Risk of Developing Non-Cancerous Central Nervous System Diseases Due to Ionizing Radiation Exposure during Adulthood: Systematic Review and Meta-Analyses

Background: High-dose ionizing radiation (IR) (>0.5 Gy) is an established risk factor for cognitive impairments, but this cannot be concluded for low-to-moderate IR exposure (<0.5 Gy) in adulthood as study results are inconsistent. The objectives are to summarize relevant epidemiological studies of low-to-moderate IR exposure in adulthood and to assess the risk of non-cancerous CNS diseases. Methods: A systematic literature search of four electronic databases was performed to retrieve relevant epidemiological studies published from 2000 to 2022. Pooled standardized mortality ratios, relative risks, and excess relative risks (ERR) were estimated with a random effect model. Results: Forty-five publications were included in the systematic review, including thirty-three in the quantitative meta-analysis. The following sources of IR-exposure were considered: atomic bomb, occupational, environmental, and medical exposure. Increased dose-risk relationships were found for cerebrovascular diseases incidence and mortality (ERRpooled per 100 mGy = 0.04; 95% CI: 0.03–0.05; ERRpooled at 100 mGy = 0.01; 95% CI: −0.00–0.02, respectively) and for Parkinson’s disease (ERRpooled at 100 mGy = 0.11; 95% CI: 0.06–0.16); Conclusions: Our findings suggest that adult low-to-moderate IR exposure may have effects on non-cancerous CNS diseases. Further research addressing inherent variation issues is encouraged.


Introduction
Recent decades have seen an increase in the exposure of the overall population to ionizing radiation (IR), especially due to the widespread use of medical imaging procedures in economically developed countries [1]. Nowadays, the average annual effective IR-dose is estimated to be around 3.0 millisievert (mSv) per person, including 20% from medical exposure [2]. The latter tends to have increased from year to year due to the use of new technological imaging from around 0.3 mSv in 1993 to 0.6 mSv in 2021 [2].
The adverse health effects following exposure to IR have been the subject of a large amount of scientific research, mainly focusing on carcinogenic risks [3]. However, several epidemiological studies have highlighted the non-cancerous detrimental impact of high or moderate IR doses on the central nervous system (CNS) [4][5][6], and radiation-induced cognitive injury is becoming an increasingly important research subject [7,8]. Non-cancerous CNS disorders are a large and complex group of diseases, including mental and behavioral disorders, diseases of the CNS, and trauma, and they have multifactorial origins. The       Individual estimate radiation doses ranged from 0 to >3 Gy (86% of the cohort members received <0. In the observational period, the SO staff were exposed to external irradiation at the dose range of 0-54.3 mSv (mean ± SD: 19.5 ± 12.8 mSv), to internal irradiation at the dose range of 0-2.4 mSv (mean ± SD: 0.4 ± 0.5 mSv), and to total irradiation at the dose range of 0-56.7 mSv (mean ± SD: 19.9 ± 13.0 mSv) Key findings of the 21 studies focusing on diseases of the nervous system can be found in Table 2.

