Article Text
PDF
Abstract
Background There is a substantial variation in life expectancy across US counties, primarily owing to differentials in chronic diseases. The authors' aim was to examine the association of life expectancy and mortality from selected diseases with altitude.
Methods The authors used data from the National Elevation Dataset, National Center for Heath Statistics and US Census. The authors analysed the crude association of mean county altitude with life expectancy and mortality from ischaemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD) and cancers, and adjusted the associations for socio-demographic factors, migration, average annual solar radiation and cumulative exposure to smoking in multivariable regressions.
Results Counties above 1500 m had longer life expectancies than those within 100 m of sea level by 1.2–3.6 years for men and 0.5–2.5 years for women. The association between altitude and life expectancy became non-significant for women and non-significant or negative for men in multivariate analysis. After adjustment, altitude had a beneficial association with IHD mortality and harmful association with COPD, with a dose–response relationship. IHD mortality above 1000 m was 4–14 per 10 000 people lower than within 100 m of sea level; COPD mortality was higher by 3–4 per 10 000. The adjusted associations for stroke and cancers were not statistically significant.
Conclusions Living at higher altitude may have a protective effect on IHD and a harmful effect on COPD. At least in part due to these two opposing effects, living at higher altitude appears to have no net effect on life expectancy.
- Geography FQ
Statistics from Altmetric.com
Introduction
There is substantial variation in life expectancy across US counties, primarily due to differentials in mortality from chronic diseases.1 A number of studies have examined variation in county all-cause and disease-specific mortality in relation to socio-demographic factors,2–4 but substantially fewer studies have considered associations with the physical environment.
Among environmental factors, a number of studies have found an association between altitude and mortality from specific chronic diseases.5–7 However, few studies have included multiple disease outcomes and their net effects on all-cause mortality,8 9 and none has examined the association with life expectancy. Few previous studies assessed dose–response relationships over a wide elevation range.
We analysed the association of life expectancy and mortality from leading chronic diseases with mean elevation of county of residence in the continental USA. We adjusted the associations for socio-demographic factors, migration, average annual solar radiation and smoking. This produces consistent and comparable analysis of the altitude–mortality association for multiple diseases in the general population and over a large elevation range.
Methods
We analysed the association between mean county elevation and (1) life expectancy and (2) age-standardised death rates from cancers and cardiopulmonary diseases in the continental USA. We focused on selected cardiopulmonary diseases and cancers because they are leading causes of death in the USA, because they have been associated with elevation in previous studies and because there are plausible mechanisms for the associations to be causal.5–13 We examined specific cardiopulmonary causes with large numbers of deaths, including ischaemic heart disease (IHD), stroke and chronic obstructive pulmonary disease (COPD) separately because they seem to have different altitude associations, with plausible mechanistic reasons for such differences. We also separately considered the group of cancers that have an established causal association with smoking14 and those that are not currently linked with smoking, to avoid residual confounding due to smoking. Results for other diseases, for example site-specific cancers, are available from the authors by request. We restricted the analysis of disease-specific mortality to those 45 years of age and older because deaths from these diseases are relatively rare in younger ages leading to unstable estimates.
We adjusted the associations for socio-demographic factors, migration, average annual solar radiation and cumulative exposure to smoking in multivariate linear regressions. The socio-demographic factors in the multivariate regressions were county per-capita income, county population density, percentage of the county population who are black, who are Hispanic, who live in urban centres, who adhere to the Mormon religion and who migrated into the county in the last 5 years. We included percentage black and percentage Hispanic because there are mortality disparities by race and ethnicity in the USA. Urban living and population density may affect disease-specific mortality through differences in environment, lifestyle or healthcare access. The percentage Mormon is relevant in our study because this group has a lower mortality, primarily owing to the effects of smoking,15 and forms a greater proportion of the population of counties at higher altitudes (9.4% above 700 m and 17.7% above 1500 m, vs 1.5% nationally); our results were not sensitive, adjusting for this variable. We used migration to adjust for population turnover and for potential confounding by selective migration—for example, if those with pre-existing disease are more likely to migrate.11 16 We used age-standardised adult lung cancer death rate, as an indicator of the accumulated exposure to smoking (including passive smoking).17
To assess residual confounding by race or ethnicity (more white populations at higher altitudes) or region of the country (lower altitudes in Eastern US), we conducted sensitivity analyses by restricting the analysis to non-Hispanic Caucasians, western regions (Pacific and Mountain Divisions) and non-Hispanic Caucasians in western regions.
