Mapping the Ecological Terrain of Stroke Prehospital Delay: A Nationwide Registry Study.

BACKGROUND: Delays in hospital presentation limit access to acute stroke treatments. While prior research has focused on patient-level factors, broader ecological and social determinants have not been well studied. We aimed to create a geospatial map of prehospital delay and examine the role of community-level social vulnerability. METHODS: We studied patients with ischemic stroke who arrived by emergency medical services in 2015 to 2017 from the American Heart Association Get With The Guidelines-Stroke registry. The primary outcome was time to hospital arrival after stroke (in minutes), beginning at last known well in most cases. Using Geographic Information System mapping, we displayed the geography of delay. We then used Cox proportional hazard models to study the relationship between community-level factors and arrival time (adjusted hazard ratios [aHR] <1.0 indicate delay). The primary exposure was the social vulnerability index (SVI), a metric of social vulnerability for every ZIP Code Tabulation Area ranging from 0.0 to 1.0. RESULTS: Of 750 336 patients, 149 145 met inclusion criteria. The mean age was 73 years, and 51% were female. The median time to hospital arrival was 140 minutes (Q1: 60 minutes, Q3: 458 minutes). The geospatial map revealed that many zones of delay overlapped with socially vulnerable areas (https://harvard-cga.maps.arcgis.com/apps/webappviewer/index.html?id=08f6e885c71b457f83cefc71013bcaa7). Cox models (aHR, 95% CI) confirmed that higher SVI, including quartiles 3 (aHR, 0.96 [95% CI, 0.93-0.98]) and 4 (aHR, 0.93 [95% CI, 0.91-0.95]), was associated with delay. Patients from SVI quartile 4 neighborhoods arrived 15.6 minutes [15-16.2] slower than patients from SVI quartile 1. Specific SVI themes associated with delay were a community's socioeconomic status (aHR, 0.80 [95% CI, 0.74-0.85]) and housing type and transportation (aHR, 0.89 [95% CI, 0.84-0.94]). CONCLUSIONS: This map of acute stroke presentation times shows areas with a high incidence of delay. Increased social vulnerability characterizes these areas. Such places should be systematically targeted to improve population-level stroke presentation times.

Association of early menarche with breast tumor molecular features and recurrence.

BACKGROUND: Early menarche is an established risk factor for breast cancer but its molecular contribution to tumor biology and prognosis remains unclear. METHODS: We profiled transcriptome-wide gene expression in breast tumors (N = 846) and tumor-adjacent normal tissues (N = 666) from women in the Nurses' Health Studies (NHS) to investigate whether early menarche (age < 12) is associated with tumor molecular and prognostic features in women with breast cancer. Multivariable linear regression and pathway analyses using competitive gene set enrichment analysis were conducted in both tumor and adjacent-normal tissue and externally validated in TCGA (N = 116). Subgroup analyses stratified on ER-status based on the tumor were also performed. PAM50 signatures were used for tumor molecular subtyping and to generate proliferation and risk of recurrence scores. We created a gene expression score using LASSO regression to capture early menarche based on 28 genes from FDR-significant pathways in breast tumor tissue in NHS and tested its association with 10-year disease-free survival in both NHS (N = 836) and METABRIC (N = 952). RESULTS: Early menarche was significantly associated with 369 individual genes in adjacent-normal tissues implicated in extracellular matrix, cell adhesion, and invasion (FDR ≤ 0.1). Early menarche was associated with upregulation of cancer hallmark pathways (18 significant pathways in tumor, 23 in tumor-adjacent normal, FDR ≤ 0.1) related to proliferation (e.g. Myc, PI3K/AKT/mTOR, cell cycle), oxidative stress (e.g. oxidative phosphorylation, unfolded protein response), and inflammation (e.g. pro-inflammatory cytokines IFN α and IFN γ ). Replication in TCGA confirmed these trends. Early menarche was associated with significantly higher PAM50 proliferation scores (ß = 0.082 [0.02-0.14]), odds of aggressive molecular tumor subtypes (basal-like, OR = 1.84 [1.18-2.85] and HER2-enriched, OR = 2.32 [1.46-3.69]), and PAM50 risk of recurrence score (ß = 4.81 [1.71-7.92]). Our NHS-derived early menarche gene expression signature was significantly associated with worse 10-year disease-free survival in METABRIC (N = 952, HR = 1.58 [1.10-2.25]). CONCLUSIONS: Early menarche is associated with more aggressive molecular tumor characteristics and its gene expression signature within tumors is associated with worse 10-year disease-free survival among women with breast cancer. As the age of onset of menarche continues to decline, understanding its relationship to breast tumor characteristics and prognosis may lead to novel secondary prevention strategies.

