Large-scale measurement of aggregate human colocation patterns for epidemiological modeling.

To understand and model public health emergencies, epidemiologists need data that describes how humans are moving and interacting across physical space. Such data has traditionally been difficult for researchers to obtain with the temporal resolution and geographic breadth that is needed to study, for example, a global pandemic. This paper describes Colocation Maps, which are spatial network datasets that have been developed within Meta's Data For Good program. These Maps estimate how often people from different regions are colocated: in particular, for a pair of geographic regions x and y, these Maps estimate the rate at which a randomly chosen person from x and a randomly chosen person from y are simultaneously located in the same place during a randomly chosen minute in a given week. These datasets are well suited to parametrize metapopulation models of disease spread or to measure temporal changes in interactions between people from different regions; indeed, they have already been used for both of these purposes during the COVID-19 pandemic. In this paper, we show how Colocation Maps differ from existing data sources, describe how the datasets are built, provide examples of their use in compartmental modeling, and summarize ideas for further development of these and related datasets. Among the findings of this study, we observe that a pair of regions can exhibit high colocation despite few people moving between those regions. Additionally, for the purposes of clarifying how to interpret and utilize Colocation Maps, we scrutinize the Maps' built-in assumptions about representativeness and contact heterogeneity.

iCARE: An R package to build, validate and apply absolute risk models.

This report describes an R package, called the Individualized Coherent Absolute Risk Estimator (iCARE) tool, that allows researchers to build and evaluate models for absolute risk and apply them to estimate an individual's risk of developing disease during a specified time interval based on a set of user defined input parameters. An attractive feature of the software is that it gives users flexibility to update models rapidly based on new knowledge on risk factors and tailor models to different populations by specifying three input arguments: a model for relative risk, an age-specific disease incidence rate and the distribution of risk factors for the population of interest. The tool can handle missing information on risk factors for individuals for whom risks are to be predicted using a coherent approach where all estimates are derived from a single model after appropriate model averaging. The software allows single nucleotide polymorphisms (SNPs) to be incorporated into the model using published odds ratios and allele frequencies. The validation component of the software implements the methods for evaluation of model calibration, discrimination and risk-stratification based on independent validation datasets. We provide an illustration of the utility of iCARE for building, validating and applying absolute risk models using breast cancer as an example.

Constrained Maximum Likelihood Estimation for Model Calibration Using Summary-level Information from External Big Data Sources.

Information from various public and private data sources of extremely large sample sizes are now increasingly available for research purposes. Statistical methods are needed for utilizing information from such big data sources while analyzing data from individual studies that may collect more detailed information required for addressing specific hypotheses of interest. In this article, we consider the problem of building regression models based on individual-level data from an "internal" study while utilizing summary-level information, such as information on parameters for reduced models, from an "external" big data source. We identify a set of very general constraints that link internal and external models. These constraints are used to develop a framework for semiparametric maximum likelihood inference that allows the distribution of covariates to be estimated using either the internal sample or an external reference sample. We develop extensions for handling complex stratified sampling designs, such as case-control sampling, for the internal study. Asymptotic theory and variance estimators are developed for each case. We use simulation studies and a real data application to assess the performance of the proposed methods in contrast to the generalized regression (GR) calibration methodology that is popular in the sample survey literature.

Breast Cancer Risk From Modifiable and Nonmodifiable Risk Factors Among White Women in the United States.

IMPORTANCE: An improved model for risk stratification can be useful for guiding public health strategies of breast cancer prevention. OBJECTIVE: To evaluate combined risk stratification utility of common low penetrant single nucleotide polymorphisms (SNPs) and epidemiologic risk factors. DESIGN, SETTING, AND PARTICIPANTS: Using a total of 17 171 cases and 19 862 controls sampled from the Breast and Prostate Cancer Cohort Consortium (BPC3) and 5879 women participating in the 2010 National Health Interview Survey, a model for predicting absolute risk of breast cancer was developed combining information on individual level data on epidemiologic risk factors and 24 genotyped SNPs from prospective cohort studies, published estimate of odds ratios for 68 additional SNPs, population incidence rate from the National Cancer Institute-Surveillance, Epidemiology, and End Results Program cancer registry and data on risk factor distribution from nationally representative health survey. The model is used to project the distribution of absolute risk for the population of white women in the United States after adjustment for competing cause of mortality. EXPOSURES: Single nucleotide polymorphisms, family history, anthropometric factors, menstrual and/or reproductive factors, and lifestyle factors. MAIN OUTCOMES AND MEASURES: Degree of stratification of absolute risk owing to nonmodifiable (SNPs, family history, height, and some components of menstrual and/or reproductive history) and modifiable factors (body mass index [BMI; calculated as weight in kilograms divided by height in meters squared], menopausal hormone therapy [MHT], alcohol, and smoking). RESULTS: The average absolute risk for a 30-year-old white woman in the United States developing invasive breast cancer by age 80 years is 11.3%. A model that includes all risk factors provided a range of average absolute risk from 4.4% to 23.5% for women in the bottom and top deciles of the risk distribution, respectively. For women who were at the lowest and highest deciles of nonmodifiable risks, the 5th and 95th percentile range of the risk distribution associated with 4 modifiable factors was 2.9% to 5.0% and 15.5% to 25.0%, respectively. For women in the highest decile of risk owing to nonmodifiable factors, those who had low BMI, did not drink or smoke, and did not use MHT had risks comparable to an average woman in the general population. CONCLUSIONS AND RELEVANCE: This model for absolute risk of breast cancer including SNPs can provide stratification for the population of white women in the United States. The model can also identify subsets of the population at an elevated risk that would benefit most from risk-reduction strategies based on altering modifiable factors. The effectiveness of this model for individual risk communication needs further investigation.

