A tool to assess risk of bias in non-randomized follow-up studies of exposure effects (ROBINS-E)

BACKGROUND

Observational epidemiologic studies provide critical data for the evaluation of the potential effects of environmental, occupational and behavioural exposures on human health. Systematic reviews of these studies play a key role in informing policy and practice. Systematic reviews should incorporate assessments of the risk of bias in results of the included studies.

OBJECTIVE

To develop a new tool, Risk Of Bias In Non-randomized Studies - of Exposures (ROBINS-E) to assess risk of bias in estimates from cohort studies of the causal effect of an exposure on an outcome.

METHODS AND RESULTS

ROBINS-E was developed by a large group of researchers from diverse research and public health disciplines through a series of working groups, in-person meetings and pilot testing phases. The tool aims to assess the risk of bias in a specific result (exposure effect estimate) from an individual observational study that examines the effect of an exposure on an outcome. A series of preliminary considerations informs the core ROBINS-E assessment, including details of the result being assessed and the causal effect being estimated. The assessment addresses bias within seven domains, through a series of 'signalling questions'. Domain-level judgements about risk of bias are derived from the answers to these questions, then combined to produce an overall risk of bias judgement for the result, together with judgements about the direction of bias.

CONCLUSION

ROBINS-E provides a standardized framework for examining potential biases in results from cohort studies. Future work will produce variants of the tool for other epidemiologic study designs (e.g. case-control studies). We believe that ROBINS-E represents an important development in the integration of exposure assessment, evidence synthesis and causal inference.

Systematic review and meta-analysis of birth weight and PFNA exposures

We used a systematic review that included risk of bias and study sensitivity analysis to identify 34 studies examining changes in birth weight (BWT) in relation to PFNA biomarker measures (e.g., maternal serum/plasma or umbilical cord samples). We fit a random effects model of the overall pooled estimate and stratified estimates based on sample timing and overall study confidence. We conducted a meta-regression to further examine the impact of gestational age at biomarker sample timing. We detected a -32.9 g (95%CI: -47.0, -18.7) mean BWT deficit per each ln PFNA increase from 27 included studies. We did not detect evidence of publication bias (p_E = 0.30) or between-study heterogeneity in the summary estimate (p_Q = 0.05; I² = 36%). The twelve high confidence studies yielded a smaller pooled effect estimate (β = -28.0 g; 95%CI: -49.0, -6.9) than the ten medium (β = -39.0 g; 95%CI: -61.8, -16.3) or four low (β = -36.9 g; 95%CI: -82.9, 9.1) confidence studies. The stratum-specific results based on earlier pregnancy sampling periods in 11 studies showed smaller deficits (β = -22.0 g; 95%CI: -40.1, -4.0) compared to 10 mid- and late-pregnancy (β = -44.2 g; 95%CI: -64.8, -23.5) studies and six post-partum studies (β = -42.9 g; 95%CI: -88.0, 2.2). Using estimates of the specific gestational week of sampling, the meta-regression showed results consistent with the categorical sample analysis, in that as gestational age at sampling time increases across these studies, the summary effect estimate of a mean BWT deficit got larger. Overall, we detected mean BWT deficits for PFNA that were larger and more consistent across studies than previous PFAS meta-analyses. Compared to studies with later sampling, BWT deficits were smaller but remained sizeable for even the earliest sampling periods. Contrary to earlier meta-analyses for PFOA and PFOS, BWT deficits that were detected across all strata did not appear to be fully explained by potential bias due to pregnancy hemodynamics from sampling timing differences.

Comparative susceptibility of children and adults to neurological effects of inhaled manganese: A review of the published literature

BACKGROUND

Manganese (Mn) is neurotoxic in adults and children. Current assessments are based on the more extensive adult epidemiological data, but the potential for greater childhood susceptibility remains a concern. To better understand potential lifestage-based variations, we compared susceptibilities to neurotoxicity in children and adults using Mn biomarker data.

METHODS

We developed a literature search strategy based on a Population, Exposures, Comparators, and Outcomes statement focusing on inhalation exposures and neurological outcomes in humans. Screening was performed using DistillerSR. Hair biomarker studies were selected for evaluation because studies with air measurements were unavailable or considered inadequate for children. Studies were paired based on concordant Mn source, biomarker, and outcome. Comparisons were made based on reported dose-response slopes (children vs. adults). Study evaluation was conducted to understand the confidence in our comparisons.

RESULTS

We identified five studies evaluating seven pairings of hair Mn and neurological outcomes (cognition and motor effects) in children and adults matched on sources of environmental Mn inhalation exposure. Two Brazilian studies of children and one of adults reported intelligent quotient (IQ) effects; effects in both comparisons were stronger in children (1.21 to 2.03-fold difference). In paired analyses of children and adults from the United States, children exhibited both stronger and weaker effects compared to adults (0.37 to 1.75-fold differences) on postural sway metrics.

