Long-term Radiation-Related Health Effects in a Unique Human Population: Lessons Learned from the Atomic Bomb Survivors of Hiroshima and Nagasaki

The paper reviews long-term radiation-related health effects among survivors of the 1945 atomic bombings of Hiroshima and Nagasaki and their offspring. Following the devastation and the recognition of unusual medical effects (eg, leukemia, cataracts, birth defects), the United States and Japan established the Atomic Bomb Casualty Commission in 1947, later reorganized as the Radiation Effects Research Foundation in 1975, to conduct a long-term, systematic study. The purpose of this review is to describe the study populations and cohorts, summarize risk estimates for cancer and noncancer outcomes derived from more than six decades of follow-up, and highlight the importance of these data for radiation risk assessment and protection standards. The work provides context and guidance for understanding health risks from acute ionizing radiation exposure in humans and has broad implications for occupational, medical, and environmental radiation exposures.

Early postwar studies largely consisted of case reports and case series documenting hematologic abnormalities, cataracts, and congenital effects, with a notable early rigorous genetic study by Neel and Schull. In 1955, the Francis Committee unified the research program, establishing the Life Span Study (LSS) for mortality and cancer incidence, the Adult Health Study (AHS) for biennial clinical examinations, an in utero cohort, and an F1 (offspring) cohort. High-quality cancer registries in Hiroshima and Nagasaki (from 1959) improved diagnostic ascertainment. Over time, successive dosimetry systems (DS86 to DS02) improved individual dose estimates. Extensive mail surveys collected data on potential confounders (eg, smoking, lifestyle). Prior publications established increased risks of leukemia and, subsequently, solid cancers, and suggested associations with noncancer diseases. These studies, along with biodosimetry (chromosome translocations, ESR in teeth), form the evidence base synthesized in this review.

The research comprises several fixed cohorts from Hiroshima and Nagasaki with long-term follow-up: (1) LSS (~120,000 participants including proximally and distally exposed survivors and nonexposed city controls) for mortality (from 1950) and cancer incidence (from 1958); (2) AHS (~20,000, a subset of LSS enriched for higher doses) undergoing biennial clinical exams since 1958, with stored biospecimens enabling laboratory studies; (3) in utero cohort (~3,300 exposed in utero) followed for mortality, cancer incidence, and, for a subset, clinical exams; and (4) F1 cohort (~77,000 offspring of exposed and unexposed parents) followed for mortality and cancer incidence, with a subset (~12,000) clinically examined for multifactorial diseases in 2002–2006. Near-complete follow-up is enabled by Japan’s koseki family registry and regional cancer registries. Individual organ doses (gammas and neutrons) are estimated using the DS02 system based on survivor location, shielding (buildings, terrain), and body self-shielding models (age/orientation-adjusted anthropomorphic phantoms). Random error in individual dose estimates (assumed ~35%) is accounted for statistically. Biodosimetry (stable chromosome translocations; electron spin resonance in tooth enamel) provides external validation in subgroups. Risk modeling uses regression to estimate excess absolute risk (EAR) and excess relative risk (ERR) as functions of dose, age at exposure, attained age, sex, city, and other covariates (eg, smoking). Analyses evaluate dose-response shapes (linear vs non-linear), temporal patterns (latency, persistence), sex differences, and effect modification.

