To weight or not to weight? The effect of selection bias in 3 large electronic health record-linked biobanks and recommendations for practice.

OBJECTIVES: To develop recommendations regarding the use of weights to reduce selection bias for commonly performed analyses using electronic health record (EHR)-linked biobank data. MATERIALS AND METHODS: We mapped diagnosis (ICD code) data to standardized phecodes from 3 EHR-linked biobanks with varying recruitment strategies: All of Us (AOU; n = 244 071), Michigan Genomics Initiative (MGI; n = 81 243), and UK Biobank (UKB; n = 401 167). Using 2019 National Health Interview Survey data, we constructed selection weights for AOU and MGI to represent the US adult population more. We used weights previously developed for UKB to represent the UKB-eligible population. We conducted 4 common analyses comparing unweighted and weighted results. RESULTS: For AOU and MGI, estimated phecode prevalences decreased after weighting (weighted-unweighted median phecode prevalence ratio [MPR]: 0.82 and 0.61), while UKB estimates increased (MPR: 1.06). Weighting minimally impacted latent phenome dimensionality estimation. Comparing weighted versus unweighted phenome-wide association study for colorectal cancer, the strongest associations remained unaltered, with considerable overlap in significant hits. Weighting affected the estimated log-odds ratio for sex and colorectal cancer to align more closely with national registry-based estimates. DISCUSSION: Weighting had a limited impact on dimensionality estimation and large-scale hypothesis testing but impacted prevalence and association estimation. When interested in estimating effect size, specific signals from untargeted association analyses should be followed up by weighted analysis. CONCLUSION: EHR-linked biobanks should report recruitment and selection mechanisms and provide selection weights with defined target populations. Researchers should consider their intended estimands, specify source and target populations, and weight EHR-linked biobank analyses accordingly.

To weight or not to weight? Studying the effect of selection bias in three large EHR-linked biobanks.

Objective: To explore the role of selection bias adjustment by weighting electronic health record (EHR)-linked biobank data for commonly performed analyses. Materials and methods: We mapped diagnosis (ICD code) data to standardized phecodes from three EHR-linked biobanks with varying recruitment strategies: All of Us (AOU; n=244,071), Michigan Genomics Initiative (MGI; n=81,243), and UK Biobank (UKB; n=401,167). Using 2019 National Health Interview Survey data, we constructed selection weights for AOU and MGI to be more representative of the US adult population. We used weights previously developed for UKB to represent the UKB-eligible population. We conducted four common descriptive and analytic tasks comparing unweighted and weighted results. Results: For AOU and MGI, estimated phecode prevalences decreased after weighting (weighted-unweighted median phecode prevalence ratio [MPR]: 0.82 and 0.61), while UKB's estimates increased (MPR: 1.06). Weighting minimally impacted latent phenome dimensionality estimation. Comparing weighted versus unweighted PheWAS for colorectal cancer, the strongest associations remained unaltered and there was large overlap in significant hits. Weighting affected the estimated log-odds ratio for sex and colorectal cancer to align more closely with national registry-based estimates. Discussion: Weighting had limited impact on dimensionality estimation and large-scale hypothesis testing but impacted prevalence and association estimation more. Results from untargeted association analyses should be followed by weighted analysis when effect size estimation is of interest for specific signals. Conclusion: EHR-linked biobanks should report recruitment and selection mechanisms and provide selection weights with defined target populations. Researchers should consider their intended estimands, specify source and target populations, and weight EHR-linked biobank analyses accordingly.

Mortality and Severe Complications Among Newly Graduated Surgeons in the United States.