Metformin and other anti-diabetic medication use and breast cancer incidence in the Nurses' Health Studies.

We aimed to examine the association between the use of metformin and other anti-diabetic medications and breast cancer incidence within two large prospective cohort studies. We followed 185,181 women who participated in the Nurses' Health Study (NHS; 1994-2016) and the NHSII (1995-2017), with baseline corresponding to the date metformin was approved for type 2 diabetes (T2D) treatment in the US Information on T2D diagnosis, anti-diabetes medications, and other covariates was self-reported at baseline and repeatedly assessed by follow-up questionnaires every 2 years. Breast cancer cases were self-reported and confirmed by medical record review. Hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between medication use and breast cancer were estimated using Cox proportional hazards regression models, adjusting for breast cancer risk factors. During 3,324,881 person-years of follow-up, we ascertained 9,192 incident invasive breast cancer cases, of which 451 were among women with T2D. Compared with women without T2D (n = 169,263), neither metformin use (HR = 0.97; 95% CI = 0.81-1.15) nor other anti-diabetic medications use (HR = 1.11; 95% CI = 0.90-1.36) associated with significantly lower breast cancer incidence. Among women with T2D (n = 15,918), compared with metformin never users, metformin ever use was not significantly inversely associated with breast cancer (HR = 0.92; 95% CI = 0.74-1.15). Although we observed that past use of metformin was inversely associated with breast cancer in the T2D population (HR = 0.67; 95% CI = 0.48-0.94), current use (HR = 1.01; 95% CI = 0.80-1.27) and longer duration of metformin use were not associated with breast cancer (each 2-year interval: HR = 1.01; 95% CI = 0.95-1.07). Overall, metformin use was not associated with the risk of developing breast cancer among the overall cohort population or among women with T2D.

Validity and Reproducibility of a Semiquantitative Food Frequency Questionnaire for Measuring Intakes of Foods and Food Groups.

We evaluated the validity and reproducibility of a semiquantitative food frequency questionnaire (FFQ) for measuring intakes of 149 foods and 25 food groups among 736 participants of the Women's Lifestyle Validation Study (WLVS, 2010-2012) and 649 participants of the Men's Lifestyle Validation Study (MLVS, 2011-2013). Validity of the FFQ compared with two 7-day dietary records measured 6 months apart and the reproducibility between 2 FFQs administered 1 year apart (FFQ1 and FFQ2) were assessed using Spearman correlations and intraclass correlation coefficients. The average 1-year reproducibility of FFQ-measured foods was 0.64 in both the WLVS and MLVS. Reproducibility of the food groups (mean = 0.71 among women and 0.72 among men) was generally higher than that for individual foods. Among women, the average validity correlation for individual foods was 0.59 when comparing FFQ2 with the 7-day dietary records. Among men, the corresponding average validity correlation was 0.61. Compared with individual foods, food groups had slightly higher validity correlations in both women (range, 0.45-0.92; mean = 0.61) and men (range, 0.46-0.88; mean = 0.65). This study reaffirms that the FFQ performs well in measuring most foods and food groups and provides data to adjust for measurement errors in epidemiologic studies of foods and food groups.

Associations of stem cell markers CD44, CD24 and ALDH1A1 with mammographic breast density in women with benign breast biopsies.

BACKGROUND: We examined associations of CD44, CD24 and ALDH1A1 breast stem cell markers with mammographic breast density (MBD), a well-established breast cancer (BCa) risk factor. METHODS: We included 218 cancer-free women with biopsy-confirmed benign breast disease within the Nurses' Health Study (NHS) and NHSII. The data on BCa risk factors were obtained from biennial questionnaires. Immunohistochemistry (IHC) was done on tissue microarrays. For each core, the IHC expression was assessed using a semi-automated platform and expressed as percent of positively stained cells for each marker out of the total cell count. MBD was assessed with computer-assisted techniques. Generalised linear regression was used to examine the associations of each marker with square root-transformed percent density (PD), absolute dense and non-dense areas (NDA), adjusted for BCa risk factors. RESULTS: Stromal CD44 and ALDH1A1 expression was positively associated with PD (≥ 10% vs. <10% ß = 0.56, 95% confidence interval [CI] [0.06; 1.07] and ß = 0.81 [0.27; 1.34], respectively) and inversely associated with NDA (ß per 10% increase = -0.17 [-0.34; -0.01] and ß for ≥10% vs. <10% = -1.17 [-2.07; -0.28], respectively). Epithelial CD24 expression was inversely associated with PD (ß per 10% increase = -0.14 [-0.28; -0.01]. Stromal and epithelial CD24 expression was positively associated with NDA (ß per 10% increase = 0.35 [0.2 × 10-2; 0.70] and ß per 10% increase = 0.34 [0.11; 0.57], respectively). CONCLUSION: Expression of stem cell markers is associated with MBD.

