Imputation strategies when a continuous outcome is to be dichotomized for responder analysis: a simulation study

Table 6 Comparison of imputation results for a clinical trial example. Treatment arm: n = 84; Placebo arm: n = 88

% Missing		NR imputation responders*		IBD imputation responders*		DTI imputation responders*
	Arm	%	Difference	%	Difference	% (95% CI)	Difference
25%	Drug	46.4	19.1 (p = 0.009)	56.3 (45.9, 68.7)	21.9 (5.3, 36.6) (p = 0.009)	56.6 (45.7, 67.5)	21.1 (5.8, 36.5) (p = 0.007)
25%	Placebo	27.3	19.1 (p = 0.009)	36.3 (25.7, 47.0)	21.9 (5.3, 36.6) (p = 0.009)	35.5 (24.6, 46.3)	21.1 (5.8, 36.5) (p = 0.007)
40%	Drug	38.1	13.1 (p = 0.064)	60.5 (48.2, 72.8)	22.6 (6.2, 39.1) (p = 0.007)	59.2 (48.1, 70.3)	24.6 (8.9, 40.5) (p = 0.002)
40%	Placebo	25.0	13.1 (p = 0.064)	37.8 (26.2, 49.5)	22.6 (6.2, 39.1) (p = 0.007)	34.6 (23.5, 45.7)	24.6 (8.9, 40.5) (p = 0.002)

*Response is defined as improvement ≥6 on the HAMD-17 total score from baseline to week 6

ISSN: 1471-2288