Can we raise the standard of care in treating schizophrenia: A cautionary note!

In an interesting and at times controversial paper (Catts and O’Toole, 2016), the authors examine an important and often overlooked area, namely, the progression of clinical and neurobiological changes in schizophrenia and how they may relate to the emergence of treatment-resistant disorder. The authors explore what underlies such changes and potential approaches to ameliorate them. We examine each of the issues considered, and raise alternative views on some of their conclusions, using the well-accepted hierarchy of evidence pyramid, which gives the highest weighting to systematic review data of randomised controlled trials (RCT) (www.nhmrc.gov.au/_files_nhmrc/file/guidelines/developers/nhmrc_levels_grades_evidence_120423.pdf).

The authors introduce the topic by using mortality rates (MR) as a measure of progress or lack of progress in the treatment of schizophrenia and infer that the outcomes of treatment are deteriorating over time. The authors note the increasing MR over time in schizophrenia as an indication that existing interventions have failed. The issue of MR is a complex one and can be divided into natural and unnatural causes. Natural causes include cardiovascular, neoplasms or other causes, and there is indeed some evidence of increased risk of cardiovascular deaths but not so for neoplasms. The strongest association, however, is between illicit drug use and unnatural deaths, which is high compared to the general population. The widening gap could also be explained by the actual life expectancy of individuals with schizophrenia remaining static, while that of the rest of the population has improved. To effectively address this apparent increase in MR gap, we need to understand the reasons better. In the subsequent paragraphs, we have addressed the key questions posed by the authors.

In response to the first two questions, the authors ask whether schizophrenia is a progressive disease and whether relapse contributes to treatment resistance; they conclude that schizophrenia is a progressive-relapsing disease, resulting in ‘emergent treatment resistance in most cases’. Although there is now considerable evidence for brain changes longitudinally, a number of key questions remain unanswered, including the role of stress, illicit drug use, inactivity and role of maturational processes (Pantelis et al., 2005). Furthermore, there are limited data informing the hypotheses that relapse is associated with neurobiological changes (Cropley and Pantelis, 2014). The term treatment-resistant schizophrenia (TRS) has been around for a number of years; however, there is no consensus in defining it and the definition may vary depending on how response is defined. An extensive review and guideline concluded that 10–30% of patients have little or no response to antipsychotics and an additional 30% may have a partial response. Most of the studies summarized by the authors are based on symptom recovery rather than functional outcomes. A 97-year (1895–1992) meta-analysis involving >300 studies (51,800 patients) demonstrated that the rate of improved outcomes varied between 35% and 48%, with broader criteria associated with better prognosis (Hegarty et al., 1994). More recent studies suggest recovery occurs in a significant proportion of cases; however, outcome is based on a variety of prognostic factors (e.g. age of onset, duration of untreated psychosis, ethnicity). Hence, it would be simplistic to generalise that schizophrenia most often leads to treatment resistance.

When should the diagnosis of schizophrenia be made? We agree with the author’s viewpoint that the diagnosis should be made at the earliest opportunity possible, and this applies to any stage of schizophrenia, not just first-episode psychosis (FEP). Indeed, diagnosis may be inconclusive during earlier illness stages. Instability in diagnosis may be related to early acute illness, wherein a clear trajectory has not yet emerged, and to drug-induced psychotic states that mimic schizophrenia. However, once diagnosis is established, it is relatively stable. It is currently unknown how neurobiological changes over time are linked to diagnostic changes. The authors argue that there has been a culture of negativity around the diagnosis of schizophrenia and lack of emphasis on medication adherence issues. They further assert that clinicians and patients are unreliable in assessing medication adherence. While it is clear that medication adherence improves outcomes in schizophrenia, the evidence for strategies to improve adherence is limited. Insight and positive attitude towards medication are the only factors that are reliably linked with improved adherence, and there is no evidence that antipsychotic class or route of administration has any impact on adherence. As clinicians, we need to be cognisant that adherence is not a simple dichotomous ‘yes’ or ‘no’ concept, but can encompass the full spectrum from complete adherence to complete non-adherence. However, we agree with the author’s view that ensuring an early diagnosis and working towards improving adherence to medication at an early stage of the illness are crucial in improving long-term outcomes, and this leads us to the next important issue as outlined below.

Should maintenance antipsychotic medication be discontinued in fully remitted FEP? The crux of the argument here is based on the author’s critique of the 2016 Royal Australian and New Zealand College of Psychiatrists (RANZCP) Clinical Practice Guidelines (CPG) and other CPGs. CPGs have been in use in Australia and New Zealand since the late 1970s and have potential benefits as well as limitations. On a positive note, they are able to distil a vast amount of evidence, which is growing all the time, into a synthesised, practice-based recommendation that is useful for clinicians as a handy tool, helpful for patients by increasing the efficacy and reducing harms and enabling policy-makers to choose the most effective interventions that offer the biggest gains and value for money. The limitations of such guidelines are that if they do not follow rigorous methodology and are not objective, they risk recommending ineffective or harmful interventions and clinicians may feel straightjacketed by the constraints imposed by them.

