LL1-03 hindsight check: Radiation: early warnings; late effects#

Late lessons from early warnings (EEA, 2001), Chapter 3, pp. 31–37. Author: Barrie Lambert. Check covers 2001 to late September 2026. Prepared 25 September 2026.

How this was done. The web search tool was unavailable in this session. Everything below was retrieved directly instead:

• Literature databases. Europe PMC, Crossref and OpenAlex.
• Official sources. The US Federal Register API, EUR-Lex, the GOV.UK content API, and primary PDFs from the ICRP, UNSCEAR, WHO, the National Academies, UKHSA and the EEA.
• Scheme and agency websites. The CSRLD, US Department of Justice and RERF sites.

Most of the historical scholarship is paywalled, including the books by Walker and Higuchi, and ICRP 60 and the full ICRP 103. For these I relied on abstracts, the ICRP’s free extract and the ICRP’s own published history. Anything stated from background knowledge that I did not re-check against a primary text is marked [not re-verified].

Second pass (later on 25 September 2026). Web search was available for a spot-check of the most time-sensitive and least-verified items:

• the full text of the NRC proposed rule;
• the status of that rulemaking;
• the UNSCEAR 2025 cancer review;
• EU infringement actions under Directive 2013/59/Euratom;
• the Mayapuri and Cikande incidents;
• the CSRLD figures;
• RERF’s institutional status.

Additions from this pass are folded into the relevant claims.

Overview#

Lambert’s chapter has aged well where it is most empirical and least well where it speculates or assigns motives.

What held up or grew stronger

• Low-dose risk. Direct evidence from nuclear workers (INWORKS 2015, 2023) and from children given CT scans (UK, Australia, EPI-CT, US/Ontario) now shows raised cancer risk at doses below 100 mGy. Risk per unit dose is at least as high as the atomic-bomb-based estimates.
• The mainstream position. BEIR VII (2006), ICRP 103 (2007), NCRP Commentary 27 (2018) and an ICRP memorandum of September 2025 all retain the linear no-threshold (LNT) model. UNSCEAR’s 2025 cancer review says modern studies “strengthen” low-dose risk assessments.
• Medical optimisation. The gap Lambert flagged got worse before it got better. UK collective medical dose rose from the NRPB’s roughly 16,000 man-Sv to 24,700 man-Sv in 2008, with CT accounting for 68%. US per-capita medical dose rose about sixfold between 1980 and 2006, then fell somewhat by 2016.
• Stewart’s pelvimetry finding. It remains the mainstream view, with a documented sceptical minority.
• Surveillance databases. His call to maintain them (p. 36) has been vindicated, both by what the databases produced and by what was lost when funding lapsed.

What weakened

• The power-line prediction (p. 34). It has not been borne out. IARC still classes ELF magnetic fields as Group 2B, WHO says the evidence is “not strong enough to be considered causal”, and the newest pooled analyses find the association fading or absent.

What held up only in part

• “Belated” (p. 36). The label is only partly fair. The evidence behind the 1990 revision came together between 1986 and 1988 (revised bomb dosimetry, longer follow-up and new projection models), and ICRP adopted it in 1990. Even so, the ICRP’s own history concedes it first responded defensively.
• “Politics entered the scene” (p. 34). Cold War historiography broadly supports the claim, but it is too simple. Limits were tightened sharply in 1950 and 1958 even as nuclear programmes expanded.

The largest post-2001 development

The “substantial lobbies” for thresholds and hormesis (p. 35) failed through the scientific route. The US NRC denied their petitions in 2021. They then gained ground through the political route:

• May 2025. Executive Order 14300 declared LNT and ALARA to “lack sound scientific basis”.
• July 2026. The NRC proposed removing ALARA from its rules, while keeping a linear dose-response model and the nominal dose limits. It would add flexibility to exceed them: multiyear occupational extensions and case-by-case alternative public limits. Comments closed on 31 August 2026. No final rule had been issued as of 25 September 2026.

The ICRP still resists these lobbies, as Lambert reported. The contest has moved from science to regulatory policy, and it is framed around the cost of protection.

Late Lessons II (2013) did not update this case in its Annex 3. Its Annex 2 re-summarised the chapter and added one line on CT over-exposure. Its Table A2.1 records “65 years of substantial inaction” (1896 to 1961–1996). Chapter 24 (Cranor) endorses the UK compensation scheme, drawing on the scheme’s own figures.

Verdicts at a glance#

Claim 1: The evidence base and the “conservative” LNT model (pp. 34–35)#

Original claim. Radiation risk estimates are “probably more quantified and more soundly based than risks from any other environmental pollutant”. But they are “derived almost exclusively from the health records of the survivors of the atomic bombings in Japan in 1945 i.e. at high dose and dose rate”. Hence “a conservative linear dose-effect relationship is assumed and it is therefore appreciated that there is a risk at all doses” (pp. 34–35).

What happened since 2001#

Authoritative reviews kept LNT

• BEIR VII (US National Academies, 2006). The report said “a comprehensive review of available biological and biophysical data supports a ‘linear-no-threshold’ (LNT) risk model … the smallest dose has the potential to cause a small increase in risk to humans”. It also found that “the preponderance of information indicates that there will be some risk, even at low doses”. It rejected the view that low doses are more harmful than linear, and it listed hormesis only as a research question. (Report in brief)
• ICRP Publication 103 (2007). The Commission found that “assuming a linear response at low doses, the combined detriment due to excess cancer and heritable effects remains unchanged at around 5% per Sv”. It kept the dose and dose-rate effectiveness factor (DDREF) at 2. (free extract)
• NCRP Commentary No. 27 (2018). After reviewing 29 low-dose and low-dose-rate studies, the NCRP judged the data “broadly supportive of the LNT model”, with “no alternative dose-response relationship … more pragmatic or prudent” (Shore et al., J Radiol Prot 2018).
• The ICRP Main Commission (September 2025). Its memorandum stated that “at low doses, the System is supported by application of the linear-no-threshold model … prudent (that is, carefully considered given existing uncertainties)” (Rühm et al., J Radiol Prot 2025).

A dissenting scientific view

• The French Academies of Science and Medicine (2005) concluded that LNT “may greatly overestimate” risks below 100 mSv, and even more below 10 mSv (Tubiana et al., IJROBP 2005; summary of the divergence, Rad Environ Biophys 2006).
• Mainstream bodies did not adopt this view.

Direct low-dose evidence arrived

This is the most important change since 2001.

• INWORKS 2015 (nuclear workers in France, the UK and the US). The study covered 308,297 workers with a mean colon dose of about 21 mGy. Solid-cancer mortality rose by 47% per Gy. That per-Gy risk was “similar to estimates derived from studies of Japanese atomic bomb survivors”, despite very low dose rates (Richardson et al., BMJ 2015). Leukaemia (excluding CLL) showed an excess relative risk (ERR) of 2.96 per Gy (Leuraud et al., Lancet Haematol 2015).
• INWORKS 2023 update. With 309,932 workers and 10.7 million person-years, the ERR was 0.52 per Gy. Restricting the analysis to 0–100 mGy “approximately doubled the estimate”. The authors call the estimate “larger than estimates currently informing radiation protection” (Richardson et al., BMJ 2023).
• Children given CT scans
• UK. Cumulative doses of about 50–60 mGy were associated with a roughly threefold risk of leukaemia and brain tumours (Pearce et al., Lancet 2012).
• Australia. Among 680,000 exposed people, cancer incidence was 24% higher, rising by 0.16 per additional scan (Mathews et al., BMJ 2013).
• Europe (EPI-CT). Brain cancer showed an ERR of 1.27 per 100 mGy (Hauptmann et al., Lancet Oncol 2023). Haematological malignancies showed an ERR of 1.96 per 100 mGy across 948,174 people (Bosch de Basea Gomez et al., Nat Med 2023).
• US and Ontario (3.7 million children). Haematological cancers showed an ERR of 2.54 per 100 mGy. An estimated 10.1% of haematological cancers in the cohort may be attributable to imaging (Smith-Bindman et al., NEJM, September 2025).
• Systematic review. An NCI-led review of 26 studies with mean doses below 100 mGy, published between 2006 and 2017, found that only a few positive studies were plausibly biased upward. It concluded the studies “directly support excess cancer risks from low-dose ionizing radiation”, at magnitudes compatible with the bomb survivors (Hauptmann et al., JNCI Monogr 2020).
• UNSCEAR’s cancer review (approved June 2025). At its 72nd session (16–20 June 2025) UNSCEAR approved a scientific annex on cancer epidemiology, its first comprehensive update since the 2006 report. The Committee’s chair said the findings from modern studies, “including new analyses of the atomic bombing survivors in Japan, strengthen contemporary assessments of radiation risks—particularly for low dose and low dose-rate exposures in medical, environmental, and occupational exposure situations” (UN Information Service, UNIS/OUS/448, 20 June 2025).
• The expert group summarised the review in Burtt et al., Lancet Oncol, May 2026. Its authors include members of the INWORKS and RERF teams. I could not access the text, so no specific findings are attributed to it here.
• A companion annex on circulatory disease “confirms increased risk of heart and vascular diseases, especially at higher doses”, with uncertainties at lower doses.

