Long-Term Outcomes After Antipsychotic Discontinuation

Several studies have demonstrated the short-term risks of antipsychotic withdrawal (Kane et al., 1982; Crow et al., 1986; Gilbert et al., 1995; Gitlin et al., 2001; Thompson et al., 2018). Although the long-term outcomes are equally important, there is a paucity of long-term studies of outcomes of antipsychotic withdrawal. Such studies are complicated to perform because of the logistics required to recruit and randomize the patients as well as the high drop-out and non-compliance rates among patients with psychosis. Whereas short-term studies pinpoint the cautionary rates of relapse, 2 long-term follow-up studies were set to explore whether favorable outcomes would take place 7 or 10 years after a trial of antipsychotic discontinuation in remitted FEP patients (Wunderink et al., 2013; Hui et al., 2018) (Table 1).

Wunderink et al. (2007) conducted an open-label RCT, for 18 months, on 128 remitted FEP patients and demonstrated relapse rates of 43% and 21% in the dose-reduction/discontinuation and maintenance groups, respectively. Similarly, in a study by Chen et al. (2010), only 1 conducted in Asia, they performed a double-blinded RCT on 178 FEP patients free of symptoms for ≥1 year and demonstrated a relapse rate of 41% with quetiapine and 79% with placebo at the 12-month follow-up. The results of these RCTs agree with those of previous studies that patients have a higher risk for relapse with antipsychotic discontinuation than with maintenance treatment.

Researchers of the 2 previously mentioned studies performed follow-up of their respective trial patients. Wunderink et al. (2013) evaluated the rates of functional remission, symptomatic remission, and a combination of both for ≥6 months (i.e., recovery). The rates of symptomatic remission did not significantly differ between the dose-reduction/discontinuation and maintenance therapy groups after 7 years of trial. However, functional remission and recovery rates were significantly higher in the dose-reduction/discontinuation than in the maintenance groups. The high relapse rates in the dose-reduction/discontinuation group in the original trial had leveled off after 3 years of follow-up. Although Wunderink et al. (2013) concluded that early antipsychotic discontinuation was advantageous for FEP patients, only 20% of the patients in dose-reduction/discontinuation group (n=13) were able to discontinue antipsychotics, while all other patients either never stopped antipsychotics or restarted them shortly after an attempt to discontinue (Wunderink et al., 2007).

Therefore, the results of small trials should be interpreted with caution. Hui et al. (2018) recently conducted a 10-year follow-up of the participants in the Chen et al. (2010) study. They found that at the 10-year follow-up, 21% and 39% of patients in the quetiapine and discontinuation groups, respectively, had poor long-term clinical outcomes (i.e., clozapine use, persistent positive symptoms, or suicide). Further analyses revealed that a recurring episode within the second or third year since onset mediates the association between antipsychotic discontinuation and poor long-term outcome. Rather than encouraging early dose reduction or discontinuation, Hui et al. (2018) present advisory data against it—that a minimum of 3 years of maintenance treatment should be followed through to prevent suboptimal clinical outcomes.

Short-term studies have demonstrated a high risk for relapse with antipsychotic discontinuation and have advised against the approach. However, long-term studies of antipsychotic discontinuation have had contradictory results. Wunderink et al. (2007, 2013) indicated potential benefits of ceasing medication early, while Chen et al. (2010) and Hui et al. (2018) found it to precede poorer clinical outcomes at 10 years. Therefore, whether early discontinuation or a minimum of 3 years on maintenance treatment as suggested by Hui et al. (2018) instigates optimal long-term outcomes is an issue of contention.

Methodology and Quality Assessment

Many attempts have been made since the 1980s to evaluate the feasibility of antipsychotic discontinuation. Viguera et al. (1997) reviewed all the relevant studies on this topic and identified 2 key methodological flaws: variable antipsychotic doses (200–1150 and 175–890 chlorpromazine equivalent mg/d for abrupt and gradual discontinuation, respectively) and variable definitions of remission (although all studies reported that patients were “clinically stable”).

