Understanding the Debate

T S Krishnan (tskrishnan@iimnagpur.ac.in) is visiting faculty, Production and Operations Management, Indian Institute of Management Nagpur. Annapureddy Rama Papi Reddy (papi@iimnagpur.ac.in) teaches at the Production and Operations Management, IIM Nagpur. M V Ramana (m.v.ramana@ubc.ca) teaches at the School of Public Policy and Global Affairs, University of British Columbia.

High natural background radiation is a constant presence in the lives of those inhabiting some coastal regions of Kerala and Tamil Nadu. While there is agreement about the existence of radiation, some studies claim that it has no impact on the health of the population, while others disagree. There is a need to examine these findings critically, because of implications for public health, and to understand some of the technical reasons for why some papers appear to find no support for evidence (lack of statistical significance) of impact on health due to high levels of background radiation.

The authors would like to thank Saparya Suresh for her insightful comments on statistical significance in developing this article; Prabasaj Paul and Jan Beyea for reviewing earlier drafts and offering useful and critical comments; and the anonymous reviewer for providing suggestions to sharpen the manuscript.

There has been a long-standing debate about the impact of nuclear radiation on human health (Beyea 2012). This debate becomes most contentious when dealing with impact of severe accidents like the ones at Fukushima in 2011 and Chernobyl in 1986. In the case of Chernobyl, for example, the estimates of the number of deaths resulting from the accident range from a few tens to nearly a million (Ramana 2006). The official Soviet Union death count for years was 31 (Marples 1988: 31–36), and this number was repeated by many, for example, in India’s annual conference on radiation protection in 1989 (Nagaratnam 1989).

Although there have been no severe accidents like Chernobyl and Fukushima in India, there has been a debate about the impact of radiation on human health. This debate is over a more quotidian source of radiation prevalent in some coastal regions in Kerala and Tamil Nadu. The sand on the beaches in these areas contains large quantities of a naturally occurring radioactive material called monazite, and this results in high levels of radiation which is a constant presence in the lives of the inhabitants of these regions. There is agreement about the occurrence of the high levels of radiation. The debate has been about whether these high levels of radiation have resulted in any observable impact on health, and this question has resulted in many academic papers and articles over the decades. Some of these studies have concluded that there are no increased risks while others disagree. In this article, we discuss some of the technical reasons for why some papers appear to find no support for evidence (lack of statistical significance) of impact on health due to high levels of background radiation.

Although these discussions might seem esoteric, understanding these intricacies is important because of their relevance to the regulation of radiation and public health. The Indian government is planning an expansion of nuclear power along with the necessary fuel cycle facilities, including uranium mines and mills and reprocessing plants. If this large-scale construction goes ahead, it will necessarily increase exposure of the public to nuclear radiation. The corollary of the arguments we lay out in this article is that any increase in radiation dose will result in a detrimental impact on the health of the exposed populations.

The Case of Kerala

Some coastal regions in Kerala are among the regions around the world that have been termed high natural background radiation (HNBR) areas. Scholars have studied the health of populations living in HNBR areas in other countries such as Brazil (Barcinski et al 1975), China (Hayata et al 2004; Zhou et al 2005), Iran (Ghiassi-Nejad et al 2004) and, of course, Kerala (Nair et al 1999; Binu et al 2005; Jaikrishan et al 2013). In the specific case of Kerala, there have been many studies over the decades that have found support for significant impact on health, including Down syndrome and mental retardation (Kochupillai et al 1976), DNA mutations (Forster et al 2002), and heritable anomalies (Padmanabhan et al 2004).

There are also other studies that do not find evidence of significant impact on health, including for cancer incidence (Nair et al 1999; Nair et al 2009) or stillbirths or congenital anomalies (Jaikrishan et al 2013). In recent years, there have been a number of papers that have explored different health metrics in the HNBR areas, and most have come up with results that suggest no significant correlation with radiation levels. Table 1 (p 40) lists a sample set of these papers along with their significant findings.

