Outcome Assessment in Epidemiological Studies of Low-Dose Radiation Exposure and Cancer Risks: Sources, Level of Ascertainment, and Misclassification

A generalisation of the method of regression calibration and comparison with Bayesian and frequentist model averaging methods

For many cancer sites low-dose risks are not known and must be extrapolated from those observed in groups exposed at much higher levels of dose. Measurement error can substantially alter the dose-response shape and hence the extrapolated risk. Even in studies with direct measurement of low-dose exposures measurement error could be substantial in relation to the size of the dose estimates and thereby distort population risk estimates. Recently, there has been considerable attention paid to methods of dealing with shared errors, which are common in many datasets, and particularly important in occupational and environmental settings. In this paper we test Bayesian model averaging (BMA) and frequentist model averaging (FMA) methods, the first of these similar to the so-called Bayesian two-dimensional Monte Carlo (2DMC) method, and both fairly recently proposed, against a very newly proposed modification of the regression calibration method, the extended regression calibration (ERC) method, which is particularly suited to studies in which there is a substantial amount of shared error, and in which there may also be curvature in the true dose response. The quasi-2DMC with BMA method performs well when a linear model is assumed, but very poorly when a linear-quadratic model is assumed, with coverage probabilities both for the linear and quadratic dose coefficients that are under 5% when the magnitude of shared Berkson error is large (50%). For the linear model the bias is generally under 10%. However, using a linear-quadratic model it produces substantially biased (by a factor of 10) estimates of both the linear and quadratic coefficients, with the linear coefficient overestimated and the quadratic coefficient underestimated. FMA performs as well as quasi-2DMC with BMA when a linear model is assumed, and generally much better with a linear-quadratic model, although the coverage probability for the quadratic coefficient is uniformly too high. However both linear and quadratic coefficients have pronounced upward bias, particularly when Berkson error is large. By comparison ERC yields coverage probabilities that are too low when shared and unshared Berkson errors are both large (50%), although otherwise it performs well, and coverage is generally better than the quasi-2DMC with BMA or FMA methods, particularly for the linear-quadratic model. The bias of the predicted relative risk at a variety of doses is generally smallest for ERC, and largest for the quasi-2DMC with BMA and FMA methods (apart from unadjusted regression), with standard regression calibration and Monte Carlo maximum likelihood exhibiting bias in predicted relative risk generally somewhat intermediate between ERC and the other two methods. In general ERC performs best in the scenarios presented, and should be the method of choice in situations where there may be substantial shared error, or suspected curvature in the dose response.

A generalisation of the method of regression calibration and comparison with Bayesian and frequentist model averaging methods

For many cancer sites low-dose risks are not known and must be extrapolated from those observed in groups exposed at much higher levels of dose. Measurement error can substantially alter the dose-response shape and hence the extrapolated risk. Recently, there has been considerable attention paid to methods of dealing with shared errors, which are particularly important in occupational and environmental settings. In this paper we test Bayesian model averaging (BMA) and frequentist model averaging (FMA) methods, the first of these similar to the so-called Bayesian two-dimensional Monte Carlo (2DMC) method, and both fairly recently proposed, against a very newly proposed modification of the regression calibration method, the extended regression calibration (ERC) method. The quasi-2DMC+BMA method performs well when a linear model is assumed, but poorly when a linear-quadratic model is assumed. FMA performs as well as quasi-2DMC+BMA when a linear model is assumed, and generally much better with a linear-quadratic model, although the coverage probability for the quadratic coefficient is uniformly too high. ERC yields coverage probabilities that are too low when shared and unshared Berkson errors are both large (50%), although otherwise it performs well, and coverage is generally better than the quasi-2DMC+BMA or FMA methods, particularly for the linear-quadratic model. The bias of predicted relative risk at a variety of doses is generally smallest for ERC, and largest for quasi-2DMC+BMA and FMA, with standard regression calibration and Monte Carlo maximum likelihood exhibiting bias in predicted relative risk generally somewhat intermediate between ERC and the other two methods. In general ERC performs best in the scenarios presented, and should be the method of choice in situations where there may be substantial shared error, or suspected curvature in the dose response.

