Introduction
Relapse constitutes a major problem in managing patients with psychotic disorders, with up to 80% of patients relapsing within 5 years following illness onset (Robinson et al., Reference Robinson, Woerner, Alvir, Bilder, Goldman, Geisler, Koreen, Sheitman, Chakos, Mayerhoff and Lieberman1999). Meta-analysis shows that the risk of relapse was up to over 90% by 2 years following first episode psychosis (Zipursky et al., Reference Zipursky, Menezes and Streiner2014). Relapse can cause substantial or even irreversible damage, particularly in young patients during the most productive period of their lives. Maintenance antipsychotic treatment for a considerable period of time is often required to prevent symptom recurrence. In the first episode schizophrenia patients whose antipsychotics are discontinued (stepwise), those responding well to antipsychotics treatment had a greater risk of relapse (Gaebel et al., Reference Gaebel, Riesbeck, Wolwer, Klimke, Eickhoff, von Wilmsdorff, de Millas, Maier, Ruhrmann, Falkai, Sauer, Schmitt, Riedel, Klingberh and Moller2016). Longitudinal studies have revealed that around 20% of patients will not relapse after their first episode of psychosis (Shepherd et al., Reference Shepherd, Watt, Falloon and Smeeton1989; Linszen et al., Reference Linszen, Dingemans and Lenior2001). However, not much long-term data exist to help characterize this subgroup of patients who do not relapse. The successful identification of factors predicting this subgroup can help tailor a better treatment approach in relapse prevention, in conjunction with maintenance antipsychotic treatment.
Few studies have examined the proportion and predictors of first-episode psychosis (FEP) patients who do not relapse in the long term. In a naturalistic follow-up study, Alverez-Jimenez et al. (Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011) found that 16.5% of FEP patients did not relapse during the 7.5-year follow-up period, and that they are characterized by having a duration of untreated psychosis (DUP) <60 days, displaying more rapid response to antipsychotic treatment, and being less likely to have parental loss at baseline (Alverez-Jimenez et al., Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011). However, other important predictors such as neurocognitive functioning were not explored. The study has a dropout rate of up to 66% at follow-up, rendering generalizability to all psychosis patients difficult. Furthermore, relapse was retrospectively recalled by patients during the follow-up assessment, which may have introduced a bias towards recalling only more severe episodes and ignoring less ‘dramatic’ ones, resulting in an underestimated rate of relapse-free patients. In another study of FEP patients who were randomized to an 18-month dose-reduction trial following symptomatic remission, Wunderink et al. (Reference Wunderink, Nieboer, Wiersma, Sytema and Nienhuis2013) found that 34.9% of patients did not relapse during a 7-year period, but no predictors of relapse were reported.
This was a 10-year follow-up study of patients who were previously enrolled in a 12-month randomized controlled trial (RCT) on medication discontinuation and relapse following complete resolution of positive symptoms from FEP (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010). Data from the RCT showed that the Kaplan–Meier estimate of the risk of relapse at 12 months nearly doubled in the placebo group (79%) compared with the maintenance group (41%) (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010). In this 10-year follow-up study, we examined (1) the proportion of patients who did not relapse over the 10 years, (2) the potential baseline predictors of not relapsing, including socio-demographic information, symptoms, functioning, as well as neurocognitive functions, and (3) the clinical and neurocognitive outcome correlates of this subgroup of relapse-free patients.
Methods
Study design
Across the Hong Kong Special Administrative Region of approximately seven million people, specialized teams of the Early Assessment Service for Young People with Psychosis provide assessment and treatment for patients with FEP (Chen, Reference Chen, Ehmann, MacEwan and Honer2004). Between 2003 and 2005, 178 FEP patients with complete resolution of positive symptoms were randomized to receive either quetiapine (400 mg/d) or placebo for 12 months (ClinicalTrials.gov Identifier: NCT00334035) (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010). Upon completion of the 12-month RCT, patients continued to receive naturalistic regular care at general adult out-patient psychiatric clinics. Between November 2013 and December 2014, this cohort was followed up prospectively after 10 years to assess their clinical outcomes, including persistent positive symptoms, the requirement of taking clozapine, and suicide (ClinicalTrials.gov Identifier: NCT01926340) (Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen, Pang, Lui, Chung, Yeung, Ng, Lo, Jones, Sham and Chen2018). The current study focused on examining the long-term outcome of relapse.