These are admittedly specialised studies and one might think that debates on these matters are best left to journals of epidemiology and public health. However, some of these studies, or the larger point they make, have been used in newspaper articles with some regularity. Most of the time, the case of the HNBR regions of Kerala is invoked to make the claim that radiation exposure is not a significant threat to health. Table 2 (p 41) lists some such instances in Indian publications and elsewhere, and the context in which these studies are invoked. When such claims move from the realm of journals of epidemiology and public health to popular media outlets, it becomes all the more important to highlight the concerns about interpretation that we lay out here.

We start with a brief overview of two such papers and then identify some problems with their conclusions. One study by Nair et al (2009) examined the impact of HNBR on a cohort of 3,85,103 residents in Karunagappally, of whom 69,958 residents were included in the statistical analysis. The researchers concluded that “the risk of major cancers such as cancers of the oropharynx, digestive tract, lung and breast showed no statistically significant association with high background radiation;” thyroid cancer “showed no significant excess related to high background radiation” and “leukemia or leukemia excluding chronic lymphocytic leukemia was not significantly related to background radiation” (Nair et al 2009: 62). In large sample statistical studies, significance refers to statistical significance as measured by p-value. As a rule of thumb, the effect with a p-value less than 5% is said to be significant and more than 5% is said to be insignificant.

A second study was carried out by the Japanese institute—Central Research Institute of Electric Power Industry—and again involved the inhabitants of Karunagappally area (CRIEPI 2009). Researchers measured the external exposure dose due to natural radiation, and observed the incidence of cancer in six districts (four districts characterised as HNBR and compared it with two other control areas). They concluded that the increase in radiation exposure increased the risk of cancers such as leukemia among the population only slightly, and that this risk was not significant even when the exposure dose increased. The study also found no correlation between total exposure dose, annual dose rate and increased risks for other cancers. Neither of these studies included radiation doses through internal pathways such as food intake or breathing in radioactive dust.

The Problem of Interpretation

Our focus here is not on the methodology (sampling, data collection and analysis) used by these researchers. Rather, we focus on the interpretation of the results and the flaws therein. Specifically, the problem is with the notion of statistical significance. Before going into the problem of interpretation, we would like to highlight the fact that there has been a movement among scholars to rethink the use of a threshold to evaluate statistical significance (Wasserstein and Lazar 2016; Amrhein et al 2019; Wasserstein et al 2019). In an editorial for the journal American Statistician, three scholars compiled 43 papers which they summarised as implying

No p-value can reveal the plausibility, presence, truth, or importance of an association or effect. Therefore, a label of statistical significance does not mean or imply that an association or effect is highly probable, real, true, or important. Nor does a label of statistical non-significance lead to the association or effect being improbable, absent, false, or unimportant. (Wasserstein et al 2019: 2)

Their recommendation: “it is time to stop using the term ‘statistically significant’ entirely” (Wasserstein et al 2019: 2).

We now move on to discussing a different problem with the HNBR studies. This problem has been highlighted by some scholars specialising in practical applications of statistics (McCloskey and Ziliak 1996; Ziliak and McCloskey 2009), who argue that the interpretation of statistical significance is flawed and effects that may not be statistically significant can still be meaningful. In particular, they point out that there are costs associated with both accepting hypotheses that are false and rejecting the hypotheses that are true.

Statistical studies (including the epidemiological ones) result in two types of error: Type 1 error and Type 2 error. In the case that we are discussing, these types of errors are: (i) Type 1 error:
A study concludes that there is evidence of cancers due to natural radiation, but actually, cancers are not caused due to natural radiation. (ii) Type 2 error: A study concludes that there is no evidence of cancers due to natural radiation, but actually, cancers are caused due to natural radiation. A Type 1 error could lead to a government compensating/spending money on the victims, when in reality the source of cancers may be something else (or might just be naturally occurring) and thus may not be easily avoided. Likewise, a Type 2 error could lead to governments promoting activities, such as building more nuclear plants, which will increase human suffering due to higher incidence of cancer that could have been avoided.

The studies of the HNBR regions that we discussed, for example Jaikrishan et al (2013), calculate p-values which only represent the probability of Type 1 error. The study does not say anything about the probability of Type 2 error. For a fixed sample size, when the probability of a Type 1 error goes down, the probability of a Type 2 error goes up and vice versa. For reasons mentioned above, we argue that committing a Type 2 error has more serious consequences than a Type 1 error. Let us take another example which will help us better understand how researchers commit two types of error while accepting or rejecting hypotheses based on sample results.

