Doubly robust estimation of the weighted average treatment effect for a target population.

The weighted average treatment effect is a causal measure for the comparison of interventions in a specific target population, which may be different from the population where data are sampled from. For instance, when the goal is to introduce a new treatment to a target population, the question is what efficacy (or effectiveness) can be gained by switching patients from a standard of care (control) to this new treatment, for which the average treatment effect for the control estimand can be applied. In this paper, we propose two estimators based on augmented inverse probability weighting to estimate the weighted average treatment effect for a well-defined target population (ie, there exists a predefined target function of covariates that characterizes the population of interest, for example, a function of age to focus on elderly diabetic patients using samples from the US population). The first proposed estimator is doubly robust if the target function is known or can be correctly specified. The second proposed estimator is doubly robust if the target function has a linear dependence on the propensity score, which can be used to estimate the average treatment effect for the treated and the average treatment effect for the control. We demonstrate the properties of the proposed estimators through theoretical proof and simulation studies. We also apply our proposed methods in a comparison of glucagon-like peptide-1 receptor agonists therapy and insulin therapy among patients with type 2 diabetes, using the UK Clinical Practice Research Datalink data.

Comparison of Stereotactic Body Radiation Therapy and Radiofrequency Ablation in the Treatment of Intrahepatic Metastases.

PURPOSE: Stereotactic body radiation therapy (SBRT) and radiofrequency ablation (RFA) are widely used therapies for the treatment of intrahepatic metastases; however, direct comparisons are lacking. We sought to compare outcomes for these 2 modalities. METHODS AND MATERIALS: From 2000 to 2015, 161 patients with 282 pathologically diagnosed unresectable liver metastases were treated with RFA (n = 112) or SBRT (n = 170) at a single institution. The primary outcome was freedom from local progression (FFLP). The effect of treatment and covariates on FFLP was modeled using a mixed-effects Cox model with application of inverse probability treatment weighting to adjust for potential imbalances in treatment modality. RESULTS: The median follow-up period was 24.6 months. Patients receiving SBRT had larger tumors than those treated with RFA (median, 2.7 cm vs 1.8 cm; P < .01). On univariate analysis, tumor size was associated with worse FFLP for RFA (hazard ratio [HR]; 1.57; 95% confidence interval [CI], 1.15-2.14; P < .01) but not for SBRT (HR, 1.38; 95% CI, 0.76-2.51; P = .3). The 2-year FFLP rate was 88.2% compared with 73.9%, favoring SBRT (P = .06). For tumors ≥2 cm in diameter, SBRT was associated with improved FFLP (HR, 0.28; 95% CI, 0.09-0.93; P < .01). On multivariate analysis, treatment with SBRT (HR, 0.21; 95% CI, 0.07-0.62; P = .005) and smaller tumor size (HR, 0.65; 95% CI, 0.47-0.91; P = .01) were associated with improved FFLP. The 2-year overall survival rate was 51.1%, with no difference between groups (P = .8). Grade ≥3 treatment-related toxicity was rare, with no difference between SBRT (n = 4) and RFA (n = 3). CONCLUSIONS: Treatment with SBRT or RFA is well tolerated and provides excellent and similar local control for intrahepatic metastases <2 cm in size. For tumors ≥2 cm in size, treatment with SBRT is associated with improved FFLP and may be the preferable treatment.

Stereotactic Body Radiation Therapy as an Alternative to Transarterial Chemoembolization for Hepatocellular Carcinoma.