Better designed studies have been published since 2000. However, a careful review of these studies identifies several issues. First, Andreasen’s criteria were most frequently used to define remission. The Andreasen’s criteria define remission as a score ≤3 on each item of Positive and Negative Syndrome Scale (PANSS) and Brief Psychiatric Rating Scale along with a score ≤2 on each item of Scale for the Assessment of Positive Symptoms (SAPS) and Scale for the Assessment of Negative Symptoms for ≥6 months (Andreasen et al., 2005). Because a systematic review reported a mean relapse rate of 77% and 90% by 1 and 2 years, respectively, following antipsychotic discontinuation (Zipursky et al., 2014), using the criteria for full recovery rather than remission criteria would be more appropriate. In the survey, Korean psychiatrists favored the definition of remission as the complete absence of positive and negative symptoms or scores ≤2 on each item of the PANSS (Lee et al., 2020). Asian psychiatrists considered “absence of any relapse following first episode” as the most important criterion for antipsychotic discontinuation (Hui et al., 2019). In this respect, defining remission while allowing a score of ≤4 on the PANSS positive subscale, as did Wunderink et al. (2007), would appear to be too liberal. Moreover, only 1 study (Mayoral van Son et al., 2016) used clinical recovery as a criterion for antipsychotic discontinuation.

Second, there was marked variation in the mean antipsychotic doses before discontinuation: long-acting risperidone injection (consta) 25–50 mg, haloperidol 2.5–4 mg, olanzapine 7.1 mg, and risperidone 2 mg. These doses are in fact comparable with the minimum effective daily dose/olanzapine 1 mg equivalents reported by Leucht et al. (2014): aripiprazole 10 mg, asenapine 10 mg, clozapine 300 mg, haloperidol 4 mg, iloperidone 8 mg, lurasidone 40 mg, olanzapine 7.5 mg, paliperidone 3 mg, quetiapine 150 mg, risperidone 2 mg, sertindole 12 mg, and ziprasidone 40 mg (1.33 mg, 1.33 mg, 40 mg, .53 mg, 1.07 mg, 5.33 mg, 1 mg, .4 mg, 20 mg, .27 mg, 1.60 mg, and 5.33 mg risperidone, respectively). These doses, however, were studied for maintenance treatment, not for discontinuation. To prepare for discontinuation, doses should be reduced to levels lower than those mentioned above. In all studies, doses were tapered over 1 to 6 months. However, 30%–40% of relapses occurred in the first 3 months after antipsychotic discontinuation (Gitlin et al., 2001; Chen et al., 2010; Boonstra et al., 2011). It is possible that the early relapses were moderated by the duration of antipsychotic intake preceding discontinuation due to the medication build-up of supersensitive dopamine receptors (Samaha et al., 2007). Therefore, antipsychotics should be tapered more gradually and cautiously. Future RCTs should evaluate the risks of relapse after dose reduction or discontinuation of typical D2 blockers and partial D2 agonists to determine when and how fast the antipsychotics should be tapered (Murray et al., 2016).

Third, relapses can be classified into mild (i.e., PANSS positive subscale score ≥3 or change in Clinical Global Impression score ≥5 for ≥1 week), moderate, or severe (i.e., PANSS positive subscale score ≥5 or change in Clinical Global Impression score ≥6 for ≥1 week). If recovery is used as the criterion for antipsychotic discontinuation, it may be justified to use the criteria for mild relapses. As for the duration of re-emerging symptoms, which justifies the decision to restart antipsychotics, the 1-week criterion could be modified depending on the availability of proactive psychosocial support, 24-hour crisis intervention, or alternative medication (e.g., benzodiazepine) (Gaebel et al., 2016).