In 2011, the Supreme Court of the United States (US) unanimously rejected a fast rule of statistical significance in the Matrixx Initiatives, Inc et al v Siracusano et al (2011) case. The case involved a homeopathic medicine called Zicam, a zinc-based common cold remedy produced by Matrixx Initiatives. Some users reported experiencing burning sensations and anosmia—the loss of ability to smell. However, the company claimed that the loss of sense of smell was not important because it was not statistically significant (that is, p-value more than 5%). The company reached this conclusion without weighing the costs of accepting or failing to reject its null hypothesis (the drug does not cause anosmia) which was false, against the costs of rejecting the alternative hypothesis (the drug causes anosmia), which was true (Labaton et al 2010).

If the side effect had not been something as serious as loss of smell, but something minor, such as a slight tingling of the nose which disappeared after a few minutes without causing any harm to the users, then no one would had sued anyone in all likelihood. The sense of smell is obviously of great importance to people and its loss has grave consequences. The bottom line is that it is important to strike a balance between statistical significance and practical importance. This point is also made in slightly different terms by two members of the American Statistical Association: “Statistical significance is not equivalent to scientific, human, or economic significance” (Wasserstein and Lazar 2016: 132). This misrepresentation of statistical significance is the main problem with how these studies of the HNBR regions of Kerala are interpreted, especially when it permeates into the popular media, as documented in Table 2.

Other Problems

There are other problems with some of these HNBR studies. There are good reasons why studies conducted in the HNBR areas, even under the best conditions, do not provide definitive results (Hendry et al 2009). Many of the areas where these studies are conducted lack well-documented health statistics and well-researched interrelationships. This not only leads to an increase in biases, but also makes it difficult for researchers of such studies to make observations that have any statistical significance. Statistical tests like multiple regression are biased towards proving the presence of an effect. The failure to prove the presence of the effect using these tests is not proof enough for the absence of such an effect. The consequences of assuming that an effect is not present when in reality it is present would be the impact on health from low-level radiation on many people living near nuclear facilities or in HNBR regions. This idea is often expressed as “absence of evidence is not evidence of absence.”

Similar problems exist in general with studies of the impact on health of low levels of radiation. As scholars Dale Hattis and David Kennedy (1986: 158) have argued: “epidemiological studies are notoriously insensitive in detecting health effects from relatively low levels of exposure.” They cite a comment by David Ozonoff, formerly at Boston University’s School of Public Health: “A good working definition of a catastrophe is an effect so large that even an epidemiological study can detect it” (Hattis and Kennedy 1986: 158). One comprehensive review of studies assessing the cancer risks associated with low-dose radiation concluded that most of the epidemiological studies conducted lacked statistical power and involved various complicated confounding factors (Ogata 2011). More generally, J M Samet (2011) identified the need to conduct multidisciplinary studies that involved not just epidemiologists, but also other professionals such as biostatisticians, health physicists, radiation biologists and risk assessors.

Conclusions

We return to the question of what explains the pattern of difference in the results of various studies on the impact on health in HNBR regions of Kerala. Our primary observation is that those studies showing null results are based on the p-value (or the statistical insignificance), which then has been interpreted as showing that the impact on health from high levels of background radiation is negligible. Put more simply, they ask whether we can have a high degree of confidence that cancers are a result of radiation in these HNBR areas. They do not ask whether we can be confident that radiation did not cause cancer. But when interpreting the results, they imply the latter without doing the statistical analysis needed to prove that assertion. Taking this interpretation seriously could have negative effects and should not be the basis for good policy formulation.

Globally, the recommendation of many scientific bodies is to base regulatory policy on the assumption that the biological risk from radiation exposure is a linear function of radiation dose at low doses without any threshold, what is often called the linear no threshold (LNT) model. The US National Research Council’s Committee to Assess Health Risks from Exposure to Low Levels of Ionizing Radiation (Biologic Effects of Ionizing Radiation Committee), for example, stated that based on a “comprehensive review of the biology data” that they had undertaken, they had concluded that “the risk would continue in a linear fashion at lower doses without a threshold and that the smallest dose has the potential to cause a small increase in risk to humans” (National Research Council 2006: 7).

