Comparison of Mammography AI Algorithms with a Clinical Risk Model for 5-year Breast Cancer Risk Prediction: An Observational Study.

Background Although several clinical breast cancer risk models are used to guide screening and prevention, they have only moderate discrimination. Purpose To compare selected existing mammography artificial intelligence (AI) algorithms and the Breast Cancer Surveillance Consortium (BCSC) risk model for prediction of 5-year risk. Materials and Methods This retrospective case-cohort study included data in women with a negative screening mammographic examination (no visible evidence of cancer) in 2016, who were followed until 2021 at Kaiser Permanente Northern California. Women with prior breast cancer or a highly penetrant gene mutation were excluded. Of the 324 009 eligible women, a random subcohort was selected, regardless of cancer status, to which all additional patients with breast cancer were added. The index screening mammographic examination was used as input for five AI algorithms to generate continuous scores that were compared with the BCSC clinical risk score. Risk estimates for incident breast cancer 0 to 5 years after the initial mammographic examination were calculated using a time-dependent area under the receiver operating characteristic curve (AUC). Results The subcohort included 13 628 patients, of whom 193 had incident cancer. Incident cancers in eligible patients (additional 4391 of 324 009) were also included. For incident cancers at 0 to 5 years, the time-dependent AUC for BCSC was 0.61 (95% CI: 0.60, 0.62). AI algorithms had higher time-dependent AUCs than did BCSC, ranging from 0.63 to 0.67 (Bonferroni-adjusted P < .0016). Time-dependent AUCs for combined BCSC and AI models were slightly higher than AI alone (AI with BCSC time-dependent AUC range, 0.66-0.68; Bonferroni-adjusted P < .0016). Conclusion When using a negative screening examination, AI algorithms performed better than the BCSC risk model for predicting breast cancer risk at 0 to 5 years. Combined AI and BCSC models further improved prediction. © RSNA, 2023 Supplemental material is available for this article.

Initiation and adherence to adjuvant endocrine therapy among urban, insured American Indian/Alaska Native breast cancer survivors.

BACKGROUND: It has been shown that racial/ethnic disparities exist with regard to initiation of and adherence to adjuvant endocrine therapy (AET). However, the relationship among American Indian/Alaska Native (AIAN) individuals is poorly understood, particularly among those who reside in urban areas. We evaluated whether AET initiation and adherence were lower among AIAN individuals than those of other races/ethnicities who were enrolled in the Kaiser Permanente of Northern California (KPNC) health system. METHODS: We identified 23,680 patients from the period 1997 to 2014 who were eligible for AET (first primary, stage I-III, hormone receptor-positive breast cancer) and used KPNC pharmacy records to identify AET prescriptions and refill dates. We assessed AET initiation (≥1 filled prescription within 1 year of diagnosis) and AET adherence (proportion of days covered ≥80%) every year up to 5 years after AET initiation. RESULTS: At the end of the 5-year follow-up period, 83% of patients were AET initiators, and 58% were AET adherent. Compared with other races/ethnicities, AIAN women had the second-lowest rate of AET initiation (non-Hispanic Black [NHB], 78.0%; AIAN, 78.6%; Hispanic, 83.0%; non-Hispanic White [NHW], 82.5%; Asian/Pacific Islander [API], 84.7%), the lowest rate of AET adherence after 1 year and 5 years of follow-up (70.3% and 50.8%, respectively), and the greatest annual decline in AET adherence during the 4- to 5-year period of follow-up (a 13.8% decrease in AET adherence [from 64.6% to 50.8%]) after initiation of AET. In adjusted multivariable models, AIAN, Hispanic, and NHB women were less likely than NHW women to be AET adherent. At the end of the 5-year period, total underutilization (combining initiation and adherence) in AET-eligible patients was greatest among AIAN (70.6%) patients, followed by NHB (69.6%), Hispanic (63.2%), NHW (58.7%), and API (52.3%) patients, underscoring the AET treatment gap. CONCLUSION: Our results suggest that AET initiation and adherence are particularly low for insured AIAN women.