Finally, patients were regularly assessed for up to 3 years after antipsychotic discontinuation in most studies. However, only 1 study described what was included in these assessments (Emsley et al., 2014). Interestingly, in the Gaebel et al. (2016) study, patients were given the antipsychotic that the patient was already taking or the benzodiazepine lorazepam for early intervention. They chose these treatment options because a previous study had found no differences between these 2 drugs (Carpenter et al., 1999). Therefore, if the severity of re-emerging positive symptoms is mild, we can try using benzodiazepine first and use antipsychotics only when symptoms do not subside after more than 1 week.

Using the GRADE approach to assess the quality of RCTs (Table 2), we found serious or very serious problems in the study design and results. The study limitations included considering drop-out patients as having no relapse, unclear definition of remission, unclear allocation concealment, open-label study designs, diagnostic heterogeneity, and failure to report outcomes for the control group. Imprecision in these studies was evident by the small number of included patients, few relapses, wide confidence intervals, and confidence intervals including the value 1. Most studies were low or very low quality, except for 2 studies that were moderate quality (Chen et al., 2010; Hui et al., 2018). All studies, except for 1 (Wunderink et al., 2013), advised against antipsychotic discontinuation because of high relapse rates (19%–82% at a mean follow-up of 1 year).

Table 2.

GRADE Evidence Profile: Studies on Antipsychotic Discontinuation

						Summary of findings
	Quality assessment					No. of patients			Absolute risk
Study	Limitations	Inconsistency	Indirectness	Imprecision	Publication bias	Discontinuationa	Maintenance	Relative risk (95% CI)	Control risk	Risk difference (95% CI)	Quality of evidence	Recommendation
Kane et al. (1982)	Very serious limitationb	No serious inconsistency	No serious indirectness	Very serious imprecisioni	Undetected	7/17	0/11	Not estimable (no events)	0 per 1000	41.18% (17.78 to 64.57)	〇〇Low	No recommendation
Crow et al. (1986)	Very serious limitationb	No serious inconsistency	No serious indirectness	Very serious imprecisionj	Undetected	41/66	25/54	1.34 (0.95 to 1.89)	463 per 1000	15.82% (−1.89 to 33.54)	〇〇Low	No recommendation
McCreadie et al. (1989)	Very serious limitationc	No serious inconsistency	No serious indirectness	Very serious imprecisioni	Industry funded	4/7	0/8	Not estimable (no events)	0 per 1000	57.14% (20.49 to 93.8)	〇〇Low	No recommendation
Wunderink et al. (2007)	Very serious limitationd	No serious inconsistency	No serious indirectness	No serious imprecision	Industry funded partially	28/65	13/63	2.09 (1.19 to 3.65)	210 per 1000	22.44% (6.80 to 38.09)	〇〇Low	No recommendation
Chen et al. (2010)	Serious limitatione	No serious inconsistency	No serious indirectness	No serious imprecision	Industry funded partially	56/89	27/89	2.07 (1.46 to 2.95)	303 per 1000	32.6% (18.73 to 46.44)	〇Moderate	No recommendation
Boonstra et al. (2011)	Serious limitationf	No serious inconsistency	No serious indirectness	Very serious imprecisionj	Industry funded partially	10/11	4/9	2.05 (0.96 to 4.35)	444 per 1000	46.46% (9.83 to 83.1)	〇〇Low	No recommendation
Gaebel et al. (2011)	Very serious limitationg	No serious inconsistency	No serious indirectness	Very serious imprecisionk	Undetected	4/21	0/23	Not estimable (no events)	0 per 1000	19.05% (2.25 to 35.84)	〇〇Low	No recommendation
Wunderink et al. (2013)	Very serious limitationd	No serious inconsistency	No serious indirectness	Serious imprecisionl	Industry funded	21/52	9/51	2.29 (21.2 to 38.56)	176 per 1000	22.74% (5.79 to 39.69)	〇〇Low	Conditional
Emsley et al. (2014)	No serious limitation	No serious inconsistency	Serious indirectness (only 2 active comparators without control)	Very serious imprecisionj	Undetected	9/12	19/21	0.83 (0.58 to 1.18)	No control	−15.48% (−43 to 12.05)	〇〇Low	No recommendation
Gaebel et al. (2016)	Very serious limitationh	No serious inconsistency	Serious indirectness (no results on control group)	Serious imprecisionm	Undetected	10/19	Not available	Not estimable	Not available	Not estimable	〇〇〇 Very low	No recommendation
Hui et al. (2018)	Serious limitatione	No serious inconsistency	No serious indirectness	No serious imprecision	Industry funded partially	35/89	19/89	1.84 (1.15–2.96)	210 per 1000	17.98% (4.73 to 31.22)	〇Moderate	No recommendation