OBJECTIVE: To evaluate severe complications and mortality over years of independent practice among general surgeons. BACKGROUND: Despite concerns that newly graduated general surgeons may be unprepared for independent practice, it is unclear whether patient outcomes differ between early and later career surgeons. METHODS: We used Medicare claims for patients discharged between July 1, 2007 and December 31, 2019 to evaluate 30-day severe complications and mortality for 26 operations defined as core procedures by the American Board of Surgery. Generalized additive mixed models were used to assess the association between surgeon years in practice and 30-day outcomes while adjusting for differences in patient, hospital, and surgeon characteristics. RESULTS: The cohort included 1,329,358 operations performed by 14,399 surgeons. In generalized mixed models, the relative risk (RR) of mortality was higher among surgeons in their first year of practice compared with surgeons in their 15th year of practice [5.5% (95% CI: 4.1%-7.3%) vs 4.7% (95% CI: 3.5%-6.3%), RR: 1.17 (95% CI: 1.11-1.22)]. Similarly, the RR of severe complications was higher among surgeons in their first year of practice compared with surgeons in their 15th year of practice [7.5% (95% CI: 6.6%-8.5%) versus 6.9% (95% CI: 6.1%-7.9%), RR: 1.08 (95% CI: 1.03-1.14)]. When stratified by individual operation, 21 operations had a significantly higher RR of mortality and all 26 operations had a significantly higher RR of severe complications in the first compared with the 15th year of practice. CONCLUSIONS: Among general surgeons performing common operations, rates of mortality and severe complications were higher among newly graduated surgeons compared with later career surgeons.

Epidemiologic Questionnaire (EPI-Q) - a scalable, app-based health survey linked to electronic health record and genotype data.

The Epidemiologic Questionnaire (EPI-Q) was established to collect broad, uniform, self-reported health data to supplement electronic health record (EHR) and genotype information from participants in the University of Michigan (UM) Precision Health cohorts. Recruitment of EPI-Q participants, who were already enrolled in 1 of 3 ongoing UM Precision Health cohorts-the Michigan Genomics Initiative, Mental Health Biobank, and Metabolism, Endocrinology, and Diabetes cohorts-began in March 2020. Of 54,043 retrospective invitations, 5,577 individuals enrolled, representing a 10.3% response rate. Of these, 3,502 (63.7%) were female, and the average age was 56.1 years (standard deviation, 15.4). The baseline survey comprises 11 modules on topics including personal and family health history, lifestyle, and cancer screening and history. Additionally, 11 optional modules cover topics including financial toxicity, occupational exposure, and life meaning. The questions are based on standardized and validated instruments used in other cohorts, and we share resources to expedite development of similar surveys. Data are collected via the MyDataHelps platform, which enables current and future participants to share non-Michigan Medicine EHR data. Recruitment is ongoing. Cohort data are available to those with institutional review board approval; for details, contact the Data Office for Clinical and Translational Research (DataOffice@umich.edu).

Prenatal per- and polyfluoroalkyl substances (PFAS) exposure in relation to preterm birth subtypes and size-for-gestational age in the LIFECODES cohort 2006-2008.

BACKGROUND: Per- and polyfluoroalkyl substances (PFAS) are a group of synthetic chemicals widely used in consumer and industrial products. Numerous studies have linked prenatal PFAS exposures to increased risks of adverse pregnancy outcomes such as preterm birth (PTB) and small-for-gestational age (SGA).However, limited evidence is available for the effects of PFAS on PTB subtypes and large-for-gestational age (LGA). OBJECTIVE: To examine the associations of PFAS with PTB [overall, placental (pPTB), spontaneous (sPTB)], BW Z-score, and size-for-gestational age (SGA, LGA). METHODS: Our nested case-control study included 128 preterm cases and 373 term controls from the LIFECODES cohort between 2006 and 2008 (n = 501). Plasma concentrations of nine PFAS were measured in early pregnancy samples. Logistic regression was used to assess individual PFAS-birth outcome associations, while Bayesian Kernel Machine Regression (BKMR) was used to evaluate the joint effects of all PFAS. Effect modification by fetal sex was examined, and stratified analyses were conducted to obtain fetal sex-specific estimates. RESULTS: Compared to term births, the odds of pPTB were higher from an interquartile range increase in perfluorodecanoic acid (PFDA) (OR = 1.60, 95% CI: 1.00-2.56), perfluorononanoic acid (PFNA) (OR = 1.67, 95% CI: 1.06-2.61), and perfluoroundecanoic acid (PFUA) (OR = 1.77, 95% CI: 1.00-3.12), with stronger associations observed in women who delivered males. BKMR analysis identified PFNA as the most important PFAS responsible for pPTB (conditional PIP = 0.78), with increasing ORs at higher percentiles of PFAS mixture. For LGA, positive associations were observed with PFDA and perfluorooctanoic acid in females only, and with PFUA in males only. BKMR analysis showed increasing, but null effects of PFAS mixture on LGA. CONCLUSIONS: The effect of prenatal exposure to single and multiple PFAS on PTB and LGA depended on fetal sex. Future studies should strongly consider examining PTB subtypes and sex-specific effects of PFAS on pregnancy outcomes.