- Leukemia: 315 leukemia deaths through 2002 in LSS; an estimated 98 excess deaths attributable to radiation among those with doses >0.005 Gy (attributable fraction ~45%), rising to ~86% at >1 Gy. Dose-response is nonlinear with upward curvature up to ~3 Gy. Radiation-related leukemia risk was highest in early years after exposure, especially for those exposed young, and declined over time, though elevated myelodysplastic syndrome risks persisted 40–60 years post-exposure. - Solid cancers: Linear dose-response fits well across multiple sites. Significant dose responses for oral cavity, esophagus, stomach, colon, liver, lung, nonmelanoma skin, female breast, ovary, urinary bladder, brain/CNS, and thyroid. No significant associations for rectum, prostate, or malignant melanoma. ERR/Gy for pooled solid cancer incidence ~0.47 at age 70 after exposure at age 30; similar for mortality. Among >100,000 LSS members, 17,448 incident solid cancers were identified; ~853 (11%) are attributable to radiation among those with >0.005 Gy (mean 0.21 Gy). The attributable fraction increases with dose, reaching ~48% at ≥1 Gy. Women have ~50% higher ERR and EAR than men (female:male ratios ~1.6 for ERR and ~1.4 for EAR). ERR declines with attained age, but EAR increases with age, suggesting lifetime persistence of excess risk. A significant linear trend is observed even in the 0–0.15 Gy range. - Site-specific magnitudes: Highest ERRs (>0.8 per Gy) for bladder, female breast, and lung; relatively high (0.5–0.8 per Gy) for brain/CNS, ovary, thyroid, colon, and esophagus. High EARs (public health impact) for female breast, stomach, colon, lung, liver, bladder, and thyroid, reflecting high baseline rates for some cancers in Japan. - Noncancer diseases: Dose-related increases observed for circulatory, digestive, and respiratory diseases. ERR for all noncancer mortality ~0.14, about one-third of that for all solid-cancer mortality; dose-response below ~0.5 Gy remains uncertain. Heart disease mortality shows an approximately linear dose response (ERR ~0.14), though risks below ~0.5 Gy are less well defined. AHS findings include dose-related elevations in blood pressure, total cholesterol and triglycerides, decreased HDL, fatty liver, aortic arch calcification, hypertension, and inflammatory markers (e.g., at 1 Gy, C-reactive protein ~+28%, interleukin-6 ~+9%). - Ocular effects: Posterior subcapsular lens opacities are characteristic of radiation exposure. Recent data on cataract surgeries indicate ~39% excess risk at 1 Gy, with a best-estimate threshold ~0.1 Gy (upper bound 0.8 Gy), challenging earlier assumptions of thresholds ≥2 Gy for lens opacities and ≥5 Gy for vision-impairing cataracts. - Thyroid and parathyroid: Thyroid nodules show a significant linear dose response with excess odds ratio per Gy ~2.01. Hyperparathyroidism prevalence increases with dose, with ERR ~3.1 at 1 Gy. - Immune system and cytogenetics: Stable chromosomal aberrations persist decades after exposure with a significant, nonlinear dose response. Radiation-related long-term immunological alterations include reduced CD4 helper and naïve T-cell proportions, increased B-cell proportions, impaired T-cell function, and elevated inflammatory cytokines, suggesting accelerated immunosenescence and links to atherosclerosis. - Life span: Median life expectancy decreased with dose (~1.3 years per Gy), with median loss ~2 months for <1 Gy and ~2.6 years for >1 Gy; at 1 Gy, roughly 60% of life lost attributable to solid cancer, 30% to noncancer diseases, 10% to leukemia. - In utero exposure: Marked susceptibility windows for neurodevelopmental effects—severe mental retardation risk increased for exposures at 8–15 weeks and 16–25 weeks postconception, with an estimated prevalence ~40% at 1 Gy for 8–15 weeks and a threshold >0.3 Gy. IQ decreased by ~25 points per Gy for exposures at 8–15 weeks. Cancer risk after in utero exposure shows significant dose dependence but is nominally smaller than for childhood exposure, and EAR did not increase with age as in childhood-exposed; further follow-up into older ages is needed. - F1 (offspring) of survivors: Large-scale studies found no evidence that parental radiation dose increased risks of congenital malformations, still births, perinatal deaths, cytogenetic abnormalities, protein variants, cancer incidence, noncancer mortality, or common adult-onset multifactorial diseases (e.g., hypertension, diabetes, hypercholesterolemia, ischemic heart disease, stroke) in midlife. To date, no discernible heritable mutation signal has been detected, though sensitive genomic technologies are being applied. - Dosimetry and validation: DS02 provides organ dose estimates for 15 organs; uncertainties from location and shielding are handled statistically (assumed ~35% random error). Biodosimetry (lymphocyte translocations, tooth enamel ESR) broadly corroborates physical dose estimates.

The long-term, well-characterized follow-up of Hiroshima and Nagasaki survivors provides uniquely robust human data on the health effects of acute, whole-body ionizing radiation across a wide dose range, both sexes, and all ages at exposure. The findings establish a largely linear dose-response for solid cancers down to low acute doses, with clear modification by sex, age at exposure, and attained age, thereby directly informing radiological protection models and risk coefficients used by international bodies. The early peaking and later decline of leukemia risk contrasts with the persistent, age-amplifying excess absolute risk of solid cancers, underscoring differing temporal dynamics and biological mechanisms. Demonstration of radiation-associated risks for noncancer outcomes, particularly circulatory diseases and lens opacities at lower-than-previously-assumed doses, expands the scope of radiation health effects and suggests vascular, inflammatory, and metabolic pathways as plausible mechanisms. Neurodevelopmental sensitivity to in utero exposure highlights critical gestational windows with substantial cognitive impacts. The absence of detectable heritable genetic or disease risks in the F1 generation to date reduces concerns about large transgenerational effects at the exposure levels experienced, while motivating continued surveillance and application of more sensitive genomic assays. Overall, these results address the central questions about long-term cancer and noncancer risks from acute radiation, provide quantitative risk estimates for policy and clinical guidance, and emphasize the need for ongoing follow-up to refine lifetime risk projections, particularly for those exposed at younger ages.

This review synthesizes more than six decades of research by ABCC/RERF, yielding the most comprehensive human-based risk estimates for cancer and noncancer outcomes following acute ionizing radiation exposure. Key contributions include demonstration of an approximately linear solid-cancer dose response to low acute doses, quantification of leukemia and solid-cancer risks by age, sex, and time since exposure, identification of radiation-related noncancer diseases and lens opacities at lower thresholds than previously assumed, delineation of neurodevelopmental risks from in utero exposure, and lack of detectable heritable disease effects in the F1 generation to date. These findings underpin international radiation protection standards and clinical risk assessment. Future directions include continued follow-up of survivors and offspring as they enter disease-prone ages to resolve lifetime risk uncertainties, refinement of low-dose risk estimates, mechanistic studies leveraging extensive biospecimens, and application of advanced genomic technologies to assess potential heritable mutations.

Key limitations include: (1) uncertainty in individual dose estimates due to survivor location and shielding histories, addressed statistically but still contributing random error; (2) limited representation and unknown doses among the most proximal survivors who survived due to massive shielding; (3) lack of LSS observation during 1945–1950, precluding direct cohort-based risk assessment for the earliest post-bomb period; (4) potential residual confounding (e.g., smoking, lifestyle) despite extensive data collection and adjustment; (5) uncertainty in dose-response at low doses (<~0.5 Gy), particularly for noncancer outcomes; (6) reliance on mortality data for many noncancer diseases due to absence of comprehensive incidence registries; and (7) generalizability primarily to acute, external, whole-body exposures, which may differ from chronic or partial-body medical/occupational exposures.