It is generally accepted that several major diseases are related to overweight and obesity. These include metabolic syndrome and type 2 diabetes mellitus, cardiovascular diseases, some tumours, gallbladder diseases, non-alcoholic steatohepatitis, sleep apnoea and osteoarthritis 123456. The endocrine activity of adipose tissue is strongly implicated in most of these diseases 78. Therefore, excess body fat rather than excess body weight is detrimental to health; hence, paradoxically, metabolically obese, normal weight individuals, not deemed obese on the basis of height and weight parameters, were found to be hyperinsulinaemic, insulin-resistant, predisposed to type 2 diabetes mellitus, hypertriglyceridaemic, or to have premature coronary heart disease; these individuals responded favourably to caloric restriction 91011.

Health professionals should assess patients on the basis of their body composition rather than their body weight 12. Thus, new clinical criteria have been introduced to evaluate body adiposity, such as percentage body fat (%BF), considered both alone 13 and in relation to metabolic syndrome risk 14, and waist circumference, as an index of abdominal fat 115. Recently, the Body Fat Mass Index (BFMI, calculated as fat mass in kg divided by height in m²) was introduced, in nutritional assessment, as an additional element for assessing adiposity 161718.

Nevertheless, treatment need is currently assessed on the basis of an individual's Body Mass Index (BMI), calculated as their body weight (in kg) divided by their height in m² 1.

The aim of our study was to determine whether application of the BMI, compared to adiposity-based criteria, results in underestimation of the number of subjects needing a lifestyle intervention.

The women enrolled in the study were slightly younger than the men (38.5 ± 14 years vs 39.2 ± 13.3 years), but this difference was not statistically significant (t = 0.201, p = 0.841).

On average, the men recorded higher body weight, waist circumference and BFMI values than the women, while the women had a higher %BF and metabolic syndrome risk (Table 4). Body fat percentage measured by DXA was significantly linearly related to that measured by BIA (r_Pearson = 0.827, p < 0.0001).

According to the classification based on BMI, 6.3% of subjects were underweight, 52.4% were normal weight, 30.2% were overweight, and 11.1% were obese. Abdominal adiposity was not increased in 63.5% of subjects, increased in 11.1%, and substantially increased in 25.4%. Total body fat, considered alone, was low in 3.2% of subjects, normal in 49.2%, increased in 19.0%, and high in 28.6%. Metabolic syndrome risk related to %BF was low in 7.1% of subjects, normal in 37.5%, increased in 21.4%, and high in 33.9%. BFMI was low in 3.6% of subjects, normal in 42.9%, increased in 32.1%, and high in 21.4%.

Metabolic syndrome risk was the only criterion to show significant gender-related differences (χ² = 9.430; p = 0.024): greater percentages of the women compared to the men showed a high metabolic syndrome risk (37.1% vs 28.6%), and a low and normal metabolic syndrome risk (8.6% vs 4.8% and 45.7% vs 23.8%, respectively); conversely, a greater proportion of the men had an increased metabolic syndrome risk (42.9% vs 8.6%).

Agreement between the BMI categories and the other classification criteria categories varied (Table 5). The most notable discrepancy emerged in the underweight and overweight categories. Of the subjects classed as underweight on the basis of BMI, 75% had normal %BF and metabolic syndrome risk values, and none had an increased waist circumference. In the BMI normal weight subjects, increased abdominal fat was found in 9.1%, increased %BF in 15.2%, and increased metabolic syndrome risk in 20%; a high %BF was found in 6.1% of the normal weight subjects and a high metabolic syndrome risk in 16.7%. In the overweight subjects, a marked discrepancy emerged between the BMI and the adiposity indices: more than 50% of the subjects had a high %BF and metabolic syndrome risk and just under 50% high abdominal fat. In the subjects rated as obese on the basis of their BMI, there was good agreement between all the criteria.

Comparing the BMI with the BFMI categories, 75% of the underweight subjects had a normal fat status. In the BMI normal weight category, 30% of the subjects were "overfat" according to the BFMI, while in the overweight category only 6.7% of subjects had a normal fat status and 40% had a severely "overfat" status. There was good agreement between BMI and BFMI in the obese subjects. Overall, the BMI showed good agreement with BFMI (tau-b Kendall = 0.722, p < 0.001) and waist circumference (abdominal fat) (tau-b Kendall = 0.704, p < 0.001). The level of agreement between the various BMI categories and the %BF and metabolic syndrome risk categories was moderate (tau-b Kendall = 0.672, p < 0.001 and tau-b Kendall = 0.563, p < 0.001 respectively).

