Tri-Ponderal Mass Index as a Screening Tool for Identifying Body Fat and Cardiovascular Risk Factors in Children and Adolescents: A Systematic Review

Because of the limitation of body mass index (BMI) in distinguishing adipose mass from muscle, the tri-ponderal mass index (TMI) has been proposed as a new indicator for better assessing adiposity in children and adolescents. However, it remains unclear whether TMI performs better than BMI or other adiposity indices in predicting obesity status in childhood and obesity-related cardiovascular risk factors (CVRFs) in childhood or adulthood. We searched PubMed, Cochrane Library, and Web of Science for eligible publications until June 15, 2021. A total of 32 eligible studies were included in this systematic review. We found that TMI had a similar or better ability to predict body fat among children and adolescents than BMI. However, most of the included studies suggested that TMI was similar to BMI in identifying metabolic syndrome although TMI was suggested to be a useful tool when used in combination with other indicators (e.g., BMI and waist circumference). In addition, limited evidence showed that TMI did not perform better than BMI for identifying specific CVRFs, including insulin resistance, high blood pressure, dyslipidemia, and inflammation in children and adolescents, as well as CVRFs in adults. Systematic Review Registration https://www.crd.york.ac.uk/prospero, CRD42021260356.


INTRODUCTION
The age-standardized prevalence of obesity among children and adolescents aged 5 to 19 years has globally increased from 0.7% in 1975 to 5.6% in 2016 among girls and 0.9% to 7.8% among boys (1). Obesity in children is a cause of several detrimental health outcomes in childhood and later in adulthood, such as left ventricular hypertrophy (2), increased carotid intima-media thickness (3), kidney disease (4), and liver disease (5), cancer, cardiovascular diseases, and death (6)(7)(8). Thus, an early and accurate diagnosis of obesity in children and adolescents is urgently needed, in order to reduce the short-term and long-term burden of pediatric obesity-related health outcomes.
Body mass index (BMI, kg/m 2 ) is the most widely used physical indicator of adiposity in both children (with overweight/obesity cutoffs based on age and sex percentiles) and among adults (overweight:  obesity BMI ≥ 30). Although BMI is strongly correlated with adiposity, the indicator cannot distinguish well between excess weight due to increased fat mass or increased muscle mass (9), especially for changes in body composition during adolescence, leading to weight increase being out of proportion of the change in height squared (10,11). Although the percentage of body fat is suggested as an accurate method for identifying obesity in children and adolescents, it is less applicable for routine health care, as well as in school-based settings (12).
Tri-ponderal mass index (TMI), calculated as weight (kg)/ height (m 3 ), is an emerging indicator, which has been suggested to predict percent body fat (10) and metabolic syndrome (MetS) (13) at least as well as, or better than BMI. However, findings in other previous studies were inconsistent (14)(15)(16)(17). For instance, the prevalence of overweight and obesity was higher when identified with BMI (based on standard deviation score, SDS) than with TMI in children and adolescents aged 6-17 years (14). It was also found that BMI (or BMI z-score or BMI-SDS) predicted MetS better than TMI among adolescents aged 10-17 years (15,16).
It is however unclear whether the emerging TMI can better identify adiposity in childhood or adolescence than the commonly used BMI (10,14,(18)(19)(20)(21)(22)(23)(24)(25)(26) and better predict obesity-related cardiovascular risk factors (CVRFs) such as high blood pressure, dyslipidemia, insulin resistance, and the MetS in childhood (13-16, 18, 21, 26-34) or adulthood (35)(36)(37). The misclassification of obesity may lead to either omissions of children who are at high risk of obesity-related diseases or excessive anxiety due to overdiagnosis and then unnecessary waste of medical resources (38). Identifying potential adiposityrelated indicators that can accurately predict body fat or related risks has significant implications for prevention, treatment, and management of pediatric obesity. Therefore, in order to assess whether TMI can be a substitute for BMI in routine pediatric clinical practice to estimate obesity and related CVRFs in children and adolescents or adults, we reviewed articles on the ability of TMI to identify increased body fat, in children and adolescents, and to predict CVRFs in both childhood and adulthood.