This study was approved by the institutional review boards at each site and carried out in accordance with Good Clinical Practice and the Declaration of Helsinki. All participants provided written informed consent.
Participants
Included participants had a diagnosis of schizophrenia or non-affective psychosis (schizophreniform disorder, schizoaffective disorder, brief psychotic disorder, or psychosis not otherwise specified) (DSM-IV) (APA, 1994), were aged 18–65 years, had been treated with antipsychotic drugs continuously for at least 1 year, had good medication compliance (missed <50% of their medication, missed <50% of their clinic visits, or had no history of medication discontinuation), and had no history of relapse (defined as no increase of positive symptoms of psychosis requiring admission to hospital or adjustment of medication). Patients had to be free of positive symptoms of psychosis for at least 8 weeks as assessed using (1) five Positive and Negative Syndrome Scale (PANSS; Kay et al., Reference Kay, Fiszbein and Opler1987) items: delusions, conceptual disorganization, hallucinations, suspiciousness, unusual thought content, and (2) the Clinical Global Impressions (CGI) scale (Guy, Reference Guy and Guy1976b) with a score of 2 (borderline or questionable) or less. Exclusion criteria were a diagnosis of drug-induced psychosis, treatment with clozapine, mood stabilizing medications (lithium, valproate or carbamazepine) or depot medication, and a risk of suicide or violence.
Diagnosis at follow-up was determined using the best-estimate consensus approach with all sources of information available, including the validated Chinese version of the Structured Clinical Interview for DSM-IV (So et al., Reference So, Kam, Leung, Chung, Liu and Fong2003), medical records, and history from research assistants during face-to-face interviews. Two experienced psychiatrists reached a consensus in the diagnosis for each subject.
Outcome measures
To obtain the outcome measures, research assistants at Master's level carried out direct face-to-face interviews with patients at 10 years and extracted monthly data using medical records and the Health Authority Clinical Management System over the entire follow-up period. Raters were blinded to the randomized trial assignment and the follow-up status of the participants.
Relapse was defined as the re-emergence or exacerbation of positive symptoms, as operationalized by a change from CGI positive (Haro et al., Reference Haro, Kamath, Ochoa, Novick, Rele, Fargas, Rodriguez, Rele, Orta, Kharbeng, Araya, Gervin, Alonso, Mavreas, Lavrentzou, Liontos, Gregor and Jones2003) scores <3 for at least 3 consecutive months to a score of ⩾3 (mild or more severe) (Haro et al., Reference Haro, Novick, Bertsch, Karagianis, Dossenbach and Jones2011; Chan et al., Reference Chan, So, Hui, Chang, Lee, Chung, Tso, Hung, Yip, Dunn and Chen2015). Relapse was assessed monthly using medical record review ratings of CGI positive symptom severity scores from the start of the RCT until the end of the 10 years in all patients. The CGI-positive has a rating from 1 (normal), 2 (borderline), 3 (mild), to 7 (most severely ill). For each relapse episode, we recorded the start and end dates (duration), and whether the event required hospitalization. Weekly consensus meetings were conducted among a clinician and research assistants during the data collection period for quality assurance and for resolving ambiguity in ratings. To ensure consistency in ratings, the CGI-positive was rated from eight independent medical records (i.e. not those cases recruited into the current study). Good agreement was found among the three raters, with an intra-class correlation of 0.7.
To validate the definition of relapse using CGI-positive, we compared the concordance between the relapses derived using CGI-positive with those using the ‘remission/relapse’ definition in the abovementioned inclusion criteria. The latter was operationalized as meeting the following criteria using medical record reviews: (1) at least one of the following in the PANSS scale: delusions >3, conceptual disorganization >4, hallucinations >3, suspiciousness >5, or unusual thought content >4, and (2) scores >3 and >5 in the CGI severity of symptoms and CGI improvement scales respectively. There was an excellent agreement between the two relapse definitions (κ = 0.842, p < 0.001).
