A critical analysis of recent data on the long-term outcome of antipsychotic treatment

Long-term antipsychotic maintenance treatment has been the goal of treatment for schizophrenia for decades based on short-term studies that show the risk of relapse is elevated after abrupt antipsychotic discontinuation. Most psychiatrists assumed that by reducing risk of relapse one would be improving long-term outcome. However, some evidence suggests that long-term outcomes of antipsychotic discontinuation are not necessarily comparable to short-term outcomes (Harrow et al., Reference Harrow, Jobe and Faull2012; Morgan et al., Reference Morgan, Lappin, Heslin, Donoghue, Lomas, Reininghaus, Onyejiaka, Croudace, Jones, Murray, Fearon, Doody and Dazzan2014; Moilanen et al., Reference Moilanen, Haapea, Jaaskelainen, Veijola, Isohanni, Koponen and Miettunene2016; Wils et al., Reference Wils, Gotfredsen, Hjorthoj, Austin, Albert, Secher, Thorup, Mors and Nordentoft2017). Together with increasing evidence of drug-related adverse effects including structural brain changes, this has led some commentators to suggest that we should question the presumption of maintenance therapy for most people with psychotic disorders (Murray et al., Reference Murray, Quattrone, Natesan, van Os, Nordentoft, Howes, Di Forti and Taylor2016).

For people who have had a single episode of psychosis, current guidelines recommend that antipsychotics should be continued for a period of at least 1–2 years (National Institute for Health and Care Excellence, 2014), the idea being that if a person remains well during that time, it might then be safe to stop the drug. Many services now offer patients the option of reducing and stopping antipsychotic treatment after a variable period of maintenance treatment.

Two recent studies appear to challenge this tendency and the authors reassert the notion that long-term antipsychotic treatment should be the goal for the majority of people who experience psychosis, even after one episode. Recent reviews by eminent schizophrenia researchers reached the same conclusions (Goff et al., Reference Goff, Falkai, Fleischhacker, Girgis, Kahn, Uchida, Zhao and Lieberman2017; Correll et al., Reference Correll, Rubio and Kane2018). The new studies are important because they are some of the few that have gathered information over extended periods of time; therefore, they merit close examination

Ten-year follow-up of antipsychotic maintenance trial

One of these recent studies consists of a 10-year follow-up of participants who entered a randomised trial of quetiapine v. placebo maintenance treatment following a first episode of psychosis (Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen, Pang, Lui, Chung, Yeung, Ng, Lo, Jones, Sham and Chen2018). The original trial, published in 2010, was conducted in Hong Kong and involved a sample of 178 patients (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010). They were followed up for a year initially or until the first instance of relapse. Relapse was defined as the reappearance of one of five pre-specified Positive and Negative Syndrome Scale (PANSS) psychotic symptoms at threshold scores of between 3 (mild) and 5 (moderately severe) along with a Clinical Global Impressions-Severity score of 3 (mildly ill) or above and a Clinical Global Impressions-Improvement score of 5 (minimally worse) or above. This definition was described by the authors as ‘more liberal’ than other studies and similar to some definitions of ‘deterioration’ (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010, p. 3).

The initial results showed a risk of relapse of 79% in the group randomised to placebo, compared with 41% for the group allocated to quetiapine maintenance. Rates of hospitalisation were 16% and 6%, respectively (Chen et al., Reference Chen, Hui, Lam, Chiu, Law, Chung, Tso, Pang, Chan, Wong, Mo, Chan, Yao, Hung and Honer2010).

Ten-year follow-up results compared the proportions of patients who had a ‘poor’ outcome, defined as a composite of persistent psychotic symptoms, being on clozapine, or death by suicide (Hui et al., Reference Hui, Honer, Lee, Chang, Chan, Chen, Pang, Lui, Chung, Yeung, Ng, Lo, Jones, Sham and Chen2018). Persistent psychotic symptoms were defined in the same way as they were for relapse in the initial analysis – in other words they could be judged as ‘mild’. Researchers attempted to do a face-to-face follow-up interview with all the participants, but could only achieve this with 142 of the original sample. The analysis also included data for 28 patients (16%) who were not interviewed at 10 years by using the last PANSS score obtained during the initial trial follow-up.