Open in a separate window

^a Include also dose reduction or intermittent treatment.

^b Drop out cases considered as non-relapse, poor definition of remission.

^c Unclear allocation concealment, poor definition of remission.

^d Open label, diagnostic heterogeneity, poor definition of remission.

^e Diagnostic heterogeneity, poor definition of remission.

^f Open label.

^g Open label, drop out cases considered as non-relapse, poor definition of remission.

^h Open label, poor definition of remission, failure to report outcome of control group.

ⁱ Few patients, few events and wide CI

^j Few patients, and CI include 1.0.

^k Few events and wide CI.

^l Wide CI.

^m Few patients.

Most patients in these studies (approximately 90%) had schizophrenia spectrum disorders (schizophrenia, schizoaffective disorder, and schizophreniform disorder), and only a small proportion (approximately 10%) had psychotic disorder not otherwise specified (NOS), delusional disorder, or brief psychotic disorder. Based on these findings, we conclude that the current evidence is against antipsychotic discontinuation in patients with remitted FES spectrum disorders. However, this conclusion should be taken in the context that most RCTs were of low-to-moderate quality based on the GRADE (Table 3).

Table 3.

Overall GRADE Evidence Profile

						Summary of findings
	Quality assessment					Relapse rate (%)
Study	Limitations	Inconsistency	Indirectness	Imprecision	Publication bias	Discontinuation	Maintenance	Relative risk (95% CI)	Risk difference (95% CI)	Quality of evidence	Recommendation
	Very serious limitation	No serious inconsistency	No serious indirectness	Very serious imprecision	Undetected	53	19	__	0.26 (0.18 to 0.34)	〇〇Low	No recommendation

Open in a separate window

Characteristics of Patients Who Discontinued Antipsychotics

It is important to identify patients who may safely discontinue antipsychotics after remission. The strongest predictor of relapse is medication non-compliance. However, relapse may also be predicted by certain psychosocial factors, including poor premorbid adjustment (Robinson et al., 1999; Larsen et al., 2000; Malla and Payne, 2005), long duration of untreated psychosis (Alvarez-Jimenez et al., 2012), expressed emotions in the household (Pourmand et al., 2005), and substance misuse (Barnett et al., 2007; Larsen et al., 2006). Decline in cognitive functions, including general memory (Verdoux and Liraud, 2000), set shifting (Chen et al., 2005), visual working memory (Hui et al., 2016), visual reproduction (Verdoux et al., 2000), verbal learning (Rund et al., 2007), attention, and processing speed (Wölwer et al., 2008), also predict relapse. However, these predictors were identified from naturalistic studies (where discontinuation is one of those predictors) that did not exclusively investigate the outcome of patients who discontinued antipsychotics. In addition, more pertinent to the issue of antipsychotic discontinuation is to identify factors that pinpoint patients who can do so without relapsing, evidence of which is currently limited.