Instead of targeting a few specific positive symptoms in defining relapse (as is the case in PANSS), the current CGI-positive definition focuses on the overall severity of the patient's positive symptoms at the time of assessment. This approach is more relevant and applicable to data extraction using medical record reviews, given that not all clinicians provide detailed psychopathology, but instead provide an overall description of the psychotic symptoms of patients. In addition, compared with retrospective recall of relapses from patients over the past 10 years (Alverez-Jimenez et al., Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011), the current medical record review of positive symptoms of relapse using CGI-positive would minimize the chance of underestimation of any relapse episodes, and ensures that relapse of positive symptoms to a mild or more severe level are all included.
Other outcome measures included information on marital status, diagnosis, antipsychotic treatment, and medication adherence. Positive and negative symptoms were assessed using the PANSS, the Scale for the Assessment of Positive Symptom (SAPS; Andreasen, Reference Andreasen1984), and the Scale for Assessment of Negative Symptoms (SANS; Andreasen, Reference Andreasen1983). Depressive symptoms were assessed using the Calgary Depression Scale for Schizophrenia (CDSS; Addington et al., Reference Addington, Addington, Maticka-Tyndale and Joyce1992). Insight into the illness was measured by the abridged Scale to Assess Unawareness of Mental Disorder (Amador et al., Reference Amador, Flaum, Andreasen, Strauss, Yale, Clark and Gorman1994). Side effect was assessed with the Simpson-Angus Scale (SAS; Simpson and Angus, Reference Simpson and Angus1970), the Abnormal Involuntary Movement Scale (AIMS; Guy, Reference Guy and Guy1976a, Reference Guy and Guy1976b), the Barnes Akathisia Rating Scale (BARNS; Barnes, Reference Barnes1989), and the Udvalg for Kliniske Undersøgelser (UKU; Lingjaerde et al., Reference Lingjaerde, Ahlfors, Bech, Dencker and Elgen1987). Functioning was assessed using the Social and Occupational Functioning Assessment Scale (SOFAS; Goldman et al., Reference Goldman, Skodol and Lave1992) and the Strauss and Carpenters’ scale (Strauss and Carpenter, Reference Strauss and Carpenter1972). Direct interviews were performed with 142 out of the 178 patients at 10 years.
Baseline predictor measures
All potential baseline predictors were evaluated at entry to the RCT, where all patients had complete resolution of positive symptoms for at least 1 year following their first episode of psychosis. Baseline variables included gender, age, years of education, employment, marital status, diagnosis (schizophrenia spectrum disorders: schizophrenia, schizophreniform disorder and schizoaffective disorder v. non-schizophrenia spectrum disorders: brief psychotic disorder and psychosis not otherwise specified), and treatment received during the RCT (quetiapine v. placebo).
Premorbid functioning was assessed using the Premorbid Adjustment Scale (Cannon-Spoor et al., Reference Cannon-Spoor, Potkin and Wyatt1982). The same symptoms, insight and functioning assessments (see above in outcome measures) were used. DUP was assessed using the Interview for the Retrospective Assessment of the Onset of Schizophrenia (IRAOS; Häfner et al., Reference Hafner, Riecher-Rossler, Hambrecht, Maurer, Meissner, Schmidtke, Fatkenheuer, Loffler and van der Heiden1992).
Neurocognitive functions assessed were as follows: Information, Arithmetic, Digit Span (forward and backward), Digit Symbol, Block Design, Trail Making Test (response time difference between the two tasks), Letter Number Span (the highest level attained), Logical Memory Test (total number of correct immediate and delayed recalls), Visual Patterns Test (the highest level attained), Semantic Fluency Test (total number of correct animals reported in one minute), and the Modified Wisconsin Card Sorting Test (total number of perseveration errors). The total number of times the patient blinked during two minutes of relaxation was also recorded.
Statistical analysis
All statistical analyses were carried out using IBM® SPSS® Version 24.0. To examine the potential baseline predictors (independent variables) for patients who never relapsed over the 10 years (dependent variable: never relapsed = 1, relapsed = 0), univariate binary logistic regression analysis was used. Univariate variables with a p value of ⩽0.1 were identified. To avoid multicollinearity, only one score for each performance test was chosen; for example, in the Logical Memory Test, immediate memory recall was used instead of both immediate and 30-min delayed recall because the two are highly correlated. The remaining identified predictors were entered into a multivariate binary logistic regression model using forward selection with a p value of ⩽0.05 indicating significance.