PURPOSE: To conduct a large single-institution comparison of transarterial chemoembolization (TACE) and stereotactic body radiation therapy (SBRT) outcomes in similar groups of patients with hepatocellular carcinoma (HCC). METHODS AND MATERIALS: From 2006 to 2014, 209 patients with 1 to 2 tumors underwent TACE (n=84) to 114 tumors or image guided SBRT (n=125) to 173 tumors. Propensity score analysis with inverse probability of treatment weighting was used to compare outcomes between treatments while adjusting for imbalances in treatment assignment. Local control (LC), toxicity, and overall survival (OS) were retrospectively analyzed. RESULTS: The TACE and SBRT groups were similar with respect to the number of tumors treated per patient, underlying liver disease, and baseline liver function. Patients treated with SBRT were older (65 vs 61 years, P=.01), had smaller tumors (2.3 vs 2.9 cm, P<.001), and less frequently underwent liver transplantation (8% vs 18%, P=.01). The 1- and 2-year LC favored SBRT: 97% and 91%, respectively, for SBRT and 47% and 23% for TACE (hazard ratio 66.5, P<.001). For patients treated with TACE, higher alpha-fetoprotein (hazard ratio 1.11 per doubling, P=.008) and segmental portal vein thrombosis (hazard ratio 9.9, P<.001) were associated with worse LC. Predictors associated with LC after SBRT were not identified. Grade 3+ toxicity occurred after 13% and 8% of TACE and SBRT treatments, respectively (P=.05). There was no difference in OS between patients treated with TACE or SBRT. CONCLUSIONS: Stereotactic body radiation therapy is a safe alternative to TACE for 1 to 2 tumors and provides better LC, with no observed difference in OS. Prospective comparative trials of TACE and SBRT are warranted.

TREE-BASED REINFORCEMENT LEARNING FOR ESTIMATING OPTIMAL DYNAMIC TREATMENT REGIMES.

Dynamic treatment regimes (DTRs) are sequences of treatment decision rules, in which treatment may be adapted over time in response to the changing course of an individual. Motivated by the substance use disorder (SUD) study, we propose a tree-based reinforcement learning (T-RL) method to directly estimate optimal DTRs in a multi-stage multi-treatment setting. At each stage, T-RL builds an unsupervised decision tree that directly handles the problem of optimization with multiple treatment comparisons, through a purity measure constructed with augmented inverse probability weighted estimators. For the multiple stages, the algorithm is implemented recursively using backward induction. By combining semiparametric regression with flexible tree-based learning, T-RL is robust, efficient and easy to interpret for the identification of optimal DTRs, as shown in the simulation studies. With the proposed method, we identify dynamic SUD treatment regimes for adolescents.

Adaptive contrast weighted learning for multi-stage multi-treatment decision-making.

Dynamic treatment regimes (DTRs) are sequential decision rules that focus simultaneously on treatment individualization and adaptation over time. To directly identify the optimal DTR in a multi-stage multi-treatment setting, we propose a dynamic statistical learning method, adaptive contrast weighted learning. We develop semiparametric regression-based contrasts with the adaptation of treatment effect ordering for each patient at each stage, and the adaptive contrasts simplify the problem of optimization with multiple treatment comparisons to a weighted classification problem that can be solved by existing machine learning techniques. The algorithm is implemented recursively using backward induction. By combining doubly robust semiparametric regression estimators with machine learning algorithms, the proposed method is robust and efficient for the identification of the optimal DTR, as shown in the simulation studies. We illustrate our method using observational data on esophageal cancer.

Diffusion tensor imaging predicts cognitive function change following partial brain radiotherapy for low-grade and benign tumors.

PURPOSE/OBJECTIVES: Radiation injury to parahippocampal cingulum white matter is associated with cognitive decline. Diffusion tensor imaging (DTI) detects micropathologic changes in white matter. Increased radial diffusion (RD) and decreased axial diffusion (AD) correspond to demyelination and axonal degeneration/gliosis respectively. We aimed to develop a predictive model for radiation-induced cognitive changes based upon DTI changes. MATERIALS/METHODS: Twenty-seven adults with benign or low-grade tumors received partial brain radiation therapy (RT) to a median dose of 54Gy. Patients underwent DTI before RT, during RT, and at the end of RT. Cognitive testing was performed before RT, and 6 and 18months after RT. Parahippocampal cingulum white matter was contoured to obtain mean values of AD and RD. RESULTS: By univariate analysis, decreasing AD and increasing RD during RT predicted declines in verbal memory and verbal fluency. By multivariate analysis, baseline neurocognitive score was the only clinical variable predicting verbal memory change; no clinical variables predicted verbal fluency change. In a multivariate model, increased RD at the end of RT significantly predicted decline in verbal fluency 18months after RT. CONCLUSIONS: Imaging biomarkers of white matter injury contributed to predictive models of cognitive function change after RT.

Predictors of Dysgeusia in Patients With Oropharyngeal Cancer Treated With Chemotherapy and Intensity Modulated Radiation Therapy.