Critiques of the low-dose evidence remain

• Confounding by indication. Children who are scanned may already be at higher risk. Boice argued that UNSCEAR’s 2013 report attributed the early CT associations possibly to this “reverse causation” (Boice, Ann ICRP 2015). A French cohort found risk estimates changed when predisposing factors were considered (Journy et al., J Radiol Prot 2016).
• Dose-response shape in the bomb survivors. The 2017 solid-cancer incidence update found a linear response for women, upward curvature for men, and a significant dose response in the 0–100 mGy range. The authors warned that “uncertainties in the shape of the dose response preclude definitive conclusions” (Grant et al., Radiat Res 2017).

Dependence on the bomb survivors persists for official risk coefficients

• The Life Span Study remains “the main technical basis” of ICRP recommendations from 1965 to 2007 and will steer the next ones (Wakeford & Laurier, Carcinogenesis 2025).
• ICRP found that DDREF is among the parameters with “a substantial impact on radiation detriment”, and that some key parameters “require updating”, such as reference population data and cancer severity (ICRP Publication 152, 2022).
• BEIR VII used a DDREF of 1.5 rather than ICRP’s 2. That choice is itself contested (Health Phys 2015).

Verdict: Strengthened#

• The core claim that there is risk at all doses now has direct epidemiological support rather than resting only on extrapolation.
• “Almost exclusively” from bomb survivors is still accurate for the official coefficients. It is no longer accurate for the evidence base.
• The word “conservative” is the weak point. If INWORKS 2023 is right, LNT combined with a DDREF of 2 may understate low-dose-rate risk rather than overstate it. Hormesis advocates and the French Academies argue the opposite. The mainstream calls LNT “prudent”, not demonstrably conservative.

Weight for the section’s lessons#

• Insight 9: risk evidence comes from harmed, atypical populations, so surveillance must be sustained (pp. 33, 35–36). Strengthened. The field did move from one extreme exposure population to routine-exposure cohorts, but only because registries existed (see Claim 7).
• Insight 10: threshold defaults persist because they set the cost of protection (pp. 32–35). Also gains support, as Claim 3 shows. The technical question of the dose-response shape at low doses remains genuinely uncertain, and that uncertainty is what the policy contest exploits.

Claim 2: The 1990 revision as “a belated response to mounting incontrovertible evidence” (p. 36; also p. 31)#

Original claim. ICRP perceived the cancer risk as “four to five times higher in 1990 as compared to 1977”. This “resulted in changes in dose limits but was a belated response to mounting incontrovertible evidence, a situation which has been a recurring theme” (p. 36).

What happened since 2001#

The size of the 1990 revision is broadly right

• ICRP 26 (1977) used a fatal-cancer risk of 1.25×10⁻² per Sv. ICRP 60 (1990) used 5×10⁻² per Sv for the whole population and 4×10⁻² for workers. That is roughly a fourfold increase [standard values; not re-verified against the primary texts].
• The ICRP’s own history confirms the order of magnitude. It quotes Publication 26 as assuming “a total risk of the order of 10⁻² Sv⁻¹” when setting the public limit. It adds that “by 1989, the Commission had itself revised upwards its estimates” (Clarke & Valentin 2009, paras 81, 84).
• UNSCEAR’s 1988 report documents the drivers. It cites three more sets of survivor mortality data, “a revised dosimetric system”, and projection of risks into the future, which “has had the combined effect of making the risk estimates … higher than before”. The lifetime fatal-cancer coefficient was 4.5% per Gy under additive projection and 7.1% per Gy under multiplicative projection. The 1977 high-dose figure had been “about 2.5 per cent per sievert” (UNSCEAR 1988 Report, paras 198, 245–247).
• Part of the increase therefore reflects a modelling choice: moving from additive to multiplicative projection. It does not reflect new observations alone.

Was it “belated”?

• Against “belated”:
• UNSCEAR deliberately postponed its review of cancer risk in its 1982 report because the bomb dosimetry “was in the process of being revised”. At that time it expected changes of no more than a factor of two (UNSCEAR 1988, para 116).
• The revised dosimetry arrived in 1986–87. UNSCEAR reported in 1988 and ICRP adopted Publication 60 in 1990.
• ICRP also issued interim statements in 1985 (Paris) and 1987 (Como) (Paris 1985; Como 1987; contents not re-verified).
• Measured from when the evidence came together, the lag was about two to four years. “Incontrovertible” overstates the certainty available before about 1987.
• For “belated”:
• The ICRP’s own commissioned history concedes that when 1980s re-evaluations suggested higher risks and “pressures began to appear for a reduction in dose limits”, ICRP’s “response was initially to emphasise the principle of optimisation and to claim that the use of collective dose and cost–benefit analysis always ensured that individual doses were sufficiently low” (Clarke & Valentin, ICRP Publication 109, 2009, paras 83–84).

No further upward revision of the cancer coefficient

• ICRP 103 held the combined detriment “unchanged at around 5% per Sv” (see Claim 1). In its words, “the overall estimates of cancer risk attributable to radiation exposure have not changed appreciably in the past 16 years”, while “the estimated risk of heritable effects is currently lower than before” (ICRP 103 free extract).
• The table values moved slightly downward: cancer detriment 5.5 versus 6.0 ×10⁻² per Sv, and heritable detriment 0.2 versus 1.3 [ICRP 103 Table 1; not re-verified].

The limit-lags-evidence pattern did recur, for the eye

• ICRP 103 itself warned that “revised judgements may be required particularly in respect of the eye”.
• In 2011 ICRP cut the recommended occupational eye-lens limit from 150 mSv to 20 mSv a year, averaged over five years, with a nominal cataract threshold of 0.5 Gy (ICRP Publication 118, 2012).
• The EU wrote it into law in Article 9(3)(a) of Directive 2013/59/Euratom, with a transposition deadline of February 2018.
• Later evidence suggests a threshold “is less evident with longer follow-up” (Hamada, Radiat Res 2023).
• Rollins had flagged cataract in about 1900 (p. 32).

Adoption of revised limits was not universal

• The US NRC never adopted ICRP 60’s 20 mSv occupational limit. It kept 50 mSv (5 rem).
• Responding to a 2012 staff paper, the NRC Commission “disapproved the staff recommendations to develop a basis for reducing the limit on occupational total effective dose”. It discontinued the ICRP 103 alignment rulemaking in 2016.
• The NRC’s own later account says most responses to its 2014 advance notice opposed alignment, “in large part because of doubts regarding the safety benefits of the proposed changes when weighed against the costs” (91 FR 43456, background section).
• In 2026 it still declined to lower the 150 mSv eye-lens limit, calling the ICRP and NCRP findings “preliminary” and “conservative” (NRC proposed rule, 91 FR 43456, 15 July 2026).

Verdict: Partly held up#

• The fact of a fourfold-to-fivefold revision: held up.
• The recurring-lag theme: held up, and the eye-lens episode reinforces it.
• “Belated response to … incontrovertible evidence”: overstated. The revision followed the consolidated evidence within a few years. Part of the increase came from modelling choices. The documented lag lies elsewhere: in ICRP’s initial defensiveness and, above all, in uneven national uptake.

