Cancer Mortality in Six Lowest Versus Six Highest Elevation Jurisdictions in the U.S.

Abstract

Low levels of background radiation exist around us continuously. These levels increase with increasing land elevation, allowing a comparison of low elevations to high elevations in regard to an outcome such as cancer death rates. The present study compares archived cancer mortality rates in six low versus six high elevation jurisdictions. The study also compares mortality rates for all causes, heart disease, and diabetes in low versus high elevation jurisdictions in an effort to see if other mortality outcomes are different in low versus high elevations. Statistically significant decreases in mortality, with very large effect sizes, were observed in high land elevation for three of the four outcomes, including cancer. One possible explanation for the decreased mortality in high elevation jurisdictions is radiation hormesis. Another possible explanation, at least in the case of heart disease mortality, is the physiologic responses that accompany higher elevations regarding decreased oxygen levels. Since this is an ecological study, no causal inferences can be made, particularly when viewpoints on possible effects of low level radiation are diametrically opposed. Further research is indicated.

INTRODUCTION

There is controversy as to whether low level radiation poses a significant health risk. One side of the argument holds to the linear no-threshold (LNT) model, which essentially states that even the smallest amounts of radiation are harmful (Preston, 2008). The other side of the argument states that low level radiation appears to be not only harmless (Nair et al, 1999), but may actually provide a health benefit through hormesis (Allright et al, 1983; Jagger, 1998; Luckey, 2006; Scott and Di Palma, 2006). What is not controversial is that the levels of natural background radiation (NBR) are lower at lower elevations than at higher elevations (NRC, 2009). Jagger (1998), for example, in an apparently non-numeric method of selection of low states, compared low and high elevation states for cancer rates and found lower cancer rates in high elevation states.

The purpose of the present study is to compare cancer mortality rates in low versus high elevations, using a numeric method of jurisdiction selection, and to compare the difference, if any, to other mortality outcomes, one of which (heart disease) is also thought to be related to elevation (altitude) while the other mortality outcomes (all causes and diabetes) seemingly would be less-related to elevation. The null hypothesis is that there will be no difference in mortality rates between low and high elevation jurisdictions while the alternate hypothesis is that at least one of the outcomes will show a statistically significant difference. If there are differences between low and high elevations, then further research would be needed to determine causality of such an association given that the present inquiry is an ecological study.

METHODS

Age-adjusted mortality rates were obtained by jurisdiction for the following four mortality outcomes: 1) all causes, 2) malignant neoplasms (“cancer”), 3) diseases of the heart (“heart disease”), and 4) diabetes mellitus (“diabetes”) for the following five years: 2001 (Arias et al, 2003), 2002 (Kochanek et al, 2004), 2003 (Hoyert et al, 2006), 2004 (Minino et al, 2007), and 2005 (Kung et al, 2008). The outcomes of cancer mortality and heart disease were selected because of they have been linked with elevation (for heart disease, the term “altitude” is typically used in the literature). The other two mortality outcomes (all causes and diabetes) were selected because of their apparent lack of being linked to elevation, in an effort to further compare the cancer mortality rates to.

Jurisdictions were identified using the U.S. Geological Survey data by viewing their table of lowest and highest elevations by jurisdiction (U.S. Geological Survey, 2005). In this table’s lowest elevation column, six jurisdictions having the lowest elevation were selected (“low elevations”). As an example of the selection process, Delaware’s highest elevation is 448 feet above sea level while its lowest elevation is at sea level. Consequently, all people in Delaware were considered to live in areas not exceeding 448 feet above sea level. Six jurisdictions having the highest elevation were selected using similar methodology (“high elevations”). As an example of the selection process of jurisdictions in the high elevation category, Colorado’s lowest elevation is 3315 feet above sea level although its highest point is 14433 feet above sea level. Colorado was therefore considered to have its entire population living at an elevation of at least 3315 feet above sea level. The highest elevation jurisdiction in the low elevation category is Rhode Island, having a maximum elevation of 812 feet above sea level. The lowest elevation jurisdiction in the high elevation category is South Dakota, having a lowest elevation of 966 feet above sea level. The next highest elevation jurisdiction after 966 was Illinois, having a highest elevation of 1235 feet above sea level and a lowest elevation of 279 feet above sea level. Consequently, this state (Illinois) would not have been an appropriate addition to either the low or high elevation category because its highest elevation crosses over into the high elevation category when its lowest elevation (of 279 feet above sea level) would disqualify it from being in the high elevation category.