Among patients who completed the face-to-face interview assessments after 10 years, the outcome correlates of whether the patient relapsed over the 10 years were explored using the independent t test for parametric continuous variables, the χ2 test for categorical variables, and the Mann–Whitney U test for non-parametric continuous variables. To handle the problem of multiple testing, the false discovery rate (FDR) (q-value) of 10% with the Benjamini–Hochberg procedure was used.
Results
Basic demographics
The study cohort was followed up for a median of 9.4 years (interquartile range, IQR 8.5–10.4) since the start of the RCT, and a median of 11.3 years (IQR 10.3–12.2) since the patient's first episode of psychosis. Of the 178 patients in the original RCT, 142 (80%) were successfully traced and interviewed, 28 (16%) declined assessment, 6 (3%) committed suicide, and 2 (1%) were unable to be contacted.
Table 1 shows the baseline (i.e. entry into the RCT) demographics and treatment characteristics during the follow-up of all patients. Forty-five percent of patients were male. They had a mean age of 24.2 years and received education for a mean of 11.8 years. The majority of them (70%) were employed. There were no statistical baseline differences between those who relapsed and never relapsed in terms of gender, age, years of education, employment, marital status, and whether they received maintenance treatment or placebo during the randomized trial. The antipsychotics medication received over the follow-up period were similar in relapsers and non-relapsers.
Table 1. Basic demographics and treatment characteristics for non-relapsers, relapsers, and all patients
a Data were available for 173 patients: 37 non-relapsers and 136 relapsers.
b Antipsychotics medication received over 10 years following the start of randomized trial.
c The mean daily dose of each antipsychotic was converted to chlorpromazine equivalent dose.
Relapse over the 10 years
Over the 10 years, 37 of 178 (21%) patients never relapsed following complete resolution of their positive symptoms from a first episode. In other words, relapse was observed in 141 of 178 (79%) patients, where 68 of 141 (48%) experienced one relapse, 44 of 141 (31%) experienced two relapses, 16 of 141 (11%) experienced three relapses, and the remaining 13 of 141 (9%) experienced four or more relapses. The mean number of relapse episodes was 1.5 (s.d. = 1.4, range 0–11).
Among the 141 relapsers, the mean aggregate time spent in the first relapse was 5.4 (s.d. = 12.6) months, and 55 (39%) of the relapsed patients required hospitalization. As for the timing of relapse, 85 of 141 (60%) patients relapsed in the first year after complete resolution of positive symptoms, 24 (17%) in the second year, 12 (6%) in the third year, 7 (5%) in the fourth year, and 4 (3%) in the fifth year.
Univariate predictors of never relapsing
Table 2 shows the univariate predictors of relapse at 10 years which had a p value of ⩽0.1. Patients who never relapsed were less likely to be diagnosed with schizophrenia spectrum disorders, more likely to have a DUP ⩽30 days, had less severe negative symptoms at baseline (especially blunted affect), had better performance in the Logical Memory (immediate and delayed recall) and Verbal Fluency Tests at baseline than those who relapsed. Other baseline predictors explored, including basic demographics, treatment received during the RCT (quetiapine or placebo), DUP, premorbid and baseline functioning, were not statistically significant in predicting relapse.
Table 2. Baseline predictors of never relapsing over the 10 years in univariate binary logistic regression analyses (p value ⩽0.1)
CI, confidence interval; DUP, duration of untreated psychosis; PANSS, Positive and Negative Syndrome Scale; SANS, Scale for Assessment of Negative Symptoms.
a Unless otherwise specified, values represent means (standard deviation). Number of missing observations are shown in brackets.
b Diagnosis was categorized into schizophrenia spectrum (including schizophrenia, schizophreniform disorder, and schizoaffective disorder [=1]) and non-schizophrenia spectrum (including brief psychotic disorder and psychosis not otherwise specified [=, reference category)).
c DUP was classified into ⩽30 days (=1) and >30 days (=2, reference category).
