Inconsistencies in Reporting Studies of Lactic Acidosis

Abstract

In Reply Many of the questions raised by Chapple and Blackston and by McDonald and colleagues about our recent Original Investigation1 are addressed in Pocock and Stone’s recent review onwhat to dowhen the primary outcome fails.2Certainly, a trial inwhich theprimaryoutcome fallsshortofstatisticalsignificance can be distressing to investigators. However, as highlighted by Chapple and Blackston, the interpretation of trial resultsmay be colored by undue attention to a single primary outcome and arbitraryPvalue cut points. These constraintsmake sense in confirmatory studies of new drugs and devices (where the consequences of falsepositivescan bedire) butnotnecessarily inmore exploratory studies (like the PersonalizedResearch forMonitoring Pain Treatment study3 ). Both letters raise a number of other methodological issues, including lack of statistical power, underemphasis of important secondary outcomes, problems with application of the n-of-1 intervention, and potentially poor patient adherence. As we noted in our article,1 the study fell 12% short of enrollment goals, but it is not clear that reaching the planned sample size of 244 would have resulted in a significant P value. Single studies rarely provide definitive estimates of effect size, and for this reason we believe further studies (and subsequent meta-analyses) are warranted. We agree that statistically significant between-group differences were seen inmedication-related shared decisionmaking and in the probability of achieving a 5-point pain interference score reduction. These findings are clinically important and deserving of further study. Likewise, although certain n-of-1 trial design choices (eg, offering nonpharmacologic treatments and relatively short treatment periods)may have contributed to the large proportion of inconclusive n-of-1 trials, our goalwas to balance experimental rigor with patient choice and convenience. Finally, although patients randomized to the n-of-1 arm adhered well to their assigned treatment regimens (averaging 1.4 on a 1-5 scale, with 1 indicating “always” following the directed treatment), we did not track adherence to the “winning” treatment following the trial. If the benefit of n-of-1 trial participation (if any) is mediated purely through the identification of clinically superior treatments, poor adherence to the “winner” in the aftermath of an n-of-1 trial could, as McDonald and colleagues suggest, limit the potential benefit. However, we suspect that other potential mechanisms are operative (eg, creating a more therapeutic physician-patient relationship, enhancing patients’ self-efficacy as autonomous agents) and may deserve more attention than previously recognized.