Several studies have attempted to identify characteristics of patients with psychosis or schizophrenia who can discontinue antipsychotics without experiencing a relapse (Table 4). Fenton and Glashan (1987) retrospectively reviewed the data from 23 patients with chronic schizophrenia who remained stable despite not taking antipsychotics for an average of 15 years. They identified better premorbid occupational and social competence, fewer hebephrenic traits, and better regulation of depressed mood during the first hospital admission in these patients compared with other patients. Another research group identified a subgroup of first- or second-episode schizophrenia patients from the Soteria Project who received specialized psychosocial interventions and minimal or no medications during a 2-year follow-up period (Bola and Mosher, 2002). These patients had favorable outcomes in terms of hospitalization rates, independent living, psychopathology, and social and occupational functioning. In a quasi-experiment, Bola et al. (2006) compared the outcomes of patients who received family-oriented psychosocial interventions during a 3-week antipsychotic-free period with those of patients who received psychosocial interventions and antipsychotics. The 2-year study was completed by 46% of the participants without the need for antipsychotics. Different variables within the same domains of good prognosis and fewer schizophrenia symptoms predicted antipsychotic-free response or nonresponse with 74% accuracy.

Table 4.

Key Results of Studies Investigating Characteristics of Patients Who Discontinued Antipsychotics

Study (design)	Diagnoses and total sample size	Definition of remission	Outcome of discontinuation	Characteristics associated with patients who were medication-free
Fenton and McGlashan (1987) (Retrospective study)	Chronic S; n=23	Clinical global outcome score of moderate or better, never re-hospitalized, and no psychotropic medication use during follow-up period	Over average of 15 y: patients who discontinued AP sustained remission	Better psychosocial performance and acquired skills, more favorable premorbid occupational and social adjustments, fewer hebephrenic traits, and better preservation of depressed mood
Bola and Mosher (2002) (Retrospective exploratory study)	S (DSM-II); n=179	No information	At 2-y follow-up: better outcomes on composite outcome scale (comprising of rehospitalization, psychopathology, independent living, social and occupational functioning) compared with those on AP continuously	Being older, higher score on Goldstein Adolescent Social Competence Scale, and fewer symptoms
Bola et al. (2006) (Quasi-experimental study)	Non-affective psychosis (DSMIII-R); n=135	No information	At 2 y: 46.3% completed 2-y study without AP use	Higher score on GRIP on Life scale, parents had no mental health treatment history, higher scores on items “wahnstimmung” (sense of inner meaning or significance) and distorted speech of Comprehensive Psychopathology Rating Scale
Harrow and Jobe (2007) (Naturalistic observational study)	S (DSM-III); n=64	(Definition of recovery) Absence of positive and negative symptoms throughout follow-up period and adequate psychosocial functioning; scoring 1 or 2 on LKP scale	At 10 y: 23% had “psychotic activity” At 15 y: 28% had “signs of psychotic activity” and 40% were in recovery	Better developmental achievements prior to onset, more favorable attitudinal approaches, more internal resources, and favorable prognostic factors
Harrow et al. (2012) (20-year longitudinal study)	S (DSM-III); n=70	(Definition of recovery) Absence of positive or negative symptoms, no rehospitalizations during follow-up year and scores ≥2 on Strauss–Carpenter scales	At 4.5-, 7.5-, 10-, 15-, 20-y follow-up: less likely to present psychotic symptoms and experienced more periods of recovery	Greater resilience, more favorable premorbid developmental achievements, more internal resources, favorable prognostic factors and better neurocognitive functioning
Larsen-Barr et al. (2018) (Cross-sectional survey)	Psychosis, with a history of antipsychotic use; n=144	No information	For varying periods of time: 55% successfully stopped all AP for some time	Patients who received support of various types were more likely to have successfully discontinued APs
Bowtell et al (2018b) (Review)	Varies between studies	Varies between studies	Proportion of individuals who experienced relapse following discontinuation ranged from 19.4% to 97% across 11 studies. No information regarding the remission rate of discontinuation group or comparing that with the maintained treatment group.	More years of education, better clinician-rated prognosis at baseline, and sustained remission at 2 y, were predictors of sustained remission following discontinuation. Several demographic and clinical factors found to be predictive of relapse. However, each predictive finding was reported only in 1 cohort of individuals such that no predictive findings were replicated across studies.
Tani et al. (2018) (Review)	Varies between studies	Varies between studies	Mean relapse rate after discontinuation: 38.3% per year	Older age, on a lower dose of AP prior to discontinuation, shorter DUP, older age at illness onset, better social functioning, less severe positive symptoms at baseline, and fewer previous relapses
Torgalsbøen et al. (2018) (Naturalistic observational study)	S (DSM-IV), n=28	For ≥2 y, scores on items of PANSS must be ≤3; scores on BPRS also ≤3; item scores on SAPS and SANS ≤2 (additionally for full recovery) Working or studying, living independently, and at least 1 weekly social activity	No information	Fully recovered first-episode schizophrenia patients (where 10% of them no longer required antipsychotic medication) had better resilience
Fu et al. (2019) (Naturalistic observational study)	S (DSM-IV), n=28	For ≥2 y, scores on items of PANSS must be ≤3; scores on BPRS also ≤3; item scores on SAPS and SANS ≤2 (additionally for full recovery) Working or studying, living independently, and at least 1 weekly social activity	Achieved recovery at earlier time point and experienced no relapses during follow-up periods	Use of active coping strategies such as being more aware of symptoms, mindful breathing, and changing of negative thought patterns promotes better recovery outcome. Larger improvements in processing speed and occupational functioning among patients not on antipsychotics than those who remained on antipsychotics.