Patterns of medication adherence in a multi-ethnic cohort of prevalent statin users diagnosed with breast, prostate, or colorectal cancer.

PURPOSE: To investigate the implications of a cancer diagnosis on medication adherence for pre-existing comorbid conditions, we explored statin adherence patterns prior to and following a new diagnosis of breast, colorectal, or prostate cancer among a multi-ethnic cohort. METHODS: We identified adults enrolled at Kaiser Permanente Northern California who were prevalent statin medication users, newly diagnosed with breast, colorectal, or prostate cancer between 2000 and 2012. Statin adherence was measured using the proportion of days covered (PDC) during the 2-year pre-cancer diagnosis and the 2-year post-cancer diagnosis. Adherence patterns were assessed using generalized estimating equations, for all cancers combined and stratified by cancer type and race/ethnicity, adjusted for demographic, clinical, and tumor characteristics. RESULTS: Among 10,177 cancer patients, statin adherence decreased from pre- to post-cancer diagnosis (adjusted odds ratio (ORadj):0.91, 95% confidence interval (95% CI):0.88-0.94). Statin adherence decreased from pre- to post-cancer diagnosis among breast (ORadj:0.94, 95% CI:0.90-0.99) and colorectal (ORadj:0.79, 95% CI:0.74-0.85) cancer patients. No difference in adherence was observed among prostate cancer patients (ORadj:1.01, 95% CI:0.97-1.05). Prior to cancer diagnosis, adherence to statins was generally higher among non-Hispanic whites and multi-race patients than other groups. However, statin adherence after diagnosis decreased only among these two populations (ORadj:0.85, 95% CI:0.85-0.92 and ORadj:0.86, 95% CI:0.76-0.97), respectively. CONCLUSIONS: We found substantial variation in statin medication adherence following diagnosis by cancer type and race/ethnicity among a large cohort of prevalent statin users in an integrated health care setting. IMPLICATIONS FOR CANCER SURVIVORS: Improving our understanding of comorbidity management and polypharmacy across diverse cancer patient populations is warranted to develop tailored interventions that improve medication adherence and reduce disparities in health outcomes.

Feasibility of analyzing DNA copy number variation in breast cancer tumor specimens from 1950 to 2010: how old is too old?

PURPOSE: The purpose of the study was to assess the feasibility of quantifying long-term trends in breast tumor DNA copy number variation (CNV) profiles. METHODS: We evaluated CNV profiles in formalin-fixed paraffin-embedded (FFPE) tumor specimens from 30 randomly selected Kaiser Permanente Northern California health plan women members diagnosed with breast cancer from 1950 to 2010. Assays were conducted for five cases per decade who had available tumor blocks and pathology reports. RESULTS: As compared to the tumors from the 1970s to 2000s, the older tumors dating back to the 1950s and 1960s were much more likely to (1) fail quality control, and (2) have fewer CNV events (average 23 and 31 vs. 58 to 69), fewer CNV genes (average 5.1 and 3.7k vs. 8.1 to 10.3k), shorter CNV length (average 2,440 and 3,300k vs. 5,740 to 9,280k), fewer high frequency Del genes (37 and 25% vs. 54 to 76%), and fewer high frequency high_Amp genes (20% vs. 56 to 73%). On average, assay interpretation took an extra 60 min/specimen for cases from the 1960s versus 20 min/specimen for the most recent tumors. CONCLUSIONS: Assays conducted in the mid-2010s for CNVs may be feasible for FFPE tumor specimens dating back to the 1980s, but less feasible for older specimens.

Metformin use and lung cancer risk in patients with diabetes.