The authors disagree with the RANZCP CPG regarding duration of treatment, concluding that maintenance medication should not be ceased in remitted FEP, which they argue on the basis of medication discontinuation leading to increased risk of relapse. While non-compliance and medication cessation can cause relapse, it is not the only cause. A Cochrane systematic review concluded that specialised care teams reduced dropouts and improved compliance; however, outcomes were not improved, including length of admissions and rehospitalisation rates at 5 years (n = 547, relative risk [RR] = 1.05, confidence interval [CI] = [0.90, 1.2]) (Marshall and Rathbone, 2011). There are not enough data from high-quality systematic reviews to make a recommendation either way. The long-term harmful effects of antipsychotics need to be carefully balanced against benefits. Care should be tailored to the individual based on previous history, risk to self and others, severity of the illness and resource availability. Clearly, the majority of patients with schizophrenia are not FEP and the issue of discontinuation is a more complex one beyond the first episode.

Can antipsychotics cause cortical grey matter changes? The authors highlight an important controversy in the field, and their conclusion that antipsychotic medication is not the cause of adverse grey matter changes is largely based on 3 key meta-analyses (Fusar-Poli et al., 2013; Haijma et al., 2013; Vita et al., 2015; for references, see Catts and O’Toole, 2016).

Undertaking a systematic review involves a number of rigorous steps and is the capstone of the hierarchy of evidence pyramid. Such reviews minimise bias through transparent methods that are available for replication. They involve formulating the correct question, developing a peer-reviewed protocol, a well-documented search strategy, refinement of the search using pre-specified criteria, quality analysis and weighting of individual studies (risk of bias assessments), transparent documented decision-making (a key ingredient) and, finally, appropriate statistical methodology for pooling comparable outcomes which makes the results scientifically acceptable and readily updatable. Evidence from systematic reviews is then used to inform guideline development, while other studies not meeting criteria (e.g. non-randomised studies) are also analysed, weighted and graded.

The three meta-analyses examined differ in methodology and rigour, including analysis of different outcomes. While emphasis was placed on the findings of Vita and colleagues, this study did not meet criteria set out above for a high-quality systematic review; for example, a quality analysis was not included and there were errors in numbers of studies reported. The Haijma et al. study failed to report a search strategy and only included English language studies; significant heterogeneity for many of the outcomes was identified (I² > 50%), but the degree of heterogeneity was not predefined; publication bias was reported but not investigated further. The study by Fusar-Poli and colleagues was better methodologically, with a documented search strategy, no language exclusion and defined protocols for quality assessment of included studies. The main conclusion from this paper was, appropriately, that ‘a different moderating effect of SGAs taken alone and of FGAs or mixed treatment on cortical GM tissue loss in schizophrenia are speculative at this time’.

Importantly, all studies reported did not aim to examine medication effects on brain structure a priori. The authors concur on this point as the studies have multiple confounders. Their higher weighting of the Vita et al. study should be reconsidered given our comments. While the authors conclude that these meta-analyses are of clinical interest but do not provide proof that medications cause brain changes, they infer that patients with less volume loss have better prognosis and hence do not require higher doses of antipsychotics. We contend that the absence of evidence is not evidence of absence (argumentum ad ignorantiam, or appeal to ignorance) in itself, and this complex issue is one that requires a key debate, which is informed by longitudinal well-controlled comparisons of antipsychotic effects in different disorders.

Do long-acting injectable antipsychotics (LAIA) reduce relapse rates? The authors argue that due to limitations of RCTs, real-world observational data should be given more weight and conclude that ‘logically, long acting injections should begin at first admission and continued indefinitely’. While they argue some of the studies are not ‘real world’, their suggestion that patients should continue on LAIA indefinitely means they are omitting a key aspect of real-world care – consumer choice and the possibility of shared decision-making with their clinician. A number of factors are important to maintaining treatment adherence, including developing therapeutic alliance and a sense of self-efficacy. We argue that such issues form the bedrock of any treatment plan devised; by contrast, enforcing indefinite invasive treatment is recipe for discontinuation.

Additionally, while observational studies are useful, they do have inherent biases. For this reason, well-thought-out grading recommendations (e.g. GRADE) do give weight to both RCTs and observational studies; however, conclusions depend on careful evaluation of the quality of such studies.

High-quality Cochrane systematic reviews that evaluated LAIA indicate that most trials do not report outcome data on relapse, and where reported, there was no difference compared with oral antipsychotic groups (references available on request). Thus, if no differences were found for highly motivated consenting adults in a trial, the real-world refusal rates for LAIA are likely to be higher, particularly if enforced. Further well-designed trials of LAIA evaluating relapse as a key outcome are required. Until such data are available, we suggest that intervention should be tailored to patient’s needs. Individualised, personalised and acceptable treatments are essential in the context of a collaborative partnership between clinicians and patients such that shared decision-making can occur and work continues towards common goals and outcomes that can be openly discussed and debated.

There is indeed much to learn about the neurobiology and treatment of schizophrenia and this debate is important and needed, but this also needs to be done thoughtfully by all concerned.