Obesity Surgery

https://doi.org/10.1007/s11695-019-04289-2

ORIGINAL CONTRIBUTIONS

Does FMI Correlate Better than BMI with the Occurrence of Metabolic
Changes in Obese Patients? Study Based on 2007 Consecutive
Mexican Patients
Carlos A. Gutiérrez-Rojas 1 & Ruth Cruz-Soto 1 & Verónica Sánchez-Muñoz 1 & Anayeli Romero 1 & Maureen Mosti-Molina 1 &
Hugo A. Sánchez-Aguilar 1 & David Velázquez-Fernández 2 & Miguel F. Herrera 1,2

# Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
Background The body mass index (BMI) is the most commonly used anthropometric indicator. However, it does not discern
among the different body components. The body fat content, expressed as fat mass index (FMI), is an accurate way to estimate
adiposity. Since most metabolic diseases are associated with excess fat tissue, our aims were to comparatively analyze the
frequency of associated metabolic abnormalities in patients with different obesity degrees based on BMI and FMI and to
determine the best cut-off value of both indicators to predict metabolic abnormalities.
Methods From a cohort of 2007 patients, BMI and FMI were calculated using DXA. Individuals were classified into the different
obesity degrees according to the reference ranges from the World Health Organization (WHO) and the National Health and
Nutrition Examination Survey (NHANES). A comparative analysis between BMI, FMI, and their correlation to the presence of
metabolic alterations was performed.
Results BMI underestimated the degree of obesity when compared with FMI. Spearman’s rank-order correlation for both indexes
resulted in very high coefficients (rho Spearman’s = 0.857; p = 0.0001). The prevalence of metabolic alterations increased as BMI
and FMI also increased. Despite the high positive statistical correlation between BMI and FMI, it was seen that some comor-
bidities were more specifically related to one particular index.
Conclusions There were no significant differences between the BMI and the FMI for predicting the degree of obesity. Likewise,
there were no significant differences between them for the prediction of metabolic alterations.

Keywords Obesity . Fatty mass index . Body mass index . DXA . Body composition

Background increased in recent years, becoming a public health problem

worldwide [1]. As greater is the BMI, the frequency of comor-
Overweight and obesity are defined as an excessive accumu- bidities such as type 2 diabetes (T2D), arterial hypertension,
lation of fat in the body [1]. According to the World Health hyperlipidemia, obstructive sleep apnea syndrome (OSA), and
Organization (WHO), individuals with body mass index cardiovascular diseases also increase [1, 2].
(BMI) equal to or greater than 30 kg/m2 are considered obese, Currently, BMI is the most commonly used anthropometric
whereas a BMI between 25 and 29.9 kg/m2 corresponds to indicator for estimating obesity. However, it does not distinguish
overweight [2]. The prevalence of overweight and obesity has between the components responsible for the excess weight, i. e.,
fat mass or muscular mass, and it does not take into consideration
gender and ancestral origin [3, 4]. Considering that the main risk
* Miguel F. Herrera factor for metabolic diseases is the excess of adipose tissue and in
miguelfherrera@gmail.com particular its central distribution as it has been demonstrated by
1 some authors [1, 4, 5], direct measurement of fat tissue content
Center for Nutrition and Obesity, The American British Cowdray
Medical Center, Mexico City, Mexico might represent a higher clinical value.
2 Dual-energy X-ray absorptiometry (DXA) is a whole-body
Department of Surgery, Instituto Nacional de Ciencias Médicas y
Nutrición Salvador Zubirán, Mexico City, Mexico imaging modality that allows analyzing the total body composi-
tion. It is based on the absorption of X-rays by the different

