Comparison of statistical methods for the analysis of recurrent adverse events in the presence of non-proportional hazards and unobserved heterogeneity: a simulation study

Table 3 Comparing sharing frailty models for analysis of recurrent AEs among pregnant women on IPTp treatment in Malawi

Estimate	ISF-PH	LSF-NPHS1	LSF-NPHS2^a	LSF-NPHS3	LSF-NPHS4^b
Log HR at 14 days (SE)	0.2281 (0.0499)	0.4758 (0.3572)	0.4343 (0.3724)	0.4648 (0.4033)	0.4404 (0.3710)
log HR at 60 days (SE)	0.2281 (0.0499)	0.0236 (0.1240)	0.2011 (0.2562)	0.2251 (0.2831)	0.1837 (0.2441)
Marginal log HR (SE)	0.2281 (0.0499)	0.2328 (0.0517)	0.3090 (0.0769)	0.3360 (0.0850)	0.3024 (0.0746)
Frailty variance	0.0324	0.0476	0.4591	0.6440	0.3977
AIC	6562.57	19,922.40	19,658.30	19,651.20	19,570.70

^aLognormal shared frailty model with non-proportional hazard and restricted cubic splines where the baseline hazard is modelled with 2 degrees of freedom and the time-dependent treatment effects is modelled with 1 degree of freedom
^bLognormal shared frailty model with non-proportional hazard and restricted cubic splines where the baseline hazard is modelled with 4 degrees of freedom and the time-dependent treatment effects is modelled with 1 degree of freedom

ISSN: 1471-2288