Declaration of Conflict of Interest Population-Based Assessment of - - PowerPoint PPT Presentation

6/9/2017 1

Population-Based Assessment of the Effect of MRI Background Parenchymal Enhancement on Future Primary Breast Cancer Risk

Vignesh Arasu, Diana Miglioretti, Brian Sprague, Nila Alsheik, Diana Buist, Louise Henderson, Sally Herschorn, Janie Lee, Tracy Onega, Garth Rauscher, Karen Wernli, Constance Lehman, Karla Kerlikowske June 9th, 2017 8th International Workshop on Breast Densitometry and Cancer Risk Assessment, San Francisco, CA

Declaration of Conflict

• f Interest
• None of the authors have any relevant conflict of interest

to declare

Background on BPE

• Breast MRI is primarily used to screen

women at very high breast cancer risk

• Background parenchymal enhancement

(BPE) describes contrast uptake of normal tissue on breast MRI

• Codified in BI-RADS MRI Atlas (5th edition)

– Four qualitative categories of increasing BPE – Qualifies degree of tumor masking on MRI

Minimal Mild Moderate Marked

Degrees of BPE on MRI

6/9/2017 2

BPE as imaging biomarker

• Recognition as novel imaging biomarker
• Reflects breast vascularity, permeability, and

peri-vascular factors present in stroma

– Whereas breast density reflects ducts and stroma

• Dynamic with hormonal variation

– Influenced by age, menstrual cycle, menopause, and hormone therapy

• Single-center studies demonstrate high BPE

is associated with increased breast cancer risk (see King et al. 2011, Dontchos et al 2015)

Research objectives

• Evaluate MRI BPE risk prediction on future

primary breast cancer risk

• Evaluate comparative risk prediction to

breast density

Methods

• Observational cohort study from six US

Breast Cancer Surveillance Consortium (BCSC) registry sites

– 40 breast imaging facilities, 123 radiologists

• Study period: 2005-2014
• Predictor: BI-RADS BPE assessment

– Cases: BPE at least 3 months prior to diagnosis

• Outcome: Cancer diagnosis ascertained

through linkage with state or regional cancer registries and pathology databases

BCSC registries

Active registries Historic registries

6/9/2017 3

Methods

• Covariates obtained prospectively via clinical

history surveys and medical record data

• Primary analysis: Cox proportional hazards

model using baseline BPE, stratified by BCSC registry site and MRI indication, adjusted for age

• Secondary analysis:

– Partly conditional repeated measures model – Dichotomized BPE measurement for improved interpretation biologically, inter-rater agreement

• (min BPE vs. mild/mod/marked BPE)

Results

• 4976 MRI examinations among 3,358

women included

• 135 women developed breast cancer

– 99 invasive (73%) – 27 DCIS (27%)

• Average of 2.5 years of follow-up
• 60% of examinations were screening MRI

Results: Baseline Characteristics

• Age: over 80% less than 60 years old
• Race/ethnicity: 80% were white
• Cases had a higher proportion of

– Pre-menopausal women (54% vs 49%) – 1st degree breast cancer family history (64%

• vs. 59%)

Table: BPE risk at baseline

BPE

Cancer Cohort Control Cohort Hazard Ratio (95% CI) (N=135 women) (N=3223 women) N % N %

Minimal 25 19% 1,080 34% Reference Mild 41 30% 982 30% 1.91 (1.12, 3.27) Moderate 42 31% 763 24% 2.23 (1.29, 3.84) Marked 27 20% 398 12% 2.69 (1.49, 4.87)

6/9/2017 4

Table: Breast Density Risk

Breast Density

Cancer Cohort Control Cohort Hazard Ratio (95% CI) (N=135 women) (N=3223 women) N % N %

Almost entirely fat 4 3% 124 5% 1.12 (0.38, 3.25) Scattered fibroglandular 24 20% 738 30% Reference Heterogeneously dense 53 45% 1,089 44% 1.35 (0.81, 2.24) Extremely dense 38 32% 550 22% 1.85 (1.06, 3.21)

Table: BPE risk, with multiple observations BPE

Cancer Cohort Control Cohort Hazard Ratio (95% CI) (N=200 obs) (N=4,776 obs) N % N %

Minimal 45 23% 1,679 35% Reference Mild 61 31% 1,435 30% 1.56 (1.05, 2.33) Moderate 53 27% 1,080 23% 1.57 (1.00, 2.48) Marked 41 21% 582 12% 1.97 (1.14, 3.39)

Table: BPE versus Breast Density risk

Not dense Dense Minimal BPE

Reference 1.49 (0.61, 3.66)

Mild, Moderate, Marked BPE

2.46 (1.04, 5.81) 3.03 (1.40, 6.53)

• Test for interaction between BPE and density is

not statistically significant (p=0.72)

Table: Stratified analysis of other risk subgroups

Other risk subgroups Hazard Ratio (95% CI) P-value Family history 2.58 (1.52, 4.36) 0.11 No family history 1.90 (1.06, 3.43) Pre-menopausal 2.76 (1.62, 4.68) 0.19 Post-menopausal 2.21 (1.25, 3.90) BCSC risk <1.67% 2.21 (1.14, 4.25) 0.054 BCSC risk >=1.67% 2.76 (1.50, 5.07)

Dichotomized BPE overall hazard ratio: 2.17 (1.34, 3.50) (Minimal vs. Mild/Moderate/Marked BPE)

6/9/2017 5 Table: Stratified analysis of invasive versus noninvasive disease

Subgroups Hazard Ratio (95% CI) DCIS 1.42 (0.60, 3.37) Invasive 2.58 (1.50, 4.44)

Dichotomized BPE overall hazard ratio: 2.17 (1.34, 3.50) (Minimal vs. Mild/Moderate/Marked BPE)

Discussion

• High levels of BPE predict increased risk
• f primary breast cancer
• In a high risk population, BPE predicted

breast cancer risk independent of breast density

• BPE had a stronger association with

cancer than breast density

Discussion (continued)

• BPE risk prediction is independent of other

breast cancer risk factors (age, family history, benign breast biopsies, menopausal status, and BCSC risk score)

• BPE is more strongly associated with

invasive cancer than DCIS

• Largest study to date in population-based

cohort among diverse facilities, replicates prior results

Limitations

• Qualitative BPE is subjective and prone to

inter and intra-observer differences

– Similar to breast density

• Unclear phase of menstrual cycle
• BPE “missing data”

– 2005: 99% – 2014: 15%

6/9/2017 6

Conclusion

• BPE is a predictor of future breast cancer

risk independent of breast density with stronger associations with invasive cancer than DCIS

• BPE should be considered for

incorporation into risk prediction models for high-risk women undergoing MRI

Acknowledgements

• Participating women
• BCSC investigators
• Radiologists and mammography facilities
• Data collection for this work was

supported by the National Cancer Institute-funded Breast Cancer Surveillance Consortium (P01 CA154292 and HHSN261201100031C) Additional slides: Sensitivity analyses for first BPE measure

Covariate Hazard Ratio (95% CI)

Family history 2.32 (1.40, 3.84) Benign breast biopsy 2.27 (1.36, 3.81) Family History and benign breast biopsy 2.37 (1.40, 4.02) Menopausal status 2.50 (1.39, 4.52) Family History, benign breast biopsy, menopausal status 2.48 (1.37, 4.49)