Predicting Postallogeneic Hematopoietic Cell Transplant Glucose Intolerance: The Role of CT-based Body Composition

In this issue of Cancer Epidemiology, Biomarkers and Prevention, Bhandari and colleagues report on the association between body composition (i.e., CT measures of muscle and adipose tissue) and glucose intolerance at 30 days after receiving allogenic hematopoietic cell transplant (HCT) among adult patients with myeloid malignancies (1). The study is based on the scientific premise that a significant proportion of patients develops glucose intolerance shortly after allogenic HCT leading to increased risk of cardiometabolic complications, impaired health-related quality of life (HRQOL) and reduced survival. However, to date, risk factors for developing glucose tolerance following HCT are not fully understood. In particular, as highlighted by the authors, the use of body mass index (BMI) alone to estimate such risks has well known shortcomings (2, 3). Because insulin resistance is largely determined by body fat and muscle distribution, and BMI, by itself, cannot distinguish between muscle and fat or bone mass, there has been a renewed interest to directly quantify these body compartments to better predict this risk (4).

To address their research question, Bhandari and colleagues have conducted a retrospective cohort study using of adults with myelodysplastic syndrome or acute leukemia undergoing allogenic HCT at their institution. Among 1,012 consecutive patients undergoing allogenic HCT over a 7-year period, the authors identified 759 eligible nondiabetic individuals who had available CT scans within 90 days of HCT. Using L3-segmental slices from archived scans, the authors were able to accurately measure body composition using previously validated methods (5, 6) and quantified muscle mass, muscle density, visceral, and subcutaneous adipose tissue mass. The authors then studied the association between those body composition indices and incident dysglycemia within 30 days of allogenic HCT. The latter was captured by reviewing medical records and defined as any dysglycemia requiring dietary or pharmacologic intervention. Using univariable and multivariable models, Bhandari and colleagues were able to show that low muscle mass and excess visceral adiposity were strongly associated with development of incident dysglycemia at 30 days; in addition, they were able to show that incident dysglycemia was associated with an increased risk of 1-year nonrelapse mortality. The authors argue that such patients could be the focus of targeted prehabilitation intervention to optimize their risk prior to undergoing HCT and mitigate the risk of such adverse outcomes (1).

So what is body composition and why does it matter? Our body is made of distinct tissue types, such as muscle and fat, with their own unique metabolic properties. A healthy balance between fat and muscle is considered essential for maintaining good health and fitness. Furthermore, the aging process is often accompanied by progressive losses of muscle mass with parallel gains in visceral adiposity which are strong risk factors for poor HRQOL, future disability, and survival. As such, accurately measuring these body compartments can help identify individuals at risk of adverse outcomes. In fact, interest in human body composition research can be dated back to the Hippocratic era. The modern phase of this field began at the turn of the 21st century, with several notable technological advancements including air displacement plethysmography, dual-energy X-ray absorptiometry, bioelectric impedance analysis, and MRI (4, 7). Yet, such tools are not readily available in routine oncology practice, thus limiting our understanding of how body composition affects outcomes of individuals with cancer. To that end, recent advancements in body composition research have enabled the use of single slices images from routinely obtained CT scans to accurately quantify skeletal muscle mass and visceral adiposity (5, 6). Such tools have led to a “reawakening” of body composition research in oncology and allowed investigators to study their impact across different cancer types and treatment regimens. In fact, a quick search of the MEDLINE database via Pubmed with search terms “Body Composition” and “Cancer” reveals 4,697 hits, with a sharp peak after mid-2000s, which somewhat coincides to the timing of the aforementioned publications (5, 6).

The study by Bhandari and colleagues is well designed and addresses an important clinical question. The study cohort comprises of a large well-characterized group of consecutive patients undergoing allogenic HCT at a single institution over a 7-year period. Whereas institutional practices vary, it appears that CT was was obtained for routine pretransplant assessment in nearly all patients, therefore minimizing the possibility of selection bias for those who had available CT scans versus those who did not. The study analytic methods are sound and the conclusions are compelling. Nevertheless, there are some shortcomings worth noting. In particular, the follow-up period is short and the event rate is relatively low (≈8%). Furthermore, it is not clear whether dysglycemia is truly an incident event or whether the subjects may have had dysglycemia or even undiagnosed Diabetes Mellitus (DM) prior to allogenic HCT. We do note that the authors excluded those with known type II DM prior to HCT as documented in medical record; however, because standardized assessment of dysglycemia was not done pretransplant and it is possible that participants with preexisting dysglycemia or even undiagnosed type II DM were included in the study. To that end, it is unclear whether body composition variables are just enabling identification of individuals who have preexisting insulin resistance versus identifying those who are developing de novo insulin resistance shortly after HCT. Finally, we also do not know what proportion of these hyperglycemic episodes were transient due to physiologic response (stress hyperglycemia), steroid induced or more sustained due to an underlying insulin resistance and development of subsequent type II DM.

So what are the future directions? In our opinion, the study results should be regarded as potentially hypothesis generating and need to be confirmed by additional reports, ideally with longer follow-up and more rigorous outcome ascertainment. In addition, the incremental value of using this assessment beyond more readily available predictors of insulin resistance such as waist circumference needs to be evaluated (8). Finally, because both skeletal muscle and visceral adiposity work at tandem in causing insulin resistance, additional research should be focused on identifying combined phenotypes rather than teasing out the individual impact of these body tissue types. Nevertheless, it is becoming increasingly clear that body composition measures can identify cardiometabolic risk above and beyond what is informed by BMI. In fact, we find it astonishing that an equation (BMI) that was described 150 years ago continues to be used to classify the health status of individuals, despite compelling evidence detailing its limitations, including the current study. We must move on!