Nuclear Workers and Uranium Miners
Out of the 21 studies that considered this outcome, 13 were on nuclear workers and uranium miners. The majority of them did not report any statistically significant results, whether the authors compared the mortality of workers to that of an external reference population (SMR) or they assessed dose-response relationships (ERR) [33,35,36,39,42,44,46].
In a cohort of 4977 U.S. mound workers potentially exposed to external or internal (polonium-210, plutonium isotopes, or tritium) radiation (mean dose from external radiation: 26.1 mSv; max: 939.1 mSv; mean lung dose from external and internal radiation combined: 100.1 mSv; max: 17.5 Sv; mean liver dose from external and internal radiation: 34.6 mSv; max: 2.3 Sv), the mortality due to diseases of the nervous system was not different from that of the general population, regardless of the radiation status of the workers or the type of radionuclides for those with intakes, but a positive trend was suggested as SMRs increased with increasing categories of occupational cumulative dose primarily due to photons (p = 0.03) [42]. In a cohort of 26,328 Los Alamos National Laboratory workers exposed to a combination of photons, neutrons, tritium, and plutonium (among which 17,053 workers were monitored for a combination of external and internal sources for plutonium; brain radiation absorbed dose: mean: 11.6 mGy; max: 760 mGy), Boice et al. (2021) reported among the whole cohort a non-significant SMR for nervous system diseases compared with national rates based on 815 deaths, but a borderline significant increase in mortality due to Parkinson's disease was observed (SMR = 1.16; 95% CI: 1.00, 1.34; n deaths = 193), and a positive dose-response relationship was suggested (ERR at 100 mGy = 0.16; 95% CI: −0.07, 0.40; n deaths = 273) [33]. In a cohort of 22,377 Mayak workers exposed to chronic IR (mean cumulative dose from external gamma rays absorbed in the brain: 0.46 Gy (max: 8.01 Gy) for men and 0.36 Gy (max: 6.14 Gy) for women), a statistically significant linear association was found between the incidence of Parkinson's disease and cumulative gamma dose after adjusting for sex and attained age (ERR per Gy = 1.02; 95% CI: 0.59, 1.63; n diseases = 300) [12]. A significant decrease in mortality was reported among 53,698 U.S. nuclear power plant industry workers (mean cumulative dose: 25.7 mSv) compared with the U.S. general population (SMR= 0.50; 95% CI: 0.31, 0.77; n deaths = 20), whereas a positive and statistically significant, but highly imprecise, dose-response relationship was observed (ERR per Sv= 46.8; 95% CI: 1.51, 242; n death = 20) [49]. Later, Boice et al. (2021) also reported a significant decrease in mortality from diseases of the nervous system compared with national rates in 135,193 U.S. nuclear power workers (mean dose to the brain: 33.2 mGy; max: 0.83 Gy) (SMR = 0.82, 95% CI: 0.76, 0.89; n deaths = 673), and the authors observed a positive non-significant dose-response relationship for Parkinson's disease mortality (ERR at 100 mGy = 0.24; 95% CI: −0.02, 0.50; n deaths = 140) [32].
In studies on uranium miner, miller, or processing workers, inconsistent results were observed, as a significantly increased mortality from diseases of the nervous system was observed in 2,930 uranium miners and millers of Grants (New Mexico) potentially exposed to radon, radon decay products, uranium dust and mill products (exposure assessment: NA) compared with the general U.S. population (SMR= 1.60; 95% CI: 1.01, 2.39; n deaths = 23) [47], while a significantly decreased mortality was found in 16,236 male Eldorado uranium workers exposed to gamma rays (dose: 52.2 mSv) or/and radon decay products (100.2 WLM) compared with the general Canadian male population (SMR = 0.66; 95% CI: 0.51, 0.85; n deaths = 61) [45] and in 35,204 male underground uranium miners of the WISMUT AG exposed to radon (mean: 364 WLM) or/and crystalline silica (mean: 7.6 mg/m 3 years) compared to the general male population in Eastern Germany (SMR = 0.73; 95% CI: 0.62, 0.85; n deaths = 163) [34].

Nuclear Weapons Test Participants
Concerning nuclear weapons test participants, a significant decrease in mortality from diseases of the nervous system was observed among 114,270 male U.S. military participants in atmospheric tests in Nevada and the Pacific from 1945 to 1962 (mean NuTRIS film badge gamma radiation dose: 6 mSv; max: 908 mSv) compared to the general male population of the U.S. (SMR = 0.84; 95% CI: 0.81, 0.88; n deaths = 1871) [53]. The 10,983 male Australian participants in the British nuclear tests conducted in Australia between 1952 and 1957 (mean radiation exposure: 2.8 mSv; max > 50 mSv) showed similar mortality to that of the general male population for diseases of the nervous system (SMR = 1.02; 95% CI: 0.78, 1.32; n deaths = 59), but showed a non-significantly higher mortality for motor neuron disease (SMR = 1.24; 95% CI: 0.71, 2.02; n deaths = 16) [54]. Rahu et al. (2014) reported an increased risk of diseases of the nervous system in a cohort of 3680 Estonian Chernobyl cleanup workers recruited between 1986 and 1991 to the Chernobyl area by the Soviet authorities for decontamination, building, and other related activities compared to a population-based cohort of 7631 unexposed Estonian men (RR = 1.13, 95% CI: 1.06, 1.21) [70]. However, the accuracy of the diagnosis and the representativeness of the unexposed cohort are an issue of this study.