components of the organism, using high- and low-energy X-ray
photons [3]. DXA has high precision, with a margin of error in
the estimation of body composition from 2 to 6% [6]. In addition,
it has the advantage of providing measurements of total and
regional body composition [3].
Since the major culprit in obesity seems to be the excess of
adipose tissue, the fat mass index (FMI), which is obtained by
dividing the fat weight in kilograms by height in meters squared,
could be a better indicator of obesity than BMI [7]. .Multiple
studies report that the distribution of body fat is different depend-
ing on several factors, including ancestral origin, age, gender,
genetic, nutritional, socioeconomic, and endocrine factors. In
addition, body conformation may vary among different geo-
graphical regions within the same country [4].
The hypothesis of our work is that the FMI might be a more
accurate criterion compared with the BMI for identifying meta-
bolic problems associated with obesity since the former is more
directly related to the adipose mass. Therefore, the aims of the
study were (1) to comparatively analyze the frequency of meta-
bolic abnormalities in patients with normal weight, overweight,
and different degrees of obesity based on BMI or FMI; and (2) to
determine the best cut-off value for both BMI and FMI to iden-
tify a higher risk for metabolic abnormalities among these
patients.
Patients and Methods
Using the prospectively fed database from our Center, a total of
2007 consecutive patients who underwent medical evaluation
between March 2015 and September 2018 were included in the
study.
All patients had a complete clinical history with physical ex-
amination and the STOP-BANG Score questionnaire for OSA.
Laboratory studies included blood count, blood chemistry, lipid
profile, and thyroid function tests. Body composition was mea-
sured by a DXA scan using the Hologic Discovery Wi (S/N
87039) software, version 13.4.1. NHANES BCA calibration
was performed to measure muscle, water, and fat contents.
BMI and FMI were calculated based on the DXA results and
individuals were classified into normal weight, overweight, or
different degrees of obesity according to the reference ranges
from the WHO and the National Health and Nutrition
Examination Survey (NHANES) from the CDC respectively.
According to BMI, patients were stratified in normal weight (
≤ 18.5–24.9 kg/m2), overweight (25–29.9 kg/m2), grade I obesity
(30–34.9 kg/m2), grade 2 (35–39.9 kg/m2), and grade 3 (≥ 40 kg/
m2). According to FMI, they were classified as follows: normal
fat mass (males < 2–6 kg/m2 and females < 3.5–9 kg/m2), excess
fat mass (males 6–9 kg/m2 and females 9.1–13 kg/m2), obesity
class 1 (males 9.1–12 kg/m2 and females 13.1–17 kg/m2), obe-
sity class 2 (males 12.1–15 kg/m2 and females 17.1–21 kg/m2)
and obesity class 3 (males > 15 kg/m2 and females > 21 kg/m2).
OBES SURG
The following metabolic alterations were investigated and
registered in each included individual: pre-type 2 diabetes and
T2D according to the criteria established by the American
Diabetes Association (ADA) [8]; arterial hypertension, according
to the American Heart Association (AHA) criteria [9]; hypertri-
glyceridemia and hypercholesterolemia using the American
Association of Clinical Endocrinologists and American College
of Endocrinology (AACE/ACE) guidelines [10], and OSA based
on the STOP-BANG Score and sleep study in patients with a
score equal to or greater than 5.
Microsoft® Excel® version 14.7.7 and IBM® SPSS®
Statistics version 21.0 were employed for the statistical analysis.
Dimensional data were expressed as mean ± SD whenever a
normal distribution was determined; otherwise, non-parametric
measures were employed such as median ± range. Univariate
analysis was used for individual variables depending upon the
intrinsic variable scaling, whereas a bivariate analysis was per-
formed to assess potential statistical associations or correlations.
Chi-square and Fisher’s exact tests were used for analyzing cat-
egorical variables. Parametric and non-parametric correlations
were performed through the use of Pearson’s, Kendall’s tau-c
for concordance paired values, and Goodman and Kruskal’s
gamma and Spearman’s correlation test for non-parametric and
rank-order correlations, respectively. Any p value equal to or less
than 0.001 (type I error) was considered statistically significant.
This protocol fulfilled the Helsinki Declaration and was previ-
ously approved by the Ethics Committee of the American British
Cowdray Medical Center.
Results
A total of 860 (43%) patients were male (M) and 1147 (57%)
female (F), with a mean ± SD age of 43.3 ± 13.6 years. Tables 1
and 2 describe the clinical and biochemical features of the studied
cohort. Figure 1 displays the frequency of the different stratifica-
tion groups according to BMI and FMI in the analyzed cohort.
As it can be seen in Table 3, the BMI tends to underestimate
the degree of obesity in comparison with the FMI.
However, there was a high rank-order correlation between
these 2 ordinal scales with a resulting gamma value of
0.94 (Goodman and Kruskal’s gamma test; p < 0.0001),
Table 1 Clinical characteristics of the studied population
Variable Mean ± SD Range
Weight, kg 90.7 ± 22.6 40.4–215
Height, m 1.67 ± 0.09 1.4–2
BMI, kg/m2 32.2 ± 6.54 16.7–66
Fat mass, kg 28.2 ± 12.9 10.2–199.6
FMI, kg/m2 13.6 ± 4.3 3.2–37.8
OBES SURG