Predicting the onset of breast cancer using mammogram imaging data with irregular boundary.

With mammography being the primary breast cancer screening strategy, it is essential to make full use of the mammogram imaging data to better identify women who are at higher and lower than average risk. Our primary goal in this study is to extract mammogram-based features that augment the well-established breast cancer risk factors to improve prediction accuracy. In this article, we propose a supervised functional principal component analysis (sFPCA) over triangulations method for extracting features that are ordered by the magnitude of association with the failure time outcome. The proposed method accommodates the irregular boundary issue posed by the breast area within the mammogram imaging data with flexible bivariate splines over triangulations. We also provide an eigenvalue decomposition algorithm that is computationally efficient. Compared to the conventional unsupervised FPCA method, the proposed method results in a lower Brier Score and higher area under the ROC curve (AUC) in simulation studies. We apply our method to data from the Joanne Knight Breast Health Cohort at Siteman Cancer Center. Our approach not only obtains the best prediction performance comparing to unsupervised FPCA and benchmark models but also reveals important risk patterns within the mammogram images. This demonstrates the importance of utilizing additional supervised image-based features to clarify breast cancer risk.

Recreational physical activity and breast cancer risk by menopausal status and tumor hormone receptor status: results from the Nurses' Health Studies.

PURPOSE: Physical activity is associated with lower breast cancer risk, especially in postmenopausal women. Associations in premenopausal women are less well established. METHODS: We evaluated recreational physical activity and breast cancer risk in the Nurses' Health Study (NHS) and NHSII (187,278 women; n = 12,785 breast cancers; follow-up: NHS = 1986-2016, NHSII = 1989-2017) by menopausal status and estrogen (ER) and progesterone (PR) receptor status. Physical activity was evaluated as updated cumulative average of metabolic equivalent of task (MET)-h/week. Cox proportional hazards models were used to estimate multivariable hazard ratios (HR) and 95% confidence intervals (CI). RESULTS: Recreational physical activity was inversely associated with breast cancer risk in pre- and postmenopausal women. Higher activity levels were associated with lower risk of ER+/PR + breast cancer in both pre- and postmenopausal women (e.g., total recreational activity, ≥ 27 vs < 3 MET-h/week, premenopausal HR = 0.83, 95%CI = (0.70-0.99), postmenopausal HR = 0.86 (0.78-0.95); pheterogeneity = 0.97). Results were attenuated with adjustment for current body mass index (BMI) among postmenopausal, but not premenopausal, women (e.g., ≥ 27 vs < 3 MET-h/week, premenopausal HR = 0.83 (0.69-0.98); postmenopausal HR = 0.95 (0.85-1.05); pheterogeneity = 0.99). In analyses of moderate-vigorous activity and breast cancer risk, no heterogeneity by menopausal status was observed (phet ≥ 0.53; e.g., ≥ 27 vs < 3 MET-h/week, ER+/PR+, premenopausal HR = 0.88 (0.69-1.11); postmenopausal HR = 0.71 (0.58-0.88). No associations were observed for ER-/PR- disease. CONCLUSIONS: Recreational physical activity was associated with lower breast cancer risk in both pre- and postmenopausal women, supporting recreational physical activity as an accessible, modifiable exposure associated with reduced breast cancer risk regardless of menopausal status.

Differences in metabolomic profiles between Black and White women in the U.S.: Analyses from two prospective cohorts.

