Total Body Irradiation and Risk of Diabetes Mellitus; A Meta-Analysis

Objective: Hematopoietic stem cell transplant (HSCT) has recently emerged as a cure for previously “incurable” diseases and is being explored and attempted in many other fields including congenital and acquired non-malignant diseases. However, the long-term side effect associated with HSCT especially Total Body Irradiation (TBI) is still understudied. Therefore, we attempted to establish association between TBI and risk of developing Diabetes Mellitus (DM) or impaired glucose metabolism (IGM). Methods: We searched for titles of articles in MEDLINE (PubMed), EMBASE, and Cochrane library in August 2018 that evaluated the association between TBI in the setting of HSCT and DM or IGM. We conducted a random effect meta-analysis of 11 studies involving a total of 13,191 participants and reported the pooled MD (mean difference) for the development of DM/IGM after TBI as part of the conditioning regimen for HSCT. Results: We found a significant increase in the risk of developing DM/IGM after TBI is used as part of the conditioning regimen compared to other types of conditioning regimen with the pooled MD being 5.42, 95% Confidence Interval (CI) 2.51-11.71, I2=92.4%. Conclusion: TBI as a conditioning regimen in the setting of HSCT significantly increases the risk of developing DM/IGM. Therefore, we recommend close monitoring and screening for diabetes mellitus in patients who underwent TBI before HSCT.


Introduction
Hematopoietic stem cell transplant (HSCT) has recently emerged as a cure for previously "incurable" diseases. Its first application was in 1957 for a leukemia patient who received a transfer of stem cell from the healthy twin conferring a "cure" state which was previously not achieved with traditional chemotherapy. (Pidala et al., 2011;Henig and Zuckerman, 2014) Since the breakthrough with the experiment, the horizon of hematopoietic stem cell transplant has expanded to involve other sources of stem cells including cells from unrelated donors or cord blood stem cell stored in the international banks (Kondo et al., 2003;Stacey and Hunt, 2006;Lin et al., 2013) Even now, newer sources such as mesenchymal tissue or other differentiated cellular lineages (Sordi and Piemonti, 2011;Keating, 2012) are also being investigated for cure of various congenital, autoimmune and malignant diseases. It is proper to say that, the potential for application is limitless (Kondo et al., 2003;Stacey and Hunt, 2006;Girlovanu et al., 2015).
However, on the same expanding frontier, the longterm side effects of transplant and accommodating treatment modalities are still understudied. It is estimated that over 8,000 HSCT was performed in the US in 2017 and the number continues to grow for both autologous as RESEARCH ARTICLE

Total Body Irradiation and Risk of Diabetes Mellitus; A Meta-Analysis
Kittika Poonsombudlert*, Nath Limpruttidham well as allogeneic HSCT (D'Souza et al., 2017). From such an astounding number; at the aftermath of the battle, what consequences are the survivors encountering? (Hilgendorf et al., 2015) Previous studies had mentioned the possibilities of developing hypertension, dyslipidemia as well as endocrinologic abnormalities with various risk factors (Baker et al., 2007;Hoffmeister et al., 2010;Wei et al., 2016;Bielorai et al., 2017). Not to mention known co-morbidities from the treatment itself such as graft versus host disease (GVHD) (Socie and Ritz, 2014;Arora et al., 2016) or infectious complications (Nakano et al., 2014;Kumar et al., 2015) which are well known to associate with the immunocompromised state rendered by the treatments (Michael et al., 2013).
Part of the treatment with HSCT is the conditioning regimen which includes a combination of chemotherapy and radiation therapy. Most adverse effects of various chemotherapy regimens including short and long-term complications have been well studied. However, for application of total body irradiation (TBI), the long-term side effect is still to be discovered. The fact that TBI can render the hematopoietic stem cell powerless signifies that other mechanisms are playing a role in other organs and metabolic pathways as well.
We aimed to explore the potential long-term outcome after HSCT and TBI on the development of diabetes

