Abstract
A previous study compared cancer mortality in the six lowest versus six highest elevations in the U.S. for all races. This study looks at a single race since death rates tend to vary by race. In this ecological study, cancer mortality rates were compared between low and high states for a race that had sufficient number of counties reporting mortality data, that is, the white race. The average cancer mortality rate for low elevation counties was 73.47 + 18.35 compared to 53.90 + 13.76 for high elevation counties, a difference that was statistically significant (p < 0.0001), with a very large effect size (of 1.2). Higher elevation counties showed less cancer mortality rates for a single race compared to lower elevation counties, suggesting the presence of radiation hormesis. Further rigorous research is indicated to verify or refute these findings.
INTRODUCTION
It is well known that levels of natural background radiation (NBR) increase with increasing land elevations (US NRC, 2009a). A previous study on this topic assessed cancer mortality in regard to all races for the six lowest versus six highest land elevations at the state level and found less cancer mortality in higher elevations (Hart, 2010). Because cancer mortality rates tend to be different between races (Albano et al, 2007), it seems helpful to study a single race when looking for differences in possible effects from a variable such as NBR. As an example of cancer mortality differences between races, Black and Hispanic women have been found to experience higher age-adjusted cervical cancer mortality (Selvin and Brett, 2003). Thus, if there is a greater percentage of Black and Hispanic women in, say, low elevation locales, then analysis of low versus high elevation could show a higher cancer mortality rate in low elevation locations, suggesting, perhaps erroneously, that radiation hormesis is involved if racial differences are not taken into account. Consequently, it behooves the researcher to analyze a single race in to further investigate the possible presence of radiation hormesis in low versus high land elevations. To this end, the present study is different from the previous similar study (Hart, 2010) on two important points: 1) the present study looks at data at the county level (instead of at the state level), thereby providing for a much larger sample size and 2) the present study looks at a single race rather than all races.
METHODS
The response variable in this study consisted of archived data from the National Cancer Institute (NCI) databases for age-adjusted cancer mortality rates, all sites cancer, age < 65, both genders, 2002–2006 by county (NCI, 2010) for the six lowest and six highest elevation jurisdictions in the U.S., including the District of Columbia (referred to now as “states). States were selected based on their mean elevations (USGS, 2005) and as previously detailed (Hart, 2010). Low elevation states consisted of Delaware, District of Columbia, Florida, Louisiana, Mississippi, and Rhode Island while high elevation states consisted of Colorado, Montana, New Mexico, South Dakota, Utah, and Wyoming. The elevation range for the low elevation states was 345 feet above sea level (Rhode Island) to 812 feet above sea level (Florida). The elevation range for the high elevation states was 966 feet above sea level (South Dakota) to 3315 feet above sea level (Colorado). Consequently, there was no overlap of land elevation between the two elevation categories. There were 210 counties having reportable data in the low elevation states and 171 counties in the high elevation states. The estimated levels of NBR by elevation were estimated to range from 51 to 74 mrem (mean = 62.5 + 12.6 standard deviations) for the low elevation states compared to 74 to 81 mrem (mean = 78.5 + 2.9 standard deviations) for high elevation states (NRC, 2009b).
If there were too few deaths for reliable statistical reporting for a given county, then NCI did not report any data for that county. The race category “Black including Hispanic” is provided by NCI but there were only six counties reporting data for the six states in the high elevation state category. Since selection of the race for this study was based on obtaining a large sample size, the Caucasian (white) race was selected in order to achieve a satisfactory sample size for this study that seeks to look at a single race. The age < 65 was used to assess mortality rates below the age of life expectancy.
Counties in the six lowest states versus six highest states were compared. Data analysis consisted determining whether the mortality rate was different between low and high elevation categories using: a) a test showing statistical differences using a two tailed alpha of 0.05 and b) effect size (Morgan et al, 2007). Since the cancer mortality data exhibited a non-normal distribution (skew = 2.37), the nonparametric Wilcoxon test was used to determine statistical differences; this was performed in SAS 9.2 (Cary, NC). Effect size was calculated in a spreadsheet with a formula outlined by Morgan et al (2007), using a pooled standard deviation.
Interpretation of effect size values were as follows: Very large = greater than or equal to 1.00; Large = 0.80; Medium = 0.50; Small = 0.20 (Morgan et al, 2007). Statistical power for comparing the mean cancer mortality between low and high elevation states was performed with an online power calculator (Researcher's toolkit, 1995–2009).
RESULTS
Summary data by county is provided in Table 1. The average cancer mortality rate for low elevation counties was 73.47 + 18.35 compared to 53.90 + 13.76 for high elevation counties. This difference was statistically significant (p < 0.0001) with a statistical power of 100% and a very large effect size (of 1.2) (Figure 1).
Mean cancer mortality rates for white persons in low versus high elevation counties.
Cancer mortality rates (“rate”) by county (NCI, 2010). Elevations from U.S. Geological Survey (USGS, 2005).
CONCLUSION
This study showed a statistically significant lower cancer mortality rate in high elevation counties compared to lower elevation counties. This suggests the presence of radiation hormesis. However, since this is an ecological study, causal inferences are less apparent than, say, case-control studies. Future, more rigorous studies will help to either verify or refute these findings.