Early Prediction of Antipsychotic Nonresponse Among Patients With Schizophrenia
Leucht S, Busch R, Kissling W, Kane JM;
Commented by , 30 Mar 2007
Aim of the study
The time course of the antipsychotic drug effect in schizophrenia is still poorly understood. For a long time there was a theory that onset of action of antipsychotic drugs is delayed. Recent reviews have clearly rejected this point of view, because they showed that antipsychotic effects can be detected at least as early as one week after the initiation of treatment (ref. 1, ref. 2).
Nevertheless, these reviews were based on the mean change of symptoms over time, i.e. they pooled responders and non-responders. To analyse which degree of early response predicts later response requires another approach. The authors tried to fill this gap by a sensitivity specificity analysis.
Methods
The authors post-hoc analysed a pooled database of seven antipsychotic drug trials with 1708 acutely ill patients with schizophrenia. The most important inclusion criterion was a considerable degree of positive symptoms at baseline. The research question basically was to establish a diagnostic test, therefore a sensitivity-specificity analysis was applied to predict which degree of initial non-response at week 1 or 2 makes it unlikely that a patient will not respond at week 4.
Several definitions for non-response at week 4 were applied, but less than 25% BPRS reduction was considered to be the most important one, because in previous analyses it had been shown that this reflects minimal improvement from a clinical point of view (ref. 3). Thus, if a patient is not even minimally improved at four weeks, he would certainly be switched.
If, however, less than 50% BPRS reduction had been used as a criterion, what would have been suggested for the patients with e.g. 45% BPRS reduction? Finally, a logistic regression model was applied to identify additional predictors of non-response.
Results
While a number of different models were presented, the main outcome was that 0% BPRS reduction at week 2 predicted less than 25% reduction at week 4 with a positive predictive power of 80%, a specificity of 98% and a sensitivity of 32%. In the logistic regression analysis this cutoff of 0% reduction remained the strongest one, but longer duration of illness, older age, and conventional antipsychotic drug use also played a certain role.
Dr Leucht's comments
The response to antipsychotic drugs in people with schizophrenia is very variable and clinically useable tests had not been available. This paper is the first one providing such a cutoff derived from a large sample. Those patients who have not improved at all or even got worse at week 2 may deserve a change of treatment (i.e. another antipsychotic drug or a high dose strategy or an augmentation strategy).
One limitation is that there is no randomised evidence showing that switching medication at week 2 is more effective than continuing with the same drug. But the non-response criterion at week 4 was less than 25% BPRS reduction meaning less than minimal improvement. Therefore, a change of treatment after 2 weeks may be justified even in the absence of such data, because keeping patients on the same drug simply means that they will remain unchanged.
Unfortunately, this cutoff of 0% does not mean that all the future non-responders will be identified. Positive predictive power and specificity were very high, meaning that only a few patients will be unnecessarily switched. But the sensitivity was low (32%), meaning that a number of patients who had more than 0% BPRS reduction at 2 weeks will also not respond.
Therefore, the 0% cutoff is a very conservative rule. Finally, the main limitation of this paper probably was that the database was composed of registrational antipsychotic drug trials. Replications in more naturalistic data are needed.
References
1. Agid O, Kapur S, Arenovich T, Zipursky RB. Delayed-onset hypothesis of antipsychotic action: a hypothesis tested and rejected. Archives of General Psychiatry 2003; 60 (12); 1228-1235
2. Leucht S, Busch R, Hamann J, Kissling W, Kane JM. Early-onset hypothesis of antipsychotic drug action: a hypothesis tested, confirmed and extended. Biological Psychiatry 2005; 57 (12); 1543-1549
3. Leucht S, Kane JM, Kissling W, Hamann J, Etschel E, Engel R. Clinical implications of Brief Psychiatric Rating Scale scores. British Journal of Psychiatry 2005; 187; 366-371