Systematic Review and Meta-analysis of Circulatory Disease from Exposure to Low-Level Ionizing Radiation and Estimates of Potential Population Mortality Risks

Background: Although high doses of ionizing radiation have long been linked to circulatory disease, evidence for an association at lower exposures remains controversial. However, recent analyses suggest excess relative risks at occupational exposure levels. Objectives: We performed a systematic review and meta-analysis to summarize information on circulatory disease risks associated with moderate- and low-level whole-body ionizing radiation exposures. Methods: We conducted PubMed/ISI Thomson searches of peer-reviewed papers published since 1990 using the terms “radiation” AND “heart” AND “disease,” OR “radiation” AND “stroke,” OR “radiation” AND “circulatory” AND “disease.” Radiation exposures had to be whole-body, with a cumulative mean dose of < 0.5 Sv, or at a low dose rate (< 10 mSv/day). We estimated population risks of circulatory disease from low-level radiation exposure using excess relative risk estimates from this meta-analysis and current mortality rates for nine major developed countries. Results: Estimated excess population risks for all circulatory diseases combined ranged from 2.5%/Sv [95% confidence interval (CI): 0.8, 4.2] for France to 8.5%/Sv (95% CI: 4.0, 13.0) for Russia. Conclusions: Our review supports an association between circulatory disease mortality and low and moderate doses of ionizing radiation. Our analysis was limited by heterogeneity among studies (particularly for noncardiac end points), the possibility of uncontrolled confounding in some occupational groups by lifestyle factors, and higher dose groups (> 0.5 Sv) generally driving the observed trends. If confirmed, our findings suggest that overall radiation-related mortality is about twice that currently estimated based on estimates for cancer end points alone (which range from 4.2% to 5.6%/Sv for these populations).


Methods
The Life Span Study mortality data of Shimizu et al. (2010) 402,404,; and (c) all other circulatory disease 401,403,405,. We shall concentrate in Supplemental Material A on analysis of deaths in which circulatory disease was mentioned as a contributory or underlying cause of death for these three endpoints.
The expected number of such deaths in the Japanese A-bomb survivor data in with disease endpoint i ( i =stroke, heart disease, other circulatory disease), stratum j and dose group d with average age at exposure, e , years, average years since exposure, y , and colon dose, D , is assumed to be given by: The stratum-specific background rates  Table S5) and also simultaneously across all three disease endpoints (stroke, heart disease, other circulatory disease) (Supplemental Material, Table S6) in order to test for heterogeneity of effect. Equation (A1) describes a linear-quadratic relative risk model.

Results
There is a highly significant increasing dose response for all three endpoints (p<0.001) (Supplemental Material, Table S5). There is a statistically significant effect of sex (p=0.022) for heart disease (Supplemental Material, Table S5), but otherwise there is no significant modification of dose response by sex or time since exposure for any endpoint (p>0.05), nor is there any significant curvature in the dose response for any endpoint (p>0.1) (Supplemental Material, Table S5). For stroke and other circulatory disease there are statistically significant (p<0.01) reductions in relative risk with increasing age at exposure, although this is not the case for heart disease (p=0.391) (Supplemental Material, Table S5). As shown by Supplemental Material, Table S6, the main effect excess relative risk coefficients are highly statistically significantly different (p<0.001) for the three endpoints. There is marginally statistically significant (p=0.053) heterogeneity also in the magnitude of the reductions of relative risk with increasing age at exposure.

Yes
See "Data and methods/Data and meta-analysis" section para 1, sentence 1: Medline/ISI Thompson search using the terms "radiation" + "heart" + "disease" or "radiation" + "stroke" or "radiation" + "circulatory" + "disease"; only peer reviewed papers from 1990 Effort to include all available studies, including contact with authors Yes Contact with the authors of all studies selected for the final analysis (as in Table 1) was attempted: however, the requested data proved to be unobtainable in all but three cases (Japanese LSS mortality, IARC 15country, EdF workers).