An inverse probability weighted regression method that accounts for right-censoring for causal inference with multiple treatments and a binary outcome.

Comparative effectiveness research often involves evaluating the differences in the risks of an event of interest between two or more treatments using observational data. Often, the post-treatment outcome of interest is whether the event happens within a pre-specified time window, which leads to a binary outcome. One source of bias for estimating the causal treatment effect is the presence of confounders, which are usually controlled using propensity score-based methods. An additional source of bias is right-censoring, which occurs when the information on the outcome of interest is not completely available due to dropout, study termination, or treatment switch before the event of interest. We propose an inverse probability weighted regression-based estimator that can simultaneously handle both confounding and right-censoring, calling the method CIPWR, with the letter C highlighting the censoring component. CIPWR estimates the average treatment effects by averaging the predicted outcomes obtained from a logistic regression model that is fitted using a weighted score function. The CIPWR estimator has a double robustness property such that estimation consistency can be achieved when either the model for the outcome or the models for both treatment and censoring are correctly specified. We establish the asymptotic properties of the CIPWR estimator for conducting inference, and compare its finite sample performance with that of several alternatives through simulation studies. The methods under comparison are applied to a cohort of prostate cancer patients from an insurance claims database for comparing the adverse effects of four candidate drugs for advanced stage prostate cancer.

A framework for assessing interactions for risk stratification models: the example of ovarian cancer.

Generally, risk stratification models for cancer use effect estimates from risk/protective factor analyses that have not assessed potential interactions between these exposures. We have developed a 4-criterion framework for assessing interactions that includes statistical, qualitative, biological, and practical approaches. We present the application of this framework in an ovarian cancer setting because this is an important step in developing more accurate risk stratification models. Using data from 9 case-control studies in the Ovarian Cancer Association Consortium, we conducted a comprehensive analysis of interactions among 15 unequivocal risk and protective factors for ovarian cancer (including 14 non-genetic factors and a 36-variant polygenic score) with age and menopausal status. Pairwise interactions between the risk/protective factors were also assessed. We found that menopausal status modifies the association among endometriosis, first-degree family history of ovarian cancer, breastfeeding, and depot-medroxyprogesterone acetate use and disease risk, highlighting the importance of understanding multiplicative interactions when developing risk prediction models.

A comparative study on behavior, awareness and belief about cervical cancer among rural and urban women in Vietnam.

Cervical cancer is the second most common gynecologic cancer in Vietnam but based on the literature, only ~25% of Vietnamese women reported ever being screened for cervical cancer. To inform strategies to reduce the cervical cancer burden in Southern Vietnam where disease incidence is higher than the national average, this study examined behaviors, awareness, barriers, and beliefs about cervical cancer screening among rural and urban women in this geographical region. In October-November 2021, we conducted a cross-sectional study among 196 rural and 202 urban women in Southern Vietnam; participants completed a cervical cancer screening questionnaire. Descriptive analyses and rural-urban differences in screening behavior, awareness, barriers, and beliefs are presented. About half of the rural and urban participants reported ever being screened for cervical cancer. Most participants showed high perceived severity of cervical cancer and benefits of screening. Further, they reported that they would screen if it was recommended by doctors and/or friends/family. However, most women showed low awareness and perceived susceptibility to cervical cancer. Logistical and psychosocial barriers to physician-based screening methods were reported. Based on our results, the World Health Organization 2030 goals for cervical cancer screening are not currently met in Southern Vietnam. Increasing health literacy and engaging doctors and family members/social networks emerged as important avenues to improve screening. HPV (Human papillomavirus) self-sampling is also a potential approach to increase uptake of cervical cancer screening given the identified psychosocial and logistical barriers.

