A framework for integrating evidence to assess hazards and risk.

To accurately characterize human health hazards, human, animal, and mechanistic data must be integrated and the relevance to the research question of all three lines of evidence must be considered. Mechanistic data are often critical to the full integration of animal and human data and to characterizing relevance and uncertainty. This novel evidence integration framework (EIF) provides a method for synthesizing data from comprehensive, systematic, quality-based assessments of the epidemiological and toxicological literature, including in vivo and in vitro mechanistic studies. It organizes data according to both the observed human health effects and the mechanism of action of the chemical, providing a method to support evidence synthesis. The disease-based component uses the evidence of human health outcomes studied in the best quality epidemiological literature to organize the toxicological data according to authors' stated purpose, with the pathophysiology of the disease determining the potential relevance of the toxicological data. The mechanism-based component organizes the data based on the proposed mechanisms of effect and data supporting events leading to each endpoint, with the epidemiological data potentially providing corroborating information. The EIF includes a method to cross-classify and describe the concordance of the data, and to characterize its uncertainty. At times, the two methods of organizing the data may lead to different conclusions. This facilitates identification of knowledge gaps and shows the impact of uncertainties on the strength of causal inference.

Estimating the Population Health Impact of Recently Introduced Modified Risk Tobacco Products: A Comparison of Different Approaches.

INTRODUCTION: Various approaches have been used to estimate the population health impact of introducing a Modified Risk Tobacco Product (MRTP). AIMS AND METHODS: We aimed to compare and contrast aspects of models considering effects on mortality that were known to experts attending a meeting on models in 2018. RESULTS: Thirteen models are described, some focussing on e-cigarettes, others more general. Most models are cohort-based, comparing results with or without MRTP introduction. They typically start with a population with known smoking habits and then use transition probabilities either to update smoking habits in the "null scenario" or joint smoking and MRTP habits in an "alternative scenario". The models vary in the tobacco groups and transition probabilities considered. Based on aspects of the tobacco history developed, the models compare mortality risks, and sometimes life-years lost and health costs, between scenarios. Estimating effects on population health depends on frequency of use of the MRTP and smoking, and the extent to which the products expose users to harmful constituents. Strengths and weaknesses of the approaches are summarized. CONCLUSIONS: Despite methodological differences, most modellers have assumed the increase in risk of mortality from MRTP use, relative to that from cigarette smoking, to be very low and have concluded that MRTP introduction is likely to have a beneficial impact. Further model development, supplemented by preliminary results from well-designed epidemiological studies, should enable more precise prediction of the anticipated effects of MRTP introduction. IMPLICATIONS: There is a need to estimate the population health impact of introducing modified risk nicotine-containing products for smokers unwilling or unable to quit. This paper reviews a variety of modeling methodologies proposed to do this, and discusses the implications of the different approaches. It should assist modelers in refining and improving their models, and help toward providing authorities with more reliable estimates.

Assessing the Likelihood and Magnitude of a Population Health Benefit Following the Market Introduction of a Modified-Risk Tobacco Product: Enhancements to the Dynamic Population Modeler, DPM(+1).

Researchers and those responsible for evaluating and implementing policies intended to reduce population harm must assess the potential for both intended and unintended consequences associated with those policies. Such assessments should be based on the combined dimensions of magnitude, and thus likelihood, of shifts in exposure patterns needed to produce a population benefit or harm, and magnitude of the expected population benefit or harm. In response to this assessment need, we provide a conceptual description of the dynamic population modeler, DPM(+1), as well as illustrative analyses that estimate the effects on all-cause mortality, life expectancy, and quality of life-adjusted life expectancy if exposure patterns in the population shift from a higher risk product (e.g., cigarettes) to a lower, or modified, risk tobacco product (MRTP) in specified ways. Estimates from these analyses indicate that, within a single birth cohort, switching completely from cigarette smoking to MRTP use is more likely to lead to a population-level survival benefit than initiating tobacco use with an MRTP instead of cigarettes. This is because tobacco initiation rarely occurs beyond young adulthood, whereas continuing smokers exist in all subsequent age categories, leading to a greater cumulative effect. In addition, complete switching to MRTP use among a small proportion of smokers in each age category offsets the survival deficit caused by unintended shifts in exposure patterns, such as MRTP initiation among never tobacco users followed by transitioning to cigarette smoking and/or cigarette smokers switching to MRTP use instead of quitting.

