The population of this study was sourced from the NHANES database—a large cross-sectional survey conducted by the National Center for Health Statistics—to investigate the health and nutritional status of the population in the United States (14–17). Its research design is complex and exquisite. The principal sample design consisted of multiyear, stratified, clustered 4-stage samples (18). According to the over-sampling standard, researchers over-sampled some subgroups of people and gave them corresponding weights so as to improve the accuracy and reliability of the overall data so that it can represent the demographic characteristics of the entire United States (19). On the official website of NHANES, we referred to the detailed survey contents, survey operations, and data-use methods (20). Personal information was mainly collected through personal interviews and mobile examination center, and all participants provided their signed informed consent (18).

In the present study, we analyzed 39,156 participants from the NHANES during 2011–2018, excluding the following patients: (1) participants aged <20 years, >59 years, and pregnant; (2) lack of data information that can be used to evaluate obesity (BMI and waist circumference); (3) non-obese participants; (4) lack of fat mass data; (5) lack of baseline data (such as income, marital status, smoking, and drinking); (6) lack of comorbidity information. Finally, we included 4,899 obese participants (60,180,984 participants after weighting). The screening process is depicted in Figure S1 .

In this study, all participants were examined by dual-energy X-ray absorptiometry (DXA) to determine the fat mass, which is the most widely accepted method of measuring body composition (21). The fat distribution in the central region includes Android, Gynoid, visceral, subcutaneous, visceral/subcutaneous (V/S), and total abdominal fat ratio (See the Supplementary materials for the definition of these areas). The fat distribution was described by the ratio (%), that is, the fat mass of each part/total fat mass × 100%. Obesity was defined as BMI ≥30 or waist circumference (wc) ≥88 cm in women or wc ≥102 cm in men (22).

Our primary study outcome was the comorbidity risk among obese participants. We obtained whether the patient also had other diseases from the medical conditions file in the NHANES questionnaire section. We included 12 obesity-related diseases in 7 categories reported previously, which included CCVD (such as hypertension, coronary heart disease, heart failure, and stroke), MD (diabetes and gout), RD (asthma, chronic bronchitis), liver disease, renal disease, and cancer and joint diseases. Among these, simple comorbidity is defined as <4 comorbidities, while the participants with ≥4 diseases were defined as complex comorbidities (23).

Considered the complex survey design of NHANES, all statistical analysis was based on sample weight, stratification, and clustering. Continuous variables were expressed by means ± standard deviation (SD), and categorical variables were expressed by percentage (%). Continuous variables were compared with the Student’s t-test or non-parametric test, and the categorical variables were compared with the Rao-Scott Chi-square test. Considering the large difference in the distribution of fat between men and women, we classified the study participants into 2 groups of men and women and applied logistic regression analysis to clarify the relationship between the distribution of fat in different portions and the risk of comorbidity. For continuous variables that did not conform to the normal distribution, we conducted a natural logarithm transformation and also described these variables in the form of sex-specific quintiles. The cutoff value was calculated from the ROC curve. In addition, we used the variance inflation factor (VIF) to detect multicollinearity among covariates. VIF >10 was considered to indicate multicollinearity.

In order to clarify the correlation between fat distribution and comorbidity risk among different subgroups, we analyzed the age subgroups (<45 and ≥45 years), comorbidity number subgroups (simple comorbidity and complex comorbidity), and comorbidity-type subgroups. In order to test the robustness of the results, we performed a sensitivity analysis, adjusted the age subgroups to <40 and ≥ 40 years old, and then performed a logistic regression analysis to clarify the relationship between fat distribution and comorbidity. Two-sided P < 0.05 was considered to indicate statistical significance, and all statistical analyses were performed by the R software (Version 4.1.2).

The mean age (SD) of these 4899 obese participants was 41.35 ± 11.16 years. The majority of them were female, with about 2950 participants. We found significant differences in total fat mass (32.84 ± 8.94 kg vs 36.56 ± 10.52 kg) and fat distribution between men and women (P < 0.001), including Android, Gynoid, visceral, subcutaneous, and abdominal fat ratio and V/S ( Table 1 ).