CONCLUSION

There is limited information on the comparative susceptibility of children and adults to inhaled Mn. We report that children may be 0.37 to 2.03 times as susceptible as adults to neurotoxic effects of Mn, thereby providing a quantitative estimate for some aspects of lifestage variation. Due to the limited number of paired studies available in the literature, this quantitative estimate should be interpreted with caution. Our analyses do not account for other sources of inter-individual variation. Additional studies of Mn-exposed children with direct air concentration measurements would improve the evidence base.

The effect of age on the relative risk of lung cancer mortality in a cohort of chromium production workers

Background

Hexavalent chromium has been found to increase the risk of lung cancer in occupational studies. It has been suggested that the relative risk of lung cancer may vary by age.

Methods

The cohort examined is the Baltimore cohort of chromium production workers. The effect of age on the lung cancer risk from hexavalent chromium exposure was examined using a conditional Poisson regression modeling approach of Richardson and Langholz (R&L) and Cox models with interaction terms of age and cumulative hexavalent chromium exposure.

Results

The inclusion of multiple age groups in the R&L approach suggests the existence of an age effect that is also supported by a Cox proportional hazard analysis. The hazard ratio in Cox models with age‐cumulative exposure interaction terms was significantly elevated for the youngest age group and significantly decreased for the oldest age group.

Conclusions

Our analyses are consistent with the observation that younger chromium production workers have a greater lung cancer risk than older workers.

Methods for evaluating variability in human health dose-response characterization

The Reference Dose (RfD) and Reference Concentration (RfC) are human health reference values (RfVs) representing exposure concentrations at or below which there is presumed to be little risk of adverse effects in the general human population. The 2009 National Research Council report Science and Decisions recommended redefining RfVs as "a risk-specific dose (for example, the dose associated with a 1 in 100,000 risk of a particular end point)." Distributions representing variability in human response to environmental contaminant exposures are critical for deriving risk-specific doses. Existing distributions estimating the extent of human toxicokinetic and toxicodynamic variability are based largely on controlled human exposure studies of pharmaceuticals. New data and methods have been developed that are designed to improve estimation of the quantitative variability in human response to environmental chemical exposures. Categories of research with potential to provide new database useful for developing updated human variability distributions include controlled human experiments, human epidemiology, animal models of genetic variability, in vitro estimates of toxicodynamic variability, and in vitro-based models of toxicokinetic variability. In vitro approaches, with further development including studies of different cell types and endpoints, and approaches to incorporate non-genetic sources of variability, appear to provide the greatest opportunity for substantial near-term advances.

Epidemiology: a foundation of environmental decision making

Many epidemiologic studies are designed so they can be drawn upon to provide scientific evidence for evaluating hazards of environmental exposures, conducting quantitative assessments of risk, and informing decisions designed to reduce or eliminate harmful exposures. However, experimental animal studies are often relied upon for environmental and public health policy making despite the expanding body of observational epidemiologic studies that could inform the relationship between actual, as opposed to controlled, exposures and health effects. This paper provides historical examples of how epidemiology has informed decisions at the U.S. Environmental Protection Agency, discusses some challenges with using epidemiology to inform decision making, and highlights advances in the field that may help address these challenges and further the use of epidemiologic studies moving forward.

Human exposure to carbon-based fibrous nanomaterials: A review

In an emerging field of nanotechnologies, assessment of exposure to carbon nanotubes (CNT) and carbon nanofibers (CNF) is an integral component of occupational and environmental epidemiology, risk assessment and management, as well as regulatory actions. The current state of knowledge on exposure to carbon-based fibrous nanomaterials among workers, consumers and general population was studied in frame of the International Agency for Research on Cancer (IARC) Monographs-Volume 111 "Some Nanomaterials and Some Fibres". Completeness and reliability of available exposure data for use in epidemiology and risk assessment were assessed. Occupational exposure to CNT/CNF may be of concern at all stages of the material life-cycle from research through manufacture to use and disposal. Consumer and environmental exposures are only estimated by modeled data. The available information of the final steps of the life-cycle of these materials remains incomplete so far regarding amounts of handled materials and levels of exposure. The quality and amount of information available on the uses and applications of CNT/CNF should be improved to enable quantitative assessment of human exposure to these materials. For that, coordinated effort in producing surveys and exposure inventories based on harmonized strategy of material test, exposure measurement and reporting results is strongly encouraged.

Influence of exposure assessment and parameterization on exposure response. Aspects of epidemiologic cohort analysis using the Libby Amphibole asbestos worker cohort

Recent meta-analyses of occupational epidemiology studies identified two important exposure data quality factors in predicting summary effect measures for asbestos-associated lung cancer mortality risk: sufficiency of job history data and percent coverage of work history by measured exposures. The objective was to evaluate different exposure parameterizations suggested in the asbestos literature using the Libby, MT asbestos worker cohort and to evaluate influences of exposure measurement error caused by historically estimated exposure data on lung cancer risks. Focusing on workers hired after 1959, when job histories were well-known and occupational exposures were predominantly based on measured exposures (85% coverage), we found that cumulative exposure alone, and with allowance of exponential decay, fit lung cancer mortality data similarly. Residence-time-weighted metrics did not fit well. Compared with previous analyses based on the whole cohort of Libby workers hired after 1935, when job histories were less well-known and exposures less frequently measured (47% coverage), our analyses based on higher quality exposure data yielded an effect size as much as 3.6 times higher. Future occupational cohort studies should continue to refine retrospective exposure assessment methods, consider multiple exposure metrics, and explore new methods of maintaining statistical power while minimizing exposure measurement error.