OBJECTIVE(S): Dysgeusia is a significant factor reducing quality of life and worsening dysphagia in patients receiving chemoradiation therapy for head and neck cancer. The factors affecting dysgeusia severity are uncertain. We investigated the effects on patient-reported dysgeusia of doses to the oral cavity, salivary output (required to dissolve food particles), and patient-reported xerostomia. METHODS AND MATERIALS: Seventy-three patients with stage III to IV oropharyngeal cancer (OPC) (N=73) receiving definitive intensity modulated radiation therapy concurrently with chemotherapy participated in a prospective, longitudinal study of quality of life (QOL), including assessment of patient-reported gustatory function by taste-related questions from the Head and Neck QOL instrument (HNQOL) and the University of Washington Head and Neck-related QOL instrument (UWQOL), before therapy and periodically after treatment. At these intervals, patients also completed a validated xerostomia-specific questionnaire (XQ) and underwent unstimulated and stimulated major salivary gland flow rate measurements. RESULTS: At 1, 3, 6, and 12 months after treatment, dysgeusia improved over time: severe dysgeusia was reported by 50%, 40%, 22%, and 23% of patients, respectively. Significant associations were found between patient-reported severe dysgeusia and radiation dose to the oral cavity (P=.005) and tongue (P=.019); normal tissue complication probability for severe dysgeusia at 3 months showed mean oral cavity D50 doses 53 Gy and 57 Gy in the HNQOL and WUQOL questionnaires, respectively, with curve slope (m) of 0.41. Measured salivary output was not statistically significantly correlated with severe taste dysfunction, whereas patient-reported XQ summary scores and xerostomia while eating scores were correlated with severe dysgeusia in the UWQOL tool (P=.04). CONCLUSIONS: Taste impairment is significantly correlated with mean radiation dose to the oral cavity. Patient-reported xerostomia, but not salivary output, was correlated with severe dysgeusia in 1 of the 2 QOL questionnaires. Reduction in oral cavity doses is likely to improve dysgeusia.

Gemcitabine Plus Radiation Therapy for High-Grade Glioma: Long-Term Results of a Phase 1 Dose-Escalation Study.

PURPOSE: To evaluate the tolerability and efficacy of gemcitabine plus radiation therapy (RT) in this phase 1 study of patients with newly diagnosed malignant glioma (HGG). PATIENTS AND METHODS: Between 2004 and 2012, 29 adults with HGG were enrolled. After any extent of resection, RT (60 Gy over 6 weeks) was given concurrent with escalating doses of weekly gemcitabine. Using a time-to-event continual reassessment method, 5 dose levels were evaluated starting at 500 mg/m(2) during the last 2 weeks of RT and advanced stepwise into earlier weeks. The primary objective was to determine the recommended phase 2 dose of gemcitabine plus RT. Secondary objectives included progression-free survival, overall survival (OS), and long-term toxicity. RESULTS: Median follow-up was 38.1 months (range, 8.9-117.5 months); 24 patients were evaluable for toxicity. After 2005 when standard practice changed, patients with World Health Organization grade 4 tumors were no longer enrolled. Median progression-free survival for 22 patients with grade 3 tumors was 26.0 months (95% confidence interval [CI] 15.6-inestimable), and OS was 48.5 months (95% CI 26.8-inestimable). In 4 IDH mutated, 1p/19q codeleted patients, no failures occurred, with all but 1 alive at time of last follow-up. Seven with IDH mutated, non-codeleted tumors with ATRX loss had intermediate OS of 73.5 months (95% CI 32.8-inestimable). Six nonmutated, non-codeleted patients had a median OS of 26.5 months (95% CI 25.4-inestimable). The recommended phase 2 dose of gemcitabine plus RT was 750 mg/m(2)/wk given the last 4 weeks of RT. Dose reductions were most commonly due to grade 3 neutropenia; no grade 4 or 5 toxicities were seen. CONCLUSIONS: Gemcitabine concurrent with RT is well-tolerated and yields promising outcomes, including in patients with adverse molecular features. It is a candidate for further study, particularly for poor-prognosis patient subgroups with HGG.

Comparisons of dysphagia and quality of life (QOL) in comparable patients with HPV-positive oropharyngeal cancer receiving chemo-irradiation or cetuximab-irradiation.

