Invited commentary: in defense of ecologic studies for testing a linear-no threshold theory

Is Population Density Associated with Non-Communicable Disease in Western Developed Countries? A Systematic Review

Over the last three decades, researchers have investigated population density and health outcomes at differing scale. There has not been a systematic review conducted in order to synthesise this evidence. Following the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines, we systematically reviewed quantitative evidence published since 1990 on population density and non-communicable disease (NCD) within Westernised countries. Fifty-four studies met the inclusion criteria and were evaluated utilising a quality assessment tool for ecological studies. High population density appears to be associated with higher mortality rates of a range of cancers, cardiovascular disease and COPD, and a higher incidence of a range of cancers, asthma and club foot. In contrast, diabetes incidence was found to be associated with low population density. High and low population density are therefore risk markers for a range of NCDs, indicating that there are unidentified factors and mechanisms underlying aetiology. On closer examination, our synthesis revealed important and complex relationships between population density, the built environment, the nature of greenspace and man-made exposures. In light of increasing rates of morbidity and mortality, future research is required to investigate these associations in order to establish causative agents for each NCD.

The relationship between minimum alcohol prices, outlet densities and alcohol-attributable deaths in British Columbia, 2002-09

AIM

To investigate relationships between periodic increases in minimum alcohol prices, changing densities of liquor stores and alcohol-attributable (AA) deaths in British Columbia, Canada.

DESIGN

Cross-section (16 geographic areas) versus time-series (32 annual quarters) panel analyses were conducted with AA deaths as dependent variables and price, outlet densities and socio-demographic characteristics as independent variables.

SETTING AND PARTICIPANTS

Populations of 16 Health Service Delivery Areas in British Columbia, Canada.

MEASUREMENTS

Age-sex-standardized rates of acute, chronic and wholly AA mortality; population densities of restaurants, bars, government and private liquor stores; minimum prices of alcohol in dollars per standard drink.

FINDINGS

A 10% increase in average minimum price for all alcoholic beverages was associated with a 31.72% [95% confidence interval (CI): ± 25.73%, P < 0.05] reduction in wholly AA deaths. Significantly negative lagged associations were also detected up to 12 months after minimum price increases for wholly but not for acute or chronic AA deaths. Significant reductions in chronic and total AA deaths were detected between 2 and 3 years after minimum price increases. Significant but inconsistent lagged associations were detected for acute AA deaths. A 10% increase in private liquor stores was associated with a 2.45% (95% CI: ± 2.39%, P < 0.05), 2.36% (95% CI: ± 1.57%, P < 0.05) and 1.99% (95% CI: ± 1.76%, P < 0.05) increase in acute, chronic and total AA mortality rates.

CONCLUSION

Increases in the minimum price of alcohol in British Columbia, Canada, between 2002 and 2009 were associated with immediate and delayed decreases in alcohol-attributable mortality. By contrast, increases in the density of private liquor stores were associated with increases in alcohol-attributable mortality.

Blindness registrations and socioeconomic factors in Canada: an ecologic study

PURPOSE

To investigate the socioeconomic factors associated with blindness registration in Canada and its regions using an ecologic approach.

METHODS

Canadian National Institute for the Blind (CNIB) blindness registration data for 1996 were divided into units of analysis using postal codes and correlated with demographic and socioeconomic information collected by the 1996 Census of Canada. A total of 1250 units were analyzed representing 28,429,519 persons (98.55% of the population of Canada). Six socioeconomic factors were examined using weighted linear multivariate regression analysis: I) Percentage of the population aged 65 years and over; 2) Median household income; 3) Percentage of the population with university education; 4) Percentage of income derived from government transfer payments; 5) Recent immigrants; and 6) Visible minorities (blacks, Chinese, South Asians). Regression models were created for Canada as well as five geographic regions within Canada.

RESULTS

For Canada as a whole, blindness registration prevalence was positively correlated with age distribution and percentage of recent immigrants, and negatively correlated with level of government assistance income and percentage ethnic Chinese population. For five regional regression models, the common predictor variables were age distribution, median household income and percentage of the population who are black. None of the regional models produced an identical set of correlations.

CONCLUSIONS

Socioeconomic factors associated with blindness registration prevalence varied across different regions. Median household income was the second most common factor after age distribution, suggesting that areas with lower incomes tend to utilize more blind services. Higher blindness registration rates were associated with areas that had a higher percentage of the population who were black. Differences in blindness registration rates may reflect under-utilization of blind services and/or variations in disease and treatment rates in different populations.

A hybrid model for reducing ecological bias

A major drawback of epidemiological ecological studies, in which the association between area-level summaries of risk and exposure is used to make inference about individual risk, is the difficulty in characterizing within-area variability in exposure and confounder variables. To avoid ecological bias, samples of individual exposure/confounder data within each area are required. Unfortunately, these may be difficult or expensive to obtain, particularly if large samples are required. In this paper, we propose a new approach suitable for use with small samples. We combine a Bayesian nonparametric Dirichlet process prior with an estimating functions' approach and show that this model gives a compromise between 2 previously described methods. The method is investigated using simulated data, and a practical illustration is provided through an analysis of lung cancer mortality and residential radon exposure in counties of Minnesota. We conclude that we require good quality prior information about the exposure/confounder distributions and a large between- to within-area variability ratio for an ecological study to be feasible using only small samples of individual data.