Alcohol consumption and breast cancer risk by estrogen receptor status: in a pooled analysis of 20 studies.

BACKGROUND: Breast cancer aetiology may differ by estrogen receptor (ER) status. Associations of alcohol and folate intakes with risk of breast cancer defined by ER status were examined in pooled analyses of the primary data from 20 cohorts. METHODS: During a maximum of 6-18 years of follow-up of 1 089 273 women, 21 624 ER+ and 5113 ER- breast cancers were identified. Study-specific multivariable relative risks (RRs) were calculated using Cox proportional hazards regression models and then combined using a random-effects model. RESULTS: Alcohol consumption was positively associated with risk of ER+ and ER- breast cancer. The pooled multivariable RRs (95% confidence intervals) comparing ≥ 30 g/d with 0 g/day of alcohol consumption were 1.35 (1.23-1.48) for ER+ and 1.28 (1.10-1.49) for ER- breast cancer (Ptrend ≤ 0.001; Pcommon-effects by ER status: 0.57). Associations were similar for alcohol intake from beer, wine and liquor. The associations with alcohol intake did not vary significantly by total (from foods and supplements) folate intake (Pinteraction ≥ 0.26). Dietary (from foods only) and total folate intakes were not associated with risk of overall, ER+ and ER- breast cancer; pooled multivariable RRs ranged from 0.98 to 1.02 comparing extreme quintiles. Following-up US studies through only the period before mandatory folic acid fortification did not change the results. The alcohol and folate associations did not vary by tumour subtypes defined by progesterone receptor status. CONCLUSIONS: Alcohol consumption was positively associated with risk of both ER+ and ER- breast cancer, even among women with high folate intake. Folate intake was not associated with breast cancer risk.

Alcohol and risk of breast cancer in postmenopausal women: an analysis of etiological heterogeneity by multiple tumor characteristics.

Alcohol consumption is an established risk factor for breast cancer. Whether associations vary by specific tumor characteristics independent of other characteristics is unclear. We evaluated the association between alcohol consumption and breast cancer risk in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial cohort (54,562 women aged 55-74 years recruited at 10 US screening centers between 1993 and 2001; median follow-up, 8.9 years; 1,905 invasive breast cancer cases). Hazard ratios and 95% confidence intervals for subtypes defined by histological type and estrogen receptor (ER)/progesterone receptor (PR) status were calculated with standard Cox models. A novel 2-stage Cox model assessed heterogeneity in risk for individual tumor characteristics while adjusting for others. Significant trends across categories of alcohol consumption were observed, with hazard ratios for those consuming 7 or more drinks per week versus never drinkers as follows: for estrogen receptor-positive (ER+) cancer, 1.48 (95% confidence interval (CI): 1.19, 1.83); for progesterone receptor-positive (PR+) cancer, 1.64 (95% CI: 1.31, 2.06); for ER+/PR+ cancer, 1.63 (95% CI: 1.30, 2.05); and for mixed ductal/lobular cancer, 2.51 (95% CI: 1.20, 5.24). For ER+ and PR+ cancers, trends were significant for ductal and mixed ductal/lobular types. PR status explained the positive association with ER status (for ER status, Pheterogeneity=0.70 after adjustment for PR status). Alcohol consumption was not associated with all breast cancer subtypes. Future work should emphasize large collaborative studies, precise definition of subtypes, and adjustment for correlated tumor characteristics.

Association of skin color, race/ethnicity, and hearing loss among adults in the USA.

Epidemiologic studies of hearing loss in adults have demonstrated that the odds of hearing loss are substantially lower in black than in white individuals. The basis of this association is unknown. We hypothesized that skin pigmentation as a marker of melanocytic functioning mediates this observed association and that skin pigmentation is associated with hearing loss independent of race/ethnicity. We analyzed cross-sectional data from 1,258 adults (20-59 years) in the 2003-2004 cycle of the National Health and Nutritional Examination Survey who had assessment of Fitzpatrick skin type and pure-tone audiometric testing. Audiometric thresholds in the worse hearing ear were used to calculate speech- (0.5-4 kHz) and high-frequency (3-8 kHz) pure-tone averages (PTA). Regression models were stratified by Fitzpatrick skin type or race/ethnicity to examine the association of each factor with hearing loss independent of the other. Models were adjusted for potential confounders (demographic, medical, and noise exposure covariates). Among all participants, race/ethnicity was associated with hearing thresholds (black participants with the best hearing followed by Hispanics and then white individuals), but these associations were not significant in analyses stratified by skin color. In contrast, in race-stratified analyses, darker-skinned Hispanics had better hearing than lighter-skinned Hispanics by an average of -2.5 dB hearing level (HL; 95% CI, -4.8 to -0.2) and -3.1 dB HL (95% CI, -5.3 to -0.8) for speech and high-frequency PTA, respectively. Associations between skin color and hearing loss were not significant in white and black participants. Our results demonstrate that skin pigmentation is independently associated with hearing loss in Hispanics and suggest that skin pigmentation as a marker of melanocytic functioning may mediate the strong association observed between race/ethnicity and hearing loss.