Misclassification of overdose events in the X:BOT study - Authors' reply
07/2023
Journal Article
Authors:
Lee, J. D.;
Nunes, E. V.;
Van Veldhuisen, P. ;
Lindblad, R.;
Rotrosen, J.
Volume:
402
Pagination:
527-528
Issue:
10401
Journal:
Lancet
PMID:
37480935
URL:
https://www.ncbi.nlm.nih.gov/pubmed/37480935
DOI:
10.1016/S0140-6736(23)00049-1
Keywords:
Humans *Drug Overdose
Abstract:
Nabarun Dasgupta and colleagues here repeat findings and conclusions from their post-hoc re-analysis of the X:BOT trial's public dataset using an alternative overdose classification methodology and a new survival model,1 compared with the original methods and results published in 2018,2 and our own post-hoc re-analysis of public and non-public safety data published in 2022.3 All three approaches yielded similar findings—ie, numerically more non-fatal overdoses (and participants having overdoses) if randomly assigned to extended-release naltrexone versus buprenorphine (found to be statistically significant with their approach, but not with ours), and numerically more fatal overdoses if randomly assigned to buprenorphine (not significant). We do not agree with Dasgupta and colleagues' call to correct or retract our initial report. We note that, in this context, The Lancet sought independent expert opinions from the original reviewers of the X:BOT paper, who were unanimously in agreement that there is no case for correcting or retracting the paper and encouraged publication of these Correspondences.
In regard to the overdose classification, although operationalised, validated criteria for defining overdose versus other adverse events are under-developed, the X:BOT protocol prespecified the classification of adverse events by the Medical Dictionary for Regulatory Activities preferred terms, a standard opioid use disorder clinical trial approach. Using a partly restricted publicly available dataset, which excluded personal health information and narrative free-text event descriptions, and an alternative analytic approach, Ajazi and colleagues1 identified possible additional overdose events. Our re-analysis of the entire X:BOT adverse events dataset using an even more inclusive classification scheme yielded more likely and possible overdoses, demonstrating that there are multiple ways to recount adverse events post hoc, and reinforcing the need for a more structured approach to identifying and characterising overdose events. In all scenarios, however, although numerically more non-fatal overdoses were found in the naltrexone group, none of these achieved statistical significance. Compared with generic adverse event assessments, new non-fatal overdose-specific assessments that we are using in ongoing trials more systematically query self-reported non-fatal overdose events.4
Dasgupta and colleagues also address naltrexone induction failures. Naltrexone for opioid use disorder is difficult to initiate, a well known limitation to its effectiveness. In the X:BOT study, overdoses disproportionately occurred in individuals in whom extended-release naltrexone induction was unsuccessful and who never started the medication. Patients who attempt but are unsuccessful in naltrexone induction should be immediately started on buprenorphine or methadone.
In regard to medication discontinuation, among X:BOT participants who initiated either medication, most overdoses occurred long after they stopped taking medication. As Ajazi and colleagues and we noted, overdoses happened earlier in patients after they stopped receiving extended-release naltrexone compared with particpants discontinuing buprenorphine. Patients receiving extended-release naltrexone should be cautioned about post-discontinuation relapse and overdose risk, as they were during the X:BOT trial and per US Food and Drug Administration-approved labelling. Patients should be encouraged to conceptualise medication for opioid use disorder as a long-term commitment as with other chronic illnesses.
In summary, we found no errors in the original X:BOT report, its reported adverse event data, the trial's acknowledged limitations, and the overall findings and conclusions. Dasgupta and colleagues' and Ajazi and colleagues' interest in the X:BOT study has led to a deeper examination of the data and clinical and methodological recommendations around overdose risk, which we greatly appreciate.
X:BOT received in-kind study medication from Indivior and was entirely funded by the National Institute on Drug Abuse. JDL has received in-kind study medication for other trials from Indivior and Alkermes, an investigator sponsored study grant from Indivior, and is a science adviser to Oar Health. EVN has served or is serving as a principal investigator or co-investigator on studies for which support in the form of donated medication or digital therapeutics has been provided by Indivior, Alkermes, Braeburn, Pear Therapeutics, and CHESS Health, and has served as an investigator on a study funded by Braeburn and Camurus. EVN has served in a non-paid capacity as a consultant to Alkermes, Pear Therapeutics, Indivior, and Camurus. JR has served as a principal investigator or a co-investigator on studies for which support in the form of donated medication or funds has been, or is, provided by Indivior, Alkermes, and Braeburn, and apps have been, or are, provided by Pear Therapeutics, CHESS Health, and Data Cubed. JR served in a non-paid capacity as a member of an Alkermes study Steering Committee in 2019–21. All other authors declare no competing interests.