Weight for the section’s lessons#

• Insight 2: well-characterised risks can be revised several-fold decades into use (pp. 31, 36). Remains strong as a fact. It is reinforced by the 7.5-fold eye-lens cut, a revision to a deterministic endpoint first warned about about 110 years earlier (p. 32).
• The causal account of the lag should be restated. Revisions depend on:
• measurement infrastructure, such as bomb dosimetry;
• modelling conventions;
• a stage in which institutions defend existing frameworks;
• jurisdictions deciding separately whether to adopt the change.

“Belated” is best treated as moderate-strength, not strong.

Claim 3: Lobbies for thresholds and hormesis “resisted by the ICRP” (p. 35)#

Original claim. “There are now substantial lobbies for changes which include both re-introducing the concept of thresholds and considerations of hormesis … these have been resisted by the ICRP” (p. 35).

What happened since 2001#

The petitions route

• 2015. Three petitions (Marcus; Miller; Doss et al.) asked the US NRC to replace LNT with a hormesis model and remove ALARA (Federal Register notice, 23 June 2015). As the NRC later summarised them, they also sought to:
• double the occupational limit from 5 rem to 10 rem (50 to 100 mSv);
• raise the public limit to match it;
• end the lower limits for pregnant women, the embryo or foetus, and children under 18.

(91 FR 43456, background section.) - August 2021 denial. The NRC found the LNT model “continues to provide a sound regulatory basis”. It noted: “None of the national and international authoritative scientific advisory bodies … support the hormesis concept as a regulatory model for radiation protection” (86 FR 45923, 17 August 2021).

The political route

• Executive Order 14300 (signed 23 May 2025). The order stated that models positing “no safe threshold” “lack sound scientific basis and produce irrational results”. It directed the NRC to “reconsider reliance on the linear no-threshold (LNT) model … and the ‘as low as reasonably achievable’ standard, which is predicated on LNT”, and to “consider adopting determinate radiation limits”. Final rules were due within 18 months (90 FR 22587, 29 May 2025).
• July 2025. An Idaho National Laboratory report recommended raising the public dose limit, “in part, to increase public acceptance of radiation exposure” (as described in the NRC proposal below).
• NRC proposed rule (published 15 July 2026; comments closed 31 August 2026). The proposal would:
• “remove references to the ALARA principle”;
• apply a graded dose-management approach with fixed thresholds;
• create a “planned occupational dose limit extension” allowing annual limits to be exceeded within five-year totals;
• allow case-by-case variances to the public dose limit.

The NRC projected savings of about $9.53 million a year to industry at a 7% discount rate, plus smaller savings to state regulators and to the NRC itself. - The NRC’s stated rationale. The NRC “has determined that the LNT model may lead to conservative implementation of radiation protection measures at low doses”. Its aim is to “definitively move away from overly conservative practices that have developed over time in the name of ALARA”. It calls ALARA “an outgrowth of applying the LNT model”. - What the proposal keeps. The NRC proposes “to continue to use the linear dose response model, as it continues to be the most appropriate model”, and says it “has also not identified a suitable alternative model”. It does not change the nominal 1 mSv (100 mrem) public limit, though it would let licensees request alternative public limits case by case. It does not adopt hormesis (91 FR 43456; full text). - Status. The comment period on docket NRC-2025-1140 closed on 31 August 2026. A public meeting on 10 August drew more than 550 participants (ANS Nuclear Newswire, 11 August 2026; trade press). No final rule was found as of 25 September 2026. The executive order’s 18-month deadline implies about November 2026. - Reactions. I did not survey the docket. As examples, the American Society of Radiologic Technologists opposed removing ALARA (AuntMinnie, 2026; secondary). Nicole Martinez, an ICRP member speaking in her professional capacity rather than for the Commission, said she was “not convinced that some of the proposed changes aren’t just a different type of subjectivity” (ANS Nuclear Newswire, 13 July 2026). I found no formal ICRP submission.

The ICRP’s current stance

• The ICRP’s September 2025 memorandum reaffirmed LNT as prudent. It also stressed that optimisation means “the appropriate level of protection, not simply the lowest dose” (Rühm et al. 2025).
• The memorandum answers a common criticism of how ALARA is applied, without conceding the threshold argument.

Verdict: Strengthened#

• ICRP resistance: Lambert’s report of it holds as of 2026.
• The lobbies: far more consequential than in 2001. Having failed through scientific bodies and petitions, the threshold and anti-ALARA position was adopted as executive policy in the largest national regulatory system. Even so, the regulator’s own proposal stopped short of abandoning a linear model.
• The framing: the US contest is openly about the cost and speed of deploying nuclear technology. It turns on how “reasonable” protection below the limits is operationalised, not on whether low doses are harmful.
• What the lobbies sought versus what they got. The 2015 petitions sought much more: higher limits and an end to special protection for the foetus and children. The 2026 proposal does not adopt those demands outright, but it loosens both kinds of limit at the margins:
• Limits. Nominal limits stay, but annual occupational limits could be exceeded under “planned occupational dose limit extensions” within multiyear totals. Licensees could request “alternative public dose limits”, a flexibility the NRC links partly to “the declared energy emergency in E.O. 14156”.
• The foetus and children. Declared pregnant women and minors would be excluded from the extensions.

The central change is the proposal to remove ALARA, the open-ended form of the optimisation duty that Lambert treats as the distinctive modern tool (pp. 34–35). The NRC would replace it with fixed-threshold “graded” dose management, which it says still meets the IAEA’s optimisation requirement (GSR Part 3, Requirement 11). Whether it does is a matter of dispute.

Weight for the section’s lessons#

Insight 10: threshold defaults set the cost of protection and so stay contested (pp. 32–35). Upgraded from moderate to strong. Two patterns are now visible:

• Challenges switch venues. When a scientific challenge fails in expert bodies, it can succeed by moving to a political venue that sets the regulator’s objectives.
• Challenges target implementation, not the model. They can aim at optimisation (ALARA) while leaving the formal dose-response model in place.

Both are technology-neutral dynamics worth carrying into any lens on standard-setting.

Claim 4: Power lines as “a similar and contemporary story” (p. 34)#

Original claim. After the pelvimetry story: “A similar and contemporary story may be unfolding in relation to the childhood leukaemia risk in proximity to overhead power lines in the United States” (p. 34).

What happened since 2001#

Formal assessments

• IARC Monograph 80 (2002). Classified ELF magnetic fields as “possibly carcinogenic to humans (Group 2B)” (IARC). No re-evaluation was found.
• WHO Environmental Health Criteria 238 (2007). The WHO noted a “consistent pattern of increased risk” above 0.3–0.4 µT. It also noted uncertainties from control-selection bias and exposure misclassification, and that “virtually all of the laboratory evidence and the mechanistic evidence fail to support a relationship”. Its conclusion: “on balance, the evidence is not strong enough to be considered causal, but sufficiently strong to remain a concern”. Assuming causality, it estimated 100–2,400 cases a year worldwide (0.2–4.9%). It advised that precautionary measures be “very low” cost and that exposure limits should not be cut “to some arbitrary level in the name of precaution” (WHO EHC 238).

Epidemiology after 2001

• England and Wales. Leukaemia risk was 1.69 within 200 m of lines. The authors noted “no accepted biological mechanism … may be due to chance or confounding” (Draper et al., BMJ 2005).
• Britain, 1962–2008. The distance association “declines over time”. The relative risk for 0–199 m fell from 4.50 in the 1960s to 0.71 in the 2000s. The authors judged this “unlikely to arise from any physical effect of the powerlines” (Bunch et al., Br J Cancer 2014).
• California, US (the setting of Lambert’s “United States” claim). The distance findings did “not clearly support an increased … risk”, and risks beyond 50 m “were not replicated” (Crespi et al., Br J Cancer 2016). There was no clear magnetic-field risk (Kheifets et al., Cancer Causes Control 2017). The findings “argue against magnetic fields as a sole explanation” (Crespi et al., Environ Res 2019).
• International pooled analysis of distance. The odds ratio was 1.33 (0.92–1.93) within 50 m of lines of 200 kV or more, with no association with calculated fields (Amoon et al., Br J Cancer 2018).
• Newest pooled field analysis. The odds ratio was 1.01 at 0.4 µT or above. The authors report “a decrease in effect to no association”. A meta-analysis of the three pooled analyses gives 1.45 (0.95–2.20) (Amoon et al., Environ Res 2022).
• Meta-analysis. The overall association (odds ratio 1.26) is “mainly explained by the studies conducted before 2000”. Studies after 2000 give 1.04 (0.84–1.29) (Brabant et al., Rev Environ Health 2023).

