Selective bias in criticism of early intervention

Dr Amos criticizes my paper (Yung, 2012) and me personally in his Commentary (Amos, 2012) and attempts to portray himself as an impartial scientific sceptic. However, while he accuses me of ‘selective use of research to support a position rather than weigh the evidence’, he uses this very technique to illustrate his points.

Specifically, Dr Amos criticizes my lack of scientific rigour, accuses me of referencing an abstract incorrectly and of wrongly citing the abstract instead of a journal article. Let us examine each of his accusations in turn.

Early intervention has some positive effects on outcome at 5 years. I illustrated this point by reference to the recently published Canadian study (Norman et al., 2011). Additionally I noted some long-term positive effects of early intervention in the Danish OPUS randomized control trial that persisted even after withdrawal of the specialized service (Bertelsen et al., 2008). I did this because critics have consistently emphasized the lack of differences between the early intervention (EI) group and the non-EI groups in the OPUS study (Bosanac et al., 2010; Castle, 2012). They then conclude that outcomes are no different between the groups. Yet ‘outcome’ should be assessed across a range of domains, including peer relations, employment, ability to live independently, level of distress and quality of life. Indeed in the recently released report on Australians living with psychotic disorders (the SHIP report) (Morgan et al., 2011), finances, social isolation and employment are noted by people with psychotic disorders and general practitioners alike as the three most important aspects in their lives. The failure to acknowledge that some important long-term benefits of early intervention (rates of living independently and days in hospital (Bertelsen et al., 2008)) persisted at 5 years is itself selective use of the literature.

Further, in his reading of the Friis abstract from the Norwegian TIPS study (Friis et al., 2011), Dr Amos makes much of the apparent higher rates of patients not capable of living independently in the early detection (ED) group compared to the group without ED. However, he fails to note the high drop-out rate of the more ill patients from the area that had no ED in contrast to a lower drop-out rate in the ED area, and with drop-outs in the ED area having fewer negative symptoms. That is, Dr Amos selectively ignores an important selection bias, one that would favour a finding of no effect of ED.

The second accusation, that I incorrectly cited an abstract, is simply wrong. Dr Amos does not seem to have even looked up the reference I supplied. If he had, then he would have discovered that I correctly reported a positive effect of ED. Rates of recovery were twice as high in the area that practised ED compared to the area that did not (McGlashan et al., 2011).

Finally he queries why I cited an abstract at all rather than a journal article. Citing published abstracts is common practice in order to present the most recent findings from a study. Dr Amos has even done it himself in his Commentary. He may be horrified to know that sometimes even ‘personal communications’ are quoted in articles if breaking research has not yet been presented at a conference. The abstract that I cited included 10 year follow-up data, not available in article form at the time of writing my commentary. Dr Amos will be relieved to discover that the data are now in press in the American Journal of Psychiatry.

Furthermore Dr Amos fails to turn his attention to Professor Castle’s article (Castle, 2012), published alongside mine, that was critical of EI. There is much to question in Professor Castle’s article, yet it seemingly escapes the scientific scepticism of Dr Amos. If Dr Amos’s agenda were to solely promote rigorous evaluation of evidence, one would think he would subject both sides of the argument to the same assessment. Yet this is not the case.

Perhaps Dr Amos could have queried why Professor Castle asks the question of who is being treated and whether they actually need treatment, when the answers to his questions lie in one of the papers Professor Castle cites (Cheng et al., 2011). In this article, Cheng et al. conducted a rigorous epidemiological study of people being treated at EI services in Cambridgeshire. The authors state explicitly that their services manage patients with first episode psychosis and do not treat people with subclinical psychotic symptoms from the general population or even individuals with an At Risk Mental States (that is, people meeting Ultra High Risk criteria, considered to be at high short-term risk of developing psychotic disorder (Yung, 2011)). They note that there was no increase in incidence of psychotic disorders in the area after the establishment of an EI service. The authors point these findings out directly in response to Professor Castle’s assertion that EI services pathologize normal developmental processes (Bosanac et al., 2010). Yet Professor Castle chooses to ignore this evidence in favour of repeating his own position. Furthermore, Professor Castle refers to his own estimate of the incidence rate of psychosis in the Early Psychosis Prevention and Intervention Centre (EPPIC) but supplies no evidence to back this up, his only reference in support of his finding being a line in a similar opinion piece (Castle, 2011). Finally in a selective omission, Professor Castle discusses the effects of duration of untreated psychosis (DUP) without reference to the biggest trial of early detection to reduce DUP, the Norwegian TIPS study, referred to above. All this escapes the attention of Dr Amos.

My article recognized that there are research gaps in the EI area. It stated that there are patients who will experience poor outcomes despite receiving EI. It is important to recognize both of these aspects so that we can continue to create knowledge, evaluate new systems of managing our patients and identify those that need additional treatment. Debate is important so that treatment, services and policy are evidence-based. But dismissing all evidence in favour of early detection and intervention by selectively quoting the literature and uncritically accepting the claims of the EI critics fails our patients and their families.

See Commentary by Amos, 2012, 46(9): 903–904. See also Debate by Yung, 2012, 46(1): 7–9