Table 2.

Sensitivity of HRs for the association between AN69ST-CRRT and mortality risk (relative to standard-only-CRRT), adjusted for unmeasured confounding

p1	p0	HRcd
		0.9	0.7	0.5	0.3	0.1
0.65	0.2	0.81	0.76	0.87	1.03	1.28
0.6	0.25	0.77	0.73	0.81	0.92	1.10
0.55	0.3	0.73	0.71	0.76	0.83	0.94
0.5	0.35	0.70	0.68	0.72	0.76	0.81
0.45	0.4	0.67	0.66	0.67	0.68	0.70
0.4	0.45	0.63	0.64	0.63	0.62	0.60
0.35	0.5	0.60	0.62	0.59	0.56	0.52
0.3	0.55	0.58	0.60	0.55	0.51	0.45
0.25	0.6	0.55	0.58	0.52	0.46	0.39
0.2	0.65	0.52	0.56	0.49	0.41	0.33

HRs are computed as point estimates only. The base model assumes an HR of 0.65, adjusted for known confounders, when unmeasured confounders are present but unaccounted for. When an unmeasured confounder of varying strength and prevalence is introduced into the models, the hazard ratios are only sensitive to strong unmeasured confounding, with an effect that remains after adjustment for only weak-moderate confounding.

HRs, hazard ratios; CRRT, continuous renal replacement therapy; HR_CD, hazard ratio for the association between unmeasured confounder and mortality risk; p1, prevalence of unmeasured confounder among those on AN69ST-CRRT; p0, prevalence of unmeasured confounder among those on to standard-only-CRRT.