A unified framework for assessing interaction effects among environmental exposures in epidemiologic studies: A case study on temperature, air pollution, and kidney-related conditions in New York state.

BACKGROUND: There are various methods to assess interaction effects. However, current methods have limitations, and quantification of interaction effects is rarely performed. This study aimed to develop a unified quantitative framework for assessing interaction effects. METHODS: We proposed a novel framework using log-linear models with a product term(s) across the exposures that generates parametric bi-variate association and interaction effect surfaces and allows flexible functional forms for exposures in the interaction term(s). In a case study, we assessed the interaction effects between temperature and air pollution (i.e., PM2.5, NO2, and O3) on risk for kidney-related conditions in New York State (2007-2016) using a case-crossover design with conditional logistic models. Our measures of exposure were the moving averages at lag 0-5 days for air pollution (linear) and daytime mean outdoor wet-bulb globe temperature (WBGT; using a natural cubic spline). RESULTS: We derived closed-form expressions for the magnitude of multiplicative interaction effects (the joint relative risk divided by the product of the two conditional relative risks) and their uncertainties. In the case study, we found a Bonferroni-corrected significant multiplicative interaction effect (IE) between outdoor WBGT at the 99th percentile (median as the reference) and (1) PM2.5 (per 5 µg/m3 increase, IE = 1.052; 95 % confidence interval [CI]: 1.019, 1.087) for acute kidney failure and (2) O3 (per 5 ppb increase; IE = 1.022; 95 % CI: 1.008, 1.036) for urolithiasis (the latter being inconclusive based on the sensitivity analysis). CONCLUSIONS: Our framework allows different functional forms of exposure variables in the interaction term, quantifies the magnitudes of entire-exposure-range (in addition to discrete exposure level) multiplicative interaction effects and their uncertainties in a categorical or continuous (linear or non-linear) manner, and harmonizes the two-way evaluation of effect modification. The case study underscores co-consideration of heat and air pollution when estimating health burden and designing heat/pollution alert systems.

Associations between short-term exposure to PM2.5, NO2 and O3 pollution and kidney-related conditions and the role of temperature-adjustment specification: A case-crossover study in New York state.

Epidemiologic evidence on the relationship between air pollution and kidney disease remains inconclusive. We evaluated associations between short-term exposure to PM2.5, NO2 and O3 and unplanned hospital visits for seven kidney-related conditions (acute kidney failure [AKF], urolithiasis, glomerular diseases [GD], renal tubulo-interstitial diseases, chronic kidney disease, dysnatremia, and volume depletion; n = 1,209,934) in New York State (2007-2016). We applied a case-crossover design with conditional logistic regression, controlling for temperature, dew point temperature, wind speed, and solar radiation. We used a three-pollutant model at lag 0-5 days of exposure as our main model. We also assessed the influence of model adjustment using different specifications of temperature by comparing seven temperature metrics (e.g., dry-bulb temperature, heat index) and five intraday temperature measures (e.g., daily mean, daily minimum, nighttime mean), according to model performance and association magnitudes between air pollutants and kidney-related conditions. In our main models, we adjusted for daytime mean outdoor wet-bulb globe temperature, which showed good model performance across all kidney-related conditions. We observed the odds ratios (ORs) for 5 µg/m3 increase in daily mean PM2.5 to be 1.013 (95% confidence interval [CI]: 1.001, 1.025) for AKF, 1.107 (95% CI: 1.018, 1.203) for GD, and 1.027 (95% CI: 1.015, 1.038) for volume depletion; and the OR for 5 ppb increase in daily 1-hour maximum NO2 to be 1.014 (95% CI; 1.008, 1.021) for AKF. We observed no associations with daily 8-hour maximum O3 exposure. Association estimates varied by adjustment for different intraday temperature measures: estimates adjusted for measures with poorer model performance resulted in the greatest deviation from estimates adjusted for daytime mean, especially for AKF and volume depletion. Our findings indicate that short-term exposure to PM2.5 and NO2 is a risk factor for specific kidney-related conditions and underscore the need for careful adjustment of temperature in air pollution epidemiologic studies.

