites Kuyken 2015 sourcePartly supported

  The Results table reports psychological quality-of-life scores for the trial’s two groups, MBCT-TS and maintenance antidepressant treatment, without a clear between-group difference. This is one randomised trial with 424 participants, not the claimed two-study random-effects analysis of 455 participants; the stated mean difference, confidence interval and I² are not reported in our copy. The indexed passages provide no page number. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

Withdrawal reactions are common in clinical practice.

“Withdrawal reactions are common in clinical practice.”

• cites Davies 2019 sourceSupported

One included study found that most participants, 60%, tapered antidepressants over six months.

“However, one included study found that most participants (60%) tapered antidepressants over six months”

• cites Bockting 2018 sourceSupported

Research suggests that misclassified withdrawal could pertain to most relapses.

“Research suggests this could pertain to most relapses (El-Mallakh 2012; Greenhouse 1991; Hengartner 2020; Recalt 2019; Rosenbaum 1988).”

• cites El-Mallakh 2012 sourceNot checked — no full text of the source
• cites Greenhouse 1991 sourceNot checked — no full text of the source
• cites Hengartner 2020 sourcePartly supported

  In “Withdrawal confounding in relapse prevention trials,” the paper says that possibly a majority of events recorded as relapses in the discontinuation arm of maintenance studies are due to withdrawal reactions. It also says withdrawal symptoms can be misdiagnosed as relapse, but distinguishes that from withdrawal triggering a genuine relapse. The claim does not preserve either distinction and extends the possibility beyond those trials. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

A four-week time frame is not feasible for many long-term users of antidepressants.

“suggesting that a time frame of four weeks is not feasible for many long-term users of antidepressants (Bockting 2018).”

• cites Bockting 2018 sourceSupported

The HAM-D has been criticised for problems in differentiating between withdrawal and depressive symptoms.

“However, this scale has been criticised because of its problems in differentiating between withdrawal and depressive symptoms (Fava 2015: Haddad 2007).”

• cites Fava 2015 sourcePartly supported

  The Discussion says withdrawal symptoms can be mistaken for signs of depressive relapse, supporting the general difficulty in distinguishing them. It does not identify the HAM-D or criticise that scale specifically (searched: HAM-D and Hamilton Depression Rating Scale). The indexed passages provide no page number. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites Haddad 2007 sourcePartly supported

  The ‘Antidepressant discontinuation symptoms’ and ‘Differential diagnosis’ sections say withdrawal symptoms may be mistaken for recurrent depression and explain how clinicians can distinguish them. They do not mention the HAM-D or criticise its ability to make that distinction. The indexed passages do not specify individual page numbers. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

A Canadian primary care study of 8119 patients found that antidepressants were prescribed longer than needed, 15 months or longer with a mean duration of 4.8 years, for 46% of patients.

“A Canadian primary care study (n = 8119) found that antidepressants were prescribed longer than needed (15 months or longer, with a mean duration of 4.8 years) for 46% of patients (Mangin 2018).”

• cites Mangin 2018 sourcePartly supported

  The Abstract and Main Study Findings report that 3,766 of 8,119 patients prescribed antidepressants (46%) met the legacy-prescribing threshold of longer than 15 months; their mean prescription duration was 4.8 years. But the Strengths and Limitations section says some prescriptions meeting the criteria may have been clinically appropriate and that individual appropriateness was beyond the study’s scope. Thus the study does not establish that treatment was longer than needed for every patient in the 46%. Page numbers were not provided in our copy. (LLM, conf=0.87)

  Experimental — accuracy not yet measured

Withdrawal symptoms may be misdiagnosed as relapse when relapse is assessed with the HAM-D scale.

“Withdrawal symptoms may be misdiagnosed as relapse when relapse is assessed with the HAM-D scale (see above, Overall completeness and applicability of evidence) (Fava 2015: Haddad 2007).”

• cites Fava 2015 sourcePartly supported

  The Discussion (page not provided) says SSRI withdrawal symptoms may be misidentified as signs of impending relapse. It does not link that risk specifically to assessing relapse with the HAM-D scale; searches for “HAM-D” and “Hamilton” found no such passage. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites Haddad 2007 sourcePartly supported

  The ‘Antidepressant discontinuation symptoms’ section says withdrawal symptoms may be diagnosed as relapse or recurrence, and ‘Nature of symptoms’ gives recurrence of depression as an example. The cited paper does not link that risk to assessment with the HAM-D scale. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

A possible reason this review did not confirm the 60-week MBCT effect is that it pooled studies with their original longer durations and included a later published study.

“possibly because we pooled studies with their original longer duration (e.g.longer than 60 weeks) (as defined in our protocol) and included a later published study (Bockting 2018).”

• cites Bockting 2018 sourceSupported

This review could not confirm Maund 2019's finding of higher withdrawal symptom risk with abrupt discontinuation versus tapering.

“but we could not confirm this finding”

• cites Maund 2019 sourceSupported
  Experimental — accuracy not yet measured

Studies without active placebo are considered to be at unclear or high risk of performance bias.

“Studies without active placebo are considered to be at unclear or high risk of performance bias (Munkholm 2019).”

• cites Munkholm 2019 sourcePartly supported

  In Results > Risk of bias (page unavailable), the authors argue that placebo-controlled antidepressant trials should arguably be rated at least unclear, or perhaps high risk of bias because adverse effects can compromise blinding. Results > Summary risk of bias assessment says they rated those trials as unclear in the blinding domains. The citing sentence drops the antidepressant-trial scope and presents this qualified judgment as a general rule for studies without active placebo. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

This review could not confirm the significant effect of MBCT versus continuation of antidepressants on reducing relapse or recurrence at 60 weeks found in an individual-patient data meta-analysis.

“Our review could not confirm this significant effect at 60 weeks”

• cites Kuyken 2016 sourceSupported

The finding from Maund 2019 that prompting general practitioners to review long-term antidepressant users was not effective for discontinuation agrees with this review's findings.

“This finding is consistent with the findings of our review.”

• cites Maund 2019 sourceSupported
  Experimental — accuracy not yet measured

Most antidepressants are prescribed by general practitioners.

“Research suggests that most antidepressants are prescribed by GPs (Johnson 2017)”

• cites Johnson 2017 sourceSupported

Horowitz 2019 found four studies showing tapering over several months results in fewer withdrawal symptoms.

“Review authors found four studies -Groot 2013;Himei 2006;Murata 2010;van Geffen 2005 -showing that tapering over several months results in fewer withdrawal symptoms (Horowitz 2019).”

• cites Horowitz 2019 sourcePartly supported

  The tapering discussion (page unavailable) reports better withdrawal outcomes for longer tapers in studies by Himei, van Geffen, and Murata. It also reports a favourable discontinuation outcome for a Groot study, but the References list that study as Groot and van Os (2018), not Groot (2013), and the reported outcome is successful discontinuation rather than fewer withdrawal symptoms. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

An individual-patient data meta-analysis found MBCT compared with continuation of antidepressants was associated with reduced relapse or recurrence risk (HR 0.77, 95% CI 0.66-0.98) over 60 weeks across four studies.

“An individual-patient data meta-analysis found evidence that MBCT compared to continuation of antidepressant treatment was associated with a reduction in the risk of relapse or recurrence, with a hazard ratio (HR) of 0.77 (95% confidence interval (CI) 0.66 to 0.98) over 60 weeks (4 studies) (Kuyken 2016).”

• cites Kuyken 2016 sourcePartly supported

  The Results section reports an individual-patient-data comparison of MBCT with antidepressant medication across four studies, with reduced relapse risk over 60 weeks and an HR of 0.77. The reported 95% CI is 0.60–0.98, not the claimed 0.66–0.98. (LLM, conf=0.92)

  Experimental — accuracy not yet measured

Across 3 RCTs with 690 participants, risk of relapse over 64 to 104 weeks had HR 0.89 (95% CI 0.66 to 1.19) for discontinuation versus continuation, with low certainty evidence.

“Risk of relapse (as defined by study authors) (HR) Follow-up: 64 to 104 weeks 556 per 1000 b 41 fewer per 1000 (from 113 to 38 more) HR 0.89 (0.66 to 1.19) 690 (3 RCTs) ⊕⊕⊝⊝ LOW c”

Across 2 RCTs with 455 participants, psychological health quality of life had MD 0.37 higher for discontinuation versus continuation.

“Mean QoL psychological health to-MD 0.37 higher -455 (2 RCTs)”

Across 2 RCTs with 484 participants, depressive symptoms on HAM-D at 64-104 weeks had MD 0.42 lower (95% CI 1.82 lower to 0.89 higher) for discontinuation versus continuation, with low certainty.

“Depressive symptoms Assessed with HAM-D Scale Follow-up: 64 to 104 weeks Mean HAM-D total endpoint score ranged from 4.7 to 8.69 MD 0.42 lower (1.82 lower to 0.89 higher) -484 (2 RCTs) ⊕⊕⊝⊝ LOW e”

None of the included studies reported anxiety symptoms.