Third mortality follow-up of the Mallinckrodt uranium processing workers, 1942-2019

INTRODUCTION

Mallinckrodt Chemical Works was a uranium processing facility during the Manhattan Project from 1942 to 1966. Thousands of workers were exposed to low-dose-rates of ionizing radiation from external and internal sources. This third follow-up of 2514 White male employees updates cancer and noncancer mortality potentially associated with radiation and silica dust.

MATERIALS AND METHODS

Individual, annualized organ doses were estimated from film badge records (n monitored = 2514), occupational chest x-rays (n = 2514), uranium urinalysis (n = 1868), radium intake through radon breath measurements (n = 487), and radon ambient measurements (n = 1356). Silica dust exposure from pitchblende processing was estimated (n = 1317). Vital status and cause of death determination through 2019 relied upon the National Death Index and Social Security Administration Epidemiological Vital Status Service. The analysis included standardized mortality ratios (SMRs), Cox proportional hazards, and Poisson regression models.

RESULTS

Vital status was confirmed for 99.4% of workers (84.0% deceased). For a dose weighting factor of 1 for intakes of uranium, radium, and radon decay products, the mean and median lung doses were 65.6 and 29.9 mGy, respectively. SMRs indicated a difference in health outcomes between salaried and hourly workers, and more brain cancer deaths than expected [SMR: 1.79; 95% confidence interval (CI): 1.14, 2.70]. No association was seen between radiation and lung cancer [hazard ratio (HR) at 100 mGy: 0.93; 95%CI: 0.78, 1.11]. The relationship between radiation and kidney cancer observed in the previous follow-up was maintained (HR at 100 mGy: 2.07; 95%CI: 1.12, 3.79). Cardiovascular disease (CVD) also increased significantly with heart dose (HR at 100 mGy: 1.11; 95%CI: 1.02, 1.21). Exposures to dust ≥23.6 mg/m3-year were associated with nonmalignant kidney disease (NMKD) (HR: 3.02; 95%CI: 1.12, 8.16) and kidney cancer combined with NMKD (HR: 2.46; 95%CI: 1.04, 5.81), though without evidence of a dose-response per 100 mg/m3-year.

CONCLUSIONS

This third follow-up of Mallinckrodt uranium processors reinforced the results of the previous studies. There was an excess of brain cancers compared with the US population, although no radiation dose-response was detected. The association between radiation and kidney cancer remained, though potentially due to few cases at higher doses. The association between levels of silica dust ≥23.6 mg/m3-year and NMKD also remained. No association was observed between radiation and lung cancer. A positive dose-response was observed between radiation and CVD; however, this association may be confounded by smoking, which was unmeasured. Future work will pool these data with other uranium processing worker cohorts within the Million Person Study.

A generalisation of the method of regression calibration and comparison with the Bayesian 2-dimensional Monte Carlo method

For many cancer sites it is necessary to assess risks from low-dose exposures via extrapolation from groups exposed at moderate and high levels of dose. Measurement error can substantially alter the shape of this relationship and hence the derived population risk estimates. Even in studies with direct measurement of low-dose exposures measurement error could be substantial in relation to the size of the dose estimates and thereby distort population risk estimates. Recently, much attention has been devoted to the issue of shared errors, common in many datasets, and particularly important in occupational settings. In this paper we test a Bayesian model averaging method, the so-called Bayesian two-dimensional Monte Carlo (2DMC) method, that has been fairly recently proposed against a very newly proposed modification of the regression calibration method, which is particularly suited to studies in which there is a substantial amount of shared error, and in which there may also be curvature in the true dose response. We also compared both methods against standard regression calibration and Monte Carlo maximum likelihood. The Bayesian 2DMC method performs poorly, with coverage probabilities both for the linear and quadratic dose coefficients that are under 5%, particularly when the magnitudes of classical and Berkson error are both moderate to large (20%-50%). The method also produces substantially biased (by a factor of 10) estimates of both the linear and quadratic coefficients, with the linear coefficient overestimated and the quadratic coefficient underestimated. By comparison the extended regression calibration method yields coverage probabilities that are too low when shared and unshared Berkson errors are both large (50%), although otherwise it performs well, and coverage is generally better than the Bayesian 2DMC and all other methods. The bias of the predicted relative risk at a variety of doses is generally smallest for extended regression calibration, and largest for the Bayesian 2DMC method (apart from unadjusted regression), with standard regression calibration and Monte Carlo maximum likelihood exhibiting bias in predicted relative risk generally somewhat intermediate between the other two methods.