Verdict: Weakened#

• The expected arc did not occur. Twenty-five years on, the pattern Lambert implied has not appeared: a real low-dose hazard, initially disbelieved, later confirmed as with Stewart.
• Where the association stands. It has attenuated over calendar time, which points toward confounding or selection. No mechanism has emerged. Official classifications are unchanged.
• Not closed. The high-exposure (>0.4 µT) signal in older studies has never been fully explained.
• Why the analogy failed. Pelvimetry involved an established carcinogen, whereas power-frequency fields are a physical agent with no demonstrated carcinogenic mechanism. That weakened the case for the analogy from the start.

Weight for the section’s lessons#

• This is the section’s only testable forward prediction, and it did not come true. It shows the risk of reasoning by analogy from one validated early warning to a structurally different agent.
• The broader lesson survives: dismissing a statistical association too quickly can be costly (p. 34). It needs a companion: persistence, dose-response, mechanism and replication across time also have to be checked.
• WHO’s handling is a useful counter-model. Low-cost precaution, proportionate to weak evidence and with explicit refusal to lower limits arbitrarily, shows a middle path that the section does not discuss.

Claim 5: Justification and optimisation in medical radiology (p. 35; Table 3.1, p. 36)#

Original claim. The NRPB estimated that “about 20% of all X-rays carried out in the United Kingdom are clinically unhelpful”. Annual collective medical dose was about 16,000 man-Sv, of which about 7,500 man-Sv, “nearly 50%”, could be avoided. Inter-hospital dose variation “can be more than an order of magnitude” (Table 3.1 says “100x”). Individual doses are perhaps “two orders of magnitude lower than 60 years ago”, but “the problem of optimisation of that dose still exists” (p. 35).

What happened since 2001#

Aggregate dose rose, driven by CT, rather than halving

• UK.
• 1997/98: 0.33 mSv per person a year.
• 2008: 24,700 man-Sv, about 0.4 mSv per person. CT’s share of collective dose rose from 40% to 68%, though CT is a small fraction of examinations; conventional radiography is 90% of examinations but only 19% of dose.
• Against Lambert’s roughly 16,000 man-Sv baseline, collective dose rose by about half instead of halving (HPA-CRCE-012, December 2010).
• US.
• Per-capita medical dose rose about sixfold, from 0.5 mSv in 1980 to 3.0 mSv in 2006 (Mettler et al., Radiology 2009, summarising NCRP Report 160 and UNSCEAR).
• In the NCRP’s 2016 reassessment (Report 184) it fell to 2.3 mSv. CT examinations nevertheless rose from 67 million to 84 million: “the trend … of increasing dose from medical radiation has reversed” (Mettler et al., Radiology 2020).
• Worldwide. There are 4.2 billion procedures a year, at 0.56 mSv per person. The US figure is 2.2 mSv, and CT use has grown faster elsewhere (Mahesh et al., Radiology 2023, drawing on UNSCEAR 2020/21).
• Formal acknowledgement. The EU’s recast directive states that “technological and scientific developments have led to a notable increase in the exposure of patients” (recital 28, Directive 2013/59/Euratom).

Per-examination optimisation for conventional radiography did improve

• UK national surveys show average falls in typical dose of 16% (1995–2000), 16% (2000–2005) and 5% (2005–2010).
• Between 2010 and 2019 there were further falls of 22% for single projections and 40% for examinations and interventional procedures.
• UKHSA attributes these to technology, technique and “the increased importance placed on dose optimisation”. Diagnostic reference levels (DRLs) have been legally required since 2000 (UKHSA 2019 review, published 8 July 2025; UK NDRLs).
• EU law now requires DRLs, dose-display devices and justification of individual exposures (Articles 55–56, 60 of Directive 2013/59/Euratom).

Variation remains wide, now documented for CT

• Across seven countries. After adjusting for patient characteristics, there was a fourfold range in mean abdominal CT dose (7.0–25.7 mSv) and a 17-fold range in the proportion of high-dose examinations. The variation was “primarily attributable to institutional decisions regarding technical parameters” (Smith-Bindman et al., BMJ 2019).
• US screening-trial sites. Screening CT effective dose ranged from 1.7 to 11 mSv across sites. Most scans exceeded professional-society dose thresholds (JACR 2025).
• High cumulative doses. These “were not rare”. England had 1.7 to 5 times fewer patients exceeding them than the average (Br J Radiol 2026).
• What this means for the Table’s figure. These data support “more than an order of magnitude” for extreme institutional comparisons. The Table’s “100x” is not supported by the text’s cited source and looks like an EEA exaggeration.

Justification gaps persist at a similar scale

• EU-JUST-CT audit. Across 6,734 CT referrals in seven EU countries, 75% were appropriate on average (58% to 86%) (Singer et al., Eur Radiol 2025). That echoes the NRPB’s 20% “unhelpful”.
• Advance justification. Fewer than half of European countries have a medical practitioner justify individual CT examinations in advance. Referral guidelines are in daily use in only 13% (Foley et al., Insights Imaging 2022).

Risk projections and their caveats

• The projection. US CT use in 2023 is projected to cause about 103,000 future cancers, with CT possibly accounting for 5% of annual diagnoses if practice persists (Smith-Bindman et al., JAMA Intern Med 2025).
• The model dependence. These figures depend on the BEIR VII models.
• UNSCEAR’s caution. UNSCEAR “does not recommend multiplying very low doses by large numbers of individuals” for doses at or below background level (UNSCEAR 2012 Report). The same caution applies, more weakly, to Lambert’s collective-dose arithmetic.

Late Lessons II (2013). Annex 2 appended: “Similar concerns are now being expressed at the frequent over-exposure to radiation from CT scans” (LL2 Annexes).

Verdict: Strengthened#

• The core lesson that principles do not ensure practice was borne out. It was borne out more strongly than Lambert could see, because a new, highly beneficial modality drove aggregate exposure up even as per-examination doses for older modalities fell.
• The specific numbers are historical. “100x” should not be relied on.

Weight for the section’s lessons#

• Insight 3: prior justification plus optimisation is a distinctive tool, but principles alone leave unjustified practice (pp. 34–35). The gap half is now strong. The tool half is better supported than in 2001, as DRL-linked declines in conventional radiography show, but only where benchmarks, surveys and legal duties exist.
• Benefit displacing caution (pp. 31–34) recurs within a mature, regulated practice, not only at a technology’s birth. Growth in use can swamp gains in per-unit safety. Governance that benchmarks dose per procedure but not procedure volume or appropriateness can show improvement while total exposure rises.

Claim 6: Stewart’s pelvimetry finding and Doll’s 5% (p. 34)#

Original claim. Stewart’s finding “was at first controversial and disbelieved but, after being repeated by others, it is now accepted that there is a significant risk of leukaemia from even small radiation doses received by the embryo or foetus”. Doll (1989) estimated that “about 5% of all cases of childhood cancer were caused by pelvimetry”: about 75 a year in the UK and 300 in the US. These cases “would have been saved had the work of Stewart et al. been acted on earlier” (p. 34).

What happened since 2001#

The mainstream view

• Doll and Wakeford (1997). Case-control studies consistently show about a 40% proportional increase. “The evidence against bias and confounding as alternative explanations … is strong.” Doses “of the order of 10 mGy” to the foetus increase childhood cancer risk (Br J Radiol 1997).
• Wakeford and Little (2003). The Oxford Survey and the bomb-survivor in utero cohort give compatible risk coefficients, supporting causation (Int J Radiat Biol 2003).
• Wakeford (2008). “On balance … the evidence points to low-level irradiation of the fetus increasing the risk of leukaemia”. The coefficients are “likely to be overestimates” (Radiat Prot Dosim 2008).
• ICRP 103 (2007). Cancer risk after prenatal exposure is judged “similar to that following irradiation in early childhood”.
• Wakeford and Bithell (2021). Studies other than the Oxford Survey show “consistent and clear elevations of risk” (Int J Radiat Biol 2021).
• Little et al. (2022). A review of 89 studies found support for excess risk. Odds ratios decline over calendar period, consistent with falling doses (Sci Total Environ 2022).

