Lianne Sheppard

Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias.

The case–crossover design has been widely used to study the association between short-term air pollution exposure and the risk of an acute adverse health event. The design uses cases only; for each individual case, exposure just before the event is compared with exposure at other control (or “referent”) times. Time-invariant confounders are controlled by making within-subject comparisons. Even more important in the air pollution setting is that time-varying confounders can also be controlled by design by matching referents to the index time. The referent selection strategy is important for reasons in addition to control of confounding. The case–crossover design makes the implicit assumption that there is no trend in exposure across the referent times. In addition, the statistical method that is used—conditional logistic regression—is unbiased only with certain referent strategies. We review here the case–crossover literature in the air pollution context, focusing on key issues regarding referent selection. We conclude with a set of recommendations for choosing a referent strategy with air pollution exposure data. Specifically, we advocate the time-stratified approach to referent selection because it ensures unbiased conditional logistic regression estimates, avoids bias resulting from time trend in the exposure series, and can be tailored to match on specific time-varying confounders.

Referent selection in case-crossover analyses of acute health effects of air pollution.

The case-crossover design was proposed for the study of a transient effect of an intermittent exposure on the subsequent occurrence of a rare acute-onset disease. This design can be an alternative to Poisson time series regression for studying the health effects of fine particulate matter air pollution. Characteristics of time-series of particulate matter, including long-term time trends, seasonal trends, and short-term autocorrelations, require that referent selection in the case-crossover design be considered carefully and adapted to minimize bias. We performed simulations to evaluate the bias associated with various referent selection strategies for a proposed case-crossover study of associations between particulate matter and primary cardiac arrest. Some a priori reasonable strategies were associated with a relative bias as large as 10%, but for most strategies the relative bias was less than 2% with confidence interval coverage within 1% of the nominal level. We show that referent selection for case–crossover designs raises the same issues as selection of smoothing method for time series analyses. In addition, conditional logistic regression analysis is not strictly valid for some case–crossover designs, introducing further bias.

Dietary fat and cancer: consistency of the epidemiologic data, and disease prevention that may follow from a practical reduction in fat consumption

International variations and national time trends in disease rates suggest major associations between dietary fat and several important cancers. In contrast, case-control and cohort studies of dietary fat in relation to the same cancers generally report weak associations, or have failed to detect any association with fat intake. This study was undertaken in an attempt to understand the apparent discrepancy between these observations. The results provide an insight into the magnitude of cancer risk reduction that may follow from a practical reduction in dietary fat.Regression analyses of international variations in cancer incidence rates were used to estimate relative risks (RR) as a function of fat intakes for both males and females. These analyses focused on cancers of the breast, colon, rectum, ovary, and endometrium in females, and colon, rectum, and prostate cancers in males.Ages 55–69 and 30–44 were considered in order to compare RR estimates between an older and younger age group, and between post- and pre-menopausal women. Corresponding RR estimates were also calculated, based on the regression of changes in disease rates from the mid-1960s to 1980 on changes in dietary fat, using data from several countries. A strong degree of consistency with the RR estimates from international comparisons was observed. The international regression analyses were also used to project changes in cancer rates among Japanese migrants to the United States. A high level of consistency with the observed disease-rate changes was noted. Similarly, the international data analyses were used to project RRs for the fat intake categories used in specific case-control and cohort studies, while acknowledging measurement error in individual dietary assessment. Although certain exceptions are noted, considerable consistency was found between the aggregate and analytic data results, leaving open the strong possibility that a practical reduction in dietary fat could result in a major reduction in the incidence of several prominent cancers in the United States and in other nations having high fat consumption.

Effects of ambient air pollution on nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994.