Databases and registries searched
Yes See "Data and methods/Data and meta-analysis" section para 1, sentence 1: Medline and ISI Thompson (Web of Knowledge) databases were used. Search software used, name and version, including special features used N/A Hand search was used. Use of hand searching (e.g., reference lists of obtained articles) Yes See "Data and methods/Data and meta-analysis" section para 1, sentence 5.

List of citations located and those excluded, including justification
No Given the numbers of searched papers (>6500), it is impossible to justify the inclusion or exclusion of each study. Among those that met preliminary requirements a secondary exclusion based on study quality was used, that would possibly have excluded some studies, but this was not in fact the case. 6 Supplemental Material, Table S1 (cont.) Category Met Where met and described, or if not met, reasons why not Method of addressing articles published in languages other than English Yes See "Data and methods/Data and meta-analysis" section para 1, sentence 7. As we state there, "Although there was no restriction to publication in English, based on assessment of the titles and abstracts the only studies meeting our criteria were published in that language.".

Method of handling abstracts and unpublished studies
No See "Data and methods/Data and meta-analysis" section para 1, sentence 3. As we state in the Discussion, para 12 "We chose to limit our results to studies published as full papers … We judge that the most important and high quality studies are likely to be published as full papers." Description of any contact with authors Yes As above, contact with the authors of all studies used in the final analysis (as in Table 1) was attempted: however, the requested data proved to be unobtainable in all but three cases (Japanese LSS mortality, IARC 15 country, EdF workers). Description of relevance or appropriateness of studies assembled for assessing the hypothesis to be tested

Yes
We discuss the appropriateness of the results of the paper to general unselected populations in Discussion para 5. Rationale for the selection and coding of data (e.g., sound clinical principles or convenience)

Yes
The data used is the obvious epidemiological data to use for the assessment of circulatory disease risk. Documentation of how data were classified and coded (e.g., multiple raters, blinding and inter-rater reliability)

Yes
There were two raters, MPL and WZ, assessing studies blind to each other. An objective scoring system was used (see Supplemental Material  Table S2), so there was 100% agreement. Assessment of confounding (e.g., comparability of cases and controls in studies where appropriate)

Yes
There is extensive discussion of the possibility of confounding, in particular by lifestyle factors, in the Discussion paras 9-10. Assessment of study quality, including blinding of quality assessors; stratification or regression on possible predictors of study results

No
We do not regard the blinding of quality assessors as a sensible requirement. There are numerous clues within each publication as to the cohort being studied and the principal authors, so that blinding would be all but impossible. The small numbers of studies, and the limited information available limit the usefulness of meta-regression; in any case we do not regard this form of analysis as very sensible given the likelihood of ecological bias. 7 Supplemental Material, Table S1 (cont.) Category Met Where met and described, or if not met, reasons why not Assessment of heterogeneity Yes This was assessed at some length (using fixed-and random-effects models), as described in the Statistical Methods for Meta-Analysis section of the main paper. Description of statistical methods (e.g., complete description of fixed or random effects models, justification of whether the chosen models account for predictors of study results, dose-response models, or cumulative metaanalysis) in sufficient detail to be replicated

Yes
The statistical methodology (fixed-and random-effects model etc) is described at some length in the Statistical Methods for Meta-Analysis section of the main paper.

Provision of appropriate tables and graphics
Yes See Tables 1-5, Supplemental Material, Tables S1-S7 and Figure 1.

Graphic summarising individual study estimates and overall estimate
Yes No (such information is given in Table 1).

Table giving descriptive information for each study included
Yes See Table 1 and Supplemental Material, Table S2.

Results of sensitivity testing (e.g., subgroup analysis)
Yes See Supplemental Material, Table S4.

Indication of statistical uncertainty of findings
Yes See Tables 1-5, Supplemental Material, Tables S4-S5. Quantitative assessment of bias (e.g., publication bias) Yes A descriptive funnel plot (see Figure 1) demonstrates adequately that there is little or no publication or selection bias. This was supplemented more formally by use of the publication/selection-bias test of Egger et al. and Steicher, and assessment of bias using the trim-and-fill method of Duval and Tweedie (Supplemental Material, Table S3), all of which also suggested little or no publication or selection bias. Justification of exclusion (e.g., exclusion of non-English language citations)

Yes
"Data and methods/Data and meta-analysis" section para 1, sentence 3. As we state in the Discussion, para 12 "We chose to limit our results to studies published as full papers and referenced in Medline or ISI Thompson. We judge that the most important and high quality studies are likely to be published as full papers."