Using residential proximity to wind turbines as an alternative exposure measure to investigate the association between wind turbines and human health.

This analysis uses data from the Community Noise and Health Study developed by Statistics Canada to investigate the association between residential proximity to wind turbines and health-related outcomes in a dataset that also provides objective measures of wind turbine noise. The findings indicate that residential proximity to wind turbines is correlated with annoyance and health-related quality of life measures. These associations differ in some respects from associations with noise measurements. Results can be used to support discussions between communities and wind-turbine developers regarding potential health effects of wind turbines.

A tissue dose-based comparative exposure assessment of manganese using physiologically based pharmacokinetic modeling-The importance of homeostatic control for an essential metal.

A physiologically-based pharmacokinetic (PBPK) model (Schroeter et al., 2011) was applied to simulate target tissue manganese (Mn) concentrations following occupational and environmental exposures. These estimates of target tissue Mn concentrations were compared to determine margins of safety (MOS) and to evaluate the biological relevance of applying safety factors to derive acceptable Mn air concentrations. Mn blood concentrations measured in occupational studies permitted verification of the human PBPK models, increasing confidence in the resulting estimates. Mn exposure was determined based on measured ambient air Mn concentrations and dietary data in Canada and the United States (US). Incorporating dietary and inhalation exposures into the models indicated that increases in target tissue concentrations above endogenous levels only begin to occur when humans are exposed to levels of Mn in ambient air (i.e. >10µg/m3) that are far higher than those currently measured in Canada or the US. A MOS greater than three orders of magnitude was observed, indicating that current Mn air concentrations are far below concentrations that would be required to produce the target tissue Mn concentrations associated with subclinical neurological effects. This application of PBPK modeling for an essential element clearly demonstrates that the conventional application of default factors to "convert" an occupational exposure to an equivalent continuous environmental exposure, followed by the application of safety factors, is not appropriate in the case of Mn. PBPK modeling demonstrates that the relationship between ambient Mn exposures and dose-to-target tissue is not linear due to normal tissue background levels and homeostatic controls.

Stroke risk among menthol versus non-menthol cigarette smokers in the United States: Analysis of the National Health and Nutrition Examination Survey (NHANES).

Though available evidence is relatively consistent in showing no additional health effects among smokers due to menthol in cigarettes, two studies reported conflicting results for stroke risk using different subsets of NHANES data. We investigated reasons for the differences in these reports by analyzing NHANES cycles conducted between 1999 and 2012, combined and in subsets. Adjusted odds ratios (AOR) and 95% confidence intervals (CI) from three different survey logistic regression models compare risk of reported stroke diagnoses among menthol and non-menthol cigarette smokers. Depending on timeframe, about 1150 to 8000 U.S. adults (aged ≥ 20 years) who smoked on ≥ 1 of the last 30 days had complete data for cigarette type and all covariates included in each model. Results were not much affected by which covariates were included in the models, but depended strongly on the NHANES cycles included in the analysis. Using NHANES 1999-2012 data combined, AORs and 95% CIs for stroke comparing menthol with non-menthol cigarette smokers were 0.95 (95% CI: 0.65, 1.37), 0.85 (95% CI: 0.59, 1.23) or 0.86 (95% CI: 0.59, 1.25). Collectively, findings illustrate the need for fully reporting research and analytical methods, especially when analyses are meant to develop evidence intended for regulatory decision-making.

Measures of initiation and progression to increased smoking among current menthol compared to non-menthol cigarette smokers based on data from four U.S. government surveys.

There are no large-scale, carefully designed cohort studies that provide evidence on whether menthol cigarette use is associated with a differential risk of initiating and/or progressing to increased smoking. However, questions of whether current menthol cigarette smokers initiated smoking at a younger age or are more likely to have transitioned from non-daily to daily cigarette use compared to non-menthol smokers can be addressed using cross-sectional data from U.S. government surveys. Analyses of nationally representative samples of adult and youth smokers indicate that current menthol cigarette use is not associated with an earlier age of having initiated smoking or greater likelihood of being a daily versus non-daily smoker. Some surveys likewise provide information on cigarette type preference (menthol versus non-menthol) among youth at different stages or trajectories of smoking, based on number of days smoked during the past month and/or cigarettes smoked per day. Prevalence of menthol cigarette use does not appear to differ among new, less experienced youth smokers compared to established youth smokers. While there are limitations with regard to inferences that can be drawn from cross-sectional analyses, these data do not suggest any adverse effects for menthol cigarettes on measures of initiation and progression to increased smoking.