A generalisation of the method of regression calibration

There is direct evidence of risks at moderate and high levels of radiation dose for highly radiogenic cancers such as leukaemia and thyroid cancer. For many cancer sites, however, it is necessary to assess risks via extrapolation from groups exposed at moderate and high levels of dose, about which there are substantial uncertainties. Crucial to the resolution of this area of uncertainty is the modelling of the dose-response relationship and the importance of both systematic and random dosimetric errors for analyses in the various exposed groups. It is well recognised that measurement error can alter substantially the shape of this relationship and hence the derived population risk estimates. Particular attention has been devoted to the issue of shared errors, common in many datasets, and particularly important in occupational settings. We propose a modification of the regression calibration method which is particularly suited to studies in which there is a substantial amount of shared error, and in which there may also be curvature in the true dose response. This method can be used in settings where there is a mixture of Berkson and classical error. In fits to synthetic datasets in which there is substantial upward curvature in the true dose response, and varying (and sometimes substantial) amounts of classical and Berkson error, we show that the coverage probabilities of all methods for the linear coefficient [Formula: see text] are near the desired level, irrespective of the magnitudes of assumed Berkson and classical error, whether shared or unshared. However, the coverage probabilities for the quadratic coefficient [Formula: see text] are generally too low for the unadjusted and regression calibration methods, particularly for larger magnitudes of the Berkson error, whether this is shared or unshared. In contrast Monte Carlo maximum likelihood yields coverage probabilities for [Formula: see text] that are uniformly too high. The extended regression calibration method yields coverage probabilities that are too low when shared and unshared Berkson errors are both large, although otherwise it performs well, and coverage is generally better than these other three methods. A notable feature is that for all methods apart from extended regression calibration the estimates of the quadratic coefficient [Formula: see text] are substantially upwardly biased.

A generalisation of the method of regression calibration

There is direct evidence of risks at moderate and high levels of radiation dose for highly radiogenic cancers such as leukaemia and thyroid cancer. For many cancer sites, however, it is necessary to assess risks via extrapolation from groups exposed at moderate and high levels of dose, about which there are substantial uncertainties. Crucial to the resolution of this area of uncertainty is the modelling of the dose-response relationship and the importance of both systematic and random dosimetric errors for analyses in the various exposed groups. It is well recognised that measurement error can alter substantially the shape of this relationship and hence the derived population risk estimates. Particular attention has been devoted to the issue of shared errors, common in many datasets, and particularly important in occupational settings. We propose a modification of the regression calibration method which is particularly suited to studies in which there is a substantial amount of shared error, and in which there may also be curvature in the true dose response. This method can be used in settings where there is a mixture of Berkson and classical error. In fits to synthetic datasets in which there is substantial upward curvature in the true dose response, and varying (and sometimes substantial) amounts of classical and Berkson error, we show that the coverage probabilities of all methods for the linear coefficient \(\alpha\) are near the desired level, irrespective of the magnitudes of assumed Berkson and classical error, whether shared or unshared. However, the coverage probabilities for the quadratic coefficient \(\beta\) are generally too low for the unadjusted and regression calibration methods, particularly for larger magnitudes of the Berkson error, whether this is shared or unshared. In contrast Monte Carlo maximum likelihood yields coverage probabilities for \(\beta\) that are uniformly too high. The extended regression calibration method yields coverage probabilities that are too low when shared and unshared Berkson errors are both large, although otherwise it performs well, and coverage is generally better than these other three methods. A notable feature is that for all methods apart from extended regression calibration the estimates of the quadratic coefficient \(\beta\) are substantially upwardly biased.