Medical Workers
In a cohort of 109,019 U.S. medical and associated radiation workers exposed to Xand gamma rays (mean cumulative absorbed dose to the brain: 18.9 mGy; max: 1.08 Gy), monitored between 1965 and 1994 and followed-up until 2016, a significant decrease in mortality from diseases of the nervous system (SMR = 0.72; 95% CI: 0.65, 0.78; n deaths = 471) was found, but not for mortality from Parkinson's disease (SMR= 0.82; 95% CI: 0.66, 1.02; n deaths = 84) when compared with the general population, and a non-significant positive dose response for Parkinson's disease was found (ERR at 100 mGy = 0.17; 95% CI: −0.20, 0.54; n deaths = 87) [58]. Furthermore, no significant increase in mortality from diseases of the nervous system nor for Alzheimer's disease was found among 34,912 U.S. male radiologists (exposure assessment: NA) when compared to 47,497 male psychiatrists or to the general population (including both men and women) [62], and no increased risk of mortality from diseases of the nervous system was shown in 41,486 male U.S. physicians who had performed fluoroscopy-guided interventional procedures (FGIP) (exposure assessment: NA) compared to 46,846 male psychiatrists [61]. However, mortality due to degenerative diseases of the nervous system (ICD-9: 331.1-337.9/ICD-10: G31-G32) appeared higher in U.S. radiologists (exposure assessment: NA) compared to the general population (SMR = 1.50; 95% CI: 1.09, 1.90; n deaths = 53) and compared to psychiatrists (RR = 1.39; 95% CI: 0.96, 2.01) even if it was not statistically significant in the latter case [62]. An increased risk of mortality from diseases of the nervous system was found in a cohort of 27,011 medical X-ray workers employed between 1950 and 1980 (average radiation exposure for the workers employed until 1969: 551 mGy; employed between 1970 and 1980: 82 mGy) in China compared to a group of 25,782 non-exposed medical specialists (RR = 2.10; 95% CI: 1.20, 3.10) [64]. On the other hand, a significant decrease in mortality from disease of the nervous system was reported among male diagnostic medical radiation workers (exposure assessment: NA) in South Korea compared with the general population (SMR = 0.54; 95% CI: 0.33, 0.88; n deaths = 16) [60].

Overall Meta-Analyses among Occupational Studies
An overall SMR was calculated based on the 16 articles presented above. A decrease in mortality due to nervous system diseases was found (SMR pooled = 0.86; 95% CI: 0.79, 0.93), with high heterogeneity between studies (Q = 65.23, p < 0.0001) and no publication bias (p = 0.91), but with a high percentage of variation across studies that is due to heterogeneity rather than chance (I 2 = 77.02%) ( Figure 2). Sensitivity analyses where studies or different workers groups were excluded one by one were conducted with persistent heterogeneity each time, as well as sensitivity analyses in which studies that did not report ICD codes were removed. bias (p = 0.91), but with a high percentage of variation across studies that is due to heterogeneity rather than chance (I 2 = 77.02%) ( Figure 2). Sensitivity analyses where studies or different workers groups were excluded one by one were conducted with persistent heterogeneity each time, as well as sensitivity analyses in which studies that did not report ICD codes were removed. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from diseases of the nervous system in IR-exposed populations compared with general populations as reference.
Regarding internal comparison, a RRpooled was computed from three studies (all including medical radiation workers [61,62,64]), showing no increased risk of death due to nervous system diseases in IR-exposed people compared to unexposed controls (RRpooled = 1.17; 95% CI: 0.85, 1.61), with high heterogeneity (Q = 9.38, p = 0.009), I 2 = 78.67%, and no suspected publication bias (p = 0.21) (Figure 3).  Figure 2. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from diseases of the nervous system in IR-exposed populations compared with general populations as reference.
Regarding internal comparison, a RR pooled was computed from three studies (all including medical radiation workers [61,62,64]), showing no increased risk of death due to nervous system diseases in IR-exposed people compared to unexposed controls (RR pooled = 1.17; 95% CI: 0.85, 1.61), with high heterogeneity (Q = 9.38, p = 0.009), I 2 = 78.67%, and no suspected publication bias (p = 0.21) (Figure 3). geneity rather than chance (I 2 = 77.02%) ( Figure 2). Sensitivity analyses where studies or different workers groups were excluded one by one were conducted with persistent heterogeneity each time, as well as sensitivity analyses in which studies that did not report ICD codes were removed. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from diseases of the nervous system in IR-exposed populations compared with general populations as reference.
Regarding internal comparison, a RRpooled was computed from three studies (all including medical radiation workers [61,62,64]), showing no increased risk of death due to nervous system diseases in IR-exposed people compared to unexposed controls (RRpooled = 1.17; 95% CI: 0.85, 1.61), with high heterogeneity (Q = 9.38, p = 0.009), I 2 = 78.67%, and no suspected publication bias (p = 0.21) (Figure 3).  Figure 3. Relative risk (RR) and 95% confidence interval (CI) for mortality from diseases of the nervous system and sense organs in the reviewed studies among IR exposed populations compared with unexposed controls.
A pooled ERR at 100 mGy from four studies assessing the dose-response relationship between IR exposure and Parkinson's disease mortality [32,33,58] and incidence [12] was calculated, showing a significant positive ERR (ERRpooled at 100 mGy = 0.11; 95% CI: 0.06, 0.16) with no heterogeneity (Q = 1.37, p = 0.71), I 2 = 0.00%, and no publication bias (p = 0.13) ( Figure 5). A sensitivity analysis was conducted by performing a meta-analysis containing only mortality data [32,33,58], and the result was consistent with the one presented above (ERRpooled at 100 mGy = 0.19; 95% CI: 0.04, 0.36). It is noted that the ERR in each of the four studies were individually adjusted on sex and age, with a 10-year lag ( Table 2).

