Lifestyle, Behavioral, and Dietary Risk Factors in Relation to Mammographic Breast Density in Women at High Risk for Breast Cancer

The Associations between Intakes of One-Carbon Metabolism-Related Vitamins and Breast Density among Young Women

BACKGROUND

Folate is the primary methyl donor and B vitamins are cofactors for one-carbon metabolism that maintain DNA integrity and epigenetic signatures implicated in carcinogenesis. Breast tissue is particularly susceptible to stimuli in early life. Only limited data are available on associations of one-carbon metabolism-related vitamin intake during youth and young adulthood with breast density, a strong risk factor for breast cancer.

METHODS

Over 18 years in the DISC and DISC06 Follow-up Study, diets of 182 young women were assessed by three 24-hour recalls on five occasions at ages 8 to 18 years and once at 25 to 29 years. Multivariable-adjusted linear mixed-effects regression was used to examine associations of intakes of one-carbon metabolism-related vitamins with MRI-measured percent dense breast volume (%DBV) and absolute dense breast volume (ADBV) at ages 25 to 29 years.

RESULTS

Folate intake in youth was inversely associated with %DBV (Ptrend = 0.006) and ADBV (Ptrend = 0.02). These inverse associations were observed with intake during post-, though not premenarche. In contrast, premenarche vitamin B2 intake was positively associated with ADBV (Ptrend < 0.001). Young adult folate and vitamin B6 intakes were inversely associated with %DBV (all Ptrend ≤ 0.04), whereas vitamins B6 and B12 were inversely associated with ADBV (all Ptrend ≤ 0.04).

CONCLUSIONS

Among these DISC participants intakes of one-carbon metabolism-related vitamins were associated with breast density. Larger prospective studies among diverse populations are needed to replicate these findings.

IMPACT

Our results suggest the importance of one-carbon metabolism-related vitamin intakes early in life with development of breast density and thereby potentially breast cancer risk later in life.

Associations of Vitamins and Related Cofactor Metabolites with Mammographic Breast Density in Premenopausal Women

BACKGROUND

Identifying biological drivers of mammographic breast density (MBD), a strong risk factor for breast cancer, could provide insight into breast cancer etiology and prevention. Studies on dietary factors and MBD have yielded conflicting results. There are, however, very limited data on the associations of dietary biomarkers and MBD.

OBJECTIVE

We aimed to investigate the associations of vitamins and related cofactor metabolites with MBD in premenopausal women.

METHODS

We measured 37 vitamins and related cofactor metabolites in fasting plasma samples of 705 premenopausal women recruited during their annual screening mammogram at the Washington University School of Medicine, St. Louis, MO. Volpara was used to assess volumetric percent density (VPD), dense volume (DV), and nondense volume (NDV). We estimated the least square means of VPD, DV, and NDV across quartiles of each metabolite, as well as the regression coefficient of a metabolite in continuous scale from multiple covariate-adjusted linear regression. We corrected for multiple testing using the Benjamini-Hochberg procedure to control the false discover rate (FDR) at a 5% level.

RESULTS

Participants' mean VPD was 10.5%. Two vitamin A metabolites (β-cryptoxanthin and carotene diol 2) were positively associated, and one vitamin E metabolite (γ-tocopherol) was inversely associated with VPD. The mean VPD increased across quartiles of β-cryptoxanthin (Q1 = 7.2%, Q2 = 7.7%, Q3 = 8.4%%, Q4 = 9.2%; P-trend = 1.77E-05, FDR P value = 1.18E-03). There was a decrease in the mean VPD across quartiles of γ-tocopherol (Q1 = 9.4%, Q2 = 8.1%, Q3 = 8.0%, Q4 = 7.8%; P -trend = 4.01E-03, FDR P value = 0.04). Seven metabolites were associated with NDV: 3 vitamin E (γ-CEHC glucuronide, δ-CEHC, and γ-tocopherol) and 1 vitamin C (gulonate) were positively associated, whereas 2 vitamin A (carotene diol 2 and β-cryptoxanthin) and 1 vitamin C (threonate) were inversely associated with NDV. No metabolite was significantly associated with DV.

CONCLUSION

We report novel associations of vitamins and related cofactor metabolites with MBD in premenopausal women.

Efficient adversarial debiasing with concept activation vector - Medical image case-studies

BACKGROUND

A major hurdle for the real time deployment of the AI models is ensuring trustworthiness of these models for the unseen population. More often than not, these complex models are black boxes in which promising results are generated. However, when scrutinized, these models begin to reveal implicit biases during the decision making, particularly for the minority subgroups.

METHOD

We develop an efficient adversarial de-biasing approach with partial learning by incorporating the existing concept activation vectors (CAV) methodology, to reduce racial disparities while preserving the performance of the targeted task. CAV is originally a model interpretability technique which we adopted to identify convolution layers responsible for learning race and only fine-tune up to that layer instead of fine-tuning the complete network, limiting the drop in performance RESULTS:: The methodology has been evaluated on two independent medical image case-studies - chest X-ray and mammograms, and we also performed external validation on a different racial population. On the external datasets for the chest X-ray use-case, debiased models (averaged AUC 0.87 ) outperformed the baseline convolution models (averaged AUC 0.57 ) as well as the models trained with the popular fine-tuning strategy (averaged AUC 0.81). Moreover, the mammogram models is debiased using a single dataset (white, black and Asian) and improved the performance on an external datasets (averaged AUC 0.8 to 0.86 ) with completely different population (primarily Hispanic patients).

CONCLUSION

In this study, we demonstrated that the adversarial models trained only with internal data performed equally or often outperformed the standard fine-tuning strategy with data from an external setting. The adversarial training approach described can be applied regardless of predictor's model architecture, as long as the convolution model is trained using a gradient-based method. We release the training code with academic open-source license - https://github.com/ramon349/JBI2023_TCAV_debiasing.