PURPOSE: Compare functional outcomes of radiotherapy (RT) concurrent with cetuximab (cet-RT) or with chemotherapy (chemo-RT) for comparable, good prognosis patients with human papillomavirus related (HPV+) oropharyngeal cancer (OPC). METHODS: Outcomes of patients with stage III/IV HPV+ OPC patients with minimal smoking history and non-T4/N3/N2C, treated on prospective protocol of RT concurrent with cetuximab (cet-RT), were compared to similar patients on prospective chemo-RT protocols. In both groups, videofluoroscopy (VF), observer rated dysphagia (ORD), and validated QOL questionnaires: xerostomia questionnaire (XQ), head and neck QOL, and University of Washington QOL, were performed periodically and compared to pretreatment. Mixed effects models with adjustment for baseline assessed differences between groups. RESULTS: 26 cet-RT patients were compared to 27 chemo-RT patients with similar baseline characteristics. In the chemo-RT group, no recurrences occurred. In the cet-RT group, 1 patient had persistent microscopic disease on salvage neck dissection and 1 distant failure. Both groups had mild VF-based swallowing dysfunction pre-treatment, worsened at 3 months (P<0.02) and persisted at 12 months, not differing between groups (P>0.11). For both groups ORD was very low pretreatment, worsened at 3 months and improved at 12 months, without differences between treatment groups (P=0.26). QOL Summary and domain scores for eating were good pretreatment, worse at 3 mo, and then improved to near baseline at 12 months, without differences between the groups in any QOL domains (P>0.10). CONCLUSION: Both groups had excellent clinical outcomes without significant differences in objective or subjective functions. These data question using cetuximab instead of chemotherapy for treatment de-intensification for HPV+ OPC.

Impact of xerostomia on dysphagia after chemotherapy-intensity-modulated radiotherapy for oropharyngeal cancer: Prospective longitudinal study.

BACKGROUND: The purpose of this study was to assess how xerostomia affects dysphagia. METHODS: Prospective longitudinal studies of 93 patients with oropharyngeal cancer treated with definitive chemotherapy-intensity-modulated radiotherapy (IMRT). Observer-rated dysphagia (ORD), patient-reported dysphagia (PRD), and patient-reported xerostomia (PRX) assessment of the swallowing mechanics by videofluoroscopy (videofluoroscopy score), and salivary flow rates, were prospectively assessed from pretherapy through 2 years. RESULTS: ORD grades ≥2 were rare and therefore not modeled. Of patients with no/mild videofluoroscopy abnormalities, a substantial proportion had PRD that peaked 3 months posttherapy and subsequently improved. Through 2 years, highly significant correlations were observed between PRX and PRD scores for all patients, including those with no/mild videofluoroscopy abnormalities. Both PRX and videofluoroscopy scores were highly significantly associated with PRD. On multivariate analysis, PRX score was a stronger predictor of PRD than the videofluoroscopy score. CONCLUSION: Xerostomia contributes significantly to PRD. Efforts to further decrease xerostomia, in addition to sparing parotid glands, may translate into improvements in PRD. © 2015 Wiley Periodicals, Inc. Head Neck 38: E1605-E1612, 2016.

Outcomes After Stereotactic Body Radiotherapy or Radiofrequency Ablation for Hepatocellular Carcinoma.

PURPOSE: Data guiding selection of nonsurgical treatment of hepatocellular carcinoma (HCC) are lacking. We therefore compared outcomes between stereotactic body radiotherapy (SBRT) and radiofrequency ablation (RFA) for HCC. PATIENTS AND METHODS: From 2004 to 2012, 224 patients with inoperable, nonmetastatic HCC underwent RFA (n = 161) to 249 tumors or image-guided SBRT (n = 63) to 83 tumors. We applied inverse probability of treatment weighting to adjust for imbalances in treatment assignment. Freedom from local progression (FFLP) and toxicity were retrospectively analyzed. RESULTS: RFA and SBRT groups were similar with respect to number of lesions treated per patient, type of underlying liver disease, and tumor size (median, 1.8 v 2.2 cm in maximum diameter; P = .14). However, the SBRT group had lower pretreatment Child-Pugh scores (P = .003), higher pretreatment alpha-fetoprotein levels (P = .04), and a greater number of prior liver-directed treatments (P < .001). One- and 2-year FFLP for tumors treated with RFA were 83.6% and 80.2% v 97.4% and 83.8% for SBRT. Increasing tumor size predicted for FFLP in patients treated with RFA (hazard ratio [HR], 1.54 per cm; P = .006), but not with SBRT (HR, 1.21 per cm; P = .617). For tumors ≥ 2 cm, there was decreased FFLP for RFA compared with SBRT (HR, 3.35; P = .025). Acute grade 3+ complications occurred after 11% and 5% of RFA and SBRT treatments, respectively (P = .31). Overall survival 1 and 2 years after treatment was 70% and 53% after RFA and 74% and 46% after SBRT. CONCLUSION: Both RFA and SBRT are effective local treatment options for inoperable HCC. Although these data are retrospective, SBRT appears to be a reasonable first-line treatment of inoperable, larger HCC.

