Glaucoma incidence risk in a cohort of Mayak PA workers occupationally exposed to ionizing radiation

Glaucoma is a major cause of visual impairment, and secondary glaucoma manifested as neovascular glaucoma has long been known to occur following high-dose fractionated radiation therapy. In contrast, little is known as to whether ionizing radiation exposure causes primary glaucoma, except that a single study in Japanese atomic bomb survivors has reported a significantly increase risk. Therefore, the effect of lower dose and lower dose rate remains unclear. Here we report that in Russian Mayak Production Association workers occupationally exposed to chronic radiation for prolonged periods, incidence of total primary glaucoma and primary open-angle glaucoma is significantly associated with various non-radiation factors (sex, attained age, and cataract diagnosed prior to glaucoma), but neither with cumulative dose from external γ-rays nor with cumulative neutron dose nor with the unweighted sum of cumulative γ and neutron doses. The present results suggest for the first time that chronic radiation exposure does not cause primary glaucoma, although the analyses need to be made in other cohorts exposed at various dose and dose rate.


Dosimetry.
For analyses, absorbed doses from external γ-rays and neutrons were used based on the Mayak Worker Dosimetry System 2008 (MWDS-2008) 17 . MWDS-2008 provides absorbed doses in 18 organs and tissues (lungs, red bone marrow, breast, bladder, kidney, skin, stomach, gonads (ovaries/testicles), thyroid, uterus, esophagus, bone surface, skeleton, liver, small intestine, large intestine, spleen), but eye dose is not available. This study therefore used individual brain absorbed doses from external γ-rays and neutrons 18 . It should be noted that there is no considerable difference among types of organ absorbed doses from external radiation. Mean cumulative brain absorbed dose from external γ-rays was 0.46 ± 0.67 Gy in males and 0.36 ± 0.56 Gy in females. Only 4083 Mayak PA workers (18.2%) were exposed to neutrons. For these workers, the mean cumulative brain absorbed external radiation dose from neutrons was 0.0016 ± 0.0043 Gy in males and 0.0016 ± 0.0050 Gy in females. For those workers who had not been exposed to neutrons, doses from neutrons were not estimated and in the present study they were considered as 'unmeasured 0.00' . The mean unweighted sum of cumulative γ and neutron brain absorbed doses was 0.46 ± 0.67 Gy in males and 0.37 ± 0.56 Gy in females. Distribution of the study cohort workers by radiation dose is presented in Table S1.

Statistical analysis.
To perform analyses of the data on the study cohort, the statistical techniques and methods were consistent with previous studies [19][20][21] . The dataset for analyses was restricted to a period of residence in Ozyorsk, because information on diseases, the results of annual eye examinations and non-radiation factors was unavailable for migrants after they had left the city. This study excluded 43 workers with acute radiation sickness due to acute single high dose-rate γ + neutron exposure, and also excluded 614 workers with missing medical information due to lost medical charts and 77 workers with glaucoma diagnosed prior to employment at the Mayak PA.
As in the previous studies [19][20][21] , the analysis provided estimates of relative risk (RR) for categories designated for one or more variables while adjusted for other variables. RR was computed based on maximum likelihood using the AMFIT module of the EPICURE software 22 . In order to test a statistical significance, 95% confidence intervals (CIs) for RRs, and p values were computed using likelihood-based techniques integrated in the AMFIT module.
Analysis of glaucoma incidence associated with dose from external γ radiation adjusted for neutron dose was set as the default analysis. It considered the whole cohort. The neutron dose adjustment was included using stratification, and the neutron dose was treated as a categorical variable. In consistence with the previous studies [19][20][21] , workers who were assumed not to have been exposed to neutrons were not excluded from the analyses; they were included in the 'unmeasured 0.00' category. Analyses of the linear trend with neutron dose and unweighted γ + neutron dose were conducted as sensitivity analyses. The analysis of the linear trend with neutron dose was restricted to workers with measured (or known, or estimated) doses from neutrons (4083 workers). The analysis of the linear trend with unweighted γ + neutron dose considered the whole cohort. To estimate the unweighted γ + neutron dose, the unmeasured neutron dose was regarded as equal to 0.00.
In addition, sensitivity analyses were performed to investigate effects of dose lagging (0, 10, 15, 20-year lag periods) and additional non-radiation factors, such as diagnosed cataract or cataract removal surgery prior to a report of glaucoma, alcohol consumption and smoking status, smoking index (<10, 10-20, >20 pack-years) rather than smoking status parameter on the observed risk estimates in relation to external radiation exposure.
Using log-linear modifications of excess relative risk per unit dose we tested whether such factors as sex, age at first employment and attained age of workers could modify the radiation risk of glaucoma incidence. All p-values were for two-sided test. Differences were judged to be significant at p < 0.05.
Since hypertension, BMI and DM are standard risk factor for glaucoma, all these factors were considered in the default analysis (with 5-year lag) via stratification. We excluded these variables as well as the variable of neutron dose, one by one thus investigating the effect of removing these variables from the model.
Definitions and categorizations of risk factors considered in this study are provided in details in Supplementary Section. Information about whether a worker had a diagnosis of DM and cataract and whether he/she underwent cataract surgery was available for every worker of the study cohort. Cases with cataract and cataract removal surgery were included in the analyses if they were registered before glaucoma was diagnosed in a worker.
An additional analysis was performed to provide odds ratios of various risk factors for primary glaucoma and POAG. The analysis was based on a logit regression and is described in details in Supplementary Section. ethical statement. The study was reviewed and approved by the SUBI Institutional Review Board who confirmed that no signed consents were needed from members of the study cohort, because the present record-based epidemiological study did not require any contact with cohort members.