Evaluating uncertainty to strengthen epidemiologic data for use in human health risk assessments

BACKGROUND

There is a recognized need to improve the application of epidemiologic data in human health risk assessment especially for understanding and characterizing risks from environmental and occupational exposures. Although there is uncertainty associated with the results of most epidemiologic studies, techniques exist to characterize uncertainty that can be applied to improve weight-of-evidence evaluations and risk characterization efforts.

METHODS

This report derives from a Health and Environmental Sciences Institute (HESI) workshop held in Research Triangle Park, North Carolina, to discuss the utility of using epidemiologic data in risk assessments, including the use of advanced analytic methods to address sources of uncertainty. Epidemiologists, toxicologists, and risk assessors from academia, government, and industry convened to discuss uncertainty, exposure assessment, and application of analytic methods to address these challenges.

SYNTHESIS

Several recommendations emerged to help improve the utility of epidemiologic data in risk assessment. For example, improved characterization of uncertainty is needed to allow risk assessors to quantitatively assess potential sources of bias. Data are needed to facilitate this quantitative analysis, and interdisciplinary approaches will help ensure that sufficient information is collected for a thorough uncertainty evaluation. Advanced analytic methods and tools such as directed acyclic graphs (DAGs) and Bayesian statistical techniques can provide important insights and support interpretation of epidemiologic data.

CONCLUSIONS

The discussions and recommendations from this workshop demonstrate that there are practical steps that the scientific community can adopt to strengthen epidemiologic data for decision making.

Increased risk of cancer mortality associated with cadmium exposures in older Americans with low zinc intake

Cadmium (Cd) exposure has been associated with increased cancer risk, and zinc (Zn) appears to reduce that risk. However, little is known about the combined influence of Cd and Zn on cancer risk. The aim of this study was to examine relationships between Cd exposure, Zn intake, and cancer mortality risks. The analyses used 5204 subjects aged 50 yr or older from the Third National Health and Nutrition Examination Survey (NHANES III, 1988-1994) and the mortality follow-up through December 31, 2006. Cox proportional hazards models were used to test associations. In total, 569 cancer deaths were recorded during an average follow-up of 12.4 yr, including 155 from lung, 61 from prostate, and 26 from breast cancer. A positive association between Cd and cancer mortality risk was identified for both genders. Despite limited cause-specific deaths, the increased risk associated with Cd was significant for lung cancer in men. All-cause cancer mortality risk was significantly elevated among women with Zn intakes below the recommended dietary allowance (RDA) compared with women who met the RDA. The effect of low dietary Zn was not observed in men. Similar trends for prostate and breast cancer deaths were not significant. There was a significant inverse association between cancer deaths and the Zn-to-Cd ratio for both genders. Cd exposure is an important independent risk factor of cancer mortality in older Americans and the risk appears exaggerated in those with inadequate dietary Zn. Additional studies are required to elucidate the mechanism(s) by which Zn participates in the carcinogenic influence of Cd.

Regression calibration for classical exposure measurement error in environmental epidemiology studies using multiple local surrogate exposures

Environmental epidemiologic studies are often hierarchical in nature if they estimate individuals' personal exposures using ambient metrics. Local samples are indirect surrogate measures of true local pollutant concentrations which estimate true personal exposures. These ambient metrics include classical-type nondifferential measurement error. The authors simulated subjects' true exposures and their corresponding surrogate exposures as the mean of local samples and assessed the amount of bias attributable to classical and Berkson measurement error on odds ratios, assuming that the logit of risk depends on true individual-level exposure. The authors calibrated surrogate exposures using scalar transformation functions based on observed within- and between-locality variances and compared regression-calibrated results with naive results using surrogate exposures. The authors further assessed the performance of regression calibration in the presence of Berkson-type error. Following calibration, bias due to classical-type measurement error, resulting in as much as 50% attenuation in naive regression estimates, was eliminated. Berkson-type error appeared to attenuate logistic regression results less than 1%. This regression calibration method reduces effects of classical measurement error that are typical of epidemiologic studies using multiple local surrogate exposures as indirect surrogate exposures for unobserved individual exposures. Berkson-type error did not alter the performance of regression calibration. This regression calibration method does not require a supplemental validation study to compute an attenuation factor.

Exposure to asbestos-containing vermiculite ore and respiratory symptoms among individuals who were children while the mine was active in Libby, Montana

BACKGROUND

Libby, Montana, was home to the largest vermiculite ore mine in the United States. The processing, use, and transport of the ore, which was contaminated with amphibole asbestos, led to generalized contamination of the community. The mine closed in 1990.

