Table 4.

The estimates of the treatment‐age interaction effects in the simulated data considering trial‐level treatment confounding.

Size*	V 1	V 2	Model	Mean (SD)		Bias	MSE	Coverage	Model	Mean (SD)			Bias		MSE		Coverage
				β T	τ	β T	β T	β T		β W	β A	τ	β W	β A	β W	β A	β W	β A
Large	20	10	(1)	0.023 (0.003)		0.013	<0.001	0.018	(3)	0.010 (0.003)	0.040 (0.006)		0	0.030	<0.001	0.001	0.958	0.003
Large	20	10	(2)	0.013 (0.003)	0.112 (0.040)	0.003	<0.001	0.848	(4)	0.010 (0.003)	0.041 (0.006)	0.027(0.019)	0	0.031	<0.001	0.001	0.958	0.010
Large	20	5	(1)	0.032 (0.004)		0.022	<0.001	0.003	(3)	0.010 (0.006)	0.039 (0.006)		0	0.029	<0.001	0.001	0.965	0.002
Large	20	5	(2)	0.021(0.006)	0.068 (0.043)	0.011	<0.001	0.418	(4)	0.010 (0.006)	0.039 (0.006)	0.025(0.018)	0	0.029	<0.001	0.001	0.964	0.005
Large	10	10	(1)	0.020 (0.003)		0.010	<0.001	0.131	(3)	0.010 (0.003)	0.068 (0.012)		0	0.058	<0.001	0.003	0.949	0.003
Large	10	10	(2)	0.011 (0.003)	0.121 (0.037)	0.001	<0.001	0.927	(4)	0.010 (0.003)	0.068 (0.012)	0.026(0.018)	0	0.058	<0.001	0.004	0.946	0.005
Large	10	5	(1)	0.035 (0.006)		0.025	0.001	0.018	(3)	0.010 (0.007)	0.06 (0.012)		0	0.058	<0.001	0.003	0.945	0.001
Large	10	5	(2)	0.015 (0.007)	0.109 (0.041)	0.005	<0.001	0.826	(4)	0.010 (0.007)	0.068 (0.012)	0.026(0.019)	0	0.058	<0.001	0.003	0.945	0.004
Small	20	10	(1)	0.022 (0.007)		0.012	<0.001	0.38	(3)	0.010 (0.007)	0.041 (0.017)		0	0.031	<0.001	0.001	0.942	0.121
Small	20	10	(2)	0.015 (0.008)	0.085 (0.078)	0.005	<0.001	0.778	(4)	0.010 (0.007)	0.041 (0.017	0.015(0.026)	0	0.031	<0.001	0.001	0.943	0.184
Small	20	5	(1)	0.030 (0.010)		0.020	0.001	0.173	(3)	0.010 (0.013)	0.039 (0.017)		0	0.029	<0.001	0.001	0.939	0.115
Small	20	5	(2)	0.024 (0.012)	0.050 (0.073)	0.014	<0.001	0.464	(4)	0.010 (0.013)	0.039 (0.017)	0.017(0.028)	0	0.029	<0.001	0.001	0.939	0.185
Small	10	10	(1)	0.019 (0.007)		0.009	<0.001	0.665	(3)	0.010 (0.006)	0.071 (0.032)		0	0.061	<0.001	0.005	0.953	0.103
Small	10	10	(2)	0.012 (0.007)	0.106 (0.069)	0.002	<0.001	0.917	(4)	0.010 (0.006)	0.071 (0.032)	0.019(0.030)	0	0.061	<0.001	0.005	0.952	0.179
Small	10	5	(1)	0.032 (0.012)		0.022	0.001	0.354	(3)	0.009 (0.013)	0.069 (0.031)		−0.001	0.059	<0.001	0.004	0.961	0.109
Small	10	5	(2)	0.019 (0.014)	0.086 (0.075)	0.009	<0.001	0.808	(4)	0.009 (0.013)	0.069 (0.031	0.016(0.027)	−0.001	0.059	<0.001	0.004	0.961	0.185

In all settings, the true interaction between the log hazard ratio (treatment effect) and age was 0.01, and so, if they reflected this, the mean estimates of β _T, β _W and β _A should be 0.01.

^* ‘Large’: J = 10 studies, N = 250 patients; ‘Small’: J = 5 studies, N = 250 patients.

MSE, mean square error; and SD, standard deviation of the 1000 parameter estimates.