Table 2 Biochemical results in our cohort increased. Despite the high statistical correlation between these
Variable Mean ± SD Range two indexes, it was seen that some comorbidities related more to
one specific index. For example, BMI showed a significant cor-
Serum glucose, mg/dL 97.1 ± 25.2 59–455 relation with OSA based on the Spearman’s rank correlation
Total cholesterol, mg/dL 202.4 ± 41.6 67–455 coefficient (rho Spearman’s = 0.377; p = 0.0001), with arterial
HDL, mg/dL 48.0 ± 12.42 18–103 hypertension (rho Spearman’s = 0.323; p = 0.0001), with pre-
LDL, mg/dL 124.6 ± 36.9 28–354 type 2 diabetes (rho Spearman’s = 0.212; p = 0.0001),with
Triglycerides, mg/dL 153.1 ± 92.1.3 20.1–983 hypertriglyceridemia (rho Spearman’s = 0.205; p = 0.0001),
HbA1c, % 5.5 ± 0.8 2.2–14 and with T2D (rho Spearman’s = 0.174; p = 0.0001), whereas
FMI was significantly associated with OSA (rho Spearman’s =
0.384; p = 0.0001), arterial hypertension (rho Spearman’s =
and a Spearman’s correlation coefficient of 0.86 (p = 0.332; p = 0.0001), pre-type 2 diabetes (rho Spearman’s =
0.0001) as displayed in Fig. 2. FMI was tested for diagnostic 0.207; p = 0.0001), hypertriglyceridemia (rho Spearman’s =
efficiency in classifying obesity versus normal/overweight indi- 0.196; p = 0.0001), and T2D (rho Spearman’s = 0.160; p =
viduals, using BMI as the gold standard. Sensitivity of FMI for 0.0001).
diagnosing any degree of obesity was 96.7% (95%CI 95.5– Knowing that gender might affect the results, we performed a
97.6), specificity was 73.3% (95%CI 69.1–75.4), PPV was comparison between gender and the different obesity degrees
83.4% (95%CI 81.8–84.9), NPV was 93.8% (95%CI 91.8– obtained by BMI and FMI which did not result in any statistically
95.4) and accuracy was 86.7% (95%CI 85.2–88.2). The overall significant differences (p > 0.39).
frequency of metabolic alterations in our studied patients is In order to explore for the best cut-off value predicting meta-
shown in Fig. 3. Mean serum glucose in patients with T2D bolic abnormalities, a receiver operating characteristic (ROC)
was 144.1 ± 26.2 mg/dL and in patients with pre T2D 96.7 ± analysis of the BMI and FMI was independently performed.
25.6 mg/dL. Mean systolic and diastolic values in patients with Table 4 shows the values with the highest sensitivity and speci-
arterial hypertension were 123.9 ± 13.6 and 81.9 ± 9.8 mmHg, ficity from this analysis. As an example, the comparative ROC
respectively. As it is shown in Fig. 4, the percentage of metabolic analysis of BMI and FMI in patients with T2D is depicted in Fig.
alterations almost linearly increased as the BMI and FMI also 5. However, the area under the curve (AUC) for BMI was

Fig. 1 Stratification of all included patients according to their BMI and FMI
 OBES SURG