“Anxiety symptoms None of the studies reported anxiety symptoms”

Across 2 studies, physical health quality of life had MD 0.22 lower (95% CI 2.16 lower to 1.73 higher) for discontinuation versus continuation.

“Mean QoL physical health total endpoint score ranged from 14.9 to 22.93 (2 studies) MD 0.22 lower (2.16 lower to 1.73 higher)”

In one comparison, the proportions with withdrawal syndrome based on DESS scores did not differ between abrupt discontinuation (31/146; 21.2%) and continuation (4/36; 11.1%) (P >= 0.06).

“There were no differences in the proportions of participants with withdrawal syndrome based on DESS scores between the abrupt discontinuation group (31/146; 21.2%) and the continuation group (4/36; 11.1%) (P ≥ 0.06).”

• AssumptionRests on an assumption: The DESS scale validly measures antidepressant withdrawal syndrome.

The review included 33 studies reported in 64 publications, comprising 4995 participants.

“We included 33 studies described in 64 publications involving 4995 participants (see Characteristics of included studies).”

In all, 13 studies compared abrupt discontinuation to continuation of long-term antidepressant treatment.

“In all, 13 studies compared abrupt discontinuation to continuation of long-term antidepressant treatment.”

Four studies compared discontinuation by tapering to continuation of antidepressant with a high-intensity psychological intervention in both treatment groups.

“Four studies compared discontinuation by”

No studies used low-intensity psychological intervention or psychosocial support to support discontinuation.

“No studies used low-intensity psychological intervention or psychosocial support to support discontinuation.”

Eight studies with no pre-randomisation phase included participants with long-term antidepressant treatment.

“Eight studies (with no pre-randomisation phase) included participants with long-term antidepressant treatment”

The definition of response or remission varied widely between studies.

“The definition of response/remission varied widely between studies”

In the multi-phase RCTs, antidepressant duration before randomisation varied widely, at 24 weeks, 26 weeks, 28 weeks, 30 weeks, 32 weeks, 34 weeks, 36 weeks, 48 weeks, 62 weeks, 76 weeks, and 86 weeks.

“In the multi-phase RCTs, antidepressant duration before randomisation varied widely, at 24 weeks (Kane 1982; Khan 2014; Klysner 2002; Kornstein 2006; Mavissakalian 1999; Montgomery 1988; Montgomery 2004; Rickels 2010; Rouillon 2000; Terra 1998; Wilson 2003), 26 weeks (Kocsis 1996), 28 weeks (Gelenberg 2003; Keller 1998; Perahia 2009; Segal 2010), 30 weeks (Stewart 1997), 32 weeks (Gilaberte 2001),”

In prevention relapse discontinuation trials, only a subgroup of antidepressant users who responded to antidepressant treatment were randomised.

“In prevention relapse discontinuation trials, only a subgroup of antidepressant users who responded to antidepressant treatment were randomised.”

We judged all studies to be at high risk of bias in at least one domain.

“We judged all studies to be at high risk of bias in at least one domain.”

Most studies excluded participants at high risk for suicide, drug abuse, anxiety disorders, and other Axis I or Axis II disorders.

“Most studies excluded participants at high risk for suicide, drug abuse, anxiety disorders, and other Axis I or Axis II disorders (see Characteristics of included studies).”

No included studies used a changed pharmacological form or additional pharmacological support.

“No included studies used a changed pharmacological form or additional pharmacological support.”

None of the studies made adjustments for withdrawal symptoms.

“None of the studies made adjustments for withdrawal symptoms.”

Thirteen studies (N = 1780) compared abrupt discontinuation with continuation of antidepressant.

“Thirteen studies (N = 1780) compared abrupt discontinuation with continuation of antidepressant (Gelenberg 2003; Gilaberte 2001 Kane 1982; Khan 2014; Klysner 2002; Kornstein 2006; Montgomery 1988 Peterson 2010; Rapaport 2001 Rouillon 2000; Streim 2012; Terra 1998; Wilson 2003);”

Risk of allocation concealment bias was unclear in 22 studies as the method of concealment is not described or is not described in sufficient detail.

“Risk of allocation concealment bias was unclear in 22 studies as the method of concealment is not described or is not described in sufficient detail to allow a definitive judgement (Bialos 1982;Cook 1986;Derubeis 2019;Gelenberg 2003;Gilaberte 2001;Kane 1982;Keller 1998;Khan 2014;Klysner 2002;Kocsis 1996;Kornstein 2006;Kupfer 1992;Montgomery 1988;Montgomery 2004;Perahia 2009;Peterson 2010;Rapaport”

Two of the 13 studies included psychological support as a co-intervention (CBT in Peterson 2010; CBASP in Gelenberg 2003).

“2 of the 13 studies included psychological support as a co-intervention (CBT in Peterson 2010; CBASP in Gelenberg 2003).”

• cites Peterson 2010Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Gelenberg 2003 sourceNot checked — no full text of the source

We assessed eight studies including psychotherapy or minimal intervention as having high risk of performance bias.

“We assessed eight studies including psychotherapy or minimal intervention as having high risk of performance bias (Bockting 2018;Derubeis 2019;Eveleigh 2018;Huijbers 2016;Kuyken 2008;Kuyken 2015;Segal 2010;Streim 2012).”

Allocation concealment was described in sufficient detail to assess risk of bias as low in only 11 studies.

“Allocation concealment was described in sufficient detail to assess risk of bias as low in only 11 studies (Bockting 2018;Eveleigh 2018;Huijbers 2016;Keller 2007;Kocsis 2007;Kuyken 2008;Kuyken 2015;Mavissakalian 1999;Mavissakalian 2001;Segal 2010;Wilson 2003).”

In all, 32 studies (97%) failed to include withdrawal symptoms.

“In all, 32 studies (97%) failed to include withdrawal symptoms, although this is a fundamental outcome in drug discontinuation trials.”

We judged the only cluster-RCT -Eveleigh 2018 -to be at low risk for cluster bias.

“We judged the only cluster-RCT -Eveleigh 2018 -to be at low risk for cluster bias (Higgins 2017).”

• cites Higgins 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

We judged eight studies to be at low risk of attrition bias because there were no missing data, all study dropouts were reported, or the number of dropout participants was low (< 20%) and was similar across treatment groups.

“We judged eight studies to be at low risk of attrition bias because there were no missing data, all study dropouts were reported, or the number of dropout participants was low (< 20%) and was similar across treatment groups (Bockting 2018;Cook 1986;Khan 2014;Kocsis 1996;Kupfer 1992;Kuyken 2015;Peterson 2010;Wilson 2003).”

None of the included studies compared discontinuation combined with low-intensity psychological intervention or psychosocial support or discontinuation with changed pharmacological form to continuation of long-term antidepressants.

“None of the included studies compared discontinuation combined with low-intensity psychological intervention or psychosocial support or discontinuation with changed pharmacological form (e.g.fluoxetine or paroxetine liquid) to continuation of long-term antidepressants.”

Data on suicide, suicide attempt, or suicidal ideation were not reported in 25 (76%) trials.

“Data on suicide, suicide attempt, or suicidal ideation were not reported in 25 (76%) trials.”

None of the studies reported successful discontinuation rate as an outcome.

“None of the studies reported successful discontinuation rate as an outcome.”

We judged all studies except one to be at high risk of reporting bias (Khan 2014).

“While following recommendations provided in the Cochrane Handbook for Systematic Reviews of Interventions, and in considering all relevant outcomes for a study-level judgement (Higgins 2017), we judged all studies except one to be at high risk of reporting bias (Khan 2014).”

• cites Higgins 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Khan 2014 sourceNot checked — no full text of the source

In the pooled analysis of 13 studies without psychological co-intervention, the discontinuation group had a higher risk of relapse than the continuation group (HR 2.97, 95% CI 2.24 to 3.93; 1546 participants).

“Participants discontinuing antidepressants (n = 776) had higher risk of relapse (as defined by study authors) compared to those in the continuation group (n = 770) (Hazard Ratio 2.97, 95% CI 2.24 to 3.93; participants = 1546; studies = 13) without specific assessment of withdrawal symptoms.”

Discontinuation group participants had a higher CGI-S severity score than continuation group participants (MD 0.31, 95% CI 0.13 to 0.49; 714 participants, 5 studies; I2 = 65%), favouring antidepressant continuation.

“Participants in the discontinuation group had higher severity score rated on the CGI-S compared to those in the continuation group (MD 0.31, 95% CI 0.13 to 0.49), favouring antidepressant continuation (714 participants, 5 studies; I = 65%; random-effects model).”

There was no difference in time to relapse between discontinuation and continuation when psychotherapy was co-intervention (HR 1.48, 95% CI 0.93 to 2.34; 182 participants, 2 studies).

“There was no difference in time to relapse between discontinuation versus continuation with psychotherapy as co-intervention in both groups (HR 1.48, 95% CI 0.93 to 2.34; 182 participants, 2 studies).”