OBJECTIVES

We examined the prevalence of respiratory symptoms in 2000-2001 and their association with history of vermiculite exposure among people who were < or = 18 years of age when the mine closed.

METHODS

Information on respiratory symptoms and exposure history was collected by questionnaire in 2000-2001, at which time participants were 10-29 years old. Logistic regression was used to model the associations between exposures and outcomes adjusted for age, sex, and tobacco smoke exposure.

RESULTS

Of the 1,003 individuals included in the study, 10.8% reported usually having a cough, 14.5% reported experiencing shortness of breath when walking up a slight hill or hurrying on level ground, and 5.9% reported having coughed up bloody phlegm in the past year. These respiratory symptoms were positively associated with frequently handling vermiculite insulation compared with never handling vermiculite insulation. We found no association between vermiculite insulation in the house and respiratory symptoms. Respiratory symptoms were associated with other vermiculite exposures as well, and the number and frequency of these activities showed a positive trend with usually having a cough. We found no association between any of the activities and abnormal spirometry.

CONCLUSIONS

These data suggest that residents of Libby, Montana, who were children when the mine closed experienced some respiratory symptoms associated with asbestos-contaminated vermiculite exposure.

A reexamination of the PPAR-alpha activation mode of action as a basis for assessing human cancer risks of environmental contaminants

BACKGROUND

Diverse environmental contaminants, including the plasticizer di(2-ethylhexyl)phthalate (DEHP), are hepatocarcinogenic peroxisome proliferators in rodents. Peroxisome proliferator-activated receptor-alpha (PPAR-alpha) activation and its sequelae have been proposed to constitute a mode of action (MOA) for hepatocarcinogenesis by such agents as a sole causative factor. Further, based on a hypothesized lower sensitivity of humans to this MOA, prior reviews have concluded that rodent hepatocarcinogenesis by PPAR-alpha agonists is irrelevant to human carcinogenic risk.

DATA SYNTHESIS

Herein, we review recent studies that experimentally challenge the PPAR-alpha activation MOA hypothesis, providing evidence that DEHP is hepatocarcinogenic in PPAR-alpha-null mice and that the MOA but not hepatocarcinogenesis is evoked by PPAR-alpha activation in a transgenic mouse model. We further examine whether relative potency for PPAR-alpha activation or other steps in the MOA correlates with tumorigenic potency. In addition, for most PPAR-alpha agonists of environmental concern, available data are insufficient to characterize relative human sensitivity to this rodent MOA or to induction of hepatocarcinogenesis.

CONCLUSIONS

Our review and analyses raise questions about the hypothesized PPAR-alpha activation MOA as a sole explanation for rodent hepatocarcinogenesis by PPAR-alpha agonists and therefore its utility as a primary basis for assessing human carcinogenic risk from the diverse compounds that activate PPAR-alpha. These findings have broad implications for how MOA hypotheses are developed, tested, and applied in human health risk assessment. We discuss alternatives to the current approaches to these key aspects of mechanistic data evaluation.

Effects of exposure to 0.06 ppm ozone on FEV1 in humans: a secondary analysis of existing data

BACKGROUND

Ozone is a potent photochemical oxidant that produces transient, reversible decrements in the lung function of acutely exposed individuals. A recent study provided previously unavailable clinical data for 30 healthy young adults exposed to O(3) at 0.06 ppm. That study showed significant effects of 0.08 ppm on lung function, confirming the findings of others. However, exposure to 0.06 ppm O(3) was not reported to significantly affect lung function.

OBJECTIVES

We conducted this analysis to reevaluate the existing lung function data of the volunteers previously exposed to 0.06 ppm O(3).

METHODS

We obtained pre- and postexposure data on forced expiratory volume in 1 sec (FEV(1)) for all subjects who were previously exposed for 6.6 hr to filtered air or to 0.06 ppm or 0.08 ppm O(3). We used standard statistical methods appropriate for paired comparisons to reanalyze FEV(1) responses after exposure to 0.06 ppm O(3) relative to filtered air.

RESULTS

Controlling for filtered air responses, 24 of the 30 subjects experienced an O(3)-induced decrement in FEV(1). On average, 0.06 ppm O(3) exposure caused a 2.85% reduction in FEV(1) (p < 0.002), which was consistent with the predicted FEV(1) response from existing models. Although the average response was small, two subjects had > 10% FEV(1) decrements.

CONCLUSIONS

Exposure to 0.06 ppm O(3) causes a biologically small but highly statistically significant decrease in mean FEV(1) responses of young healthy adults.