Open in a separate window

Abbreviations: AP, antipsychotic; BPRS, Brief Psychiatric Rating Scale; DSM, Diagnostic and Statistical Manual of Mental Diseases; DUP, duration of untreated psychosis; LKP, Levenstein-Klein-Pollack scale; PANSS, Positive and Negative Syndrome Scale; S, schizophrenia; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms.

Using a different approach, Harrow and Jobe (2007) conducted a 15-year prospective longitudinal study to compare the premorbid and personality factors between schizophrenia patients on and off antipsychotics. They found that patients not on antipsychotics had longer periods of recovery and better functioning than those on antipsychotics. Patients who did not relapse after antipsychotic discontinuation had better premorbid functional development, more favorable personality traits, better prognostic factors, and greater resilience than patients who relapsed. Similarly, Torgalsbøen et al. (2018) identified resilience to be associated with fully recovered FES patients, with 10% of patients not on any treatment.

A secondary analysis performed using the protocol of Chen et al. (2010) evaluated the predictors of relapse for remitted FEP patients, both on and off antipsychotics (Hui et al., 2013). For patients who discontinued antipsychotics, relapse over the next year was predicted by a diagnosis of schizophrenia, poor semantic fluency, and higher blink rate.

In a longitudinal study by Harrow et al. (2012), 139 patients with psychosis from the Chicago Follow-up Study were followed-up 7 times over 20 years. Schizophrenia patients who remained stable without antipsychotics did not relapse more frequently than those who remained on maintenance antipsychotics. Greater resilience, better prognostic factors, premorbid functional development, less vulnerability towards anxiety, and better cognition were seen more frequently in patients off antipsychotics than on antipsychotics.

A recent cross-sectional survey by Larsen-Barr et al. (2018) observed that 105 out of 144 current or past antipsychotic users have attempted antipsychotic discontinuation at least once. In 55% of these attempts, patients received professional, service user, or peer support; this support had a positive association with successful attempts and negative association with relapses during withdrawal and need for antipsychotics. Hui et al. (2018) examined the 10-year clinical outcomes of the 178 FEP patients included in a previous trial (Chen et al., 2010) and evaluated the factors that predict successful antipsychotic discontinuation. Antipsychotics were successfully discontinued for 2 years, without any positive symptoms resurfacing, in 16% of FEP patients. Absence of relapses is, therefore, the most significant predictor of successful antipsychotic discontinuation. The other predictors include male sex, shorter duration of untreated psychosis, a diagnosis of non-schizophrenia spectrum disorder, better functioning at the initial phase of treatment, lower interference effect during the Stroop task, and remaining relapse-free while on placebo during RCT.