Maintaining physical activity during head and neck cancer treatment: Results of a pilot controlled trial.

BACKGROUND: Concurrent chemoradiotherapy (concurrent CRT) to treat head and neck cancer is associated with significant reductions of weight, mobility, and quality of life (QOL). An intervention focusing on functional exercise may attenuate these losses. METHODS: We allocated patients to a 14-week functional resistance and walking program designed to maintain physical activity during cancer treatment (MPACT group; n = 11), or to usual care (control group; n = 9). Outcomes were assessed at baseline, and 7 and 14 weeks. RESULTS: Compared to controls, the MPACT participants had attenuated decline or improvement in several strength, mobility, physical activity, diet, and QOL endpoints. These trends were statistically significant (p < .05) in knee strength, mental health, head and neck QOL, and barriers to exercise. CONCLUSION: In this pilot study of patients with head and neck cancer undergoing concurrent CRT, MPACT training was feasible and maintained or improved function and QOL, thereby providing the basis for larger future interventions with longer follow-up. © 2015 Wiley Periodicals, Inc. Head Neck 38: E1086-E1096, 2016.

Impact of retropharyngeal adenopathy on distant control and survival in HPV-related oropharyngeal cancer treated with chemoradiotherapy.

PURPOSE: Retropharyngeal adenopathy (RPA) is poor prognostic factor in head and neck (HN) cancer. However, the prognostic significance of RPA in Human Papillomavirus-related (HPV+) oropharyngeal cancer (OPC) is unknown. PATIENTS AND METHODS: 185 patients with HPV+OPC were assessed. Pre-therapy images reviewed by a HN radiologist to determine presence of RPA. Doses to the RPAs were determined from treatment plans. Outcomes analyzed using Kaplan-Meier method, log-rank tests, and correlations determined using Spearman's rank analyses. RESULTS: 29 (16%) of the HPV+patients had RPA. At median follow-up 49months, 5-year overall survival (OS), failure-free survival (FFS) and distant failure-free survival (DFFS) were 57% vs. 81% (P=0.02), 63% vs 80% (P=0.015) and 70% vs 91% (P=0.002) for patients with/without RPA, respectively. No differences observed in local/ regional control rates, exceeding 90% in both groups, and No RPA recurrences were observed. In multivariable analysis, stages T4 or N3, and RPA, were independently, statistically significantly associated with both OS and distant failure, while N2c, age, disease site, and smoking status, were not. CONCLUSION: RPA in HPV+OPC is an independent prognostic factor for distant failure, translating into worse OS. Patients with RPA may not be suitable candidates for trials of systemic treatment de-escalation.

Latent variable models for gene-environment interactions in longitudinal studies with multiple correlated exposures.

Many existing cohort studies designed to investigate health effects of environmental exposures also collect data on genetic markers. The Early Life Exposures in Mexico to Environmental Toxicants project, for instance, has been genotyping single nucleotide polymorphisms on candidate genes involved in mental and nutrient metabolism and also in potentially shared metabolic pathways with the environmental exposures. Given the longitudinal nature of these cohort studies, rich exposure and outcome data are available to address novel questions regarding gene-environment interaction (G × E). Latent variable (LV) models have been effectively used for dimension reduction, helping with multiple testing and multicollinearity issues in the presence of correlated multivariate exposures and outcomes. In this paper, we first propose a modeling strategy, based on LV models, to examine the association between repeated outcome measures (e.g., child weight) and a set of correlated exposure biomarkers (e.g., prenatal lead exposure). We then construct novel tests for G × E effects within the LV framework to examine effect modification of outcome-exposure association by genetic factors (e.g., the hemochromatosis gene). We consider two scenarios: one allowing dependence of the LV models on genes and the other assuming independence between the LV models and genes. We combine the two sets of estimates by shrinkage estimation to trade off bias and efficiency in a data-adaptive way. Using simulations, we evaluate the properties of the shrinkage estimates, and in particular, we demonstrate the need for this data-adaptive shrinkage given repeated outcome measures, exposure measures possibly repeated and time-varying gene-environment association.