Children's response to air pollutants

It is important to focus on children with respect to air pollution because (1) their lungs are not completely developed, (2) they can have greater exposures than adults, and (3) those exposures can deliver higher doses of different composition that may remain in the lung for greater duration. The undeveloped lung is more vulnerable to assault and less able to fully repair itself when injury disrupts morphogenesis. Children spend more time outside, where concentrations of combustion-generated air pollution are generally higher. Children have higher baseline ventilation rates and are more physically active than adults, thus exposing their lungs to more air pollution. Nasal breathing in adults reduces some pollution concentrations, but children are more typically mouth-breathers--suggesting that the composition of the exposure mixture at the alveolar level may be different. Finally, higher ventilation rates and mouth-breathing may pull air pollutants deeper into children's lungs, thereby making clearance slower and more difficult. Children also have immature immune systems, which plays a significant role in asthma. The observed consequences of early life exposure to adverse levels of air pollutants include diminished lung function and increased susceptibility to acute respiratory illness and asthma. Exposure to diesel exhaust, in particular, is an area of concern for multiple endpoints, and deserves further research.

Panel discussion review: session three--issues involved in interpretation of epidemiologic analyses--statistical modeling

The Clean Air Act mandates that the US Environmental Protection Agency (EPA) develop National Ambient Air Quality Standards for criteria air pollutants and conduct periodic reviews of the standards based on new scientific evidence. In recent reviews, evidence from epidemiologic studies has played a key role. Epidemiologic studies often provide evidence for effects of several air pollutants. Determining whether there are independent effects of the separate pollutants is a challenge. Among the many issues confronting the interpretation of epidemiologic studies of multi-pollutant exposures and health effects are those specifically related to statistical modeling. The EPA convened a workshop on 13 and 14 December 2006 in Chapel Hill, North Carolina, USA, to discuss these and other issues; Session Three of the workshop was devoted specifically to statistical modeling. Prominent statistical modeling issues in epidemiologic studies of air pollution include (1) measurement error across the co-pollutants; (2) correlation and multi-collinearity among the co-pollutants; (3) the timing of the concentration-response function; (4) confounding; and (5) spatial analyses.

Smallpox vaccination and ischemic coronary events in healthy adults

Although smallpox vaccine-associated myopericarditis has been reported, the risk of cardiac ischemic events remains uncertain. We identified personnel receiving the smallpox vaccination and compared them to a historical referent population. The rate of cardiac ischemia diagnoses in the 30 days following smallpox vaccination was 140.1 per 100,000 person-years, compared to 143.5 per 100,000 person-years in referent group (RR 1.0 [95% CI: 0.7-1.4]). The rate of cardiac ischemic events in vaccinees was 121.4 per 100,000 person-years before and 175.7 after adopting pre-vaccination cardiac screening (RR 1.4 [95% CI: 0.8-2.7]). Implementation of pre-vaccination cardiac risk factor screening was not associated with a reduction in cardiac events.

Particulate air pollution and the rate of hospitalization for congestive heart failure among medicare beneficiaries in Pittsburgh, Pennsylvania

The authors used a case-crossover approach to evaluate the association between ambient air pollution and the rate of hospitalization for congestive heart failure among Medicare recipients (aged > or =65 years) residing in Allegheny County (Pittsburgh area), Pennsylvania, during 1987-1999. They also explored effect modification by age, gender, and specific secondary diagnoses. During follow-up, 55,019 patients were admitted with a primary diagnosis of congestive heart failure. In single-pollutant models, particulate matter with an aerodynamic diameter of <10 microm (PM(10)), carbon monoxide, nitrogen dioxide, and sulfur dioxide-but not ozone-were positively and significantly associated with the rate of admission on the same day. The strongest associations were observed with carbon monoxide, nitrogen dioxide, and PM(10). The associations with carbon monoxide and nitrogen dioxide were the most robust in two-pollutant models, remaining statistically significant even after adjusting for other pollutants. Patients with a recent myocardial infarction were at greater risk of particulate-related admission; otherwise, there was no significant effect modification by age, gender, or other secondary diagnoses. These results suggest that short-term elevations in air pollution from traffic-related sources may trigger acute cardiac decompensation in heart failure patients and that those with certain comorbid conditions may be more susceptible to these effects.

A sensitivity analysis of bias in relative risk estimates due to disinfection by-product exposure misclassification

We conducted a sensitivity analysis of relative risk estimates using local area mean disinfection by-product exposures. We used Monte Carlo simulations to generate data representing 100 towns, each with 100 births (n=10,000). Each town was assigned a mean total trihalomethane (TTHM) exposure value (mean=45, SD=28) based on a variable number of sampling locations (range 2-10). True maternal TTHM exposure was randomly assigned from a lognormal distribution using that town's true mean value. We compared the effect of a 20 microg/l increase in TTHM exposure on the risk of small-for-gestational age infancy using the true maternal exposure compared to various weighting measures of the town mean exposures. The exposure metrics included: (1) unweighted town mean, (2) town mean weighted by the inverse variance of the town mean, (3) town mean weighted by the inverse standard deviation of the town mean, (4) town mean weighted by 1-(standard deviation of sites per town/mean across all towns), and (5) a randomly selected value from one of the sites within the town of residence. To estimate the magnitude of misclassification bias from using the town mean concentrations, we compared the true exposure odds ratios (1.00, 1.20, 1.50, and 2.00) to the mean exposure odds ratios from the five exposure scenarios. Misclassification bias from the use of unweighted town mean exposures ranged from 19 to 39%, increasing in proportion to the size of the true effect estimates. Weighted town mean TTHM exposures were less biased than the unweighted estimates of maternal exposure, with bias ranging from 0 to 23%. The weighted town mean analyses showed that attenuation of the true effect of DBP exposure was diminished when town mean concentrations with large variability were downweighted. We observed a trade-off between bias and precision in the weighted exposure analyses, with the least biased effects estimates having the widest confidence intervals. Effect attenuation due to intrasystem variability was most evident in absolute and relative terms for larger odds ratios.

