Is Less than 50% More Narcotics Really Noninferior?

I read with great interest the study by Kim et al.¹  presenting data on the superiority of an adductor canal block compared to a femoral nerve block preserving muscle strength and the noninferiority in analgesia. I believe the noninferiority “conclusion” deserves some discussion.

One of the most important aspects of designing a noninferiority study is the choice of the noninferiority margin^*—here a 50% increase in narcotic consumption. The validity of any conclusion from the noninferiority study depends on the choice of margin.^* I do not think—statistically formulated but from a clinical perspective—that an intervention is noninferior if less than 50% more narcotics are used.

From a statistical perspective, the margin was chosen based on a small study with the risk of overestimation of a treatment effect.²  Further, as discussed in the same article, there are conflicting results of the narcotic-sparing effect of a femoral nerve block.²  A conservative adjustment would have been necessary—choosing a smaller margin to accommodate this uncertainty.^* The upper limit of the CI is 1.3–1.4 depending on the time interval examined.¹  This means a more conservative margin of 30% (30% increase in narcotic consumption) would have failed to show noninferiority at all time intervals.

Another factor deserves illustration: the constancy assumption.^* In a noninferiority study, you are trying to compare the treatment group indirectly to a historical placebo group. But there are substantial differences in both studies, namely, epidural analgesia and small-dose acetaminophen, that make comparison difficult. These confounding interventions can reduce differences between study (adductor canal block) and control (femoral nerve block), making it “easier” to show noninferiority.

Also influencing the interpretation of the results is the substantial loss or simply the variability of treatment effect of the femoral nerve block. In the study by Allen et al.,²  approximately 18 mg morphine equivalent dose was used (only a figure provided) compared to 36 mg in the current study. This could lead to the incorrect result of “proving” noninferiority when in fact there could be inferiority or potentially no effect at all.

Additionally, in a noninferiority study, the power analysis becomes crucial—the smaller the margin, the greater n has to be. In a superiority trial, you bias toward a type 2 error if you “under power” a study; in a noninferiority trial, a type 1 error is introduced. As discussed above, there are strong clinical and statistical arguments to choose a more conservative margin. This would have required a greater number of patients for adequate power.

I agree with the authors that the superiority conclusion is very sound (muscle strength), but a noninferiority conclusion cannot be drawn from the data at hand—so it remains to be determined if an adductor canal block is noninferior to a femoral nerve block with regard to analgesia. Here, the margin needs to be reevaluated using more than one small study, adjusted and clinically evaluated, answering the question: How much more narcotics are you willing to give to provide greater muscle strength and call it noninferior knowing now that there is at present no evidence for an increased fall risk with regional anesthesia after total knee arthroplasty?³ 

The author received speaking fees and honoraria from Cadence^® (now Mallinckrodt Pharmaceuticals^®) (St. Louis, Missouri) and Baxter^® (Deerfield, Illinois). The author is a shareholder in Johnson & Johnson^® (New Brunswick, New Jersey).