Data sources
Mortality and population
We used mortality and population data for 2001 through 2005 for this analysis, with a total of 12.1 million death records. Mortality data, including county of residence and cause of death, certified and coded according to the International Classification of Diseases system, were from the National Center for Health Statistics (NCHS), adjusted for comparability of cause of death assignment as described in detail elsewhere.18 19 County population data by age and bridged race categories were from the NCHS.20
Elevation and solar radiation
County elevation was calculated using the National Elevation Dataset, maintained by the US Geological Survey and accessed via the USGS Seamless Data Portal (http://seamless.usgs.gov/). The National Elevation Dataset data, which provide elevation for all 30 m×30 m geographic units in the USA, were combined with county boundaries from the National Atlas in Geographical Information System software. The mean elevation for each county was calculated as the average elevation of all 30 m×30 m units that constituted a county, using the Spatial Analyst Zonal Calculation tool. County population density was based on land area from the National Oceanic and Atmospheric Administration National Coastal Assessment Framework (http://coastalgeospatial.noaa.gov/). County annual average solar radiation (daily normal insolation, DNI) was obtained from the National Renewable Energy Laboratory, operated by the Midwest Research Institute for the US Department of Energy.
Socio-demographic variables
The proportion of county population by race and Hispanic status for 2003 (midpoint of analysis years) was from the US Census Population Estimates (http://www.census.gov/popest/). County median household income was from the US Census Small Area Income and Poverty Estimates (http://www.census.gov/did/www/saipe/).
Migration
The US Census provides data on in-migration in the 5 years prior to the census. The IRS uses Taxpayer Identification Number to estimate county-to-county migration. In the IRS data, the cross-county correlation between in- and out-migration is >0.96. To avoid collinearity, we included only one migration measure in the multivariate analysis. We used the 5-year migration from the 2000 US Census because older people and the poor may be under-represented in the IRS data.
Statistical analysis
Based on previous analyses,1 we arranged the 3115 counties and county equivalents (eg, independent cities, parishes and boroughs) in the continental US into 2063 units, each consisting of one or multiple individual counties. Merged county units were formed to avoid unstable death rates in small counties and to account for changes in county definitions and lines over time. For each county unit, we calculated life expectancy and cause-specific death rates from death and population data in 5-year age groups using standard life-table techniques.21 We pooled death and population counts over 5 years (2001–2005) to have sufficient number of deaths in all counties.
In a multivariate analysis, we grouped counties into elevation bands to allow for a non-linear relationship. The 14 elevation bands were in 100 m (328 ft) increments up to 1000 m and in 500 m increments above 1000 m because there are fewer counties at higher altitudes.
All analyses were done separately by sex. All analyses were conducted using Stata version 10.0 and ESRI ArcGIS version 9.2.
Results
Life expectancy in US counties in 2001–2005 pooled data ranged from 61.3 to 81.1 years for men and from 70.2 to 86.0 years for women. Age-standardised cardiovascular disease (CVD) and cancer mortality among people ≥45 years had a three- to fivefold variation across counties: CVD mortality ranged from 47 to 213 per 10 000 for men and from 50 to 179 for women; cancer mortality ranged from 25 to 134 per 10 000 for men and from 18 to 104 for women. Counties at mean altitudes 500–1000 m, 1000–1500 m and ≥1500 m covered 18%, 13% and 19% of the landmass of continental USA, respectively and in 2003 had populations of 32.5 million (11% of the continental US), 7.7 million (2.7%) and 10.1 million (3.5%), respectively.
Altitude and life expectancy
Of the 20 counties with the highest life expectancy, 11 for men and five for women are in Colorado and Utah, all at mean elevations of 1819 m or higher. Men and women who lived in counties at a higher altitude in the USA had a longer life expectancy (r=0.24 for men and 0.22 for women; p<0.001) (figure 1). Male life expectancy at the three elevation bands above 1500 m ranged from 75.8 to 78.2 years; female life expectancy in these three elevation bands ranged from 80.5 to 82.5 years (figure 2). These were longer than life expectancy within 100 m of sea level by 1.2–3.6 years for men and 0.5–2.5 years for women (p<0.001). Life expectancy did not rise monotonically at 500–900 m possibly due to socio-demographic composition of these counties, namely (1) the proportion of county population that are Asian (with longer life expectancy) or black (with shorter life expectancy) and (2) the balance between Caucasians from the Northern Plains (with longer life expectancy) or Appalachia (with shorter life expectancy).2
(A) Mean elevation, (B) male life expectancy and (C) female life expectancy by county in the continental USA.