Terra 1998 reported two participants in the antidepressant group (n=110) withdrew due to adverse events, and zero in the discontinuation group (n=94).

“Terra 1998 (n = 204) reported that two participants in the antidepressant group (n = 110) withdrew during the trial due to adverse events, along with zero participants in the discontinuation group (n = 94).”

• cites Terra 1998 sourceNot checked — no full text of the source

Study authors reported no statistically significant difference between groups in proportions with withdrawal syndrome based on DESS scores (21.2% vs 11.1%; P ≥ 0.06).

“Study authors reported there were no statistically significant differences between groups in the proportions of participants with withdrawal syndrome based on DESS scores (abrupt discontinuation 21.2%; continuation 11.1%; P ≥ 0.06).”

• cites Khan 2014 sourceNot checked — no full text of the source

Funnel plots for the 13-study analysis showed low risk of publication bias.

“These funnel plot showed low risk of publication bias.”

Funnel plots for the 16-study analysis showed low risk of publication bias.

“These funnel plots showed low risk of publication bias.”

Relapse was based on criteria defined by study authors, which varied widely between studies.

“Relapse was based on the criteria defined by study authors, which varied widely between studies.”

Most studies used clinical scale scores covering domains such as depressed mood, insomnia, agitation, gastrointestinal somatic symptoms, anxiety, general somatic symptoms, and sexual dysfunction, all of which are withdrawal symptoms.

“Most studies were based on clinical scale scores that include domains for depressed mood, insomnia, agitation, gastrointestinal somatic symptoms, anxiety, general somatic symptoms, and sexual dysfunction -all of which are symptoms of withdrawal.”

• AssumptionRests on an assumption: Clinical scale scores include domains that are symptoms of withdrawal.

In the pooled analysis of 4 studies with psychological co-intervention, the discontinuation group had a higher risk of relapse than the continuation group (HR 1.90, 95% CI 1.42 to 2.53; 570 participants).

“We found higher risk of relapse (as defined by study authors) in the discontinuation group compared to continuation of antidepressant with psychological treatment as co-intervention (HR 1.90, 95% CI 1.42 to 2.53; 570 participants, 4 studies).”

In the pooled analysis of 16 studies, participants who discontinued antidepressants had a higher risk of relapse than those who continued, with a hazard ratio of 2.59 (95% CI 2.07 to 3.25) among 2120 participants.

“Participants discontinuing antidepressants had higher risk of relapse (as defined by study authors) compared to those in the continuation group (Hazard Ratio 2.59, 95% CI 2.07 to 3.25; participants = 2120; studies = 16) without specific assessment of withdrawal symptoms.”

In Khan 2014, 31 of 146 participants in the venlafaxine abrupt discontinuation group experienced a withdrawal syndrome, compared with 4 of 36 in the continuation group.

“In all, 31 of the 146 participants in the discontinuation group experienced a withdrawal syndrome, along with 4 of the 36 participants in the continuation group.”

• cites Khan 2014 sourceNot checked — no full text of the source

In one study of participants with anxiety disorder, there was no difference in depression as an adverse event between discontinuation and continuation groups (9/89 vs 9/92; P = 1).

“Study authors reported no difference in depression between the discontinuation group (9/89; 10.1%) and the continuation group (9/92; 9.8%) (P = 1) (Rapaport 2001).”

• cites Rapaport 2001 sourceNot checked — no full text of the source

The most common adverse events (≥10%) among participants who continued their antidepressant included headache, asthenia, pain, flu-like symptoms, infection, back pain, nausea, traumatic injury, dyspepsia, somnolence and insomnia, upper airway infection, anxiety, decreased libido, nausea, abdominal pain, weight loss, psychiatric disorder, dizziness, and fatigue.

“The most common adverse events amongst participants who continued their antidepressant included headache, asthenia, pain, flu-like symptoms, infection, back pain, nausea, traumatic injury, dyspepsia, somnolence and insomnia, upper airway infection, anxiety, decreased libido, nausea, abdominal pain, weight loss, psychiatric disorder, dizziness, and fatigue.”

No study measured quality of life.

“No study measured quality of life.”

The most common adverse events (≥10%) among antidepressant discontinuers included headache, asthenia, malaise, depression, pain, abdominal pain, flu-like symptoms, infection, back pain, dyspepsia, insomnia, upper airway infection, anxiety, libido decreased, nausea, dysmenorrhoea (only in women), psychiatric disorder, dizziness, and malaise.

“The most common adverse events (≥ 10%) among antidepressant discontinuers included headache, asthenia, malaise, depression, pain, abdominal pain, flu-like symptoms, infection, back pain, dyspepsia, insomnia, upper airway infection, anxiety, libido decreased, nausea, dysmenorrhoea (only in women), psychiatric disorder, dizziness, and malaise.”

No study measured social and occupational functioning.

“No study measured social and occupational functioning.”

No studies without psychological support reported a successful discontinuation rate.

“No studies reported a successful discontinuation rate.”

Rouillon 2000 reported two participants with two suicide attempts in each group.

“Rouillon 2000 (n = 214) reported two participants with two suicide attempts in each group.”

• cites Rouillon 2000Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In Khan 2014, there was no difference in proportions of participants with withdrawal syndrome based on DESS scores between discontinuation (30/139; 21.6%) and continuation (4/36; 11.1%) groups (P ≥ 0.06).

“Data show no differences in the proportions of participants with withdrawal syndrome based on DESS scores between the discontinuation group (30/139; 21.6%) and the continuation group (4/36; 11.1%) (P ≥ 0.06).”

• cites Khan 2014 sourceNot checked — no full text of the source

No difference in severity of depressive symptoms measured by HAM-D between discontinuation and continuation groups (MD 0.44, 95% CI -1.12 to 2.00; 330 participants, 3 studies; I2 = 85%).

“No difference in the severity of depressive symptoms was measured by the HAM-D scale between participants in the discontinuation group and those in the continuation group (mean difference (MD) 0.44, 95% CI -1.12 to 2.00; 330 participants, 3 studies; I = 85%; random-effects model).”

Terra 1998 found higher endpoint COVI Anxiety Scale score in discontinuation group (5.8, SD 3.0, n=94) than continuation group (4.7, SD 2.4, n=109; P = 0.003), favouring continuation.

“Participants in the discontinuation group (n = 94) had a higher endpoint score (5.8 (SD 3.0)) on the COVI Anxiety Scale than participants in the continuation group (n = 109) (4.7 (SD 2.4); P = 0.003), favouring continuation of long-term antidepressants.”

• cites Terra 1998 sourceNot checked — no full text of the source

In all 11 pooled studies, abrupt discontinuation increased the risk of relapse compared with continuation (HR 1.97, 95% CI 1.56 to 2.50; 1555 participants).

“Abrupt discontinuation increased the risk of relapse (as defined by study authors) (HR 1.97, 95% confidence interval (CI) 1.56 to 2.50; 1555 participants, 11 studies) compared to discontinuation, without specific assessment of withdrawal symptoms (Analysis 1.1).”

Peterson 2010 found no significant differences in change scores on BAI nor SQ Anxiety between groups.

“Study authors concluded that there were no significant differences in change scores on BAI nor on SQ Anxiety between groups (BAI -0.31 (SD 0.15) discontinuation (n = 16) and 0.4 (SD 0.64) antidepressant (n = 14) in treatment arms without CBT, and 0.66 (SD 2.04) discontinuation (n = 11) and -0.34 (SD 0.65) antidepressant (n = 11) in treatment arms with CBT; and SQ Anxiety 1.53 (SD 1.10) discontinu”

• cites Peterson 2010Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In Khan 2014, the DESS item most commonly reported as severe was trouble sleeping/insomnia (2.2% discontinuation, 4.2% continuation).

“The DESS item most commonly reported as severe was trouble sleeping/insomnia (2.2% of participants in the discontinuation group and 4.2% in the continuation group).”

• cites Khan 2014 sourceNot checked — no full text of the source

Kocsis 2007 reported one participant in the continuation group with suicidal thoughts, considered not related to treatment.

“Kocsis 2007 (n = 336) reported that one participant in the continuation group experienced suicidal thoughts; study authors considered this not related to treatment.”

• cites Kocsis 2007 sourceNot checked — no full text of the source

In the pooled analysis of 3 studies, the discontinuation group had lower QoL social functioning subscale scores than the antidepressant group (MD -6.44, 95% CI -12.10 to -0.77; 502 participants).

“Participants in the discontinuation group reported lower scores on the QoL social functioning subscale (MD -6.44, 95% CI -12.10 to -0.77; 502 participants, 3 studies; I = 43%; a decrease in score indicates worsening, favouring continuation of antidepressants)”

Sensitivity analysis excluding three studies that measured only serious adverse events did not substantially alter the adverse events meta-analysis result (OR 1.00, 95% CI 0.73 to 1.37).

