Confounding of Antipsychotic Discontinuation Studies by Withdrawal-Related Relapse

Schoretsanitis et al conclude from their reanalysis of individual patient data that most people relapse when they stop antipsychotics but the risk of relapse is greater by about a factor of 4 in those who stop oral medication compared with those who stop depot medication.1 It is perhaps no accident that depot antipsychotic concentrations slowly fall over months in an exponential manner, partially mimicking the gradual, hyperbolic antipsychotic reduction recently proposed.2 There are 2 possible explanations for a lower relapse rate from depot medication compared with oral and one or both might be relevant. The first is that the small amount of antipsychotic that remains for months after stopping depots is enough to prevent relapse in some patients. This explanation is of interest because it suggests that very low concentrations of antipsychotic (after 5 half-lives, drug levels will only be 3% of original) are enough to prevent relapse. This is consistent with previous findings3 and the knowledge that low concentrations of antipsychotic produce higher receptor occupancy than might be expected.4 It is also possible that the small residual doses may function to delay the onset of withdrawal-related effects. The second explanation, favored by the authors, is that gradual tapering is beneficial beyond simply the period over which the antipsychotic is reduced, and leads to less relapse overall—and not just delayed relapse. The authors surmise that this may be because of a “more homeostatic transition to a new equilibrium.” There is some support for this in their analysis, as the risk for relapse is 5 times greater in the oral cessation group than in the depot group even when analysis is restricted to those patients for whom drugs released from depots are largely cleared from their systems (ie, at more than 5 half-lives of follow-up). This finding is also consistent with a recent meta-analysis where patients tapered off antipsychotics over months had less than half the extent of relapse than those stopped abruptly (31% vs 77%).5 However, there still remains the possibility that these relapse curves might converge at a more distant time point. These observations lead us to the conclusion that there is something about the nature of the process of drug discontinuation itself that is causally related to the risk of relapse. If stopping medication were simply a case of revealing the underlying, untreated condition then the manner in which medication was stopped would be of no consequence to the rate and extent of relapse. The fact that the manner of stopping is significant suggests that withdrawal effects may play a part: either that psychotic symptoms themselves are withdrawal effects, as has been noted in case studies,2,6 or that the destabilization caused by withdrawal effects (which can include insomnia, anxiety, and agitation) can lead to genuine relapse (which has been termed withdrawal-associated relapse).2,7 It is possible that rapid withdrawal does not give sufficient time for full homeostatic reset and so relapse results. The full implications of the role of the discontinuation process in causing relapse have not been considered in the current study. Although stopping 3-monthly paliperidone will allow a slow fall in plasma levels over a year or more, this may still be too quick to eliminate all withdrawal-related effects. One study considered them still evident when antipsychotics were reduced over 18 months.8 Note also that long lasting changes to the brain (as manifest in tardive dyskinesia which is thought to be due to dopamine supersensitization) can persist for years or longer.2 Therefore, it is possible that with an even slower reduction of antipsychotic plasma concentrations, the relapse rate would be lower than that seen in the paliperidone palmitate cessation group and the extent of relapse lower still. If this is true, it may not be accurate to say that most patients relapse on stopping their antipsychotics—only that this is the case with current cessation methods, throwing into question the authors’ conclusion that there is a high risk of relapse after antipsychotic discontinuation. We must also question the idea that “rebound psychosis” (a synonym for withdrawal-associated relapse) should occur within 1 month following drug cessation. The increased relapse rate present in the oral group, which must at least partially be due to rebound psychosis, is present for several months after abrupt stopping (the curves continue to have diverging gradients for this period). Indeed, with the understanding that withdrawal effects are caused by neuroadaptations (thought primarily to be dopaminergic hypersensitivity) to the presence of an antipsychotic which persist after stopping the drug, it should not be a surprise that the consequences including withdrawal-associated relapse should persist for the same time period.2 Finally, given the evidence presented in this study, and elsewhere,5 that the rate of drug discontinuation affects the rate and extent of relapse, implicating withdrawal-related effects, it may not be accurate to say that long-acting injections (LAIs) reduce the relapse rate. It may be some or all of the benefit in this respect is due to reduction of withdrawal-related effects that are (mis-)categorized as relapse. Therefore, the major conclusion to be drawn from this study is that antipsychotics should be stopped much more gradually than they are currently (even more gradually than that entailed in stopping a depot abruptly), not that there should be more widespread prescription of LAIs (because of intrinsically greater relapse prevention properties on discontinuation). In general, the rate of stopping antipsychotics in discontinuation trials should be seen as an important factor in affecting rate of relapse. Consequently, tapering rate should be reported when discontinuation trials are analyzed and should be taken into consideration in interpretation of results. In order to minimize this confound antipsychotic discontinuation trials should employ very gradual tapers (probably in a hyperbolic pattern) over months or years especially for patients who have been on medication long term. Given the confounds affecting antipsychotic discontinuation trials,9 as well as the discrepancy between discontinuation and existing continuation trials, we should be more cautious in interpretation of the relapse prevention properties of antipsychotics. M.H. declares no conflict of interest. D.T. reports grants and personal fees from Janssen and Recordati, personal fees from Mylan, Sunovion, and Recordati, as well as stock in Saladax and Psychiatric Genetic Testing, outside the submitted work. R.M. has received payments for nonpromotional lectures from Janssen, Lundbeck, Otsuka, Sunovian, outside the submitted work. No specific funding was provided for this work. M.H. is supported by a clinical research fellowship through the North East London NHS Foundation Trust (NELFT). This funding source had no role in the writing of the manuscript or the decision to submit it for publication. M.H. conceived the manuscript idea and wrote the manuscript. D.T. and R.M. helped develop the manuscript idea, and revised and edited the manuscript. Ethical approval was not required for this work because no patients were involved in the production of the work.