Table 3 Patient distribution

according to BMI and FMI BMI (n) FMI

Normal Excess fat Obesity 1 Obesity 2 Obesity 3 Total

Normal (187) 102 81 2 2 0 0

Overweight (634) 18 392 215 6 2
Obesity 1(616) 0 36 401 163 13 3
Obesity 2 (342) 0 3 51 209 79 0
Obesity 3 (228) 0 0 3 31 193
Total (2007) 120 512 672 411 287 5

slightly greater than FMI (0.694 versus 0.643). Based on this
analysis, we may infer that both indexes are strikingly efficient
for predicting metabolic abnormalities among our patients, with
an increasing probability as their grades escalate too without any
decisive differences as we previously thought, but with gender.
Discussion
BMI is the most commonly used measurement to estimate the
ideal weight and to classify obesity despite the fact that the asso-
ciation between obesity and fat mass is not perfect mathemati-
cally [1, 3, 11–13]. In addition, the classification of obesity based
on merely BMI does not take into consideration the variability
related to gender, ethnicity, and ancestry, which may require
adjustments of the standard cut-off values [12–14]. However,
several authors have documented a close relationship between
BMI and the presence or development of metabolic disorders as
well as risk factors for both cardiovascular diseases and mortality
[15]. It has been demonstrated that median survival is reduced by
2–4 years in obese patients when it is compared with individuals
with optimal BMI [16].
The relationship between the body fat content and the devel-
opment of comorbidities is clear [4], especially with respect to
central obesity, which has been associated with metabolic syn-
drome and the increased risk of mortality due to cardiovascular

Fig. 2 Non-parametric correlation between the BMI and FMI in our analyzed patients
OBES SURG
Fig. 3 Overall frequency of metabolic alterations in our analyzed cohort
diseases [1, 4, 5]. Considering FMI as a theoretically better mark-
er of adiposity, it emerges as a superior option to classify the
degree of obesity and potentially a more preferable predictor of
an elevated risk for metabolic and cardiovascular comorbidities
[7].
In this study, we performed a comparative analysis in a
cohort-based approach of consecutive patients between the
BMI versus the FMI for predicting the occurrence of metabolic
alterations among these individuals. Our primary hypothesis was
to investigate which somatometric index (BMI or FMI) displays
a higher correlation and prognostic efficiency for determining the
occurrence of the most common obesity-associated comorbidi-
ties such as OSA, arterial hypertension, T2D, and dyslipidemia.
Based on our results, we found that the number of comorbidities
significantly increases as the BMI and FMI also increase
(Kendall’s tau-b; p ≤ 0.0001). Although BMI seemed to under-
estimate obesity when compared with FMI, none of them dem-
onstrated superiority as a classificatory tool. It was interesting to
observe in our study that hypercholesterolemia did not follow the
same pattern of increase when both BMI and FMI were also
increased.
A discrepancy between BMI and the fat mass content has
been documented in different published series [12, 17–19] and
the phenomenon that the BMI tends to underestimate the degree
of obesity according to the FMI has been also observed and
described by different authors in different populations [18–20].
FMI is able to identify individuals with an elevated BMI but
without excess fat mass. It can also identify subjects with normal
BMI but with an elevated fat mass [18, 19, 21]. Despite the
evidence of this discrepancy, in a similar way to our study, the
significant correlation between BMI and body fat mass has been
confirmed by other authors [1, 20]. It is important to highlight
that the relationship between BMI and FMI is closely related to
the idiosyncratic characteristics of the studied population. At
least, there are two potential factors that may have a great impact
on the correlation between BMI and FMI: age and gender [1, 20].
The correlation has shown to be higher in patients with obesity
than in patients with normal weight or overweight [20]. Studies
in high-performance athletes have documented that the BMI
tends to overestimate the degree of overweight and obesity up
to 6 times [22], and as it would be expected, the number of
metabolic alterations in those subjects is very low and does not
correlate with the BMI. Gender is a well-known factor associated
not only with adiposity but with a specific phenotype distribu-
tion. Considering that gender might affect our results, we per-
formed a comparison between gender and the different obesity
grades determined by BMI and FMI. This analysis did not result
in any statistically significant value (p > 0.39).
The relationship between obesity and metabolic alterations
is not linear but it is mathematically interdependent. It has
been recognized that more than the weight or the increased
fat tissue by themselves, the type of fat and the distribution of
the excessive fat tissue may be important. One recognized
phenotype is the metabolically healthy but obese individual,
sometimes referred to as “uncomplicated” obesity; these indi-
viduals appear to be relatively resistant to the development of
the adiposity-associated cardiometabolic abnormalities that
increase the cardiovascular risk. The second phenotype
 OBES SURG

