An early treatment marker can predict a later outcome without proving that acting on the marker improves care. A 2012 analysis of bupropion trials for methamphetamine dependence is a useful example: it proposed early decision rules, but never randomized participants to follow those rules.
Which patients supplied the prediction data?
Brensilver and colleagues retrospectively combined 55 participants from the active-bupropion arms of two earlier randomized trials. All reported methamphetamine use on fewer than 19 of the preceding 30 days. They represented 48% of the combined bupropion-treated samples, rather than every participant in the parent trials.
The historical regimen used oral sustained-release bupropion, 150 mg daily initially and then 150 mg twice daily, during 12 weeks of treatment. Both trials included cognitive behavioral therapy and incentives, but the counseling schedules and incentive contingencies differed. This secondary analysis was not a new placebo-controlled efficacy trial.
What counted as later treatment failure?
End-of-treatment abstinence required no positive methamphetamine urine results in the final two weeks, allowing only one missing specimen per week. Participants with a positive result or more extensive missingness were classified as persistent users. Fourteen of 55 met the abstinence definition; only 28 submitted any specimen in the final week.
That endpoint combined evidence of drug use with remaining available for testing. Someone who left the study could fail the endpoint without a positive final specimen. Describing every failure as chemically confirmed continued use would therefore misstate the measurement.
How well did early results discriminate?
For thrice-weekly urine testing, the area under the receiver operating characteristic curve was 0.801 after two weeks, with a 95% confidence interval of 0.675–0.926. Three weeks produced an AUC of 0.808, without a significant improvement over two weeks. An AUC summarizes ranking discrimination between later outcome groups; it is not the percentage of patients correctly classified by every possible decision threshold.
The authors reported a greater than 90% likelihood of endpoint failure among people not achieving the proposed early-negative-sample pattern. With only the first specimen of each week, failure to provide at least two negative specimens in the first three weekly visits had a reported positive predictive value of 88% for failure. Predicting success was weaker: all-negative results in those three weekly samples had a reported negative predictive value of 51%.
Why can those predictive values change in another clinic?
The underlying failure rate was high in this selected sample. Positive and negative predictive values depend on how common the target outcome is in the population being tested. They are not fixed properties of a urine assay.
Missing early samples were counted as positive. Testing frequency also changed the information available: six opportunities over two weeks are not equivalent to three opportunities over three weeks. The parent studies used somewhat different laboratory procedures, including different confirmation practices. A proposed clinical rule needs those details, not just a memorable percentage.
Did the analysis validate medication-specific resistance?
No. Early response had similar predictive characteristics in placebo-treated participants. That observation suggests the marker may reflect broader engagement or treatment responsiveness, rather than a specific biological inability to benefit from bupropion.
The small pooled sample, empirically selected threshold and absence of an independent prospective validation limit transportability. Most importantly, the analysis did not compare continuing treatment against switching or augmenting it when the marker was unfavorable. It cannot establish that discontinuation improves outcomes, and should not be presented as a current treatment instruction.
Source
Brensilver M, Heinzerling KG, Swanson AN, Shoptaw SJ. A Retrospective Analysis of Two Randomized Trials of Bupropion for Methamphetamine Dependence: Suggested Guidelines for Treatment Discontinuation/Augmentation. Drug and Alcohol Dependence. 2012;125:169–172. doi:10.1016/j.drugalcdep.2012.03.027. Historical retrospective prediction analysis.
Leave a Reply