Methodologic biases may explain why observational studies examining metformin use in relation to lung cancer risk have produced inconsistent results. We conducted a cohort study to further investigate this relationship, accounting for potential biases. For 47,351 patients with diabetes ages ≥40 years, who completed a health-related survey administered between 1994 and 1996, data on prescribed diabetes medications were obtained from electronic pharmacy records. Follow-up for incident lung cancer occurred from January 1, 1997, until June 30, 2012. Using Cox regression, we estimated lung cancer risk associated with new use of metformin, along with total duration, recency, and cumulative dose (all modeled as time-dependent covariates), adjusting for potential confounding factors. During 428,557 person-years of follow-up, 747 patients were diagnosed with lung cancer. No association was found with duration, dose, or recency of metformin use and overall lung cancer risk. Among never smokers, however, ever use was inversely associated with lung cancer risk [HR, 0.57; 95% confidence interval (CI), 0.33-0.99], and risk appeared to decrease monotonically with longer use (≥5 years: HR, 0.48; 95% CI, 0.21-1.09). Among current smokers, corresponding risk estimates were >1.0, although not statistically significant. Consistent with this variation in effect by smoking history, longer use was suggestively associated with lower adenocarcinoma risk (HR, 0.69; 95% CI, 0.40-1.17), but higher small cell carcinoma risk (HR, 1.82; 95% CI, 0.85-3.91). In this population, we found no evidence that metformin use affects overall lung cancer risk. The observed variation in association by smoking history and histology requires further confirmation.

Analyzing historical trends in breast cancer biomarker expression: a feasibility study (1947-2009).

BACKGROUND/OBJECTIVES: Determining long-term trends in tumor biomarker expression is essential for understanding aspects of tumor biology amenable to change. Limiting the availability of such data, currently used assays for biomarkers are relatively new. For example, assays for the estrogen receptor (ER), which are the oldest, extend back only to the 1970s. METHODS: To extend scant knowledge about the feasibility of obtaining long-term data on tumor biomarkers, we randomly selected 60 breast cancer cases (10 per decade) diagnosed between 1947-2009 among women members of the Kaiser Permanente Northern California health plan to obtain and analyze their formalin-fixed paraffin-embedded (FFPE) tumor specimens. For each tumor specimen, we created duplicate tissue microarrays for analysis. RESULTS: We located tumor blocks and pathology reports for 50 of the 60 cases (83%), from which we randomly sampled 5 cases per decade for biomarker analysis (n = 30). All 30 cases displayed excellent morphology and exhibited biomarkers compatible with histologic type and grade. Test-retest reliability was also excellent: 100% for ER; 97% for human epidermal growth factor receptor 2 and epidermal growth factor receptor; 93% for progesterone receptor and cytokeratin 5/6; and 90% for Ki67 and molecular phenotype; the kappa statistic was excellent (>0.9) for 4 of the 7 biomarkers, strong (0.6-0.8) for 2, and fair for only 1 (owing to low prevalence). CONCLUSIONS: These results indicate immunostaining for biomarkers commonly used to evaluate breast cancer biology and assign surrogate molecular phenotypes can reliably be employed on archival FFPE specimens up to 60 years old.

Serial glycosylated hemoglobin levels and risk of colorectal neoplasia among patients with type 2 diabetes mellitus.