Fig. 4 Frequency of metabolic alterations according to different BMI and FMI degrees

includes individuals with normal or slightly increased weight,
who express cardiometabolic abnormalities [23, 24]. Several
publications with up to a 30-year follow-up have demonstrat-
ed that even patients with apparently uncomplicated obesity
present a higher risk for T2D, cardiovascular events, and
increased mortality [15, 25] when compared with individuals
in healthy weights [15]. In the ROC analysis, we found that
the cut-off values with the higher sensitivity and specificity to
predict the occurrence of different metabolic alterations were
in the range of obesity grade 1 or 2 (Table 4). This may lead us
OBES SURG

Table 4 Values with the highest

sensitivity and specificity for Metabolic alteration BMI Obesity FMI Obesity
metabolic diseases from the ROC Sensitivity/specificity, degree Sensitivity/specificity, degree
analysis % BMI % FMI

OSA 33.64 1 12.65 M: 2

(74/74) (75/52) W: excess fat
Arterial hypertension 31.57 1 12.15 M: 2
(70/61) (72/48) W: excess fat
Pre-type 2 diabetes 30.05 1 12.5 M: 2
(70/50) (63/52) W: excess fat
T2D 31.55 1 13.5 M: 2
(75/55) (61/57) W: 1
Hypertriglyceridemia 31.35 1 12.25 M: 2
(59/59) (62/47) W: excess fat
Hypercholesterolemia 30.37 1 12.95 M: 2
(57/46) (50/50) W: excess fat

to initiate our treatment efforts on patients with early stages of the fact that Mexican mestizos are a very genetically admixed
obesity diagnosed in terms of BMI or FMI by DXA. and heterogeneous population [26]. Since most of our patients
Our study has several limitations: (a) a relatively limited are from a private hospital in one of the largest cities, Mexican
sample size, (b) the retrospective nature of the design, and (c) mestizo representativeness cannot be claimed and (d) our