Patients and method
We searched for titles of articles in the MEDLINE (PubMed), EMBASE, and Cochrane library. We performed a search in August 2018 and did not restrict publication dates. The following main search terms were used: hematopoietic stem cell transplant, total/whole body irradiation and diabetes mellitus/impaired glucose metabolism. The full search strategy is detailed in Item S1 in the Supplementary Material.
All published nonrandomized trials that evaluated the association between TBI in the setting of HSCT and DM or IGM were included. Observational studies−prospective cohort, retrospective cohort and cross-sectional studies−were included. Review articles, case reports, letters, commentaries, abstracts, unpublished studies and studies in languages other than English were not included.
The primary outcome was the diagnosis of DM/ IGM, which was measured by standardized blood tests recommended by American Diabetes Association (ADA) (American Diabetes, 2018) or self-report questionnaire given to patients or their primary care physicians. We compared the risks of DM/IGM diagnosis between patients who did and did not receive TBI as part of the conditioning regimen for HSCT.

Data extraction and quality assessment
Two investigators independently extracted the following data: Authors, publication year, country of origin, study design, baseline patient's characteristics, interventions, and outcomes. Any conflicting opinions on data extraction were resolved by consensus of the investigators.
The Newcastle − Ottawa Quality Assessment Scale (NOS) was used to assess the quality of the nonrandomized studies based on the selection of the study groups, comparability of study groups, and ascertainment of exposure/outcome. Studies with total scores of >6 and <4 were considered to be of high and low quality, respectively. We excluded any studies that the meta-analysis indicated poor quality. There were no randomized or blinded control trials due to the obvious nature of the intervention.

Statistical methods
The primary outcome of the study is the pooled mean difference (MD) of the risk of development of DM/ IGM between patients who received TBI compared with patients who did not receive TBI. We used a randomeffects model as included studies are observational study.
We conducted sensitivity analysis and subgroup analysis to explore heterogeneity of the included studies (Higgins et al., 2003). We used Funnel plot (as detailed in Item S2 in the Supplementary Material) and Egger's test to assess for publication bias. All analyses were performed using Stata 13 software.

Description of included studies
The initial search yielded 1,388 articles; 1,365 were excluded from title and abstract review as they were obviously not pertaining to our study, did not involve TBI as part of the conditioning regimen, did not evaluate association between TBI and DM/IGM, were not conducted in patients receiving HSCT, conducted in animal models, or were published in languages other than English. A total of 23 articles underwent full-length review. 12 of them were excluded because they did not have a TBI control group or did not have an eligible study population. Finally, we included 11 studies: Five retrospective cohort studies, three prospective cohort studies and two cross-sectional studies. They included 13,191 patients with indications for HSCT, 2,695 who had undergone TBI and 10,266 who had not (controls). Figure  S3 In the supplementary material outlined our search methodology and selection process. Table 1 described the characteristics of the extracted studies. The total sample size range was 22-8,599 patients. The mean age range was 7-44 years. The follow-up duration ranged from 2-30 years. DM/IGM was diagnosed by blood sample per ADA (American Diabetes, 2018) guideline, from self-report or contact with patients' primary care physicians.

Meta-analysis results
There were 11 studies included in this meta-analysis. Using a random-effects model, the pooled MD of diagnosis of DM/IGM was found to be significantly higher in the TBI group than in the control group (MD = 5.42, 95% Confidence Interval (CI) 2.51-11.71). The heterogeneity among studies (I 2 ) was 92.4%. We found publication bias from Funnel plot which was not statistically significant from Egger's test (P = 0.466). This is likely due to less likelihood of negative studies being selected for publication.

See details for meta-analysis in S4 in Supplementary Material Subgroup analysis
We conducted a subgroup analysis to assess the effect of age, study design and indication for HSCT on the primary outcome.
First, for age group, we categorized the studies to include only pediatric population and the other for adult patients. Pooled MD among adult group (Baker et al., 2007;Armenian et al., 2012) was 2.09, 95% CI 0.95-4.61, I 2 =71.7% while for the pediatric group (Taskinen et al., 2000;Hoffmeister et al., 2004;Neville et al., 2006;Oudin et al., 2011;Michael et al., 2013;Hirabayashi et al., 2014;Oudin et al., 2015;Nakagawa et al., 2018;Shalitin et al., 2018), the pooled MD was 8.43, 95% CI 5.51-12.91, cross-sectional study design has limitations to accurately diagnose DM/IGM after patients received TBI as the diagnosis is based on a single timeframe rather than a prolonged period of follow-up time in cohort studies. The association remained statistically significant after exclusion of two cross-sectional studies. Therefore, we concluded that a diagnosis of DM/IGM is markedly increased in patients who received TBI as part of the conditioning regimen compared to other conditioning regimens in the setting of HSCT. This observed association was stronger among prospective studies, pediatric population, non-US based studies and studies done exclusively for a malignant hematologic condition.