The majority of patients had comorbidity (54.1%), and there was no significant difference in the comorbidity rates between men and women (P = 0.626). The comorbidity rate of men was about 53.6%, while women had about 54.4%. The comorbidity types of the two groups were similar, including hypertension (men, 32.6%, women, 26.5%), arthritis (men, 14.9%, women, 20.9%) and asthma (men, 12.8%, women, 19%).

After adjusting for age, race, education level, marital status, income, medical insurance, alcohol drinking, smoking, BMI, wc, and arm circumference, the Android fat ratio (OR, 4.21, 95% CI, 1.54–11.50, P_trend=0.007), visceral fat ratio (OR, 2.16, 95% CI, 1.42–3.29, P_trend<0.001), and V/S (OR, 2.07, 95% CI, 1.43–2.99, P_trend<0.001) were independent risk factors for comorbidity in men. The Android fat ratio is “J” type related to comorbidity risk, while the visceral fat ratio and V/S were very significantly linear type related to comorbidity risk, that is, compared to Quintile 1, the OR values of Quintiles 2, 3, 4, and 5 exhibited progressive growth. Simultaneously, the Gynoid fat ratio (OR, 0.87, 95%CI, 0.80–0.95, P_trend=0.001) and the subcutaneous fat ratio (OR, 0.81, 95%CI, 0.67–0.98, P_trend=0.016) played a protective role in the risk of comorbidity, and this trend was still visible after dividing by the cutoff value. After dividing by sex-specific quintiles, we discovered a significant linear inversely dose-response relationship between Gynoid fat ratio, subcutaneous fat ratio, and comorbidity risk, implying that their protective effects were accumulated as fat ratio increases ( Table 2 ).

We also adjusted for covariates among women. The results showed that as continuous variables, Android fat ratio (OR, 4.65, 95% CI, 2.11–10.24, P_trend=0.020), visceral fat ratio (OR, 1.83, 95% CI, 1.31–2.56, P_trend=0.001), and V/S (OR, 1.80, 95% CI, 1.32–2.45, P_trend=0.020) were also independent risk factors for comorbidity. And their dose-response association showed the same trend as in males. Unlike the male results, however, only the Gynoid fat ratio (OR, 0.93, 95% CI, 0.87–0.99, P_trend=0.016) played a protective role ( Table 3 ). Although there was a protective trend in the subcutaneous fat ratio (OR, 0.69, 95% CI, 0.31–1.53), it was not statistically significant (P = 0.350, P_trend=0.208).

We separated men and women into two groups(<45 and≥45 years old) to explore the differences in fat distribution and comorbidity across age groups. We found variations in all fat ratio among participants of two groups, regardless of sex ( Figure S2 ). Further logistic regression analysis showed that ( Tables S1, S2 ; Figure 1 ), in contrast to the results of the total male population, the Android fat ratio (OR, 2.77, 95%CI, 0.87–8.80, P = 0.082) and the Gynoid fat ratio (OR, 0.91, 95%CI, 0.83–1.00, P = 0.056) of men with <45 were not statistically significant with the risk of comorbidity. But in men with ≥45, Android fat ratio (OR, 7.24, 95%CI, 1.25-41.49, P = 0.020) and Gynoid fat ratio (OR, 0.78, 95%CI, 0.68–0.90, P = 0.001) were significantly associated with comorbidity risk, and this trend also existed in women ( Tables S3, S4 ; Figure 2 )

We reclassified the patients according to the number of comorbidities. Complex comorbidities were defined as four or more comorbidities. Based on this, we studied the relationship between fat distribution as a continuous variable and various degrees of comorbidity ( Figures S3, S4 ). The forest plot in Figures 3 , 4 clearly showed that, with the emergence of complex comorbidity, Android fat ratio, visceral fat ratio, and V/S have significantly increased the risk of comorbidity, while the protective effect of Gynoid fat ratio and the subcutaneous fat ratio on the risk of comorbidity had also increased.

