Association Between Body Composition and Development of Glucose Intolerance after Allogeneic Hematopoietic Cell Transplantation

This article is featured in Highlights of This Issue, p. 1993

Allogeneic hematopoietic cell transplantation (HCT) is an established curative treatment for malignant and nonmalignant hematologic diseases (1). However, HCT is associated with short- and long-term toxicities, including the development of diabetes mellitus (DM) with its attendant impact on the quality and quantity of survival after HCT (2, 3). Development of glucose intolerance and subsequent DM in the early post-HCT period has been associated with additional cardiometabolic comorbidities (e.g., hypertension, cardiovascular disease), lower disease-free survival, and increased risk of death or relapse (3–7). Recipients of allogeneic HCT have a higher risk of developing glucose intolerance and/or DM than recipients of autologous HCT (2). Potential allogeneic transplant-related risk factors include conditioning with total body irradiation (TBI), use of unrelated donor stem cell source, development of acute graft-versus-host disease (aGVHD), and associated high cumulative corticosteroid exposure (2, 3, 6, 8). In these patients, glucose intolerance generally develops early (<30 days) in the post-HCT course, and studies have reported that up to 60% may develop glucose intolerance and/or DM by day 100 (3–5). It is therefore necessary to identify which patients are at risk for developing glucose intolerance prior to HCT to inform targeted screening, prevention, and management during and immediately after HCT.

In the general population, overweight and obesity are the strongest risk factors for developing glucose intolerance and/or DM (9). Historically, body mass index (BMI), defined as the ratio of weight to square of height (kg/m²), has been utilized to classify individuals as overweight or obese. However, it is well-established that BMI is an indirect measure of obesity that does not fully account for sex- or age-specific differences in body composition (10). With age, body fat increases while muscle mass decreases, but these changes are not completely reflected by BMI (10). Aging-related loss of lean muscle mass is associated with DM, as muscle is a key endocrine organ responsible for a majority of the body's glucose disposal (11–13). Individuals with loss of skeletal muscle mass may be considered normal or even underweight by BMI, but still carry an increased risk of developing DM, potentially suggesting a different underlying pathophysiology in these individuals compared with those who are considered overweight or obese by BMI. This is especially relevant in the oncology population, in whom the prevalence of skeletal muscle loss is increased due to the effects of cancer, its treatment, and the associated side effects of these treatments (14–16). There is a paucity of information regarding the association between pre-HCT body composition and the development of DM in patients undergoing HCT, in part due to limitations in methods to accurately measure body composition. Recent advances in imaging technology have enabled distinction of body tissues (e.g., muscle, fat) based on signal attenuation from frequently utilized clinical studies such as CT images (17, 18). The overall aims of this study were to describe the association between body composition measures obtained from pre-HCT CT images and early-onset (≤30 days) de novo glucose intolerance after allogeneic HCT, and to examine its impact on survival after HCT.

This was a retrospective cohort study of patients who underwent allogeneic HCT for acute leukemia or myelodysplastic syndrome (MDS) as adults (≥18 years old) at City of Hope (COH; Duarte, CA) between January 1, 2007 and December 31, 2014. The institutional review board at COH approved the protocol and written informed consent was obtained from study participants in accordance with the Declaration of Helsinki. Medical records were abstracted for demographic data (age at HCT, sex, ethnicity), BMI [calculated as weight (kg)/height (m²)], diagnosis, HCT details (donor source, conditioning, GVHD prophylaxis, development of grade 2 to 4 aGVHD), and variables necessary to derive pre-HCT–comorbidity index (HCT-CI), including history of DM prior to HCT. Patients with acute leukemia in first complete remission, MDS with refractory anemia, or refractory anemia with ringed sideroblasts were considered at low risk of relapse at HCT; all others were considered high risk. Conditioning regimen was classified as myeloablative (MA) with TBI, MA without TBI, or non-MA. High HCT-CI was defined as having a pre-HCT comorbidity-age index of 3 or more (19, 20). BMI at HCT was used to categorize individuals as overweight/obese (BMI ≥ 25 kg/m²) or nonoverweight (BMI < 25 kg/m²) per Centers for Disease Control (CDC) guidelines (21). Information on vital status was obtained from the National Death Index and COH medical records.

