Table 3. Analysis of ITT-LOCF, LOCF, and completers for handling missing data in the 12 raw datasets using ordinary least squares.

	Actual Analysis			Null	Imposed Treatment Effect
ITT-LOCF	Observed Mean Difference	Observed p-value	Permuted p-value	Empirical α	Mean Difference	Power
RCT 1a	4.07	1×10−5	<10−5	.0479	1.99	.539
RCT 2 a	2.60	.0103	.0113	.0522	2.13	.528
RCT 3	0.62	.2827	.2837	.0510	1.19	.515
RCT 4	5.01	<10−5	<10−5	.0505	2.26	.489
RCT 5	7.17	.0022	.0005	.0469	5.29	.482
RCT 6	0.03	.9551	.9561	.0510	1.17	.414
RCT 7	1.66	.2344	.2252	.0478	2.55	.429
RCT 8	1.66	.4133	.4270	.0485	3.71	.413
RCT 9	0.71	.3881	.3971	.0517	1.57	.443
RCT10	1.90	<10−5	<10−5	.0474	0.56	.537
RCT 11 a	1.30	.0009	.0006	.0531	0.82	.541
RCT 12	1.05	.0001	.0003	.0496	0.53	.481
LOCF
RCT 1a	4.07	1×10−5	<10−5	.0479	1.99	.539
RCT 2 a	2.60	.0103	.0113	.0522	2.13	.528
RCT 3	0.61	.2964	.2976	.0509	1.21	.515
RCT 4	4.95	<10−5	<10−5	.0497	2.34	.502
RCT 5	9.12	.0005	<10−5	.0459	6.09	.511
RCT 6	0.39	.5906	.5868	.0498	1.50	.498
RCT 7	1.81	.2075	.2015	.0471	2.87	.508
RCT 8	2.09	.3731	.3825	.0471	4.48	.457
RCT 9	0.83	.3702	.3730	.0483	1.87	.500
RCT10	1.93	<10−5	<10−5	.0468	0.57	.537
RCT 11 a	1.30	.0009	.0006	.0531	0.82	.541
RCT 12	1.10	.0002	.0004	.0492	0.57	.484
Completers
RCT 1	4.72	2×10−5	<10−5	.0474	1.90	.390
RCT 2	2.71	.0140	.0133	.0513	2.18	.480
RCT 3	0.65	.3767	.3759	.0523	1.34	.424
RCT 4	5.32	<10−5	<10−5	.0470	2.37	.451
RCT 5	9.34	.0038	.0011	.0484	6.12	.375
RCT 6	0.44	.6058	.6052	.0505	1.66	.468
RCT 7	2.14	.1880	.1842	.0475	2.74	.389
RCT 8	0.56	.8261	.8531	.0379	5.28	.512
RCT 9	1.58	.4435	.4403	.0512	2.27	.176
RCT10	0.88	.3411	.3497	.0500	0.65	.107
RCT 11	1.39	.0029	.0021	.0537	0.92	.490
RCT 12	1.42	.0030	.0038	.0500	0.66	.279

Abbreviations: ITT-LOCF, intent-to-treat-last observation carried forward; RCT, randomized controlled trial.

^aIndicates missing data pattern is the same for ITT-LOCF and LOCF. Each permutation test is based on 10,000 permutations of each dataset.