“Sensitivity analysis excluding these three studies did not substantially alter the metaanalysis result (OR 1.00, 95% CI 0.73 to 1.37).”

Kocsis 2007 reported no suicide attempts during the trial.

“No suicide attempts during the trial were reported.”

• cites Kocsis 2007 sourceNot checked — no full text of the source

In Perahia 2009, there was no difference in the SF-36 physical component summary score (0.33 (SE 0.76) discontinuation vs -0.45 (SE 0.7) continuation; P=0.415).

“and no difference in the SF-36 physical component summary score: 0.33 (SE 0.76) in the discontinuation group and -0.45 (SE 0.7) in the continuation group (P = 0.415).”

• cites Perahia 2009 sourceNot checked — no full text of the source

Keller 2007 reported one participant (1/40) in the discontinuation group with suicidal ideation, considered possibly related to treatment.

“Keller 2007 (n = 83) reported that one (1/40) participant in the discontinuation group experienced suicidal ideation; study authors considered this possibly related to study treatment.”

• cites Keller 2007 sourceNot checked — no full text of the source

In Kuyken 2008, 75% (46/61) of participants in the MBCT arm discontinued their medication at the end of the six-month tapering window.

“Kuyken 2008 reported that at the end of the six-month window for tapering, 46 of 61 (75%) participants in the MBCT arm of the trial discontinued their medication.”

• cites Kuyken 2008 sourceNot checked — no full text of the source

Kocsis 2007 reported one participant in the discontinuation group with suicidal ideation.

“Suicidal ideations were also reported by one participant in the discontinuation group.”

• cites Kocsis 2007 sourceNot checked — no full text of the source

In Perahia 2009, the discontinuation group reported a higher score for Sheehan Disability Scale global functioning compared to the antidepressant group (mean change 2.06 (SE 0.77) vs -0.05 (SE 0.71); P=0.029).

“Participants in the discontinuation group (n = 142) reported a higher score for Sheehan Disability Scale global functioning compared to those in the antidepressant group (n = 145) (mean change (SE) 2.06 (0.77) in the discontinuation group and -0.05 (0.71) in the antidepressant group; P = 0.029; increase in score signifies worsening).”

• cites Perahia 2009 sourceNot checked — no full text of the source

Keller 2007 reported no other suicide-related events in either group.

“No other suicide-related events in the continuation group or in the discontinuation group were reported.”

• cites Keller 2007 sourceNot checked — no full text of the source

In Perahia 2009, there was a difference in the SF-36 mental component summary mean change score favouring continuation (-5.74 (SE 1.2) discontinuation vs -1.11 (SE 1.11) continuation; P=0.002).

“Perahia 2009 reported a difference in the SF-36 mental component summary mean change score, favouring continuation of antidepressant: -5.74 (SE 1.2) in the discontinuation group (n = 142) and -1.11 (SE 1.11) in the continuation group (n = 146) (P = 0.002)”

• cites Perahia 2009 sourceNot checked — no full text of the source

In the pooled analysis of 6 studies, the discontinuation group had significantly higher HAM-D scores than the antidepressant group (MD 3.50, 95% CI 2.31 to 4.68; 1017 participants).

“In the pooled discontinuation group, participants reported significantly higher scores on the HAM-D scale compared to those in the antidepressant group (MD 3.50, 95% CI 2.31 to 4.68; 1017 participants, 6 studies; I = 29%).”

In the pooled analysis of 6 studies, the discontinuation group had higher CGI-S scores compared to the antidepressant group, favouring antidepressant (MD 0.61, 95% CI 0.44 to 0.79; 1187 participants).

“In the pooled discontinuation group, participants had higher scores on CGI-S compared to those in the pooled antidepressant group, favouring the antidepressant group (lower score indicates less impairment) (MD 0.61, 95% CI 0.44 to 0.79; 1187 participants, 6 studies; I = 29%).”

Removing Derubeis 2019 (where the antidepressant was not replaced by placebo) from the adverse events meta-analysis did not change overall results (OR 1.04, 95% CI 0.77 to 1.41).

“When this study was removed from the meta-analysis, the overall results did not change (OR 1.04, 95% CI 0.77 to 1.41).”

• cites Derubeis 2019Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In the pooled analysis of 3 studies, the discontinuation group had lower QoL emotional functioning subscale scores than the antidepressant group (MD -18.81, 95% CI -26.66 to -10.97; 502 participants).

“and on the QoL emotional functioning subscale (MD -18.81, 95% CI -26.66 to -10.97; 502 participants, 3 studies; I = 0%; lower scores indicate impairment of functioning, favouring continuation of antidepressant) compared to participants in the antidepressant group.”

In Kuyken 2008, ten (16%) participants in the continuation group withdrew from the trial due to discontinuation of antidepressants.

“Ten (16%) participants in the continuation group withdrew from the trial due to discontinuation of antidepressants.”

• cites Kuyken 2008 sourceNot checked — no full text of the source

In the pooled analysis of 3 studies, the discontinuation group had higher HAM-A scores than the antidepressant group (MD 3.53, 95% CI 1.92 to 5.14; 526 participants).

“Participants in the pooled discontinuation group (n = 257) showed higher scores on the HAM-A scale compared to those in the pooled antidepressant group (n = 269) (MD 3.53, 95% CI 1.92 to 5.14; 526 participants, 3 studies; I = 53%; higher score signifies more anxiety symptoms).”

In the pooled analysis of 7 studies, the difference in adverse events (including withdrawal symptoms) between discontinuation and continuation groups was not significant (OR 1.06, 95% CI 0.82 to 1.38; 1479 participants).

“The difference between discontinuation and continuation groups was not significant (odds ratio (OR) 1.06, 95% CI 0.82 to 1.38; 1479 participants, 7 studies; I = 0%).”

Perahia 2009 reported no suicide attempts.

“Perahia 2009 (n = 288) reported no suicide attempts.”

• cites Perahia 2009 sourceNot checked — no full text of the source

In the pooled analysis of 3 studies, the discontinuation group had higher total scores on the SAS-SR compared to the antidepressant group, favouring continuation (MD 0.19, 95% CI 0.11 to 0.28; 502 participants).

“In the pooled discontinuation group, participants had higher total scores on the SAS-SR compared to participants in the pooled antidepressant group, favouring continuation of antidepressants (MD 0.19, 95% CI 0.11 to 0.28; 502 participants, 3 studies; I = 0%; lower scores indicate less impairment).”

In the pooled analysis of 3 studies, there were no differences between groups on the QoL physical health problems subscale (MD -2.08, 95% CI -5.66 to 1.49; 502 participants).

“There were no differences between groups on the subscale quality of life (QoL) physical health problems (MD -2.08, 95% CI -5.66 to 1.49; 502 participants, 3 studies; I = 0%).”

In the tapering advice group, 4 of 70 (6%; 95% CI 2 to 14) participants successfully stopped their antidepressant compared to 6 of 76 (8%; 95% CI 4 to 16) participants in the usual care group who spontaneously discontinued after one year.

“In the tapering advice group, 4 of 70 (6%; 95% CI 2 to 14) participants successfully stopped their antidepressant compared to 6 of 76 (8%; 95% CI 4 to 16) participants in the usual care group who spontaneously discontinued a er one year (Eveleigh 2018).”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

No studies measured severity of illness.

“No studies measured this outcome.”

No studies measured anxiety symptoms.

“No studies measured this outcome.”

In the tapering advice group, 18 of 70 (26%) participants relapsed compared to 10 of 76 (13%) in the usual care group.

“In the tapering advice group, 18 of 70 (26%) participants relapsed compared to 10 of 76 (13%) in the usual care group (Eveleigh 2018).”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

There was no difference in quality of life between discontinuation with tailored recommendation and usual care (MD -0.02, 95% CI -0.05 to 0.10).

“There was no difference in quality of life between discontinuation with tailored recommendation and usual care (MD (with multiple imputation for missing values) -0.02, 95% CI -0.05 to 0.10; higher scores indicates better quality of life).”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

There was no statistically significant difference in mean CESD score between the tapering group (n = 51; mean 13.7, SD 8.9) and the usual care group (n = 55; mean 12.6, SD 7.9) at the end of the trial (P = 0.51).

“Study authors reported a higher mean CESD score in the tapering group (n = 51) (mean CESD total endpoint score 13.7 (SD 8.9)) compared to the usual care group (n = 55) (mean CESD total endpoint score 12.6 (SD 7.9)) at the end of the trial but no differences between groups (P = 0.51)”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The estimated effect for depressive disorder was HR 2.62 (95% CI 2.01 to 3.42).

“and for depressive disorder HR 2.62 (95% CI 2.01 to 3.42).”

Participants in the tapering advice group (n = 70) had a mean of 0.70 QALYs (SD 0.25), and in the usual care group (n = 76) a mean of 0.72 QALYs (SD 0.26).

“Participants in the tapering advice (n = 70) group had a mean of 0.70 QALYs (SD 0.25), and in the usual care group (n = 76) a mean of 0.72 QALYs (SD 0.26).”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Of the 36 (51%) participants who agreed to try, only 4 (6%) successfully stopped the antidepressant.