Medical records were reviewed for development of de novo glucose intolerance within 30 days after HCT. The decision to limit our outcome of interest to the first 30 days was based on the latency of glucose intolerance in the peri-HCT period as well as the concern for differential clinical monitoring/screening that may occur once individuals are discharged from their inpatient admission. For this study, individuals were required to meet the Common Terminology Criteria for Adverse Events version 5.0 (22) criteria of grade 2 or higher [i.e., requiring dietary modification or pharmacologic (i.e., oral antiglycemic agent or insulin indicated)] to be considered as having glucose intolerance after HCT. Individuals with a history of DM prior to HCT were excluded from the current study.

Body composition was assessed from existent screening CT scans of the chest, abdomen, and pelvis that were performed as part of the pre-HCT evaluation. CT scans completed ≤90 days prior to stem cell infusion were deemed to accurately represent patients’ body composition at HCT and were therefore selected for analysis. Three adjacent cross-sectional images obtained from a spiral CT scanner were used to quantify muscle area at the level of the third lumbar vertebra (L3) as previously described (23). Muscle area was normalized for height (cm²/m²; ref. 24) and reported as lumbar skeletal muscle index (SMI). Skeletal muscle density (SMD), typically used to determine muscle quality, was measured using the muscle radiation attenuation rate (in Hounsfield Units). Skeletal muscle gauge (SMG), which combines SMI and SMD (SMI × SMD), with arbitrary units (AU; refs. 25–27), was also determined for all participants. Visceral adipose tissue (VAT) was quantified with Automatic Body composition Analyzer using CT image segmentation (ABACS; ref. 17), which is a commercially available software that automatically segments skeletal muscle and adipose tissue (Fig. 1). All ABACS VAT segmentation was manually reviewed by members of the research team who were blinded to patient characteristics and outcomes (R. Bhandari, A. Iukuridze, J. Berano Teh). Because of a lack of well-established definitions for adiposity and skeletal muscle measures, we used the lowest sex-specific tertile with the cohort stratified by BMI as the cut-off for abnormal SMI [≤38.20 cm²/m² (female, BMI < 25 kg/m²); ≤41.98 cm²/m² (female, BMI ≥ 25 kg/m²); ≤45.18 cm²/m² (male, BMI < 25 kg/m²); ≤51.84 cm²/m² (male, BMI ≥ 25 kg/m²)], SMD [≤41.86 HU (female, BMI < 25 kg/m²); ≤34.55 HU (female, BMI ≥ 25 kg/m²); ≤44.89 HU (male, BMI < 25 kg/m²); ≤40.11 HU (male, BMI ≥ 25 kg/m²)], and SMG [<1664.9 AU (female, BMI < 25 kg/m²); ≤1551.7 AU (female, BMI ≥ 25 kg/m²); ≤2046.0 AU (male, BMI < 25 kg/m²); ≤2127.4 AU (male, BMI ≥ 25 kg/m²)]. Individuals were considered to have abnormal adiposity according to sex-based VAT tertile cut-offs, again with the cohort stratified by BMI [≥48.78 cm² (female, BMI < 25 kg/m²); ≥122.71 cm² (female, BMI ≥ 25 kg/m²); ≥72.58 cm² (male, BMI < 25 kg/m²); ≥174.62 cm² (male, BMI ≥ 25 kg/m²)].

The primary study outcome was de novo glucose intolerance by day +30 after stem cell infusion. Descriptive statistics were generated for the entire cohort, and by outcome of interest (developed glucose intolerance, did not develop glucose intolerance). Patient, disease, and HCT characteristics were summarized with median and range for continuous variables, and frequencies and percentages for categorical variables.

Cumulative incidence of glucose intolerance by day +30 was calculated taking into consideration competing risk of death for right-censored data (28). Gray test (29) was used to compare the cumulative incidence of glucose intolerance between various subgroups. Multivariable regression using Fine–Gray proportional subdistribution hazard models (28) was used to determine the relationship between clinically relevant variables [demographics: age at HCT, sex, ethnicity; diagnosis: acute myeloid leukemia (AML), acute myeloid leukemia (ALL)/other leukemia, MDS)], risk of relapse at HCT, HCT-CI severity, stem cell source, conditioning intensity, body composition, prophylaxis for and development of aGVHD (time-dependent variable), and glucose intolerance, taking into consideration death as competing risk. Body composition measures considered for the multivariable model were limited to VAT and SMG due to its ability to provide a composite measure (quantity and quality) of skeletal muscle health (25–27). This analysis was performed for each BMI group (nonoverweight, overweight/obese). HRs and their 95% confidence intervals (CI) were determined to quantify magnitude of risk. Univariable Fine–Gray regression was conducted for each independent variable. In the nonoverweight BMI group, AML and ALL/other were combined because there was no significant difference in their HRs. MA with TBI and MA without TBI were also combined for the same reason in both BMI groups. Variables with P ≤ 0.10 in univariable analyses were all included in the multivariable model. Backward stepwise elimination was used to remove nonsignificant variables one at a time, starting with the least significant with P > 0.05, and the model reestimated. This was repeated until no more variables could be removed. Of note, BMI (continuous) was included in each of the multivariable models, to show the independent association between body composition measures and risk of glucose intolerance.

