Statistical methods for elimination of guarantee-time bias in cohort studies: a simulation study

Table 3 Bias and MSE of Hazard ratio for estimation of the association between cumulative drug dose and outcome

β ₀	Cumulative dose	Cox regression		Time-dependent Cox regression		Landmark analyses
						τ = 5		τ = 7
		Bias	MSE	Bias	MSE	Bias	MSE	Bias	MSE
log(0.015)	Low	0.0732	0.0035	0.0009	0.0041	0.0673	0.0062	0.0416	0.0105
	Moderate	0.2735	0.0019	0.0005	0.0043	0.0641	0.0043	0.0469	0.0068
	High	0.6210	0.0004	0.0419	0.0034	0.0437	0.0174	0.0429	0.0061
log(5^∗0.015)	Low	0.0822	0.0009	0.0007	0.0012	0.0614	0.0020	0.0352	0.0034
	Moderate	0.3002	0.0004	−0.0024	0.0012	0.0585	0.0013	0.0455	0.0023
	High	0.6263	0.0001	0.0392	0.0011	0.0377	0.0048	0.0366	0.0019
log(10^∗0.015)	Low	0.0944	0.0006	0.0011	0.0008	0.0565	0.0016	0.0300	0.0029
	Moderate	0.3349	0.0002	−0.0032	0.0009	0.0524	0.0011	0.0461	0.0020
	High	0.6331	0.0001	0.0360	0.0009	0.0360	0.0038	0.0338	0.0016

β ₀ values represent disease incidence rate, MSE Mean squared error, τ Landmark time

ISSN: 1471-2288