Who is sensitive to the effects of particulate air pollution on mortality? A case-crossover analysis of effect modifiers

BACKGROUND

Populations susceptible to the effects of particulate matter have begun to be characterized, but the independent contributions of specific factors have not been explored.

METHODS

We used a case-crossover study to examine PM10-associated mortality risk during 1988-1991 among 65,180 elderly residents of Cook County, Illinois, who had a history of hospitalization for heart or lung disease. We assessed how the effect was independently modified by specific diagnoses and personal characteristics.

RESULTS

We found a 1.14% (95% confidence interval [CI] = 0.44% to 1.85%) increased risk of death per 10 microg/m3 increase in ambient PM10 concentration. Persons with heart or lung disease-but no specific diagnosis of myocardial infarction, diabetes, congestive heart failure, chronic obstructive pulmonary disorder, or conduction disorders-were at 0.74% (-0.29% to 1.79%) increased risk. Persons with a history of myocardial infarction had a 2.7-fold higher risk (CI = -2.1 to 7.4). Those with diabetes carried a 2.0-fold higher risk (CI = -1.5 to 5.5). Risk appeared to decrease with age among elderly men and increase with age among elderly women, but the estimated 3-way interaction was not precise enough to exclude the null. We found no indication that susceptibility varied by group-level socioeconomic measures.

CONCLUSION

Among a frail population, individuals diagnosed with myocardial infarction or diabetes were at greatest risk of death associated with high concentrations of PM10. These results suggest that their susceptibility may derive from prior vascular damage to the heart.

Selection bias and confounding in case-crossover analyses of environmental time-series data

The case-crossover study design is a popular analytic tool for estimating the effects of triggers of acute outcomes by environmental exposures. Although this approach controls for time-invariant confounders by design, it may allow for selection bias and confounding by time-varying factors. We conducted a simulation study of the sensitivity of the symmetric bidirectional case-crossover design to time-varying patterns in exposure and outcome. We identified the effects of selection bias and confounding on symmetric bidirectional case-crossover results and offer strategies to eliminate or reduce these biases. Selection bias results when exposure in the reference periods is not identically representative of exposure in the hazard periods, even when the distribution of exposure is stationary. This bias can be estimated and removed. Selection bias also occurs when the distribution of exposure is nonstationary, but the adjusted symmetric bidirectional case-crossover methodology substantially controls for this. Confounding results from a common temporal pattern in the exposure and the outcome time series, but can also be the result of patterns in exposure and outcome that, although asymptotically uncorrelated, are correlated at finite series lengths. All three biases are reduced by choosing shorter referent-spacing lengths. This effect is illustrated using data on air pollution and daily deaths in Chicago.

Control for seasonal variation and time trend in case-crossover studies of acute effects of environmental exposures

The case-crossover study design is used to study the triggers of acute outcomes in populations. It controls for all measured and unmeasured time-invariant confounders by design. Studies of environmental triggers of morbidity are potentially confounded by temporal trends in the outcome owing to omitted covariates. We conducted a simulation study of the case-crossover design's ability to control for temporal confounding patterns by design rather than through modeling. We compared five case-crossover control sampling strategies including the matched pair, a symmetric bi-directional, a total history approach, and two approaches proposed by Navidi (Biometrics 1998;54:596-605). We simulated true relative risks (RR) of 1.10 and 2.00 and induced confounding by seasonal patterns as well as linear and nonlinear long-term trends to yield estimated RR values as high as 3.18. The symmetric bi-directional approach was compared across four lag times and controlled for temporal confounding best when the lag was shortest. With a 1-week lag, it estimated the RR values as 1.10 and 2.01. The four other approaches failed to control for the temporal trends. Our simulations show that the symmetric bi-directional case-crossover design can substantially control for temporal confounding by design although it is not as efficient (66%) as Poisson regression analysis.

