Abstract
The validity of the linear-no-threshold (LNT) model in radiation regulation remains contested. Although extensive experimental evidence challenges it, epidemiological studies—especially cohort and case-control designs—are still used to justify its application. This article focuses on a key methodological limitation of individual-level epidemiological studies, particularly the reported link between residential radon and lung cancer. Pooled case-control analyses suggest a linear dose-response relationship consistent with the LNT model, while ecological studies show mixed results. Case-control data are often deemed more reliable than ecological studies, but their validity requires reexamination. A central issue is the neglect of radon’s role within the broader context of indoor air pollution. Because more than 90% of radon decay products adhere to airborne particles, measured radon levels can effectively serve as a proxy for indoor particulate matter (PM), which is a complex mixture of harmful compounds. Since PM2.5 is a well-established lung carcinogen, the observed radon-lung cancer association may reflect PM2.5 effects rather than radon itself. This confounding is weaker in ecological studies, which use regional averages less directly tied to individual homes. When experimental evidence, evolutionary reasoning, and ecological findings contradict individual-level studies, it is possible that the latter are inherently flawed by design. Continued reliance on epidemiological studies to uphold the LNT model should therefore be critically reconsidered.
Introduction
The linear-no-threshold (LNT) model—assuming that excess cancer risk increases linearly from the lowest to highest doses, without a threshold—has long served as the foundation of radiation regulation. However, most experimental studies in cells and animals have failed to support this model. Its persistence is due largely to continued reliance on epidemiological evidence. Notably, epidemiological studies have repeatedly reported no elevated cancer risk among populations living in areas with very high natural background radiation, such as Ramsar (Iran), Kerala (India), and Yangjiang (China).1-4 Nevertheless, these findings, which challenge the LNT model, are often dismissed as being derived from low-quality epidemiological studies. 5
Within the “hierarchy of evidence,” individual-level epidemiological designs—such as case-control and cohort studies—are ranked above ecological studies. The distinction lies in their units of observation: individuals versus groups. Individual-level designs typically contain more detailed information on exposures and major confounders, whereas ecological studies rely on group-level exposure data and often cannot adequately control for confounding. Thus, when their findings conflict, individual-level studies are regarded as more credible.
This article argues, however, that individual-level epidemiological studies may suffer from a serious but unrecognized bias—one that is less common in ecological studies—as exemplified by the case of residential radon and lung cancer. Because radon decay products readily attach to airborne particles, 6 measured residential radon exposure in such studies can serve as a surrogate marker for indoor air pollution, which is an established carcinogen. Ignoring this confounding may distort the true link between radon and lung cancer.
Residential Radon and Lung Cancer: Conflicting Epidemiology
Radon is a recognized human carcinogen and the second leading cause of lung cancer after smoking. 7 Its carcinogenicity was clarified by studies of underground miners exposed to high concentrations. 8 To assess risks at lower doses, multiple pooled case-control studies in the general population have been conducted, and they have suggested a linear dose-response relationship consistent with the LNT model.9-11
In contrast, ecological studies have reported mixed or even inverse associations. For example, an analysis of 1601 U.S. counties found an inverse relationship between radon exposure and lung cancer mortality, even after adjusting for smoking. 12 Similar null or inverse findings have been observed in Sweden and South Korea.13,14 Because ecological analyses rely on group-level exposure data, they are prone to ecological fallacy—drawing inferences about individuals from aggregate measures—and are generally considered more vulnerable to confounding and model misspecification. 15 Consequently, when findings conflict, greater weight is typically given to case-control studies.
A Swedish study that directly compared both approaches illustrates this hierarchy: the individual-level analysis showed a positive association, whereas the ecological analysis did not, despite both adjusting for the same covariates, including smoking. 13 The authors concluded that ecological analyses may be misleading in studies of weak associations, even when major confounders are controlled. Nevertheless, this default preference for case-control evidence may be problematic if such studies systematically capture radon as a surrogate for indoor air pollution.
More recently, a large case-control study examined the dose-response relationship between radon and lung cancer using spatial estimates of indoor radon exposure at the residential-area level, employed as a proxy for individual radon exposure. 16 Although designed as an individual-level study, its exposure assessment resembled that of ecological studies. Interestingly, unlike earlier case-control studies based on individual radon measurements, it reported a J-shaped or U-shaped dose-response relationship for radon concentrations below 200 Bq/m3, deviating from the LNT model.
