A landmark contribution to radiation and cancer epidemiology—and a call for revisiting linear no-threshold assumption

We are grateful to the editors of the Annals of Cancer Epidemiology for the opportunity to comment on the recent article by Richardson et al. (1), “Site-specific cancer mortality after low-level exposure to ionizing radiation: findings from an update of the International Nuclear Workers Study (INWORKS).” This large, international cohort study represents a major contribution to the science of low-dose ionizing radiation and solid cancer mortality, and we commend the authors for their rigorous approach and thoughtful discussion.

Note on units: in this commentary, dose values are reported in mrem, mSv, and mGy, as used in the respective studies. For low linear energy transfer (low-LET) radiation such as X-radiations (X-rays) and gamma-rays (γ-rays), 1 mSv ≈1 mGy, given a radiation weighting factor of 1. The following conversions apply: 1 Sv =100 rem, and therefore 1 mSv =100 mrem.

Advancing radiation risk estimation through cohort expansion and hierarchical modeling

Richardson et al. have extended the INWORKS cohort to include over 10.7 million person-years of follow-up among 309,932 workers across France, the United Kingdom, and the United States. Their analysis, combining traditional maximum likelihood estimation with Markov chain Monte Carlo method, stabilizes risk estimates across cancer sites, particularly for less-common outcomes where statistical power is often limited.

The study offers site-specific excess relative rate (ERR) estimates per gray (Gy) for a wide array of solid tumors, updating prior work with improved precision and scope. Notably, the positive association between low-dose ionizing radiation and lung, colon, and prostate cancer mortality adds weight to the body of evidence supporting radiation protection standards grounded in the linear no-threshold (LNT) model. The robustness of findings across multiple sensitivity analyses—including restrictions on internal contamination, neutron exposure, and early hires—further strengthens the credibility of the conclusions.

Methodologically, the use of Markov chain Monte Carlo-based hierarchical regression represents an important innovation. This approach balances statistical stability with the heterogeneity of cancer biology, offering more interpretable and reliable site-specific risk estimates than traditional maximum likelihood models alone.

Recognizing the role of dose timing and time-dependent modeling

Radiation epidemiology consensus holds that dose should be calculated in a time-dependent manner whenever possible. Although Richardson et al. (1) did not explicitly state this, their tabulation of person-years by attained age and use of multiple lag periods indicate that dose accumulation was implemented time-dependently. Their sensitivity analyses across different lag structures further strengthen the robustness of their approach.

Time-dependent dose models allow exposure to be modeled as it accumulates over time. This enables differentiation between recent vs. remote exposures, evaluation of latency effects, and exploration of dose-time interactions—issues of increasing relevance in both scientific research and radiation protection policy. These methods are particularly well-suited for exploring nonlinearities and low-dose thresholds, which remain areas of active scientific and regulatory interest.

Using different lag times for radiation doses in epidemiological studies of cancer is a critical methodological approach that provides a more nuanced understanding of the disease’s biology and a radiation exposure’s effect. It moves beyond simple assumptions of a fixed latency period by allowing researchers to explore how the risk of cancer evolves over time following exposure.

Our study: a complementary perspective from U.S. shipyard workers

We recently published a study titled “Low Dose Radiation and Solid Tumors Mortality Risk” (2), which examined 437,937 workers, including 153,930 radiation workers, employed between 1957 and 2004 at eight U.S. shipyards. As of 2011, 35,646 solid tumor deaths were observed. Using time-dependent cumulative dose models within a Poisson regression framework, we evaluated associations between annual cumulative radiation exposure (lagged 5 years) and solid tumor mortality. The result for lagged 10 years was provided as a supplemental material.

In a categorical dose-response analysis restricted to radiation workers, we observed that the >0–<25 mSv dose group had a significantly lower relative risk (RR) of solid tumor mortality [RR =0.95; 95% confidence interval (CI): 0.91–0.99] compared to the 0 mSv group (Fig. 1) (2). While this counterintuitive finding must be interpreted cautiously and requires further validation, it raises important questions about dose-response nonlinearity and may suggest conditions under which the LNT assumption could be reconsidered.

Importantly, the use of time-dependent exposure metrics enabled us to examine dose-time dynamics with greater granularity—capturing potential risk patterns that may be obscured when using fixed, lifetime cumulative dose measures.