Decomposing interaction and mediating effects of race/ethnicity and circulating blood levels of cystatin C on cognitive status in the United States health and retirement study.

Background and objectives: Elevated circulating cystatin C is associated with cognitive impairment in non-Hispanic Whites, but its role in racial disparities in dementia is understudied. In a nationally representative sample of older non-Hispanic White, non-Hispanic Black, and Hispanic adults in the United States, we use mediation-interaction analysis to understand how racial disparities in the cystatin C physiological pathway may contribute to racial disparities in prevalent dementia. Methods: In a pooled cross-sectional sample of the Health and Retirement Study (n = 9,923), we employed Poisson regression to estimate prevalence ratios and to test the relationship between elevated cystatin C (>1.24 vs. ≤1.24 mg/L) and impaired cognition, adjusted for demographics, behavioral risk factors, other biomarkers, and chronic conditions. Self-reported racialized social categories were a proxy measure for exposure to racism. We calculated additive interaction measures and conducted four-way mediation-interaction decomposition analysis to test the moderating effect of race/ethnicity and mediating effect of cystatin C on the racial disparity. Results: Overall, elevated cystatin C was associated with dementia (prevalence ratio [PR] = 1.2; 95% CI: 1.0, 1.5). Among non-Hispanic Black relative to non-Hispanic White participants, the relative excess risk due to interaction was 0.7 (95% CI: -0.1, 2.4), the attributable proportion was 0.1 (95% CI: -0.2, 0.4), and the synergy index was 1.1 (95% CI: 0.8, 1.8) in a fully adjusted model. Elevated cystatin C was estimated to account for 2% (95% CI: -0, 4%) for the racial disparity in prevalent dementia, and the interaction accounted for 8% (95% CI: -5, 22%). Analyses for Hispanic relative to non-white participants suggested moderation by race/ethnicity, but not mediation. Discussion: Elevated cystatin C was associated with dementia prevalence. Our mediation-interaction decomposition analysis suggested that the effect of elevated cystatin C on the racial disparity might be moderated by race/ethnicity, which indicates that the racialization process affects not only the distribution of circulating cystatin C across minoritized racial groups, but also the strength of association between the biomarker and dementia prevalence. These results provide evidence that cystatin C is associated with adverse brain health and this effect is larger than expected for individuals racialized as minorities had they been racialized and treated as non-Hispanic White.

Integrating Information from Existing Risk Prediction Models with No Model Details.

Consider the setting where (i) individual-level data are collected to build a regression model for the association between an event of interest and certain covariates, and (ii) some risk calculators predicting the risk of the event using less detailed covariates are available, possibly as algorithmic black boxes with little information available about how they were built. We propose a general empirical-likelihood-based framework to integrate the rich auxiliary information contained in the calculators into fitting the regression model, to make the estimation of regression parameters more efficient. Two methods are developed, one using working models to extract the calculator information and one making a direct use of calculator predictions without working models. Theoretical and numerical investigations show that the calculator information can substantially reduce the variance of regression parameter estimation. As an application, we study the dependence of the risk of high grade prostate cancer on both conventional risk factors and newly identified molecular biomarkers by integrating information from the Prostate Biopsy Collaborative Group (PBCG) risk calculator, which was built based on conventional risk factors alone.

Insérer votre résumé ici. We will supply a French abstract for those authors who can't prepare it themselves.

A synthetic data integration framework to leverage external summary-level information from heterogeneous populations.