The Future of Radioactive Medicine

The discovery of X rays in the late 19th century heralded the beginning of a new age in medicine, and the advent of channeling the power of radiation to diagnose and treat human disease. Radiation has been leveraged in medicine in a multitude of ways and is a critical element of cancer care including screening, diagnosis, surveillance, and interventional treatments. Modern radiotherapy techniques include a multitude of methodologies utilizing both externally and internally delivered radiation from a variety of approaches. This review provides a comprehensive overview of contemporary radiotherapy methodologies, the field of radiopharmaceuticals and theranostics, effects of low dose radiation and highlights the phenomena of fear of exposure to radiation and its impact in modern medicine.

Systematic review of the potential carcinogenicity of bisphenol A in humans

Bisphenol A (BPA) is a synthetic chemical to which humans are exposed through a variety of environmental sources. We have conducted a comprehensive, systematic review of 29 epidemiology studies and 27 experimental animal studies, published through May 2022, evaluating the potential carcinogenicity of BPA to contribute to the understanding of whether BPA is carcinogenic in humans. We conducted this review according to best practices for systematic reviews and incorporating established frameworks for study quality evaluation and evidence integration. The epidemiology studies have many limitations that increase the risk of biased results, but overall, the studies do not provide clear and consistent evidence for an association between BPA exposure and the development of any type of cancer. The experimental animal studies also do not provide strong and consistent evidence that BPA is associated with the induction of any malignant tumor type. Some of the proposed mechanisms for BPA carcinogenicity are biologically plausible, but the relevance to human exposures is not clear. We conclude that there is inadequate evidence to support a causal relationship between BPA exposure and human carcinogenicity, based on inadequate evidence in humans, as well as evidence from experimental animal studies that suggests a causal relationship is not likely.

Biological and cellular responses of humans to high-level natural radiation: A clarion call for a fresh perspective on the linear no-threshold paradigm

There remains considerable uncertainty in obtaining risk estimates of adverse health outcomes of chronic low-dose radiation. In the absence of reliable direct data, extrapolation through the linear no-threshold (LNT) hypothesis forms the cardinal tenet of all risk assessments for low doses (≤ 100 mGy) and for the radiation protection principle of As Low As Reasonably Achievable (ALARA). However, as recent evidences demonstrate, LNT assumptions do not appropriately reflect the biology of the cell at the low-dose end of the dose-response curve. In this regard, human populations living in high-level natural radiation areas (HLNRA) of the world can provide valuable insights into the biological and cellular effects of chronic radiation to facilitate improved precision of the dose-response relationship at low doses. Here, data obtained over decades of epidemiological and radiobiological studies on HLNRA populations is summarized. These studies do not show any evidence of unfavourable health effects or adverse cellular effects that can be correlated with high-level natural radiation. Contrary to the assumptions of LNT, no excess cancer risks or untoward pregnancy outcomes have been found to be associated with cumulative radiation dose or in-utero exposures. Molecular biology-driven studies demonstrate that chronic low-dose activates several cellular defence mechanisms that help cells to sense, recover, survive, and adapt to radiation stress. These mechanisms include stress-response signaling, DNA repair, immune alterations and most importantly, the radiation-induced adaptive response. The HLNRA data is consistent with the new evolving paradigms of low-dose radiobiology and can help develop the theoretical framework of an alternate dose-response model. A rational integration of radiobiology with epidemiology data is imperative to reduce uncertainties in predicting the potential health risks of chronic low doses of radiation.

Age effects on radiation response: summary of a recent symposium and future perspectives

One of the principal uncertainties when estimating population risk of late effects from epidemiological data is that few radiation-exposed cohorts have been followed up to extinction. Therefore, the relative risk model has often been used to estimate radiation-associated risk and to extrapolate risk to the end of life. Epidemiological studies provide evidence that children are generally at higher risk of cancer induction than adults for a given radiation dose. However, the strength of evidence varies by cancer site and questions remain about site-specific age at exposure patterns. For solid cancers, there is a large body of evidence that excess relative risk (ERR) diminishes with increasing age at exposure. This pattern of risk is observed in the Life Span Study (LSS) as well as in other radiation-exposed populations for overall solid cancer incidence and mortality and for most site-specific solid cancers. However, there are some disparities by endpoint in the degree of variation of ERR with exposure age, with some sites (e.g., colon, lung) in the LSS incidence data showing no variation, or even increasing ERR with increasing age at exposure. The pattern of variation of excess absolute risk (EAR) with age at exposure is often similar, with EAR for solid cancers or solid cancer mortality decreasing with increasing age at exposure in the LSS. We shall review the human data from the Japanese LSS cohort, and a variety of other epidemiological data sets, including a review of types of medical diagnostic exposures, also some radiobiological animal data, all bearing on the issue of variations of radiation late-effects risk with age at exposure and with attained age. The paper includes a summary of several oral presentations given in a Symposium on "Age effects on radiation response" as part of the 67th Annual Meeting of the Radiation Research Society, held virtually on 3-6 October 2021.

