A Validation of Methods for the Evaluation of Observational Studies of Screening Mammography: An Exploratory Analysis Based on Simulating Screening Cohorts

Figures & data

Figure 1 Cumulative probability curves of simulated cohort, null model.

Table 1 Simulated Cohort of 100,000 Women Under the Null Model

	Value	Total	Ever Mammogram	Never Mammogram	P-value
		100,000	45,545 (45.5%)	54,455 (54.5%)
Breast cancer	Yes	11,109 (11.1%)	3812 (8.4%)	7297 (13.4%)	<0.0001
Age at breast cancer diagnosis	<50	2369 (21.3%)	0 (0.0%)	2369 (32.5%)	<0.0001
	50–59	2622 (23.6%)	749 (19.6%)	1873 (25.7%)
	60–69	3268 (29.4%)	1684 (44.2%)	1584 (21.7%)
	70–80	2850 (25.7%)	1379 (36.2%)	1471 (20.2%)
Age at first mammogram	Mean (SD)	54.9 (2.9)	54.9 (2.9)	–	n/a
	Median (IQR)	54.9 (52.4–57.4)	54.9 (52.4–57.4)	–
Age at last mammogram	Mean (SD)	67.8 (3.2)	67.8 (3.2)	–	n/a
	Median (IQR)	68.8 (68.2–69.4)	68.8 (68.2–69.4)	–
Vital status	Alive	58,807 (58.8%)	28,668 (62.9%)	30,139 (55.3%)	<0.0001
	Breast	2268 (2.3%)	637 (1.4%)	1631 (3.0%)
	Other COD	38,925 (38.9%)	16,240 (35.7%)	22,685 (41.7%)
Age at death from breast cancer	Mean (SD)	61.7 (12.0)	70.0 (6.5)	58.5 (12.1)	<0.0001
	Median (IQR)	63.0 (53.3–71.7)	70.4 (65.1–75.7)	58.3 (49.6–68.2)	<0.0001

Table 2 Regression Results Using Simulated Cohort of 3 Million Women

Analysis	Description	Null Model (HR = 1.00)	Protective Model (HR = 0.80)
		Hazard Ratio	Hazard Ratio
Analysis 1	Crude analysis	0.57	0.47
Analysis 2	Time-dependent exposure	0.70	0.63
Analysis 3	Matched ever/never mammogram	0.82	0.67
Analysis 4	Matched through pseudo-randomization (controls sampled with replacement)	0.99*	0.86

Note: *Not statistically significant (p > 0.05).

Figure 2 Age-specific breast cancer mortality rates by mammography group, analysis 2 (null model).