Evaluating the association between menthol cigarette use and the likelihood of being a former versus current smoker.

Menthol in cigarettes has been examined for its potential to affect smoking dependence, measured primarily as number of cigarettes smoked per day and time to first cigarette after waking; the ability to quit smoking constitutes an additional measure of dependence. Successful quitting among menthol compared to non-menthol cigarette smokers is difficult to determine from the literature, due in part to the various definitions of quitting used by researchers. Nevertheless, intervention and follow-up studies of smoking cessation treatments generally indicate no differences in quitting success among menthol compared to non-menthol smokers, while cross-sectional studies suggest some differences within race/ethnicity groups. The association between menthol cigarette use and likelihood of being a former versus current smoker was examined based on data from the National Health Interview Survey and Tobacco Use Supplement to the Current Population Survey. Analyses stratified by race/ethnicity and limited to smokers who had quit at least one year prior to survey participation provided inconsistent results with regard to menthol cigarette use and quitting, both within surveys (i.e., comparing race/ethnicity groups) and between surveys (i.e., same race/ethnicity group across surveys). Evidence suggesting the existence or direction of an association between menthol in cigarettes and quitting depended on the data source.

Primary measures of dependence among menthol compared to non-menthol cigarette smokers in the United States.

Previously published studies provide somewhat inconsistent evidence on whether menthol in cigarettes is associated with increased dependence. The National Health and Nutrition Examination Survey, National Survey on Drug Use and Health, National Health Interview Survey, and Tobacco Use Supplement to the Current Population Survey collect data on current cigarette type preference and primary measures of dependence, and thus allow examination of whether menthol smokers are more dependent than non-menthol smokers. Analyses based on combined data from multiple administrations of each of these four nationally representative surveys, using three definitions for current smokers (i.e., smoked ⩾1day, ⩾10days and daily during the past month), consistently demonstrate that menthol smokers do not report smoking more cigarettes per day than non-menthol smokers. Moreover, two of the three surveys that provide data on time to first cigarette after waking indicate no difference in urgency to smoke among menthol compared to non-menthol smokers, while the third suggests menthol smokers may experience a greater urgency to smoke; estimates from all three surveys indicate that menthol versus non-menthol smokers do not report a higher Heaviness of Smoking Index. Collectively, these findings indicate no difference in dependence among U.S. smokers who use menthol compared to non-menthol cigarettes.

Patterns of menthol cigarette use among current smokers, overall and within demographic strata, based on data from four U.S. government surveys.

The National Health and Nutrition Examination Survey, National Survey on Drug Use and Health, National Health Interview Survey and Tobacco Use Supplement to the Current Population Survey provide estimates of the proportions of U.S. smokers who currently use menthol cigarettes, overall and within demographic strata. Among adult past-month, regular and daily smokers, menthol cigarette use ranges from 26% to 30%, with statistically higher proportions of female versus male smokers (8-11 percentage points higher) currently using menthol cigarettes. Compared to adult smokers overall, statistically higher proportions of non-Hispanic Black smokers (72-79%) and statistically lower proportions of non-Hispanic White smokers (19-22%) currently use menthol cigarettes, with no differences among smokers of other race/ethnicity groups (18-20% to 28-30%, depending on the survey). Higher proportions of younger adult past-month, regular and daily smokers (aged 18-25years) currently use menthol cigarettes compared to older adult smokers (aged 26-29years and/or ⩾30years); however, differences are small in magnitude, with the vast majority of adult smokers (70-75%) who currently use menthol cigarettes being aged ⩾30years. Comparisons between youth and adult smokers are provided, although data for youth smokers are less available and provide less consistent patterns of menthol cigarette use.

A dynamic population model for estimating all-cause mortality due to lifetime exposure history.

We developed a comprehensive, flexible dynamic model that estimates all-cause mortality for a hypothetical cohort. All model input is user-specified. In the base case, members of the cohort may be exposed to a high risk product as they age. The counterfactual scenario includes exposure to both a high risk and a lower risk product. The model sorts the population into age and exposure categories, and applies the appropriate mortality rates to each category. The model tracks individual exposure histories, and estimates, at the end of each modeled age category, the number of survivors in the two exposure scenarios (base case and counterfactual), and the difference between them. Markov Chain Monte Carlo techniques are used to estimate the variability of the results. Model output was compared against US and Swedish life tables using population-specific tobacco exposure transition probabilities derived from the literature, and it produced similar survival estimates.