Review of the risk of cancer following low and moderate doses of sparsely ionising radiation received in early life in groups with individually estimated doses

BACKGROUND

The detrimental health effects associated with the receipt of moderate (0.1-1 Gy) and high (>1 Gy) acute doses of sparsely ionising radiation are well established from human epidemiological studies. There is accumulating direct evidence of excess risk of cancer in a number of populations exposed at lower acute doses or doses received over a protracted period. There is evidence that relative risks are generally higher after radiation exposures in utero or in childhood.

METHODS AND FINDINGS

We reviewed and summarised evidence from 60 studies of cancer or benign neoplasms following low- or moderate-level exposure in utero or in childhood from medical and environmental sources. In most of the populations studied the exposure was predominantly to sparsely ionising radiation, such as X-rays and gamma-rays. There were significant (p < 0.001) excess risks for all cancers, and particularly large excess relative risks were observed for brain/CNS tumours, thyroid cancer (including nodules) and leukaemia.

CONCLUSIONS

Overall, the totality of this large body of data relating to in utero and childhood exposure provides support for the existence of excess cancer and benign neoplasm risk associated with radiation doses < 0.1 Gy, and for certain groups exposed to natural background radiation, to fallout and medical X-rays in utero, at about 0.02 Gy.

Cancer risk following low doses of ionising radiation - Current epidemiological evidence and implications for radiological protection

Recent studies suggest that every year worldwide about a million patients might be exposed to doses of the order of 100 mGy of low-LET radiation, due to recurrent application of radioimaging procedures. This paper presents a synthesis of recent epidemiological evidence on radiation-related cancer risks from low-LET radiation doses of this magnitude. Evidence from pooled analyses and meta-analyses also involving epidemiological studies that, individually, do not find statistically significant radiation-related cancer risks is reviewed, and evidence from additional and more recent epidemiological studies of radiation exposures indicating excess cancer risks is also summarized. Cohorts discussed in the present paper include Japanese atomic bomb survivors, nuclear workers, patients exposed for medical purposes, and populations exposed environmentally to natural background radiation or radioactive contamination. Taken together, the overall evidence summarized here is based on studies including several million individuals, many of them followed-up for more than half a century. In summary, substantial evidence was found from epidemiological studies of exposed groups of humans that ionizing radiation causes cancer at acute and protracted doses above 100 mGy, and growing evidence for doses below 100 mGy. The significant radiation-related solid cancer risks observed at doses of several 100 mGy of protracted exposures (observed, for example, among nuclear workers) demonstrate that doses accumulated over many years at low dose rates do cause stochastic health effects. On this basis, it can be concluded that doses of the order of 100 mGy from recurrent application of medical imaging procedures involving ionizing radiation are of concern, from the viewpoint of radiological protection.

Avoidance or adaptation of radiotherapy in patients with cancer with Li-Fraumeni and heritable TP53-related cancer syndromes