The dissent

• Boice and Miller (1999). Evidence “derives almost entirely from case-control studies, whereas practically all cohort studies find no association”, including the bomb survivors exposed in utero. The uniform relative risk of about 1.5 across childhood cancer types “suggests an underlying bias” (Teratology 1999).
• The UK Childhood Cancer Study. It found a non-significant odds ratio of 1.36 for leukaemia after in utero X-rays, reflecting lower modern doses (Rajaraman et al., BMJ 2011).
• The bomb survivors exposed in utero. They show a solid-cancer excess in adulthood. Lifetime risk may be “considerably lower than for early childhood exposure” (Preston et al., JNCI 2008).

Timing of acceptance

UNSCEAR’s own 1988 retrospective (paras 103, 115) shows that its mid-1960s and 1970s reports already treated in utero cancer risk as real and quantified it. That supports the Table’s note that the finding was “not generally accepted until the 1970s” (UNSCEAR 1988).

Ongoing review. ICRP Task Group 121 is reviewing intrauterine risk estimates for the next recommendations (Wakeford & Hande, Ann ICRP 2026).

The 5% figure. I found no later primary source restating Doll’s attributable fraction. It applies to the era of routine obstetric radiography and would be far lower today.

Verdict: Held up#

• “Now accepted” remains an accurate description of the mainstream. A minority dissent, based on cohort-study nulls, persists.
• Doll’s numbers are historical estimates, not a current burden.
• “Would have been saved” is a plausible but unquantified counterfactual.
• Later evidence suggests the original risk coefficients may be overestimates.

Weight for the section’s lessons#

• Insight 7: harm evidence is resisted through familiar moves such as demanding replication and proposing alternative explanations (pp. 32–34). Still supported. The Stewart episode ended with the finding accepted and practice changed.
• Balance note. The same case shows that scepticism based on design (case-control versus cohort) was scientifically legitimate and never fully resolved. The lesson should not be read as “the dissenters were simply wrong”.

Claim 7: Fund long-term epidemiological databases “even when an immediate need is not perceived” (p. 36)#

Original claim. “The precautionary principle suggests that epidemiological databases of long-term effects must be funded and maintained for the future even when an immediate need is not perceived” (p. 36).

What happened since 2001#

Payoffs from maintained databases

Almost every major advance under Claims 1, 5 and 6 came from records-based infrastructure built decades earlier.

• UK National Registry for Radiation Workers (set up 1976). Now more than 270,000 workers.
• On 1 January 2024 the separate BNFL, UKAEA and AWE worker studies were merged into it, and a 25-year Sellafield biological-sample archive passed to UKHSA (GOV.UK, April 2026).
• A governance group including unions was established in 2022 (NRRW guidance, updated June 2026).
• It is a core INWORKS cohort.
• The CT cohorts depended on hospital radiology information systems and national registries (NHS; Australian Medicare; nine European countries).
• The Radiation Effects Research Foundation (RERF) continues its Life Span Study and offspring work into 2026 (RERF). Eighty years on, the survivors “provide important information on risks decades after exposure at a young age” (Wakeford & Laurier 2025).
• Funding. RERF is still jointly funded by Japan’s Ministry of Health, Labour and Welfare and the US Department of Energy (disclosure in Burtt et al. 2026).
• New premises. Its Hiroshima laboratory “is scheduled to relocate to the Hiroshima University Kasumi Campus next year, 2027” (RERF notice, 2026). That is a long-term institutional commitment rather than a wind-down.
• Biosamples. RERF has archived “over 2.3 million biosamples from more than 30,000 study participants”, which its reviewers see as the route to new low-dose questions as the cohort dies out (Gray & Feinberg, Carcinogenesis, October 2025).

Losses when need was “not perceived”

• The Oxford Childhood Cancer Research Group grew out of Stewart’s Oxford Survey and supported the government’s Committee on Medical Aspects of Radiation in the Environment (COMARE) from its inception. It closed in 2014. “The effect of the closure on the work of COMARE has already been noted.” Its registry data went to an archive, and childhood cancer registration passed to Public Health England (Kendall et al., Br J Cancer 2018; Muirhead, Br J Cancer 2018; Stiller et al., BMC Cancer 2019).
• The US Department of Energy’s Low Dose Radiation Research Program (1999–2016) “was terminated owing to BER’s change of focus”. The National Academies found that since 2016 there has been “a lack of leadership and scientific activity” in the area. They recommended about $100 million a year for 10–15 years, against congressionally authorised levels of $30 million (2023) and $40 million (2024) (NASEM 2022, Summary). Actual appropriations were not verified.

Verdict: Strengthened#

• The value of the recommendation has been demonstrated. The direct low-dose risk estimates exist only because worker and hospital records were kept for decades.
• The warning has also been borne out by counter-example: well-regarded surveillance units were closed or defunded in quiet periods.

Weight for the section’s lessons#

• Insight 9 (pp. 33, 35–36) and the section’s single explicit recommendation deserve high weight.
• A technology-neutral corollary. Surveillance capacity is most vulnerable to cuts exactly when no crisis is visible, and rebuilding it later costs far more than keeping it, if it can be rebuilt at all.
• Institutional form matters. The UK registry survived through several arrangements: hosting within a public-health agency, formal data-sharing agreements with each employer, legal authority to process records without individual consent, renewed ethics approvals, and joint employer-union governance. That durability is worth noting as a design feature.

Claim 8: Directive 96/29 “will (eventually) be implemented”; uneven implementation; incidents like Goiânia (p. 35)#

Original claim. Directive 96/29 “will (eventually) be implemented throughout Europe … However, it has been found difficult to ensure that radiation protection legislation is implemented uniformly and there continue to be examples of careless or irresponsible attitudes towards radiation sources and waste which have resulted in horrendous injuries and death, such as the caesium-137 incident at Goiania” (p. 35).

What happened since 2001#

Legislation

• Directive 96/29 was implemented and then replaced. Directive 2013/59/Euratom (5 December 2013) consolidated and repealed 96/29, the medical directive 97/43 and the high-activity sources directive 2003/122. Transposition was due by 6 February 2018.
• The recast directive’s recital 51 states: “Unresolved problems with orphan sources remain, and there have been significant cases of contaminated metal being imported from third countries.”

Uneven transposition and compliance

• Late transposition. Italy transposed by Legislative Decree 101 of 31 July 2020, about two and a half years late.
• Commission enforcement.
• On 29 September 2022, more than four and a half years after the deadline, the Commission referred Latvia to the Court of Justice for incomplete transposition. Latvia had not set optimised protection strategies for contaminated areas, had no strategies for existing exposure situations, and had not adopted a radon action plan (European Commission, IP/22/5406). The case outcome was not verified.
• In April 2023 the Commission opened further infringement procedures against Ireland, France and Cyprus “for failing to correctly transpose” the directive (European Commission, DG Energy, 19 April 2023).
• Compliance in radiology departments. European Society of Radiology surveys in 2018 and 2021 found “variable compliance” and “a mixed picture” on key requirements, including clinical audit (Insights Imaging 2021).

Source incidents continued

• Mayapuri, Delhi (2010). A cobalt-60 research irradiator was “improperly disposed of” by the University of Delhi’s chemistry department and dismantled in a scrapyard. One worker died of multiple organ failure on 26 April 2010, and six more people were hospitalised (World Nuclear News, 20 May 2010). It was “the worst radiation accident India has yet dealt with” (PLoS One 2014; Curr Radiopharm 2022).
• Seattle, University of Washington (2019). A 2,900 Ci caesium-137 source was breached during decommissioning. Thirteen people were contaminated, and estimated clean-up costs and lost revenue exceeded $150 million (Health Phys 2023).
• Cikande industrial estate, Indonesia (2025). A caesium-137 contamination incident (Appl Radiat Isot 2026).
• Detection abroad. The contamination was first noticed abroad. In August 2025 the US FDA found Cs-137 in frozen shrimp from one Indonesian firm (about 68 Bq/kg), and in September in cloves from another (732 Bq/kg). Both were below the FDA’s intervention level of 1,200 Bq/kg. The FDA placed both firms on an import alert and, from 31 October 2025, required import certification for shrimp and spices from affected regions (FDA, updated 28 May 2026).
• Source. The FDA page does not name one. Indonesian officials reportedly traced it to contaminated scrap metal at a metal-smelting plant in the Cikande estate, with contamination at 22 facilities (Food Safety Magazine; CBS News; secondary).
• Why it matters here. Like Goiânia and Mayapuri, the harm pathway was a source that escaped regulatory control into the scrap-metal stream.