There is a growing need for flexible general frameworks that integrate individual-level data with external summary information for improved statistical inference. External information relevant for a risk prediction model may come in multiple forms, through regression coefficient estimates or predicted values of the outcome variable. Different external models may use different sets of predictors and the algorithm they used to predict the outcome Y given these predictors may or may not be known. The underlying populations corresponding to each external model may be different from each other and from the internal study population. Motivated by a prostate cancer risk prediction problem where novel biomarkers are measured only in the internal study, this paper proposes an imputation-based methodology, where the goal is to fit a target regression model with all available predictors in the internal study while utilizing summary information from external models that may have used only a subset of the predictors. The method allows for heterogeneity of covariate effects across the external populations. The proposed approach generates synthetic outcome data in each external population, uses stacked multiple imputation to create a long dataset with complete covariate information. The final analysis of the stacked imputed data is conducted by weighted regression. This flexible and unified approach can improve statistical efficiency of the estimated coefficients in the internal study, improve predictions by utilizing even partial information available from models that use a subset of the full set of covariates used in the internal study, and provide statistical inference for the external population with potentially different covariate effects from the internal population.

Association of Surgical Resident Competency Ratings With Patient Outcomes.

PURPOSE: Accurate assessment of clinical performance is essential to ensure graduating residents are competent for unsupervised practice. The Accreditation Council for Graduate Medical Education milestones framework is the most widely used competency-based framework in the United States. However, the relationship between residents' milestones competency ratings and their subsequent early career clinical outcomes has not been established. It is important to examine the association between milestones competency ratings of U.S. general surgical residents and those surgeons' patient outcomes in early career practice. METHOD: A retrospective, cross-sectional study was conducted using a sample of national Medicare claims for 23 common, high-risk inpatient general surgical procedures performed between July 1, 2015, and November 30, 2018 (n = 12,400 cases) by nonfellowship-trained U.S. general surgeons. Milestone ratings collected during those surgeons' last year of residency (n = 701 residents) were compared with their risk-adjusted rates of mortality, any complication, or severe complication within 30 days of index operation during their first 2 years of practice. RESULTS: There were no associations between mean milestone competency ratings of graduating general surgery residents and their subsequent early career patient outcomes, including any complication (23% proficient vs 22% not yet proficient; relative risk [RR], 0.97, [95% CI, 0.88-1.08]); severe complication (9% vs 9%, respectively; RR, 1.01, [95% CI, 0.86-1.19]); and mortality (5% vs 5%; RR, 1.07, [95% CI, 0.88-1.30]). Secondary analyses yielded no associations between patient outcomes and milestone ratings specific to technical performance, or between patient outcomes and composites of operative performance, professionalism, or leadership milestones ratings ( P ranged .32-.97). CONCLUSIONS: Milestone ratings of graduating general surgery residents were not associated with the patient outcomes of those surgeons when they performed common, higher-risk procedures in a Medicare population. Efforts to improve how milestones ratings are generated might strengthen their association with early career outcomes.

Improving the Prediction of Death from Cardiovascular Causes with Multiple Risk Markers.