Search Strategy
This review was performed according to the recommendation from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) (http://www.prismastatement.org/). We searched relevant articles in PubMed, Cochrane Library, and Web of Science until June 15, 2021 using the following search strategy: ("Triponderal mass index" OR "Triponderal mass index" OR "Tri-ponderal index") AND ("children" OR "childhood" OR "adolescents" OR "adolescence" OR "teenagers" OR "youth" OR "students") AND ("body mass index" OR "obesity" OR "body fat" OR "cardiovascular disease risk" OR "hypertension" OR "dyslipidemia" OR "insulin resistance" OR "impaired glucose" OR "metabolic syndrome" OR "MetS" OR "inflammation"). We also identified eligible papers from the lists of references in the identified papers. We have registered on PROSPERO (available at: https://www.crd.york. ac.uk/prospero/#aboutpage), and the ID is CRD42021260356.

Inclusion Criteria and Exclusion Criteria
Inclusion criteria were as follows: 1) original article; 2) body fat or adiposity assessed using TMI and BMI in childhood or adolescence; 3) the paper described the association of TMI and BMI measured in childhood and adult with selected CVRFs [i.e., hypertension; dyslipidemia; insulin resistance (IR) or impaired glucose; MetS; and inflammation] measured either in childhood (e.g., at the same time of measurement of the BMI/TMI, e.g., in cross-sectional surveys) or in adulthood (e.g., cohort studies) or both; and 4) cross-sectional, cohort, or retrospective studies. Exclusion criteria were as follows: 1) obviously irrelevant articles; 2) TMI measured in adulthood; 3) other languages rather than English; 4) letter or comment; and 5) studies without data of interest.

Identification of Relevant Studies and Data Extraction
Two independent authors (JS and RY) performed the literature search and extracted the data. In case of disagreement between the two authors, a third expert (BX) was consulted to reach an agreement. The information on the first author, publication year, country of origin, study design, sample size, age and sex distribution of the study population, exposures, outcome definition, adjusted covariates, and results was extracted from each eligible study.

Study Quality Assessment
An 11-item checklist of the cross-sectional study evaluation scale recommended by the Agency for Healthcare Research and Quality (AHRQ) was used to evaluate the quality of crosssectional studies (39), with answers coded as "Yes" (1) or "No or not clear" (0). The total score of the scale is 11 points. A score of 8 to 11 points is rated as high quality, 4 to 7 points as moderate quality, and less than 4 points as low quality. The Newcastle-Ottawa Scale (NOS) star system (range 0 to 9 stars) was used to evaluate the quality of cohort and case-control studies (40). In brief, four items related to the selection of participants, two items to the comparability of participants, and three items to the exposures or outcomes. A score of "0-3" was regarded as low quality, "4-6" as moderate quality, and "7-9" as high quality.
It has been shown that percent of body fat (BF%) as a gold standard was better predicted by TMI than by BMI (10,25), although one study reported that both relative fat mass pediatric (RFMp) based on height and waist circumference [WC], and waist-to-height ratio (WHtR) performed better than both TMI and BMI (25). When WHtR was used to define central obesity as the gold standard, three articles showed that TMI was better than BMI correlated with central fat accumulation in both preschoolaged children aged 2-5 years (24) and children and adolescents aged 5-17 years (19,43).
When BMI and TMI were used as continuous variables, TMI correlated similarly or better than BMI with BF% in children and adolescents (18,19,22,23,41,42). Although TMI and BMI among children and adolescents aged 5-18 years explained a similar proportion of the variability for BF%, TMI was recommended to replace BMI z-score in children and adolescents due to its lower false-positive rate of obesity than the BMI z-score (boys: 2.9% vs. 21.8%; girls: 17.5% vs. 28.5%) (18,19,22). TMI presented a higher area under the curve (AUC) value than BMI for predicting high BF% (0.96 vs. 0.93, p < 0.001) measured by dual-energy X-ray absorptiometry (DEXA) among children and adolescents aged 8-17 years (23) or more strongly correlated with BF% compared to BMI in adolescents (41,42).
The remaining four studies could not conclude about a possible advantage of either TMI or BMI to identify overweight or obesity status because of the lack of a gold standard (to objectively assess adiposity) and inconsistent cutoffs (14,20,21,26). Akcan et al. found that TMI identified a lower prevalence of overweight and obesity among children aged 6-17 years compared to BMI-SDS (14), which was contrary to the finding among children aged 9-13 years, independent of sex (20). When considering overweight and       (21), inversely to the findings among children and adolescents aged 10-20 years reported by Park et al. (26).
Overall, studies using a gold standard for comparison and using BMI and TMI as continuous variables suggested that TMI performed equally or better than the widely used BMI to predict BF% and central fat among children and adolescents. TMI was preferred in adolescence due to its stability.  (2) The misclassification of overweight was less than BMI z-score (8.4% vs. 19.4%, p < 0.001) but equal to updated BMI percentiles based on the same data set (8.4% vs. 8.0%, p = 0.62). However, TMI was preferred due to its simplicity with no complicated percentiles.
(3) The results were similar when stratified by sex.
(2) TMI and WHtR were more applicable for public health use than BMI, WC, and BF% due to their simplicity in calculating and identifying obesity. The AUCs of these indicators remained similar when stratified by sex.
Age, sex, treatment, and puberty 6 Akcan, 2019 TMI revealed less overweight and obesity than BMI.
The prevalence of overweight defined by TMI was lower than that defined by BMI (15% vs. 18%), but the prevalence of obesity defined by TMI was higher than that defined by BMI (9.7% vs. 8.9%) None 4 Zaniqueli, 2019 Although TMI (R 2 = 0.73 for boys and R 2 = 0.75 for girls) and BMI (R 2 = 0.74 for boys and R 2 = 0.75 for girls) performed similar in the portion of the variability for BF%, TMI was recommended to replace the BMI z-score in children and adolescents due to a lower falsepositive rate of obesity Park, 2020 The prevalence of overweight defined by TMI was slightly higher than that defined by BMI (10.6% vs. 10.2%), but the prevalence of obesity defined by TMI was lower than that defined by BMI (5.3% vs. 10.6%), similar in both sex.

