Tri-ponderal mass index in survivors of childhood brain tumors: A cross-sectional study

Survivors of childhood brain tumors (SCBT) face a higher risk of cardiometabolic disorders and premature mortality compared to the general population. Excess adiposity is a known risk factor for these comorbidities. However, while SCBT have higher adiposity compared to healthy controls, measuring adiposity in clinical practice involves access to specialized equipment and may impact busy clinical services. Tri-ponderal Mass Index (TMI; kg/m3) may be a superior measure of adiposity when compared to Body Mass Index (BMI; kg/m2). However, its use in determining adiposity in SCBT has not been assessed. This study aims to validate TMI as a clinical measure of adiposity in SCBT. This was a cross-sectional study including 44 SCBT (n = 20 female) and 137 (n = 64 female) non-cancer control children, 5–17 years of age. BMI and TMI were calculated from height and weight measurements. Fat mass percentage was assessed using bioelectrical impedance analysis and waist to hip and waist to height ratios were used to assess central adiposity. Regression analyses were adjusted for age, sex, puberty and treatment. TMI demonstrated strong correlations to measures of total and central adiposity and predicted adiposity in SCBT and non-cancer controls, with stronger trends in the latter group. TMI may serve as a reliable clinical measure of adiposity in both SCBT and healthy children.

Index (BMI, kg/m 2 ) and BMI z-score when validated against DXA scans in adolescents 31 . An important advantage of TMI is that it uses the same height and weight measurements used to calculate BMI. Thus, TMI may offer a feasible clinical measure to assess adiposity. However, TMI has not been validated as a measure of adiposity in SCBT. Our aim is to validate TMI as a clinical measure of adiposity in SCBT and compare this group to non-cancer controls.
Due to the age and pubertal staging differences noted, we validated the results of the analysis of the full cohort by performing a subgroup analysis that included age-and sex-matched controls. In the latter analysis age, sex, and puberty adjusted analyses revealed identical trends to those reported for the full study cohort. The data for the matched subgroup analyses are reported in Supplementary Tables S1-S3.
Tumor characteristics and treatments. The details of tumor characteristics and treatment methods are reported in Table 2.
The association of TMI with body mass and adiposity. To assess if TMI correlates with body mass measures and adiposity, we used Spearman's correlation test. We conducted unadjusted and age, sex, and puberty-adjusted analyses (Table 3).
We calculated unstandardized (B) and standardized (β) coefficients, and both trended in the same direction. Moving forward, we report on the latter coefficient (Table 4).

Variables
No. (%)  Table 3. Spearman's correlation of TMI and BMI z-score with adiposity measures in SCBT and controls. *p-value < 0.05, **p-value < 0.001. Abbreviations: SCBT, survivors of childhood brain tumors; BMI, body mass index; TMI, tri-ponderal mass index; %FM, percent fat mass; WHR, waist-to-hip ratio, WHtR, waist-to-height ratio. The above threads of data indicate that TMI is a strong predictor of total adiposity and WHtR. The association between TMI and adiposity appears to be stronger in controls compared to SCBT for BMI z-score, %FM and WHtR.