Using these definitions, a poor outcome was present in 21% of the group originally allocated to quetiapine maintenance and 39% of the placebo group (p = 0.012). However, looking at the components of the composite outcome in more detail reveals that most of those who were classified as having poor outcome were so defined by virtue of reaching threshold scores on the pre-specified psychotic symptoms. Most of these involved P1 (delusions) where the threshold criteria was 3 (mild). Inspecting other outcomes shows no difference in the overall PANSS score, the positive, negative and general subscales of the PANSS, rates of being in work and social functioning as measured by the Social and Occupational Functioning Assessment Scale (SOFAS). Slightly more people died by suicide or were prescribed clozapine in the group originally randomised to antipsychotic discontinuation, but numbers were small. It appears that the composite outcome may therefore have been constituted out of criteria selected post hoc on the basis of their differential distribution between the groups.

Overall this study indicates that there was little difference in outcome between the groups 10 years after their enrolment in the antipsychotic discontinuation trial. This is not surprising given that the median duration of the original trial was only 145 days for the quetiapine group and 106 for the placebo group. Thus although it was described as a 12-month trial, in fact follow-up for most participants was much shorter, due to withdrawal or ‘relapse’. Moreover, the inclusion of data from the original trial for some participants means the follow-up results do not reflect 10-year outcomes for all participants, and means the results presented are likely to mirror the results of the original trial.

Retrospective database study

The second study was a cohort study using routinely collected data from national registers concerning people who had a first admission for schizophrenia in Finland (Tiihonen et al., Reference Tiihonen, Tanskanen and Taipale2018). The main outcome was a composite variable called ‘treatment failure’ which consisted of a combination of rehospitalisation or death. Treatment failure was least common among continuous users of antipsychotics compared with non-users or people who discontinued. People who never started antipsychotics or who discontinued early were less likely to show treatment failure than those who took medication for longer periods and then stopped.

Again, the composite outcome, as well as lack of information about cause of death, makes the primary results difficult to interpret. Moreover, the observational nature of the data, and lack of matching or control for potential confounders makes it difficult to draw any causal inferences about the role of antipsychotic treatment.

It is recognised that observational data can be misleading because interventions are not allocated at random, even producing results that contradict those of randomised controlled trials (Freemantle et al., Reference Freemantle, Marston, Walters, Wood, Reynolds and Petersen2013). Careful matching or controlling potential confounders in the analysis may improve the reliability of the findings. In the present study, people were matched on the length of time they had been taking antipsychotics and age, sex and length of initial hospitalisation were adjusted for in the subsequent analysis. In a previous analysis of mortality data by the same group, the analysis also controlled for previous suicide attempt, cancer and ischaemic heart disease (Tiihonen et al., Reference Tiihonen, Lonnqvist, Wahlbeck, Klaukka, Niskanen, Tanskanen and Haukka2009). The authors were still criticised for not including other variables that might have important effects on mortality such as socio-economic status, substance misuse, indicators of unhealthy lifestyle and detailed cardiovascular risk factors (De Hert et al., Reference De Hert, Correll and Cohen2010).

Evidence shows that people who are compliant with any treatment, including placebo, have better outcomes than those who are not across a range of conditions. People who stick to recommended treatment may, for example, be more physically or mentally healthy to begin with, have more stable lifestyles, be more likely to engage with other aspects of treatment and they may have more social support. In the Women's Health Initiative, for example, women who showed high compliance with placebo had lower rates of hip fractures, heart attacks, cancer death and death from all causes compared with those who showed poor placebo compliance (Curtis et al., Reference Curtis, Larson, Delzell, Brookhart, Cadarette, Chlebowski, Judd, Safford, Solomon and Lacroix2011).

Clinicians’ decisions to prescribe or to stop medication are also not random. In particular, clinicians may avoid prescribing a drug with known physical complications to someone who already has risk factors for those conditions. For example, evidence shows that clinicians prescribe a strongly weight-inducing drug like olanzapine to people with lower initial body weight than people to whom they prescribe other antipsychotics, such that olanzapine can be associated with lower body weight in observational data (Osborn et al., Reference Osborn, Petersen, Beckley, Walters, Nazareth and Hayes2018). Reverse causality is also possible, since people who develop life-threatening conditions may have their prescription stopped subsequent to the onset of the illness but prior to actual death.

