Table 1.

Improvement in the Cox regression model by allowing different vitamin C effects for two consecutive time periods.

Periods of vitamin C effect (days)	Improvement*
	${\chi }_{(1\text{df})}^2$	P
0-28	Reference
0-1 and 2-28	4.54	0.033
0-2 and 3-28	4.82	0.028
0-3 and 4-28	6.52	0.011
0-4 and 5-28	7.65	0.006
0-5 and 6-28	7.22	0.007
0-6 and 7-28	3.49	0.061
0-7 and 8-28	1.41	0.236

^*The improvement in the Cox regression model over the reference model was tested by the likelihood ratio test. The reference model estimates that vitamin C decreases mortality uniformly by RR = 0.57 (95% CI: 0.34-0.95; P = 0.028). Compared with the model with no vitamin C effect, the best-fitting two-period model (0–4 days and 5–28 days) improves the Cox regression with ${\chi }_{(2df)}^{\mathit{\text{2}}}$ = 12.5 (P = 0.002). See the Supplementary File for the extraction of mortality data from Figure 3 of the CITRIS-ALI trial report (1) and the calculations for this table.