At Issue

Cochrane, Early Intervention, and Mental Health Reform

Analysis, Paralysis, or Evidence-informed Progress?

Patrick McGorry


Schizophr Bull. 2012;38(2):221-224. 

In This Article

The Cochrane Review on Early Intervention in Psychosis

The recent Cochrane review on EIP[9] has prompted renewed debate regarding the effectiveness of the EIP model of care, particularly in the context of the increased implementation of this model in Australia and internationally. While acknowledging the need for further studies to expand the evidence base, the review, while encouraging, remained equivocal regarding the effectiveness of EIP. Unfortunately, this stance is largely a function of the RCTs selected for the review, which were fundamentally inadequate "on their own" to answer the review's central question as to whether EIP is effective and warrants investment. The well-known constraints of the Cochrane methodology inevitably results in the exclusion of crucially relevant evidence—even clozapine has so far failed to receive strong endorsement by Cochrane. In the case of EIP, while 18 RCTs were included, 68 other studies, most of which were relevant to the key question regarding the value and impact of early intervention and should influence the conclusion that the review has sought to comment upon, were considered but excluded. Crucial exclusions were 2 positive health service level RCTs (which it should be noted are notoriously difficult and expensive to conduct, given their evaluation of models of service delivery, in contrast with simpler treatment/placebo trials) as well as a number of positive quasi-experimental[10] and historical controlled[11] studies. The landmark Early Treatment and Intervention in Psychosis (TIPS) study,[12] which showed that reducing the duration of untreated psychosis was feasible and effective in improving long-term outcome (notably negative symptoms), used the best possible (ethical) methodology for that research question yet was excluded by the orthodoxy of Cochrane. Similarly, at least 5 cost-effectiveness studies, with results that clearly favor investment in EIP models, could not be included.

Crucially, the majority of research studies included in the review were not health services research designs: that is, with the exception of the OPUS trial, the review did not include studies comparing EIP with standard mental health care. Rather, the review focused largely on trials comparing the "component interventions" of EIP services, such as cognitive behavioral therapy, suicide prevention, family therapy, vocational intervention, and interventions to improve physical health outcomes, with "standard care" (yet chose to exclude studies of dose and generation of antipsychotic medications). Such interventions were typically studied against the backdrop of the range of care provided within streamed EIP services. It is not surprising that some of these individual trials were "ineffective" given that the control groups in these trials were the recipients of an already comprehensive model of care.[13] The review blended RCTs of individual treatment components with health service level RCTs, problematic because the methodological standards cannot be the same, and only the service level studies are relevant to decisions on health service reform. Thus, the review muddied the waters on the salient question.

Despite positive 1- and 2-year outcomes in the OPUS trial,[14–16] the Cochrane review focused upon the study's nonsignificant 5-year outcomes and rather than interpret the partial erosion of benefit as a failure of standard care to maintain real benefits and disease modification, the review appeared to regard this erosion as a limitation of EIP. The recent paper from Norman et al[17] illustrates that EIP services, if extended even in dilute form for 5 years, can maintain the initial hard-won gains. Two other positive health services RCTs comparing EIP and standard care were excluded.[18,19]

Finally, the authors perpetuate the myth that UHR patients are not in need of care of some kind too, despite the fact that they are actively seeking help and manifestly distressed and impaired with average Global Assessment of Functioning scores in the low 50s. Hence the ethical situation is no different from that in first-episode psychosis, with the exception that the content of interventions, in line with the predictions of the clinical staging model,[20] should be different because of risk-benefit considerations and because it appears that antipsychotic medications are not needed or appropriate as first line treatment in this clinical stage. The Cochrane authors failed to carry out a meta-analysis of the UHR studies, which, when done, strongly favors intervention of some kind in this group rather than monitoring or neglect,[21] at the very least on the basis of need for care but on the secondary basis of delayed transition. Both are valuable and achievable goals.