BACKGROUND: Hyperglycemia may increase the risk of colorectal neoplasia by serving as an energy source for neoplastic growth. We sought to determine whether glycemic control measured by serial hemoglobin A1c (HbA1c) was associated with the risk of colorectal adenoma. METHODS: Among a cohort of patients with type 2 diabetes mellitus who received health care within the Kaiser Permanente Northern California from 1994 to 2005, we conducted 2 case-control analyses. Cases had at least 1 colorectal adenoma identified at either colonoscopy (analysis 1) or sigmoidoscopy (analysis 2). Controls had no colorectal neoplasia identified at the corresponding endoscopic examination. Serial HbA1c levels between the cases and the controls were compared using a longitudinal model. RESULTS: Case-control analysis 1 included 4,248 patients, of whom 1,296 (31%) had at least 1 adenoma. The adjusted mean HbA1c levels among those without any adenomas was 8.20% versus 8.26% among those with at least 1 adenoma, a difference of 0.06% (95% CI = -0.02 to 0.14, P = 0.16). Case-control analysis 2 included 9,813 patients, of whom 951 (10%) had at least 1 distal adenoma. The adjusted mean HbA1c levels among those without any distal adenomas was 8.32% versus 8.37% among those with at least 1 distal adenoma, a difference of 0.05% (95% CI = -00.04 to 0.14, P = 0.25). The results were similar for advanced adenomas. CONCLUSIONS: Glycemic control was not associated with the risk of colorectal adenoma among diabetic persons. IMPACT: These results would suggest that glycemic control is unlikely to confound the reported association between diabetes medications and the risk of colorectal cancer.

Mammographic density and risk of second breast cancer after ductal carcinoma in situ.

BACKGROUND: We examined whether mammographic density predicts risk of second breast cancers among patients with ductal carcinoma in situ (DCIS). METHODS: The study included DCIS patients diagnosed during 1990 to 1997 and treated with breast-conserving surgery at Kaiser Permanente Northern California. Medical records were reviewed for clinical factors and subsequent breast cancers (DCIS and invasive). Ipsilateral mammograms from the index DCIS were assessed for density without knowledge of subsequent cancer status. Cox regression modeling was used to examine the association between mammographic density and risk of breast cancer events. RESULTS: Of the 935 eligible DCIS patients, 164 (18%) had a subsequent ipsilateral breast cancer, and 59 (6%) had a new primary cancer in the contralateral breast during follow-up (median, 103 mo). Those with the greatest total area of density (upper 20% of values) were at increased risk for invasive disease in either breast [hazard ratio (HR), 2.1; 95% confidence interval (95% CI), 1.2-3.8] or any cancer (DCIS or invasive) in the ipsilateral (HR, 1.7; 95% CI, 1.0-2.9) or contralateral (HR, 3.0; 95% CI, 1.3-6.9) breast compared with those with the smallest area of density (bottom 20%). HRs for these same end points comparing those in the highest with those in the lowest American College of Radiology Breast Imaging Reporting and Data System category were 1.6 (95% CI, 0.7-3.6), 1.3 (95% CI, 0.7-2.6), and 5.0 (95% CI, 1.4-17.9), respectively. There was a suggestion of increasing risk of contralateral, but not ipsilateral, cancer with increasing percent density. CONCLUSIONS: Women with mammographically dense breasts may be at higher risk of subsequent breast cancer, especially in the contralateral breast. IMPACT: Information about mammographic density may help with DCIS treatment decisions.

Treatment of ductal carcinoma in situ among patients cared for in large integrated health plans.

OBJECTIVE: To examine whether use of adjuvant therapy varies by race/ethnicity among patients with ductal carcinoma in situ (DCIS) at 3 integrated health plan delivery sites based in California and Massachusetts. STUDY DESIGN: Cross-sectional study nested within a cohort of women diagnosed as having DCIS between 1990 and 2001. METHODS: We reviewed medical records of 3000 non-Hispanic white (69%), black (10%), Hispanic (9%), and Asian or Pacific Islander (12%) women diagnosed as having DCIS between 1990 and 2001 and treated with breast-conserving therapy. chi(2) Test and multinomial logistic regression analysis were used to examine the association between race/ethnicity and use of adjuvant treatments after controlling for patient and clinical variables, including certain pathologic factors. RESULTS: We found no significant differences in DCIS adjuvant treatment among racial/ethnic groups in bivariate or multinomial analyses after adjusting for demographic characteristics, comorbidity, and clinical factors. Minority women were as likely to undergo adjuvant radiation therapy as non-Hispanic white women. However, women 70 years or older (odds ratio, 0.40; 95% confidence interval, 0.31-0.51) and women who lived in areas with low geocoded median family income (odds ratio, 0.65; 95% confidence interval, 0.48-0.89) were less likely to receive adjuvant radiation therapy. Tumor size and comedo histologic growth pattern were associated with increased likelihood of receiving radiation therapy. CONCLUSION: Use of adjuvant therapy by minority women in these managed care plans is similar to that by non-Hispanic white women, although use was less among older women and among women who lived in poorer neighborhoods.

