A systematic review is only as informative as the relationship between the evidence it includes and the conclusions it draws. Writing in the British Journal of General Practice, Froentjes and colleagues conclude that opioids provide little or no benefit over longer periods and may worsen pain beyond twelve weeks. Yet their principal analysis excludes a substantial body of randomized evidence involving patients who initially benefited from opioid treatment.
The fundamental problem is the mismatch between the authors’ restricted evidence base and their broad conclusion. Their analysis can describe outcomes in the trials they selected. It cannot establish that sustained opioid benefit is absent in populations whose evidence they excluded.
Selecting the population matters
In oncology, a targeted therapy may benefit patients whose tumors possess a particular molecular characteristic while offering little benefit to a population without those same molecular properties. Including patients unlikely to respond can dilute the average treatment effect. We would not accept that as proof that therapy lacks efficacy in patients whose biology makes them unlikely responders.
Opioid pharmacology also provides reasons to expect heterogeneous responses. One illustration is incomplete cross-tolerance: when a patient develops tolerance to one opioid, that tolerance does not fully extend to another, so some patients obtain better pain relief after switching. Pasternak described a possible biological explanation: alternative splicing allows the same μ-opioid receptor gene to produce multiple receptor variants with different signaling properties, which contribute to differences in how individual opioids produce their effects.
The parallel with targeted cancer therapy is that biological responsiveness matters when evaluating treatment benefit. Opioid receptor research does not yet provide a validated molecular test for selecting the best medication for an individual patient, nor does incomplete cross-tolerance establish long-term efficacy. It does, however, support taking differences in response seriously. Failure to benefit from one opioid does not establish that another will fail, and an average result in a broad population does not establish the absence of benefit among demonstrated responders.
Excluding enriched enrollment randomized withdrawal trials (EERWs) skews the evidence base relative to the clinical question. EERW trials identify patients who initially benefit from and tolerate a specific opioid before testing whether continued treatment preserves that response. Removing these trials disproportionately removes evidence about successful treatment response. Generalizing the remaining findings to all patients—including demonstrated responders—introduces selection bias. Any broad conclusion about sustained benefit must account for studies specifically designed to evaluate its maintenance in that population.
The review’s own results undermine its broad conclusion
Froentjes and colleagues excluded ten EERW trials from their principal analysis. Nevertheless, their secondary analysis of nine such trials, involving 3,829 randomized participants, found significantly greater clinically meaningful pain response with continued opioid treatment at twelve weeks or longer: risk ratio 1.29, with a 95% confidence interval of 1.22–1.37.
The risk ratio of 1.29 means that patients assigned to continued opioid treatment were 29% more likely to achieve the study-defined clinically meaningful pain response than those switched to placebo. This is a relative increase in the likelihood of responding, not a 29% reduction in pain or a 29-percentage-point jump in overall response rates. The 95% confidence interval of 1.22–1.37 indicates an estimated relative advantage of 22%–37%. Because the entire interval exceeds 1.0, the finding is statistically significant.
The result matters because it directly addresses whether patients who initially benefit from and tolerate an opioid can maintain benefit with continued treatment. It requires any broad conclusion about the absence of sustained benefit to acknowledge the positive evidence in demonstrated responders rather than relegating it to a secondary discussion.
Not every trial needs to enroll demonstrated responders, as trials in broader populations can inform decisions about initiating treatment. But a broad claim about sustained efficacy must account for evidence from patients who initially respond. Otherwise, a review risks answering whether treatment benefits an average participant but excluding those who are likely responders.
An inconclusive average is not proof of no benefit
Even within the selected conventional trials, the long-duration responder estimate was inconclusive: risk ratio 0.91, with a confidence interval of 0.73–1.14. The risk ratio of 0.91 means that patients receiving opioids were estimated to be 9% less likely to achieve a clinically meaningful pain response than those receiving the comparison treatment. However, the 95% confidence interval of 0.73–1.14 spans possibilities ranging from a 27% lower to a 14% higher likelihood of response with opioids. Because that interval includes 1.0—the value indicating no difference—the result does not establish either superiority or inferiority, nor does it demonstrate equivalence.
The appropriate conclusion is that these trials did not establish a clear average advantage, not that opioids provide no sustained benefit or that responsive subgroups cannot benefit. The suggestion of worsening pain in their analysis is weaker still. The secondary pain-score estimate favored opioids, although the average difference was small and statistically inconclusive.
These findings justify a narrow statement: the selected conventional trials did not demonstrate a clear average responder advantage at twelve weeks or longer. They do not justify a general conclusion that opioids cease to provide meaningful analgesia. As targeted cancer therapy illustrates, identifying the relevant population is central to interpreting efficacy. Excluding evidence from that population cannot establish the absence of benefit within it.
References
- Froentjes LS, Barrington CAM, Craig RA, et al. Impact of treatment duration on the effectiveness of opioid analgesia: a systematic review and meta-analysis. Br J Gen Pract. 2026;76–e643. doi:10.3399/BJGP.2025.0705.
- Pasternak GW. Incomplete cross tolerance and multiple mu opioid peptide receptors. Trends Pharmacol Sci. 2001;22(2):67–70. doi:10.1016/S0165-6147(00)01616-3.
- Pasternak GW. Multiple opiate receptors: déjà vu all over again. Neuropharmacology. 2004;47(Suppl 1):312–323. doi:10.1016/j.neuropharm.2004.07.004.
Lynn Webster, MD
Lynn R. Webster, MD, is a physician, researcher, and author specializing in pain and addiction medicine. He is a Senior Fellow at the Center for U.S. Policy and co-author of Deconstructing Toxic Narratives: Data, Disparities, and a New Path Forward in the Opioid Crisis.