MetS
Children aged <10 years: only BMI z-score was associated with MetS (b = 2.21, p < 0.05) Children aged ≥10 years: BMI z-score (b = 2.67, p < 0.001), TMI (b = 0.19, p < 0.01), conicity index (b = 9.02, p < 0.001) and WHR (b = 12.32, p < 0.001) were associated with MetS. Similar results were found among males, whereas only conicity index (b = 7.37, p < 0.05) and WHR (b = 7.94, p < 0.05) were associated with MetS among females. High BP: BMI z-score was the best predictor of high BP in both children and adolescents, whereas TMI performed better among males. High TG: conicity index was the best predictor for high TG in females and WHR was best in males.
Low HDL-C: BMI z-score was the best indicator for low HDL-C.
Age and sex 8 Calcaterra, 2021 (45) Among children and adolescents with obesity, TMI was associated with IR indicators only in females while BMI correlated with all IR indicators except for triglyceride and glucose index in females and BMI z score correlated with all IR indicators except for HOMA-b in males.
(1) Youth TMI, BMI, and subscapular skinfold thickness were significantly associated with adult T2D, obesity, high carotid intima-media thickness, and high LDL-C level.
(2) Between the ages of 14 and 16, BMI outperformed TMI for elevated insulin levels (difference in AUC = 0.018 and 0.025) and IR (difference in AUC = 0.018-0.024). At age 16-20 years, BMI outperformed TMI for hypertension (difference in AUC = 0.017-0.022). For other outcomes of impaired FPG, high aortic intima-media thickness, high LDL-C, low HDL-C, and high TG, the predictive utilities were similar.

TMI and Obesity-Related Cardiovascular Risk Factors in Children and Adolescents
Twenty articles on the association between TMI and MetS and its components were included in this systematic review ( Table 2) (13-18, 21, 26-34, 42-45).