“Of the 36 (51%) participants who agreed to try, only 4 (6%) successfully stopped the antidepressant.”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Of participants who continued their antidepressant, there was a non-statistically significant higher relapse rate in the tapering advice group versus the usual care group (25% versus 11%; P = 0.07).

“Of participants who continued their antidepressant, there was a nonstatistically significant higher relapse rate in the tapering advice group versus the usual care group (25% versus 11%; P = 0.07).”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Of the 70 participants in the tapering group, 34 (49%) did not comply with the advice to stop their antidepressant medication.

“They also reported that of the 70 participants in the tapering group, 34 (49%) did not comply with the advice to stop their antidepressant medication.”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

No studies assessed withdrawal symptoms.

“No studies assessed this outcome.”

There was no statistically significant difference in mean PAS score between the tapering group (n = 50; mean 4.1, SD 7.2) and the usual care group (n = 51; mean 3.6, SD 7.1) at the end of the study (P = 0.71).

“Study authors reported a higher mean PAS score in the tapering group (n = 50) (mean PAS 4.1 (SD 7.2)) compared to the usual care group (n = 51) (mean PAS 3.6 (SD 7.1)) at the end of the study but no differences between groups (P = 0.71)”

• cites Eveleigh 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Kuyken 2008 (n = 123) reported that no serious adverse events were recorded through the oversight of the Trial Steering Committee.

“Kuyken 2008 (n = 123) reported that no serious adverse events were recorded through the oversight of the Trial Steering Committee.”

• cites Kuyken 2008 sourceNot checked — no full text of the source

For tapering versus continuation, the test for subgroup differences for relapse was significant only for the indication of antidepressant (P = 0.02; I2 = 81.7%).

“For the outcome relapse, the test for subgroup differences was significant only for the indication of antidepressant (P = 0.02; I = 81.7%).”

The estimated effect for anxiety disorder was HR 7.06 (95% CI 3.22 to 15.52).

“The estimated effect for anxiety disorder was HR 7.06 (95% CI 3.22 to 15.52)”

Tapering may increase anxiety symptoms compared with continuation, based on three studies with very low-certainty evidence.

“but may increase severity of depressive symptoms (6 studies) and anxiety symptoms (3 studies).”

The review included 33 randomised controlled trials with 4995 participants prescribed antidepressants long term (24 weeks or longer).

“In this review, we included 33 randomised controlled trials (RCTs) with 4995 participants prescribed antidepressants long term (24 weeks or longer).”

Thirteen included studies, or 39%, discontinued antidepressants abruptly.

“Thirteen (39%) studies discontinued the antidepressant abruptly.”

Discontinuation rates in the discontinuation group varied widely: 40%, 59%, and 75%, based on three studies with very low-certainty evidence.

“Discontinuation rates in the discontinuation group varied widely: 40%, 59%, and 75% (3 studies; very low-certainty evidence).”

The predefined subgroup analysis based on age did not show a difference in relapse.

“The predefined subgroup analysis based on age did not show a difference in relapse.”

Evidence is insufficient on the effect of tapering on the quality of life subdomain physical health problems, based on three studies with very low-certainty evidence.

“There is insufficient evidence of its effect on the quality of life subdomain physical health problems (3 studies; very lowcertainty evidence)”

In one study, 55% of the discontinuation group and 12% of the continuation group discontinued after 15 months.

“Discontinuation rates in the discontinuation group varied: 55% in the discontinuation group and 12% in the continuation group a er 15 months in one study; and 40% in the discontinuation group a er 6 months in the other study (2 studies; very low-certainty evidence).”

Withdrawal symptoms were measured in only one included study.

“Withdrawal symptoms were measured in only one study.”

Tapering may have little or no effect on adverse events compared with continuation, based on seven studies with very low-certainty evidence.

“or adverse events (7 studies)”

Tapering may increase severity of depressive symptoms compared with continuation, based on six studies with very low-certainty evidence.

“but may increase severity of depressive symptoms (6 studies) and anxiety symptoms (3 studies).”

Compared with continuation, abrupt discontinuation may increase severity of illness, based on five studies with very low-certainty evidence.

“Compared to continuation of antidepressants, abrupt discontinuation may increase severity of illness (5 studies; very low-certainty evidence).”

Tapering with PCT and MBCT may have little or no effect on the quality of life subdomain psychological functioning, based on two studies with very low-certainty evidence.

“and may have little or no effect on the quality of life subdomains psychological or physical or social functioning (2 studies; very low-certainty evidence) compared to continuation of long-term antidepressants.”

Pooling discontinuation rate was not possible because of insufficient reporting of data.

“Again, pooling was not possible for discontinuation rate due to insufficient reporting of data.”

One subgroup analysis suggested that discontinuing antidepressants prescribed for anxiety disorder leads to more relapse than when prescribed for depressive disorders.

“One subgroup analysis suggested that discontinuation of antidepressants prescribed for anxiety disorder leads to more relapse than when prescribed for depressive disorders.”

Evidence is insufficient on the effect of abrupt discontinuation with psychological support on relapse compared with continuation, based on two studies with very low-certainty evidence.

“There is insufficient evidence of the effect of abrupt discontinuation with psychological support as a co-intervention on relapse compared to continuation of antidepressants (2 studies; very lowcertainty evidence).”

In another study, 40% of the discontinuation group discontinued after 6 months.

“Discontinuation rates in the discontinuation group varied: 55% in the discontinuation group and 12% in the continuation group a er 15 months in one study; and 40% in the discontinuation group a er 6 months in the other study (2 studies; very low-certainty evidence).”

Low study numbers prevented sensitivity analysis for tapering regimens longer than four weeks.

“Low study numbers prohibited sensitivity analysis based on tapering regimens longer than four weeks.”

Eligible studies in the high-intensity psychological intervention comparison did not report withdrawal symptoms, anxiety symptoms, social and occupational functioning, or severity of illness.

“None of the eligible studies in this comparison category reported withdrawal symptoms, anxiety symptoms, social and occupational functioning, or severity of illness.”

Pooling discontinuation outcome data was not possible for the two studies of tapering with psychological support as co-intervention.

“Insufficient data pooling of results for the discontinuation outcome was not possible for the two studies in which”

Discontinuation with a letter to the GP may make little or no difference in relapse, based on one study with low-certainty evidence.

“Discontinuation with a letter to the GP may make little or no difference in relapse (1 study; low-certainty evidence), depressive symptoms (1 study; lowcertainty evidence), and anxiety (1 study; low-certainty evidence)”

Subgroup analyses by antidepressant type and duration did not show a difference.

“Subgroup analyses based on type of antidepressant and duration of antidepressant did not show a difference.”

Evidence on the effects of abrupt discontinuation on withdrawal symptoms is very uncertain, based on one study.

“Evidence about the effects of discontinuation on withdrawal symptoms (1 study) is very uncertain.”

Discontinuation with a letter to the GP may make little or no difference in depressive symptoms, based on one study with low-certainty evidence.

“Discontinuation with a letter to the GP may make little or no difference in relapse (1 study; low-certainty evidence), depressive symptoms (1 study; lowcertainty evidence), and anxiety (1 study; low-certainty evidence)”

Discontinuation with a letter to the GP may make little or no difference in anxiety, based on one study with low-certainty evidence.

“Discontinuation with a letter to the GP may make little or no difference in relapse (1 study; low-certainty evidence), depressive symptoms (1 study; lowcertainty evidence), and anxiety (1 study; low-certainty evidence)”

Discontinuation with a letter to the GP may have little or no effect on quality of life, based on one study with very low-certainty evidence.

“and may have little or no effect on quality of life (1 study; very lowcertainty evidence) compared to usual care.”

Discontinuation with a letter to the GP may make little or no difference in discontinuation rate, based on one study with low-certainty evidence.

“Discontinuation with a letter to the GP may make little or no difference in the discontinuation rate (1 study; low-certainty evidence).”

Tapering with PCT and MBCT may have little or no effect on the quality of life subdomain physical functioning, based on two studies with very low-certainty evidence.

“and may have little or no effect on the quality of life subdomains psychological or physical or social functioning (2 studies; very low-certainty evidence) compared to continuation of long-term antidepressants.”

Pooling adverse events was not possible because of insufficient information, based on three studies with very low-certainty evidence.

“Pooling was not possible for adverse events due to insufficient information (3 studies; very low-certainty evidence).”

No included studies made adjustments for withdrawal symptoms.

“Nor did any of the included studies make adjustments for withdrawal symptoms.”

Tapering with PCT and MBCT may have little or no effect on the quality of life subdomain social functioning, based on two studies with very low-certainty evidence.

“and may have little or no effect on the quality of life subdomains psychological or physical or social functioning (2 studies; very low-certainty evidence) compared to continuation of long-term antidepressants.”

Most tapering regimens in the included studies lasted four weeks or less.