Declining recurrence among ductal carcinoma in situ patients treated with breast-conserving surgery in the community setting.

INTRODUCTION: Randomized trials indicate that adjuvant radiotherapy plus tamoxifen decrease the five-year risk of recurrence among ductal carcinoma in situ patients treated with breast-conserving surgery from about 20% to 8%. The aims of this study were to examine the use and impact of these therapies on risk of recurrence among ductal carcinoma in situ patients diagnosed and treated in the community setting. METHODS: We identified 2,995 patients diagnosed with ductal carcinoma in situ between 1990 and 2001 and treated with breast-conserving surgery at three large health plans. Medical charts were reviewed to confirm diagnosis and treatment and to obtain information on subsequent breast cancers. On a subset of patients, slides from the index ductal carcinoma in situ were reviewed for histopathologic features. Cumulative incidence curves were generated and Cox regression was used to examine changes in five-year risk of recurrence across diagnosis years, with and without adjusting for trends in use of adjuvant therapies. RESULTS: Use of radiotherapy increased from 25.8% in 1990-1991 to 61.3% in 2000-2001; tamoxifen increased from 2.3% to 34.4%. A total of 245 patients had a local recurrence within five years of their index ductal carcinoma in situ. The five-year risk of any local recurrence decreased from 14.3% (95% confidence interval 9.8 to 18.7) for patients diagnosed in 1990-1991 to 7.7% (95% confidence interval 5.5 to 9.9) for patients diagnosed in 1998-1999; invasive recurrence decreased from 7.0% (95% confidence interval 3.8 to 10.3) to 3.1% (95% confidence interval 1.7 to 4.6). In Cox models, the association between diagnosis year and risk of recurrence was modestly attenuated after accounting for use of adjuvant therapy. Between 1990-1991 and 2000-2001, the proportion of patients with tumors with high nuclear grade decreased from 46% to 32% (P = 0.03) and those with involved surgical margins dropped from 15% to 0% (P = 0.03). CONCLUSIONS: The marked increase in the 1990s in the use of adjuvant therapy for ductal carcinoma in situ patients treated with breast-conserving surgery in the community setting only partially explains the 50% decline in risk of recurrence. Changes in pathology factors have likely also contributed to this decline.

Adherence to long-term surveillance mammography among women with ductal carcinoma in situ treated with breast-conserving surgery.

PURPOSE Breast-conserving surgery (BCS) is an effective treatment for ductal carcinoma in situ (DCIS) but women who undergo BCS remain at risk for recurrences. Whether mammographic surveillance after BCS occurs and by whom is not known. METHODS We reviewed medical records of women diagnosed with DCIS between 1990 and 2001 and treated with BCS. Using descriptive statistics, generalized estimating, and logistic regression modeling, we examined the rates and predictors of surveillance mammography over a 10-year period after BCS. Results The cohort included 3,037 women observed for a median of 4.8 years (range, 0.5 to 15.7). Of the 2,676 women observed for at least 1 year after BCS, most (79%) had at least one surveillance mammogram during the first year of follow-up; 69% in year 5 and 61% in year 10. Among those observed for 5 years, surveillance mammograms were more likely among women age 60 to 69 years (odds ratio [OR], 1.72; 95% CI, 1.26 to 2.34), users of menopausal hormone therapy at diagnosis (OR, 1.26; 95% CI, 1.01 to 1.57) as well as those treated with adjuvant radiation (OR, 1.28; 95% CI, 1.08 to 1.53) and adjuvant radiation with tamoxifen (OR, 1.61; 95% CI, 1.13 to 2.30). Surveillance mammograms were less likely among obese women (OR, 0.70; 95% CI, 0.56 to 0.86). The findings were similar among women observed for 10 years. Only 34% and 15% of women observed for 5 and 10 years, respectively, had a surveillance mammogram during each year of follow-up. CONCLUSION Surveillance mammography after BCS among insured women with DCIS often did not occur yearly and declined over time after treatment. Patients and providers must remain vigilant about surveillance after BCS.