There is growing interest in incorporating metabolomics into public health practice. However, Black women are under-represented in many metabolomics studies. If metabolomic profiles differ between Black and White women, this under-representation may exacerbate existing Black-White health disparities. We therefore aimed to estimate metabolomic differences between Black and White women in the U.S. We leveraged data from two prospective cohorts: the Nurses' Health Study (NHS; n = 2077) and Women's Health Initiative (WHI; n = 2128). The WHI served as the replication cohort. Plasma metabolites (n = 334) were measured via liquid chromatography-tandem mass spectrometry. Observed metabolomic differences were estimated using linear regression and metabolite set enrichment analyses. Residual metabolomic differences in a hypothetical population in which the distributions of 14 risk factors were equalized across racial groups were estimated using inverse odds ratio weighting. In the NHS, Black-White differences were observed for most metabolites (75 metabolites with observed differences ≥ |0.50| standard deviations). Black women had lower average levels than White women for most metabolites (e.g., for N6, N6-dimethlylysine, mean Black-White difference = - 0.98 standard deviations; 95% CI: - 1.11, - 0.84). In metabolite set enrichment analyses, Black women had lower levels of triglycerides, phosphatidylcholines, lysophosphatidylethanolamines, phosphatidylethanolamines, and organoheterocyclic compounds, but higher levels of phosphatidylethanolamine plasmalogens, phosphatidylcholine plasmalogens, cholesteryl esters, and carnitines. In a hypothetical population in which distributions of 14 risk factors were equalized, Black-White metabolomic differences persisted. Most results replicated in the WHI (88% of 272 metabolites available for replication). Substantial differences in metabolomic profiles exist between Black and White women. Future studies should prioritize racial representation.

Associations of alcohol consumption with breast tissue composition.

BACKGROUND: We investigated the associations of alcohol with percentage of epithelium, stroma, fibroglandular tissue (epithelium + stroma), and fat in benign breast biopsy samples. METHODS: We included 857 cancer-free women with biopsy-confirmed benign breast disease within the Nurses' Health Study (NHS) and NHSII cohorts. Percentage of each tissue was measured on whole slide images using a deep-learning algorithm and then log-transformed. Alcohol consumption (recent and cumulative average) was assessed with semi-quantitative food frequency questionnaires. Regression estimates were adjusted for known breast cancer risk factors. All tests were 2-sided. RESULTS: Alcohol was inversely associated with % of stroma and fibroglandular tissue (recent ≥ 22 g/day vs. none: stroma: ß = - 0.08, 95% Confidence Interval [CI] - 0.13; - 0.03; fibroglandular: ß = - 0.08, 95% CI - 0.13; - 0.04; cumulative ≥ 22 g/day vs. none: stroma: ß = - 0.08, 95% CI - 0.13; - 0.02; fibroglandular: ß = - 0.09, 95% CI - 0.14; - 0.04) and positively associated with fat % (recent ≥ 22 g/day vs. none: ß = 0.30, 95% CI 0.03; 0.57; cumulative ≥ 22 g/day vs. none: ß = 0.32, 95% CI 0.04; 0.61). In stratified analysis, alcohol consumption was not associated with tissue measures in premenopausal women. In postmenopausal women, cumulative alcohol use was inversely associated with % of stroma and fibroglandular tissue and positively associated with fat % (≥ 22 g/day vs. none: stroma: ß = - 0.16, 95% CI - 0.28; - 0.07; fibroglandular: ß = - 0.18, 95% CI - 0.28; - 0.07; fat: ß = 0.61, 95% CI 0.01; 1.22), with similar results for recent alcohol use. CONCLUSION: Our findings suggest that alcohol consumption is associated with smaller % of stroma and fibroglandular tissue and a greater % of fat in postmenopausal women. Future studies are warranted to confirm our findings and to elucidate the underlying biological mechanisms.

A model for predicting both breast cancer risk and non-breast cancer death among women > 55 years old.

BACKGROUND: Guidelines recommend shared decision making (SDM) for mammography screening for women ≥ 75 and not screening women with < 10-year life expectancy. High-quality SDM requires consideration of women's breast cancer (BC) risk, life expectancy, and values but is hard to implement because no models simultaneously estimate older women's individualized BC risk and life expectancy. METHODS: Using competing risk regression and data from 83,330 women > 55 years who completed the 2004 Nurses' Health Study (NHS) questionnaire, we developed (in 2/3 of the cohort, n = 55,533) a model to predict 10-year non-breast cancer (BC) death. We considered 60 mortality risk factors and used best-subsets regression, the Akaike information criterion, and c-index, to identify the best-fitting model. We examined model performance in the remaining 1/3 of the NHS cohort (n = 27,777) and among 17,380 Black Women's Health Study (BWHS) participants, ≥ 55 years, who completed the 2009 questionnaire. We then included the identified mortality predictors in a previously developed competing risk BC prediction model and examined model performance for predicting BC risk. RESULTS: Mean age of NHS development cohort participants was 70.1 years (± 7.0); over 10 years, 3.1% developed BC, 0.3% died of BC, and 20.1% died of other causes; NHS validation cohort participants were similar. BWHS participants were younger (mean age 63.7 years [± 6.7]); over 10-years 3.1% developed BC, 0.4% died of BC, and 11.1% died of other causes. The final non-BC death prediction model included 21 variables (age; body mass index [BMI]; physical function [3 measures]; comorbidities [12]; alcohol; smoking; age at menopause; and mammography use). The final BC prediction model included age, BMI, alcohol and hormone use, family history, age at menopause, age at first birth/parity, and breast biopsy history. When risk factor regression coefficients were applied in the validation cohorts, the c-index for predicting 10-year non-BC death was 0.790 (0.784-0.796) in NHS and 0.768 (0.757-0.780) in BWHS; for predicting 5-year BC risk, the c-index was 0.612 (0.538-0.641) in NHS and 0.573 (0.536-0.611) in BWHS. CONCLUSIONS: We developed and validated a novel competing-risk model that predicts 10-year non-BC death and 5-year BC risk. Model risk estimates may help inform SDM around mammography screening.