Verdict: Held up#

• “Eventually” implemented: yes, though the framework was replaced by a broader directive that was itself transposed late in places.
• Uneven implementation: persists.
• Careless handling of sources and waste: caused serious incidents in low-, middle- and high-income settings throughout the period, and EU law itself acknowledges the problem is unresolved.

Weight for the section’s lessons#

• Insight 8: voluntary recommendations leave misuse unchecked; binding law lags (pp. 34–35). Extends naturally. Even binding law has a long implementation tail.
• Hazards outlive their useful life. Harm concentrates at the end of a hazardous item’s life cycle, when it is disused, orphaned or scrapped, and falls outside the institutions that managed its use. This is a durable, technology-neutral pattern.

Claim 9: The UK joint compensation scheme “has been extremely successful” (p. 36)#

Original claim. The scheme, “run jointly by trades unions and the nuclear industry, has been extremely successful in providing an alternative to litigation” (p. 36).

What happened since 2001#

Scheme data (CSRLD 2024–25 Annual Statement, June 2025)

• The scheme has run since 1982. It now “provides nearly complete coverage of the UK nuclear industry employees”, including defence and dockyard employers.
• It pays on a sliding scale from 20% probability of causation, versus the courts’ 50% “all or nothing”.
• Its dose-probability schedules were updated in 2009 to reflect BEIR VII and UNSCEAR 2006.
• 1,823 cases have been considered since 1982; 202 succeeded (about 11%).
• In 2024/25, 23 new claims were received and none succeeded (CSRLD annual statement; scheme history). The figures were re-checked against the live statement on 25 September 2026.
• The same statement reports “a successful year”, a review “to maintain Scheme capability”, and consideration of new ICRP internal-dose models.

Evaluations

• Late Lessons II, Chapter 24 (Cranor, 2013, pp. 598–599). Endorses the scheme, citing 106 people compensated for £5.3 million in total (about £50,000 each), most at causation probabilities below 50%. It concludes “the Scheme had therefore achieved its goals”, but the source is the scheme itself (CSRLD, 2010) (LL2 Part D).
• Leigh and Wakeford (2001). The scheme “has worked well and is held up as a model of alternative dispute resolution”. Wakeford then worked for the nuclear industry (Health Phys 2001).
• No independent evaluation was found, for example of claimant experience, timeliness or payment adequacy.

Parallel development

The US Radiation Exposure Compensation Act (RECA), “designed as a non-adversarial alternative to litigation” that “does not require claimants to prove causation”, was reauthorized and amended on 4 July 2025 (Pub. L. 119-21). It pays $100,000 lump sums and added a Manhattan Project waste category (US DOJ, RECA).

Verdict: Partly held up#

• What is confirmed: the scheme’s durability (more than 40 years), expansion, generosity relative to tort, and apparent success in displacing litigation.
• “Extremely successful” rests on self-reported and insider sources. The scheme’s own measure of success is process continuity, not claimant outcomes. Its 2024/25 statement calls a year with no successful claims “a successful year”.
• Scale is modest: about 200 successful claims in 43 years, with many claims unsuccessful because most cancers in workers are not attributable to their doses.

Weight for the section’s lessons#

Insight 13: no-fault or probability-weighted compensation suits long-latency harm (pp. 35–36). Moves from “asserted” to “supported but modest”.

• Where it applies. The model depends on well-characterised dose records and dose-response models, which radiation uniquely has.
• What it does not show. It is evidence that scheme design can reduce adversarial cost. It is not evidence that such schemes deliver large-scale redress.

Claim 10: Post-war limits set so as not to restrict the nuclear industries; the ICRP took no stand on atmospheric testing (p. 34)#

Original claim. “The radiation protection community was faced with the problem of setting dose limits that did not appear to restrict the expansion of these industries — politics entered the scene.” The ICRP “took no stand about the testing of nuclear weapons in the atmosphere” (p. 34). The claim is unsourced in the chapter.

What happened since 2001#

What the ICRP’s own history concedes

The ICRP’s commissioned history (Clarke & Valentin 2009) documents several things:

• 1950s public protection. Public-exposure protection was introduced in 1954 amid concern about fallout, with “nuclear energy expected to be an expanding industry”.
• The 1977 framework. Publication 26 aimed to protect people “while still allowing necessary activities from which radiation exposure might result”. Collective dose was introduced partly because “a global expansion of nuclear power reactors … was foreseen”.
• The 1980s. The Commission initially responded defensively to higher risk estimates (Claim 2).

Where the same history cuts against Lambert

• Limits tightened during the expansion years.
• 1950: from about 500 to about 150 mSv a year.
• 1958: to 5 rem (50 mSv) a year for workers and 0.5 rem (5 mSv) for the public, driven largely by genetic concerns.

Historical scholarship

• Boudia (2007). The international structure for radiation expertise and regulation was built “as a result of American foreign policy, international relations in the context of the Cold War, public mobilization against nuclear weapons … as well as scientific research interests and professional legitimatization” (History and Technology 2007).
• Higuchi (2017, 2020). Expert bodies evaluated fallout’s “scientific uncertainty and moral ambiguity” in incongruent ways, which “fueled the fallout controversy”. Fallout became a global environmental issue that fed into the 1963 Partial Test Ban Treaty (IRAP 2017; Political Fallout, Stanford UP 2020).
• Accounts not read in full here.
• Hamblin on the negotiation of the first US National Academy of Sciences study of radiation effects (J Hist Biol 2007).
• Walker’s Permissible Dose (UC Press 2000).
• A US National Council on Radiation Protection (NCRP) founder’s own lecture, “Some nonscientific influences on radiation protection standards and practice” (Taylor, Health Phys 1980).
• British government downplaying of fallout against dissenting scientists (J Cold War Stud 2018).

I did not read these in full, so no specific findings are attributed to them.

On testing

• I found nothing indicating the ICRP took a position on atmospheric testing, which is consistent with its self-limited remit.
• The UN created UNSCEAR in 1955 (resolution 913(X)) (UNSCEAR 1988).
• The US later officially acknowledged harm to “downwinders” and test participants through RECA (1990; amended 2025).

Verdict: Partly held up#

• What is confirmed. Standard-setting was shaped by Cold War programmes, the institutions’ sense that nuclear expansion should proceed, and public pressure. The ICRP’s own account confirms that expansion was a design context, and it did stay silent on testing.
• What is too simple. Occupational limits were cut about tenfold between the late 1940s and 1958 (from about 500 to about 150 mSv a year in 1950, then to 50 mSv in 1958). This happened during rapid expansion, partly because of genetic concerns and fallout-driven public concern. That contradicts a crude reading.
• Better statement of the claim. Protection bodies set limits they judged compatible with continued expansion, and tightened them when scientific and public pressure made that necessary.

Weight for the section’s lessons#

Insight 12: a mandate to expand an industry bends standards, and non-candid reassurance erodes trust (p. 34). Moves from “asserted” to “moderately supported, with nuance”.

Technology-neutral mechanisms that are defensible:

• Enabling baselines. Standard-setters’ framing of “necessary activities” sets the baseline that protection is optimised against.
• Public controversy tightens standards. Public controversy can push standards tighter against a promotional mandate.
• Scope silence. Expert bodies may stay silent on the most politically charged exposure sources by defining them as outside their remit.

Cross-cutting assessment#

Where hindsight moves the weight up

• Latent harm and miscalibrated governance (insight 1, pp. 33, 171, 174). Reinforced by eye-lens cataract (a deterministic effect with a far lower threshold than assumed) and circulatory disease.
• Several-fold revisions decades into use (insight 2). Reinforced by the eye-lens cut.
• The optimisation gap (insight 3). Grew with CT before partial correction.
• Sustained surveillance (insight 9). Vindicated, and shown to be fragile.
• Contested thresholds that set the cost of protection (insight 10). Now visible at the level of national executive policy.