Predicting Birth-Related Levator Ani Tear Severity in Primiparous Women: Evaluating Maternal Recovery from Labor and Delivery (EMRLD Study).

OBJECTIVE: To determine which maternal characteristics or birth events independently predict severity of levator ani muscle (LA) tears at first vaginal birth in a longitudinal/observational investigation in a tertiary care hospital. SAMPLE: Ninety primiparas with at least one at risk for LA tear inclusion factor at vaginal birth: maternal age ≥ 33 years, second stage ≥ 150 minutes, macrosomia, instrumented delivery, and/or anal sphincter laceration were studied. METHODS: Magnetic Resonance Imaging (MRI) was obtained early postpartum (mean ± sd 48.9 ± 21.6 days) to identify LA tear. Severity of LA muscle fiber loss was graded on an ordinal scale of: "0" as no loss, "1" as <50% unilateral loss, "2" as ≥50% unilateral or <50% bilateral loss, and "3" as ≥50% bilateral loss. Data were analyzed using proportional odds modeling. Inclusion factors were explored as predictors of LA tear severity and at analysis episiotomy, time spent actively pushing, epidural, and oxytocin were also considered. The main outcome measures of interest included grading of severity of LA muscle fiber loss on an ordinal scale. RESULTS: Respective counts/percentages of women within each 0 thru 3 ordered category of LA tear severity were: "0" = 58(64%), "1" = 9(10%), "2" = 15(17%), and "3" = 8(9%). Estimates and 95% CI for significant demographic or obstetric univariate predictors of LA tear severity level were age, OR 1.093 (CI 1.012 - 1.180), p = 0.023; and time spent in active pushing, OR 1.089 (CI 1.005 - 1.180), p = 0.038. The other factors considered were not significant. There were too few women with forceps deliveries to analyze. CONCLUSION: In our enriched sample of primiparous women, 26% showed a significant LA tear. Maternal age and time spent actively pushing independently predict LA tear severity.

Environmental risk score as a new tool to examine multi-pollutants in epidemiologic research: an example from the NHANES study using serum lipid levels.

OBJECTIVE: A growing body of evidence suggests that environmental pollutants, such as heavy metals, persistent organic pollutants and plasticizers play an important role in the development of chronic diseases. Most epidemiologic studies have examined environmental pollutants individually, but in real life, we are exposed to multi-pollutants and pollution mixtures, not single pollutants. Although multi-pollutant approaches have been recognized recently, challenges exist such as how to estimate the risk of adverse health responses from multi-pollutants. We propose an "Environmental Risk Score (ERS)" as a new simple tool to examine the risk of exposure to multi-pollutants in epidemiologic research. METHODS AND RESULTS: We examined 134 environmental pollutants in relation to serum lipids (total cholesterol, high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL) and triglycerides) using data from the National Health and Nutrition Examination Survey between 1999 and 2006. Using a two-stage approach, stage-1 for discovery (n = 10818) and stage-2 for validation (n = 4615), we identified 13 associated pollutants for total cholesterol, 9 for HDL, 5 for LDL and 27 for triglycerides with adjustment for sociodemographic factors, body mass index and serum nutrient levels. Using the regression coefficients (weights) from joint analyses of the combined data and exposure concentrations, ERS were computed as a weighted sum of the pollutant levels. We computed ERS for multiple lipid outcomes examined individually (single-phenotype approach) or together (multi-phenotype approach). Although the contributions of ERS to overall risk predictions for lipid outcomes were modest, we found relatively stronger associations between ERS and lipid outcomes than with individual pollutants. The magnitudes of the observed associations for ERS were comparable to or stronger than those for socio-demographic factors or BMI. CONCLUSIONS: This study suggests ERS is a promising tool for characterizing disease risk from multi-pollutant exposures. This new approach supports the need for moving from a single-pollutant to a multi-pollutant framework.

Statistical strategies for constructing health risk models with multiple pollutants and their interactions: possible choices and comparisons.