Background: Traditional risk factors including demographics, blood pressure, cholesterol, and diabetes status are successfully able to predict a proportion of cardiovascular disease (CVD) events. Whether including additional routinely measured factors improves CVD prediction is unclear. To determine whether a comprehensive risk factor list, including clinical blood measures, blood counts, anthropometric measures, and lifestyle factors, improves prediction of CVD deaths beyond traditional factors. Methods: The analysis comprised of 21,982 participants aged 40 years and older (mean age=59.4 years at baseline) from the National Health and Nutrition Examination Survey (NHANES) from 2001 to 2016 survey cycles. Data were linked with the National Death Index mortality data through 2019 and split into 80:20 training and testing sets. Relative to the traditional risk factors (age, sex, race/ethnicity, smoking status, systolic blood pressure, total and high-density lipoprotein cholesterol, antihypertensive medications, and diabetes), we compared models with an additional 22 clinical blood biomarkers, 20 complete blood counts, 7 anthropometric measures, 51 dietary factors, 13 cardiovascular health-related questions, and all 113 predictors together. To build prediction models for CVD mortality, we performed Cox proportional hazards regression, elastic-net (ENET) penalized Cox regression, and random survival forest, and compared classification using C-index and net reclassification improvement. Results: During follow-up (median, 9.3 years), 3,075 participants died; 30.9% (1,372/3,075) deaths were from cardiovascular causes. In Cox proportional hazards models with traditional risk factors (C-index=0.850), CVD mortality classification improved with incorporation of clinical blood biomarkers (C-index=0.867), blood counts (C-index=0.861), and all predictors (C-index=0.871). Net CVD mortality reclassification improved 13.2% by adding clinical blood biomarkers and 12.2% by adding all predictors. Results for ENET-penalized Cox regression and random survival forest were similar. No improvement was observed in separate models for anthropometric measures, dietary nutrient intake, or cardiovascular health-related questions. Conclusions: The addition of clinical blood biomarkers and blood counts substantially improves CVD mortality prediction, beyond traditional risk factors. These biomarkers may serve as an important clinical and public health screening tool for the prevention of CVD deaths.

COVID-19 Outcomes by Cancer Status, Site, Treatment, and Vaccination.

BACKGROUND: Studies have shown an increased risk of severe SARS-CoV-2-related (COVID-19) disease outcome and mortality for patients with cancer, but it is not well understood whether associations vary by cancer site, cancer treatment, and vaccination status. METHODS: Using electronic health record data from an academic medical center, we identified a retrospective cohort of 260,757 individuals tested for or diagnosed with COVID-19 from March 10, 2020, to August 1, 2022. Of these, 52,019 tested positive for COVID-19 of whom 13,752 had a cancer diagnosis. We conducted Firth-corrected logistic regression to assess the association between cancer status, site, treatment, vaccination, and four COVID-19 outcomes: hospitalization, intensive care unit admission, mortality, and a composite "severe COVID" outcome. RESULTS: Cancer diagnosis was significantly associated with higher rates of severe COVID, hospitalization, and mortality. These associations were driven by patients whose most recent initial cancer diagnosis was within the past 3 years. Chemotherapy receipt, colorectal cancer, hematologic malignancies, kidney cancer, and lung cancer were significantly associated with higher rates of worse COVID-19 outcomes. Vaccinations were significantly associated with lower rates of worse COVID-19 outcomes regardless of cancer status. CONCLUSIONS: Patients with colorectal cancer, hematologic malignancies, kidney cancer, or lung cancer or who receive chemotherapy for treatment should be cautious because of their increased risk of worse COVID-19 outcomes, even after vaccination. IMPACT: Additional COVID-19 precautions are warranted for people with certain cancer types and treatments. Significant benefit from vaccination is noted for both cancer and cancer-free patients.

Lifestyle and personal factors associated with having macroscopic residual disease after ovarian cancer primary cytoreductive surgery.

OBJECTIVE: The presence of macroscopic residual disease after primary cytoreductive surgery (PCS) is an important factor influencing survival for patients with high-grade serous ovarian cancer (HGSC). More research is needed to identify factors associated with having macroscopic residual disease. We analyzed 12 lifestyle and personal exposures known to be related to ovarian cancer risk or inflammation to identify those associated with having residual disease after surgery. METHODS: This analysis used data on 2054 patients with advanced stage HGSC from the Ovarian Cancer Association Consortium. The exposures were body mass index, breastfeeding, oral contraceptive use, depot-medroxyprogesterone acetate use, endometriosis, first-degree family history of ovarian cancer, incomplete pregnancy, menopausal hormone therapy use, menopausal status, parity, smoking, and tubal ligation. Logistic regression models were fit to assess the association between these exposures and having residual disease following PCS. RESULTS: Menopausal estrogen-only therapy (ET) use was associated with 33% lower odds of having macroscopic residual disease compared to never use (OR = 0.67, 95%CI 0.46-0.97, p = 0.033). Compared to nulliparous women, parous women who did not breastfeed had 36% lower odds of having residual disease (OR = 0.64, 95%CI 0.43-0.94, p = 0.022), while there was no association among parous women who breastfed (OR = 0.90, 95%CI 0.65-1.25, p = 0.53). CONCLUSIONS: The association between ET and having no macroscopic residual disease is plausible given a strong underlying biologic hypothesis between this exposure and diagnosis with HGSC. If this or the parity finding is replicated, these factors could be included in risk stratification models to determine whether HGSC patients should receive PCS or neoadjuvant chemotherapy.

