Faculty Bibliography

BACKGROUND:Incidence of premenopausal breast cancer (BC) has risen in recent years, though most existing BC prediction models are not generalizable to young women due to underrepresentation of this age group in model development. METHODS:Using questionnaire-based data from 19 prospective studies harmonized within the Premenopausal Breast Cancer Collaborative Group (PBCCG), representing 783,830 women, we developed a premenopausal BC risk prediction model. The data were split into training (2/3) and validation (1/3) datasets with equal distribution of cohorts in each. In the training dataset variables were chosen from known and hypothesized risk factors: age, age at menarche, age at first birth, parity, breastfeeding, height, BMI, young adulthood BMI, recent weight change, alcohol consumption, first-degree family history of BC, and personal history of benign breast disease (BBD). Hazard ratios (HR) and 95% confidence intervals (CI) were estimated by Cox proportional hazards regression using age as time scale, stratified by cohort. Given that complete information on all risk factors was not available in all cohorts, coefficients were estimated separately in groups of cohorts with the same available covariate information, adjusted to account for the correlation between missing and non-missing variables and meta-analyzed. Absolute risk of BC (in situ or invasive) within 5 years, was determined using country-, age-, and birth cohort-specific incidence rates. Discrimination (area under the curve, AUC) and calibration (Expected/Observed, E/O) were evaluated in the validation dataset. We compared our model with a literature-based model for women < 50 years (iCARE-Lit). RESULTS:Selected model risk factors were age at menarche, parity, height, current and young adulthood BMI, family history of BC, and personal BBD history. Predicted absolute 5-year risk ranged from 0% to 5.7%. The model overestimated risk on average [E/O risk = 1.18 (1.14-1.23)], with underestimation of risk in lower absolute risk deciles and overestimation in upper absolute risk deciles [E/O 1st decile = 0.59 (0.58-0.60); E/O 10th decile = 1.48 (1.48-1.49)]. The AUC was 59.1% (58.1-60.1%). Performance was similar to the iCARE-Lit model. CONCLUSION/CONCLUSIONS:In this prediction model for premenopausal BC, the relative contribution of risk factors to absolute risk was similar to existing models for overall BC. The discriminatory ability was nearly identical (< 1% difference in AUC) to the existing iCARE-Lit model developed in women under 50 years. The inability to improve discrimination highlights the need to investigate additional predictors to better understand premenopausal BC risk.

BACKGROUND:Among premenopausal women, higher body mass index (BMI) is associated with lower breast cancer risk, although the underlying mechanisms are unclear. Investigating adiposity distribution may help clarify impacts on breast cancer risk. This study was initiated to investigate associations of central and peripheral adiposity with premenopausal breast cancer risk overall and by other risk factors and breast cancer characteristics. METHODS:We used individual-level data from 14 prospective cohort studies to estimate hazard ratios (HRs) for premenopausal breast cancer using Cox proportional hazards regression. Analyses included 440,179 women followed for a median of 7.5 years (interquartile range: 4.0-11.3) between 1976 and 2017, with 6,779 incident premenopausal breast cancers. RESULTS:All central adiposity measures were inversely associated with breast cancer risk overall when not controlling for BMI (e.g. for waist circumference, HR per 10 cm increase: 0.92, 95% confidence interval (CI): 0.90-0.94) whereas in models adjusting for BMI, these measures were no longer associated with risk (e.g. for waist circumference: HR 0.99, 95% CI: 0.95-1.03). This finding was consistent across age categories, with some evidence that BMI-adjusted associations differed by breast cancer subtype. Inverse associations for in situ breast cancer were observed with waist-to-height and waist-to-hip ratios and a positive association was observed for oestrogen-receptor-positive breast cancer with hip circumference (HR per 10 cm increase: 1.08, 95% CI: 1.10-1.14). For luminal B, HER2-positive breast cancer, we observed an inverse association with hip circumference (HR per 10 cm: 0.84, 95% CI: 0.71-0.98), but positive associations with waist circumference (HR per 10 cm: 1.18, 95% CI: 1.03-1.36), waist-to-hip ratio (HR per 0.1 units: 1.29, 95% CI: 1.15-1.45) and waist-to height ratio (HR per 0.1 units: 1.46, 95% CI: 1.17-1.84). CONCLUSIONS:Our analyses did not support an association between central adiposity and overall premenopausal breast cancer risk after adjustment for BMI. However, our findings suggest associations might differ by breast cancer hormone receptor and intrinsic subtypes.

The benefits and challenges of electronic health records (EHRs) as data sources for clinical and epidemiologic research have been well described. However, several factors are important to consider when using EHR data to study novel, emerging, and multifaceted conditions such as postacute sequelae of SARS-CoV-2 infection or long COVID. In this article, we present opportunities and challenges of using EHR data to improve our understanding of long COVID, based on lessons learned from the National Institutes of Health (NIH)-funded RECOVER (REsearching COVID to Enhance Recovery) Initiative, and suggest steps to maximize the usefulness of EHR data when performing long COVID research.