To evaluate the association between development of DM by day +30 and the risk of nonrelapse mortality (NRM) by 1 year after HCT for the entire cohort, we initially conducted a univariable Fine–Gray regression analysis in HCT patients who were alive at day +30 using BMI, glucose intolerance (yes, no), and the aforementioned clinically relevant variables as independent variables, and considering relapse-related mortality as competing risk. Multivariable regression was conducted as described earlier. All statistical analyses were two-sided, and a P < 0.05 was considered statistically significant in the final multivariable models. SAS 9.4 (SAS Institute Inc.) was used.

The data generated in this study are available upon request from the corresponding author.

Between 2007 and 2014, 1,012 adult patients underwent allogeneic HCT for acute leukemia or MDS. Of these, 110 were excluded from the current study due to a history of diabetes pre-HCT, 72 were excluded because they had undergone a previous HCT, 73 were excluded because abdominal CT images were obtained more than 90 days from HCT, and 8 were excluded because of poor image quality. The clinical characteristics of the 749 patients included in the current study are detailed in Table 1. The median age at HCT was 50.2 years (range 18–74 years). The majority (53.1%) were male, non-Hispanic (58.6%), had a diagnosis of AML (54.3%), and were overweight/obese (58.7%) by BMI. Regarding HCT-specific characteristics, most were at high risk of relapse (53.4%), had an unrelated donor (55.4%), and received tacrolimus/sirolimus-based GVHD prophylaxis (86.0%).

Overall, 8.1% of patients undergoing HCT developed glucose intolerance. The cumulative incidence of glucose intolerance among nonoverweight patients was 5.2%, and the cumulative incidence was 10.2% for those who were overweight/obese by BMI (Table 1). In univariable analysis for the overall cohort, abnormal VAT was associated with a greater than 2-fold (HR = 2.13; 95% CI, 1.28–3.58) risk of glucose intolerance; additional risk factors included Hispanic ethnicity (HR = 1.91; 95% CI, 1.14–3.2) and development of grade 2 to 4 aGVHD (HR = 1.78; 95% CI, 1.07–2.93; Table 2). Among nonoverweight patients (n = 309), abnormal SMG was associated with a greater than 3-fold (HR = 3.50; 95% CI, 1.28–9.60) risk of glucose intolerance;patients with MDS were also noted to have an increased risk (HR = 6.47; 95% CI, 1.96–21.34). Among patients who were overweight/obese by BMI (n = 440), abnormal VAT was associated with a greater than 2-fold (HR = 2.29; 95% CI, 1.27–4.13) risk of glucose intolerance; additional risk factors included Hispanic ethnicity (HR = 2.29; 95% CI, 1.28–4.11) and development of grade 2 to 4 aGVHD (HR = 2.03; 95% CI, 1.13–3.67; Table 2). The final multivariable regression model for the nonoverweight BMI group included abnormal SMG [yes, no (referent)] and diagnosis [MDS vs. AML, ALL, other (referent)], while the final model for the overweight/obese BMI group included ethnicity [Hispanic, non-Hispanic (referent)], VAT [normal (referent), anormal], and aGVHD [yes, no (referent)]. Of note, both models were adjusted for BMI (continuous). In multivariable regression analysis, pre-HCT abnormal SMG was an independent predictor of glucose intolerance among nonoverweight patients (HR = 3.00; 95% CI, 1.15–7.81), while abnormal VAT was an independent predictor among overweight/obese patients (HR = 2.26; 95% CI, 1.24–4.12; P = 0.008; Table 3.)

Of the 749 patients included in the study, 731 survived more than 30 days. Among patients alive at day 30, those who developed glucose intolerance had a significantly higher incidence of NRM at 1 year after HCT, compared with those who did not [21.3% (glucose intolerance) vs. 12.7% (no glucose intolerance), P = 0.032). The final multivariable regression model included glucose intolerance [yes, no (referent)], risk of relapse [low (referent), high], and GVHD prophylaxis [tacrolimus/sirolimus-based (referent), other]. Development of glucose intolerance was significantly and independently associated with risk of NRM (HR = 1.88; 95% CI, 1.05–3.39) at 1 year after HCT.