MetS
Ten articles have evaluated the association of TMI and other anthropometric indicators with MetS, metabolic risk score, or cardio-metabolic risk (13, 15-18, 27-30, 33). Three of the 10 articles showed that TMI was not better than other indicators such as BMI (or BMI z-score or BMI-SDS) among children and adolescents aged 10-17 years to predict MetS and a metabolic risk score (15)(16)(17). However, the other seven articles suggested that TMI could be a useful screening tool or similar to BMI in predicting MetS or cardio-metabolic risks in children and adolescents aged 5.3 to 25 years (13,18,(27)(28)(29)(30)33). TMI was found to be associated with obesity-related CVRFs, including MetS and its components [elevated blood pressure (BP), elevated WC, low high-density lipoprotein cholesterol (HDL-C), and elevated triglycerides (TG)] in late adolescence (28). It was reported that TMI performed similarly to FMI (27) or BMI, or was an auxiliary indicator in addition to BMI, to identify MetS, a metabolic risk score, or CVRFs among children and adolescents aged 5-18 years (13,18,29,33). However, the tri-ponderal body shape index [WC z-score/(TMI 2/3 *height 1/2 )] including TMI and WC z-score components performed more accurately in predicting MetS than BMI and TMI (Youden index: 0.85 vs. 0.73 vs. 0.70) among children and adolescents aged 7-18 years, suggesting that the combination of TMI and a WC z-score could be considered as a useful predictor for MetS in children and adolescents (30).
Overall, TMI performed similarly as compared to BMI and other indicators in predicting MetS in most of the included studies, and TMI was also suggested to be a useful tool when used in combination with other adiposity indicators (e.g., BMI and WC) for identifying MetS.

Insulin Resistance
Eight articles compared TMI and BMI for identifying insulin resistance (IR) or impaired glucose in children and adolescents ( Table 2) (13,14,21,29,31,32,34,45). Among these eight articles, seven reported that BMI (used as a continuous variable) performed similarly or marginally better than TMI for identifying IR (13,21,29,31,32,34,45). In addition, compared to BMI, TMI was more likely to underestimate IR (14). The inconsistent cutoffs of TMI and BMI for identifying overweight might lead to different identification of IR. When restricted to children and adolescents aged 4-18 years with overweight or obesity, WHtR or WC, used as continuous variables, seemed to perform best among the four obesityrelated indicators (TMI, WC z-score, BMI z-score, and BMI) to predict IR (31,32).
Overall, TMI did not seem to be superior to BMI for predicting IR in children and adolescents. However, WHtR or WC could be a useful indicator for identifying IR among children and adolescents with overweight and obesity.

Blood Pressure
Only five studies compared the correlation of TMI and BMI with BP, with inconsistent results ( Table 2) (13,17,21,33,43). Although BMI correlated with BP levels stronger than TMI (17,33), one study based on 5,814 children and adolescents aged 6-19 years showed that, similar to BMI using a continuous variable, TMI (used as a continuous variable) had a low ability to identify high BP, with an AUC of only 0.66 to predict systolic BP and 0.60 to predict diastolic BP (21); similar findings were found among 57,201 Chinese children and adolescents aged 7-18 years, among 10,441 American adolescents aged 12-18 years (13) and among Italian adolescents (43).
Overall, only a few studies examined the question and they tended to suggest that either TMI or BMI performed poorly in identifying high BP in children and adolescents, and the ability varied in different populations.

Dyslipidemia
As shown in Table 2, three articles showed that both TMI and BMI poorly predicted dyslipidemia (13,14,21). Although using the same BMI classification, total cholesterol (TC) in boys and HDL-C and TG in girls were worse among children with overweight defined by TMI than among those with normal TMI (26), BMI (as a continuous variable) seemed to be more strongly associated with TC (R 2 : 0.32 vs. 0.27), HDL (-0.52 vs. -0.46), and TG (0.32 vs. 0.27) compared to TMI, while TMI (as a continuous variable) was more strongly associated with lowdensity lipoprotein-cholesterol (LDL-C) than BMI (0.38 vs. 0.33) (29), similar to findings on low HDL-C reported by Leone et al. (17), but inversely to findings by Alfaraidi et al. (42).
Overall, there are limited studies on the association of TMI and BMI with dyslipidemia components, and findings suggest that BMI performs better than TMI to identify high TC and TG, whereas TMI is superior to BMI to identify high LDL-C. This will need further evaluation.

Inflammatory and Liver Function Markers
As shown in Table 2, for C-reactive protein (CRP), the prediction accuracy of TMI and BMI z-score was similar (AUC: 0.74 vs. 0.73) (21), whereas other inflammatory markers including leptin, IL-6, and fetuin-A were more closely correlated with BMI than TMI (29). For liver enzymes, overweight and obese status based on TMI could significantly predict elevated serum glutamic oxaloacetic transaminase or elevated serum glutamic pyruvic transaminase, compared with overweight and obesity status based on BMI. However, different cutoffs were defined for BMI vs. TMI, which limits direct comparison (14). For non-alcoholic fatty liver, the discriminating ability of TMI was similarly poor as BMI, with AUC values of only 0.58 (44).
Overall, there is only limited evidence about the performance of TMI and BMI to predict inflammatory markers, which needs further research.