Effectiveness of an external ankle brace in reducing parachuting-related ankle injuries.

Outside-the-boot parachute ankle braces (PABs) worn during US Army paratrooper training have been shown to reduce the risk of severe ankle injuries. In spite of evidence to the contrary, anecdotal reports continue to suggest increases in risk of other types of injury, and the cost of obtaining and periodically replacing the PAB has been used to justify its discontinued use. The authors identified inpatient and outpatient treatment for injuries during US Army paratrooper training. Those undergoing training during two periods when PAB use was mandated had 40% lower risks of ankle injury (brace I, RR=0.60 (95% CI 0.47 to 0.75); brace II, RR=0.62 (95% CI 0.49 to 0.78)), with no difference in risks of other types of injury. There were no differences in risk of ankle injury during periods when brace use was not mandated. The PAB is safe, effective and cost effective.

Risk assessment of an essential element: manganese.

Manganese (Mn) is an essential element for humans, animals, and plants and is required for growth, development, and maintenance of health. Mn is present in most tissues of all living organisms and is present naturally in rocks, soil, water, and food. High-dose oral, parenteral, or inhalation exposures are associated with increased tissue Mn levels that may lead to development of adverse neurological, reproductive, or respiratory effects. Manganese-induced clinical neurotoxicity is associated with a motor dysfunction syndrome commonly referred to as manganism. Because Mn is an essential element and absorption and excretion are homeostatically regulated, a reasonable hypothesis is that there should be no adverse effects at low exposures. Therefore, there should be a threshold for exposure, below which adverse effects may occur only rarely, if at all, and the frequency of occurrence of adverse effects may increase with higher exposures above that threshold. Lowest-observed-adverse-effect levels (LOAELs), no-observed-adverse-effect levels (NOAELs), and benchmark dose levels (BMDs) have been derived from studies that were conducted to evaluate subclinical neurotoxicity in human occupational cohorts exposed to Mn. Although there is some uncertainty about the predictive value of the subclinical neuromotor or neurobehavioral effects that were observed in these occupational cohort studies, results of the neurological tests were used in risk assessments to establish guidelines and regulations for ambient air levels of Mn in the environment. A discussion of the uncertainties associated with these tests is provided in this review. The application of safety and uncertainty factors result in guidelines for ambient air levels that are lower than the LOAELs, NOAELs, or BMDs from occupational exposure studies by an order of magnitude, or more. Specific early biomarkers of effect, such as subclinical neurobehavioral or neurological changes or magnetic resonance imaging (MRI) changes, have not been established or validated for Mn, although some studies attempted to correlate certain biomarkers with neurological effects. Pharmacokinetic studies with rodents and monkeys provide valuable information about the absorption, bioavailability, and tissue distribution of various Mn compounds with different solubilities and oxidation states in different age groups. These pharmacokinetic studies showed that rodents and primates maintain stable tissue Mn levels as a result of homeostatic mechanisms that tightly regulate absorption and excretion of ingested Mn and limit tissue uptake at low to moderate levels of inhalation exposure. In addition, physiologically based pharmacokinetic (PBPK) models are being developed to provide for the ability to conduct route-to-route extrapolations, evaluate nasal uptake to the central nervous system (CNS), and determine life-stage differences in Mn pharmacokinetics. Such models will facilitate more rigorous quantitative analysis of the available human pharmacokinetic data for Mn and will be used to identify situations that may lead to increased brain accumulation related to altered Mn kinetics in different human populations, and to develop quantitatively accurate predictions of elevated Mn levels that may serve as a basis of dosimetry-based risk assessments. Such dosimetry-based risk assessments will permit for the development of more scientifically refined and robust recommendations, guidelines, and regulations for Mn levels in the ambient environment and occupational settings.

A review of health effects of carbon disulfide in viscose industry and a proposal for an occupational exposure limit.