Indoor Air Pollution as a Missing Link
Air pollution, particularly fine particulate matter (PM2.5, aerodynamic diameter <2.5 μm), is recognized as a major global health hazard and a Group 1 human carcinogen. 17 PM2.5 consists of a complex mixture of harmful compounds, including polycyclic aromatic hydrocarbons, persistent organic pollutants, endocrine disruptors, heavy metals, and micro- and nanoplastics.18-22 While research has focused primarily on outdoor PM2.5, substantial exposure occurs indoors. Urban residents spend most of their time indoors, where air quality is influenced both by outdoor infiltration and by indoor sources such as cooking, smoking, heating, cleaning, and consumer product use. 23 As a result, indoor air pollution is more complex and more dynamic in both composition and concentration than outdoor air.24,25
Radon is an important natural contributor to this mixture. Radon decay products readily attach to airborne particles, and in typical indoor environments more than 90% of radon progeny are particle-bound. 6 Thus, measured radon levels in homes reflect not only radon gas but also particle-associated exposures. Individual-level case-control studies commonly adjust for confounders such as age, sex, and smoking, but no epidemiological study—whether at the individual or population level—has accounted for indoor PM exposure when assessing the relationship between residential radon and lung cancer. One study did examine whether radon modified the association between outdoor PM2.5 and mortality. 26 However, because it considered outdoor rather than indoor PM2.5, and overall mortality rather than lung cancer, it cannot be regarded as addressing the role of indoor PM2.5 in radon-related lung cancer.
This limitation is particularly relevant for individual-level case-control studies. In these studies, radon is often measured directly in participants’ homes over extended periods—sometimes with personal samplers—making the measurements highly correlated with indoor PM2.5. Because most radon progeny exist in a particle-bound form, disentangling the effects of radon from those of indoor air pollution may be very difficult, if not impossible, in epidemiological studies.
By contrast, ecological studies typically rely on regional averages, radon maps, or predictive models, which are less directly tied to actual indoor air quality. Moreover, even when a case-control design is used, if the exposure variable is defined in a manner similar to ecological studies, the results may be less affected by indoor air pollution. This may explain why the recent large case-control study, which used spatial residential radon levels as the exposure variable, did not support the LNT model.
16
Therefore, even if residential radon exposure itself were harmless, case-control studies could still yield a spurious positive association with lung cancer due to their inability to adequately separate radon from indoor PM2.5-related exposures (Figure 1). Can Indoor Air Pollution, as an Unmeasured Confounder, Explain the Positive Associations Between Residential Radon Exposure and Lung Cancer Observed in Individual-Level Case-Control Studies?
Can Low-Dose Radon Really Increase Cancer Risk?
The LNT model implies that even current natural background radiation levels may be unsafe and should therefore be reduced. In fact, radon and its progeny are the main sources of natural background radiation, accounting for about half of total human exposure. 6 Although residential radon levels may not fully align with estimates of natural background radiation because radon accumulates in poorly ventilated indoor spaces, individual-level epidemiological studies of residential radon and lung cancer are widely cited as human evidence supporting the LNT model.
This assumption, however, raises questions about biological plausibility, given that life evolved under much higher levels of ionizing radiation than those present today. 27 Experimental studies further demonstrate that removing natural background radiation impairs normal growth and DNA repair capacity in organisms,28,29 suggesting that low-level radiation may be essential for normal biological functioning. In contrast to radon, most synthetic chemicals found in PM2.5—introduced only since the 20th century—are novel to human biology and physiology. From an evolutionary perspective, this raises the possibility that indoor PM2.5 may exert harmful effects independently of radon, while the carcinogenic potential of low-dose residential radon exposure alone remains uncertain.
Conclusion
Despite strong experimental evidence challenging the LNT model, epidemiological findings continue to justify its use in radiation regulation. Yet, as discussed above, individual-level epidemiological studies face important methodological limitations. When well-integrated conclusions that draw on evidence from so-called lower-quality epidemiological studies, evolutionary biology, and experimental research conflict with case-control or cohort studies, it is entirely plausible that the latter are compromised by inherent methodological biases.