Where hindsight moves the weight down or adds nuance

• “Belated” and “incontrovertible” (p. 36). Overstated.
• The power-line analogy (p. 34). Failed.
• Institutional motive claims (p. 34). Partly supported but too simple.
• “Extremely successful” compensation (p. 36). Rests on insider data.
• “Conservative” LNT (p. 35). Contested from both sides. The best current direct evidence suggests the model is not obviously conservative at low dose rates.

Late Lessons II (2013)

• No Annex 3 update for this case.
• Annex 2 added the CT concern and a “65 years of substantial inaction” figure.
• Chapter 24 endorsed the compensation scheme.
• Chapter 18 on Chernobyl and Fukushima extends the radiation material; it was not reviewed here.

Sources#

All retrieved or checked in September 2026 unless stated. The date given is the publication date.

Scientific assessments and recommendations

• National Academies, BEIR VII Phase 2, report in brief (2005/2006): https://nap.nationalacademies.org/resource/11340/beir_vii_final.pdf
• ICRP Publication 103, 2007 Recommendations (free extract, 2007): https://www.icrp.org/docs/ICRP_Publication_103-Annals_of_the_ICRP_37(2-4)-Free_extract.pdf
• ICRP Publication 118, tissue reactions (2012): https://doi.org/10.1016/j.icrp.2012.02.001
• ICRP Publication 152, detriment methodology (2022): https://www.icrp.org/publication.asp?id=ICRP%20Publication%20152
• Rühm W et al., “Essentials of the system of radiological protection”, J Radiol Prot (September 2025): https://doi.org/10.1088/1361-6498/ae02a2
• Clarke RH, Valentin J, The History of ICRP and the Evolution of its Policies, ICRP Publication 109 (2009): https://www.icrp.org/docs/The%20History%20of%20ICRP%20and%20the%20Evolution%20of%20its%20Policies.pdf
• ICRP, Statement from the 1985 Paris meeting (1985): https://doi.org/10.1088/0260-2814/5/2/407
• ICRP, Statement from the 1987 Como meeting (1987): https://doi.org/10.1016/0146-6453(87)90006-6
• UNSCEAR 1988 Report to the General Assembly (1988): https://www.unscear.org/unscear/uploads/documents/unscear-reports/UNSCEAR_1988_Report.pdf
• UNSCEAR 2012 Report (2012/2015 corrigendum): https://www.unscear.org/unscear/uploads/documents/unscear-reports/UNSCEAR_2012_Report-CORR.pdf
• Shore RE et al., NCRP Commentary 27 summary, J Radiol Prot (2018): https://doi.org/10.1088/1361-6498/aad348
• Tubiana M et al., French Academies joint report, IJROBP (2005): https://doi.org/10.1016/j.ijrobp.2005.06.013
• “Recent reports on the effect of low doses…”, Radiat Environ Biophys (2006): https://doi.org/10.1007/s00411-006-0032-9
• Wakeford R, Laurier D, Carcinogenesis (2025): https://doi.org/10.1093/carcin/bgaf062
• “Comments on the DDREF estimate of the BEIR VII Committee”, Health Phys (2015): https://doi.org/10.1097/HP.0000000000000189
• Hamada N, Radiat Res (2023): https://doi.org/10.1667/RADE-23-00030.1
• UN Information Service, “UN Scientific Committee approves major scientific updates on radiation risks”, UNIS/OUS/448 (20 June 2025): https://unis.unvienna.org/unis/en/pressrels/2025/unisous448.html
• Burtt JJ et al., “Ionising radiation and cancer: a UN review of the recent epidemiological evidence”, Lancet Oncol 27(5):534–537 (May 2026): https://doi.org/10.1016/S1470-2045(26)00115-4
• National Academies, Leveraging Advances in Modern Science to Revitalize Low-Dose Radiation Research in the United States (2022), Summary: https://nap.nationalacademies.org/read/26434/chapter/2 (DOI https://doi.org/10.17226/26434)

Low-dose epidemiology

• Richardson DB et al., INWORKS, BMJ (October 2015): https://doi.org/10.1136/bmj.h5359
• Leuraud K et al., INWORKS leukaemia, Lancet Haematol (2015): https://doi.org/10.1016/S2352-3026(15)00094-0
• Richardson DB et al., INWORKS update, BMJ (August 2023): https://doi.org/10.1136/bmj-2022-074520
• Pearce MS et al., Lancet (June 2012): https://doi.org/10.1016/S0140-6736(12)60815-0
• Mathews JD et al., BMJ (May 2013): https://doi.org/10.1136/bmj.f2360
• Hauptmann M et al., EPI-CT brain, Lancet Oncol (2023): https://doi.org/10.1016/S1470-2045(22)00655-6
• Bosch de Basea Gomez M et al., EPI-CT haematological, Nat Med (2023): https://doi.org/10.1038/s41591-023-02620-0
• Smith-Bindman R et al., NEJM (September 2025): https://doi.org/10.1056/NEJMoa2502098
• Hauptmann M et al., JNCI Monogr (2020): https://doi.org/10.1093/jncimonographs/lgaa010
• Boice JD, Ann ICRP (2015): https://doi.org/10.1177/0146645315575877
• Journy N et al., J Radiol Prot (2016): https://doi.org/10.1088/0952-4746/36/1/N1
• Grant EJ et al., Radiat Res (2017): https://doi.org/10.1667/RR14492.1

US regulatory trajectory

• NRC, notice of LNT petitions, 80 FR (23 June 2015): https://www.federalregister.gov/documents/2015/06/23/2015-15441/linear-no-threshold-model-and-standards-for-protection-against-radiation
• NRC, denial of petitions, 86 FR 45923 (17 August 2021): https://www.federalregister.gov/documents/2021/08/17/2021-17475/linear-no-threshold-model-and-standards-for-protection-against-radiation
• Executive Order 14300, 90 FR 22587 (signed 23 May 2025; published 29 May 2025): https://www.federalregister.gov/documents/2025/05/29/2025-09798/ordering-the-reform-of-the-nuclear-regulatory-commission
• NRC, “Reforming and Modernizing the NRC’s Radiation Protection Framework”, proposed rule, 91 FR 43456 (15 July 2026; docket NRC-2025-1140; comments closed 31 August 2026): https://www.federalregister.gov/documents/2026/07/15/2026-14208/reforming-and-modernizing-the-nrcs-radiation-protection-framework (full text: https://www.federalregister.gov/documents/full_text/text/2026/07/15/2026-14208.txt)
• ANS Nuclear Newswire, “A closer look at NRC’s proposed rule eliminating ALARA” (13 July 2026): https://www.ans.org/news/2026-07-13/article-8199/a-closer-look-at-nrcs-proposed-rule-eliminating-alara/
• ANS Nuclear Newswire, “NRC holds public meeting for proposed rule eliminating ALARA” (11 August 2026): https://www.ans.org/news/2026-08-11/article-8284/nrc-holds-public-meeting-for-proposed-rule-eliminating-alara/
• AuntMinnie, “ASRT opposes proposal to remove ALARA standard from radiation rules” (2026; secondary, headline only): https://www.auntminnie.com/clinical-news/radiation-oncology-therapy/news/15835538/asrt-opposes-proposal-to-remove-alara-standard-from-radiation-rules

Power lines

• IARC Monograph 80 (2002): https://publications.iarc.who.int/98
• WHO, Environmental Health Criteria 238 (2007): https://iris.who.int/handle/10665/43646
• Draper G et al., BMJ (2005): https://doi.org/10.1136/bmj.330.7503.1290
• Bunch KJ et al., Br J Cancer (2014): https://doi.org/10.1038/bjc.2014.15
• Crespi CM et al., Br J Cancer (2016): https://doi.org/10.1038/bjc.2016.142
• Kheifets L et al., Cancer Causes Control (2017): https://doi.org/10.1007/s10552-017-0951-6
• Amoon AT et al., Br J Cancer (2018): https://doi.org/10.1038/s41416-018-0097-7
• Crespi CM et al., Environ Res (2019): https://doi.org/10.1016/j.envres.2019.01.022
• Amoon AT et al., Environ Res (2022): https://doi.org/10.1016/j.envres.2021.111993
• Brabant C et al., Rev Environ Health (2023): https://doi.org/10.1515/reveh-2021-0112