BACKGROUND:The American Cancer Society recommends physicians inform average risk women about endometrial cancer (EC) risk on reaching menopause, but new diagnoses are rising fastest in women <50 years. Educating these women about EC risks requires knowledge of risk factors. However, EC in young women is rare and challenging to study in single study populations. METHODS:We included 13,846 incident EC patients (1,639 < 50 years) and 30,569 matched control individuals from the Epidemiology of Endometrial Cancer Consortium. We used generalized linear models to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for 6 risk factors and EC risk. We created a risk score to evaluate the combined associations and population attributable fractions of these factors. RESULTS:In younger and older women, we observed positive associations with BMI and diabetes, and inverse associations with age at menarche, oral contraceptive use, and parity. Current smoking was associated with reduced risk only in women ≥50 years (PHet<0.01). BMI was the strongest risk factor [OR≥35 vs <25 kg/m2=5.57 (95% CI:4.33-7.16) for <50 years; OR≥35 vs <25 kg/m2=4.68 (95% CI : 4.30-5.09) for ≥50 years; PHet=0.14]. Possessing ≥4 risk factors was associated with ∼9-fold increased risk in women <50 years and ∼4-fold increased risk in women ≥50 years (PHet<0.01). Together, 59.1% of ECs in women <50 and 55.6% in women ≥50 were attributable to these factors. CONCLUSIONS:Our data confirm younger and older women share common EC risk factors. Early educational efforts centered on these factors may help mitigate the rising EC burden in young women.

Children's mental health problems are pressing social, economic, and public health concerns in the U.S. While pediatric primary care offers important venues to integrate mental health services for children and their families, new challenges, including widening educational, economic, and health disparities in the context of structural racism and COVID-related social isolation, underscore the need for innovative approaches. The authors reviewed 6 innovative methods in pediatric care that have helped address these issues and amplify intervention efforts focused on children's mental health. Limitations and future directions for research and clinical practice in pediatric mental health services are also discussed.

PURPOSE/OBJECTIVE:To examine the association between cumulative exposure to neighborhood walkability (NW) and diabetes risk. METHODS:A total of 11,037 women free of diabetes at enrollment were included. We constructed a 4-item NW index at baseline, and a 2-item average annual NW across years of follow-up that captured both changes in neighborhood features and residential moves. We used multivariable Cox PH regression models with robust variance to estimate the hazard ratios (HRs) of diabetes by NW scores. RESULTS:Compared with women living in areas with lowest NW (Q1), those living in areas with highest NW (Q4) had 33 % (26 %-39 %) reduced risk of incident diabetes, using baseline NW, and 25 % (95 % CI 11 %-36 %), using average annual NW. Analysis using time-varying exposure showed that diabetes risks decreased by 13 % (10 %-16 %) per -standard deviation increase in NW. The associations remained similar when using inverse probability of attrition weights and/or competing risk models to account for the effect of censoring due to death or non-response. The associations of average annual NW with incident diabetes were stronger in postmenopausal women as compared to premenopausal women. CONCLUSION/CONCLUSIONS:Long-term residence in more walkable neighborhoods may be protective against diabetes in women, especially postmenopausal women.

The collapse of the World Trade Center (WTC) buildings in New York City generated a large plume of dust and smoke. WTC dust contained human carcinogens including metals, asbestos, polycyclic aromatic hydrocarbons (PAHs), persistent organic pollutants (POPs, including polychlorinated biphenyls (PCBs) and dioxins), and benzene. Excess levels of many of these carcinogens have been detected in biological samples of WTC-exposed persons, for whom cancer risk is elevated. As confirmed in this structured literature review (n studies = 80), all carcinogens present in the settled WTC dust (metals, asbestos, benzene, PAHs, POPs) have previously been shown to be associated with DNA methylation dysregulation of key cancer-related genes and pathways. DNA methylation is, therefore, a likely molecular mechanism through which WTC exposures may influence the process of carcinogenesis.