A more recent review of 29 epidemiologic studies of “total solid cancer, leukemia, breast cancer, and thyroid cancer, as well as heritable effects and a few non-malignant conditions” carried out by an international author group that can be described as a “who’s who” of radiation epidemiology researchers underscores this point (Shore et al 2019: 235). That review concluded that although

the possible risks from very low doses of low linear-energy-transfer radiation are small and uncertain and … it may never be possible to prove or disprove the validity of the linear no-threshold assumption by epidemiologic means. Nonetheless, the preponderance of recent epidemiologic data on solid cancer is supportive of the continued use of the linear no-threshold model for the purposes of radiation protection. This conclusion is in accord with judgments by other national and international scientific committees, based on somewhat older data. Currently, no alternative dose-response relationship appears more pragmatic or prudent for radiation protection purposes than the linear no-threshold model. (Shore et al 2019: 235)

The implication of the LNT model is that it is prudent to make policies that assume that even at radiation dose levels that are much lower than those in the HNBR areas of Kerala, there would be an impact on health. As some of the articles listed in Table 2 exemplify, some studies about the Kerala HNBR areas are routinely used to make claims about the harmless or benign nature of low-level radiation. This is not justified. In contrast, if one goes by the LNT model, then any expansion of nuclear power or other nuclear activities will result in deleterious effects on public health.

References

Amrhein, V, S Greenland and B McShane (2019): “Scientists Rise Up against Statistical Significance,” Nature, Vol 567, pp 305–07, https://www.nature.com/articles/d41586-019-00857-9.

Anam, P (2019): “There Is No Escaping Radiation Anywhere, No Matter the Location,” DNA, 24 April, https://www.dnaindia.com/analysis/column-there-is-no-escaping-radiation-anywhere-no-matter-the-location-2742603#.

Barcinski, M A, M D C Abreu, J C C De Almeida, J M Naya, L G Fonseca and L E Castro (1975): “Cytogenetic Investigation in a Brazilian Population Living in an Area of High Natural Radioactivity,” American Journal of Human Genetics, Vol 27, No 6, pp 802–06.

Beyea, Jan (2012): “The Scientific Jigsaw Puzzle: Fitting the Pieces of the Low-level Radiation Debate,” Bulletin of the Atomic Scientists, Vol 68, No 3, pp 13–28.

Bhaduri, Amita (2016): “Mines Radiate Disaster,” India Water Portal, 4 August, https://www.indiawaterportal.org/articles/mines-radiate-disaster.

Binu, V S, P Gangadharan, P Jayalekshmi, R R K Nair, M K Nair, B Rajan et al (2005): “The Risk of Lung Cancer in HBR Area in India — A Case Control Study,” High Levels of Natural Radiation and Radon Areas: Radiation Dose and Health Effects, T Sugahara and Y Sasaki (eds), Amsterdam: Elsevier, pp 236–37.

Cheriyan, V D, C J Kurien, B Das, E N Ramachandran, C V Karuppasamy, M V Thampi et al (1999): “Genetic Monitoring of the Human Population from High-level Natural Radiation Areas of Kerala on the Southwest Coast of India. II. Incidence of Numerical and Structural Chromosomal Aberrations in the Lymphocytes of Newborns,” Radiation Research, Vol 152, No 6s, pp S154–S158.

CRIEPI (2009): “Trying to Elucidate the Effects of Low Dose Radiation—From the Results of Epidemiological Studies on High Background Radiation Areas,” Central Research Institute of Electric Power Industry, https://criepi.denken.or.jp/research/news/pdf/den451e.pdf.

Das, Birajalaxmi and C V Karuppasamy (2009): “Spontaneous Frequency of Micronuclei among the Newborns from High Level Natural Radiation Areas of Kerala in the Southwest Coast of India,” International Journal of Radiation Biology, Vol 85, No 3, pp 272–80.

