Accuracy of alternative indexes for assessing the nutritional status of men and women Acurácia de índices alternativos para avaliar o estado nutricional de homens e mulheres

A good evaluation of the nutritional status requires knowledge on body composition, casting doubts on the accuracy of some indexes. herefore, the aim of this study was to analyze the accuracy of the following nutritional status indexes: Body Mass Index (BMI), BMI elevated to 2.5 (BMI2.5), Fat Mass Index (FMI) and BMI adjusted for fat mass (BMIfat). Participated of this study 280 subjects (aged 17-48 years), from which the results of BMI, BMI2.5, FMI and BMIfat indexes were analyzed, having the Hydrostatic Weighing method as reference. FMI presented the highest concordance value, but classiied as discrete (k=0.21). he other indexes presented small concordance with results of the reference method (k<0.20). In conclusion, none of the indexes investigated has good accuracy in assessing the nutritional status of the study group, considering that, although they show results of correlation with the reference method, they do not reach the minimum agreement criterion.


INTRODUCTION
here is a growing need to develop good health indicators, since these indicators are used by agencies responsible for monitoring the health conditions of the population.For the Interagency Health Information Network (RIPSA) 1 , morbidity and risk factors for diseases make up an important item on the list of basic health indicators in Brazil.
Nutritional status is one of these indicators, since it is a tool for the establishment of the "overweight prevalence rate", and the drastic change contributes to the increase of morbidity and mortality; in addition, the imbalance tending both to malnutrition and for overweight and obesity can trigger risk factors for a varied number of health problems 2 , strongly inluencing the physical itness of afected individuals 3 .According to Anjos 4 , a good evaluation of the nutritional status requires knowledge on the energy reserves and the metabolically active mass of individuals to be evaluated, which should be obtained by assessing body composition, casting doubts on some indexes that do not take this into account.
One of the indexes most widely used for assessing nutritional status is the Quetelet equation, or body mass index (BMI), which was named by Keys et al. 5 years after its creation.Adopted by the World Health Organization, BMI is considered the simplest nutritional status indicator, involving conventional anthropometric dimensions such as body weight (BW) and height (HEI), but it has the limitation of not estimating the amount of body fat 6 .
Despite the wide use, some authors 4,7 recommend caution, since it is fundamental to emphasize that the fact that BW has good correlation with HEI is not enough to recommend universal use.It is important to correlate BMI values with other body composition measures such as body fat percentage (BF %) 4 .
With the premise that BMI represents body dimensions in the wrong way, since people of higher HEI values have larger structural and physiological compartments than those with lower HEI values, Trefethen 8 developed a new BMI, called BMI 2.5 (BMI elevated to 2.5). he most recent formula was created by the researcher of the University of Oxford (UK) and uses, in addition to the already known BW and HEI, also a numerical correction and the power of 2.5; allowing, according to the researcher, placing people in more appropriate categories to HEI. here are some academic papers demonstrating the strong correlation between traditional BMI and new BMI, but this index still needs to be tested through scientiic research in diferent populations with larger samples and using as a reference a method for assessing the nutritional status with better accuracy.
Another index that promises more reliable results when assessing nutritional status is the Fat Mass Index (FMI) developed by Van Itallie et al. 9 .he equation proposes a better determination of the actual body fat variation.FMI takes into account fat mass (kg) and HEI (m) for determination.FMI classiication values are cited by Kelly et al. 10 , from a validation study with 1195 adult individuals using DEXA as the reference method.However, it should be noted that one of the limitations of the index is that, for use, fat mass value must be obtained by some other validated method.
Also questioning older indexes and seeking a better accuracy when determining the nutritional proile of an individual, Mialich et al. 11 developed BMI adjusted for fat mass (BMIfat) by means of a study with 200 individuals of both sexes.BMIfat is an equation that takes into account, in addition to the BW and HEI values, fat body mass (FM) of the individual (expressed in%). he nutritional status classiication follows the standards cited by Mialich et al. 12 .he new index proposes to characterize speciic populations and / or delimitations of cutof points for classiication of normal weight, overweight and obesity.
In view of the above, the aim of the present study was to analyze the accuracy of BMI, BMI 2.5 , FMI and BMIfat, as alternative indexes to evaluate the nutritional status of adults of both sexes using Hydrostatic Weighing (HW) as the reference method.