Medical exposure

• HPA, Frequency and collective dose for medical and dental X-ray examinations in the UK, 2008 (HPA-CRCE-012, December 2010): https://www.gov.uk/government/publications/medical-and-dental-x-rays-frequency-and-collective-doses-in-the-uk
• UKHSA, Patient doses from radiographic and simple fluoroscopic X-ray imaging procedures in the UK: 2019 review (8 July 2025): https://www.gov.uk/government/publications/radiographic-and-simple-fluoroscopic-x-ray-imaging-patient-doses
• UKHSA, National Diagnostic Reference Levels (updated 11 December 2025): https://www.gov.uk/government/publications/diagnostic-radiology-national-diagnostic-reference-levels-ndrls
• Mettler FA et al., Radiology (2009; NCRP 160/UNSCEAR summary): https://doi.org/10.1148/radiol.2532082010
• Mettler FA et al., Radiology (2020; NCRP 184 assessment): https://doi.org/10.1148/radiol.2020192256
• Mahesh M et al., Radiology (2023; US and worldwide 2009–2018): https://doi.org/10.1148/radiol.221263
• Smith-Bindman R et al., BMJ (2019): https://doi.org/10.1136/bmj.k4931
• Smith-Bindman R et al., JAMA Intern Med (April 2025): https://doi.org/10.1001/jamainternmed.2025.0505
• Watch the Spot trial radiation doses, J Am Coll Radiol (2025): https://doi.org/10.1016/j.jacr.2025.10.010
• Malekhedayat M et al., Br J Radiol (2026): https://doi.org/10.1093/bjr/tqag115
• Singer C et al., EU-JUST-CT audit, Eur Radiol (2025): https://doi.org/10.1007/s00330-024-11083-x
• Foley SJ et al., Insights Imaging (2022): https://doi.org/10.1186/s13244-022-01325-1
• European Society of Radiology, BSS uptake survey, Insights Imaging (2021): https://doi.org/10.1186/s13244-021-01078-3

In utero exposure

• Doll R, Wakeford R, Br J Radiol (1997): https://doi.org/10.1259/bjr.70.830.9135438
• Boice JD, Miller RW, Teratology (1999): https://doi.org/10.1002/(SICI)1096-9926(199904)59:4%3C227::AID-TERA7%3E3.0.CO;2-E
• Wakeford R, Little MP, Int J Radiat Biol (2003): https://doi.org/10.1080/0955300031000114729
• Preston DL et al., JNCI (2008): https://doi.org/10.1093/jnci/djn045
• Wakeford R, Radiat Prot Dosim (2008): https://doi.org/10.1093/rpd/ncn272
• Rajaraman P et al., BMJ (2011): https://doi.org/10.1136/bmj.d472
• Wakeford R, Bithell JF, Int J Radiat Biol (2021): https://doi.org/10.1080/09553002.2021.1906463
• Little MP et al., Sci Total Environ (2022): https://doi.org/10.1016/j.scitotenv.2022.154723
• Wakeford R, Hande MP, Ann ICRP (2026): https://doi.org/10.1177/01466453251412160

Surveillance infrastructure

• UKHSA, NRRW guidance (updated 15 June 2026): https://www.gov.uk/guidance/national-registry-for-radiation-workers-guidance
• UKHSA, Merger of BNFL, UKAEA and AWE studies into NRRW (21 April 2026): https://www.gov.uk/guidance/merger-of-the-bnfl-ukaea-and-awe-worker-epidemiology-studies-into-nrrw
• Radiation Effects Research Foundation, website (accessed September 2026): https://www.rerf.or.jp/en/
• RERF, notice on the pause in facility tours ahead of the 2027 relocation (2026): https://www.rerf.or.jp/en/contact/form_tourh-en/
• Gray JW, Feinberg AP, “Radiation effects research foundation—a view to the future”, Carcinogenesis (27 October 2025): https://doi.org/10.1093/carcin/bgaf061
• Kendall GM et al., Br J Cancer (2018): https://doi.org/10.1038/s41416-018-0182-y
• Muirhead CR, Br J Cancer (2018): https://doi.org/10.1038/s41416-018-0222-7
• Bithell JF et al., Br J Cancer (2018): https://doi.org/10.1038/s41416-018-0180-0
• Stiller CA et al., BMC Cancer (2019): https://doi.org/10.1186/s12885-019-5344-7

EU law and source incidents

• Council Directive 2013/59/Euratom (5 December 2013; OJ L 13, 17 January 2014): https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32013L0059
• Italy, Legislative Decree 101 of 31 July 2020 (Gazzetta Ufficiale, 12 August 2020): https://www.gazzettaufficiale.it/eli/id/2020/08/12/20G00121/sg
• European Commission, “Commission refers LATVIA to Court”, IP/22/5406 (29 September 2022): https://ec.europa.eu/commission/presscorner/detail/en/ip_22_5406
• European Commission, DG Energy, “April infringements package: key decisions on energy” (19 April 2023): https://energy.ec.europa.eu/news/april-infringements-package-key-decisions-energy-2023-04-19_en
• Mayapuri: World Nuclear News, “Contamination removal in Delhi scrap market” (20 May 2010): https://www.world-nuclear-news.org/Articles/Contamination-removal-in-Delhi-scrap-market
• Mayapuri: PLoS One (2014): https://doi.org/10.1371/journal.pone.0091579
• Mayapuri: Curr Radiopharm (2022): https://doi.org/10.2174/1874471014666210118123424
• University of Washington caesium-137 breach, Health Phys (2023): https://doi.org/10.1097/HP.0000000000001660
• Cikande caesium-137 incident, Appl Radiat Isot (2026): https://doi.org/10.1016/j.apradiso.2026.112896
• US FDA, “FDA Response to Imported Foods Potentially Contaminated with Cesium-137” (updated 28 May 2026): https://www.fda.gov/food/environmental-contaminants-food/fda-response-imported-foods-potentially-contaminated-cesium-137
• Food Safety Magazine, “Indonesian Officials Find Cause of Cs-137 Contamination Behind Radioactive Shrimp” (2025; secondary): https://www.food-safety.com/articles/10757-indonesian-officials-find-cause-of-cs-137-contamination-behind-radioactive-shrimp
• CBS News, “Indonesia finds radiation at clove farm…” (2025; secondary): https://www.cbsnews.com/news/shrimp-radiation-indonesia-cesium-clove-farm-us-fda-spices-contaminated/

Compensation

• CSRLD, 2024–25 Annual Statement (June 2025): http://www.csrld.org.uk/index.php/annual-statement
• CSRLD, Scheme History (accessed September 2026): http://www.csrld.org.uk/index.php/scheme-history
• Leigh WJ, Wakeford R, Health Phys (2001): https://doi.org/10.1097/00004032-200112000-00014
• US Department of Justice, Radiation Exposure Compensation Act programme page (accessed September 2026): https://www.justice.gov/civil/common/reca

History of standard-setting

• Boudia S, History and Technology (2007): https://doi.org/10.1080/07341510701527443
• Higuchi T, International Relations of the Asia-Pacific (2017): https://doi.org/10.1093/irap/lcx024
• Higuchi T, Political Fallout, Stanford UP (2020): https://doi.org/10.1515/9781503612907
• Hamblin JD, J Hist Biol (2007): https://doi.org/10.1007/s10739-005-6531-8
• Walker JS, Permissible Dose, UC Press (2000): https://doi.org/10.1525/california/9780520223288.001.0001
• Taylor LS, 1980 Sievert Lecture, Health Phys (1980): https://pubmed.ncbi.nlm.nih.gov/7204045/
• “Dissenting Scientists in Early Cold War Britain”, J Cold War Stud (2018): https://doi.org/10.1162/jcws_a_00801

Late Lessons II (EEA Report 1/2013)

• Annexes (Annex 2 overview of 2001 cases; Annex 3 updates, which do not include radiation): https://www.eea.europa.eu/en/analysis/publications/late-lessons-2/late-lessons-2-full-report/late-lessons-2-annexes
• Part D (Chapter 24, Cranor, on compensation, pp. 598–599): https://www.eea.europa.eu/en/analysis/publications/late-lessons-2/late-lessons-2-full-report/part-d-costs-justice-and-innovation-1