The management of patients with cancer and Li-Fraumeni or heritable TP53-related cancer syndromes is complex because of their increased risk of developing second malignant neoplasms after genotoxic stresses such as systemic treatments or radiotherapy (radiosusceptibility). Clinical decision making also integrates the risks of normal tissue toxicity and sequelae (radiosensitivity) and tumour response to radiotherapy (radioresistance and radiocurability). Radiotherapy should be avoided in patients with cancer and Li-Fraumeni or heritable TP53 cancer-related syndromes, but overall prognosis might be poor without radiotherapy: radioresistance in these patients seems similar to or worse than that of the general population. Radiosensitivity in germline TP53 variant carriers seems similar to that in the general population. The risk of second malignant neoplasms according to germline TP53 variant and the patient's overall oncological prognosis should be assessed during specialised multidisciplinary staff meetings. Radiotherapy should be avoided whenever other similarly curative treatment options are available. In other cases, it should be adapted to minimise the risk of second malignant neoplasms in patients who still require radiotherapy despite its genotoxicity, in view of its potential benefit. Adaptations might be achieved through the reduction of irradiated volumes using proton therapy, non-ionising diagnostic procedures, image guidance, and minimal stray radiation. Non-ionising imaging should become more systematic. Radiotherapy approaches that might result in a lower probability of misrepaired DNA damage (eg, particle therapy biology and tumour targeting) are an area of investigation.

Epidemiological studies of CT scans and cancer risk: the state of the science

20 years ago, 3 manuscripts describing doses and potential cancer risks from CT scans in children raised awareness of a growing public health problem. We reviewed the epidemiological studies that were initiated in response to these concerns that assessed cancer risks from CT scans using medical record linkage. We evaluated the study methodology and findings and provide recommendations for optimal study design for new efforts. We identified 17 eligible studies; 13 with published risk estimates, and 4 in progress. There was wide variability in the study methodology, however, which made comparison of findings challenging. Key differences included whether the study focused on childhood or adulthood exposure, radiosensitive outcomes (e.g. leukemia, brain tumors) or all cancers, the exposure metrics (e.g. organ doses, effective dose or number of CTs) and control for biases (e.g. latency and exclusion periods and confounding by indication). We were able to compare results for the subset of studies that evaluated leukemia or brain tumors. There were eight studies of leukemia risk in relation to red bone marrow (RBM) dose, effective dose or number of CTs; seven reported a positive dose-response, which was statistically significant (p < 0.05) in four studies. Six of the seven studies of brain tumors also found a positive dose-response and in five, this was statistically significant. Mean RBM dose ranged from 6 to 12 mGy and mean brain dose from 18 to 43 mGy. In a meta-analysis of the studies of childhood exposure the summary ERR/100 mGy was 1.78 (95%CI: 0.01-3.53) for leukemia/myelodisplastic syndrome (n = 5 studies) and 0.80 (95%CI: 0.48-1.12) for brain tumors (n = 4 studies) (p-heterogeneity >0.4). Confounding by cancer pre-disposing conditions was unlikely in these five studies of leukemia. The summary risk estimate for brain tumors could be over estimated, however, due to reverse causation. In conclusion, there is growing evidence from epidemiological data that CT scans can cause cancer. The absolute risks to individual patients are, however, likely to be small. Ongoing large multicenter cohorts and future pooling efforts will provide more precise risk quantification.

Lymphoma and multiple myeloma in cohorts of persons exposed to ionising radiation at a young age

There is limited evidence that non-leukaemic lymphoid malignancies are radiogenic. As radiation-related cancer risks are generally higher after childhood exposure, we analysed pooled lymphoid neoplasm data in nine cohorts first exposed to external radiation aged <21 years using active bone marrow (ABM) and, where available, lymphoid system doses, and harmonised outcome classification. Relative and absolute risk models were fitted. Years of entry spanned 1916-1981. At the end of follow-up (mean 42.1 years) there were 593 lymphoma (422 non-Hodgkin (NHL), 107 Hodgkin (HL), 64 uncertain subtype), 66 chronic lymphocytic leukaemia (CLL) and 122 multiple myeloma (MM) deaths and incident cases among 143,136 persons, with mean ABM dose 0.14 Gy (range 0-5.95 Gy) and mean age at first exposure 6.93 years. Excess relative risk (ERR) was not significantly increased for lymphoma (ERR/Gy = -0.001; 95% CI: -0.255, 0.279), HL (ERR/Gy = -0.113; 95% CI: -0.669, 0.709), NHL + CLL (ERR/Gy = 0.099; 95% CI: -0.149, 0.433), NHL (ERR/Gy = 0.068; 95% CI: -0.253, 0.421), CLL (ERR/Gy = 0.320; 95% CI: -0.678, 1.712), or MM (ERR/Gy = 0.149; 95% CI: -0.513, 1.063) (all p-trend > 0.4). In six cohorts with estimates of lymphatic tissue dose, borderline significant increased risks (p-trend = 0.02-0.07) were observed for NHL + CLL, NHL, and CLL. Further pooled epidemiological studies are needed with longer follow-up, central outcome review by expert hematopathologists, and assessment of radiation doses to lymphoid tissues.