Figure 5.
Excess relative risk (ERR) at 100 mGy and 95% confidence interval (CI) for morbidity and mortality from Parkinson's disease in relation to IR exposure (° incidence; * mortality).

Cerebrovascular Diseases (ICD-10: I60-I69)
Key findings of the 39 studies focusing on cerebrovascular diseases can be found in (Table 3). . Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from Parkinson's disease in IR-exposed populations compared with general populations as reference.
A pooled ERR at 100 mGy from four studies assessing the dose-response relationship between IR exposure and Parkinson's disease mortality [32,33,58] and incidence [12] was calculated, showing a significant positive ERR (ERR pooled at 100 mGy = 0.11; 95% CI: 0.06, 0.16) with no heterogeneity (Q = 1.37, p = 0.71), I 2 = 0.00%, and no publication bias (p = 0.13) ( Figure 5). A sensitivity analysis was conducted by performing a meta-analysis containing only mortality data [32,33,58], and the result was consistent with the one presented above (ERR pooled at 100 mGy = 0.19; 95% CI: 0.04, 0.36). It is noted that the ERR in each of the four studies were individually adjusted on sex and age, with a 10-year lag ( Table 2). pooled was computed, showing no significant overall difference in mortality from Parkinson's disease between the IR-exposed populations presented above and the general populations (SMRpooled = 0.96; 95% CI: 0.78, 1.18), with moderate heterogeneity (Q = 8.64, p = 0.013), I 2 = 78.86% and no publication bias (p = 0.34) (Figure 4). . Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from Parkinson's disease in IR-exposed populations compared with general populations as reference.
A pooled ERR at 100 mGy from four studies assessing the dose-response relationship between IR exposure and Parkinson's disease mortality [32,33,58] and incidence [12] was calculated, showing a significant positive ERR (ERRpooled at 100 mGy = 0.11; 95% CI: 0.06, 0.16) with no heterogeneity (Q = 1.37, p = 0.71), I 2 = 0.00%, and no publication bias (p = 0.13) ( Figure 5). A sensitivity analysis was conducted by performing a meta-analysis containing only mortality data [32,33,58], and the result was consistent with the one presented above (ERRpooled at 100 mGy = 0.19; 95% CI: 0.04, 0.36). It is noted that the ERR in each of the four studies were individually adjusted on sex and age, with a 10-year lag ( Table 2).

Figure 5.
Excess relative risk (ERR) at 100 mGy and 95% confidence interval (CI) for morbidity and mortality from Parkinson's disease in relation to IR exposure (° incidence; * mortality).

Cerebrovascular Diseases (ICD-10: I60-I69)
Key findings of the 39 studies focusing on cerebrovascular diseases can be found in (Table 3).

Figure 5.
Excess relative risk (ERR) at 100 mGy and 95% confidence interval (CI) for morbidity and mortality from Parkinson's disease in relation to IR exposure ( • incidence; * mortality).

Cerebrovascular Diseases (ICD-10: I60-I69)
Key findings of the 39 studies focusing on cerebrovascular diseases can be found in (Table 3).

Nuclear Weapons Test Participants
Regarding nuclear weapons test series, a higher risk of mortality for cerebrovascular diseases was found in 21,357 servicemen and male civilians who participated in the U.K.'s atmospheric nuclear weapon tests and experimental programs compared with 22,312 controls (RR = 1.12; 95% CI: 1.03, 1.21), but mortality in this cohort was not significantly different than in the general population (SMR = 0.91; 95% CI: 0.85, 0.97; n deaths = 816) [52]. Decreased mortality due to cerebrovascular diseases was reported among U.S. military participants compared to the general population (SMR = 0.86; 95% CI: 0.83, 0.89; n deaths = 3161) [53] and among Australian participants to the British nuclear test in Australia compared with the general population (SMR = 0.86; 95% CI: 0.76, 0.98; n deaths = 243) [54].

Chernobyl Cleanup Workers
An increased risk of acute (RR = 1.40; 95% CI: 1.30, 1.50) and chronic (RR = 1.23; 95% CI: 1.00, 1.50) cerebrovascular diseases was shown in 198 Ukrainian Chernobyl liquidators (mean external dose exposure: 456 mSv) compared to 42 controls exposed to <50 mSv [65]. Among 957 evacuees from the 30 km zone of Chernobyl aged 18-60 years at the time of the accident, a significantly increased risk of cerebrovascular diseases was reported in those with a thyroid 131 I dose of 0.31-0.75 Gy compared to those with a dose below 0.30 Gy (RR = 2.16; 95% CI: 1.45, 3.22) [66]. A statistically significant dose-response relationship was reported between external gamma doses and the incidence of cerebrovascular diseases among 53,772 male Russian Chernobyl emergency workers who arrived in the zone of the Chernobyl accident within the first year after it (26 April 1986-26 April 1987) and who were followed from 1986 to 2012 (mean external whole body dose: 0.161 Gy; max: 1.42 Gy) (ERR per Gy = 0.45; 95% CI: 0.28, 0.62; n diseases = 23,264) [67], whereas the risk of cerebrovascular diseases in an Estonian cohort of Chernobyl cleanup workers was not statistically different from an unexposed comparison cohort of 7631 men (RR = 1.05; 95% CI: 0.91, 1.20) [70].