Associations between short-term temperature exposure and kidney-related conditions in New York State: The influence of temperature metrics across four dimensions.

BACKGROUND: Evidence describing the relationship between short-term temperature exposure and kidney-related conditions is insufficient. It remains unclear how temperature specification affects estimation of these associations. This study aimed to assess associations between short-term temperature exposure and seven kidney-related conditions and to evaluate the influence of temperature specification. METHODS: We obtained data on hospital encounters in New York State (2007-2016). We assessed associations with a case-crossover design using conditional logistic regression with distributed lag non-linear models. We compared model performance (i.e., AIC) and association curves using 1) five temperature spatial resolutions; 2) temperature on an absolute versus relative scale; 3) seven temperature metrics incorporating humidity, wind speed, and/or solar radiation; and 4) five intraday temperature measures (e.g., daily minimum and daytime mean). RESULTS: We included 1,209,934 unplanned adult encounters. Temperature metric and intraday measure had considerably greater influence than spatial resolution and temperature scale. For outcomes not associated with temperature exposure, almost all metrics or intraday measures showed good model performance; for outcomes associated with temperature, there were meaningful differences in performance across metrics or intraday measures. For parsimony, we modelled daytime mean outdoor wet-bulb globe temperature, which showed good performance for all outcomes. At lag 0-6 days, we observed increased risk at the 95th percentile of temperature versus the minimum morbidity temperature for acute kidney failure (odds ratio [OR] = 1.36, 95% confidence interval [CI]: 1.09, 1.69), urolithiasis (OR = 1.41, 95% CI: 1.16, 1.70), dysnatremia (OR = 1.26, 95% CI: 1.01, 1.59), and volume depletion (OR = 1.88, 95% CI: 1.41, 2.51), but not for glomerular diseases, renal tubulo-interstitial diseases, and chronic kidney disease. CONCLUSIONS: High-temperature exposure over one week is a risk factor for acute kidney failure, urolithiasis, dysnatremia, and volume depletion. The differential model performance across temperature metrics and intraday measures indicates the importance of careful selection of exposure metrics when estimating temperature-related health burden.

Relationships between short-term ambient temperature exposure and kidney disease hospitalizations in the warm season in Vietnam: A case-crossover study.

BACKGROUND: Under a warming climate, adverse health effects of heat are an increasing concern. We evaluated associations between short-term ambient temperature exposure and hospital admission for kidney disease in Vietnam. METHODS: We linked province-level meteorologic data with admission data from 14 province-level hospitals (2003-2015). We used a case-crossover design to evaluate associations between daily ambient temperature metrics (mean, maximum, and minimum temperature and mean heat index) and risk of hospitalization for four kidney disease subtypes: glomerular diseases, renal tubulo-interstitial diseases, chronic kidney disease, and urolithiasis, including lagged (≤lag 14 days) and cumulative (≤lag 0-6 days) associations, during the warm season. We also evaluated independent associations with extreme heat days (defined as days with daily maximum temperature >95th percentile of the provincial daily maximum temperature distribution). Akaike's information criterion and patterns of risk estimates across cumulative exposure time windows and single-day lags informed our selection of final models. RESULTS: We included 58,330 hospital admissions during the warm season. Daily mean temperature averaged over the same day and the previous six days (lag 0-6 days) was associated with risk of hospitalization for each kidney disease outcome with odds ratios (per 1 °C increase in daily mean temperature) of 1.07 (95% confidence interval [CI]: 0.99, 1.16) for glomerular diseases, 1.06 (95% CI: 0.96, 1.17) for renal tubulo-interstitial diseases, 1.12 (95% CI: 1.00, 1.24) for chronic kidney disease, and 1.09 (95% CI: 1.02, 1.16) for urolithiasis. We found no additional independent associations with extreme heat. Results for the four temperature metrics were similar. CONCLUSIONS: High ambient temperature was associated with increased risk of hospitalization for each kidney disease subtype, with the most convincing associations for chronic kidney disease and urolithiasis. Further laboratory and epidemiologic research is needed to confirm the findings and disentangle the underlying mechanisms.

