To the Editor:
Although the medical community is appropriately alarmed and galvanized to action by the current opioid epidemic in our nation, associations can be easily ascribed and portrayed as scientific certainty. In a recent large database study, Gupta et al. posited in their title that “Opioid Abuse or Dependence Increases 30-day Readmission Rates after Major Operating Room Procedures.”1 This title, however, is at odds with more cautiously crafted “association” statements in the abstract conclusion, discussion, and concluding paragraph, which more appropriately describe the retrospective findings.
I applaud the authors for the sheer magnitude of the study and the data describing the presumed inherent risk for this patient cohort. “Adjusted analysis” cannot obfuscate, however, the message apparent from a careful reading of the presented data. I defer to the statistical expertise and skill of the authors, but the presented medical conditions and risks of the opioid abuse/dependence group are significantly different (in exponentially large effect) in 29 of the 32 presented comorbidities spanning every clinically relevant major organ system (table 2). It is not surprising that the odds ratio of the multivariate model (table 4) has several positive predictors of readmission of greater or equal magnitude than opioid abuse/dependence (odds ratio, 1.26), including the following:
Renal failure: 1.37
Deep vein thrombosis: 1.34
Acquired immunodeficiency syndrome: 1.32
Diabetes mellitus with complications: 1.29
Peptic ulcer: 1.28
Liver disease: 1.27
Fluid/electrolyte disorders: 1.26
Although a couple of these predictors may, in turn, be associated with opioid abuse, several have no plausible explanation for a relationship.
Discussion of the study limitations and concluding that “we describe an association between a history of opioid abuse and increased risk of readmission after perioperative hospitalization” appears to be a fairer reflection of the presented data than the original title, especially when the magnitude of other, likely independent predictors is considered, even when accounted for as potential confounders. Acknowledgment of the retrospective association appears to be missing from the title, which to the casual reader may be accepted as cause and effect.
The author declares no competing interests.