Confidence interval of risk difference by different statistical methods and its impact on the study conclusion in antibiotic non-inferiority trials

Table 2 Confidence interval based on the five commonly used statistical methods

	CI width in % risk difference Median IQR	CI width difference compared to Wald method Median IQR^a	Non-inferiority shown^b N (%)	Most conservative CI^c
Wald	13.0 (10.8, 17.4)	Reference	170 / 216 (78.7%)	15 (6.7%)
Agresti-Caffo	13.3 (10.9, 18.5)	0.05 (0, 0.2)	169 / 216 (78.2%)	1 (0.5%)
SCAS	13.4 (11.1, 18.9)	0.2 (0.05, 0.4)	166 / 216 (76.9%)	53 (23.7%)
Newcombe	13.6 (11.1, 18.9)	0.1 (-0.01, 0.4)	165 / 216 (76.4%)	10 (4.5%)
Miettinen-Nurminen	13.6 (11.1, 19.0)	0.2 (0.04, 0.5)	165 / 216 (76.4%)	145 (64.7%)

CI confidence interval, IQR interquartile range, SCAS skewness-corrected asymptotic score method
^aDifference calculated as CI width minus CI width by the Wald method, so a positive number suggests a CI that is wider than the CI by the Wald method
^bThere were 8 cases where non-inferiority margin was not specified
^cMost conservative CI defined by the widest CI of the five methods for a given study

ISSN: 1745-6215