Predicting spatiotemporally-resolved mean air temperature over Sweden from satellite data using an ensemble model.

Mapping of air temperature (Ta) at high spatiotemporal resolution is critical to reducing exposure assessment errors in epidemiological studies on the health effects of air temperature. In this study, we applied a three-stage ensemble model to estimate daily mean Ta from satellite-based land surface temperature (Ts) over Sweden during 2001-2019 at a high spatial resolution of 1 × 1 km2. The ensemble model incorporated four base models, including a generalized additive model (GAM), a generalized additive mixed model (GAMM), and two machine learning models (random forest [RF] and extreme gradient boosting [XGBoost]), and allowed the weights for each model to vary over space, with the best-performing model for each grid cell assigned the highest weight. Various spatial predictors were included as adjustment variables in all the base models, including land cover type, normalized difference vegetation index (NDVI), and elevation. The ensemble model showed high performance with an overall R2 of 0.98 and a root mean square error of 1.38 °C in the ten-fold cross-validation, and outperformed each of the four base models. Although each base model performed well, the two machine learning models (RF [R2 = 0.97], XGBoost [R2 = 0.98]) had better performance than the two regression models (GAM [R2 = 0.95], GAMM [R2 = 0.96]). In the machine learning models, Ts was the dominant predictor of Ta, followed by day of year, NDVI, latitude, elevation, and longitude. The highly spatiotemporally-resolved Ta can improve temperature exposure assessment in future epidemiological studies.

Short-term associations between particulate matter air pollution and hospital admissions through the emergency room for urinary system disease in Beijing, China: A time-series study.

Evidence on the relationship between particulate matter air pollution and urinary system disease (UD) is scarce. This study aims to evaluate the associations between short-term exposures to PM2.5 and PM10 and risk of daily UD inpatient hospital admissions through the emergency room (ER-admissions) in Beijing. We obtained 41,203 weekday UD ER-admissions for secondary and tertiary hospitals in all 16 districts in Beijing during 2013-2018 from the Beijing Municipal Health Commission Information Center and obtained district-level air pollution concentrations based on 35 fixed monitoring stations in Beijing. We conducted a two-stage time-series analysis, with district-specific generalized linear models for each of Beijing's 16 districts, followed by random effects meta-analysis to obtain pooled risk estimates. We evaluated lagged and cumulative associations up to 30 days. In single-pollutant models, for both PM2.5 and PM10, cumulative exposure averaged over the day of admission and the previous 10 days (lag 0-10 days) showed the strongest association, with per interquartile range increases of PM2.5 or PM10 concentrations associated with a 7.5 % (95 % confidence interval [CI]: 3.0 %-12.2 %) or 6.0 % (95 % CI: 1.1 %-11.2 %) increased risk of daily UD hospital admissions, respectively. The risk estimates were robust to adjustment for co-pollutants and to a variety of sensitivity analyses. However, due to the strong correlation between PM2.5 and PM10 concentrations, we were unable to disentangle the respective relationships between these two exposures and UD risk. In this study, we found that short-term exposures to PM2.5 and PM10 are risk factors for UD morbidity and that cumulative exposure to PM pollution over a period of one to two weeks (i.e., 11 days) could be more important for UD risk than transient exposure during each of the respective single days.

Evaluation of confounding in epidemiologic studies assessing alcohol consumption on the risk of ischemic heart disease.

