. 2021 Nov 16;4(11):e2134627. doi: 10.1001/jamanetworkopen.2021.34627

Table 2. Estimated Comparative Treatment Outcomes of Natalizumab vs Rituximab and Dimethyl Fumarate vs Fingolimod Based on Registry-Annotated Treatment Groups and Adjustment for Full EHR Features.

Treatment	Estimate (95% CI)	P value^a	E-value (E-value*)^b
Natalizumab vs rituximab ^c
Outcome^d
Difference in 1-y relapse rate^e	0.080 (0.013 to 0.137)	.02 (.02)	1.50 (1.13)
Difference in 2-y relapse rate^e	0.132 (0.043 to 0.189)	.004 (.004)	2.26 (1.31)
Relative risk of 2-y non-relapse (from time-to-relapse analysis)^e	0.903 (0.822 to 0.944)	<.001 (.01)	1.11 (1.06)
Dimethyl fumarate vs fingolimod ^f
Outcome^d
Difference in 1-y relapse rate^g	0.028 (–0.031 to 0.084)	.38 (.38)	NA
Difference in 2-y relapse rate^g	0.071 (0.008 to 0.128)	.03 (.08)	NA
Relative risk of 2-y non-relapse (from time-to-relapse analysis)^g	0.957 (0.884 to 1.035)	.28 (.50)	NA

Abbreviations: EHR, electronic health record; NA, not applicable.

^{^a}

P values in parentheses are adjusted for multiple testing among the 3 analyses with the same treatment groups and feature set (eMethods and eResults in the Supplement).

^{^b}

E-values assess the strength of the unmeasured confounding that would change the direction of association, while E-values* assess the strength of the unmeasured confounding that would negate the significance of the observed associations. Thus, an E-value (or E-value*) indicates that residual confounding could explain the observed association if there exists an unmeasured covariate with a relative risk association at least as large as the E-value. E-values were computed for significant associations and were NA for nonsignificant findings (eMethods and eResults in the Supplement).

^{^c}

Rituximab was used as the reference group.

^{^d}

For each relapse outcome, we applied 2 adjustments, outcome regression and propensity scores, to derive the doubly robust estimation.

^{^e}

With rituximab as the reference, a positive difference in the 1-year or 2-year relapse rate or a relative risk (ratio) of non-relapse rates less than 1 would indicate higher relapse probability of natalizumab.

^{^f}

Dimethyl fumarate was used as the reference group.

^{^g}

With dimethyl fumarate as the reference, a positive difference in the 1-year or 2-year relapse rate or a relative risk (ratio) of non-relapse rates less than 1 would indicate higher relapse probability of fingolimod.