Medical Patients
In the only available study on patients, there was no significantly increased risk of cerebrovascular diseases mortality among a cohort of 77,275 tuberculosis patients in Canada and Massachusetts according to X-ray fluoroscopic diagnostic repeated exposures (ERR per Gy = 0.441; 95% CI: −0.119, 1.090; n deaths = 1561 for cumulative doses restricted to 0-0.5 Gy) [73].

Environmental Radiation
Only one study among those included in this review focused on indoor radon exposure, and it found a slight but significant association with indoor radon level towards an increased incidence of stroke among a South Korean cohort of 28,557 selected inhabitants based on demographic criteria and aged over 40 years (OR = 1.004; 95% CI: 1.001, 1.007) [57].

Overall Meta-Analyses
From the 23 studies reporting SMRs, an SMR pooled was computed showing a statistically significant lower mortality from cerebrovascular diseases in IR-exposed populations compared with general populations (SMR pooled = 0.70; 95% CI: 0.62, 0.80), with high heterogeneity (Q = 672.95, p < 0.0001), I 2 = 96.73%, and a publication bias (p = 0.03), suggesting that small studies with negative results were published less often ( Figure 6). Sensitivity analyses where studies or different worker groups were excluded one by one resulted in no change in heterogeneity, as well as sensitivity analyses in which studies that did not report ICD codes were removed. (ERR per Gy = 0.441; 95% CI: −0.119, 1.090; ndeaths = 1561 for cumulative doses restricted to 0-0.5 Gy) [73].

Environmental Radiation
Only one study among those included in this review focused on indoor radon exposure, and it found a slight but significant association with indoor radon level towards an increased incidence of stroke among a South Korean cohort of 28,557 selected inhabitants based on demographic criteria and aged over 40 years (OR = 1.004; 95% CI: 1.001, 1.007) [57].

Overall Meta-Analyses
From the 23 studies reporting SMRs, an SMRpooled was computed showing a statistically significant lower mortality from cerebrovascular diseases in IR-exposed populations compared with general populations (SMRpooled = 0.70; 95% CI: 0.62, 0.80), with high heterogeneity (Q = 672,95, p < 0.0001), I 2 = 96.73%, and a publication bias (p = 0.03), suggesting that small studies with negative results were published less often ( Figure 6). Sensitivity analyses where studies or different worker groups were excluded one by one resulted in no change in heterogeneity, as well as sensitivity analyses in which studies that did not report ICD codes were removed. Figure 6. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from cerebrovascular diseases in IR exposed populations compared with general populations as reference. Figure 6. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from cerebrovascular diseases in IR exposed populations compared with general populations as reference.

Figure 7.
Relative risk (RR) and 95% confidence interval (CI) for mortality from cerebrovascular diseases in the reviewed studies among IR exposed populations compared with unexposed controls. * 90% CI.

Figure 7.
Relative risk (RR) and 95% confidence interval (CI) for mortality from cerebrovascular diseases in the reviewed studies among IR exposed populations compared with unexposed controls. * 90% CI.

Mental and Behavioral Disorders (ICD-10: F00-F99)
Key findings of the 22 studies focusing on mental and behavioral disorders can be found in Table 4. diseases became significant in sensitivity analyses where studies that did not report ICD codes were removed (i.e., the Eldorado uranium workers cohort study [45]) (ERRpooled at 100 mGy = 0.13; 95% CI: 0.03, 0.22).

Mental and Behavioral Disorders (ICD-10: F00-F99)
Key findings of the 22 studies focusing on mental and behavioral disorders can be found in Table 4.  Figure 10. Excess relative risk (ERR) per 100 mGy and 95% confidence interval (CI) for morbidity from cerebrovascular diseases in relation to IR exposure.

Mental and Behavioral Disorders (ICD-10: F00-F99)
Key findings of the 22 studies focusing on mental and behavioral disorders can be found in Table 4. Table 4. Key findings of the included studies on mental and behavioral disorders.

Nuclear Weapons Test Participants
Among U.S. military participants in nuclear weapons test series, a significant decrease in mortality due to mental and behavioral disorders was reported as compared with the general population (SMR = 0.92; 95% CI: 0.87, 0.98; n deaths = 1021) [53].