To avoid the problem of multicollinearity, the single variable with the smaller p value in each assessment was selected for inclusion into the multivariate regression model. They included baseline diagnosis, DUP ⩽30 days, SANS sum of all items, logical memory (immediate recall), and verbal fluency (correct recall).
Multivariate predictors of never relapsing
The multivariate logistic regression analysis suggested that patients who did not relapse over the 10 years were less likely to be diagnosed with schizophrenia spectrum disorders (Odds ratio, OR = 0.23, 95% CI 0.13 to 0.90, p = 0.030), more likely to have a DUP ⩽30 days (OR = 4.60, 95% CI 1.65 to 12.84, p = 0.004), and performed better in the logical memory immediate recall at baseline (OR = 1.10, 95% CI 1.00 to 1.22, p = 0.050) (Table 3). The model explained 26.9% (Nagelkerke R2) of the variance, with an overall correct classification of 81.8%.
Table 3. Significant predictors of never relapsing over the 10 years in multivariate binary logistic regression analysis (n = 137)
CI, confidence interval; DUP, duration of untreated psychosis.
a DUP was classified into ⩽30 days (=1) and >30 days (=2, reference category).
Outcome correlates of never relapsing
As shown in Table 4, more non-relapsers were married/in a stable relationship and diagnosed with non-schizophrenia spectrum disorders at 10 years. They took a lower mean dose of daily antipsychotics chlorpromazine equivalent over the follow-up period. More of them were able to stop taking antipsychotics in the 2 years prior to follow up, reported having fewer medication side effects (BARNS, UKU neurologic, and UKU others), and had a lower BMI than those who relapsed. Non-relapsers also had better social and occupational functioning at 10 years (higher SOFAS score, higher SCS score, and had more months in open employment over the 2 years prior to follow up). All these findings remained significant at a FDR rate (q value) of 10%. When the FDR rate was set to 5%, the significant variables included medication discontinuation in the past 2 years, diagnosis with schizophrenia spectrum disorders, UKU neurologic side effect, BARNS side effect, BMI, and being married/in a stable relationship.
Table 4. Outcome measures for non-relapsers, relapsers, and all patients
PANSS, Positive and Negative Syndrome Scale; CDSS, Calgary Depression Scale for Schizophrenia; SAS, Simpson-Angus Scale; AIMS, Abnormal Involuntary Movement Scale; BARNS, Barnes Akathisia Rating Scale; UKU, Udvalg for Kliniske Undersøgelser; BMI, Body Mass Index; SOFAS, Social and Occupational Functioning Assessment Scale.
a Unless otherwise specified, values represent means (standard deviation). Fishers exact test was used for cells that have count <5.
b Data were available in patients who had an end-of-follow-up assessment: 30 non-relapsers, 111 relapsers, and 141 patients overall.
c Schizophrenia spectrum disorders (schizophrenia, schizophreniform disorder, schizoaffective disorder) v. non-schizophrenia spectrum (including psychosis not otherwise specified, brief psychotic disorder, delusional disorder).
d The mean daily dose of each antipsychotic was converted to a chlorpromazine equivalent dose. Data were available in patients who had an end-of-follow-up assessment: 21, 112, and 133 patients.
e Medication compliance was measured using the modified Adherence Rating Scale. Higher scores indicate better adherence behavior. Data were available in patients who had an end-of-follow-up assessment: 20, 94, and 114 patients.
f Medication attitude was measured using the modified Adherence Rating Scale. Higher scores indicate a more positive attitude. Data were available in patients who had an end-of-follow-up assessment: 22, 94, and 116 patients.
g Data were available in patients who had an end-of-follow-up assessment: 30, 111, and 141 patients.
h Data were available in patients who had an end-of-follow-up assessment: 24, 106, and 130 patients.
i Data were available in patients who had an end-of-follow-up assessment: 24, 105, and 129 patients.
j Data were available in patients who had an end-of-follow-up assessment: 24, 105, and 129 patients.
k Data were available in patients who had an end-of-follow-up assessment: 17, 102, and 119 patients.
l BMI is the weight in kilograms divided by the square of the height in meters. Data were available in patients who had an end-of-follow-up assessment: 28, 103, and 131.
m Data were available in patients who had an end-of-follow-up assessment: 30, 106, and 136 patients.
n Data were available in patients who had an end-of-follow-up assessment: 28, 107, and 135 patients.