Percentage of the US population residing in each elevation band (bars) and elevation-band life expectancies (points).
After adjustment for socio-demographic factors, migration, average annual solar radiation and smoking, the association between altitude and life expectancy became non-significant for women and non-significant or negative for men (table 1).
Coefficients, CIs and p values for multivariable regression analysis (life expectancy)
Altitude and mortality from cardiopulmonary diseases and cancers
Age-standardised adult cardiopulmonary and cancer death rates at higher altitudes were less than those within 100 m of sea level, by 6–16% at altitudes of 1500–2000 m, by 10–22% at 2000–2500 m and by 17–34% above 2500 m. In crude associations, IHD mortality declined almost monotonically, and COPD mortality increased with altitude for both sexes.
After adjustment, altitude continued to have a marked beneficial association with IHD and a detrimental association with COPD; the association had a dose–response relationship for IHD and, up to 1500 m, for COPD (table 2, figure 3). The beneficial effects for IHD were more consistent for women than for men. For example, the four elevation bands above 1000 m had four to eight fewer IHD deaths per 10 000 adult men and 10–14 fewer IHD deaths per 10 000 adult women than those below 100 m. In contrast, the same elevation bands experienced three to four more COPD deaths per 10 000 adults. The adjusted associations with total stroke were not statistically significant below 1500 m, and there was little evidence of a dose–response relationship.
Coefficients to CIs and p values for multivariable regression analysis (cause-specific adult mortality)
Adjusted association of elevation band with age-standardised death rates from leading chronic diseases for adults ≥45 years. Vertical lines show the 95% CIs. COPD, chronic obstructive pulmonary disease; IHD, ischaemic heart disease.
Both smoking-related cancers and other cancers were inversely associated with altitude in crude analyses. This association mostly disappeared after controlling for socio-demographic variables and cumulative smoking, except possibly some increased risk of non-smoking related cancers among women above 800 m (table 2, figure 3). Solar radiation had a significant beneficial association with mortality from cancers after adjustment for elevation and other variables.
Sensitivity analyses
The correlation between county elevation and life expectancy remained significant when the analysis was restricted to non-Hispanic Caucasians (r=0.19 for men and 0.16 for women; p<0.001). In the multivariate analysis, the associations of elevation bands with life expectancy and disease-specific mortality had the same direction and magnitude as the main analysis, except for COPD mortality, where the adjusted elevation effect was reduced by half for women.
When the analysis was restricted to states in western regions, the correlation between county elevation and life expectancy was not significant. In the multivariate analysis, the associations of elevation bands with various outcomes were similar to the main analysis, except for IHD mortality, where the elevation effect was reduced by half for both sexes.
Finally, when restricted to non-Hispanic Caucasians in western regions, the correlation between county elevation and life expectancy was not significant. Adjusted elevation band coefficients were unchanged from the main analysis.
Discussion
Principal findings
In this study of altitude–mortality relationships using national data, we found that living at a higher altitude was associated beneficially with IHD but harmfully with COPD, with a dose–response relationship for all (IHD) or part (COPD) of the elevation range. At least in part owing to these two opposing effects, after adjustment for socio-demographic factors, migration, solar radiation and smoking, there was no significant effect of living at a higher altitude on life expectancy for women and an apparently detrimental effect for men.