TMI in Childhood or Adolescence for Prediction of Specific CVRFs in Adulthood
Only four articles focused on the association of TMI vs. other obesity-related indicators in childhood or adolescence with CVRFs in adulthood (35)(36)(37)46) (Table 2). BMI at ages 2 to 20 years predicted the presence of CVRFs in young adults aged 20 years as well or better than TMI. For example, the ability to predict adult IR, elevated insulin levels, and hypertension seemed to be stronger for BMI vs. TMI (as assessed in childhood), but similar for the prediction in adults of impaired fasting plasma glucose (FPG), low HDL-C, high LDL-C, high TG, and high aortic intima-media thickness (37). Similarly, another study showed that the AUC values for TMI, or for combination of TMI and BMI, did not outperform BMI alone in predicting adult obesity, diabetes, high carotid intimamedia thickness, high LDL-C, and hypertension (35). The AUCs were low for TMI (0.53, 0.56, and 0.54), as well as for other adiposity indicators such as abdominal volume index (0.61, 0.62, and 0.61), BMI (0.59, 0.60, and 0.59), and WC (0.61, 0.61, and 0.61) in childhood to predict adult homeostasis model assessment 2insulin resistance (HOMA2-IR), HOMA2-b, and high fasting insulin (36). However, when considering growth trajectory instead of a single measurement in childhood, a persistently high TMI during adolescence had predicted diabetes quite well in adults (AUC value as high as 0.81) (46).
Overall, TMI in childhood or adolescence seems to have a lower ability than BMI and other adiposity indicators to predict specific CVRFs in adulthood, whereas TMI trajectory has a higher ability than BMI trajectory in predicting diabetes in adulthood.

Main findings
To the best of our knowledge, this is the first review to summarize the evidence regarding TMI as a screening tool for body fat and CVRFs in childhood and adulthood. TMI seemed to perform similarly or better than BMI for predicting body fat and central fat and performed similarly well as BMI in identifying MetS. However, the available evidence on the comparison of TMI and BMI (measured in childhood) for identifying specific CVRFs (in childhood or later in adulthood) including IR, high BP, dyslipidemia, and inflammation was limited and not compelling.

TMI Performed Better Than BMI to Estimate Body Fat in Children and Adolescents
Unlike for adults, no standard BF% cutoff was established to define excess adiposity among children and adolescents until now (47), and objective measurements of fat mass [e.g., DEXA, doubly-labeled water (48), and isotope dilution technique (49)] were much complex and expensive. The components of TMI or BMI (weight and height) can be simply measured using the weight scale and the stadiometer that are widely used for routine pediatric clinical practice. Therefore, in this review, we compared the performance of TMI and BMI and our study suggested that TMI performed better than BMI to estimate body fat in children and adolescents at clinical practice.
The disadvantage of BMI and the advantage of TMI to estimate body fat are as follows. First, although BMI z-score seemed to predict well total fat mass, it predicted BF% weakly with altered body composition among adolescents (50). Second, the definition of overweight and obesity using BMI should be based on sex-and age-specific percentile values in childhood, but this requires using complex tables (10), which may overestimate the actual prevalence of adiposity in children, excessively worrying families and patients (50)(51)(52)(53), particularly for adolescents who may be more prone to fat-shaming and weight bias (54). Third, TMI (which is defined independently of age and sex) could be simpler to use compared to age-and sexstratified BMI cutoffs and a specific cutoff of TMI has been proposed (10). A better relation of TMI with body fat mass across age may be consistent with the fact that BF% may change largely during adolescence (possibly more among girls) due to the height spurt in this age range (55). Fourth, compared to BMI, TMI was more correlated with WHtR, which is a reliable clinical measure of abdominal obesity and is consistently associated with CVRFs (56). TMI could therefore help identify children and adolescents who are overweight or obese based on BMI but also have central obesity and increased risk of CVRFs.