Fig. 5 ROC analysis for T2D in both BMI and FMI


study population is biased (selection bias) since it included
individuals who seek medical attention at a Nutrition and
Obesity Center by themselves and consequently the number
of patients in ideal weight was very low (9.3%).
Conclusions
Based on our analysis we have been able to validate the clin-
ical utility of either BMI or FMI for predicting the occurrence
of the most frequently associated comorbidities in obese pa-
tients. There is a clear positive correlation with a strong linear
tendency between these clinimetric indexes and their related
pathologies which resulted statistically significant for all con-
ditions. There were no significant differences between both
instruments for the prediction of metabolic alterations, al-
though BMI exhibited slightly greater AUCs than FMI for
some specific associated morbidities. Both somatometric mea-
surements could be indistinctly employed for categorizing
obesity and predicting the occurrence of the most commonly
associated metabolic disturbances among obese patients.
Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no conflict of
interest.
Ethical Approval The study does not include animal or human experi-
ments and all personal information was codified to maintain
confidentiality.
Consent Statements Non-applicable.
References
1. Ranasinghe C, Gamage P, Katulanda P, et al. Relationship between
body mass index and body fat percentage, estimated by bioelectri-
cal impedance, in a group of Sri Lankan adults: a cross sectional
study. BMC Public Health. 2013;13:1–8.
2. Hernández CP, Flores CA, Rocha MAD, et al. Composición corpo-
ral en mujeres derechohabientes del IMSS que acuden al Servicio
de Densitometría. El Residente. 2012;7:6–15.
3. Lorente RRM, Azpeitia AJ, Arévalo GN, et al. Absorciometría con
rayos X de doble energía. Fundamentos, metodología y
aplicaciones clínicas, vol. 54: Elsevier; 2012. p. 410–23.
4. Charon GR, Berger M, Garg RK, et al. Measures of adiposity and
cardiovascular disease risk factors, New York City Health and
Nutrition Examination Survey, 2004. Prev Chronic Dis. 2011;8:1–
10.
5. Schuster J, Vogel P, Eckhardt C, et al. Applicability of the visceral
adiposity index (VAI) in predicting components of metabolic syn-
drome in young adults. Nutr Hosp. 2014;30:806–12.
6. Albanese CV, Diessel E, Genant HK. Clinical applications of body com-
position measurements using DXA. J Clin Densitom. 2003;6:75–85.
7. Kelly TL, Wilson KE, Heymsfield SB. Dual energy X-Ray absorp-
tiometry body composition reference values from NHANES. PLoS
ONE. 2009;4:1–8.
OBES SURG
8. American Diabetes Association (ADA). Standard Medical Care in
Diabetes 2018. J Clin Appl Res Educ. 2018;41:1–150.
9. Cifu AS, Davis AM. Prevention, detection, evaluation, and manage-
ment of high blood pressure in adults. JAMA. 2017;318:2132–4.
10. Jellinger PS, Handelsman Y, Rosenblit PD, et al. (2017). AACE
AND ACE guidelines for management of dyslipidemia and preven-
tion of cardiovascular disease. Endocr Pract. 2017;23:479–97.
11. Roubenoff R, Dallal GE, Wilson PWF. Predicting body fatness: the
body mass index vs estimation by bioelectrical impedance. Am J
Public Health. 1995;85:726–8.
12. Fatima SS, Rehman R, Chaudhry B. Body mass index or body fat!
Which is a better obesity scale for Pakistani population? J Pak Med
Assoc. 2014;64:1225–8.
13. Fernández-Real JM, Vayreda M, Casamitjana R, et al. Body mass
index (MBI) and percent fat mass. A BMI > 27.5 Kg/m2 could be
indicative of obesity in the Spanish population. Med Clin.
2001;117:681–4.
14. Razak F, Anand SS, Shannon H, et al. Defining obesity cut points in
a multiethnic population. Circulation. 2007;115:2111–8.
15. Ärnlöv J, Sundström J, Ingelsson E, et al. Impact of BMI and the
metabolic syndrome on the risk of diabetes in middle-aged men.
Diabetes Care. 2011;34:61–5.
16. MacMahon S, Baigent C, Duffy S, et al. Body mass index and
cause-specific mortality in 900 000 adults: collaborative analyses
of 57 prospective studies. Lancet. 2009;373:1083–96.
17. Deurenberg P, Yap M, Van Staveren WA. Body mass index and
percent body fat: a meta analysis among different ethnic groups.
Int J Obes. 1998;22:1164–71.
18. Schutz Y, Kyle UUG, Pichard C. Fat-free mass index and fat mass
index percentiles in caucasians aged 18-98 y. Int J Obes. 2002;26:
953–60.
19. Deurenberg-Yap M, Schmidt G, van Staveren WA, et al. The para-
dox of low body mass index and high body fat percentage among
Chinese, Malays and Indians in Singapore. Int J Obes Relat Metab
Disord. 2000;24:1011–7.
20. De Schutter A, Lavie CJ, Arce K, et al. Correlation and discrepan-
cies between obesity by body mass index and body fat in patients
with coronary heart disease. J Cardiopulm Rehabil Prev. 2013;33:
77–83.
21. Lu Y, Shu H, Zheng Y, et al. Comparison of fat-free mass index and
fat mass index in Chinese adults. Eur J Clin Nutr. 2012;66:1004–7.
22. Provencher MT, Chahla J, Sanchez G, et al. Body mass index ver-
sus body fat percentage in prospective national football league ath-
letes: overestimation of obesity rate in athletes at the national foot-
ball league scouting combine. J Strength Cond Res. 2018;32:1013–
9.
23. Wildman RP. The obese without cardiometabolic risk factor clus-
tering and the normal weight with cardiometabolic risk factor clus-
tering. Arch Intern Med. 2008;168:1617.
24. Flowers E, Molina C, Mathur A, et al. Adiposity and cardiovascular
risk clustering in South Asians. Metab Syndr Relat Disord.
2013;11:434–40.
25. Arnlov J, Ingelsson E, Sundstrom J, et al. Impact of body mass
index and the metabolic syndrome on the risk of cardiovascular
disease and death in middle-aged men. Circulation. 2010;121:
230–6.
26. Silva-Zolezzi I et al. Analysis of genomic diversity in Mexican
Mestizo populations to develop genomic medicine in Mexico.
Proc Natl Acad Sci U S A. 2009;106:8611–6.
Publisher’s Note Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.