Results
By the end of the follow-up period, 476 cases of primary glaucoma were registered over 544,236 person-years of follow-up. Of these, RR of glaucoma incidence was calculated each for primary glaucoma and POAG, but not for PACG due to a small number of cases (15 cases). www.nature.com/scientificreports www.nature.com/scientificreports/ Risk analysis started with assessment of its associations with known non-radiation factors promoting glaucoma (sex, attained age, smoking, alcohol consumption, cataract and cataract removal surgery, birth cohort, calendar period of diagnosis date, age at first employment, hypertension, BMI, and DM). Tables 2 and S2 summarize the results of the analysis of glaucoma incidence in relation to non-radiation factors.
RR for incidence of both primary glaucoma and POAG was significantly lower in females than in males. The incidence risks of primary glaucoma and POAG increased significantly with attained age of workers of both sexes. Incidence of primary glaucoma and POAG was not significantly associated with a calendar period of first employment except for the significantly lower incidence among females hired in 1969-1972 (RR = 0.30 (95% CIs: 0.05, 0.97) and 0.31 (95% CIs: 0.05, 0.99), respectively), which could likely be explained by a small number of cases falling into this group (2 cases). Incidence of primary glaucoma and POAG was significantly associated with certain calendar periods for workers of both sexes. RR for incidence of primary glaucoma and POAG was insignificant in groups of workers by age of first employment at the Mayak PA except for decreased risk in males hired at age 25-30 years compared to those hired before age 20 years. None of smoking status, alcohol consumption status, smoking index and BMI affected incidence of primary glaucoma and POAG in both males and females. Incidence of primary glaucoma and POAG was notably higher in workers diagnosed with cataracts (preceding glaucoma) than those without cataracts: RR in males was 12.09 (95% CIs: 9.08, 16.23) and 12.56 (95% CIs: 9.40, 16.95), respectively; RR in females was 8.43 (95% CIs: 5.54, 13.19) and 9.45 (95% CIs: 6.09, 15.16), respectively. Increased albeit insignificant risk estimates were demonstrated for incidence of primary glaucoma and POAG among workers after cataract removal surgery as well as those with diagnosed DM.
Next, RRs were analyzed for incidence of primary glaucoma and POAG in relation to cumulative dose from external γ-rays, neutrons and γ + neutron exposure. RRs for incidence of primary glaucoma and POAG resulting from all these analyses were insignificant in all dose groups (Table 3). Using a linear model, the primary glaucoma incidence risk, in particular, risk of POAG, was analyzed in relation to external γ-dose (Table 4, Fig. 1). Incidence of primary glaucoma and POAG was not significantly associated with cumulative dose from external γ-rays, regardless of adjustment for neutron dose with ERR/Gy of 0.01 (95% CIs: −0.13, 0.19) and 0.0003 (95% CIs: −0.13, 0.19) when adjusted and 0.01 (95% CIs: −0.12, 0.20) and 0.01 (95% CIs: −0.13, 0.19) when unadjusted, respectively. Moreover, no significant associations were found with cumulative neutron dose and with cumulative γ + neutron dose for either primary glaucoma or POAG (Table 4, Fig. 1  www.nature.com/scientificreports www.nature.com/scientificreports/ periods), inclusion of additional adjustments for non-radiation factors (smoking index, cataract and cataract surgery, smoking status and alcohol consumption) did not significantly modify the observed findings. Exclusion of adjustments for hypertension, BMI and DM did not considerably affect the risk estimate either. Sex, attained age and age at first employment at the enterprise did not significantly modify risk for primary glaucoma (p = 0.27, p = 0.38 and p = 0.27, respectively) and for POAG (p = 0.28, p > 0.5 and p = 0.31, respectively).