Chernobyl Cleanup Workers
A statistically significant dose-dependent increase in the level of mental disorders (as assessed by the Brief Psychiatric Rating Scale (BPRS) [74]) was found among 326 Ukrainian cleanup workers exposed to dose under 500 mSv [69]. Higher risks of organic psychoses and non-psychotic organic brain damages were found among 198 Ukrainian Chernobyl liquidators who intervened in 1986-1987, relative to 42 internal controls exposed to doses < 50 mSv (RR = 3.15; 95% CI: 2.60, 3.70 and RR = 1.99; 95% CI: 1.60, 2.50, respectively) [65]. Regarding schizophrenia spectrum disorders, the incidence increased dramatically among 100 Chernobyl exclusion zone personnel with acute radiation sickness as compared to the general Ukrainian population in 1990, just after the disaster (5.4/10,000 vs. 1.1/10,000, respectively) [72].
Furthermore, a statistically significant increased frequency of mild cognitive disorders was observed among 196 men workers involved in the Chernobyl "Shelter Object" (total irradiation mean: 19.9 mSv; max: 56.7 mSv) between baseline (T0) and after completion of their period of work on-site (T1) (3.6% vs. 11.2% (p < 0.01), respectively). Nevertheless, this increase was not found in workers who had already been exposed to IR before this task [68]. Loganovsky et al. (2013) found a higher level of depression assessed by the self-rating depression scale [75] in 219 people with post-traumatic stress disorder (PTSD) affected by the Chernobyl disaster, whether they were diagnosed with acute radiation sickness (mean and standard deviations: 52.3 ± 12.9) or not (58.6 ± 12.6), compared with 28 war veterans (47.8 ± 12.6) and a group of 22 healthy unexposed people (39.6 ± 7.3) [71]. On the other hand, no increase in the incidence of mental and behavioral disorders was found in Estonian Chernobyl clean-up workers compared to a cohort of unexposed men (RR = 1.00; 95% CI: 0.94, 1.07) but was found for mental disorders due to alcohol (RR = 1.21; 95% CI: 1.06, 1.39) [70].

Overall Meta-Analysis
Out of the 22 studies included in this section, 13 had available SMRs to be integrated in a meta-analysis, showing a statistically significant lower mortality from mental and behavioral disorders between the IR-exposed populations presented above and the general populations (SMR pooled = 0.86; 95% CI: 0.74, 0.99, Q = 103.72 (p < 0.0001), I 2 = 88.43%, no publication bias (p = 0.97) (Figure 11). Sensitivity analyses where studies or different workers groups were excluded one by one resulted in no change in heterogeneity, as well as sensitivity analyses in which studies that did not report ICD codes were removed.

Overall Meta-Analysis
Out of the 22 studies included in this section, 13 had available SMRs to be integrated in a meta-analysis, showing a statistically significant lower mortality from mental and behavioral disorders between the IR-exposed populations presented above and the general populations (SMRpooled = 0.86; 95% CI: 0.74, 0.99, Q = 103.72 (p < 0.0001), I 2 = 88.43%, no publication bias (p = 0.97) (Figure 11). Sensitivity analyses where studies or different workers groups were excluded one by one resulted in no change in heterogeneity, as well as sensitivity analyses in which studies that did not report ICD codes were removed. Figure 11. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from mental and behavioral disorders in IR exposed populations compared with general populations as reference.

Discussion
The risk of non-cancerous CNS diseases after exposure to low-to-moderate doses of IR in adulthood was analyzed based on 45 studies. Meta-analyses show reduced mortality due to nervous system diseases, cerebrovascular diseases, and mental and behavioral disorders in radiation workers compared to general populations and suggest a higher risk of cerebrovascular diseases and Parkinson's disease that may be dose-dependent in people exposed during adulthood (radiation workers, A-bomb survivors). For cerebrovascular diseases, a significant dose-risk relationship is reported for incidence, but it was non-significant for mortality. These findings are consistent with the previous meta-analysis by Little et al. (2012), who found a significant positive relationship between cumulative IR dose and cerebrovascular diseases, studying mortality and morbidity outcomes together [13]. We conducted separate analyses for mortality and incidence outcomes because mortality analyses are often based on the underlying cause-of-death, and it has been shown Figure 11. Standardized mortality ratio (SMR) and 95% confidence interval (CI) for mortality from mental and behavioral disorders in IR exposed populations compared with general populations as reference.