METHODOLOGICAL PROCEDURES
his is descriptive-quantitative study, with the participation of individuals of both sexes, living in Santa Maria-RS, Brazil, which is part of a macro project approved by the ethics research committee with human beings of the Federal University of Santa Maria (CAEE -11511112.8.0000.5346).Data come from collections with volunteers from the community in general, carried out over a period of two years by two technically qualiied evaluators and in a specialized laboratory. he study included individuals who had data regarding chronological age (years), ethnicity, physical activity level, BW, HEI and BF%.
hus, the study group consisted of 280 subjects aged 17-48 years, analyzing the results from diferent indexes for nutritional status evaluation (BMI, BMI 2.5 , FMI and BMIfat) and HW.
BW was determined with a Marte® digital scale (Santa Rita do Sapucaí, Brazil), with resolution of 0.1 kg and capacity of 180 kg, and HEI with Cardiomed® stadiometer (Curitiba, Brazil), with resolution of 0.1 cm (according to procedures of Stewart et al. 13 ).BMI was calculated by dividing BW (kg) by squared HEI (m).BMI 2.5 was determined using variables BW (kg) and HEI (m) in an equation consisting of the multiplication of BW by 1.3 and the division of this result by HEI elevated to the power of 2.5.For the classiication of individuals both by BMI and by BMI 2.5 , the WHO reference values were used 6 .
For the FMI determination, variables FM (Kg) and HEI (m) were used in an equation in which FM is divided by squared HEI.FM was obtained through a Maltron® bioelectric impedance analyzer (BI) (Rayleigh, United Kingdom), model BF-906.For the classiication of the FMI results, the Kelly et al. 10 reference table was used.
Finally, to calculate BMI adjusted for FM or BMIfat, BW (kg) was multiplied by 3, FM (%) (obtained by BI) was multiplied by 4, dividing the value by HEI (cm). he classiication criteria are those of Mialich et al. 12 .
As the reference method of the present study to evaluate the nutritional status through HW, a tank designed and appropriate for this purpose was used, on which a Filizola® scale (São Paulo, Brazil) was installed, with capacity of 6 kg and resolution of 0.01 kg, to verify the underwater weight (UW). he water temperature was set between 32°C and 36°C. he procedures used to verify underwater weight are those described by Salem 14 .To determine Body Density (BD) through HW, an equation that considers variables BW, UW, water density (WD) and residual volume (RV) (Goldman and Becklake equation 15 ) was used.After BD determination, the equations to estimate BF% proposed by Heyward and Stolarczyk 16 were used, which can be visualized in Box 1. BF% was classiied according to cutof points of Lohman 17 (Box 1).For nutritional status classiication based on results of the indexes investigated and HW, the authors previously mentioned were used in the irst moment 6,10,12,17 .Secondly, in view of the analysis using the Cohen's kappa coeicient 18 to determine the diagnostic consistency from the results obtained, and that such analysis is only possible if the number of categories for comparison is the same, the allocation of proposals for the interpretation of each index into three categories was performed (below reference, normal weight and above reference).Box 2 shows the diferent indexes and their respective categories of interpretation with corresponding reference, as well as the framing of each of these in the three categories of adequacy proposed in the present study (right column).
It is noteworthy that all data analyses were performed considering the speciic characteristics of each subject, such as age, ethnicity, sex, body composition and physical activity level.Information regarding the physical activity level of the study group was obtained through the International Physical Activity Questionnaire -short version (IPAQ ).Descriptive analysis of data was used, the Kolmogorov-Smirnov test was used for the analysis of normality; the Pearson correlation coeicient was used to deine the degree of association among nutritional status evaluation indexes; and the Cohen's kappa coeicient 18 was used for concordance analysis.he kappa coeicient results were interpreted according to the following parameters 19 : <0 as absence of concordance; small concordance from 0.00 to 0.20; discrete concordance from 0.21 to 0.40; regular concordance from 0.41 to 0.60; good concordance from 0.61 to 0.80; very good concordance from 0.81 to 0.92; and excellent concordance from 0.93 to 1.00; being acceptable, at least, regular concordance.GraphPad Prism 5.00.288 statistical program was used for the elaboration of graphs; and for data analysis, the Statistical Package for the Social Sciences (SPSS, 21.0, Inc., Chicago, IL, USA), adopting 5% signiicance level.

RESULTS
Table 1 shows the general characteristics of the study group, stratiied by sex.All variables presented signiicant correlation with the results of the reference method, being considered high with FMI and moderated with BMIfat, when considering the Male Group (GM), according to categories proposed by Mukaka 20 .In the Female Group (GF), a moderate correlation was found between BF% results from HW and FMI, BMIfat and BMI 2.5 results. he other indexes (BMI 2.5 and BMI for GM and GF, respectively) presented low correlation (Table 2).Since the main focus of the present study was to analyze the diagnostic concordance between indexes used to evaluate the nutritional state and the HW results, FMI was the one that presented the highest nutritional status diagnostic concordance result, being considered as discreet concordance.he other indexes presented little concordance with results of the reference method.However, none of the indexes were able to reach the minimum concordance level (≥0.41) (Table 3).

