The relationship between abdominal fat and change in left ventricular ejection fraction in cancer patients

Abstract Objectives Prior studies have identified a relationship between body mass index (BMI) and intraperitoneal (IP) fat with heart failure; however, in prior studies of cancer patients receiving potentially cardiotoxic chemotherapy, elevations in BMI have not necessarily been associated with decrements in heart function. This study tested the hypothesis that IP fat may be associated with left ventricular ejection fraction (LVEF) decline among cancer patients receiving potentially cardiotoxic chemotherapy. Methods In this prospective study of 61 cancer patients (23 breast cancer, 32 lymphoma, and 6 sarcoma), IP fat and other assessments of body composition, and changes in LVEF from pre‐ to postcancer treatment using noninvasive magnetic resonance imaging was ascertained. Results After accounting for age, baseline LVEF, and confounding variables, pre‐ to 24‐month post‐treatment LVEF changes were inversely correlated with IP fat (r = −0.33; p = 0.02) and positively correlated with measures of subcutaneous (SQ) fat (r = 0.33; p = 0.01). These LVEF changes were not correlated with BMI (r = 0.12; p = 0.37). Conclusion Among patients receiving potentially cardiotoxic chemotherapy, pretreatment IP fat was associated with subsequent declines in LVEF. There was no association between BMI and LVEF decline. These findings may be related to a potential protective effect of SQ fat.


| INTRODUCTION
Obesity is defined not by excess weight but by excessive fat accumulation, and is an independent risk factor for cardiovascular diseases (CVDs), including heart failure. 1 While many studies employ body mass index (BMI) as a proxy for obesity, it is an imprecise measure of adiposity. 2 This may not be problematic in the general population where BMI remains a strong predictor for CVD 3 ; however, in individuals with chronic diseases, including cancer, it may be a less satisfactory metric. 4,5 One reason is that BMI does not differentiate different body composition compartments, such as fat mass and skeletal muscle mass. 2,6,7 As a result, in cancer survivors, in whom sarcopenic obesity may occur in response to cancer treatment, BMI may be a particularly poor measure of adiposity. [7][8][9] Nor does BMI discriminate the location of adiposity. While measures of abdominal adiposity, such as waist to hip ratio, typically provide an improved risk stratification compared to BMI alone, 10 further improvement may be gained through deep phenotyping of adiposity as differing fat depots convey disparate physiologic impacts depending on anatomical location. 11 This has been well established, as visceral adipose tissue (VAT) is consistently linked with increased cardiometabolic risk, in contrast with a lack of association or potentially reduced risk with subcutaneous (SQ) fat. [11][12][13] Still more precision can be gained from separating VAT into its components of intraperitoneal (IP) and retroperitoneal (RP) fat. 14 IP fat, which contains blood high in lipids potentially due to its location near the hepatic portal vein, 15 is of particular interest because it is an independent predictor of cardiometabolic risk. 16 While the distinction of anatomical distribution of adiposity has been appreciated for some time in the cardiovascular literature, 10 it has received less attention in studies of cancer patients focused on CVD. With CVD being a primary contributor to morbidity and mortality in cancer survivors treated with potentially cardiotoxic chemotherapy, studies have sought to better understand the role of patient risk factors. 17 Some have shown that BMI modifies the effect of anthracyclines on left ventricular ejection fraction (LVEF), 18,19 a marker often used to define cardiotoxicity, 20,21 while others have not. [22][23][24][25][26][27][28] A potential reason for the inconsistent findings is the imprecision with which BMI approximates obesity and the varying sample sizes of the studies. The goal of this analysis was to investigate the relationships between obesity, IP fat, and LV dysfunction associated with the administration of potentially cardiotoxic chemotherapy through the assessment of postdiagnosis LVEF decline. As such, assessing IP fat as well as BMI may provide the information necessary to help identify the relationship between obesity and LVEF decline at 24-month postdiagnosis. CMR images to assess LVEF were acquired at baseline (i.e., prior to the first cycle of chemotherapy) and at 24 months postbaseline.

| Patient population
In addition, at baseline body composition was ascertained using measures of height, weight, and abdominal MRI measures of intraabdominal and SQ fat. This study was approved by the Wake Forest Health Sciences Institutional Review Board and all participants provided written, witnessed informed consent.

| CMR analysis
To measure LVEF, images were acquired using a 1.5 Tesla Avanto (Siemens Healthcare) MRI scanner. MR imaging was chosen to assess LVEF due to its accuracy and prior use in US National Institutes of REDING ET AL. -83 Health-funded initiatives such as the Multi-Ethnic Study of Atherosclerosis (MESA). 29 LVEF measurements were obtained using previously published methods, 30 that included cine bright blood steadystate free precision techniques with 160 � 120 matrix, a 42 cm field of view, an 8-mm-thick slice with a 2-mm inter-slice gap, and a 33-ms temporal resolution.
The CMR cine slices were manually analyzed using QMASS  For adjustment of confounders, variables were considered that a priori were considered to be a risk factor for LVEF decline. In addition, the variable "time between visits" was used to account for varying time from baseline to the 24-month visit (for which the mean timeframe for the 24-month visit was 24.72 months from baseline [interquartile range: 24.12-24.96]) and was included in all models.