As part of the Clean Air Act, Congress has directed EPA to set air quality standards to protect sensitive population groups from air pollutants in the ambient environment. People with asthma represent one such group. We undertook a study of the relation between measured ambient air pollutants in Seattle and nonelderly hospital admissions with a principal diagnosis of asthma. We regressed daily hospital admissions to local hospitals for area residents from 1987 through 1994 on particulate matter less than 10 and 2.5 microm in aerodynamic diameter (PM10 and PM2.5, respectively); coarse particulate mass; sulfur dioxide (SO2); ozone (O3); and carbon monoxide (CO) in a Poisson regression model with control for time trends, seasonal variations, and temperature-related weather effects. With the exception of seasonally monitored O3, we supplemented incomplete pollutant measures in a multiple imputation model to create a complete time series of exposure measures. We found an estimated 4-5% increase in the rate of asthma hospital admissions associated with an interquartile range change in PM (19 microg/m3 PM(10),11.8 microg/m3 PM2.5, and 9.3 microg/m3 coarse particulate mass) lagged 1 day; relative rates were as follows: for PM10, 1.05 [95% confidence interval (CI) = 1.02-1.08]; for PM2.5, 1.04 (95% CI = 1.02-1.07); and for coarse particulate mass, 1.04 (95% CI = 1.01-1.07). In single-pollutant models we also found that a 6% increase in the rate of admission was associated with an interquartile range change in CO (interquartile range, 924 parts per billion; 95% CI = 1.03-1.09) at a lag of 3 days and an interquartile range change in O3 (interquartile range, 20 parts per billion; 95% CI = 1.02-1.11) at a lag of 2 days. We did not observe an association for SO2. We found PM and CO to be jointly associated with asthma admissions. We estimated the highest increase in risk in the spring and fall seasons.

Worker recovery expectations and fear-avoidance predict work disability in a population-based workers' compensation back pain sample

Study Design. Prospective, population-based cohort study. Objectives. To examine whether worker demographic, pain, disability, and psychosocial variables, assessed soon after work-related back pain disability onset, predict 6-month work disability. Summary of Background Data. Greater age, pain, and physical disability, and certain psychosocial characteristics may be risk factors for prolonged back pain-related work disability, although many studies have been small, findings have been inconsistent, and some psychosocial variables have not been examined prospectively. Methods. Workers (N = 1,068) completed telephone interviews assessing demographic, pain, disability, and psychosocial variables 18 days (median) after submitting Workers’ Compensation back pain disability claims. Administrative measures of work disability 6 months after claim submission were obtained. Results. At 6 months, 196 workers (18.4%) were receiving work disability compensation. Age, race, education, and baseline pain and disability were significant predictors of 6-month disability. Adjusting for baseline demographics, pain, disability, and other psychosocial variables, high work fear-avoidance (odds ratio, 4.6; 95% confidence interval, 1.6–13.7) and very low recovery expectations (odds ratio, 3.1, 95% confidence interval, 1.5–6.5) were significant independent predictors. Conclusions. Among individuals with acute work-related back pain, high pain and disability, low recovery expectations, and fears that work may increase pain or cause harm are risk factors for chronic work disability.

Source identification of PM2.5 in an arid Northwest U.S. City by positive matrix factorization

Spokane, WA is prone to frequent particulate pollution episodes due to dust storms, biomass burning, and periods of stagnant meteorological conditions. Spokane is the location of a long-term study examining the association between health effects and chemical or physical constituents of particulate pollution. Positive matrix factorization (PMF) was used to deduce the sources of PM2.5 (particulate matter V2.5 Am in aerodynamic diameter) at a residential site in Spokane from 1995 through 1997. A total of 16 elements in 945 daily PM2.5 samples were measured. The PMF results indicated that seven sources independently contribute to the observed PM2.5 mass: vegetative burning (44%), sulfate aerosol (19%), motor vehicle (11%), nitrate aerosol (9%), airborne soil (9%), chlorinerich source (6%) and metal processing (3%). Conditional probability functions were computed using surface wind data and the PMF deduced mass contributions from each source and were used to identify local point sources. Concurrently measured carbon monoxide and nitrogen oxides were correlated with the PM2.5 from both motor vehicles and vegetative burning. D 2003 Elsevier Science B.V. All rights reserved.

Fine Particulate Air Pollution and the Progression of Carotid Intima-Medial Thickness: A Prospective Cohort Study from the Multi-Ethnic Study of Atherosclerosis and Air Pollution

In a prospective cohort study, Sara Adar and colleagues find that decreasing levels of fine particulate matter in multiple US urban areas are associated with slowed progression of intima-medial thickness, a surrogate measure of atherosclerosis.

Relation between short-term fine-particulate matter exposure and onset of myocardial infarction.