“Most tapering regimens lasted four weeks or less.”

• Also stated in: § Overall completeness and applicability of evidence · plain

  “Nearly all tapering studies used a tapering schedule of four weeks or less.”

The four studies including people prescribed antidepressants for anxiety differed in study characteristics, type of anxiety disorder, and intervention.

“The four studies that included people prescribed antidepressants for anxiety differed in terms of study characteristics, type of anxiety disorder, and intervention.”

Twenty-five of the 33 included studies focused on participants with recurrent depression.

“Twenty-five included of the 33 studies focused on participants with recurrent depression.”

Many study authors used one continuous scale measuring depressive symptoms to assess relapse, then dichotomised the outcome as a threshold value.

“Many study authors used one continuous scale measuring depressive symptoms (e.g.Hamilton Depression Rating Scale (HDRS), Montgomery-Åsberg Depression Rating Scale (MADRS)) to assess relapse, then dichotomised the outcome as a threshold value (e.g.HAM-D score > 17).”

Few studies focused on de-prescribing long-term antidepressants.

“Few studies had a focus on de-prescribing longterm antidepressants.”

Most studies in the review included participants in remission from severe depression with a history of more than two previous episodes.

“Most of the studies in our review included participants in remission from severe depression with a history of more than two previous episodes.”

Only two studies provided sufficient information to assess successful discontinuation rate.

“Only two studies provided sufficient information to assess successful discontinuation rate”

Limited information on most primary and secondary outcomes restricted investigation of the benefits and harms of discontinuing long-term antidepressant use.

“Limited information on the majority of our primary and secondary outcomes restricted investigation of the benefits and harms of discontinuing long-term antidepressant use.”

Most included studies had a limited duration of antidepressant treatment, between six and twelve months, before discontinuation.

“Furthermore, most studies included in our review had a limited duration of antidepressant treatment (between six and twelve months) before discontinuation.”

The certainty of evidence for primary and secondary outcomes was very low to low due to limitations in study design, indirectness, and imprecision.

“We found the certainty of evidence for primary and secondary outcomes to be very low to low due to limitations in study design, indirectness, and imprecision.”

A minimal intervention such as a letter with tapering advice sent to the GP was investigated in only one trial.

“A minimal intervention such as a letter with tapering advice sent to the GP was investigated in only one trial.”

Some studies included dysthymia, chronic depression, or double depression.

“Some studies included dysthymia, chronic depression, or double depression.”

Overall, data on discontinuation were either not reported or insufficiently reported.

“Overall, data on discontinuation either were not reported or were insufficiently reported.”

Nearly all studies were designed to examine how continuation of long-term antidepressants could avoid relapse and recurrence rather than discontinuation success.

“Nearly all studies were designed to examine how continuation of long-term antidepressants could avoid relapse and recurrence rather than discontinuation success.”

None of the included studies used an active placebo that mimics side effects of antidepressants.

“None of the included studies used an active placebo that mimics side effects of antidepressants (e.g.drugs with anticholinergic actions, such as atropine, mimic TCA side effects).”

This review found no differences between subgroups based on total duration of antidepressant use in the tapering versus continuation comparison.

“we did not find any differences between subgroups on the basis of total duration of antidepressant use in the comparison tapering versus continuation”

Only a few studies reported blinding integrity.

“Only a few studies reported blinding integrity.”

This review has not been able to show that relapse rates in primary care are lower than in secondary care.

“Our review has been unable to demonstrate that the relapse rate in primary care is lower than in secondary care.”

Most studies eligible for this review involved people with recurrent depression.

“In addition, most studies involved people with recurrent depression.”

Relapse rather than discontinuation success was the primary outcome for the vast majority of studies eligible for this review.

“Relapse rather than discontinuation success was the primary outcome for the vast majority of studies eligible for inclusion in this review.”

This review found no difference between subgroups based on total duration of antidepressant use.

“we did not find any difference between subgroups on the basis of total duration of antidepressant use”

There was no statistically significant difference in adverse events between abrupt discontinuation and continuation (OR 1.11, 95% CI 0.62 to 1.99; 7 studies, 1012 participants).

“1.2 Adverse events 7 1012 Odds Ratio (M-H, Random, 95% CI) 1.11 [0.62, 1.99]”

In the subgroup without co-intervention and with follow-up longer than 24 weeks, abrupt discontinuation compared with continuation was associated with a higher risk of relapse (HR 2.09, 95% CI 1.59 to 2.74; 10 studies, 1373 participants).

“1.1.2No co-intervention follow-up > 24 weeks 10 1373 Hazard Ratio (IV, Random, 95% CI) 2.09 [1.59, 2.74]”

In adults with depressive or anxiety disorders taking long-term antidepressants, abrupt discontinuation compared with continuation was associated with a higher risk of relapse (HR 1.97, 95% CI 1.56 to 2.50; 11 studies, 1555 participants).

“1.1.1All studies 11 1555 Hazard Ratio (IV, Random, 95% CI) 1.97 [1.56, 2.50]”

In the subgroup with co-intervention and follow-up longer than 24 weeks, the risk of relapse with abrupt discontinuation compared with continuation was not statistically significant (HR 1.48, 95% CI 0.93 to 2.34; 2 studies, 182 participants).

“1.1.3With co-intervention follow-up > 24 weeks 2 182 Hazard Ratio (IV, Random, 95% CI) 1.48 [0.93, 2.34]”

This review identified one RCT that compared tapering with CBT to tapering without CBT for people with a remitted anxiety disorder.

“We identified one RCT that compared tapering with CBT to tapering without CBT for people with a remitted anxiety disorder.”

For CGI-S total score, abrupt discontinuation was associated with greater severity of illness than continuation (MD 0.60, 95% CI 0.27 to 0.93; 2 studies, 343 participants).

“1.4.1 CGI-S total score 2 343 Mean Difference (IV, Random, 95% CI) 0.60 [0.27, 0.93]”

For HAM-D mean change, there was no statistically significant difference between abrupt discontinuation and continuation (MD -0.11, 95% CI -1.54 to 1.32; 2 studies, 190 participants).

“1.3.2HAM-D mean change 2 190 Mean Difference (IV, Random, 95% CI) -0.11 [-1.54, 1.32]”

Abrupt discontinuation was associated with greater severity of illness than continuation (MD 0.31, 95% CI 0.13 to 0.49; 5 studies, 714 participants).

“1.4 Severity of illness 5 714 Mean Difference (IV, Random, 95% CI) 0.31 [0.13, 0.49]”

For CGI-S mean change, abrupt discontinuation was associated with greater severity of illness than continuation (MD 0.24, 95% CI 0.05 to 0.43; 3 studies, 371 participants).

“1.4.2CGI-S mean change 3 371 Mean Difference (IV, Random, 95% CI) 0.24 [0.05, 0.43]”

In the included studies, relapse was reported as time-to-event data using hazard ratios.

“In the included studies, authors reported relapse as time to events by using a hazard ratio (time-to-event data).”

There was no statistically significant difference in depressive symptoms between abrupt discontinuation and continuation (MD 0.44, 95% CI -1.12 to 2.00; 3 studies, 330 participants).

“1.3 Depressive symptoms 3 330 Mean Difference (IV, Random, 95% CI) 0.44 [-1.12, 2.00]”

Data could not be pooled because insufficient data were reported.

“Data could not be pooled due to insufficient data reported”

Evidence showed no association between relapse rates and duration of treatment.

“Moreover, evidence showed no association between relapse rates and duration of treatment (Geddes 2003)”

• cites Geddes 2003 sourceNot checked — no full text of the source

The included studies did not distinguish relapse from withdrawal symptoms.

“Studies did not distinguish relapse from symptoms of withdrawal”

There was no evidence to justify the clinical distinction between continuation treatment (six to nine months) and maintenance treatment (two years or longer).

“and there was no evidence to justify the defined clinical distinction between continuation treatment (six to nine months) and maintenance treatment (two years or longer; Kaymaz 2008).”

• cites Kaymaz 2008Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Continuation of antidepressant treatment reduces risk of relapse and recurrence by 50% to 70%.

“Evidence suggests that continuation of antidepressant treatment is effective, as it reduces risk of relapse and recurrence by 50% to 70% (Geddes 2003;Glue 2010;Kaymaz 2008)”

• cites Geddes 2003 sourceNot checked — no full text of the source
• cites Glue 2010 sourceNot checked — no full text of the source
• cites Kaymaz 2008Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Cognitive therapy and MBCT combined with antidepressant discontinuation are potential alternatives to antidepressants for preventing relapse or recurrence.

“These findings suggest that cognitive therapy and MBCT in combination with discontinuation of antidepressants are potential alternatives to antidepressants for preventing relapse or recurrence.”

Overall, strong evidence indicating whether long-term antidepressants can be discontinued effectively and safely is lacking.

“Overall, strong evidence indicating whether long-term antidepressants can be discontinued effectively and safely is lacking.”