The epidemiology of infantile hypertrophic pyloric stenosis

Infants with infantile hypertrophic pyloric stenosis (IHPS) born from 1983 to 1988 and recorded in the California Birth Defects Monitoring Program (CBDMP) database were compared with their birth cohort by demographic characteristics and selected associated birth defects. We identified 1963 cases of IHPS for a cumulative incidence of 1.9 per 1000 livebirths. The cumulative incidence per 1000 livebirths was 2.4 in White, 1.8 in Hispanic, 0.7 in Black, and 0.6 in Asian infants. Between weeks 3-12 after birth, 1871 (95%) IHPS cases were diagnosed. Premature infants were diagnosed with IHPS later than term or post-term infants. The incidence of IHPS declined for those born to maternal age groups of > or = 25 years and, independently, for successive birth ranks. The probandwise concordance rate for IHPS in monozygous twins was less than unity (0.25-0.44), although higher than the concordance for dizygous twins (0.05-0.10). The incidence of Smith-Lemli-Opitz syndrome (SLO) diagnosed in infants with IHPS (3 of 1963) was 157-fold higher than the incidence of SLO diagnosed in the CBDMP population. IHPS occurs in all of the largest racial and ethnic groups in California, most frequently in White and Hispanic infants. Pyloric stenosis presents only within a brief phase of development, which may be delayed in premature infants. A predominant discordance of disease state in monozygous twins implies an aetiological role for undetermined environmental factors. The association between SLO, caused by deficient cholesterol synthesis, and IHPS deserves additional study. Infants with suspected SLO require close observation for the onset of IHPS.

Maternal medications and environmental exposures as risk factors for gastroschisis

In a case-control study of gastroschisis, we evaluated the risks associated with mother's first-trimester use of medications and with hobby or occupational exposures for 110 cases and 220 controls without a birth defect. Mothers of cases and controls were age-matched. For hobby or occupational exposures, we found significantly elevated risks for high levels of solvents (odds ratio (OR) = 3.8; 95% confidence interval (CI) = 1.6-9.2) and for colorants (OR = 2.3; 95% CI = 1.3-4.0). For medications, we found significantly elevated risks for two strong cyclooxygenase inhibitors, aspirin (OR = 4.7; 95% CI = 1.2-18.1) and ibuprofen (OR = 4.0; 95% CI = 1.0-16.0), but not for acetaminophen, a weak cyclooxygenase inhibitor. Periconceptional exposure to X rays was also associated with gastroschisis (OR = 2.5; 95% CI = 1.2-5.5), but exposure to antibiotics, antinauseants, sulfonamides, or oral contraceptives was not. We also found elevated risks for two decongestants, pseudoephedrine (OR = 2.1; 95% CI = 0.8-5.5) and phenylpropanolamine (OR = 10.0; 95% CI = 1.2-85.6). For the group of all decongestants, including also oxymetazoline and ephedrine, the risk was significantly elevated (OR = 2.4; 95% CI = 1.0-5.4). Controlling in multivariate analyses for several demographic and pregnancy variables associated with gastroschisis in a previous analysis [Torfs et al. (1994) Teratology 50: 44-53] did not substantially change the level or direction of the associations. Most of these associations are for vasoactive substances, which supports a vascular hypothesis for the pathogenesis of gastroschisis.

Population-based study of tracheoesophageal fistula and esophageal atresia

From a multiracial population of 1,035,384 births monitored by the California Birth Defects Monitoring Program (CBDMP) from 1983 to 1988, we ascertained 34 cases of esophageal atresia (EA), 204 cases of tracheoesophageal fistula (TEF) with an EA (TEF/EA), and 54 cases of TEF without EA. The total prevalence rate was 2.82 per 10,000 live births and stillbirths and showed no secular trend nor marked seasonal variation. Rates of multiple birth were high for each defect (TEF 3.7%, TEF/EA 4.9%, EA 8.8%; population 1.1%). Non-Hispanic whites were overrepresented for TEF and TEF/EA, but not for EA. Excluding trisomies, mean maternal age was above the population mean (26.8 years) for TEF (27.9 years) and TEF/EA (28.0 years), but not for EA (26.2 years). The proportion of trisomies was significantly higher for EA (23.5%) than for TEF (9.3%; P < 0.02) or for TEF/EA (7.4%; P < 0.003). We subdivided each defect into five mutually exclusive types: isolated, multiple, syndromic, chromosomal, and trisomic. Male-to-female (M/F) ratios varied considerably between types, both within and between defects. The highest M/F ratios within each defect were for the multiple type (TEF 2.29, TEF/EA 1.44, EA 1.33; population 1.05), and the lowest for trisomies (TEF 0.25; TEF/EA 0.25, and EA 0.60). All types except trisomies were significantly associated with a single umbilical artery. We found an association of EA and TEF/EA with dextrocardia (3.1% of cases), and confirmed the association of primary hydrocephalus with TEF and TEF/EA (2.7% of cases). The proportion of cases with additional midline defects or VATER or VACTERL type anomalies was similar for all three defects, suggesting a common developmental pathogenesis. Epidemiologic differences between defects, and between types within defects, may reflect differences in timing of the pathological process or differences in susceptibilities (e.g., by sex or aneuploidy) and emphasize the need to evaluate each defect and its types separately in epidemiologic studies.