BACKGROUND: As public awareness of consequences of environmental exposures has grown, estimating the adverse health effects due to simultaneous exposure to multiple pollutants is an important topic to explore. The challenges of evaluating the health impacts of environmental factors in a multipollutant model include, but are not limited to: identification of the most critical components of the pollutant mixture, examination of potential interaction effects, and attribution of health effects to individual pollutants in the presence of multicollinearity. METHODS: In this paper, we reviewed five methods available in the statistical literature that are potentially helpful for constructing multipollutant models. We conducted a simulation study and presented two data examples to assess the performance of these methods on feature selection, effect estimation and interaction identification using both cross-sectional and time-series designs. We also proposed and evaluated a two-step strategy employing an initial screening by a tree-based method followed by further dimension reduction/variable selection by the aforementioned five approaches at the second step. RESULTS: Among the five methods, least absolute shrinkage and selection operator regression performs well in general for identifying important exposures, but will yield biased estimates and slightly larger model dimension given many correlated candidate exposures and modest sample size. Bayesian model averaging, and supervised principal component analysis are also useful in variable selection when there is a moderately strong exposure-response association. Substantial improvements on reducing model dimension and identifying important variables have been observed for all the five statistical methods using the two-step modeling strategy when the number of candidate variables is large. CONCLUSIONS: There is no uniform dominance of one method across all simulation scenarios and all criteria. The performances differ according to the nature of the response variable, the sample size, the number of pollutants involved, and the strength of exposure-response association/interaction. However, the two-step modeling strategy proposed here is potentially applicable under a multipollutant framework with many covariates by taking advantage of both the screening feature of an initial tree-based method and dimension reduction/variable selection property of the subsequent method. The choice of the method should also depend on the goal of the study: risk prediction, effect estimation or screening for important predictors and their interactions.

[Time-series analysis on the acute mortality affected by air pollution, in the city of Guangzhou, 2004-2008].

OBJECTIVE: To study the associations between daily mortality and the status of exposure to air pollution. METHODS: A time-series analysis was conducted to assess the relations between acute mortality and exposure to respiratory particulate matter (PM(10)), sulfur-dioxide (SO2) and nitrogen dioxide (NO2) in urban residents of Guangzhou (2004 - 2008), using Poisson regression. RESULTS: Through controlling the factors as temperature, relative humidity, age, gender and time, significant increases were observed in all-cause mortality of 0.94% (0.79 - 1.09) for PM(10), 1.55% (1.31 - 1.78) for NO2, and 1.09% (0.91 - 1.27) for SO2, per 10 µg/m(3), when increase of the lagging 2-day average concentrations of air pollution was seen, in Guangzhou. Stronger effects of exposure to air pollution were found on cardiovascular and respiratory mortality, as well as in elderly (≥ 65 years) and female population. CONCLUSION: Our results suggested that exposure to ambient pollution was significantly associated with the increase of excess risks, on total and cardio-respiratory mortality in the residents of Guangzhou.

Seasonal variation of chemical species associated with short-term mortality effects of PM(2.5) in Xi'an, a Central City in China.

The authors conducted a time-series analysis to examine seasonal variation of mortality risk in association with particulate matter less than 2.5 µm in aerodynamic diameter (PM(2.5)) and chemical species in Xi'an, China, using daily air pollution and all-cause and cause-specific mortality data (2004-2008). Poisson regression incorporating natural splines was used to estimate mortality risks of PM(2.5) and its chemical components, adjusting for day of the week, time trend, and meteorologic effects. Increases of 2.29% (95% confidence interval: 0.83, 3.76) for all-cause mortality and 3.08% (95% confidence interval: 0.94, 5.26) for cardiovascular mortality were associated with an interquartile range increase of 103.0 µg/m(3) in lagged 1-2 day PM(2.5) exposure. Stronger effects were observed for the elderly (≥65 years), males, and cardiovascular diseases groups. Secondary components (sulfate and ammonium), combustion species (elemental carbon, sulfur, chlorine), and transition metals (chromium, lead, nickel, and zinc) appeared most responsible for increased risk, particularly in the cold months. The authors concluded that differential association patterns observed across species and seasons indicated that PM(2.5)-related effects might not be sufficiently explained by PM(2.5) mass alone. Future research is needed to examine spatial and temporal varying factors that might play important roles in modifying the PM(2.5)-mortality association.