Occupational exposure limits (OELs) for carbon disulfide vary between 1 and 10 ppm worldwide. They are generally based on health effects observed in viscose industry. Publications after the mid-1970s are reviewed to determine whether there is a scientific justification for an OEL below 10 ppm. The exposure situation in viscose industry is governed by long exposure durations, high exposures in past decades, high peak exposures, former analytical procedures underestimating exposure, and shift work. Three approaches were used to define an OEL based on workplace data: (1) Division of a cumulative exposure index by lifetime exposure duration. This approach ignores the possible existence of a threshold and fails to differentiate between brief high and sustained low exposures. (2) Defining the NOEL/LOEL by mean exposure levels. With a wide range of exposures, effects observed at the mean are driven by high exposures underestimating the true NOEL. (3) Assessment of effects observed at workplaces complying with a predefined exposure limit. Without adverse effects at such a limit this should be the starting point to define the OEL. The most important health effects for carbon disulfide are coronary heart disease, coronary risk factors, retinal angiopathy, color discrimination, effects on peripheral nerves, psychophysiological effects, morphological and other central nervous system (CNS) effects, and fertility and hormonal effects. The data generally support an OEL of 10 ppm. Some uncertainties exist for effects on electrocardiogram (ECG), heart rate, retinal microaneurysms (in Japanese workers), peripheral nerve conduction velocities, some psychophysiological parameters, brain magnetic resonance imaging (MRI; hyperintensive spots), and hearing function. Further investigations on workers under defined long-term exposure conditions might help to come to a final conclusion. Finally, the reproductive capacity of female workers may not be adequately protected at exposures around 10 ppm.

Prolonged acute mechanical ventilation, hospital resource utilization, and mortality in the United States.

OBJECTIVE: Adjusted costs of mechanical ventilation (MV) are $1,500 per patient-day. We compared the prevalence, characteristics, and outcomes of MV < 96 hrs (MV < 96) and prolonged acute MV (PAMV) of > or = 96 hrs' duration in a representative sample of U.S. hospital discharges. DESIGN: A multicenter cross-sectional study. SETTING: Nationally representative sample of U.S. hospital discharges. PATIENTS: Adult hospital discharges were identified from the 2003 Nationwide Inpatient Sample, Healthcare Cost and Utilization Project from the Agency for Healthcare Research and Quality (AHRQ). PAMV was based on the presence of ICD-9 code 96.72, and MV < 96 hrs based on ICD-9 codes 96.70 and 96.71. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Of 31,340,578 discharges for adults (> or = 18 yrs), 2.4% had any MV, of which 469,168 (61%) had MV < 96, and 294,333 (39%) had PAMV. Patient demographics were similar for MV < 96 and PAMV. With the exception of acute myocardial infarction and chronic and end-stage renal disease without dialysis, the prevalence of coexisting conditions was higher in the PAMV group. Median length of stay (17 vs. 6 days) and hospital costs ($40,903 vs. $13,434) also were higher with PAMV vs. MV < 96. Although Agency for Healthcare Research and Quality disease severity and mortality probability were higher in the PAMV than MV < 96 group, actual mortality was similar between the two groups (34% vs. 35%). CONCLUSIONS: There were nearly 300,000 PAMV discharges in the United States in 2003 at an annual aggregated hospital cost of > $16 billion, or nearly two thirds of the cost for all of the MV discharges. Despite a higher predicted mortality, patients requiring PAMV had the same likelihood of being discharged alive as those on shorter-term MV. These analyses will help inform health care decision-making and resource planning in the face of an aging population.

Risk Factors for Training-Related Injuries During U.S. Army Basic Combat Training.