BACKGROUND: Among different investigators studying the same exposures and outcomes, there may be a lack of consensus about potential confounders that should be considered as matching, adjustment, or stratification variables in observational studies. Concerns have been raised that confounding factors may affect the results obtained for the alcohol-ischemic heart disease relationship, as well as their consistency and reproducibility across different studies. Therefore, we assessed how confounders are defined, operationalized, and discussed across individual studies evaluating the impact of alcohol on ischemic heart disease risk. METHODS: For observational studies included in a recent alcohol-ischemic heart disease meta-analysis, we identified all variables adjusted, matched, or stratified for in the largest reported multivariate model (i.e. potential confounders). We recorded how the variables were measured and grouped them into higher-level confounder domains. Abstracts and Discussion sections were then assessed to determine whether authors considered confounding when interpreting their study findings. RESULTS: 85 of 87 (97.7%) studies reported multivariate analyses for an alcohol-ischemic heart disease relationship. The most common higher-level confounder domains included were smoking (79, 92.9%), age (74, 87.1%), and BMI, height, and/or weight (57, 67.1%). However, no two models adjusted, matched, or stratified for the same higher-level confounder domains. Most (74/87, 85.1%) articles mentioned or alluded to "confounding" in their Abstract or Discussion sections, but only one stated that their main findings were likely to be affected by residual confounding. There were five (5/87, 5.7%) authors that explicitly asked for caution when interpreting results. CONCLUSION: There is large variation in the confounders considered across observational studies evaluating the impact of alcohol on ischemic heart disease risk and almost all studies spuriously ignore or eventually dismiss confounding in their conclusions. Given that study results and interpretations may be affected by the mix of potential confounders included within multivariate models, efforts are necessary to standardize approaches for selecting and accounting for confounders in observational studies.

Consideration of confounding in epidemiologic studies assessing alcohol consumption on the risk of breast cancer: A brief report.

BACKGROUND: Individual observational studies have suggested null, weak, linear, and J-shaped associations between alcohol consumption and breast cancer risk. However, observational studies are susceptible to confounders, which can obscure the true impact of an exposure on an outcome. Given the uncertainty regarding the association between alcohol consumption and breast cancer, and the challenges of identifying, measuring, and accounting for all potential confounders, we assessed whether and how authors of observational studies evaluating the impact of alcohol consumption on the risk of breast cancer considered bias when interpreting their main study findings. METHODS: We identified all observational studies included in a recent alcohol-breast cancer meta-analysis. The Abstract and/or Discussion sections were reviewed to determine whether authors considered confounding. RESULTS: Among 101 eligible studies, 73 (72.3%) mentioned confounding explicitly in the Abstract and Discussion sections. There were 33 (32.7%) studies that included statements regarding specific confounders that were not adjusted for in the analyses and 60 (59.4%) studies without any statements about the impact of residual confounding on their main findings. Although none of the studies outlined that their main findings were "likely" to be affected by residual confounding, 25 (24.8%) mentioned a "possible" impact and 16 (15.8%) claimed an "unlikely" impact. Only one (1.0%) article explicitly stated that caution was needed when interpreting their findings due to confounding. CONCLUSION: These results highlight the need for more adequate consideration of the potential impact of residual confounding in observational studies evaluating the impact of alcohol consumption on the risk of breast cancer.

Vibration of effects in epidemiologic studies of alcohol consumption and breast cancer risk.

BACKGROUND: Different analytical approaches can influence the associations estimated in observational studies. We assessed the variability of effect estimates reported within and across observational studies evaluating the impact of alcohol on breast cancer. METHODS: We abstracted largest harmful, largest protective and smallest (closest to the null value of 1.0) relative risk estimates in studies included in a recent alcohol-breast cancer meta-analysis, and recorded how they differed based on five model specification characteristics, including exposure definition, exposure contrast levels, study populations, adjustment covariates and/or model approaches. For each study, we approximated vibration of effects by dividing the largest by the smallest effect estimate [i.e. ratio of odds ratio (ROR)]. RESULTS: Among 97 eligible studies, 85 (87.6%) reported both harmful and protective relative effect estimates for an alcohol-breast cancer relationship, which ranged from 1.1 to 17.9 and 0.0 to 1.0, respectively. The RORs comparing the largest and smallest estimates in value ranged from 1.0 to 106.2, with a median of 3.0 [interquartile range (IQR) 2.0-5.2]. One-third (35, 36.1%) of the RORs were based on extreme effect estimates with at least three different model specification characteristics; the vast majority (87, 89.7%) had different exposure definitions or contrast levels. Similar vibrations of effect were observed when only extreme estimates with differences based on study populations and/or adjustment covariates were compared. CONCLUSIONS: Most observational studies evaluating the impact of alcohol on breast cancer report relative effect estimates for the same associations that diverge by >2-fold. Therefore, observational studies should estimate the vibration of effects to provide insight regarding the stability of findings.