The effectiveness of interventions aimed at discontinuation of long-term antidepressant use is unknown.

“The effectiveness of interventions aimed at discontinuation of long-term antidepressant use is unknown.”

Most guidelines for optimal duration of continuation and maintenance treatment are based on expert opinion because evidence is lacking.

“Due to lack of evidence for the optimal duration of continuation and maintenance treatment, most guidelines are based on expert opinion.”

We judged not measuring withdrawal symptoms as introducing a large source of bias.

“Therefore we judged not measuring withdrawal symptoms as introducing a large source of bias.”

• AssumptionRests on an assumption: Withdrawal symptoms are a fundamental outcome in drug discontinuation trials.

The evidence for the 16-study relapse outcome was downgraded by two levels for risk of bias due to poor description of randomisation and blinding and withdrawal confounding bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (three studies included participants with panic disorder; the other studies included recurrent depression or chronic depression).”

The evidence for the 13-study relapse outcome was downgraded by two levels for risk of bias due to poor description of randomisation and blinding and withdrawal confounding bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (3 studies included participants with panic disorder; the other studies included participants with recurrent depression or chronic depression).”

The certainty of evidence for the severity of illness outcome was very low, downgraded two levels for risk of bias and one level for indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, attrition bias, and withdrawal confounding bias; CGI-S rating of the clinician may be influenced by withdrawal symptoms, which include psychological symptoms and may worsen participants' condition) and by one level for indirectness (nearly”

• AssumptionRests on an assumption: Withdrawal symptoms including psychological symptoms may impair severity of illness.

The certainty of evidence for successful discontinuation rate was very low, downgraded one level each for imprecision, indirectness, and risk of bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for imprecision (no meta-analysis), by one level for indirectness (studies included participants with recurrent depression in remission), and by one level for risk of bias (incomplete outcome reporting, withdrawal confounding bias).”

The certainty of evidence for the anxiety symptoms outcome was very low, downgraded one level each for imprecision, risk of bias, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for imprecision (no pooling and small number of participants), by one level for risk of bias (confounding withdrawal bias, poor description of randomisation and blinding), and by one level for indirectness (only one of the three studies included participants with anxiety disorder (panic disorder)).”

• AssumptionRests on an assumption: The HAM-A scale contains domain symptoms of psychic anxiety and somatic anxiety that are symptoms of withdrawal.
• AssumptionRests on an assumption: The COVI Anxiety Scale may register anxiety and other symptoms of withdrawal.

The certainty of evidence for the relapse outcome in the 13-study no-co-intervention analysis was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (3 studies included participants with panic disorder; the other studies included participants with recurrent depression or chronic depression).”

None of the 4 co-intervention studies assessed withdrawal symptoms.

“None of these studies assessed withdrawal symptoms.”

The certainty of evidence for the relapse outcome in the 16-study pooled analysis was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (three studies included participants with panic disorder; the other studies included recurrent depression or chronic depression).”

The certainty of evidence for the withdrawal syndrome outcome was very low, downgraded one level each for imprecision, risk of bias, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level due to imprecision (single study included a small number of participants), by one level for risk of bias (randomisation and blinding were poorly described, severity of withdrawal symptoms was not scored in the outcome, and the period for observing withdrawal symptoms was relatively short), and by one level”

Suicidal ideas and attempts are well recognised as withdrawal effects.

“However,”

• cites Valuck 2009 sourceNot checked — no full text of the source

The certainty of evidence for the adverse events outcome was very low, downgraded one level each for risk of bias, imprecision, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for risk of bias (poor description of randomisation and blinding), high risk of attrition bias, and withdrawal confounding bias (withdrawal symptoms may be misdiagnosed as adverse events), by one level for imprecision (wide 95% confidence interval, which includes the null effect of no difference), and by”

The certainty of evidence for the relapse outcome in all studies was very low, downgraded two levels for risk of bias and one level for indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, and withdrawal confounding bias) and by one level for indirectness (majority of participants had recurrent depression, and one study included panic disorder).”

• Also stated in: § See Analysis 1.1. · plain

  “We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, and withdrawal confounding bias) and by one level for indirectness (majority of participants had recurrent depression, and one study included panic disorder).”

None of the 17 studies measuring relapse measured withdrawal symptoms separately or adjusted for withdrawal.

“None of these studies measured withdrawal symptoms separately nor made adjustments for withdrawal.”

The certainty of evidence for the relapse outcome with psychotherapy was very low, downgraded one level each for risk of bias, imprecision, and indirectness.

“We assessed the certainty of the outcome as very low, downgraded by one level for risk of bias (poor description of randomisation process and blinding, and withdrawal confounding bias), by one level for imprecision (small number of studies, small number of participants, and wide 95% confidence interval, which includes the null effect of no difference), and by one level for indirectness (studies in”

The certainty of evidence for the depressive symptoms outcome was very low, downgraded one level each for risk of bias, imprecision, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for risk of bias (poor description of randomisation and blinding, attrition bias, and withdrawal confounding bias), by one level for imprecision (wide 95% confidence interval, which includes the null effect of no differences between treatments), and by one level for indirectness (participants with recurrent”

• AssumptionRests on an assumption: The HAM-D scale contains domains that are symptoms of withdrawal.

The evidence for the 16-study relapse outcome was downgraded by one level for indirectness because three studies included panic disorder and the others recurrent or chronic depression.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (three studies included participants with panic disorder; the other studies included recurrent depression or chronic depression).”

The evidence for the 13-study relapse outcome was downgraded by one level for indirectness because three studies included panic disorder and the others recurrent or chronic depression.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, withdrawal confounding bias) and by one level for indirectness (3 studies included participants with panic disorder; the other studies included participants with recurrent depression or chronic depression).”

The evidence for SAS-SR social functioning was downgraded by two levels for risk of bias due to poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

• AssumptionRests on an assumption: Withdrawal symptoms may influence social and occupational functioning.

The certainty of evidence for the SAS-SR social functioning outcome was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

The evidence for CGI-S severity of illness was downgraded by two levels for risk of bias due to poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

• AssumptionRests on an assumption: Withdrawal symptoms including psychological symptoms may impair the severity of illness.

The evidence for HAM-A anxiety symptoms was downgraded by one level for indirectness because two studies included generalised anxiety disorder and one included recurrent depression.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and binding, attrition bias, and confounding withdrawal bias) and by one level for indirectness (two studies included participants with generalised anxiety disorder; one study included recurrent depression).”

The certainty of evidence for the CGI-S severity of illness outcome was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

The evidence for SAS-SR social functioning was downgraded by one level for indirectness because studies included participants with recurrent or chronic/double depression.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

The evidence for CGI-S severity of illness was downgraded by one level for indirectness because studies included participants with recurrent or chronic/double depression.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias) and by one level for indirectness (studies included participants with recurrent or chronic/double depression).”

All four studies in Comparison 3 included participants with recurrent depression in remission.

“All four studies included participants with recurrent depression in remission.”

The certainty of evidence for the three QoL subdomains was very low.

“We assessed the certainty of evidence for the three QoL subdomains as very low, downgraded by one level for risk of bias (poor description of blinding outcome assessors, attrition bias, and withdrawal confounding bias; withdrawal symptoms may impair QoL measures) and by one level for indirectness (studies included participants with recurrent depression or chronic/double depression) and imprecision”

The evidence for adverse events was downgraded by one level for indirectness because three studies reported only serious adverse events.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, high risk of attrition bias, confounding withdrawal bias; withdrawal symptoms may be misdiagnosed as adverse events) and by one level for indirectness (three studies reported only serious adverse events).”

The evidence for HAM-A anxiety symptoms was downgraded by two levels for risk of bias due to poor description of randomisation and blinding, attrition bias, and confounding withdrawal bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and binding, attrition bias, and confounding withdrawal bias) and by one level for indirectness (two studies included participants with generalised anxiety disorder; one study included recurrent depression).”

• AssumptionRests on an assumption: HAM-A scales contain symptoms from domains of psychic anxiety and somatic anxiety, all of which are symptoms of withdrawal.

The evidence for QoL subdomains was downgraded by one level for indirectness (studies included recurrent or chronic/double depression) and imprecision (wide 95% CI).

“We assessed the certainty of evidence for the three QoL subdomains as very low, downgraded by one level for risk of bias (poor description of blinding outcome assessors, attrition bias, and withdrawal confounding bias; withdrawal symptoms may impair QoL measures) and by one level for indirectness (studies included participants with recurrent depression or chronic/double depression) and imprecision”

The certainty of evidence for successful discontinuation rate was very low, downgraded for withdrawal confounding bias, imprecision, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for bias (withdrawal confounding bias) and by one level for imprecision (no pooling due to insufficient information) and indirectness (studies included participants with recurrent depression in remission).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The evidence for adverse events was downgraded by two levels for risk of bias due to poor description of randomisation and blinding, high risk of attrition bias, and confounding withdrawal bias (withdrawal symptoms may be misdiagnosed as adverse events).