The annual NBR for all jurisdictions consisted a minimum value of 26 mrem for cosmic radiation at sea level + the additional radiation according to feet above sea level, as follows: up to 1000 feet above sea level= 2 additional mrem; 1000–2000 feet above sea level = 5 additional mrem; 2000–4000 feet above sea level = additional 9 mrem (NRC, 2009). The radiation levels from terrestrial sources (i.e., radon) were added as follows: 23 mrem for states bordering the Gulf coast (Louisiana, Mississippi, and Florida); 46 mrem for all other states (NRC, 2009). Ninety mrem can be assigned to the Colorado Plateau areas but since entire states were assessed, and some areas would not be part of the Colorado Plateau, the addition of the 90 mrem value was omitted. Thus, six jurisdictions in each of the low and high categories were obtained (Table 1). Five years of mortality data were analyzed for each of the 12 jurisdictions for the four mortality outcomes (Table 2). This meant that each mortality outcome in each of the two categories (low versus high elevation) had a sub-total of 30 observations each or a total of 60 observations for each mortality outcome (Table 2). The population for each jurisdiction, for each year, was estimated by the following calculation where total number of case, crude rate per 100,000 persons was known:

$$\frac{\text{Total}\hspace{0.17em}\text{number}\hspace{0.17em}\text{of}\hspace{0.17em}\text{cases}}{\text{Population}}=\frac{\text{Crude}\hspace{0.17em}\text{rate}}{100,000}$$

TABLE 1.

Elevations, means and standard deviations in low versus high jurisdictions.

LOW	HIGHEST ELEVATION
Elevation/jurisdiction	Feet above sea level	Estimated mrem
Delaware	448	74
District of Columbia	410	74
Florida	345	51
Louisiana	535	51
Mississippi	806	51
Rhode Island	812	74
Mean	559.3	62.5
Standard deviation	209.9	12.6
HIGH	LOWEST ELEVATION
Elevation/jurisdiction	Feet above sea level	Estimated mrem
Colorado	3315	81
Montana	1800	77
New Mexico	2842	81
South Dakota	966	74
Utah	2000	77
Wyoming	3099	81
Mean	2337.0	78.5
Standard deviation	902.6	2.9
p-value*	0.004	0.025
Effect size*	2.7	1.8

^*p-value and effect size pertain to comparison of low versus high elevation and NBR.

TABLE 2.

Age-adjusted mortality for four outcomes, five years, 12 jurisdictions (six in each of low and high elevation categories.

