Factors Associated With Successful Medication Discontinuation After a Randomized Clinical Trial of Relapse Prevention in First-Episode Psychosis

Abstract

This follow-up study of a randomized clinical trial examines a cohort of patients who had participated in the trial 10 years ago to establish what factors were associated with successful discontinuation of antipsychotic medication regimens.

Not all patients with psychotic disorders require a lifetime of antipsychotic treatment. A study of patients on the schizophrenia spectrum found that after a mean duration of illness of 10.4 years, 24 of 70 (34%) were free of antipsychotic medication for at least 3 months, of whom 20 were medication-free for longer than a year.¹ Another study of first-episode psychosis reported that at 7-year follow-up, 17 of 103 patients (16.5%) had stopped antipsychotic medication during the previous 2 years.² Identifying clinical features of illness associated with lesser or greater likelihood of successfully discontinuing antipsychotic medication is of considerable importance. We report the proportion and factors associated with successful discontinuation during long-term follow-up of patients with first-episode psychosis.

Methods

We carried out a 10-year follow-up study of 178 patients with first-episode psychosis from a 1-year randomized clinical trial (RCT) of medication maintenance vs discontinuation to prevent relapse after complete resolution of initial psychotic symptoms in Hong Kong (NCT00334035).³ After the RCT, all patients received usual psychiatric care, including medication at the discretion of patients and treating psychiatrists. Clinical outcomes were examined at 10 years in the patients who provided written informed consent (NCT01926340).⁴

Baseline measures (Table 1) were assessed using standardized assessments. Patients who had successfully discontinued medication were defined as having not received any antipsychotic drugs during the final 2 years of follow-up and being free of key positive symptoms of psychosis (Positive and Negative Syndrome Scale scores: delusions ≤2, conceptual disorganization ≤3, hallucinations ≤2, suspiciousness ≤4, and unusual thought content ≤3)³ during the final month as assessed in a direct interview. Relapse during the follow-up period was assessed monthly through medical record review of a clinical global impression (CGI) of positive symptom severity. A relapse was defined as a change from a CGI positive score of less than 3 for at least 3 consecutive months to a score of 3 or more. (Apart from weekly consensus meetings between a clinician and raters, a CGI of positive symptoms was rated by the 3 raters from 8 independent medical records to ensure reliability; the intraclass correlation was 0.70, indicating good agreement.)

Table 1. Demographic and Clinical Features of Patients and Illness Course.

Variable	No./Total No. (%)
Age, mean (SD), y	24.5 (6.9)
Men	65/142 (46)
Education, mean (SD), y	11.9 (2.8)
Duration of untreated psychosis, median (IQR), d	91 (40-209)
Premorbid functioning score, mean (SD)a	2.7 (0.9)
Type of disorder at final reviewb
Schizophrenia spectrum
Schizophrenia	108/142 (76)
Schizophreniform disorder	17/142 (12)
Schizoaffective disorder	4/142 (3)
Other psychotic disorder
Brief psychotic disorder	7/142 (5)
Delusional disorder	4/142 (3)
Psychosis not otherwise specified	2/142 (1)
Severity of symptoms at RCT entry per Positive and Negative Syndrome scale score, mean (SD)
Positive	7.2 (3.2)
Negative	8.9 (3.2)
Total	33.2 (5.2)
Any relapse during RCT	63/142 (44)
Any relapse over 10-y follow-up from baseline	85/142 (60)
Antipsychotic medication at follow-up visit
None	28/138 (20)
Typical antipsychotic only	21/138 (15)
Atypical antipsychotic only	83/138 (60)
Both typical and atypical antipsychotic	6/138 (4)
Social and Occupational Functioning Assessment Scale at 2 y, mean (SD)c	60.6 (6.3)
Stroop interference, mean (SD)d	69.0 (23.2)
Successful discontinuation during RCTe	26/142 (18)

Abbreviations: IQR, interquartile range; RCT, randomized clinical trial.

^aPremorbid adjustment scale was used to assess patients’ premorbid functioning, including isolation, peer relations, scholastic performance, adaptation to school, and interests during childhood and adolescence using a 7-point scale, with 1 indicating the patient was very well adjusted and 7 indicating poor adjustment.

^bA lifetime diagnosis was made by a research psychiatrist using all available information over the full longitudinal course of illness, including a structured clinical interview for DSM-IV.

^cThis score ranges from 1 to 100; a higher score indicates poorer functioning. The score in the first 2 years after the start of the trial was averaged as a mean score of performance.

^dThe Stroop interference effect is defined as an increase in reaction time spent when the word meaning and the stimulus color do not match, calculated as Reaction Time of the Color Task – [(Reaction Time of Word Task + Reaction Time of Color Task) / 2]. Data were obtained in 83 patients (71 patients taking medication and 12 patients who were not taking medication).

^eSuccessful discontinuation refers to patients who received placebo and did not relapse in the RCT.