Associations of low-carbohydrate diets with breast cancer survival.

BACKGROUND: The objective of this study was to evaluate the role of low-carbohydrate diets after breast cancer diagnosis in relation to breast cancer-specific and all-cause mortality. METHODS: For 9621 women with stage I-III breast cancer from two ongoing cohort studies, the Nurses' Health Study and Nurses' Health Study II, overall low-carbohydrate, animal-rich low-carbohydrate, and plant-rich low-carbohydrate diet scores were calculated by using food frequency questionnaires collected after breast cancer diagnosis. RESULTS: Participants were followed up for a median 12.4 years after breast cancer diagnosis. We documented 1269 deaths due to breast cancer and 3850 all-cause deaths. With the use of Cox proportional hazards regression and after controlling for potential confounding variables, we observed a significantly lower risk of overall mortality among women with breast cancer who had greater adherence to overall low-carbohydrate diets (hazard ratio for quintile 5 vs. quintile 1 [HRQ5vsQ1 ], 0.82; 95% CI, 0.74-0.91; ptrend  = .0001) and plant-rich low-carbohydrate diets (HRQ5vsQ1 , 0.73; 95% CI, 0.66-0.82; ptrend < .0001) after breast cancer diagnosis but not animal-rich low-carbohydrate diets (HRQ5vsQ1 , 0.93; 95% CI, 0.84-1.04; ptrend  = .23). However, greater adherence to overall, animal-rich, or plant-rich low-carbohydrate diets was not significantly associated with a lower risk of breast cancer-specific mortality. CONCLUSIONS: This study showed that greater adherence to low-carbohydrate diets, especially plant-rich low-carbohydrate diets, was associated with better overall survival but not breast cancer-specific survival among women with stage I-III breast cancer.

Consumption of olive oil and risk of breast cancer in U.S. women: results from the Nurses' Health Studies.

BACKGROUND: Olive oil consumption may reduce breast cancer risk, but it is unclear whether olive oil is beneficial for breast cancer prevention in populations outside of Mediterranean regions, namely in the U.S., where the average consumption of olive oil is low compared with Mediterranean populations. We examined whether olive oil intake was associated with breast cancer risk in two prospective cohorts of U.S. women. METHODS: We used multivariable-adjusted time-varying Cox proportional hazards models to estimate hazard ratios (HR) and 95% confidence interval (CI) for breast cancer among 71,330 (Nurses' Health Study, 1990-2016) and 93,295 women (Nurses' Health Study II, 1991-2017) who were free of cancer at baseline. Diet was assessed by a validated semi-quantitative food frequency questionnaire every 4 years. RESULTS: During 3,744,068 person-years of follow-up, 9,638 women developed invasive breast cancer. The multivariable-adjusted HR (95% CI) for breast cancer among women who had the highest consumption of olive oil (>1/2 tablespoon/d or >7 g/d) compared with those who never or rarely consumed olive oil, was 1.01 (0.93, 1.09). Higher olive oil consumption was not associated with any subtype of breast cancer. CONCLUSION: We did not observe an association between higher olive oil intake and breast cancer risk in two large prospective cohorts of U.S. women, whose average olive oil consumption was low. Prospective studies are needed to confirm these findings and to further investigate whether different varieties of olive oil (e.g., virgin and extra virgin olive oil) may play a role in breast cancer risk.

Validation of a breast cancer risk prediction model based on the key risk factors: family history, mammographic density and polygenic risk.