		all causes age-adjusted rate	cancer age-adjusted rate	heart age-adjusted rate	diabetes age-adjusted rate	population
	Low elevation
1	Delaware 2001	891.9	219.3	257.2	27.2	796595.0
2	Delaware 2002	838.2	193.8	236.0	25.9	807366.4
3	Delaware 2003	844.4	201.4	243.1	28.2	817530.1
4	Delaware 2004	823.3	207.4	232.3	24.1	830388.3
5	Delaware 2005	830.5	197.8	224.3	25.9	843531.3
6	District of Columbia 2001	1038.2	235.3	308.5	37.0	573811.6
7	District of Columbia 2002	1021.4	230.0	291.7	33.7	570885.0
8	District of Columbia 2003	982.3	199.7	287.3	32.2	564354.4
9	District of Columbia 2004	974.0	207.0	274.9	40.2	553537.0
10	District of Columbia 2005	971.4	206.0	268.2	34.6	550502.0
11	Florida 2001	799.7	187.2	233.2	22.0	16373238.1
12	Florida 2002	786.4	183.4	222.0	21.4	16712877.2
13	Florida 2003	776.0	181.4	212.7	21.8	17018869.8
14	Florida 2004	763.6	179.9	204.9	21.6	17396603.2
15	Florida 2005	749.4	178.1	194.6	22.6	17790729.2
16	Louisiana 2001	1005.9	227.4	280.1	41.7	4470292.3
17	Louisiana 2002	1000.5	222.9	269.6	42.1	4482596.6
18	Louisiana 2003	1004.6	221.9	274.2	40.8	4496263.6
19	Louisiana 2004	986.1	216.7	256.6	39.9	4515939.2
20	Louisiana 2005	1020.8	209.3	255.7	38.7	4523712.4
21	Mississippi 2001	1023.2	216.3	329.0	24.1	2859643.8
22	Mississippi 2002	1036.3	218.3	326.6	24.2	2871802.5
23	Mississippi 2003	1014.0	211.1	310.3	24.1	2881169.1
24	Mississippi 2004	998.2	209.8	300.1	23.6	2902926.8
25	Mississippi 2005	1026.9	208.4	306.8	23.5	2921060.5
26	Rhode Island 2001	806.4	200.6	240.5	21.7	1059638.4
27	Rhode Island 2002	809.5	196.9	239.1	21.2	1069743.2
28	Rhode Island 2003	786.9	190.2	227.7	20.2	1076222.1
29	Rhode Island 2004	741.1	194.1	216.0	21.9	1080641.6
30	Rhode Island 2005	747.3	184.1	213.8	21.6	1076137.2
	mean	903.3	204.5	257.9	28.3	4482953.6
	High elevation
1	Colorado 2001	787.8	169.6	181.0	18.7	4431323.4
2	Colorado 2002	790.2	171.8	178.9	17.7	4506325.2
3	Colorado 2003	784.3	169.2	178.0	19.0	4550586.1
4	Colorado 2004	736.4	160.1	162.7	18.2	4601593.0
5	Colorado 2005	742.8	159.6	162.0	19.2	4664934.7
6	Montana 2001	840.3	198.8	197.9	23.1	905356.6
7	Montana 2002	849.7	190.7	191.6	21.0	909440.8
8	Montana 2003	828.1	180.9	190.7	25.5	917633.0
9	Montana 2004	778.6	180.2	173.6	22.6	926829.3
10	Montana 2005	798.4	184.4	169.4	26.5	935703.3
11	New Mexico 2001	825.4	167.0	203.4	31.4	1830892.8
12	New Mexico 2002	815.0	172.5	193.3	32.9	1855147.4
13	New Mexico 2003	823.8	169.6	191.5	33.0	1874525.2
14	New Mexico 2004	778.9	161.9	180.4	31.7	1903354.6
15	New Mexico 2005	795.0	162.6	184.5	31.2	1928314.0
16	South Dakota 2001	784.8	187.9	218.1	24.2	758352.5
17	South Dakota 2002	771.2	182.4	209.7	22.6	761032.7
18	South Dakota 2003	790.5	189.6	208.0	22.9	764333.9
19	South Dakota 2004	746.4	176.8	187.1	25.2	770870.9
20	South Dakota 2005	757.0	180.9	182.1	25.8	775952.7
21	Utah 2001	776.8	143.4	185.2	31.9	2278567.6
22	Utah 2002	782.0	143.3	185.1	31.4	2316086.9
23	Utah 2003	782.3	144.1	183.5	31.2	2351332.4
24	Utah 2004	760.4	141.2	175.0	28.6	2389068.1
25	Utah 2005	731.2	139.4	162.6	30.4	2469571.6
26	Wyoming 2001	851.7	192.9	209.5	25.5	493750.0
27	Wyoming 2002	864.3	175.1	210.0	30.0	498685.8
28	Wyoming 2003	849.9	188.4	199.5	27.7	501261.6
29	Wyoming 2004	789.7	170.8	191.0	22.1	506531.8
30	Wyoming 2005	801.4	167.7	186.9	25.7	509319.1
	mean	793.8	170.8	187.7	25.9	1829555.9

Analysis consisted of assessing the data as follows: a) skew, performed in a spreadsheet, with values between - 2 and + 2 considered to be not significantly different from normal (Garson, 2009); b) statistical differences between low and high elevation regions for the mortality outcomes, performed in SAS 9.2 (Cary, NC), elevation and NBR (performed in a spreadsheet), and c) effect size, performed in a spreadsheet for low versus high elevation groups for each variable, using a pooled standard deviation (Morgan et al, 2007). Interpretation of effect size ranges 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). Since multiple tests for statistical significance were performed (n = 4, for the four mortality outcomes and n = 2 for elevations and NBR), a stringent Bonferroni-adjusted alpha was applied where the number of tests was divided by the traditional alpha of 0.05 (0.05/4 for outcomes) resulting in an adjusted alpha of 0.0125 (for outcomes) and 0.025 (0.05/2) for land and NBR, below or equal to which, statistical significance was considered to be present.

RESULTS

The data fell between skew values and −2 and + 2. Consequently, a t-test was used (two tails) to assess statistical differences. The mean elevation above sea level for low elevation jurisdictions was 559.3 compared to 2337 feet for high elevation jurisdictions, a statistically significant difference (p = 0.004) with a very large effect size (2.7). The mean annual NBR level for low elevation states was 62.5 mrem compare to 78.5 mrem for high elevation states, a statistically significant difference (p = 0.025; Figure 1), and therefore considered to be not statistical significant, while exhibiting a very large effect size (1.8) (Table 1).

FIGURE 1.

NBR comparison between low and high elevation jurisdictions. The difference between low and high elevation jurisdictions (p = 0.025) was statistically significant along with a very large effect size between the two NBR means (effect size = 1.8).