Discussion
This is a retrospective cohort study with a number of advantages, such as the large size of the study cohort, information on both sexes (with 25.4% of females), the extensive follow-up period (up to 70 years), availability of medical information and data on non-radiation risk factors over the entire follow-up period. The main advantage of the present study are the mandatory annual health examinations over the entire follow-up. The annual health examinations were performed using a standard protocol and included a mandatory check by an ophthalmologist and an ophthalmological examination using conventional techniques regardless of workers' sex, age, working site, occupation, radiation dose, etc. www.nature.com/scientificreports www.nature.com/scientificreports/ As expected, the present study found significant associations between primary glaucoma and some known non-radiation risk factors such as sex, attained age, cataract diagnosed before glaucoma manifestation, and cataract removal surgery. These results are in line with those observed in other studies 3,4,[23][24][25] .
The present study did not reveal any significant associations between primary glaucoma incidence and doses from external γ + neutron radiation exposure and its separate components (γ-rays, neutrons) whether adjusted for each other or not. Further one-by-one exclusion of known risk factors (hypertension, BMI, DM) in the model did not modify the ERR/Gy estimate either for primary glaucoma or for POAG (Table 4). While investigating effects following fractionated exposures at dose above tens Gy (each fraction delivered at high dose rate), a number of studies have demonstrated neovascular glaucoma as one of the normal tissue complications following ocular or head and neck radiotherapy, which have been mediated by plaque formation in microcirculation and further neovascularization [5][6][7][8][9][10][11] . As shown in Table 5, following single, acute exposure at several Gy, glaucoma risk has been reported in Japanese atomic bomb survivors: Yamada et al. first reported that incidence of glaucoma in aggregate (i.e., without distinguishing any type of glaucoma) significantly decreases with increasing radiation dose 26 . In a subsequent study, and as shown in Table 5, Kiuchi et al. reported an Odds ratio at 1 Gy of 1.31 (95% CIs: 1.11, 1.53, p = 0.001) for normal-tension glaucoma, albeit with no significant associations between radiation doses and other types of glaucoma 12 , and the authors in their recent paper 13 confirmed the significant odds ratio at 1 Gy for NTG for both right and left eyes (Table 5). However, Kiuchi et al. 12 note that it was a screening study and the cohort member participation rate in that study was relatively low, underlining a need to interpret these findings with caution. Following occupational protracted exposures at low dose, there was no literature available until Little et al. very recently reported an insignificantly decreased risk for self-reported glaucoma in aggregate in the US radiologic technologists (USRT) cohort 14 (and see also Table 5), which was based solely on questionnaire information obtained from medically literate workers, without any ophthalmological validation. Accordingly, it remains unclear whether members of the USRT cohort and Mayak workers exhibit a significantly increased risk for normal-tension glaucoma as reported in atomic bomb survivors. As such, our future studies include evaluation of the POAG risk separately for normal-tension type and high-tension type based on the measured IOP data.
Biological mechanisms underlying radiation-induced glaucoma remain unclear 27 . The most likely mechanisms that have been suggested thus far involve impaired retinal circulation since some data have demonstrated the impaired circulation at 30-45 years after radiation exposure 28 as well as retinal vessel atherosclerosis 29 . In addition, other possible mechanisms of radiation glaucoma may include neural excitotoxicity 30 , inflammation 31 , genetic predisposition [32][33][34][35] , autoimmune reactions 36 and retinal circulation disorders 37,38 . Some investigators believe that the most likely of them might be retinal circulation disorders 12   www.nature.com/scientificreports www.nature.com/scientificreports/ radiation-induced blood flow decrease in the retina 30 or 45 years after the exposure 28 , as well as of the significant association of retinal arteriosclerosis with radiation exposure in atomic bomb survivors 55 years after the exposure 29 .
In conclusion, significant association was found neither between cumulative dose from external γ-rays nor cumulative neutron dose nor the unweighted sum of cumulative γ and neutron doses and incidence of primary glaucoma and POAG in the cohort of Mayak PA workers. Glaucoma incidence was significantly associated with non-radiation risk factors (sex, attained age, and cataract diagnosed prior to glaucoma). Further analyses to distinguish normal-tension type and high-tension type are needed.

Data Availability
The dataset is the intellectual property of the Southern Urals Biophysics Institute, Ozyorsk, Chelyabinsk Region, 456780, Russia. For privacy reasons it is not publicly available. Any access to the Mayak Workers Cohort must be approved by SUBI's Institutional Review Board. These restrictions on data availability are imposed by Federal Act No. 323 of 21 November 2011 on the basics of health care for Russian citizens and Federal Act No. 152 of 27 July 2014 on personal data. To request the data used in the presented analyses, please, contact Dr. Tamara Azizova, the head of the clinical department of the Southern Urals Biophysics Institute. Any access to the Mayak Workers Cohort must be approved by SUBI's Institutional Review Board. Please, contact Dr. Valentina Rybkina, MD, leading researcher of SUBI, IAB member (rybkina@subi.su; +73513029953).