The percentages of subjects for whom weight loss treatment would be recommended and strongly recommended, on the basis of each criterion, are summarised in figure 1. The proportion of subjects for whom clinicians would recommend weight loss on the basis of their BMI (41.3%) was not significantly different from the proportions in whom it would be recommended on the basis of %BF (p = 0.344) and abdominal fat (p = 0.508). Conversely, a significant difference emerged in relation to the criteria metabolic syndrome risk (p = 0.022) and BFMI (p = 0.021). The picture changes when analysing the proportion of subjects for whom weight loss would be strongly recommended, with the proportion identified on the basis of BMI (11.1%) differing significantly from the proportion identified by abdominal fat (p = 0.004), %BF (p = 0.003), and metabolic syndrome risk (p = 0.002). Instead, no significant difference was found between the proportions of subjects whom clinicians, on the basis of BMI versus BFMI, would be strongly urged to lose weight (p = 0.125).

Obesity is a metabolic disorder characterised by excess body fat, which is an important risk factor for disease, not only because it is a volume-filling organ, but primarily due to the endocrine activity of adipose tissue 7820.

Although the BMI is easily calculated and can be readily used in population studies, it does not discriminate between fat mass and fat-free mass, or reflect the fat mass distribution in the body 2122232425. In our study, the BMI compared to almost all of the other criteria, based on adiposity, identified a lower percentage of subjects for whom treatment would be recommended and strongly recommended. In particular, the difference was statistically significant when the BMI was compared to metabolic syndrome risk. These findings suggest that a certain proportion of subjects classified as normal weight on the basis of their BMI would not be recommended for treatment even though they harbour excess body fat which could have clinical and metabolic consequences. This is also true if we compare the BMI to the BFMI, the latter an index that denotes the amount of body fat in relation to stature. Indeed, the BFMI, which can be considered a qualitative evaluation of BMI, can result in a better clinical assessment of subjects. From a clinical point of view, basing an individual's treatment needs on their BMI can place at risk those subjects who, on the basis of their metabolic syndrome risk, %BF, and waist circumference, would be strongly advised to undergo weight-loss treatment.

As regards the underweight subjects, a large proportion of them had normal (not low) %BF, waist circumference, metabolic syndrome risk, and BFMI values. However, in these subjects, the misclassification does not have important clinical consequences, since treatment would not, on the basis of any of the criteria, be recommended for any of them.

Thus, our data confirm that indirect estimates of body composition (i.e. BMI) are useful for groups but unreliable in individuals 26. Furthermore, our data also underline the importance of discriminating between lean mass and fat mass, and of relating these parameters to body height, in order to obtain a better nutritional assessment. In addition, the clinical consequences of an altered body composition should be taken into account. The use of the BMI alone to evaluate overweight and obese individuals leads to undesirable misclassifications. Of note, our study sample consisted of people with a sedentary lifestyle, which may predispose them to increased adiposity even before an increase in body weight becomes evident; this implies that normal weight subjects can be at risk of excessive adiposity.

Thus, there is a need to replace the BMI or to supplement it with other diagnostic criteria, in particular, ones that focus on body adiposity, considered both alone and with regard to its distribution. If confirmed by data from larger studies, our results highlight the need to investigate the clinical consequences of excess body fat in normal weight subjects. Several studies have focused on the relationship between the BMI and morbidity and mortality 2327. Unfortunately, to date, few studies have focused on the relationship between %BF and morbidity and mortality, and those that have been conducted were restricted to specific ethnic groups 141628293031. Thus, precise %BF cut-offs that can be used in clinical settings to evaluate an individual's health remain to be determined.

Since %BF has important clinical consequences, and treatment costs and drop-out rates among "overfat" subjects are high, there is clearly a pressing need for precise and unambiguous guidelines. In particular, %BF cut-offs should be defined that can be used in addition to the BMI and waist circumference values already published for the diagnosis and treatment of at-risk subjects 1.

The use of the BMI alone, as opposed to an assessment based on body composition, to identify individuals needing lifestyle intervention may lead to unfortunate misclassifications. Population-specific data on the relationships between body composition, morbidity, and mortality are needed to improve the diagnosis and treatment of at-risk individuals.

The author(s) declare that they have no competing interests.

OC: conception and design of the study; generation, collection and assembly of data; interpretation of data; drafting of the manuscript. SV: analysis and interpretation of data. GP: conception and design of the study; generation, collection and assembly of data. CT: conception and design of the study. MT: collection, assembly and interpretation of data. OT: collection, assembly and interpretation of data. AT: conception and design of the study; generation, collection and assembly of data; interpretation of data; drafting of the manuscript. All the authors have read and approved the final manuscript.