Exposure to heavy metals and hormone levels in midlife women: The Study of Women's Health Across the Nation (SWAN).

Exposure to heavy metals may alter the circulating levels of sex hormones. However, epidemiologic studies on heavy metals and sex hormones have been limited, and results have been inconsistent. We assessed the associations of heavy metals assayed in urine, including arsenic, cadmium, lead, and mercury, with repeated measures of serum estradiol (E2), follicle-stimulating hormone (FSH), testosterone, and sex hormone-binding globulin (SHBG) levels in the Study of Women's Health Across the Nation Multi-Pollutant Study. The sample included 1355 White, Black, Chinese, and Japanese women, aged 45-56 years at baseline (1999-2000), whose serum hormone levels were repeatedly measured through 2017. Urinary metal concentrations were measured at baseline. Linear mixed effect models were used to calculate percent changes in serum hormone levels per doubling of urinary metal concentrations, adjusting for demographics, socioeconomic status, lifestyle, health-related factors, and urinary creatinine. After multivariable adjustment, a doubling of urinary metal concentration was associated with lower E2 levels by 2.2% (95% CI: 4.0%, -0.3%) for mercury and 3.6% (95% CI: 5.7%, -1.6%) for lead; higher FSH levels by 3.4% (95% CI: 0.9%, 5.9%) for lead; and higher SHBG levels by 3.6% (95% CI: 1.3%, 5.9%) for cadmium. The overall joint effect using the Bayesian kernel machine regression showed that metal mixtures were inversely associated with E2 and positively associated with FSH levels. No association was found between metals and testosterone levels. Results from this prospective cohort study demonstrate that environmental heavy metal exposures, including cadmium, mercury, and lead, may disturb circulating levels of E2, FSH, and SHBG in midlife women.

Predicting cumulative lead (Pb) exposure using the Super Learner algorithm.

Chronic lead (Pb) exposure causes long term health effects. While recent exposure can be assessed by measuring blood lead (half-life 30 days), chronic exposures can be assessed by measuring lead in bone (half-life of many years to decades). Bone lead measurements, in turn, have been measured non-invasively in large population-based studies using x-ray fluorescence techniques, but the method remains limited due to technical availability, expense, and the need for licensing radioactive materials used by the instruments. Thus, we developed prediction models for bone lead concentrations using a flexible machine learning approach--Super Learner, which combines the predictions from a set of machine learning algorithms for better prediction performance. The study population included 695 men in the Normative Aging Study, aged 48 years and older, whose bone (patella and tibia) lead concentrations were directly measured using K-shell-X-ray fluorescence. Ten predictors (blood lead, age, education, job type, weight, height, body mass index, waist circumference, cumulative cigarette smoking (pack-year), and smoking status) were selected for patella lead and 11 (the same 10 predictors plus serum phosphorus) for tibia lead using the Boruta algorithm. We implemented Super Learner to predict bone lead concentrations by calculating a weighted combination of predictions from 8 algorithms. In the nested cross-validation, the correlation coefficients between measured and predicted bone lead concentrations were 0.58 for patella lead and 0.52 for tibia lead, which has improved the correlations obtained in previously-published linear regression-based prediction models. We evaluated the applicability of these prediction models to the National Health and Nutrition Examination Survey for the associations between predicted bone lead concentrations and blood pressure, and positive associations were observed. These bone lead prediction models provide reasonable accuracy and can be used to evaluate health effects of cumulative lead exposure in studies where bone lead is not measured.