Discussion
To date, this is the first 10-year follow-up study examining the proportion and predictors of patients who did not experience a psychotic relapse in the 10-year period following complete resolution of positive symptom from FEP. Our data suggested that the proportion of relapse-free patients after 10 years was 21%. This subgroup of relapse-free patients was characterized by being less likely to be diagnosed with a schizophrenia spectrum disorder, having a DUP ⩽30 days, and having better short-term verbal memory following their first episode. These observed differences between relapsers and non-relapsers were unlikely to be a result of the difference in antipsychotic medications received over the 10 years. Such findings not only provide an updated perspective on the characterization of patients who did not relapse after 10 years but also expand the predictors of not relapsing to include a neurocognitive dimension. Identification of potentially modifiable predictors of not relapsing would facilitate timely intervention at the early stages of the illness and thus produce a better long-term outcome.
Relapse rates over the 10 years
Consistent with previous longitudinal studies (Shepherd et al., Reference Shepherd, Watt, Falloon and Smeeton1989; Linszen et al., Reference Linszen, Dingemans and Lenior2001), the proportion of relapse-free patients in this study was 21%. This was slightly higher than the rate of 16.8% reported in a different 7.5-year follow-up study of FEP (Alverez-Jimenez et al., Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011). We believe the discrepancy could well be reflecting the differences in the study populations, as the previous study included FEP patients under a naturalistic setting, whereas the current study focused on FEP patients who were entirely free from any psychotic symptoms and had received maintenance treatment for at least 1 year under a controlled study setting. This might explain why our good prognostic first-episode cohort would have a slightly higher proportion of relapse-free patients after 10 years.
It was interesting to note that in a similar follow-up study of FEP patients who had been randomized to either dose reduction or maintenance treatment for 18 months following stabilization, the proportion of relapse-free patients at 7 years was up to 35% (Wunderink et al., Reference Wunderink, Nieboer, Wiersma, Sytema and Nienhuis2013). One reason for this higher rate of relapse-free patients in the Dutch study may be related to their use of a dose reduction arm instead of complete medication discontinuation (as in our case) during the initial clinical trial. The fact that a lower relapse rate in the initial trial was found in the Dutch study (43% in the dose reduction group; Wunderink et al., Reference Wunderink, Nienhuis, Sytema, Slooff, Knegtering and Wiersma2007) when compared with the Hong Kong study (79% in the discontinuation group; Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010), could explain the higher proportion of patients in the Dutch study who did not subsequently relapse in the long term.
Neurocognition and relapse
More research has begun to explore the link between neurocognitive dysfunction and relapse in psychosis, as the neurocognitive function is a potentially modifiable factor. Shorter-term follow-up studies have identified baseline memory function (Verdoux et al., Reference Verdoux, Lengronne, Liraud, Gonzales, Assens, Abalan and van Os2000), delayed visual reproduction (Verdoux et al., Reference Verdoux, Lengronne, Liraud, Gonzales, Assens, Abalan and van Os2000), Wisconsin Card Sorting Test perseverative errors (Chen et al., Reference Chen, Hui, Dunn, Miao, Yeung, Wong, Chan and Tang2005b), Trail Making Test-B (Wölwer et al., Reference Wölwer, Brinkmeyer, Riesbeck, Freimüller, Klimke, Wagner, Möller, Klingberg and Gaebel2008), working memory, and verbal learning (Rund et al., Reference Rund, Melle, Friis, Johannessen, Larsen, Midbøe, Opjordsmoen, Simonsen, Vaglum and McGlashan2007) to be predictive of relapse in psychosis. More recently, visual working memory deterioration was found to occur preceding a psychotic relapse (Hui et al., Reference Hui, Li, Li, Lee, Chang, Chan, Lam, Thornton, Sham, Honer and Chen2016).