DISCUSSION
One of the diiculties found for the discussion of results was the fact that no studies were found in literature proposing to test the diagnostic accuracy of FMI, which is the aim of this study.In the study by Schutz et al. 21with 5635 European adults (18-98 years) aimed at establishing the distribution of percentiles according to age groups and sex for FMI, it was observed that the majority of individuals of both sexes is classiied as eutrophic according to reference values 10 . he same occurred in the present study, also considering similar BF% values among studies, which demonstrates the ability to evaluate the independent index of the investigated group.
Researchers claim that FMI may be more efective than BMI because it takes into account body fat mass and has a greater practical value for clinical evaluation.It could be inferred that in a way this argument is true, since FMI presented high correlation with BF% (HW) and the results related to the diagnostic power showed discrete concordance; however, this association of results was evidenced only in GM.
BMIfat also seems to be a good option when evaluating the nutritional status of adult individuals, considering correlation results.Mialich et al. 12 validated this index by performing a study with 501 individuals of both sexes aged 17-38 years.he determination coeicient found was high (R 2 > 0.81), in relation to the traditional BMI, considering the satisfactory validation.In the present study, moderate correlation values (between 0.5 and 0.7) and small concordance regarding the diagnosis of nutritional status for both sexes were found, but it should be taken into account that, unlike the study by Mialich et al. 12 , which used BMI, HW, a method considered as a reference 22,23 .he same author also points out that BMIfat, developed by Mialich et al. 11 , was superior for diagnosing obesity in relation to BMI, even when applied in a new study population.
It is important to emphasize that, although low, BMI 2.5 presented better concordance value than BMI, casting doubt on the ability of the latter index to classify the adequately evaluated nutritional status.Considering that BMI 2.5 is still a little known index, without much scientiic dissemination, it is diicult to ind scientiic studies that seek to test its accuracy and applicability.In one of the few studies found in literature, Ribas Junior 24 , when correlating BMI and BMI 2.5 results, presented correlation coeicient values indicative of positive linearity between both equations.In addition, some academic papers have indicated that the new index classiies a greater percentage of individuals as being overweight, a fact that was also observed in the GF of the present study.
When analyzing the concordance among indexes, when the results of nutritional status diagnosis are associated, having HW as reference, it is evident that both BMI 2.5 and BMI present unsatisfactory results, indicating that they are inadequate for this purpose.
As the index that presented the worst results in the present study, BMI tends to classify a higher percentage of individuals as eutrophic, and in fact a large part of them should be classiied as above the reference, that is, overweight or obese.Studies have shown that BMI is not the best option when evaluating certain populations 7,25 ; however, some authors indicate that this index shows good practical applicability in epidemiological studies 26 and with groups of individuals presenting some pathology 27 .
he diiculty in obtaining a large number of volunteers, especially for the female group, is characterized as a study limitation.It is important to emphasize that when investigating human beings, there is a wide range of variables that need to be controlled and failure to comply with some pre-collection guidelines may result in signiicant sample loss, such as the change in routine water intake and the use of diuretics, which inluence BI results; or the consumption of fermentative foods, which may have an impact on HW results.

CONCLUSIONS
It could be concluded that none of the indexes investigated has adequate accuracy to evaluate the nutritional status of the study population, considering that, although they show signiicant correlation results with the reference method, they do not reach the minimum concordance criterion.herefore, such indexes should be used with caution, since they can lead to a wrong determination of the nutritional status, and, consequently, can have direct impact in the planning of a possible program of physical exercises and in the health of individuals.
As a solution, we suggest the use of other strategies to evaluate body composition such as anthropometric equations that use the results of skinfolds, since they are easy to apply and have relatively low cost.In addition, FMI tended to present a good result, being more efective than BMI, which is widely used, casting doubt that such index can be eicient in the evaluation of more speciic populations.
HW = hydrostatic weighing; BMI = body mass index; BMI2.5 = new body mass index; FMI = fat mass index; BMIfat = body mass index adjusted for fat mass; ABR = below reference; E = normal weight; ACR = above reference; K = Kappa index; T = total number of subjects.

Box 1 .
Equations for converting BD into BF% according to Heyward and Stolarczyk16 and cutoff points for BF% classification according to Lohman17

Box 2 .
Adequacy of categories for nutritional status classification.

Table 2 .
Correlation values between nutritional status index and BF% (HW) results.Groups PH

Table 3 .
Diagnostic concordance of nutritional status between study indexes and the reference method (HW).