Mortality rates for the outcomes were as follows. All causes: low elevation = 903.3; high elevation = 793.8, a statistically significant difference (p < 0.0001) with a very large effect size (1.3); heart disease: low elevation = 257.9; high elevation = 187.7, a statistically significant difference (p < 0.0001) with a very large effect size (2.5); Cancer: low elevation = 204.5; high elevation = 170.8, a statistically significant difference (p < 0.0001) with a very large effect size (2.1); Diabetes: low elevation = 28.3; high elevation = 25.9, a statistically insignificant difference (p < 0.1) with a small-medium effect size (0.4) (Figure 2).

FIGURE 2.

Mortality rate means, differences, and effect sizes in low versus high elevation jurisdictions. Statistically significant differences, along with very large effect sizes were found between low and high elevation jurisdictions for all-causes mortality rates (p < 0.0001, effect size = 1.3), heart disease (p < 0.0001, effect size = 2.5), cancer (p < 0.0001, effect size = 2.1). Diabetes mortality was lower in higher elevation jurisdictions but this difference was not statistically significant (p = 0.1) with a small-medium effect size (0.4).

DISCUSSION

Cancer mortality rates were significantly lower in high elevation jurisdictions, which of course consist of higher NBR levels. This suggests the presence of radiation hormesis. However, heart disease mortality (as well as mortality from all causes) was also significantly lower in the high elevation jurisdictions. The finding of lower cancer mortality in the present study is consistent with similar previous studies. Jagger (1998) for example found that cancer death rates were 1.26 times that of Rocky Mountain states. Tao et al (2000) also found that in high NBR areas in China cancer mortality was also lower in high background radiation regions. Conversely, Cardis et al (2005) found a small increased risk of cancer for low dose exposure to nuclear workers.

In high elevations, decreased heart disease mortality may be related to decreased oxygen levels. In high altitudes effects of decreased oxygen can range from being factors in conditions such as pulmonary edema and right cardiac failure with peripheral edema (Ausiello et al, 2004), to being innocuous (Erdmann, 1998) to providing a protective effect against cardiovascular disease and cancer (Weinberg et al, 1987). In addition, this beneficial effect is thought to be related not to NBR but rather, due to thinner oxygen levels at higher altitudes (Weinberg et al, 1987). Mortimer et al (1977) theorize that the benefit of higher altitudes on arteriosclerotic heart disease may relate to the increased levels of exertion needed at higher altitudes for the same work done at lower altitudes, hence more exercise at the higher altitudes compared to lower altitudes. Whether the decreased heart disease mortality is due to increase NBR (i.e., radiation hormesis) or to physiologic responses that accompany decreased oxygen at high altitudes, or some other factor(s), or due to interaction of all of the above, is a question beyond the scope of this paper.

Mortality from all causes is an outcome that perhaps is too general to attempt to suggest possible explanations as to why it would be less in higher elevations. The outcome of diabetes, which was theorized by the author to not be related to elevation, showed the weakest relationship with elevation.

The question may arise as to the possibility of cancer clusters being a factor in the mortality rates. This was considered but no clusters were found for the jurisdictions and years in this study (CDC, 2009).

Ecological studies such as the present one do not allow causal inferences to be made but can be a first step for inquiry into new areas of research (Grimes and Schulz, 2002). Furthermore,there remains a lack of clear understanding in regard to the possible effects of exposure to low level radiation (Hendry et al, 2009). There is certainly a divergence of opinion as to what effects, if any, result from exposure to low level radiation. On the one hand there is the viewpoint that low level radiation provides a protective (hormetic) effect (Allright et al, 1983; Jagger, 1998; Luckey, 2006; Scott and Di Palma, 2006) while on the other hand there is the view that supports the LNT model (Brenner et al, 2003) as reported in BEIR VII (Health Physics, 2005). Clearly there is a need for further research, particularly they type that follows a given population over time such as done by Nair et al (2009), as well as research that assesses confounding variables.

CONCLUSION

Statistically significant decreases, with very large effect sizes were observed in high elevation jurisdictions compared to low elevation jurisdictions for the mortality outcomes of all causes, heart disease, and cancer. Diabetes mortality was also lower in high elevation jurisdictions compared to low elevation jurisdictions but the difference was not statistically significant and exhibited a small-medium effect size. These findings suggest the presence of radiation hormesis, at least in regard to the cancer mortality from a plausibility standpoint (i.e., hormesis theory). However, causal inferences cannot be made since this is an ecological study. Future study should include other variables that are thought to be related to cancer death rates, such as educational attainment, smoking, and diet.