Discussion
This meta-analysis investigated the effect of TBI on the development of DM/IGM in patients indicated for HSCT. Our main findings suggested that TBI was significantly associated with an increased risk of the development of DM/IGM compared to those who did not undergo TBI as part of the conditioning regimen for HSCT.
There have been many theories as to why TBI can cause DM/IGM (Chemaitilly et al., 2009). Starting from the fact that TBI; as other types of radiation do is causing inflammation. Most radiation is site-specific, thus limiting the extent of the inflammation to a targeted I 2 =44.4%.
According to this regression meta-analysis, differences in age group and indications for HSCT potentially had effect modification to the association between TBI and the diagnosis of DM/IGM, which could be used to identify high risk population. In addition, we recognized that   Table 1. Study Characteristics region, however, as the name suggested, TBI; total body irradiation can cause systemic inflammation and interfere with molecular signaling, cascade pathways and more significantly can induce hormonal disturbances. The central result of all the inflammation caused by the TBI is insulin resistance. This is theorized to be from growth hormone deficiency (Martin and Jeanrenaud, 1985;Bakker et al., 2007;Nakagawa et al., 2018) or abnormal adipose tissue redistribution. (Rajendran et al., 2013;Bizzarri et al., 2015) Other more, site-specific issues are from reduced pancreatic beta-cell reserve and an overall decrease in pancreatic volume directly induced by the TBI (Wei et al., 2015;Arora et al., 2016). All of these theories could well explain how TBI can be associated with the development of DM/IGM.
On the contrary, there has long been speculation that various treatments modalities associated with HSCT itself can increase the risk of developing DM/ IGM. (Griffith et al., 2010) including, but not limited to chemotherapy, prolonged use of corticosteroid and various immunosuppressants (Vantyghem et al., 2014;Abou-Mourad et al., 2010;Mostoufi-Moab et al., 2016), which are well known to be associated with truncal obesity and potentially DM/IGM when used for prolonged period of time as well as physical inactivity resulting from prolonging hospitalization (Inoue et al., 2010;Al Tunaiji et al., 2014).
In addition, we found potential publication bias visualized from the Funnel plot, although it did not meet statistical significance in Egger's test. We believed that they were fewer published negative studies, which contributed to publication bias. Although it is generally recognized that small negative studies were difficult to get published in high quality scientific journals, we used general broad search terms as described in supplemental material S1 to include small negative studies as much as we could to minimize publication bias.
There are some limitations to this study. First, only 11 studies met our inclusion criteria, most of which were conducted in the USA, possibly limiting the external validity of this meta-analysis. Secondly, patients of varying age and pre-transplant co-morbidity were included in the study which could be confounding the test result. Thirdly, some studies included patients with known endocrinologic disorders such as hypothyroidism, hypogonadism and growth hormone deficiency which could influence the development of DM/IGM. And lastly, we recognized that there is heterogeneity in the study which is likely due to differences in study designs and background characteristics of the subjects, but according to the regression meta-analysis, we found that change in age, study type, a location of the study and indication for HSCT could potentially explain the heterogeneity.
We believed that there is still more to study regarding an association between TBI, HSCT and development of DM/IGM as DM is one of the most debilitating diseases with severe consequences if left untreated (White, 2015). Survivors might not realize that while they are being "cured" of one life-threatening condition; after the treatment, their risk of developing other chronic co-morbidity is not infinitesimal, they should be aware that they are at higher risk of developing DM/IGM and frequent monitoring is warranted (Wei et al., 2016). Our meta-analysis of clinical trials demonstrated that TBI significantly increased the risk of development of DM/ IGM, but this is still an undiscovered field and further research is needed to clarify if the association truly stems directly from TBI or treatment with HSCT as a whole. We suggested that further large, prospective, controlled trials to investigate the long-term outcomes and treatment option for DM/IGM in the future as this is a rapidly expanding field of treatment potentials.