| Statistical analysis
Thus, our multivariable model included age (in years), gender (male or female), baseline LVEF, receipt of anthracyclines (yes or no), cancer site (sarcoma, lymphoma, or breast cancer), cardiovascular risk factors (a summary variable of hypertension, diabetes and coronary artery disease [CAD]), time between visits, as well as adjustment for IP fat when investigating SQ fat and adjustment for SQ when investigating VAT, IP, and RP fat.
Next, the impact of the IP fat and SQ fat on the 24-month LVEF change was examined using a general linear model (GLM) adjusted for multivariable model described above. Then the relationship of IP fat and SQ fat with a 24-month drop in LVEF of 5%, equivalent to the mean decline in LVEF, was examined using logistic regression and a receiver operator curve (ROC) analysis. Secondarily, a LVEF decline was modeled for a commonly used threshold of either a 10% decline from pre-to post-treatment or a decline to 50%. 21 To assess the overall model fit, the likelihood ratio test was used. A multivariable model was fit using the same set of adjustment variables described above. Then the depots of IP fat and SQ fat were added to this model. Lastly, exploratory analyses were conducted using a variable that combined IP and SQ fat depots modeled as a ratio of IP: SQ fat. This variable was used as the independent variable of the GLM models, and selected a parsimonious model of adjustment (i.e., those significantly associated with LVEF decline which included baseline LVEF and anthracycline use) plus age, gender, and time between visits. All statistical analyses were performed using SAS (SAS Institute).

| RESULTS
In this dataset of cancer survivors, participants had a mean age of 53.37 years (SD: 15.25), mean weight of 86.11 kg (SD: 17.87) and a mean BMI of 30.14 kg/m 2 (SD: 5.73; Table 1). Nearly one-third of our sample was men and approximately three-quarters were Caucasian.

| DISCUSSION
In this study among patients treated with potentially cardiotoxic treatment, LVEF decline at 2 years was inversely correlated with abdominal fat in the IP region of the viscera and positively correlated with SQ fat at cancer diagnosis. This is contrasted with our finding that BMI, which does not differentiate body fat distribution, was not associated with 24-month postchemotherapy declines in LVEF. In addition, our data suggest that addition of prediagnostic IP and SQ BMI has been investigated in relation to cardiotoxicity based on its role in the development of CVD and due to the use of BSA as a factor guiding chemotherapy dosing. [35][36][37] Our findings of a lack of association with BMI and LVEF decline differ from those in a 2016 meta-analysis involving breast cancer patients who were treated with anthracyclines alone or sequential anthracyclines and trastuzumab. 35 While the meta-analysis showed a 1.38-fold increased risk of cardiotoxicity associated with a BMI ≥25 kg/m 2 , the meta-analysis study design did not allow for control of obesity-related cardiovascular risks factors such as diabetes and hypertension, thus leaving the possibility that unmeasured confounding by cardiovascular risk factors were responsible for the elevated risk. Moreover, methodologic issues have been raised pertaining to the pooling of studies with different designs, sample sizes, and definitions of cardiotoxicity that could influence the overall study results. 38 Our analysis, by contrast, used precise CMR-derived measures of LVEF, and would not necessarily suffer from this concern.
These findings showed relationships of IP fat and SQ fat with LVEF decline in opposing directions. For interpretability, two sample patients were modeled with contrasting IP and SQ fat depots, showing that a patient with low IP and high SQ fat had a LVEF decline less than -87 5%, the mean decline in cancer patients. This was contrasted with a patient with high IP and low SQ whose LVEF decline equaled 15.68, corresponding to a 24-month LVEF of 46.34%, which exceeds a commonly used threshold of LVEF decline for cardiotoxicity. 20,21 This concept of a body composition phenotype of high visceral fat with low SQ fat is well described in the literature. 11,13,39,40 Mechanistically, in the presence of excessive adipose accumulation, SQ fat may be redirected to ectopic fat storage, including in the IP region. 11,39 When the SQ fat depot is overwhelmed by an overabundance of lipids, it shunts lipids away from SQ to a visceral placement of adipose tissue, thereby creating a phenotype of low SQ in the presence of high VAT. 10 The implications of this are described in an extensive set of literature showing that an accumulation of VAT, particularly in the IP region, is associated with increased cardiometabolic risk. 13

| CONCLUSIONS
Our data suggests that assessment of locations of adipose storage is useful in understanding the relationship between obesity and late effects of potentially cardiotoxic cancer therapy. This could be due to the additional information provided by the comparison of SQ fat to ectopic fat storage in the IP region of the viscera. While additional studies are needed to replicate these findings prior to these results being actionable in a clinical setting, the physiologic basis of our findings are supported by the literature in obesity and CVD that point to a role of elevated IP in relation to reduced SQ fat. 13,15,16 However, given the challenge of measuring depots of fat through imaging on all patients, the development of predictive statistical models to estimate IP and SQ fat depots, such as those developed in adolescents, 46