Thiazolidinedione therapy is not associated with increased colonic neoplasia risk in patients with diabetes mellitus.

BACKGROUND & AIMS: Thiazolidinedione ligands for peroxisome proliferator-activated receptor gamma (PPARgamma), are used to treat diabetes. PPARgamma is highly expressed in the colon, and exposure to thiazolidinediones has been proposed to affect the risk for colorectal neoplasia. In vitro models suggest that thiazolidinediones have antineoplastic effects, whereas in vivo studies have produced mixed results: Some indicate an increased risk for intestinal tumors. This study examined the association between PPARgamma-targeted therapies and the risk of colonic neoplasia in patients with diabetes. METHODS: We conducted 3 retrospective case-control studies nested within the cohort of diabetic patients who were cared for within the Kaiser Permanente of Northern California system from 1994 to 2005. Case subjects were those with colonic neoplasia identified at the time of colonoscopy (study 1), sigmoidoscopy (study 2), or at follow-up lower endoscopy (study 3). Controls had no neoplasia identified at the endoscopic examination. A minimum of 1 year of therapy was used to define medication exposure. RESULTS: Fourteen thousand eighty-six patients were included. Among patients undergoing colonoscopy, there was an inverse association between thiazolidinedione exposure and prevalence of neoplasia (adjusted odd ratio [OR], 0.73; 95% confidence interval [CI], 0.57-0.92); however, this was not evident among patients without anemia (adjusted OR, 0.97; 95% CI, 0.64-1.49). Significant associations between any or long-term thiazolidinedione use and colonic neoplasia were not observed among patients undergoing sigmoidoscopy or serial lower endoscopies. CONCLUSIONS: These results indicate that thiazolidinedione therapy is not associated with an increased risk for colonic neoplasia.

Medical therapy for diabetes is associated with increased use of lower endoscopy.

BACKGROUND: Diabetes mellitus is associated with an increased risk of colorectal neoplasia and diabetes medications may further influence the risk. Observational studies of the effect of diabetes medications on colonic neoplasia may be biased if use of diabetes medications is associated with undergoing lower endoscopy. This study examined the association between diabetes therapies and use of lower endoscopy. METHODS: This retrospective cohort study included patients with diabetes in an integrated, prepaid health plan. The primary exposure variables were use of sulfonylureas, metformin, thiazolidinediones (TZDs), and insulin. The outcome measure was completion of a flexible sigmoidoscopy or colonoscopy. Cox proportional hazards modeling, accounting for the time-varying nature of the medication exposures, was used to generate estimates of the relative hazard (HR) of lower endoscopy with different medications. RESULTS: The study included 44 169 patients followed for a mean duration of 4.2 years (SD = 2.5 years); 34% underwent at least one lower endoscopy. Patients who filled a diabetes medication prescription were more likely to undergo lower endoscopy (HR = 1.13, 95%CI 1.06-1.21). Compared to those taking only sulfonylureas, patients receiving sulfonylureas and metformin (HR = 1.12, 95%CI 1.06-1.18) or metformin alone (HR = 1.17, 95%CI 1.07-1.26) were more likely to undergo lower endoscopy. For all medications, new use was associated with undergoing lower endoscopy (p < 0.05 for all comparisons). CONCLUSIONS: Diabetic patients receiving medications are more likely to undergo lower endoscopy than those on diet control alone, particularly in the first year after initiating a new medication class and if taking metformin.