Interpretation, relation to other studies and research needs
The observed associations between altitude and mortality may have several explanations. Most interestingly, the results may reflect real physiological mechanisms. The dose–response relationships for IHD and COPD support a physiological mechanism related to environmental factors that have an altitude gradient, such as decreased partial pressure of oxygen at higher altitude which seems to be cardioprotective22 but detrimental for COPD.23 24 Possible cardioprotective mechanisms for our findings and those of previous studies include enhanced cardiac metabolic efficiency and substrate utilisation at the cellular and transcriptional levels, increased angiogenesis, improved lipid profiles and structural vascular changes that lower blood pressure.5 7–10 12 22 25–31 Some of these mechanisms (eg, increased angiogenesis and structural vascular changes) are consistent with our finding of a beneficial association with IHD and no consistent association with stroke. An alternative possible mechanism for the potential benefits of altitude for IHD, and for some cancers, may be increased solar radiation and vitamin D synthesis.32–37 In our multivariable regression, average annual solar radiation had a significant beneficial association with mortality from cancer, no association with IHD and a harmful association with COPD, stroke and life expectancy. At least one plausible mechanism could also lead to increased COPD mortality at high altitude, observed in our results and also previous studies13 38–40: namely, even modestly lower ambient oxygen levels in persons with already-impaired breathing and gas exchange may exacerbate hypoxia and pulmonary hypertension and potentiate cor pulmonale as a cause of death.11 Less plausibly, there is mixed evidence on whether hypoxia inhibits or stimulates cancer growths.9 41–43
A second explanation for the observed results may be differences in lifestyle risk factors. The data from the Behavioral Risk Factor Surveillance System show that, after correction for the self-report bias,44–46 body mass index, diabetes, smoking and possibly systolic blood pressure, were all inversely associated with altitude (detailed results available from authors by request). While these risk factors may partially explain the observed negative association between altitude and IHD, none would explain the increase in COPD at higher altitudes. Further, the absence of a consistent association with stroke suggests that residual confounding is a less likely explanation of the beneficial effect on IHD because IHD and stroke have in common several of these risk factors. Other hypothesised altitude-dependent environmental conditions include temperature for CVD47 and mining pollutants for COPD.11 48 49 It has also been found that ambient air pollution is associated with county life expectancy in the USA.50
Finally, migration may influence the altitude–mortality association, especially if migration is related to the health effects of altitude—for example, if people with heart and lung conditions leave high-altitude areas.11 16 We adjusted for migration in our analysis, but it was not possible to examine the recent versus long-term effects of living at higher altitudes. If people with incident disease out-migrate from high altitudes long before death, our results might underestimate the harmful effects of altitude on COPD and overestimate the beneficial IHD effects.
Strengths and limitations
This study has a number of innovations and strengths. To our knowledge, this is the first study to examine the relationship of altitude with life expectancy and with mortality from leading chronic diseases using consistent and comparable data and methods. Using data for all counties in the continental USA allowed dose–response relationships to be estimated over a larger elevation range than most previous studies. The large number of deaths in the vital registration data led to robust estimates of death rates and life expectancy. Finally, we used data from the US Census and health surveys to adjust for socio-demographic factors and migration and to investigate the potential role of health risks.
The study also has a number of limitations. First, we did not have any data on how population and deaths are distributed by elevation within each county. Using mean county elevation is equivalent to measuring the true location of residence with random error, which commonly biases the association towards the null (the so-called ‘regression dilution bias’). Therefore, the actual association with altitude may be stronger than seen in our results. Second, data on the cause of death may be affected by regional variability in certification. The relatively broad disease categories used in our analysis mitigate the effect of regional incomparability in cause-of-death assignment. Finally, following previous analyses,17 we used lung cancer as the indicator of accumulated population exposure to smoking. This adjustment for lung cancer in multivariable regressions may have overadjusted, if altitude has a beneficial effect on lung cancer.
Conclusions
Four hundred and seven million people worldwide live at elevations above 1500 m, 12% of whom live above 3000 m (http://ciesin.org/). Our population-based national study identifies relationships between altitude and important causes of death in the USA. It also raises questions regarding the roles of socio-demographic factors, dietary, physiological and behavioural risk factors, other environmental risks and migration, which should be examined using longitudinal or linked data at the individual level (eg, using prospective cohorts such as the American Cancer Society Cancer Prevention Study or National Health and Nutrition Examination Survey follow-up) and studies in other populations.
What is already known on this subject
There is a large variation in life expectancy across counties in the USA.
Some studies have found an association between altitude and mortality from specific chronic diseases, including cancers, and cardiovascular and respiratory causes.
Previous studies had not examined net effects on life expectancy and on different diseases using consistent methods. Few previous studies were national or covered large elevation ranges to assess dose–response relationships.
What this study adds
This study examined the association between altitude and life expectancy in a study of all counties in the continental USA, adjusting for important potential confounders. It also examined how associations between altitude and mortality from major chronic diseases help explain the relationship with life expectancy.
Our study suggests that living at a higher altitude may have a protective effect on IHD and a harmful effect on COPD; it appears to have no net effect on life expectancy, possibly partly due to these opposing effects.
Acknowledgments
G Danaei and J Marcus provided valuable advice on statistical analysis. D Mozaffarian, E Rimm and G Roth provided valuable comments on the mechanisms for observed effects.
References
Footnotes
Funding The study was supported by the Centers for Disease Control and Prevention (CDC) through the Association of Schools of Public Health (ASPH) (Grant No U36/CCU300430-23), the Altitude Research Center and MRC-HPA Centre for Environmental and Health, Imperial College London.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.