Binary logistic regression analysis was used to identify potential associated factors in this study (age, sex, education, duration of untreated psychosis, premorbid functioning, diagnosis, positive and negative symptoms, relapse during the RCT, relapse in the follow-up period, mean Social and Occupational Functioning Assessment Scale [SOFAS] scores at 2 years, cognitive function at baseline [information, digit span, arithmetic, digit backward, digit forward, Wisconsin Card Sorting Test perseveration errors, logical memory, verbal fluency, and Stroop test], and whether medication could be successfully discontinued during the RCT). Univariate regression analysis was first performed on each factor; apart from the Stroop interference variable, which was missing for 59 of 142 patients (42%), univariate variables with a P value of .05 or less (ie, sex, duration of untreated psychosis, never-relapsed status, mean SOFAS score, diagnosis, and successful discontinuation during RCT) were entered into a multivariate regression model concurrently for exploration of independent factors. Analysis was carried out with SPSS version 24.0 (IBM) from January to August 2018.

Results

Of 178 patients, 142 (80%) were interviewed at 10 years. Unavailable patients (n = 36) were excluded but had similar baseline demographic and clinical features.

In the final 2 years of follow-up, 23 of 142 patients (16.2%) successfully discontinued antipsychotic medications. Univariate regression analyses indicated this group included more men (15 of 23 [65%]) than the group continuing to receive medication or experience persistent positive symptoms (50 of 119 [42%]) and were more likely to have a duration of untreated psychosis less than or equal to 30 days (8 of 23 [35%] vs 16 of 119 [13%]; odds ratio [OR], 3.43 [95% CI, 1.25-9.40]; P = .02) and be relapse-free from baseline (15 of 23 [65%] vs 15 of 119 [13%]; OR, 13.00 [95% CI, 4.71-35.86]; P < .001). The group that successfully discontinued antipsychotic therapy also had a higher SOFAS mean score in the first 2 years after the start of the RCT (mean [SD], 63.5 [5.6] vs 60.1 [6.3]; OR, 1.10 [95% CI, 1.02-1.19]; P = .02), exhibited a lower Stroop interference result at baseline (mean [SD], 54.5 [13.7] vs 71.5 [23.6]; OR, 0.95 [95% CI, 0.92-0.99]; P = .02), had fewer patients with schizophrenia spectrum disorders (17 [74%] vs 112 [94%]; OR, 0.18 [95% CI, 0.05-0.59]; P = .01), and had more patients with successful discontinuation of antipsychotics during the RCT (8 [35%] vs 18 [15%]; OR, 0.33 [95% CI, 0.12-0.90]; P = .03) (Table 2). Multivariate regression analysis showed that being relapse-free from baseline was the only statistically significant factor (odds ratio, 8.65 [95% CI, 2.65-28.27]; P < .001). The model explained 35.3% of the variance.

Table 2. Baseline and Course of Illness Measures Associated With Successful Antipsychotic Discontinuation^a.

Measuresb	Patients, No. (%)		Univariate Logistic Regression
	Receiving Medication and/or Persistent Positive Symptoms (n = 119)	Successful Discontinuation (n = 23)
			Odds Ratio (95% CI)	P Value
Men	50 (42)	15 (65)	2.59 (1.02-6.57)	.046
Duration of untreated psychosis ≤30 d	16 (13)	8 (35)	3.43 (1.25-9.40)	.02
Never relapsed	15 (13)	15 (65)	13.00 (4.71-35.86)	<.001
Social and Occupational Functioning Assessment scale in first 2 y, mean (SD)	60.1 (6.3)	63.5 (5.6)	1.10 (1.02-1.19)	.02
Stroop interference reaction time, mean (SD), s	71.5 (23.6)	54.5 (13.7)	0.95 (0.92-0.99)	.02
Schizophrenia spectrum disorders	112 (94)	17 (74)	0.18 (0.05-0.59)	.01
Successful discontinuation during randomized clinical trial	18 (15)	8 (35)	0.33 (0.12-0.90)	.03

^aPatients with successful discontinuation were defined as not receiving any antipsychotic drugs during the final 2 years of follow-up and were free from 5 key positive symptoms of psychosis.

^bDefinitions are the same as in Table 1.

Discussion

After 10 years, 16% of patients with first-episode psychosis (with complete initial response to treatment) could successfully stop medication in the final 2 years of follow-up and remain free of positive symptoms of psychosis. We note that this definition of successful discontinuation applies only to the final 2 years and the symptom remission criterion only to the final month. However, in Hong Kong, the presence of mild but definite delusions or hallucinations (Positive and Negative Syndrome Scale delusion and hallucinations scores equal to 3⁵) would likely be an indication for restarting antipsychotic medication, so it is reasonable to posit that this group was relatively free of psychosis for much if not all of the preceding 2 years. Interestingly, using a different definition of successful discontinuation, a study in the Netherlands reported a similar rate (16%).² Being relapse-free, being male, having a shorter duration of untreated psychosis and a diagnosis other than one on the schizophrenia spectrum, having better functioning in the early phase of treatment, and remaining relapse-free in the placebo group during the RCT were individual factors associated with successful discontinuation, consistent with previous studies.^1,2

Diagnoses were longitudinal. Making an earlier, accurate diagnosis is challenging, and the present findings do not provide a test of the value of an accurate initial diagnosis in individual patients. Despite the small effect size, meta-analysis indicates an increased Stroop interference effect in schizophrenia.⁶

The present findings suggest early evaluation of this measure could contribute to assessment of prognosis. Remaining relapse-free was the central most important predictor of successful long-term discontinuation, supporting interventions to prevent relapse.