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, high risk of attrition bias, confounding withdrawal bias; withdrawal symptoms may be misdiagnosed as adverse events) and by one level for indirectness (three studies reported only serious adverse events).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misdiagnosed as adverse events.

The evidence for QoL subdomains was downgraded by one level for risk of bias due to poor description of blinding of outcome assessors, attrition bias, and withdrawal confounding bias (withdrawal symptoms may impair QoL measures).

“We assessed the certainty of evidence for the three QoL subdomains as very low, downgraded by one level for risk of bias (poor description of blinding outcome assessors, attrition bias, and withdrawal confounding bias; withdrawal symptoms may impair QoL measures) and by one level for indirectness (studies included participants with recurrent depression or chronic/double depression) and imprecision”

• AssumptionRests on an assumption: Withdrawal symptoms may impair quality of life measures.

The certainty of evidence for the HAM-A anxiety symptoms outcome was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and binding, attrition bias, and confounding withdrawal bias) and by one level for indirectness (two studies included participants with generalised anxiety disorder; one study included recurrent depression).”

The certainty of evidence for the adverse events outcome was very low.

“We assessed the certainty of evidence for this outcome as very low, downgraded by two levels for risk of bias (poor description of randomisation and blinding, high risk of attrition bias, confounding withdrawal bias; withdrawal symptoms may be misdiagnosed as adverse events) and by one level for indirectness (three studies reported only serious adverse events).”

The certainty of evidence for relapse in Comparison 4 was low, downgraded for withdrawal confounding bias and imprecision.

“We assessed the certainty of evidence for this outcome as low, downgraded by one level for withdrawal confounding bias (study measured DESS symptoms but did not analyse data) and by one level for imprecision (single study had a small number of participants).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The certainty of evidence for quality of life in Comparison 4 was very low, downgraded for imprecision and risk of bias.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for imprecision (single study with a small number of participants) and by two levels for risk of bias (withdrawal confounding bias, large number of missing values, risk of performance bias).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The certainty of evidence for adverse events was very low, downgraded for risk of bias, imprecision, and indirectness.

“We assessed the certainty of evidence for this outcome as very low, downgraded by one level for risk of bias (withdrawal confounding bias; adverse events may include withdrawal symptoms), by one level for imprecision (no meta-analysis possible; low numbers of events and insufficient data supported study authors' conclusions), and by one level for indirectness (studies measured only serious adverse”

• AssumptionRests on an assumption: Adverse events may include withdrawal symptoms.

The certainty of evidence for successful discontinuation rate in Comparison 4 was low, downgraded for imprecision and risk of bias.

“We assessed the certainty of evidence for this outcome as low, downgraded by one level for imprecision (single study with a small number of participants) and by one level for risk of bias (withdrawal confounding bias; withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The certainty of evidence for depressive symptoms was low, downgraded for risk of bias and indirectness.

“We assessed the certainty of evidence for this outcome as low, downgraded by one level due to risk of bias (withdrawal confounding bias) and by one level due to indirectness (studies included recurrent depressive disorder).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The certainty of evidence for quality of life was very low, downgraded for risk of bias, imprecision, and indirectness.

“We assessed the certainty of evidence for quality of life as very low, downgraded by one level due to risk of bias (withdrawal confounding bias, attrition bias, performance bias; outcome assessed with self-report questionnaire and likely to be influenced by lack of blinding of participants), by one level due to imprecision (wide 95% confidence interval includes the null effect), and by one level”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

The certainty of evidence for relapse was low, downgraded for withdrawal confounding bias and indirectness.

“We assessed the certainty of evidence for this outcome as low, downgraded by one level for bias (risk of withdrawal confounding bias; studies did not use very slow tapering regimens or low doses) and by one level for indirectness (studies included participants with recurrent depression in remission).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

Despite inclusion of 10 studies in the meta-analysis, subgroup analysis for setting, disorder, type of antidepressant, or duration of antidepressant treatment could not be performed due to unevenly distributed covariates.

“Despite inclusion of 10 studies in the meta-analysis, we were unable to perform subgroup analysis for setting, disorder, type of antidepressant, or duration of antidepressant treatment due to unevenly distributed covariates”

The certainty of evidence for depressive symptoms in Comparison 4 was low, downgraded for risk of bias and imprecision.

“We assessed the certainty of evidence for this outcome as low, downgraded by one level due to risk of bias (withdrawal confounding bias, attrition bias) and by one level due to imprecision (a single study).”

• AssumptionRests on an assumption: Withdrawal symptoms may be misclassified as relapse and this may lead to restart of the antidepressant.

Sensitivity analysis to assess the potential impact of a tapering regimen of four weeks or less could not be performed because none of the studies without co-intervention reported a tapering regimen longer than four weeks.

“We could not perform sensitivity analysis to assess the potential impact of a tapering regimen of four weeks or less because none of the studies (without co-intervention) reported a tapering regimen longer than four weeks.”

It is unclear to what degree misclassified withdrawal symptoms contributed to relapse rates.

“Consequently, it is unclear to what degree misclassified withdrawal symptoms contributed to”

It is not possible for the review to draw conclusions about the best way to taper antidepressants.

“It is therefore not possible for our review to draw conclusions about the best way to taper antidepressants.”

Hyperbolic tapering may be of particular benefit for those unable to tolerate withdrawal symptoms.

“This approach may be of particular benefit for those unable to tolerate withdrawal symptoms.”

Sensitivity analysis for depressive symptoms or severity of illness in abrupt discontinuation versus continuation could not be performed because all included studies had non-commercial sponsorship or failed to report the source of funding.

“We could not perform sensitivity analysis for depressive symptoms or severity of illness because all included studies had noncommercial sponsorship or failed to report the source of funding.”

Withdrawal symptoms may affect almost every outcome, including adverse events, quality of life, social functioning, severity of illness, and anxiety and depression scores.

“Moreover, withdrawal symptoms may have an effect on almost every outcome including adverse events, quality of life, social functioning, severity of illness, and anxiety and depression scores.”

The review did not compare abrupt discontinuation with tapering per se.

“First, in our review, we did not compare abrupt discontinuation with”

Withdrawal symptoms may be misdiagnosed as relapse, a phenomenon known as withdrawal confounding bias.

“Withdrawal symptoms may be misdiagnosed as relapse, known as”

Sensitivity analysis for adverse events, depressive symptoms, anxiety symptoms, quality of life, social functioning, or severity of illness in tapering versus continuation could not be performed because all studies had non-commercial sponsorship or failed to report the source of funding.

“We could not perform sensitivity analysis for adverse events, depressive symptoms, anxiety symptoms, quality of life, social functioning, or severity of illness because all studies had non-commercial sponsorship or failed to report the source of funding.”

None of the included studies distinguished relapse from withdrawal symptoms.

“None of the studies distinguished relapse from symptoms of withdrawal.”

• Also stated in: § Discontinuation with high-intensity psychological interventions · plain

  “but none of the studies distinguished relapse from symptoms of withdrawal.”

• Also stated in: § Overall completeness and applicability of evidence · plain

  “the finding that trials did not distinguish relapse from symptoms of withdrawal.”

Thirty-two studies had high risk of bias due to selective reporting.

“Thirty-two studies were assessed as having high risk of bias due to selective reporting.”

Because such factors may influence the benefits and harms of long-term antidepressant use, this limits the applicability of review results to clinical practice.

“Given that such factors may influence the benefits and harms of long-term antidepressant use, this limits the applicability of review results to clinical practice.”

The review could not draw conclusions about whether it is effective and safe to discontinue long-term antidepressant use for patients with anxiety disorders.

“Therefore, we were not able to draw conclusions about whether it is effective and safe to discontinue long-term antidepressant use for patients with anxiety disorders.”

Overall, blinding was poorly described.

“Overall, blinding was poorly described.”

The review illustrates that initiation of high-intensity interventions such as PCT and MBCT may support discontinuation in patients with recurrent depression, with very low certainty of evidence.

“This review illustrates that initiation of high-intensity interventions such as PCT and MBCT may support discontinuation in patients with recurrent depression (very-low certainty of evidence)”

Applicability of the evidence in clinical practice is limited because most studies recruited participants from specialist outpatient mental healthcare settings.

“Applicability of the evidence in clinical practice is limited because most studies recruited participants from specialist outpatient mental healthcare settings.”

Many depressive episodes, 50% to 80%, treated in primary care are self-limiting within three to twelve months, based on six studies.

“Many depressive episodes (50% to 80%) treated in primary care are self-limiting within three to twelve months (6 studies) (Ormel 2020).”

• cites Ormel 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Use of different definitions of relapse, evaluated at different time points, compounds the difficulty and further limits generalisability.

“Use of different definitions of relapse, evaluated at different time points, compounds the difficulty and further limits generalisability.”

All studies had methodological limitations and were judged at unclear or high risk of bias in at least one domain.

“All studies had methodological limitations and were judged to be at unclear or high risk of bias in at least one domain.”