It is particularly difficult to interpret findings on mortality without a breakdown by cause of death. A similar registry data study conducted in Sweden showed that the majority of deaths, which were due to cardiovascular disease, showed a U-shaped relationship with antipsychotic exposure. People who were non-users had high rates of death, but there was a dose–response relationship between antipsychotic exposure and death for those who had low, moderate and high exposure. The high mortality among non-users also held for cancer, suggesting that non-users had higher underlying risk factors for cardiovascular mortality than users (Torniainen et al., Reference Torniainen, Mittendorfer-Rutz, Tanskanen, Bjorkenstam, Suvisaari, Alexanderson and Tiihonen2015).

In the current study, figures presented in the supplemental information suggest that people who never used antipsychotics after discharge had the highest number of deaths, followed by those who discontinued antipsychotics within 1 year. Those who discontinued antipsychotics after longer periods did not, in fact, show higher rates of death compared with those who continued antipsychotics, although numbers of deaths for these groups were small. However, the figures exclude deaths that occurred after 24 h of hospitalisation (to any hospital – personal communication). Since most people die in hospital, the data must considerably underestimate deaths in general, and possibly deaths in antipsychotic users in particular, as adherence is related to engagement with services which may make admission for a physical health event more likely.

As far as readmissions are concerned, the unmatched comparison (no matched group is provided for this comparison) found the risk of rehospitalisation was only slightly higher among non-users compared with continuous users of antipsychotic medication (hazard ratio 1.24, confidence interval 1.18–1.30).

In the comparison of ‘matched’ groups of continuous users and discontinuers, the risk of rehospitalisation increased with longer duration of treatment prior to discontinuation. The analysis consisted of a ‘survival’ analysis and follow-up ended at the time of rehospitalisation or when people switched antipsychotic status, i.e. when discontinuers started using antipsychotics again, or when users stopped taking them. In fact the numbers and proportions of people rehospitalised were comparable between continuous users and discontinuers, but length of follow-up was shorter with people who discontinued antipsychotics being followed up for a median of between 122 and 134 days compared with 376 for continuous users. What the analysis shows, therefore, is that readmissions occurred earlier in people who had discontinued antipsychotics, but not necessarily that they were more common. The result is consistent with other evidence showing that stopping antipsychotics brings forward risk of relapse but may not influence risk in the long-run (Wunderink et al., Reference Wunderink, Nieboer, Wiersma, Sytema and Nienhuis2013).

The focus on short-term events may explain the discrepancy between the present study and the findings of long-term cohort studies, which suggest that people who avoid long-term continuous antipsychotic treatment have better outcomes, including lower rates of relapse and higher levels of functioning and recovery, compared with continuous users (Harrow et al., Reference Harrow, Jobe and Faull2012; Morgan et al., Reference Morgan, Lappin, Heslin, Donoghue, Lomas, Reininghaus, Onyejiaka, Croudace, Jones, Murray, Fearon, Doody and Dazzan2014; Moilanen et al., Reference Moilanen, Haapea, Jaaskelainen, Veijola, Isohanni, Koponen and Miettunene2016; Wils et al., Reference Wils, Gotfredsen, Hjorthoj, Austin, Albert, Secher, Thorup, Mors and Nordentoft2017). These studies are also subject to the biases that arise because treatment is not randomly allocated, but they do provide face-to-face follow-up over long periods with all participants who can be located.

The benefits that non-users of antipsychotics showed in one cohort study that tracked progress every few years only started to become apparent at the 4-year follow-up point (Harrow et al., Reference Harrow, Jobe and Faull2012). Although the present study is reported as a 20-year follow-up, this is the maximum duration, but the median follow-up was only 17–19 weeks for those who discontinued antipsychotics, 54 weeks for continuous users and 58 weeks for non-users (408 days).

Reverse causality may also operate in relation to the outcome of rehospitalisation in circumstances when a deterioration in mental state leads to discontinuation of antipsychotics that pre-dates hospitalisation. This is especially likely given that most instances of antipsychotic discontinuation in this study are likely to have been unsupervised.

Given the caveats about observational studies, the present study may indicate an increased risk of readmission to hospital following antipsychotic discontinuation, but it is not a study of a controlled and gradual reduction of antipsychotics done with professional support. It also provides no data on other outcomes such as quality of life and social functioning or on the burden of drug-related adverse effects that patients can suffer.