In another longitudinal study, Fu et al. (2019) identified 10 out of 28 FES patients who had recovered at the eighth follow-up during 6–8 years of follow-up. The processing speed and occupational functioning improved significantly in patients who were not on antipsychotics compared with those who remained on antipsychotics. Importantly, many patients used active coping strategies for recovery (e.g., changing negative thought patterns, improving understanding of symptoms, and practicing mindful breathing).

Tani et al. (2018) conducted a systematic review of 37 studies and observed that safe and successful antipsychotic discontinuation was predicted by lower antipsychotic dose prior to discontinuation, shorter duration of untreated psychosis, older age at disease onset, older age at present, better social functioning, milder positive symptoms at onset, and infrequent relapses. However, another review of studies conducted on FEP patients did not identify any replicated predictors of remission after antipsychotic discontinuation (Bowtell et al., 2018b). In patients who responded well to the initial antipsychotics and had sustained remission with maintenance treatment, longer duration of remission predicted lower risk for relapse following antipsychotic discontinuation (Hui et al., 2019a).

Further research is needed to identify factors that can identify patients who will and will not relapse following antipsychotic discontinuation. Although some of these factors have been identified, findings from the previous studies need to be confirmed in future studies and biological markers need to be identified to understand the mechanisms underlying relapses. It is important to identify patients who are likely to benefit from antipsychotic discontinuation. It is evident that not all patients with psychosis require life-long treatment (Jablensky and Sartorius, 2008; Harrow et al., 2012; Zipursky et al., 2014), and many patients discontinue antipsychotics without taking professional advice (Lacro et al., 2002; Zygmunt et al., 2002; Fleischhacker et al., 2003). It is therefore essential to determine the safest method of discontinuing antipsychotics.

Review of Clinical Practice Guidelines

More than 20 clinical practice guidelines and algorithms have been proposed since 2000 for the treatment of schizophrenia (Table 5). A systematic review reported that 10 out of 11 guidelines and algorithms on the long-term antipsychotic use in the maintenance phase of schizophrenia, published between 2000 and 2011, advised against antipsychotic discontinuation within 5 years of treatment (Takeuchi et al., 2012). While exclusively for patients with FES, 6 of the 11 guidelines partially recommended antipsychotic discontinuation after 1 to 2 years of treatment. Intermittent or targeted antipsychotic use (i.e., stepwise drug discontinuation and early drug use in case of prodromal or early symptoms of relapse) is a potential alternative to antipsychotic discontinuation, and 9 guidelines and algorithms mentioned this approach, but none recommended it. For maintenance of remission, only 1 guideline suggested dose reduction (not lower than one-half of the effective dose for the acute phase) (Verma et al., 2011), while none of the others mentioned about or simply did not recommend dose reduction or lower dose therapy during this phase. Another systematic review conducted by Hui et al. (2019b) included 18 clinical practice guidelines for the treatment of maintenance phase of FES, and none of them recommended antipsychotic discontinuation.

Shimomura et al. (2020) updated their previous systematic review of 11 guidelines and algorithms published after 2013 and added newly published guidelines/algorithms to the review. They suggested that in terms of considering antipsychotic discontinuation during maintenance phase, these newly published guidelines/algorithms seemed to be inclined to shift from “not recommended” to “partially recommended,” not only for FES (7 newly published guidelines endorsed, 4 not mentioned), but also for schizophrenia in general (also 7 newly published guidelines endorsed, 4 not mentioned) after a period of 1 to 5 years of treatment, possibly due to the recent reports on favorable long-term outcome in FEP: (1) the dose reduction/discontinuation patients experienced twice the recovery rate of the maintenance treatment patients at 7 years of follow-up (Wunderink et al., 2013), and (2) pooled prevalence of recovery among 9642 individuals with FEP was 38% (mean follow-up 7.2 years) (Lally et al., 2017). It is noteworthy that only the APA guideline (Lehman et al., 2004) did not recommend antipsychotic discontinuation in FES patients. Although intermittent or targeted antipsychotic use was not endorsed by the new guidelines and algorithms, antipsychotic dose reduction was recommended by 1, partially recommended by 5, and not mentioned by 5 out of 11 guidelines and algorithms published since 2013. In comparison, among the 9 guidelines published before 2012, 2 recommended, 3 partially recommended, 2 did not recommend, and 1 recommended a lower dose strategy. Therefore, the newer guidelines are more accepting of dose reduction/lower dose strategy during maintenance phase.