In this large, contemporary cohort of adults who underwent allogeneic HCT, we found significant associations between body composition measurements obtained from pre-HCT CT imaging studies and early-onset glucose intolerance after HCT. The components of body composition that increased risk for glucose intolerance differed by BMI; in nonoverweight patients, skeletal muscle loss increased this risk, while visceral adiposity was associated with increased risk in overweight/obese patients. These findings suggest that multiple mechanisms underlie the development of glucose intolerance and/or DM after HCT, and highlight the importance of improving DM risk characterization according to pre-HCT body composition phenotype. Furthermore, we report on the increased risk of NRM associated with de novo glucose intoleranc that is independent of established risk factors such as aGVHD severity and risk of relapse at HCT. Taken together, the findings from the current study may help set the stage for personalized risk assessments prior to HCT and for early screening and interventions in patients at risk for developing glucose intolerance and/or DM, to ultimately improve health outcomes in this patient population.

While there is a well-established relationship between BMI and glucose intolerance, studies in non-oncology and oncology populations have found that more representative measures of body composition such as skeletal muscle loss and abdominal adiposity may be better predictors of glucose intolerance and diabetes risk (30–32). For example, a recent study in HCT patients found no significant association between pre- or post-HCT BMI and glucose intolerance within the first 100 days after HCT (5). This reflects the limitations of BMI, which is solely a ratio of a person's height to weight that is not sex-specific and does not account for actual body composition.

A complex interplay exists between skeletal muscle health and glucose intolerance, as skeletal muscle is key for insulin-stimulated glucose disposal and one role of insulin is to prevent protein catabolism in muscle (13). Accordingly, glucose intolerance is characterized by decreased proportions of muscle fibers retaining sensitivity towards insulin-stimulated glucose uptake, and interventions directed towards improving muscle mass and quality and function can in turn improve diabetes risk (13, 33). Patients with cancer are at particular risk for skeletal muscle loss, with contributing factors including decreased activity, poor nutrition, and treatment-induced (e.g., chemotherapy, radiation) myopathy. Historically, measurement of muscle mass and quality have not been readily accessible in the clinical setting. Prior use of CT imaging to evaluate body composition has been limited by the need for manual identification and segmentation of images (34), a time-intensive process requiring personnel with training in anatomy and software applications. However, advancements in imaging technology and automated software have enabled routine clinical exams such as CT to be segmented and analyzed, providing real-time measures of body composition that can be readily incorporated into clinical care (17). Additional studies utilizing such automated software may improve our understanding of the relationship between body composition and health outcomes in real-time, findings which could be applicable to the non-oncology population as well. Given the lack of well-established definitions for adiposity and skeletal muscle measures, clinicians may need to contextualize their interpretation of body composition measures according to the populations of interest, due to likely differences in pathophysiology and phenotypes.

Previous studies have reported conflicting findings regarding transplant-specific risk factors, such as aGVHD, corticosteroid exposure, conditioning regimen, and calcineurin exposure for developing post-HCT glucose intolerance or DM (2, 3, 5, 6, 8, 35). In our study, conditioning regimen and GVHD prophylaxis were not associated with risk of glucose intolerance in the overall cohort, or in the two subcohorts. Interestingly, aGVHD was associated with risk of post-HCT glucose intolerance in the overweight/obese group, but not in the nonoverweight group. Potential mechanisms for this are that glucocorticoids, which are a first-line treatment for aGVHD, can increase VAT, alter lipid metabolism, and may also modify adipokine levels towards a more diabetogenic profile (36). Together, these studies and our findings suggest that comprehensive determination of body composition, specifically in nonoverweight patients, prior to HCT may allow for improved early detection and tailored intervention for those at greatest risk of developing glucose intolerance or DM. For example, overweight patients with abnormal VAT may require interventions to decrease adiposity, while nonoverweight patients with skeletal muscle loss may necessitate measures to increase skeletal muscle health to achieve the same goal of optimizing glucose and insulin regulation.