Das, Birajalaxmi, D Saini and M Seshadri (2009): “Telomere Length in Human Adults and High Level Natural Background Radiation,” PloS ONE, Vol 4, No 12, e8440.

Economic Times (2017): “Is Kerala Experiencing an Over-diagnosis of Thyroid Cancer?” 16 April, https://economictimes.indiatimes.com/industry/healthcare/biotech/healthcare/is-kerala-experiencing-an-over-diagnosis-of-thyroid-cancer/articleshow/58204624.cms?from=mdr.

Forster, L, F Peter, S Lutz-Bonengel, H Willkomm and B Brinkmann (2002): “Natural Radioactivity and Human Mitochondrial DNA Mutations,” Proceedings of the National Academy of Sciences of the United States of America, Vol 99, No 21, pp 13, 950–54.

Ghiassi-Nejad, M, F Zakeri, R G Assaei and A Kariminia (2004): “Long-term Immune and Cytogenetic Effects of High Level Natural Radiation on Ramsar Inhabitants in Iran,” Journal of Environmental Radioactivity, Vol 82, Nos 1–3, pp 107–16.

Hattis, Dale and David Kennedy (1986): “Assessing Risks from Health Hazards: An Imperfect Science,” Technology Review, Vol 89, No 4, pp 60–71.

Hayata, I, C Wang, W Zhang, D Chen, M Miamihisamatsu, H Morishima et al (2004): “Effect of High-level Natural Radiation on Chromosomes of Residents in Southern China,” Cytogenetic and Genome Research, Vol 104, Nos 1–4, pp 237–39.

Hendry, J H, S L Simon, A Wojcik, M Sohrabi, W Burkart, E Cardis et al (2009): “Human Exposure to High Natural Background Radiation: What Can It Teach Us about Radiation Risks?” Journal of Radiological Protection, Vol 29, No 2A, pp A29–A42.

IOL (2013): “RBM to Bury Its Radioactive Headache,” Independent Online, 13 February, https://www.iol.co.za/news/rbm-to-bury-its-radioactive-headache-1469648.

Jaikrishan, G, V J Andrews, M V Thampi, P K M Koya, V K Rajan and P S Chauhan (1999): “Genetic Monitoring of the Human Population from High-level Natural Radiation Areas of Kerala on the Southwest Coast of India. I. Prevalence of Congenital Malformations in Newborns,” Radiation Research, Vol 152, No 6s, pp S149–S153.

Jaikrishan, G, K R Sudheer, V J Andrews, P K M Koya, M Madhusoodhanan, C K Jagadeesan et al (2013): “Study of Stillbirth and Major Congenital Anomaly among Newborns in the High-level Natural Radiation Areas of Kerala, India,” Journal of Community Genetics, Vol 4, No 1, pp 21–31.

Kochupillai, N, I C Verma, M S Grewal and V Ramalingaswami (1976): “Down’s Syndrome and Related Abnormalities in an Area of High Background Radiation in Coastal Kerala,” Nature, Vol 262, No 5563, pp 60–61.

Labaton, E, I Schochet and C J McDonald (2010): “Brief of Amici Curiae Statistics Experts Professors Deirdre N McCloskey and Stephen T Ziliak in Support of Respondents,” No 09-1156, Supreme Court of the United States, Washington, DC, http://www.deirdremccloskey.com/docs/amicus.pdf.

Marples, David R (1988): The Social Impact of the Chernobyl Disaster, Edmonton: University of Alberta Press.

Matrixx Initiatives, Inc, et al v Siracusano et al (2011): Supreme Court of the United States, Certiorari to the United States Court of Appeals for the Ninth Circuit, No 091156, https://www.supremecourt.gov/opinions/10pdf/09-1156.pdf.

McCloskey, D N and S T Ziliak (1996): “The Standard Error of Regressions,” Journal of Economic Literature, Vol 34, No 1, pp 97–114.

Nagaratnam, A (1989): “Bhopal and Chernobyl: Some Comparisons,” Bulletin of Radiation Protection, Vol 12, Nos 1–2, pp 25–33.

Nair, M K, K S V Nambi, N S Amma, P Gangadharan, P Jayalekshmi, S Jayadevan et al (1999): “Population Study in the High Natural Background Radiation Area in Kerala, India,” Radiation Research, Vol 152, No 6s, pp S145–S148.

