Collider and reporting biases involved in the analyses of cause of death associations in death certificates: an illustration with cancer and suicide

Table 2 Suicide ORs by cancer site in observed and simulated mortality data, and estimated bias magnitudes

Cancer site	French mortality data		Simulation #1: independence		Simulation #2: RR from Fang et al. [4]		Fang et al. study[4]	Collider bias^b	Reporting bias^c
Cancer site	OR	[95% CI]	OR	[95% CI]	OR	[95% CI]	RR	Collider bias^b	Reporting bias^c
No cancer^a	1.00		1.00		1.00		1.0	1.0	1.0
Prostate	0.24	[0.22;0.26]	0.71	[0.68;0.75]	1.14	[1.10;1.18]	1.9	1.7	4.7
Cutaneous melanoma	0.01	[0.01;0.02]	0.61	[0.53;0.70]	0.77	[0.68;0.86]	1.4	1.8	64
Breast	0.04	[0.03;0.04]	0.58	[0.52;0.64]	0.85	[0.78;0.92]	1.6	1.9	24
Colorectal	0.05	[0.05;0.06]	0.35	[0.33;0.37]	0.50	[0.48;0.52]	1.6	3.2	9.4
Oesophageal, liver, pancreatic	0.03	[0.03;0.03]	0.17	[0.16;0.19]	0.55	[0.53;0.58]	4.5	8.1	20
Lung	0.02	[0.02;0.02]	0.14	[0.13;0.15]	0.36	[0.35;0.38]	3.3	9.1	17
Central nervous system	0.01	[0.01;0.01]	0.11	[0.09;0.14]	0.25	[0.22;0.28]	2.3	9.3	35

Logistic regression models adjusted for age (B-spline of degree 3), gender, and region of death; mainland France, 2000–2010
OR Odds ratio, RR Relative risk, 95% CI 95% Confidence interval
^aExcluding other cancer sites, multiple cancers, and haematological malignancies. The magnitude of the biases was estimated from the following ratios
^bCollider bias = RR from Fang et al./OR estimated from the data of simulation #2
^cReporting bias = OR estimated from the data of simulation #2/OR estimated from observed deaths

ISSN: 1478-7954