Here, we presented a novel finding that patients with better short-term verbal memory (immediate logical memory) at baseline were more likely to be relapse-free after 10 years. Similarly, Rund et al. (Reference Rund, Barder, Evensen, Haahr, Hegelstad, Joa, Johannessen, Langeveld, Larsen, Melle, Opyjordsmoen, Rossberg, Simonsen, Sundet, Vaglum, McGlashan and Friis2016) have found that remitted patients who did not relapse during the first year had a better neurocognitive function at 10 years than those with unstable remission or continuously psychotic during the first year, implying a link between early relapse and long-term cognitive outcome. Our data further clarified that early neurocognitive functioning is also associated with subsequent relapse status after 10 years.
DUP and relapse
DUP is an important marker for the delay in first effective psychiatric treatment. The identification of DUP as a predictor is of high clinical interest due to its ability to improve clinical outcome and reduce costs, since considerable long-term studies (with follow-up durations >7.5 years) have confirmed the negative impact of a long DUP on overall symptomatic outcome (Thara and Eaton, Reference Thara and Eaton1996; Bottlender et al., Reference Bottlender, Sato, Jager, Wegener, Wittmann, Strauss and Moller2003; White et al., Reference White, Stirling, Hopkins, Morris, Montague, Tantam and Lewis2009; Tang et al., Reference Tang, Chang, Hui, Wong, Chan, Lee, Yeung, Wong, Tang, Chan, Pang, Tso, Ng, Hung, Dunn, Sham and Chen2014). In this study, the status of non-relapse at 10 years was predicted by DUP ⩽30 days. This is consistent with previous short-term follow-up studies (Crow et al., Reference Crow, MacMillan, Johnson and Johnstone1986; Larsen et al., Reference Larsen, Moe, Vibe-Hansen and Johannessen2000), as well as the longer-term follow-up study of FEP (Alverez-Jimenez et al., Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011).
Given how there is no standardized method of handling the positively skewed DUP data, the adoption of the threshold value of DUP ⩽30 days for categorizing subgroups made no assumption about the linear relationship between DUP and outcome. The threshold was also based on our local study, which showed that a DUP as short as 30 days is sufficient to demonstrate a significant negative impact on the remission outcome at 13 years in Chinese patients with FEP (Tang et al., Reference Tang, Chang, Hui, Wong, Chan, Lee, Yeung, Wong, Tang, Chan, Pang, Tso, Ng, Hung, Dunn, Sham and Chen2014).
Schizophrenia spectrum disorders and relapse
Although it is widely known that a diagnosis of schizophrenia (as opposed to other psychoses) is associated with poorer outcome, the exact relationship between a diagnosis of schizophrenia and the short-term outcome of relapse remains contentious: for example, while one study found that schizophrenia patients who discontinued medication experience more relapses (Hui et al., Reference Hui, Wong, Tang, Chang, Chan, Lee, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Hung, Honer and Chen2013), other short-term follow-up studies have not found such a relationship (Crow et al., Reference Crow, MacMillan, Johnson and Johnstone1986; Lenior et al., Reference Lenior, Dingemans, Schene and Linszen2005). On the other hand, the available evidence for a diagnosis of schizophrenia and the long-term outcome of relapse appears more consistent. Alverez-Jimenez et al. (Reference Alverez-Jimenez, Gleeson, Henry, Harrigan, Harris, Amminger, Killackey, Yung, Herrman, Jackson and McGorry2011) found schizophrenia spectrum disorders (including schizophrenia, schizophreniform, schizoaffective, delusional, psychotic disorder not otherwise specified, and brief psychotic disorders) to be a significant univariate predictor of relapse at 7.5 years in FEP. In our 10-year study, we also found that those diagnosed at baseline with schizophrenia spectrum disorders (including schizophrenia, schizoaffective disorder, and schizophreniform disorder) were more likely to predict relapse.
Given that a non-schizophrenia diagnosis and a DUP ⩽30 days were found to be significant multivariate predictors of not relapsing, one might ask if these relapse-free patients were composed of individuals diagnosed with brief psychotic disorders in the first place. Out of 37 relapse-free patients, 18 (49%) were diagnosed with schizophrenia spectrum disorders, 10 (27%) diagnosed with psychosis not otherwise specified, and the remaining 9 (24%) diagnosed with the brief psychotic disorder. The postulation that relapse-free patients were simply brief psychotic disorder cases with a shorter DUP is therefore unlikely.
