Overweight and obesity have been consistently and strongly associated with increased risk of common chronic disease, such as type 2 diabetes (T2D), cardiovascular disease (CVD), hypertension, and some cancers [1]. Yet, a recent large-scale study by Flegal et al. [2] suggests that the obesity-associated disease risk not necessarily translates in higher mortality. Their meta-analysis of 97 studies, representing more than 2.88 million individuals, showed that, while obesity class 2 and 3 (BMI ≥35 kg/m2) is associated with a significantly higher all-cause mortality compared to normal weight individuals, no difference was observed between obesity class 1 (BMI 30 < 35 kg/m2) and normal weight individuals. Most intriguingly, overweight (BMI 25 < 30 kg/m2) was associated with a significantly lower all-cause mortality compared to normal weight.

Similar observations have been made in some other large-scale studies before and are prone to misrepresentation, causing confusion about the risks of overweight and obesity among the general public, media, and clinicians, some of whom might like to believe that being overweight or obese is not as much of a health burden as scientists and policy makers have warned it is. After all, a substantial number of overweight and obese individuals seem to be protected from developing metabolic complications. On the other hand, some normal weight individuals exhibit a metabolically obese profile.

Could the existence of subgroups of metabolically healthy overweight and obese (MHO) individuals, and of metabolically obese but normal weight (MONW) individuals help explain the relationship between the BMI-categories and all-cause mortality reported by Flegal et al.? This would require each of these subgroups to be at least (i) sufficiently large and (ii) their metabolic profile to be associated with all-cause mortality, and for these subgroups to be more prevalent in the studies included in Felgal's et al.'s meta-analysis than in other studies that suggest a J-shaped relationship between BMI and mortality with a nadir among the normal weight category.

Even though MHO and MONW have been described since the 1980s, there is currently no uniform definition for diagnosis, resulting in a wide range of reported prevalences. An estimated 15-45% of obese individuals are considered metabolically healthy, having favorable glucose, lipid, liver enzyme, hormone, and inflammation levels, despite excess adiposity, and about 5-30% of normal weight individuals demonstrate one or more metabolic abnormalities [3-6]. For example, in a representative sample of 5,440 US adults from NHANES 1999-2004, 31.7% of the obese (equivalent to 10% of the total population) were considered MHO when “healthy” was defined by the presence of maximum one of six common cardiometabolic risk factors (Figure 1) [3]. With a more stringent definition of “healthy,” i.e. no risk factors allowed, the prevalence of MHO dropped to 17% [3]. When “unhealthy” was defined as having two or more risk factors, 23.5% of normal weight individuals (equivalent to 8% of the total population) were classified as MONW, which dropped to 8.6% when “unhealthy” was defined as having at least three risk factors (Figure 1) [3]. About half of the overweight individuals were considered “healthy” when using the less stringent definitions (Figure 1) [3].


Figure 1. Prevalence of metabolically healthy obesity and metabolically obese normal weight in 5,440 US adults from the NHANES study [3] as a proportion of each BMI-category (Panel A) and as a proportion of the total population (Panel B) (metabolically “healthy” is defined as presence of <2 cardiometabolic risk factors; metabolically “unhealthy” is ≥2 risk factors).

Download figure to PowerPoint

Age, sex, race/ethnicity, BMI, waist circumference, physical activity, smoking, and alcohol intake are the main correlates of MHO and MONW [3-9]. The MHO individuals tend to be younger, female, more likely to be non-hispanic black, have a relatively lower BMI and waist circumference, and live a healthier lifestyle compared to their unhealthy obese peers, whereas MONW individuals are more often older, male, less likely to be non-hispanic black, have a larger waist circumference and live a less healthy lifestyle compared to healthy normal weight individuals [3-9]. However, even after adjusting for these correlates, the MHO continue to have a more favorable metabolic risk profile than the unhealthy obese, and the MONW continue to be at greater risk than the healthy normal weight [4, 6-8], suggesting that innate mechanisms also contribute to the metabolic difference among obese and normal weight individuals.

It has been speculated that peripheral physiological pathways implicated in adipogenesis, adipose tissue expandability, adipocytes differentiation, fat distribution, ectopic fat accumulation, and responsiveness to endocrine secretions explain, at least in part, the metabolic heterogeneity among obese and among normal weight individuals [7, 9]. As such, variations in genes encoding proteins that constitute these pathways are considered potential determinants of MHO and of MONW. The most convincing evidence for such physiological mechanisms has come from experimental studies in animal models [10-12], whereas human physiological studies examining potential causal pathways are limited. Using a large-scale genome-wide association study (GWAS), we identified a locus near IRS1 that is highly significantly associated with body fat percentage (BF%) [13]. Most intriguingly, it was the BF%-increasing allele that is robustly associated with a favorable metabolic profile, including decreased insulin resistance, dyslipidemia, and increased adiponectin levels, and with reduced risk of T2D and coronary heart disease (CHD) [13]. In targeted follow-up analyses, we showed that the BF%-increasing allele was associated with increased IRS1 expression in fat tissue and with preferential storage of fat in the subcutaneous rather than in the harmful visceral adipose tissue [13]. As large-scale GWAS continue to identify more genetic loci for a wide range of adiposity and metabolic traits, the integration of these discoveries promises to reveal more genes for MHO and MONW, and thus new physiological insights into the metabolic heterogeneity among obese and normal weight individuals.

While one would speculate that the MHO individuals would have a lower disease risk and mortality, possibly similar to that of healthy normal weight individuals, and that the MONW would have an increased disease risk and mortality compared to the normal weight, the evidence for such associations is inconclusive. For example, some report that risk of T2D and CVD in MHO is similar to that of healthy normal weight, and lower than that of unhealthy obese, others found the MHO to have the same increased disease risk as the unhealthy obese [4, 14]. In a large-scale prospective study of >22,000 men and women, who were followed for an average of 7 years, all-cause mortality in MHO [hazard ratio (HR) 0.94] was not significantly different from healthy normal weight individuals, whereas it was significantly increased in MONW (HR 1.59) and unhealthy obese (HR 1.79), even after adjusting for relevant correlates [6]. However, in 6,000 individuals of NHANES III with 9 years of follow-up, all-cause mortality was significantly higher in all obese individuals, even in the MHO (HR 2.74), compared to healthy normal weight [15]. Similar findings were reported in a 30 year-follow-up study of 1,758 Swedish men, i.e. both the unhealthy obese (HR 2.43) and MHO (HR 1.65) had a significantly higher mortality than healthy normal weight. None of the studies reported diseases risks and mortality for the obesity subclasses. Inconsistencies between studies may be due to difference in the definition of “(un)healthy,” adiposity traits used to define obesity, population characteristics, and duration of follow-up.

Thus, even though the MHO and MONW represent a non-negligible proportion of the population, their representation among the obese and among the normal weight respectively may not be sufficiently large to be the cause of the attenuated all-cause mortality among the obese class 1 and the increased all-cause mortality among the normal weight, as reported by Flegal et al., in particular given the uncertainty about the beneficial impact of the MHO and detrimental influence of MONW on disease risk and mortality. Along the same lines, even though half of the overweight individuals are metabolically healthy, it is unlikely that this would result in a lower all-cause mortality compared to the normal weight individuals of whom ¾ is metabolically healthy. Taken together, the metabolic heterogeneity observed in each of the BMI-categories is unlikely to underlie the right-shifted and J-shaped association with all-cause mortality.


  1. Top of page
  2. References