Studies have previously shown that glucose intolerance and/or DM is associated with worse overall survival after HCT, attributed in part to increased inflammation or immune dysregulation (3–5, 37). Similarly, in our study, development of glucose intolerance was significantly and independently associated with increased risk of 1 year NRM. While the exact mechanisms underlying this increased risk are unclear, it may reflect an increased burden of infectious complications and associated comorbidities (38). Current recommendations for diabetes screening in HCT recipients largely reflect those of the general population (i.e., screening every 3 years in those ≥45 years of age or with hypertension, or every 3 to 6 months for high-risk patients starting 3 months after HCT; ref. 39). However, as most patients develop glucose intolerance within 1 month of HCT (4, 5, 8), improved characterization of risk factors for glucose intolerance in this unique patient population would enable timely identification of particularly at-risk patients. This may be especially relevant for certain higher risk populations identified in this study, including patients with Hispanic race/ethnicity or those with MDS undergoing allogeneic HCT. Additional studies with longer follow-up and external validation are needed to better delineate demographic and pre-HCT treatment-related modifiers of risk.

The findings from this study should be considered in the context of its limitations. As data regarding glucose intolerance diagnosis and onset date were obtained through retrospective chart review, it is possible the actual incidence of glucose intolerance by day +30 was higher than that captured by medical record abstraction. That said, the study included clinically significant measures during a time when patients are closely monitored primarily in the inpatient setting. As such, the proportion of patients with glucose intolerance who were not captured in our current study is likely to be very low. We also acknowledge that glucose intolerance, as captured in this study, is not necessarily DM. Additional studies with longer follow-up are needed to help characterize the association between body composition, pre-HCT glucose intolerance, and long-term DM risk. The CT images that were utilized for this study were obtained from routine pre-HCT scans and were not for the purposes of measuring muscle and adipose tissue. However, we used validated automated software to quantify skeletal muscle and VAT with quality control conducted by research team members who were blinded to the outcomes of interest. While we did not assess the relationship between the intensity and duration of specific medication exposures (e.g., corticosteroids) and development of glucose intolerance, we did evaluate severity of aGVHD, which served as a surrogate for corticosteroids since they are often the initial line of treatment. Our findings that aGVHD was an independent predictor of glucose intolerance in patients with overweight/obese BMI provide further evidence regarding the role of iatrogenic risk factors in glucose intolerance and/or DM risk after HCT. Despite these limitations, to our knowledge this is the first study to evaluate risk factors for development of glucose intolerance after HCT taking into consideration the potentially different pathophysiology amongst patients with overweight/obese versus nonoverweight BMI. Future studies may need to evaluate the role blood biomarkers (e.g., inflammatory markers) and readily available clinical assessments (e.g., handgrip strength, short physical performance battery) can play in facilitating timely detection of pre-HCT skeletal muscle loss, while minimizing patient burden. Such biomarkers and assessments may ultimately be more widely translated to patients without planned CT imaging.

In conclusion, we found that the development of glucose intolerance post-HCT was associated with skeletal muscle loss in nonoverweight patients and with excess visceral adiposity in overweight/obese patients, and that glucose intolerance was associated with an increased risk of 1 year NRM. Nonoverweight patients with skeletal muscle loss and overweight patients with increased visceral adiposity may benefit from prehabilitation and nutritional optimization prior to HCT as well as tailored medical approaches during their HCT admission to mitigate the subsequent development of glucose intolerance and/or DM. The development of these targeted screening approaches and interventions are necessary to improve overall morbidity and mortality for HCT recipients.

R. Nakamura reports personal fees from Bluebird, Sanofi, Magenta, and Viracor; grants and personal fees from Omeros; other support from Miyarisan; and personal fees from Kadmon outside the submitted work. No disclosures were reported by the other authors.

R. Bhandari: Conceptualization, data curation, formal analysis, validation, writing–original draft, writing–review and editing, data interpretation. J. Berano Teh: Data curation, validation, writing–review and editing. T. He: Formal analysis, writing–review and editing. K. Peng: Data curation, writing–review and editing. A. Iukuridze: Data curation, validation, writing–review and editing. L. Atencio: Data curation, writing–review and editing. R. Nakamura: Writing–review and editing, data interpretation. S. Mostoufi-Moab: Writing–review and editing, data interpretation. S. McCormack: Writing–review and editing. K. Lee: Conceptualization, resources, data curation. F.L. Wong: Formal analysis, supervision, writing–review and editing. S.H. Armenian: Conceptualization, formal analysis, supervision, writing–original draft, writing–review and editing, data interpretation.

This study was supported, in part, by grants from the Leukemia and Lymphoma Society Scholar Award for Clinical Research (grant no. 2315–17 to S.H. Armenian) and NIH/NCI (grant no. R01CAHL150069 to S.H. Armenian).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.