Background: Epidemiologic studies have reported increases in the incidence of cardiovascular morbidity and myocardial infarction (MI) associated with increases in short-term and daily levels of fine-particulate matter air pollution, suggesting a role for particulate matter in triggering an MI. Methods: We studied the association between onset time of MI and preceding hourly measures of fine-particulate matter using a case-crossover study of 5793 confirmed cases of acute MI. We linked data from a community-wide database on acute MI from 1988–1994 in King County, Washington, with central site air pollution monitoring data on fine-particulate matter determined by nephelometry. We compared air pollution exposure levels averaged 1 hour, 2 hours, 4 hours, and 24 hours before MI onset to a set of time-stratified referent exposures from the same day of the week in the month of the case event. Results: The estimated relative risk for a 10-μg/m3 increase in fine-particulate matter the hour before MI onset was 1.01 (95% CI = 0.98–1.05). Analyses of pollutant levels at the other time points demonstrated a similar lack of association. No increased risk was found in all cases with preexisting cardiac disease (odds ratio = 1.05; 0.95–1.16). Stratification by known cardiovascular risk factors (hypertension, diabetes, and smoking status) also did not modify the relation between fine-particulate matter and MI onset. Conclusion: Although a very small effect cannot be excluded, there was no consistent association between ambient levels of fine-particulate matter and risk of MI onset.

Effects of ambient air pollution on symptom severity and medication use in children with asthma

BACKGROUND Exposure to air pollutants has been investigated as a possible cause of asthma attacks in children. OBJECTIVE To investigate the short-term effects of air pollutants on a panel of 133 children with asthma who enrolled in the Childhood Asthma Management Program. METHODS During screening, the children completed daily diary cards for an average of 58 days to indicate their medication use and asthma severity. We used ordinal logistic regression to compare the odds of a more serious relative to a less serious asthma attack, and we used a Poisson model to analyze medication use. In both analyses we accommodate dependence in the data and different periods of observation for study subjects. RESULTS Our results indicate that a 10-microg/m3 increase in particulate matter less than or equal to 2.5 microm (PM2.5) lagged 1 day was associated with a 1.20 times increased odds of having a more serious asthma attack [95% confidence interval (CI), 1.05 to 1.37] and a 1.08-fold increase in medication use (95% CI, 1.01 to 1.15). A 10-microg/m3 increase in particulate matter less than or equal to 10 microm (PM10) increased the odds of a more serious asthma attack (odds ratio = 1.12; 95% CI, 1.04 to 1.22) and also increased medication use (relative risk = 1.05; 95% CI, 1.00 to 1.09). CONCLUSIONS Increases in PM2.5 and PM10 are significantly associated with an increased risk of more severe asthma attacks and medication use in Seattle area children with asthma. We also found associations with carbon monoxide, but we believe that carbon monoxide is a marker for exposure to combustion byproducts.

Predictors of mortality in patients diagnosed with probable Alzheimer's disease

Objective: We sought to identify factors associated with mortality in persons recently diagnosed with probable Alzheimers disease (AD). Background: Predicting mortality in AD is needed both in patient care and public health planning. Previous studies have identified several factors which contribute to mortality in AD, but few longitudinal studies of population-based cohorts exist. Methods: In a longitudinal follow-up study 327 patients with newly diagnosed probable AD (mean Mini-Mental State Examination [MMSE] score of 20) from a large, stable health maintenance organization were identified. Demographic characteristics, dementia severity, and comorbid conditions were identified at enrollment. Patients were followed longitudinally (median 3.3 years, total 898 person-years). Baseline characteristics were used to predict survival in univariate and multivariate models. Results: Increased mortality was seen in patients with probable AD (9.0 deaths per 100 person-years) compared with the community population adjusted for age and gender (4.3 deaths per 100 person-years). On univariate analysis we found increased age, male gender, impairment on MMSE or Blessed dementia rating scale (DRS), rate of MMSE decline, wandering or agitation, vascular disease, and sensory impairment affecting the ability to read or hear to be moderately associated with decreased survival. After adjusting for age and gender in a multivariate model, Blessed DRS score and sensory impairment affecting the ability to read were independently associated with decreased survival. Conclusions: Short-term mortality is increased in patients newly diagnosed with probable AD. Measures of dementia severity, measures of general debility, and vascular disease are associated with increased mortality. Of these, general debility and sensory impairment were more strongly associated with shortened survival. NEUROLOGY 1996;47: 433-439