Case studies in bias reduction and inference for electronic health record data with selection bias and phenotype misclassification.

Electronic health records (EHR) are not designed for population-based research, but they provide easy and quick access to longitudinal health information for a large number of individuals. Many statistical methods have been proposed to account for selection bias, missing data, phenotyping errors, or other problems that arise in EHR data analysis. However, addressing multiple sources of bias simultaneously is challenging. We developed a methodological framework (R package, SAMBA) for jointly handling both selection bias and phenotype misclassification in the EHR setting that leverages external data sources. These methods assume factors related to selection and misclassification are fully observed, but these factors may be poorly understood and partially observed in practice. As a follow-up to the methodological work, we demonstrate how to apply these methods for two real-world case studies, and we evaluate their performance. In both examples, we use individual patient-level data collected through the University of Michigan Health System and various external population-based data sources. In case study (a), we explore the impact of these methods on estimated associations between gender and cancer diagnosis. In case study (b), we compare corrected associations between previously identified genetic loci and age-related macular degeneration with gold standard external summary estimates. These case studies illustrate how to utilize diverse auxiliary information to achieve less biased inference in EHR-based research.

ExPRSweb: An online repository with polygenic risk scores for common health-related exposures.

Complex traits are influenced by genetic risk factors, lifestyle, and environmental variables, so-called exposures. Some exposures, e.g., smoking or lipid levels, have common genetic modifiers identified in genome-wide association studies. Because measurements are often unfeasible, exposure polygenic risk scores (ExPRSs) offer an alternative to study the influence of exposures on various phenotypes. Here, we collected publicly available summary statistics for 28 exposures and applied four common PRS methods to generate ExPRSs in two large biobanks: the Michigan Genomics Initiative and the UK Biobank. We established ExPRSs for 27 exposures and demonstrated their applicability in phenome-wide association studies and as predictors for common chronic conditions. Especially the addition of multiple ExPRSs showed, for several chronic conditions, an improvement compared to prediction models that only included traditional, disease-focused PRSs. To facilitate follow-up studies, we share all ExPRS constructs and generated results via an online repository called ExPRSweb.

Polygenic Liability to Depression Is Associated With Multiple Medical Conditions in the Electronic Health Record: Phenome-wide Association Study of 46,782 Individuals.

BACKGROUND: Major depressive disorder (MDD) is a leading cause of disease-associated disability, with much of the increased burden due to psychiatric and medical comorbidity. This comorbidity partly reflects common genetic influences across conditions. Integrating molecular-genetic tools with health records enables tests of association with the broad range of physiological and clinical phenotypes. However, standard phenome-wide association studies analyze associations with individual genetic variants. For polygenic traits such as MDD, aggregate measures of genetic risk may yield greater insight into associations across the clinical phenome. METHODS: We tested for associations between a genome-wide polygenic risk score for MDD and medical and psychiatric traits in a phenome-wide association study of 46,782 unrelated, European-ancestry participants from the Michigan Genomics Initiative. RESULTS: The MDD polygenic risk score was associated with 211 traits from 15 medical and psychiatric disease categories at the phenome-wide significance threshold. After excluding patients with depression, continued associations were observed with respiratory, digestive, neurological, and genitourinary conditions; neoplasms; and mental disorders. Associations with tobacco use disorder, respiratory conditions, and genitourinary conditions persisted after accounting for genetic overlap between depression and other psychiatric traits. Temporal analyses of time-at-first-diagnosis indicated that depression disproportionately preceded chronic pain and substance-related disorders, while asthma disproportionately preceded depression. CONCLUSIONS: The present results can inform the biological links between depression and both mental and systemic diseases. Although MDD polygenic risk scores cannot currently forecast health outcomes with precision at the individual level, as molecular-genetic discoveries for depression increase, these tools may augment risk prediction for medical and psychiatric conditions.