TMI Was More Simple and Accurate Than Other Indicators to Estimate Body Fat in Children and Adolescents
Although the RFMp calculated based on WC and height performed better to estimate BF% than TMI (25), the interoperator variability between WC measurements is significant, which may cause more misclassification of MetS (57). Furthermore, for tall and thin people, WHtR may be unusually high, causing RFMp and RFM to tend to be 0 or negative (25). Therefore, considering the accuracy and simplicity of the use of TMI in primary health care services and its constancy in predicting adiposity at adolescence, TMI may be useful to evaluate body fat in adolescents. Yet, definite answers about the performance of BMI vs. TMI to predict adiposity in chidden and adolescents needs further studies using objective measurement of body fat mass (e.g., DEXA, isotope dilution) as the gold standard for comparisons, and do so in several populations, and within different ethnic, age, and sex groups.

TMI Performed Similarly as Compared to BMI and Other Indicators in Predicting MetS and Its Components
Although TMI was superior to BMI to screen central fat (19,24), in this review, it was similar to or not better than BMI to identify MetS and specific CVRFs. One possible reason might be that adiposity defined according to TMI or BMI only accounts for one of the MetS criteria. Another reason might be the inconsistent performances of three indicators (TMI, BMI, and WC) in identifying specific CVRFs including IR, high BP, dyslipidemia, and inflammation (13,14,21,29,31,32), which are the main components of MetS.
Age and Trajectory Influence the Association Between TMI vs. BMI in Childhood or Adolescence and Obesity-Related Morbidity in Adulthood Although BMI in childhood or adolescence seemed to perform marginally better than TMI to predict obesity-related morbidity in adulthood (35,36), the difference disappeared after adjusting for age (35), suggesting that age might be an important confounding factor that influences the association between BMI in childhood and obesity-related morbidity in adulthood. BMI was better than TMI only in late adolescence to predict adult IR and hypertension, suggesting that the variation of BMI during adolescence influences the strength of the association (10,37). When considering trajectories, persistently high TMI during 13 and 18 years performed better than the BMI trajectory to predict adult diabetes (46), suggesting that, during adolescence, TMI trajectory (i.e., repeated measurements) may better reflect growth and predict adult CVRD outcomes. Therefore, further prospective studies with large sample sizes, multiethnic populations, and repeated measurements of anthropometric indicators are needed to confirm these findings.

Strengths and Limitations
To the best of our knowledge, this is the first comprehensive review that compared TMI with BMI or other indicators in children and adolescents to predict obesity-related morbidity in both childhood and adulthood. Several limitations should be noted in this review. First, there was high heterogeneity between studies in the considered variables and how the adiposity cutoffs were defined, which limits direct comparisons. Second, most studies on the identification of CVRFs in childhood and adolescence were cross-sectional, which cannot prove causality (55). It must be however mentioned that a marker does not necessarily need to be causally related to an outcome to enable a good prediction. Third, a majority of the included studies came from Western countries, which limits the extrapolation of the results to other populations. Further studies with various ethnic/ race groups are needed to confirm the predictive ability of TMI to predict adiposity in children and adolescents. Fourth, although TMI seems better than BMI to predict concomitant fat mass in children and adolescents, neither TMI nor BMI can distinguish fat mass from non-fat mass, and these indicators cannot replace objective measurement of fat mass (e.g., DEXA, isotope dilution). Again, an ultimate fully valid method to compare how BMI or TMI predicts adiposity should rely on objectively measured adiposity as the gold standard (e.g., DEXA, isotope dilution methods) and use a similar dichotomization of categories of elevated BMI or elevated TMI (e.g., using the same percentile cutoffs, e.g., p80 or p90) to enable valid comparisons; this was only rarely performed in the considered studies.

CONCLUSIONS
In conclusion, TMI only requires a single threshold according to sex (i.e., no need for sex-and age-specific thresholds) and TMI seems to predict adiposity similarly or better in children and adolescents than BMI. In addition, TMI seemed to perform similarly as BMI for identifying MetS. However, the clinical use of TMI vs. BMI in childhood, in order to predict specific elevated CVRFs in childhood or later in adulthood, is still not definitive and needs further studies, particularly those with a longitudinal design.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material. Further inquiries can be directed to the corresponding author.