Discussion
Survivors of childhood brain tumors are facing multiple comorbidities including cardiovascular disease and type 2 diabetes, which can impact their quality of life and lifespan [13][14][15]27,[32][33][34][35][36] . The identification of predictors and markers of cardiometabolic risk may offer a path to stratify those in need of close observation and early intervention. In this study, we identified TMI as one such measure. TMI was an equally strong predictor as BMI z-score, total adiposity and WHtR, the latter being a stronger predictor of cardiometabolic risk compared to WHR [37][38][39] . As central adiposity is associated with adverse cardiometabolic outcomes [37][38][39][40] , this is of great clinical significance, as it allows the stratification of children with higher central adiposity to a care stream with closer cardiometabolic health monitoring and early more aggressive interventions.
TMI has been validated against DXA scan-measured adiposity in the general pediatric population 31 . Our data adds to the value of TMI in the general pediatric population and in a population with chronic health needs that has not been studied previously. TMI is a promising marker of adiposity that is clinically feasible and informative. The most widely used clinical measure of body mass, BMI, is a useful population-based measure to report the presence of obesity, and is used interchangeably to report adiposity 41 . However, one of the limitations of BMI is that it may not be adequate to diagnose obesity in certain populations, including adolescents 42,43 , hence the use of BMI z-score and percentile data to assess overweight and obesity in children. In addition, BMI misclassifies muscular individuals as being overweight or obese, which does not necessarily reflect their future risk of cardiometabolic disorders 44 . Furthermore, BMI has a weaker association with cardiovascular risk when compared to measures of adiposity including waist circumference and WHtR 37,45,46 .
Children require special consideration in using BMI to classify obesity and adiposity. Ratios of weight over height to various powers of rho, ρ, (weight/height ρ ) have been explored to account for the effects of children's growth during puberty, and the adjustment for ρ is critical because incorrect values misclassify tall or physically advanced children as overweight 47,48 . A ρ value equal to two as used in BMI is sufficient when height is constant, however during puberty changes in height increase the ρ value 48 .
In pre-school children, weight over height squared is adequate for assessing adiposity 47 . Adiposity generally declines between the ages 5-7 years before it begins to rise again, the adiposity rebound phase 49 . An earlier adiposity rebound than expected is associated with an increased risk of obesity and type 2 diabetes in adults [49][50][51] .
As children approach the peripubertal phase of growth and development, their body composition changes with increased adiposity, especially in girls 52,53 . The value of ρ gradually rises from two to three; children who have undergone a growth spurt due to puberty tend to be heavier when compared to less mature children at the same height 47 . The increase in weight accompanying growth spurts results in higher BMI values, therefore greater values of ρ are required to offset the weight gain experienced in physically advanced children.  However, body composition changes in children become more constant as they get older and ρ decreases back to two 47,48 . Therefore, during puberty, TMI may be a more accurate measure of adiposity in children 47 . For this reason, the use of TMI is more relevant to assess body fat mass in children and, when evaluated in adults, TMI is less reliable compared to BMI when correlated with skinfold thickness 41 .
Measures of adiposity in children have relied on technologies that may not be readily available in the clinical setting but have demonstrated accuracy in estimating the fat mass. DXA estimates of trunk and abdominal fat have demonstrated a strong association to total abdominal fat [28][29][30] , while Bioelectrical impedance analysis (BIA) has been validated as a measure of adiposity against DXA [54][55][56] . Our results demonstrate that TMI is a strong predictor of adiposity measured using BIA, which is congruent with adiposity assessments using DXA. This is another strength of this study, as it validates TMI against BIA, a common measure of adiposity that is more easily accessible than DXA.
Our results indicate that TMI may offer a better estimate of fat mass and is a potential tool for predicting adiposity in children compared to BMI z-score. While some studies have found TMI to be an appropriate measure of adiposity in children, others have suggested BMI may still be an equivalent measure 31,57 . However, these studies often include children as young as two years of age, and weight scales over height squared in this age group that makes validating TMI as a measure of adiposity in this group a future goal of research 47,57 .
There are several strengths of our study. The inclusion of non-cancer controls for comparison to the SCBT group offers validation of this measure in the general pediatric population as well as SCBT. The description of the association of TMI with measures of central adiposity is another strength, as central adiposity is not routinely assessed in clinical practice. Furthermore, the inclusion of an age and sex matched subgroup analysis validates our findings.
A larger sample size of SCBT is required to validate these results further, and to define their associations. In addition, TMI needs to be validated as a predictor of cardiometabolic outcomes which should be part of longitudinal studies.
In conclusion, TMI represents a clinically feasible measure that uses the same variables measuring BMI but demonstrate higher correlation with adiposity.
The availability of TMI as a clinical measure of adiposity will allow the stratification of patients at risk of excess adiposity to be prioritized for targeted interventions. This is critical, as these survivors are facing cardiometabolic diseases that are important emergent determinants of outcomes.

Participants. Participants were consecutively recruited from McMaster Children's Hospital (Hamilton),
Ontario, Canada from November 2012 to November 2017 to partake in the Canadian Study of Determinants of Endometabolic Health in Children (CanDECIDE study) 58,59 .
Children between 5-17 years of age were included. Parental informed consent was obtained for participants less than 7 years of age, and parental consent and participant assent were obtained for those 7-15 years of age. Participants provided written informed consent if they were 16 years or older. This is a secondary analysis of the CanDECIDE cohort study data and this analysis has been approved by the Hamilton Integrated Research Ethics Board. Study procedures were performed in accordance with the relevant guidelines and legal regulations.
Anthropometric and clinical measurements. Standardized questionnaires were used to collect data on age, sex, puberty, and ethnicity 58,59 . Medical records were consulted to verify and collect data regarding tumor type, location, sidedness and treatment modalities for SCBT.
Anthropometric measurements included weight measured to the nearest 0.1 kg using an electronic weighing scale (Seca, USA) and height using a stadiometer measured to the nearest 0.1 cm. BMI (kg/m 2 ) and TMI (kg/m 3 ) were determined using height and weight measurements from all participants. BMI percentile was determined from the Children's BMI Tool for Schools 60 . BMI z-scores were determined from the Centers for Disease Control and Prevention (CDC) growth chart 61 .
Fat mass percentage (%FM) was used to determine adiposity in the participants that was measured with the Tanita body fat monitor (Tanita Corporation, Illinois, USA) 62 .
Statistical Analysis. Statistical analyses were performed using PASW version 18 statistical package 63 . Data are presented as counts with percentages for categorical variables and means with standard deviation for continuous variables. Only participants with complete datasets were included. Box plots and visual inspection were used to identify any outliers for removal from the analysis. Normality of the data distribution was assessed using the Shapiro-Wilk test 64 . In the case that variables had non-normal distributions, data were log-transformed. Sample size was calculated using the method proposed by Norman and Streiner. We calculated that we need eight subjects per variable to detect significant differences between groups 65 .
Spearman's correlations were used to assess the relationship between TMI, BMI z-score, %FM, WHR and WHtR. To assess this relationship adjusted for age, sex and puberty, we ran a Partial Correlations test. Multivariable linear regression analyses were performed to determine the association between TMI with BMI z-score and adiposity measures (%FM, WHR and WHtR), in SCBT and controls. The dependent variable was set as %FM, WHtR, WHR or BMI z-score and the independent variables included TMI, age, sex, puberty and treatment. Results were reported as standardized β coefficients and associated p-values, with statistical significance set to alpha of 0.05. A model summary, including the adjusted R Square and the Standard Error of the Estimate, were also reported.
In order to validate that age and puberty differences did not affect the results, analyses were also repeated using an age-and sex-matched control group. Control participants were matched in terms of sex distribution to SCBT participants on a one-to-one ratio and age was matched to closest value, within three years of the SCBT participants.

Data Availability
The data for the current study used for statistical analysis are available from the corresponding author upon reasonable justification.