Limitations
Owing to the restrictive inclusion criteria used (must have complete resolution of positive symptoms, good compliance, and been on maintenance medication for at least 1 year after stabilization), the employment rate for our cohort at baseline was up to 70%. Therefore, our cohort may not have been representative of all FEP patients. Given that the cohort must have had good medication compliance to enter the study before randomization to either maintenance treatment or placebo, it is not possible to explore the effect of medication non-compliance at baseline on relapse after 10 years. However, exploratory analyses on the initially five yearly medication compliance data extracted from medical records did not predict relapse. Further, the existing medication compliance data do not allow for the differentiation between relapses that occur in patients receiving ongoing treatment and those relapsing following medication discontinuation over the 10 years. A more rigorous adherence assessment to identify the temporal relationship to the relapse events would be helpful. We did not include measures on expressed emotion and stress, which may be relevant for relapse prediction and long-term prognosis of first episode psychosis. In addition, as the prevalence of substance misuse is relatively low in the Hong Kong population, those with significant alcohol or other substance misuse in the past 3 months were excluded. Such exclusion enabled a more homogeneous sample but leaves results incomparable to those of Western cohorts. It should also be noted that relapse in this study referred to the re-emergence or exacerbation of positive symptoms to a mild or more severe level, which may be different from the rest of the literature. Although whether patients were randomized to quetiapine/placebo during the RCT was not significantly different between relapsers and non-relapsers, we cannot entirely discard the effect of randomization on subsequent relapse after 10 years.
Clinical and research implications
The existing prediction model had 27% of the variance explained. Testable and modifiable predictors of relapse such as DUP and cognitive function would be an important addition to the long-term management of patients with psychosis and could lead to effective strategies for relapse prevention. In a study of Chinese FEP patients in Hong Kong, previous family experiences and knowledge about psychiatric illnesses were found to be important factors related to a shorter DUP (Chen et al., Reference Chen, Dunn, Miao, Yeung, Wong, Chan, Chen, Chung and Tang2005a, Reference Chen, Hui, Dunn, Miao, Yeung, Wong, Chan and Tang2005b). This is particularly relevant to our Chinese patients because a predominant proportion of them live with their families, and the decision to seek help heavily relies on immediate family members. Early detection and psychoeducation programs could target family members to increase their awareness of signs related to the onset of psychosis during the prodromal period, thereby shortening the delay in receiving timely psychiatric treatment.
The identification of neurocognitive function as a marker for a long-term clinical outcome such as relapse is important, because neurocognitive markers are clearly defined, relatively easy to administer, reproducible, and can objectively be measured. There is now a growing focus in improving the outcome of psychosis by cognitive enhancement through remediation and other non-invasive approaches. More research is needed to replicate this finding before this marker is used as a screening tool for relapse in actual clinical practice.
Finally, we found that never relapsing during the 10-year period was associated with various clinical correlates, including an increased likelihood of being able to stop taking medication in the 2 years prior to follow-up, being married/in a stable relationship, having fewer medication side effects, and having a lower BMI. Shortening the delay in receiving timely psychiatric treatment and treatment for cognitive symptoms, especially among patients with schizophrenia spectrum disorders, thus appears to be important for improving the long-term clinical outcome of relapse.
Author ORCIDs
Christy LM Hui http://orcid.org/0000-0001-6912-208X
Acknowledgements
The randomized-treatment phase of the study was supported by the Research Grants Council of Hong Kong (7655/05M), and AstraZeneca (investigator-initiated study award). AstraZeneca prepared the quetiapine and the placebo, packaged the study medications according to the randomization schedule. The follow-up study was supported by the Food and Health Bureau of Hong Kong (10111101). WGH was supported by the Jack Bell Chair in Schizophrenia.
Conflict of interest
WGH received consultation fees from Otsuka/Lundbeck, AphaSights and Eli Lilly. WTLL participated as a paid consultant for Jansen. EYHC received speaker honoraria from Otsuka and DSK BioPharma; received research funding from Otsuka; participated in paid advisory boards for Jansen and DSK BioPharma; received funding to attend conferences from Otsuka and DSK BioPharma. The remaining authors declare no competing interests.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
