Intention-to-treat analysis may be more conservative than per protocol analysis in antibiotic non-inferiority trials: a systematic review

Table 3 Comparison of ITT to PP outcomes in terms of ARR

	PP Median (IQR)	ITT Median (IQR)	Difference ITT – PP Median (IQR)	Wilcoxon signed-rank test p-value	PP analysis is more conservative N (%)
Point estimate	−0.2 (−2.6, 2.2)	0.04 (− 2.6, 2.6)	− 0.01 (− 1.6, 1.9)	0.7025	81 (49.4%)
CI width	12.4 (9.7, 16.6)	13.3 (11.2, 17.5)	0.9 (−0.4, 2.0)	< 0.0001	58 (35.4%)
Lower CI limit	−6.9 (− 10.0, −4.0)	−7.5 (− 10.3, −4.7)	−0.5 (− 1.8, 1.2)	0.0402	72 (43.9%)

A positive value for the difference in CI width indicates less precise estimation of the ARR with ITT analysis. A negative difference for the lower CI limit signifies that the PP lower CI limit lies above the ITT CI limit, so ITT analysis has a more conservative result
ARR Absolute risk reduction, CI Confidence interval, IQR Interquartile range, ITT Intention-to-treat, PP Per-protocol

ISSN: 1471-2288