Discussion
The risk of non-cancerous CNS diseases after exposure to low-to-moderate doses of IR in adulthood was analyzed based on 45 studies. Meta-analyses show reduced mortality due to nervous system diseases, cerebrovascular diseases, and mental and behavioral disorders in radiation workers compared to general populations and suggest a higher risk of cerebrovascular diseases and Parkinson's disease that may be dose-dependent in people exposed during adulthood (radiation workers, A-bomb survivors). For cerebrovascular diseases, a significant dose-risk relationship is reported for incidence, but it was nonsignificant for mortality. These findings are consistent with the previous meta-analysis by Little et al. (2012), who found a significant positive relationship between cumulative IR dose and cerebrovascular diseases, studying mortality and morbidity outcomes together [13]. We conducted separate analyses for mortality and incidence outcomes because mortality analyses are often based on the underlying cause-of-death, and it has been shown that cerebrovascular diseases could be poorly captured when using only underlying causes of death [76].
In the present study, the decreases in mortality found in radiation workers when compared to general population rates for nervous system diseases, cerebrovascular diseases, and mental and behavioral diseases can be explained by the healthy-worker effect [77], meaning that worker populations usually present a better health condition than the general population. The calculation of relative risks using a control occupational group that is supposed to be more comparable to the exposed group (except for exposure) allows the healthy-worker effect to be avoided. This can be observed with the pooled analysis of the results of one study on nuclear weapons test participants [52] and three studies of medical radiation workers compared to groups of unexposed populations [61,62,64] that yields nonsignificantly increased relative risks of mortality from nervous or cerebrovascular diseases. Internal dose-risk analyses are even more informative for investigating the impact of IR exposure on the risk of non-cancerous CNS diseases.
Most of the studies included in this review dealt with occupational exposure, mainly among nuclear workers and uranium miners, who are subject to different types of exposure depending on their activities. For example, uranium miners are repeatedly exposed to a mix of radon gas and its progenies, external gamma rays, and uranium dusts [78], and cycle nuclear workers are exposed to external gamma rays, possibly combined with tritium, uranium, or plutonium, depending on their activity [79,80]. However, external exposures are more commonly reported in studies, although internal contamination was often mentioned among nuclear workers. Due to the low rate of workers monitored for internal exposure, or the fact that only the status "exposed to internal contamination" was known [39], few studies performed separate exposure-based, adjusted, or sensitivity analyses to disentangle the share of risk attributable to each type of exposure [30,36,41,42,44], which did not allow specific meta-analyses for internal exposures in this work. Nevertheless, some studies have addressed the issue of co-exposure by treating it as a confounding factor [41], and it did not significantly change the result (ERR per Gy = 0.05; 95% CI: −0.03, 0.16 vs. ERR per Gy = > 0; 95% CI: −0.10, 0.16). In addition, differences in results between studies may be related to the characteristics of the exposure such as brief or prolonged; the type of radiation field (e.g., external low-LET photons vs. high-LET alpha particles); or possible biases in dosimetry [32].
Furthermore, exposure to IR in the occupational setting is often accompanied by coexposure to other health risk factors (e.g., chemical substances, pesticides, heavy metals, nitro compounds, non-ionizing radiations, air pollution, tobacco use, etc.) that may confound and/or modify the relationship between IR exposure and a health outcome. For example, medical radiation workers are predominantly exposed to X-rays, but can possibly be exposed to chemicals or drugs (such as hydroquinone, aldehydes, acetic acid, ammonia, etc.) [61,81]. Nevertheless, research on the health effects of co-exposures to two or more risk factors (exposome) is a very dynamic area of research, and synergies or antagonisms following co-exposure to different environmental agents have been shown [82,83]. However, the interaction of various factors and associated health outcomes are poorly characterized to date. In the present work, few studies have considered co-exposures, with limited evidence on their impact on the dose-response relationship.
A high number of risk factors for non-cancerous CNS diseases have been identified in the scientific literature, with varying degrees of evidence depending on the outcome [84,85]. Several risk factors have been considered in the dose-response analyses by each study separately. However, socio-economic factors (often used in occupational studies to control for possible confounding factors that are not available on an individual basis and may influence mortality and disease occurrence) were included in only half of the studies that report dose-response analyses in the present work (Tables 2-4). In order to take risk factors into account as much as possible, we performed our meta-analysis on adjusted estimates even if the estimates were not systematically adjusted on the same risk factors, as is usually recommended in meta-analysis methodology.
Finally, mental and behavioral disorders are known to be influenced by individual characteristics, but also by the socio-economic and environmental circumstances in which people live [86]. Indeed, specific disorders such as Alzheimer's disease and dementia spectrum disorders are known to be influenced by environmental and/or genetic factors [87,88].
