Methodological Challenges and Updated Findings from a Meta-analysis of the Association between Mammographic Density and Breast Cancer

DOI: 10.1158/1055-9965.EPI-17-1175 Published January 2019

Article Figures & Data

Download figure
Open in new tab
Download powerpoint
Figure 1.
A forest plot summarizing all continuous PMD and continuous AMD studies. MLO, mediolateral oblique.
Download figure
Open in new tab
Download powerpoint
Figure 2.
A forest plot summarizing all categorical PMD studies analyzing the 10% to 25%, 25% to 50%, and 50% to 75% PMD categories. African-Am., African American; CC, craniocaudal; FHBC, family history of breast cancer; MLO, mediolateral oblique.
Download figure
Open in new tab
Download powerpoint
Figure 3.
A forest plot summarizing all categorical PMD studies analyzing the greater than 75% PMD category and BIRADS studies.
Download figure
Open in new tab
Download powerpoint
Figure 4.
A forest plot summarizing all categorical AMD studies. CC, craniocaudal.
Download figure
Open in new tab
Download powerpoint
Figure 5.
Comparison across data sets and estimation methods for categorical PMD. The figure compares the log ORs for studies conducted since 2006 (clockwise line pattern) with those from before 2006 (counterclockwise line pattern). It also depicts the widths of the standard confidence intervals (solid lines) and the Bayesian CIs (dotted lines) for the estimates.

Table 1.

Meta-analysis results

	Overall OR estimate (95% CI)	Heterogeneity estimate (95% CI)	I2 statistic in %	Probability of	Predicted OR	Predicted	Number of samples (studies) included	Ratio of width 95% CI and OR
Continuous PMD	1.1317 (1.0656–1.2140)	0.0092 (0.0051–0.0168)	89.32%	99.95%	1.1352	0.0105	15 (8)	0.0131
Continuous AMD	1.1315 (1.0232–1.2725)	0.0082 (0.0032–0.0247)	83.38%	98.65%	1.1372	0.0120	7 (5)	0.0218
Categorical PMD: 10%–25%	1.4366 (1.3250- 1.5549)	0.0679 (0.0072–0.1715)	15.44%	100%	1.4360	0.0852	24 (21)	0.1600
Categorical PMD: 25%–50%	1.9845 (1.7427- 2.2632)	0.1826 (0.0944–0.3229)	51.20%	100%	1.9857	0.2018	21 (19)	0.2622
Categorical PMD: 50%–75%	2.5115 (2.0969- 3.0397)	0.2855 (0.1670–0.4848)	64.92%	100%	2.5197	0.3158	21 (19)	0.3755
Categorical PMD: >75%	3.8436 (2.1917–7.2227)	0.2660 (0.0112–1.2550)	41.00%	99.77%	3.9190	0.5035	6 (6)	1.389
BIRADS–B	1.625 (1.2964–2.1292)	0.1650 (0.0153–0.5304)	36.66%	99.88%	1.6503	0.2444	11 (10)	0.5125
BIRADS–C	1.9536 (1.2228–2.9829)	0.4890 (0.2041–1.0696)	80.57%	99.39%	1.9234	0.5939	11 (10)	0.901
BIRADS–D	2.0044 (1.1221–3.4231)	0.5788 (0.1559–1.3264)	74.1%	98.69%	1.9727	0.6982	11 (10)	1.148
Categorical AMD: 20–35 cm²	1.4849 (1.2289–1.7695)	0.1325 (0.0148–0.3888)	43.53%	99.89%	1.4785	0.1832	9 (8)	0.3641
Categorical AMD: 35–50 cm²	1.7625 (1.5521–2.0296)	0.0686 (0.0044–0.2468)	14.4%	100%	1.7716	0.1097	11 (9)	0.2709
Categorical AMD: >50 cm²	2.2117 (1.9208–2.6363)	0.1160 (0.0144–0.3247)	35.74%	100%	2.2376	0.1562	11 (9)	0.3235

Table 2.

Meta-regression results

		Matching^a	Thresholding^b	Adjustment^c	Invasive^d
Continuous PMD	(95% CI)	0.0114 (0.0012–0.0206)
	OR intercept (95% CI)	1.0847 (1.0422–1.1549)
	OR incl. (95% CI)	1.0972 (1.0435–1.1790)
Categorical PMD: 25%–50%	(95% CI)	0.4077 (0.1726–0.6443)	0.3051 (0.0213–0.5408)
	OR intercept (95% CI)	1.4475 (1.173–1.7678)	1.5928 (1.2995–2.0009)
	OR incl. (95% CI)	2.1760 (1.3873–3.3904)	2.1611 (1.3256–3.4764)
Categorical PMD: 50%–75%	(95% CI)	0.5118 (0.1698–0.8712)	0.4242 (0.0277–0.7616)	0.4284 (0.0453–0.7673)	−0.3970 (-0.8094–0.0001)
	OR intercept (95% CI)	1.6468 (1.223–2.2797)	1.9034 (1.4004–2.5696)	1.9073 (1.3548–2.5157)	3.3720 (2.3980–4.7325)
	OR incl. (95% CI)	2.7465 (1.4359–5.4943)	2.9090 (1.4584–5.6765)	2.9273 (1.4044–5.4690)	2.2670 (1.0612–4.8141)

↵^aMatching is defined as use of matched case–control study setup; reference category: no matching.
↵^bThresholding is defined as use of an automatic thresholding program; reference category: no use of automatic thresholding.
↵^cAdjustment is defined as inclusion of confounders, at a minimum age and/or body mass index (BMI); reference category: no adjustment.
↵^dInvasive is defined as invasive cancer type; reference category: noninvasive cancer.

Supplementary Data

Tables S1 and S2 - Table S1: List of included studies and their effect sizes; Table S2: List of candidate studies
Figures S1, S2 and S3 - Figure S1: Risk gradients for categorical PMD, BIRADS and categorical AMD; Figure S2: Centered posterior density of the overall effect for categorical PMD, BIRADS and categorical AMD; Figure S3: Probability distributions of the effect estimate for categorical PMD
Tables S3 and S4, Figures S4 and S5 - Table S3: Meta-analysis results for re-estimated McCormack and dos Santos (2006) data set; Figure S4: Cumulative probability distributions for categorical PMD; Table S4: Meta-regression results for year of mammography as moderator variable; Figure S5: Predicted effect (change and OR) of PMD (category 10-25%) on BC by year of mammography