BACKGROUND:The mechanisms underlying alcohol-induced breast carcinogenesis are not fully understood but may involve hormonal changes. METHODS:Cross-sectional associations were investigated between self-reported alcohol intake and serum or plasma concentrations of estradiol, estrone, progesterone (in premenopausal women only), testosterone, androstenedione, dehydroepiandrosterone sulfate, and sex hormone binding globulin (SHBG) in 45 431 premenopausal and 173 476 postmenopausal women. Multivariable linear regression was performed separately for UK Biobank, European Prospective Investigation into Cancer and Nutrition, and Endogenous Hormones and Breast Cancer Collaborative Group, and meta-analyzed the results. For testosterone and SHBG, we also conducted Mendelian randomization and colocalization using the ADH1B (alcohol dehydrogenase 1B) variant (rs1229984). RESULTS:Alcohol intake was positively, though weakly, associated with all hormones (except progesterone in premenopausal women), with increments in concentrations per 10 g/day increment in alcohol intake ranging from 1.7% for luteal estradiol to 6.6% for postmenopausal dehydroepiandrosterone sulfate. There was an inverse association of alcohol with SHBG in postmenopausal women but a small positive association in premenopausal women. Two-sample randomization identified positive associations of alcohol intake with total testosterone (difference per 10 g/day increment: 4.1%; 95% CI, 0.6-7.6) and free testosterone (7.8%; 4.1-11.5), and an inverse association with SHBG (-8.1%; -11.3% to -4.9%). Colocalization suggested a shared causal locus at ADH1B between alcohol intake and higher free testosterone and lower SHBG (posterior probability for H4, 0.81 and 0.97, respectively). CONCLUSIONS:Alcohol intake was associated with small increases in sex hormone concentrations, including bioavailable fractions, which may contribute to its effect on breast cancer risk.

UNLABELLED:With the development of sequencing technology and analytic tools, studying within-species variations enhances the understanding of microbial biological processes. Nevertheless, most existing methods designed for strain-level analysis lack the capability to concurrently assess both strain proportions and genome-wide single nucleotide variants (SNVs) across longitudinal metagenomic samples. In this study, we introduce LongStrain, an integrated pipeline for the analysis of large-scale metagenomic data from individuals with longitudinal or repeated samples. In LongStrain, we first utilize two efficient tools, Kraken2 and Bowtie2, for the taxonomic classification and alignment of sequencing reads, respectively. Subsequently, we propose to jointly model strain proportions and shared haplotypes across samples within individuals. This approach specifically targets tracking a primary strain and a secondary strain for each subject, providing their respective proportions and SNVs as output. With extensive simulation studies of a microbial community and single species, our results demonstrate that LongStrain is superior to two genotyping methods and two deconvolution methods across a majority of scenarios. Furthermore, we illustrate the potential applications of LongStrain in the real data analysis of The Environmental Determinants of Diabetes in the Young study and a gastric intestinal metaplasia microbiome study. In summary, the proposed analytic pipeline demonstrates marked statistical efficiency over the same type of methods and has great potential in understanding the genomic variants and dynamic changes at strain level. LongStrain and its tutorial are freely available online at https://github.com/BoyanZhou/LongStrain. IMPORTANCE/OBJECTIVE:The advancement in DNA-sequencing technology has enabled the high-resolution identification of microorganisms in microbial communities. Since different microbial strains within species may contain extreme phenotypic variability (e.g., nutrition metabolism, antibiotic resistance, and pathogen virulence), investigating within-species variations holds great scientific promise in understanding the underlying mechanism of microbial biological processes. To fully utilize the shared genomic variants across longitudinal metagenomics samples collected in microbiome studies, we develop an integrated analytic pipeline (LongStrain) for longitudinal metagenomics data. It concurrently leverages the information on proportions of mapped reads for individual strains and genome-wide SNVs to enhance the efficiency and accuracy of strain identification. Our method helps to understand strains' dynamic changes and their association with genome-wide variants. Given the fast-growing longitudinal studies of microbial communities, LongStrain which streamlines analyses of large-scale raw sequencing data should be of great value in microbiome research communities.

There has been growing evidence suggesting that diabetes may be associated with increased liver cancer risk. However, studies conducted in Asian countries are limited. This project considered data of 968,738 adults pooled from 20 cohort studies of Asia Cohort Consortium to examine the association between baseline diabetes and liver cancer incidence and mortality. Cox proportional hazard model and competing risk approach was used for pooled data. Two-stage meta-analysis across studies was also done. There were 839,194 subjects with valid data regarding liver cancer incidence (5654 liver cancer cases [48.29/100,000 person-years]), follow-up time and baseline diabetes (44,781 with diabetes [5.3%]). There were 747,198 subjects with valid data regarding liver cancer mortality (5020 liver cancer deaths [44.03/100,000 person-years]), follow-up time and baseline diabetes (43,243 with diabetes [5.8%]). Hazard ratio (HR) (95% confidence interval [95%CI]) of liver cancer diagnosis in those with vs. without baseline diabetes was 1.97 (1.79, 2.16) (p < .0001) after adjusting for baseline age, gender, body mass index, tobacco smoking, alcohol use, and heterogeneity across studies (n = 586,072; events = 4620). Baseline diabetes was associated with increased cumulative incidence of death due to liver cancer (adjusted HR (95%CI) = 1.97 (1.79, 2.18); p < .0001) (n = 595,193; events = 4110). A two-stage meta-analytic approach showed similar results. This paper adds important population-based evidence to current literature regarding the increased incidence and mortality of liver cancer in adults with diabetes. The analysis of data pooled from 20 studies of different Asian countries and the meta-analysis across studies with large number of subjects makes the results robust.