Nair, R R K, B Rajan, S Akiba, P Jayalekshmi, M K Nair, P Gangadharan et al (2009): “Background Radiation and Cancer Incidence in Kerala, India-Karanagappally Cohort Study,” Health Physics, Vol 96, No 1, pp 55–66.

National Research Council (2006): Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII, Phase 2, Washington, DC: National Academies Press.

Noack, M (2014): “Radioactive Sands Known for 50 Years,” Half Moon Bay Review, 16 January, https://www.hmbreview.com/news/radioactive-sands-known-for-50-years/article_1c513cde-7ed7-11e3-8c62-001a4bcf887a.html.

Ogata, H (2011): “A Review of Some Epidemiological Studies on Cancer Risk from Low-dose Radiation or Other Carcinogenic Agents,” Radiation Protection Dosimetry, Vol 146, Nos 1–3, pp 268–71.

Padmanabhan, V T, A P Sugunan, C K Brahmaputhran, K Nandini and K Pavithran (2004): “Heritable Anomalies among the Inhabitants of Regions of Normal and High Background Radiation in Kerala: Results of a Cohort Study, 1988–1994,” International Journal of Health Services, Vol 34, No 3, pp 483–515.

Parthasarathy, K S (2011): “Background Radiation and Radioactivity in India,” Hindu, 5 May, https://www.thehindu.com/sci-tech/agriculture/background-radiation-and-radioactivity-in-india/article1991190.ece#.

— (2012): “No Major Birth Defects Found in High Level Natural Radiation Areas of Kerala,” Hindu, 4 October, https://www.thehindu.com/sci-tech/No-major-birth-defects-found-in-high-level-natural-radiation-areas-of-Kerala/article12543736.ece#.

Ramana, M V (2006): “Twenty Years after Chernobyl: Debates and Lessons,” Economic & Political Weekly, Vol 41, No 18, pp 1743–47.

Samet, J M (2011): “Radiation and Cancer Risk: A Continuing Challenge for Epidemiologists,” Environmental Health, Vol 10, No 1, pp S4.

Shore, R E, Harold L Beck, John D Boice, Emily A Caffrey, Scott Davis, Helen A Grogan et al (2019): “Recent Epidemiologic Studies and the Linear No-threshold Model for Radiation Protection-Considerations Regarding NCRP Commentary 27,” Health Physics, Vol 116, No 2, pp 235–46.

Thampi, M V, V D Cheriyan, G Jaikrishan, B Das, C J Kurien, E N Ramachandran et al (2005): “Investigations on the Health Effects of Human Population Residing in the High-level Natural Radiation Areas in Kerala in the Southwest Coast of India,” International Congress Series, Vol 1276, pp 8–12.

Times of India (2011): “No Radiation Hazard from Metal Plant: Rtd Scientist,” 25 March, https://timesofindia.indiatimes.com/city/mysuru/No-radiation-hazard-from-metal-plant-Rtd-scientist/articleshow/7782465.cms#.

Wasserstein, R L, A L Schirm and N A Lazar (2019): “Moving to a World Beyond ‘p?<?0.05,’” American Statistician, Vol 73, No S1, pp 1–19.

Wasserstein, R L and N A Lazar (2016): “The ASA Statement on p-values: Context, Process, and Purpose,” American Statistician, Vol 70, No 2, pp 129–33.

Zhou, J, Z Tao, Q Sun, S Akiba and Y Zha (2005): “Cancer and Non-cancer Epidemiological Study in the High Background Radiation Area of Yangjiang China,” High Levels of Natural Radiation and Radon Areas: Radiation Dose and Health Effects, T Sugahara, H Morishima, M Sohrabi, Y Sasaki, I Hayata and S Akiba (eds), Amsterdam: Elsevier, pp 97–101.

Ziliak, S T and D N McCloskey (2009): The Cult of Statistical Significance: How the Standard Error Costs Us Jobs, Justice, and Lives, Ann Arbor: University of Michigan Press.

Source: https://epw.in/journal/2020/37/insight/impact-natural-background-radiation-health.html