Background: Injuries during basic combat training (BCT) impact military health and readiness in the U.S. Army. Identifying risk factors is crucial for injury prevention, but few Army-wide studies to identify risk factors for injury during BCT have been completed to date. This study examines associations between individual and training-related characteristics and injuries during Army BCT. Methods: Using administrative data from the Total Army Injury and Health Outcomes Database (TAIHOD), we identified individuals who apparently entered BCT for the first time between 1 January 2002 and 30 September 2007, based on review of administrative records. Injuries were identified and categorized based on coded medical encounter data. When combined with dates of medical services, we could count injuries per person, identify unique injuries, and identify the quantity and type of medical care delivered. Regression models produced odds ratios (ORs) and 95% confidence intervals (CIs) to identify risk factors for injury during BCT (yes/no), adjusted for potential confounders. Results: Of the 278,045 (83.4%) men and 55,302 (16.6%) women who were apparently first-time trainees, 39.5% (n = 109,760) of men and 60.9% (n = 33,699) of women were injured during training based on over 2 million recorded medical encounters entries. The large cohort yielded statistically significant, small magnitude associations between injury and all individual and training-related covariates for men, and all but medical accession waivers and weight for women. After adjustment, largest magnitude effects among men were due to age > 25 yr vs. 17-18 yr (OR = 1.83, 95% CI: 1.75, 1.91); having been married in the past vs. being single (OR = 1.36, 95% CI: 1.24, 1.49); rank E4-E7 vs. E1 (OR = 0.56, 95% CI: 0.53. 0.59); training at Ft. Jackson (OR = 0.66, 95% CI: 0.64, 0.69), Ft. Leonard Wood (OR = 0.67, 95% CI: 0.65, 0.70), or Ft. Knox (OR = 0.69, 95% CI: 066, 0.72) vs. Ft. Benning. Odds of injury were highest during 2005, 2006, and 2007. After adjustment for weight and body mass index, taller men had higher odds of BCT injury than average height men (OR = 1.08, 95% CI: 1.05, 1.11). Among women, short stature (OR = 1.11; 95% CI: 1.04, 1.19), training at Ft Leonard Wood (OR = 1.10; 95% CI: 1.04, 1.16) and evidence of injury prior to training based on accession waiver (OR = 1.12; 95% CI: 1.00, 1.26) increased injury risk. Conclusions: This Army-wide analysis reveals higher BCT-related injury rates for both men and women than prior studies and identifies risk factors for injuries during BCT. The large data set allows adjustment for many covariates, but because statistical analysis may yield significant findings for small differences, results must be interpreted based on minimally important differences determined by military and medical professionals. Results provide information that may be used to adapt training or medical screening and examination procedures for basic trainees.

Data on cardiovascular and pulmonary diseases among smokers of menthol and non-menthol cigarettes compiled from the National Health and Nutrition Examination Survey (NHANES), 1999-2012.

This Data in Brief contains results from three different survey logistic regression models comparing risks of self-reported diagnoses of cardiovascular and pulmonary diseases among smokers of menthol and non-menthol cigarettes. Analyses employ data from National Health and Nutrition Examination Survey (NHANES) cycles administered between 1999 and 2012, combined and in subsets. Raw data may be downloaded from the National Center for Health Statistics. Results were not much affected by which covariates were included in the models, but depended strongly on the NHANES cycles included in the analysis. All three models returned elevated risk estimates for three endpoints when they were run in individual NHANES cycles (congestive heart failure in 2001-02; hypertension in 2003-04; and chronic obstructive pulmonary disease in 2005-06), and all three models returned null results for these endpoints when data from 1999-2012 were combined.

Disability among U.S. Army personnel vaccinated against anthrax.

This study was conducted to examine whether U.S. Army personnel receiving > or =1 dose of anthrax vaccine adsorbed (AVA) between March 1998 and February 2002 were at higher risk of disability than unvaccinated personnel. We studied a historical cohort study of 716,833 active-duty soldiers (154,456 vaccinated) followed for 4.25 years to determine rates of evaluation for disability discharge. Cox proportional hazards models compared estimated risk of evaluation for disability, accounting for occupation and sociodemographics. Adjusted hazard ratio (HR) and 95% confidence interval (CI) was 0.96 (CI = 0.92-0.99). Separate adjusted HRs for men, women, permanent and temporary disability, musculoskeletal and neurologic conditions were similar, ranging from 0.90 to 1.04. Latency assumptions did not affect results. Anthrax vaccination does not increase risk of disability. This finding may be partially the result of factors influencing selection for vaccination or vaccine tolerance.

Cause-specific mortality among Kelly Air Force Base civilian employees, 1981-2001.

In response to concerns about occupational and environmental exposures, and a perceived cluster of amyotrophic lateral sclerosis (ALS) in the community, the mortality experience among 31,811 civilian employees who worked for at least 1 year between 1981 and 2000 at Kelly Air Force Base, Texas was ascertained. A total of 3264 deaths occurred through October 31, 2001. Overall, significant deficits in mortality were observed for all causes of death and all cancers combined. An excess of breast cancer [standardized mortality ratio (SMR) = 216; 95% confidence interval (CI) = 128-341] among blue-collar women was identified, and remained elevated after adjusting for race and ethnicity [rate ratio (RR) = 2.83; 95% CI = 1.50-5.34]. Mortality from motor neuron disease, which includes ALS deaths, was not increased overall (SMR = 0.98; 95% CI = 0.52-1.68), and was lower among blue-collar employees and higher among white-collar employees than expected, based on small numbers. Overall, mortality patterns indicated a healthy worker population and no large increased mortality associated with employment at Kelly Air Force Base.