PURPOSE: We compared a simple breast cancer risk prediction model, BRISK (which includes mammographic density, polygenic risk and clinical factors), against a similar model with more risk factors (simplified Rosner) and against two commonly used clinical models (Gail and IBIS). METHODS: Using nested case-control data from the Nurses' Health Study, we compared the models' association, discrimination and calibration. Classification performance was compared between Gail and BRISK for 5-year risks and between IBIS and BRISK for remaining lifetime risk. RESULTS: The odds ratio per standard deviation was 1.43 (95% CI 1.32, 1.55) for BRISK 5-year risk, 1.07 (95% CI 0.99, 1.14) for Gail 5-year risk, 1.72 (95% CI 1.59, 1.87) for simplified Rosner 10-year risk, 1.51 (95% CI 1.41, 1.62) for BRISK remaining lifetime risk and 1.26 (95% CI 1.16, 1.36) for IBIS remaining lifetime risk. The area under the receiver operating characteristic curve (AUC) was improved for BRISK over Gail for 5-year risk (AUC = 0.636 versus 0.511, P < 0.0001) and for BRISK over IBIS for remaining lifetime risk (AUC = 0.647 versus 0.571, P < 0.0001). BRISK was well calibrated for the estimation of both 5-year risk (expected/observed [E/O] = 1.03; 95% CI 0.73, 1.46) and remaining lifetime risk (E/O = 1.01; 95% CI 0.86, 1.17). The Gail 5-year risk (E/O = 0.85; 95% CI 0.58, 1.24) and IBIS remaining lifetime risk (E/O = 0.73; 95% CI 0.60, 0.87) were not well calibrated, with both under-estimating risk. BRISK improves classification of risk compared to Gail 5-year risk (NRI = 0.31; standard error [SE] = 0.031) and IBIS remaining lifetime risk (NRI = 0.287; SE = 0.035). CONCLUSION: BRISK performs better than two commonly used clinical risk models and no worse compared to a similar model with more risk factors.

Long-term Alcohol Consumption and Risk of Exfoliation Glaucoma or Glaucoma Suspect Status among United States Health Professionals.

PURPOSE: To assess the association between intakes of total alcohol and individual alcoholic beverages and the incidence of exfoliation glaucoma/glaucoma suspect (XFG/XFGS) status. DESIGN: Prospective cohort study. PARTICIPANTS: A total of 195 408 participants in the Nurses' Health Study (1980-2018), the Health Professionals Follow-up Study (1986-2018), and the Nurses' Health Study II (1991-2019) were followed biennially. Eligible participants at each 2-year risk period were ≧ 40 years and free of XFG/XFGS status with available data on diet and ophthalmic examination findings. METHODS: Cumulatively averaged total (primary exposure) and individual alcoholic beverage (beer, wine, and liquor) intakes from validated dietary information every 2-4 years. MAIN OUTCOME MEASURES: Confirmed incident XFG/XFGS status using medical records. We used per-eye Cox proportional hazards models, accounting for intereye correlations, to estimate multivariate-adjusted relative risks (MVRRs) and 95% confidence intervals (CIs). RESULTS: During 6 877 823 eye-years of follow-up, 705 eyes with XFG/XFGS status were documented. Greater total alcohol consumption was associated significantly with higher XFG/XFGS status risk: the MVRR for XFG/XFGS status for cumulatively averaged alcohol consumption of ≧15 g/day or more versus nondrinking was 1.55 (95% CI, 1.17-2.07; P = 0.02 for trend). Long- and short-term alcohol intake was associated significantly with XFG/XFGS status risk, with the strongest associations with cumulatively averaged alcohol intake as of 4 years before diagnosis (MVRR ≥ 15 g/day vs. nondrinking, 1.65; 95% CI, 1.25-2.18; P = 0.002 for trend). Compared with nondrinkers, consuming ≧ 3.6 drinks of beer, wine, or liquor per week was associated with the following MVRRs for XFG/XFGS status: 1.26 (95% CI, 0.89-1.77; P = 0.40 for trend), 1.30 (95% CI, 1.00-1.68; P = 0.15 for trend), and 1.46 (95% CI, 1.15-1.85; P = 0.01 for trend), respectively. We did not observe interactions by age, latitude, residential tier, or intakes of folate or vitamin A (P > 0.40 for interaction); however, the association between alcohol and XFG/XFGS status was suggestively stronger for those without a family history of glaucoma (P = 0.10 for interaction). CONCLUSIONS: Long-term alcohol consumption was associated with a higher risk of XFG/XFGS status. Our findings provide further clues regarding the XFG/XFGS etiology.

Supervised two-dimensional functional principal component analysis with time-to-event outcomes and mammogram imaging data.