Although Richardson et al. did not cite our 2024 study, we view both investigations as complementary and mutually reinforcing. The INWORKS update offers unparalleled statistical power and broad international coverage, while our work of a single industry benefited from greater internal consistency in radiation type and exposure conditions, avoiding complexities that have limited comparability in other settings. Notably, Richardson et al. also observed a nonlinear dose-response pattern for lung cancer that is strikingly similar to ours (see INWORKS Fig. 1A) (1). In their analysis, the ERR for lung cancer drops below 1 at the second cumulative dose category, then plateaus around 200 mGy, and subsequently declines. This mirrors the lung cancer RR pattern in our Table 3, as well as the solid tumors (all cancers less lymphohematopoietic cancers) pattern in our Fig. 1 (2). The convergence of these independent datasets strengthens the observation that, in the 0–100 mSv low cumulative dose region, the dose-response relationship may deviate from strict linearity.

Our discussion emphasizes the concordance of findings between Tao et al. (2) and Richardson et al. (1). While Richardson et al. [2025] focused on site-specific cancers, their results align with Tao et al. (2), and the convergence across studies strengthens confidence in the observed associations.

Moving forward: revisiting LNT assumption and other challenges

We see the current INWORKS study as a landmark achievement that significantly deepens our understanding of radiation-related solid cancer mortality. Its size, methodological rigor, and transparency make it a valuable resource for researchers and regulators alike. At the same time, we encourage future pooled analyses and cohort extensions to carefully evaluate the shape of dose-response relationships within the ≤100 mSv range to determine whether the LNT model continues to hold.

The need to revisit and refine radiation dose-response models has also been emphasized in a recent technical review: “Reevaluation of Radiation Protection Standards for Workers and the Public” by the Idaho National Laboratory (3). The report indicates that “epidemiological studies have consistently failed to demonstrate statistically significant adverse health effects at doses below 10,000 mrem” (100 mSv or mGy) and “multiple major professional organizations acknowledge significant limitations and uncertainties in the LNT model at low doses”. The report notes that while LNT remains the default framework for radiation protection, emerging epidemiological and mechanistic evidence points to potential deviations from strict linearity, especially in the low-dose range most relevant to occupational and environmental exposures. The report suggests a strong need in further investigation into the applicability of the LNT model at low doses and dose rates, as well as research into biologically based, dose-time integrated models that incorporate latency, repair, and adaptive responses. Aligning with these evidence-based findings, we see value in extending large-scale epidemiologic studies—such as INWORKS or our shipyard cohort—to further integrate time-dependent exposure metrics and dose-time interaction terms. Such efforts could improve risk estimation accuracy, better inform protective standards, and address current uncertainties in the extrapolation of risk from higher to lower doses.

Moving forward, methodological improvements are needed when evaluating low-dose radiation and cancer risk. For example, most occupational studies—including INWORKS and our US shipyard cohort—do not capture non-work-related exposures such as medical radiation. Expanding data sources to include these exposures would strengthen future research. Another priority is the use of cancer incidence rather than mortality as the primary endpoint. Incidence-based data would improve dose-response analyses, especially given the substantial improvements in survival for many solid tumors in recent decades. Unfortunately, most occupational radiation studies, including INWORKS and our shipyard cohort, are restricted to mortality outcomes. Interpretation of site-specific risks remains complicated by unmeasured confounding, particularly smoking in lung and pleural cancer analyses. This underscores a continuing methodological challenge for occupational studies.

Acknowledgments

We thank the authors for their substantial contribution to the field and the editors of ACE for the opportunity to reflect on the significance and future direction of radiation epidemiology research.

Footnote

Provenance and Peer Review: This article was commissioned by the Editorial Office, Annals of Cancer Epidemiology. The article has undergone external peer review.

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://ace.amegroups.com/article/view/10.21037/ace-2025-6/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Richardson DB, Laurier D, Leuraud K, et al. Site-specific cancer mortality after low-level exposure to ionizing radiation: findings from an update of the International Nuclear Workers Study (INWORKS). Am J Epidemiol 2025;194:1285-94. [Crossref] [PubMed]
2. Tao XG, Curriero FC, Mahesh M. Low Dose Radiation and Solid Tumors Mortality Risk. J Occup Environ Med 2024;66:e230-7. [Crossref] [PubMed]
3. Wagner J, Kanter S, Case R, et al. Reevaluation of Radiation Protection Standards for Workers and the Public Based on Current Scientific Evidence. Idaho National Laboratory 2025. Available online: https://inl.gov/content/uploads/2023/07/INLRPT-25-85463_Reevaluation-of-Radiation-Protection-Standards-R0-Final.pdf