Risk of Bias Assessments and Evidence Syntheses for Observational Epidemiologic Studies of Environmental and Occupational Exposures: Strengths and Limitations

BACKGROUND

Increasingly, risk of bias tools are used to evaluate epidemiologic studies as part of evidence synthesis (evidence integration), often involving meta-analyses. Some of these tools consider hypothetical randomized controlled trials (RCTs) as gold standards.

METHODS

We review the strengths and limitations of risk of bias assessments, in particular, for reviews of observational studies of environmental exposures, and we also comment more generally on methods of evidence synthesis.

RESULTS

Although RCTs may provide a useful starting point to think about bias, they do not provide a gold standard for environmental studies. Observational studies should not be considered inherently biased vs. a hypothetical RCT. Rather than a checklist approach when evaluating individual studies using risk of bias tools, we call for identifying and quantifying possible biases, their direction, and their impacts on parameter estimates. As is recognized in many guidelines, evidence synthesis requires a broader approach than simply evaluating risk of bias in individual studies followed by synthesis of studies judged unbiased, or with studies given more weight if judged less biased. It should include the use of classical considerations for judging causality in human studies, as well as triangulation and integration of animal and mechanistic data.

CONCLUSIONS

Bias assessments are important in evidence synthesis, but we argue they can and should be improved to address the concerns we raise here. Simplistic, mechanical approaches to risk of bias assessments, which may particularly occur when these tools are used by nonexperts, can result in erroneous conclusions and sometimes may be used to dismiss important evidence. Evidence synthesis requires a broad approach that goes beyond assessing bias in individual human studies and then including a narrow range of human studies judged to be unbiased in evidence synthesis. https://doi.org/10.1289/EHP6980.

Epidemiological Studies of Low-Dose Ionizing Radiation and Cancer: Summary Bias Assessment and Meta-Analysis

BACKGROUND

Ionizing radiation is an established carcinogen, but risks from low-dose exposures are controversial. Since the Biological Effects of Ionizing Radiation VII review of the epidemiological data in 2006, many subsequent publications have reported excess cancer risks from low-dose exposures. Our aim was to systematically review these studies to assess the magnitude of the risk and whether the positive findings could be explained by biases.

METHODS

Eligible studies had mean cumulative doses of less than 100 mGy, individualized dose estimates, risk estimates, and confidence intervals (CI) for the dose-response and were published in 2006-2017. We summarized the evidence for bias (dose error, confounding, outcome ascertainment) and its likely direction for each study. We tested whether the median excess relative risk (ERR) per unit dose equals zero and assessed the impact of excluding positive studies with potential bias away from the null. We performed a meta-analysis to quantify the ERR and assess consistency across studies for all solid cancers and leukemia.

RESULTS

Of the 26 eligible studies, 8 concerned environmental, 4 medical, and 14 occupational exposure. For solid cancers, 16 of 22 studies reported positive ERRs per unit dose, and we rejected the hypothesis that the median ERR equals zero (P = .03). After exclusion of 4 positive studies with potential positive bias, 12 of 18 studies reported positive ERRs per unit dose (P  = .12). For leukemia, 17 of 20 studies were positive, and we rejected the hypothesis that the median ERR per unit dose equals zero (P  = .001), also after exclusion of 5 positive studies with potential positive bias (P  = .02). For adulthood exposure, the meta-ERR at 100 mGy was 0.029 (95% CI = 0.011 to 0.047) for solid cancers and 0.16 (95% CI = 0.07 to 0.25) for leukemia. For childhood exposure, the meta-ERR at 100 mGy for leukemia was 2.84 (95% CI = 0.37 to 5.32); there were only two eligible studies of all solid cancers.