A population-based study of gastroschisis: demographic, pregnancy, and lifestyle risk factors

Gastroschisis, an abdominal wall defect, most often occurs in infants of young mothers. To identify risk factors for gastroschisis, we conducted a case-control study in the population surveyed by the California Birth Defects Monitoring Program (CBDMP). From structured questionnaire data, we compared sociodemographic, reproductive, and lifestyle factors for 110 mothers of infants with gastroschisis with those for 220 age-matched mothers of normal infants. Univariate matched-pair analysis showed significant associations of gastroschisis with mother's education, yearly family income, marital status, a history of mother's mother smoking, mother's father's absence from home during the mother's youth, more than one elective abortion, a short interval between menarche and first pregnancy, siblings from different fathers, and use of either a recreational drug (either cocaine, amphetamine, marijuana, or LSD), alcohol, or tobacco during the trimester preceding pregnancy. For cocaine, amphetamine, and marijuana, use of more than one drug showed a stronger association than single drug use. The association was stronger if both parents used drugs. Although many variables were correlated, odds ratios (OR) were significant (95% confidence intervals) in multivariate conditional logistic analysis for: yearly family income < $10,000 [OR = 4.34 (1.54, 12.22)] or $10,000-$49,999 [OR = 3.93 (1.43, 10.80)]; mother's mother's smoking status not known [OR = 3.99 (1.66, 9.56)]; mother's father's absence from home during her youth [OR = 3.11 (1.14, 8.46)]; and drug use by mother [OR = 2.21 (1.21, 4.03)], father [OR = 1.66 (1.02, 2.69)], or both [OR = 3.05 (1.48, 6.28)]. The best predictive model explained 32% of the deviance. Young, socially disadvantaged women with a history of substance use were at highest risk for a child with a gastroschisis.

A population-based study of congenital diaphragmatic hernia

From 1983 through 1987, in a California population of 718,208 births, 237 infants were born with a congenital diaphragmatic hernia (CDH), a birth prevalence of 3.30 per 10,000 total births (live births and stillbirths). We proposed that the various types of this defect, characterized by their different pathogeneses, would be reflected in differences in their descriptive epidemiologies. We evaluated various demographic, maternal, and infant characteristics for three major types of defects, the Morgagni hernia, the pars sternalis hernia, and the posterolateral hernia, categorizing the latter type into isolated defect (N = 129), multiple congenital anomalies including nonchromosomal syndromes (N = 86), trisomies (N = 10), and chromosomal anomalies other than trisomies (N = 2). For the posterolateral hernia, we present the distribution of associated anomalies (43%) and specifically of midline defects (19%). Although the number of cases for the Morgagni hernia (N = 5) and the pars sternalis hernia (N = 5) were small, comparisons with the posterolateral hernia suggested lower sex ratios, of borderline significance for the pars sternalis hernia (P < 0.09), and higher mean maternal ages for both groups. Within the posterolateral type, we found a significantly higher male to female ratio (M/F = 1.58) only for the isolated subgroup compared to the population (P < 0.03), and a borderline significant rural/urban difference in prevalences (2.12 vs. 1.45 per 10,000) (P < 0.06). Additionally, the distribution of monthly prevalence rates adjusted for gestational age suggested opposite seasonal trends between the isolated and the other posterolateral hernias; within this latter subgroup the difference between the highest monthly rate (1.68) and the lowest (0.96) was of borderline significance (P < 0.09). Our results suggest the need to consider the respective types and subgroups of CDH separately in epidemiologic studies.

A population-based study of congenital diaphragmatic hernia

From 1983 through 1987, in a California population of 718,208 births, 237 infants were born with a congenital diaphragmatic hernia (CDH), a birth prevalence of 3.30 per 10,000 total births (live births and stillbirths). We proposed that the various types of this defect, characterized by their different pathogeneses, would be reflected in differences in their descriptive epidemiologies. We evaluated various demographic, maternal, and infant characteristics for three major types of defects, the Morgagni hernia, the pars sternalis hernia, and the posterolateral hernia, categorizing the latter type into isolated defect (N = 129), multiple congenital anomalies including nonchromosomal syndromes (N = 86), trisomies (N = 10), and chromosomal anomalies other than trisomies (N = 2). For the posterolateral hernia, we present the distribution of associated anomalies (43%) and specifically of midline defects (19%). Although the number of cases for the Morgagni hernia (N = 5) and the pars sternalis hernia (N = 5) were small, comparisons with the posterolateral hernia suggested lower sex ratios, of borderline significance for the pars sternalis hernia (P < 0.09), and higher mean maternal ages for both groups. Within the posterolateral type, we found a significantly higher male to female ratio (M/F = 1.58) only for the isolated subgroup compared to the population (P < 0.03), and a borderline significant rural/urban difference in prevalences (2.12 vs. 1.45 per 10,000) (P < 0.06). Additionally, the distribution of monthly prevalence rates adjusted for gestational age suggested opposite seasonal trends between the isolated and the other posterolateral hernias; within this latter subgroup the difference between the highest monthly rate (1.68) and the lowest (0.96) was of borderline significance (P < 0.09). Our results suggest the need to consider the respective types and subgroups of CDH separately in epidemiologic studies.