Residential radon exposure and lung cancer risk: commentary on Cohen's county-based study

The large United States county-based study () in which an inverse relationship has been suggested between residential low-dose radon levels and lung cancer mortality has been reviewed. While this study has been used to evaluate the validity of the linear nonthreshold theory, the grouped nature of its data limits the usefulness of this application. Our assessment of the study's approach, including a reanalysis of its data, also indicates that the likelihood of strong, undetected confounding effects by cigarette smoking, coupled with approximations of data values and uncertainties in accuracy of data sources regarding levels of radon exposure and intensity of smoking, compromises the study's analytic power. The most clear data for estimating lung cancer risk from low levels of radon exposure continue to rest with higher-dose studies of miner populations in which projections to zero dose are consistent with estimates arising from most case-control studies regarding residential exposure.

Social disparities in housing and related pediatric injury: a multilevel study

OBJECTIVES

We conducted an ecologic analysis to determine whether housing characteristics mediate the associations between concentration of poverty and pediatric injury and between concentration of racial minorities and pediatric injury and whether the association between housing conditions and pediatric injury is independent of other risks.

METHODS

We created a hierarchical data set by linking individual-level data for pediatric injury with census data. Effect sizes were estimated with a Poisson model.

RESULTS

After adjustment for owner occupancy and the percentage of housing built before 1950, the association between concentration of poverty and pediatric injury was attenuated. For concentration of racial minorities, only percentage of owner occupancy had some mediating effect. In hierarchical models, housing characteristics remained independent and significant predictors of pediatric injury.

CONCLUSIONS

The association between community characteristics and pediatric injury is partially mediated by housing conditions. Risk of pediatric injury associated with housing conditions is independent of other risks.

Ecologic versus individual-level sources of bias in ecologic estimates of contextual health effects

A number of authors have attempted to defend ecologic (aggregate) studies by claiming that the goal of those studies is estimation of ecologic (contextual or group-level) effects rather than individual-level effects. Critics of these attempts point out that ecologic effect estimates are inevitably used as estimates of individual effects, despite disclaimers. A more subtle problem is that ecologic variation in the distribution of individual effects can bias ecologic estimates of contextual effects. The conditions leading to this bias are plausible and perhaps even common in studies of ecosocial factors and health outcomes because social context is not randomized across typical analysis units (administrative regions). By definition, ecologic data contain only marginal observations on the joint distribution of individually defined confounders and outcomes, and so identify neither contextual nor individual-level effects. While ecologic studies can still be useful given appropriate caveats, their problems are better addressed by multilevel study designs, which obtain and use individual as well as group-level data. Nonetheless, such studies often share certain special problems with ecologic studies, including problems due to inappropriate aggregation and problems due to temporal changes in covariate distributions.

Migration bias in ecologic studies

Differential migration may provoke bias in an epidemiological assessment of the public health risks from exposure to environmental agents, particularly in ecologic studies of health outcomes with a long latency or induction period. The potential impact of migration bias on epidemiological research is complex, and it depends not only on the direction of the factor-related migration, but also on its extent. This study shows that even a small amount of differential migration can bias the assessment of the exposure outcome relationship. Migration bias may result from a number of circumstances that are related to the way in which 'populations' are defined and ascertained. It is important to understand and minimise this type of bias in epidemiological research.

Socioeconomic and environmental covariates of premature mortality in Ontario

This paper contributes to debates on the broad determinants of health and on the policy shift from curative to preventive and protective interventions. It addresses empirically the relative importance of influences on health with a multiple regression analysis of ecologic data from the 49 counties of Ontario. One model achieved high predictive power (that is, Adj R2 > 75%, p < 0.0001). Educational levels were a strong predictor of population health, showing a consistent inverse relationship with premature mortality ratios for both sexes and it was the strongest predictor for females. A low income variable supplied the strongest prediction for male mortality. This variable displayed a positive association with male mortality. Municipal expenditures on environmental protection exerted a negative effect on male mortality. These findings raise questions about the current directions of health policy in Ontario where the provincial government has reduced funding to social and environmental programs, while promising to maintain health care funding.

How dangerous is low level radiation?

Problems in the theoretical basis for the linear-no threshold theory of radiation carcinogenesis are reviewed, and it is shown that they very strongly suggest that the theory greatly overestimates the risk of low level radiation. A direct test of the theory, based on the radon-lung cancer relationship is described; it strongly reinforces that conclusion. However, it is shown that even if the linear-no threshold theory is valid, the public's fear of low level radiation, at least in some contexts, is grossly exaggerated.

Violence by male partners against women during the childbearing year: a contextual analysis

OBJECTIVES

Many contextual analyses that bridge the micro-level-macro-level gap in identifying risk factors for adverse outcomes have not used methods appropriate for multilevel data. The purpose of this paper is to illustrate the application of appropriate multi-level analytic methods and discuss their implications for public health.

METHODS

A previously published individual-level model of physical violence perpetrated by male partners during the childbearing year was reanalyzed to include variables describing the neighborhoods where the women resided. Logistic regression with estimation methods of the generalized estimating equation was used for the contextual analysis. To assess the advantages of the generalized estimating equation over conventional logistic regression, both were used for the two-level model.

RESULTS

The regression coefficients from the contextual model differed from the betas obtained in the individual-level model. Not only were neighborhood-level variables related to the risk of partner-perpetrated violence, but the presence of these macro-level variables in the models modified the relationships of the individual-level variables to the risk of violence.

CONCLUSIONS

Two-level models that include individual- and community-level factors may be beneficial for purposes of explanation in public health research.