In addition to considering the stage of illness (first episode vs multiple episodes), long-term treatment strategy (maintenance vs intermittent treatment), and dosage for maintenance (the same as that used in acute phase vs lower dose strategy), more details should be considered for achieving appropriate and successful antipsychotic discontinuation. As seen in most guidelines, the minimum duration of stabilization under long-term antipsychotic treatment, usually recognized as a pre-requisite of discontinuation, was not uniformly defined, varying from 1, 1.5, 1 to 2, 2, or even up to 2 to 5 years (Shimomura et al., 2020). In general, the clinical treatment guidelines for FEP recommended the duration of maintenance treatment to be between 1 and 2 years (International Early Psychosis Association Writing Group, 2005; Early Psychosis Guidelines Working Group, 2010; Abidi et al., 2017; Remington et al., 2017), while a recent study has proposed that a minimum of 3 years promotes optimal long-term outcomes (Hui et al., 2018). The discontinuation strategy (discontinuation within a few weeks vs gradual tapering down antipsychotic dosage) as well as the tempo of gradual reduction (that is, the proportion of dose reduced at a time and the interval between dose reductions) is arbitrarily determined, mainly based on doctor’s clinical judgment or shared decision-making with the patients. Only 1 guideline specifically advised to “stop gradually over at least 3–6 months with close follow-up” (Galletly et al., 2016). This variation in strategy based on clinician judgment may affect the outcomes after drug discontinuation (National collaborating centre for mental health, 2014) or dose reduction but has not received adequate attention (Liu et al., 2021). Moreover, only 2 guidelines suggested monitoring 2 years after discontinuation (Grover et al., 2017, National collaborating centre for mental health, 2014). Gaebel et al. (2019) reported that the clinical course following antipsychotic discontinuation is heterogeneous and complex. There was wide variation in the clinical outcomes of patients, including no relapses, acute or insidious relapses, or clinical deterioration following drug discontinuation. Nonetheless, there was a minority of patients who remained clinically stable after antipsychotic discontinuation, but no individual-level variables predicted successful antipsychotic discontinuation or higher risk for relapse following discontinuation (Bowtell et al., 2018a, 2018b).

Correll et al. (2018) critically appraised the risk-benefit ratio of long-term antipsychotic use in patients with schizophrenia and identified a small group of patients with no relapse despite prolonged periods of no antipsychotic use. However, they concluded that the evidence favored continuous antipsychotic use. In response to that study, several prominent researchers advised to “encourage a sense of curiosity about the possibility of dose reduction and discontinuation in appropriate patients (Marder and Zito, 2018),” to have “guidelines to be developed on when and how slowly to reduce antipsychotics, and in whom it is appropriate to eventually stop them (Murray and Di Forti, 2018),” and if guidelines are to be changed, “due to the subjectivity of psychiatric outcomes—there is room for interpretation, in the future the evidence will have to be presented such that patients can decide themselves” (Leucht, 2018).

Ongoing Discontinuation Trials

Four prominent studies of medication discontinuation are underway (Table 6). The participants of these studies have been or will be randomized into maintenance or dose reduction strategy groups. If the dose is successfully reduced, participants can discontinue their antipsychotics completely. The primary outcomes of interest in these studies are the functional outcomes.

The Asian College of Neuropsychopharmacology (AsCNP) Guideline Committee provided the recommendation to endorse the manuscript and the permission for the authors to insert the recommendation from AsCNP in their manuscript.

This study was supported by a grant from the Korean Mental Health Technology R&D Project, Ministry of Health and Welfare, Republic of Korea (HL19C0015) and a grant of the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (HI18C2383).