Screening mammography aims to identify breast cancer early and secondarily measures breast density to classify women at higher or lower than average risk for future breast cancer in the general population. Despite the strong association of individual mammography features to breast cancer risk, the statistical literature on mammogram imaging data is limited. While functional principal component analysis (FPCA) has been studied in the literature for extracting image-based features, it is conducted independently of the time-to-event response variable. With the consideration of building a prognostic model for precision prevention, we present a set of flexible methods, supervised FPCA (sFPCA) and functional partial least squares (FPLS), to extract image-based features associated with the failure time while accommodating the added complication from right censoring. Throughout the article, we hope to demonstrate that one method is favored over the other under different clinical setups. The proposed methods are applied to the motivating data set from the Joanne Knight Breast Health cohort at Siteman Cancer Center. Our approaches not only obtain the best prediction performance compared to the benchmark model, but also reveal different risk patterns within the mammograms.

A repeated measures approach to pooled and calibrated biomarker data.

Participant-level meta-analysis across multiple studies increases the sample size for pooled analyses, thereby improving precision in effect estimates and enabling subgroup analyses. For analyses involving biomarker measurements as an exposure of interest, investigators must first calibrate the data to address measurement variability arising from usage of different laboratories and/or assays. In practice, the calibration process involves reassaying a random subset of biospecimens from each study at a central laboratory and fitting models that relate the study-specific "local" and central laboratory measurements. Previous work in this area treats the calibration process from the perspective of measurement error techniques and imputes the estimated central laboratory value among individuals with only a local laboratory measurement. In this work, we propose a repeated measures method to calibrate biomarker measurements pooled from multiple studies with study-specific calibration subsets. We account for correlation between measurements made on the same person and between measurements made at the same laboratory. We demonstrate that the repeated measures approach provides valid inference, and compare it to existing calibration approaches grounded in measurement error techniques in an example describing the association between circulating vitamin D and stroke.

Analytical method for detecting outlier evaluators.

BACKGROUND: Epidemiologic and medical studies often rely on evaluators to obtain measurements of exposures or outcomes for study participants, and valid estimates of associations depends on the quality of data. Even though statistical methods have been proposed to adjust for measurement errors, they often rely on unverifiable assumptions and could lead to biased estimates if those assumptions are violated. Therefore, methods for detecting potential 'outlier' evaluators are needed to improve data quality during data collection stage. METHODS: In this paper, we propose a two-stage algorithm to detect 'outlier' evaluators whose evaluation results tend to be higher or lower than their counterparts. In the first stage, evaluators' effects are obtained by fitting a regression model. In the second stage, hypothesis tests are performed to detect 'outlier' evaluators, where we consider both the power of each hypothesis test and the false discovery rate (FDR) among all tests. We conduct an extensive simulation study to evaluate the proposed method, and illustrate the method by detecting potential 'outlier' audiologists in the data collection stage for the Audiology Assessment Arm of the Conservation of Hearing Study, an epidemiologic study for examining risk factors of hearing loss in the Nurses' Health Study II. RESULTS: Our simulation study shows that our method not only can detect true 'outlier' evaluators, but also is less likely to falsely reject true 'normal' evaluators. CONCLUSIONS: Our two-stage 'outlier' detection algorithm is a flexible approach that can effectively detect 'outlier' evaluators, and thus data quality can be improved during data collection stage.

Development and Validation of a Risk Model for Predicting Contrast-Associated Acute Kidney Injury in Patients With Cancer: Evaluation in Over 46,000 CT Examinations.