It is therefore important to highlight that disasters (natural and human-made) can inflict psychological damage on the affected populations. It has been reported that a major health impact of the Chernobyl nuclear power plant accident was the fear about potential upcoming health problems [89]. Furthermore, the Hiroshima and Nagasaki bombings have had long-lasting effects on mental health, such as post-traumatic stress, depression, anxiety, and somatization. However, there were few well-designed studies (i.e., evaluation of exposure, confounding factors) on mental health following the Chernobyl catastrophe. It appears very difficult to determine which part of mental health disorders is due to radiation directly, and which is due to the psychological consequence of having experienced such a disaster [90]. In a broader sense, when studying non-cancerous CNS diseases, such as cognitive disorders, many factors such as exposure time vulnerability, mechanism, and susceptibility factors are important to consider [11].
One major limitation of this systematic review was related to the definition of outcomes. Although most of the studies are based on the ICD coding, whether for causes of death or diagnostics (which still ensures a certain homogeneity and reliability in our analyses), the fact remains that classification errors may have been made during coding. For example, the rules of the ICD lead to the selection of suicide as the initial cause of death, even if the physician has indicated another sequence (e.g., depression leading to suicide) [91]. Thus, the number of deaths due to depression may have been underestimated in these studies. In Russia, a high degree of inconsistency across the region was found for mental and behavioral disorders, diseases of the nervous system, and certain cardiovascular diseases, suggesting differences in coding practices [92]. Then, analyses could be performed on broad categories, as the level of consistency improves when causes of death are grouped into broader diagnostic categories, but could not be performed for subcategories, when the classification bias might be higher. Finally, 10 studies did not even mention ICD coding, which has occasionally made it difficult to classify the outcomes of these studies within the causes of disease/death used in this work, possibly leading to a classification bias. Sensitivity analyses performed by removing the studies without ICD codes showed no change in the results, except the positive dose-response relationship for cerebrovascular diseases, which became significant.
Heterogeneity seems unavoidable because of different populations, various types of radiation exposure, and chronical or acute exposure. Sensitivity analyses in which the pooled SMR was calculated excluding each study one at a time and each group (e.g., aircrews, nuclear workers, medical workers, etc.) revealed no substantial alteration of the aggregate SMR for the three studied outcomes. Nevertheless, we used random effects models to calculate our estimates (SMR, RR, and ERR), which account for potential heterogeneity between and within studies.
The consideration of bioindicators and biomarkers in epidemiological studies could be very informative in improving the accuracy of the outcomes and the reconstruction of actual IR exposure of participants. In the long run, this could also help to better understand the mechanisms of these neurodegenerative disorders. For example, Borghini et al. (2017) have shown that the expression profiles of circulating brain miR-134 (a brain-specific miRNA that has been shown to be dysregulated in pathologies such as Alzheimer disease, bipolar disorder, and glioblastomas [93]) and miR-2392 were significantly downregulated in interventional cardiologists compared with controls [94]. Complementary studies are needed to confirm these findings and to further explore the potential of circulating miRNAs to be used clinically as novel biomarkers to identify early, disease-related perturbations caused by long-term radiation exposure.
According to this work, the effects of low doses of IR on non-cancerous CNS diseases cannot be excluded. Compared to the 2.7 billion people who had neurological disorders in 2019 [9], the estimated increased risk of 17% would result in a significant public health impact. In addition, due to the fact that human populations are increasingly exposed to IR from various sources (e.g., cosmic rays, environmental radionuclides), along with the continued growth and evolution of IR imaging technologies, the resulting dose in the general population is increasing. Moreover, in a context where exposure to IR is steadily increasing in some groups of workers (e.g., medical radiation workers), new studies avoiding the biases mentioned in this work are justified: the use of precise dosimetry, an indisputable definition of the outcomes, and adjusted dose-response calculations are encouraged. To our knowledge, this work is the first systematic review and meta-analysis of the literature assessing the risk of non-cancerous CNS diseases and mortality in populations exposed to IR during adulthood only. We included a broad range of endpoints, resulting in a large number of studies covered. All included studies met the previously defined criteria according to the PRISMA recommendations, allowing robust and exhaustive analysis while maintaining a focus on the main research question. The quality score between 4 and 9 on the Newcastle Ottawa scale for all studies included in this review provides a good quality rating for this work.

Conclusions
The present review and meta-analyses did not suggest higher risk of mortality due to non-cancerous CNS diseases after adult IR-exposure compared with unexposed controls. However, some of the studies reviewed had low statistical power to detect an effect and inadequate dosimetry, if any. Furthermore, a significant positive excess relative risk was found for cerebrovascular disease morbidity and for Parkinson's disease. Nevertheless, we caution against drawing firm conclusions due to methodological issues, including uncertainties related to the classification of the diseases, dosimetry assessment, and potential confusion bias. Further studies, ideally large-scale studies with individual dose reconstruction and collection of information on potential confounding factors, will be essential to expand our knowledge of the risk of non-cancerous CNS diseases following exposure to low-dose IR.
Supplementary Materials: The following supporting information can be downloaded at https: //www.mdpi.com/article/10.3390/brainsci12080984/s1, Table S1: Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2020 checklist; Table S2: List of causes of deaths studied and corresponding codes according to the International Classification of Diseases (ICD). Ref. [95] is cited in Supplementary Materials.