Assessment of quality of included studies
Yes See Supplemental Material, Table S2.

Consideration of alternative explanations for observed results
Yes There is extensive discussion of the possibility of confounding, in particular by lifestyle factors, in the Discussion paras 9-10. 8 Supplemental Material, Table S1 (cont.) Category Met Where met and described, or if not met, reasons why not Generalization of the conclusions (i.e., appropriate for the data presented and within the domain of the literature review)

Yes
There is extensive discussion of the generalization of the conclusions to unselected general populations, given the possibility of (a) confounding, in particular by lifestyle factors, and (b) selection in the study cohorts considered, in the Discussion paras 5, 9-10.

Guidelines for future research
Yes See Discussion final para (16).

Dosimetry (out of 5)
Starting with a score of 5 points: subtract 1 point if dosimetry is not based on concurrent registry-derived records; subtract 1 point if some substantial component of dose is not assessed (e.g., neutrons); subtract 2 points if dosimetry is based on area-based assessments of exposure; subtract 1 point if no attempt is made to correct for dose error.

Endpoint (out of 5)
Starting with a score of 5 points: subtract 1 point if the follow-up is substantially incomplete (>5% of deaths/cases in cohort are lost to follow-up); subtract 1 point if follow-up is not based on local (regional), national or cohort-based registers; subtract 1 point if date of ascertainment of disease incidence/mortality may be substantially in error (> 1 year); subtract 1 point if there is no clinical review of pathology data to verify diagnosis of mortality/morbidity.

Selection criteria (out of 5)
Starting with a score of 5 points: subtract 1 point if the selection may result in omission of potentially highly exposed persons; subtract 2 points if the selection does not stringently exclude workers with missing dose records; subtract 2 points if the selection does not exclude workers working for a short time (< 6 months).

Lifestyle/circulatory disease risk factors assessed (out of 5)
Starting with a score of 5 points: subtract 1 point if there is no information on (and adjustment for) socioeconomic status; subtract 1 point if there is no information on (and adjustment for) cigarette smoking and alcohol consumption; subtract 1 point if there is no information on (and adjustment for) obesity; subtract 1 point if there is no information on (and adjustment for) diabetes; subtract 1 point if there is no information on (and adjustment for) blood pressure.

Statistical analysis (out of 5)
Starting with a score of 5 points: subtract 1 point if there is no attempt to assess interactions of dose response with age at exposure, attained age or time since exposure; subtract 1 point if the method of analysis is unclear; subtract 1 point if an inappropriate lag period is used (outside the range 5-10 years).
The mean score was obtained as the arithmetic average of these five component scores.  Analysis based on mortality from heart failure. c Analysis based on mortality from heart failure and other heart disease. d Analysis based on morbidity from cerebrovascular disease, with a 10-year lag. e Analysis based on morbidity from hypertension, hypertensive heart disease and aortic aneurysm.

Table S3. Assessments of publication/selection bias using methods of Egger et al. (1997) and Steichen (1998), and bias-corrected ERR coefficients using method of Duval and Tweedie (2000).
f Analysis based on morbidity from hypertension, disease of arteries, arterioles and capillaries, veins, lymphatic vessels and lymph nodes. g Analysis based on mortality from rheumatic heart disease and circulatory disease apart from heart disease and cerebrovascular disease.     Little et al. (2008;). All data are in relation to underlying cause of death, unless otherwise indicated.

Data
Reference Average heart/brain dose (range) (Sv) Numbers in cohort (person years follow-up) Endpoint (mortality unless otherwise indicated) Excess relative risk Sv -1 (and 95% CI) French-UK childhood cancer study Tukenova et al. 2010