Interpreting a Non-Inferiority Trial for Formulary Decision-Making
A pharmacy and therapeutics committee reviews a randomized, double-blind, non-inferiority trial of a new oral antibiotic (drug X) versus intravenous ertapenem for complicated urinary tract infection. The primary outcome was clinical cure at 10 days, with a non-inferiority margin of 10 percentage points. Results: cure 82% with drug X (n=310) vs 86% with ertapenem (n=305); difference −4% (95% CI −10.5% to 2.5%). The modified intention-to-treat population excluded 38 patients (12%) in the drug X arm and 15 patients (5%) in the ertapenem arm who received prohibited antibiotics or had no follow-up culture. Secondary: 30-day readmission 9% vs 7% (95% CI −1% to 5%). The manufacturer's press release states the drug is 'as effective as IV therapy and will reduce hospital admissions.'
1. Has non-inferiority been established? Explain using the trial results.
Read the worked answer
No. For a non-inferiority claim, the entire 95% confidence interval for the difference must lie above the prespecified margin of −10 percentage points.
Observed difference: −4% (drug X minus ertapenem). 95% CI: −10.5% to 2.5%.
The lower bound of −10.5% crosses the −10% margin, so the trial fails to demonstrate non-inferiority. The point estimate of −4% looks favorable, but the confidence interval includes a clinically meaningful loss of efficacy greater than the acceptable margin. A nonsignificant p-value is not evidence of equivalence. The press release's claim of 'as effective' is not supported by the prespecified analysis.
2. How does the differential exclusion of patients affect interpretation?
Read the worked answer
The exclusion rates are asymmetric and concerning: 12% in the drug X arm versus 5% in the ertapenem arm. Patients excluded for prohibited antibiotics or missing follow-up cultures are often treatment failures or complicated cases.
Removing a larger proportion of drug X patients can bias the results toward showing non-inferiority by removing poor outcomes from the new-drug arm. In non-inferiority trials, both intention-to-treat and per-protocol analyses matter, but differential attrition weakens the modified ITT result. The committee should request:
- Full ITT analysis including all randomized patients.
- Sensitivity analyses imputing excluded patients as failures.
- Reasons for exclusion by arm.
- Whether exclusion decisions were blinded.
If the result is not robust under these analyses, the non-inferiority claim is further weakened.
3. What formulary recommendation would you make to the P&T committee?
Read the worked answer
Recommend not adding drug X to formulary at this time, or adding it only with restriction pending stronger data.
Rationale:
- The primary non-inferiority endpoint was not met.
- Differential exclusion biases toward the new drug.
- The secondary readmission finding is imprecise (95% CI −1% to 5%) and does not support a claim of reduced hospital admissions.
- The press release overstates the evidence.
Committee actions: request the full trial report and protocol, ask for ITT and sensitivity analyses, assess whether a confirmatory trial is planned, and consider total cost including outpatient parenteral therapy. If a formulary pathway is created, specify patient selection, culture follow-up, and reassessment after real-world data are available.