BACKGROUND. Patients with cancer undergo frequent CT examinations with iodinated contrast media and may be uniquely predisposed to contrast-associated acute kidney injury (CA-AKI). OBJECTIVE. The purpose of this study was to develop and validate a model for predicting the risk of CA-AKI after contrast-enhanced CT in patients with cancer. METHODS. This retrospective study included 25,184 adult patients (12,153 men, 13,031 women; mean age, 62.3 ± 13.7 [SD] years) with cancer who underwent 46,593 contrast-enhanced CT examinations between January 1, 2016, and June 20, 2020, at one of three academic medical centers. Information was recorded regarding demographics, malignancy type, medication use, baseline laboratory values, and comorbid conditions. CA-AKI was defined as a 0.3-mg/dL or greater increase in serum creatinine level from baseline within 48 hours after CT or a 1.5-fold or greater increase in the peak measurement within 14 days after CT. Multivariable models accounting for correlated data were used to identify risk factors for CA-AKI. A risk score for predicting CA-AKI was generated in a development set (n = 30,926) and tested in a validation set (n = 15,667). RESULTS. CA-AKI occurred after 5.8% (2682/46,593) of CT examinations. The final multivariable model for predicting CA-AKI included hematologic malignancy, diuretic use, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use, chronic kidney disease (CKD) stage 3a, CKD stage 3b, CKD stage 4 or 5, serum albumin level less than 3.0 g/dL, platelet count less than 150 × 103/µL, 1+ or greater proteinuria on baseline urinalysis, diabetes mellitus, heart failure, and contrast medium volume 100 mL or greater. A risk score (range, 0-53 points) was generated with these variables. The most points (13) were for CKD stage 4 or 5 and for albumin level less than 3 g/dL. The frequency of CA-AKI progressively increased in higher risk categories. For example, in the validation set, CA-AKI occurred after 2.2% of CT examinations in the lowest risk category (score ≤ 4) and after 32.7% of CT examinations in the highest risk category (score ≥ 30). The Hosmer-Lemeshow test result indicated that the risk score was a good fit (p = .40). CONCLUSION. A risk model in which readily available clinical data are used to predict the likelihood of CA-AKI after contrast-enhanced CT in patients with cancer was developed and validated. CLINICAL IMPACT. The model may help facilitate appropriate implementation of preventive measures in the care of patients at high risk of CA-AKI.

Defining Daily Disposable Contact Lens Wear in a Clinical Study.

SIGNIFICANCE: The U.S. Food and Drug Administration regulates contact lenses as prescription medical devices and defines daily disposable lenses for single use; however, safety comparisons between daily disposable and reusable lenses rely on the lens-wearing regimen. When inappropriately discerned, studies may erroneously report replacement regimen, resulting in inaccurate risk rates. PURPOSE: This study aimed to explore different measures for defining daily disposable wearers in the context of a clinical study. METHODS: A secondary analysis of data from five multisite fieldings (n = 1059) from the Contact Lens Risk Survey was performed. Descriptive statistics were used to examine self-reported lens replacement, use of lens case, and manufacturer's recommended replacement frequency as defined by the participants' selection of their habitual lenses using a photographic aide. Daily disposable wearers were identified as reporting daily replacement (by self-report and lens identification) and not using a lens case. If there was a discrepancy among these three factors, the lens assignment was assessed as a majority response (two of three) or as missing if conflicting information was reported. RESULTS: The cohort was approximately two-thirds (68.7%) female and ranged from 12 to 33 years of age. A total of 154 participants (14.5%) were classified as daily disposable wearers and 896 (84.6%) as reusable wearers. Congruence was observed among all three daily disposable assignment factors for only 106 wearers (68.8%). The greatest discrepancy among daily disposable wearers was the report of using a lens case (n = 32 [20.8%]). In contrast, reusable lens wearers were more likely to report agreement across all three factors (n = 798 [89.1%], P < .001). CONCLUSIONS: This analysis suggests that the report of not using a lens case is a more conservative surrogate for true single-use lens wear, as one in five may be storing and reusing their daily disposable lenses.

Estimation and testing for clustered interval-censored bivariate survival data with application using the semi-parametric version of the Clayton-Oakes model.

The Kaplan-Meier estimator is ubiquitously used to estimate survival probabilities for time-to-event data. It is nonparametric, and thus does not require specification of a survival distribution, but it does assume that the risk set at any time t consists of independent observations. This assumption does not hold for data from paired organ systems such as occur in ophthalmology (eyes) or otolaryngology (ears), or for other types of clustered data. In this article, we estimate marginal survival probabilities in the setting of clustered data, and provide confidence limits for these estimates with intra-cluster correlation accounted for by an interval-censored version of the Clayton-Oakes model. We develop a goodness-of-fit test for general bivariate interval-censored data and apply it to the proposed interval-censored version of the Clayton-Oakes model. We also propose a likelihood ratio test for the comparison of survival distributions between two groups in the setting of clustered data under the assumption of a constant between-group hazard ratio. This methodology can be used both for balanced and unbalanced cluster sizes, and also when the cluster size is informative. We compare our test to the ordinary log rank test and the Lin-Wei (LW) test based on the marginal Cox proportional Hazards model with robust standard errors obtained from the sandwich estimator. Simulation results indicate that the ordinary log rank test over-inflates type I error, while the proposed unconditional likelihood ratio test has appropriate type I error and higher power than the LW test. The method is demonstrated in real examples from the Sorbinil Retinopathy Trial, and the Age-Related Macular Degeneration Study. Raw data from these two trials are provided.