CONCLUSIONS

Our systematic assessments in this monograph showed that these new epidemiological studies are characterized by several limitations, but only a few positive studies were potentially biased away from the null. After exclusion of these studies, the majority of studies still reported positive risk estimates. We therefore conclude that these new epidemiological studies directly support excess cancer risks from low-dose ionizing radiation. Furthermore, the magnitude of the cancer risks from these low-dose radiation exposures was statistically compatible with the radiation dose-related cancer risks of the atomic bomb survivors.

Evaluation of Confounding and Selection Bias in Epidemiological Studies of Populations Exposed to Low-Dose, High-Energy Photon Radiation

BACKGROUND

Low-dose, penetrating photon radiation exposure is ubiquitous, yet our understanding of cancer risk at low doses and dose rates derives mainly from high-dose studies. Although a large number of low-dose cancer studies have been recently published, concern exists about the potential for confounding to distort findings. The aim of this study was to describe and assess the likely impact of confounding and selection bias within the context of a systematic review.

METHODS

We summarized confounding control methods for 26 studies published from 2006 to 2017 by exposure setting (environmental, medical, or occupational) and identified confounders of potential concern. We used information from these and related studies to assess evidence for confounding and selection bias. For factors in which direct or indirect evidence of confounding was lacking for certain studies, we used a theoretical adjustment to determine whether uncontrolled confounding was likely to have affected the results.

RESULTS

For medical studies of childhood cancers, confounding by indication (CBI) was the main concern. Lifestyle-related factors were of primary concern for environmental and medical studies of adult cancers and for occupational studies. For occupational studies, other workplace exposures and healthy worker survivor bias were additionally of interest. For most of these factors, however, review of the direct and indirect evidence suggested that confounding was minimal. One study showed evidence of selection bias, and three occupational studies did not adjust for lifestyle or healthy worker survivor bias correlates. Theoretical adjustment for three factors (smoking and asbestos in occupational studies and CBI in childhood cancer studies) demonstrated that these were unlikely to explain positive study findings due to the rarity of exposure (eg, CBI) or the relatively weak association with the outcome (eg, smoking or asbestos and all cancers).

CONCLUSION

Confounding and selection bias are unlikely to explain the findings from most low-dose radiation epidemiology studies.

Issues in Interpreting Epidemiologic Studies of Populations Exposed to Low-Dose, High-Energy Photon Radiation

This article addresses issues relevant to interpreting findings from 26 epidemiologic studies of persons exposed to low-dose radiation. We review the extensive data from both epidemiologic studies of persons exposed at moderate or high doses and from radiobiology that together have firmly established radiation as carcinogenic. We then discuss the use of the linear relative risk model that has been used to describe data from both low- and moderate- or high-dose studies. We consider the effects of dose measurement errors; these can reduce statistical power and lead to underestimation of risks but are very unlikely to bring about a spurious dose response. We estimate statistical power for the low-dose studies under the assumption that true risks of radiation-related cancers are those expected from studies of Japanese atomic bomb survivors. Finally, we discuss the interpretation of confidence intervals and statistical tests and the applicability of the Bradford Hill principles for a causal relationship.

Epidemiological Studies of Low-Dose Ionizing Radiation and Cancer: Rationale and Framework for the Monograph and Overview of Eligible Studies

Whether low-dose ionizing radiation can cause cancer is a critical and long-debated question in radiation protection. Since the Biological Effects of Ionizing Radiation report by the National Academies in 2006, new publications from large, well-powered epidemiological studies of low doses have reported positive dose-response relationships. It has been suggested, however, that biases could explain these findings. We conducted a systematic review of epidemiological studies with mean doses less than 100 mGy published 2006-2017. We required individualized doses and dose-response estimates with confidence intervals. We identified 26 eligible studies (eight environmental, four medical, and 14 occupational), including 91 000 solid cancers and 13 000 leukemias. Mean doses ranged from 0.1 to 82 mGy. The excess relative risk at 100 mGy was positive for 16 of 22 solid cancer studies and 17 of 20 leukemia studies. The aim of this monograph was to systematically review the potential biases in these studies (including dose uncertainty, confounding, and outcome misclassification) and to assess whether the subset of minimally biased studies provides evidence for cancer risks from low-dose radiation. Here, we describe the framework for the systematic bias review and provide an overview of the eligible studies.