We reclassified the patients according to their comorbidity types, mainly including CCVD, MD, RD, liver disease, renal disease, cancer, and joint disease, and studied the relationship between fat distribution and different types of comorbidities. As continuous and categorical variables, the quintile OR value of CCVD comorbidity risk of Android fat ratio, visceral fat ratio, and V/S showed a significant increase. Similarly, the quintiles OR value of Gynoid fat ratio and subcutaneous fat ratio, and CCVD comorbidity risk showed a decreasing trend, indicating that its protective effect was gradually increasing ( Figure S5 ). Except for subcutaneous fat ratio, Android fat ratio, Gynoid fat ratio, visceral fat ratio, and V/S were the same in MD as they were in CCVD ( Figure S6 ). However, when fat distribution was associated with the risk of RD, liver disease, renal disease, cancer, and joint disease, these relationships were less regular ( Figures S7–S11 ).

In addition, we also conducted some sensitivity analyses and discovered that the results were stable. To overcome the bias caused by age grouping, we reset the age boundary and investigated the role of fat distribution in obese participants at the age of 40. Android fat ratio (OR, 2.60, 95% CI, 0.56–12.12, P = 0.219) and Gynoid fat ratio (OR, 0.96, 95% CI, 0.86–1.08, P = 0.501) were not significantly associated with comorbidity among male obese adults aged <40 years ( Figure S12 ). Conversely, the risk effect of Android fat ratio (OR, 5.30, 95% CI, 1.33–21.13, P = 0.019) and the protective effect of Gynoid fat ratio (OR, 0.80, 95% CI, 0.71–0.91, P = 0.006) were obvious for male obese people aged ≥40 years. In women, the results were similar ( Figure S13 ). Considering the impact of menstrual status on female fat distribution and some disease risks (CCVD), the results of subgroups of people with menstrual status by age showed that the comorbidity risk of postmenopausal (older) participants seemed more likely to be affected by abnormal fat distribution ( Figure S14 ). In addition, in order to avoid the impact of estrogen use, we made additional adjustments to the estrogen use of participants with complete estrogen use information, and the results were still stable ( Table S5 ).

This study initially investigated the differences in fat distribution among obese participants of different sexes and ages. To begin, males had greater Android fat, visceral fat, and V/S compared to women, but less Gynoid fat and subcutaneous fat, which may be related to hormone levels, eating habits, living habits, and genetic differences (24). Second, this difference was mirrored in fat function. We also discovered that in men, both visceral fat ratio and subcutaneous fat ratio were strongly linked with comorbidity risk, but in women, only visceral fat ratio was significantly associated with comorbidity risk. This result was completely consistent with the results of Mutsert et al. (25). As a result, in women, just variations in visceral fat may need to be assessed for stratification of comorbidity risk, but in men, the potential effects of subcutaneous fat may need to be additionally assessed. Second, age was an important reason for the differences in fat distribution among participants. With advancing age, Android fat, visceral fat, and abdominal fat increased, but Gynoid fat and subcutaneous fat decreased. This also coincided with previous research results. Aging promotes fat redistribution, that is, loss of subcutaneous fat and growth of visceral fat, and hormonal imbalance can also invert the distribution of Android and Gynoid fat (26). In terms of fat function, older participants were more susceptible to fat than younger participants, which was consistent with previous studies. Preis et al. also found a stronger correlation between fat distribution and metabolic diseases in older participants (27).

For obese participants, complex comorbidities are a difficult public health prevention target (28). Although some studies have noted the relationship between fat distribution or obesity degree and various comorbidities, for example, a recent study by Mika et al. found a dose-response relationship between obese individuals’ BMI and complex comorbidities, obese participants exhibited a 5-fold greater risk of simple comorbidity and a 12-fold increased risk of complex comorbidity compared to healthy weight participants (23). However, few studies have elucidated the relationship between fat distribution and complex comorbidity. When compared to people with simple comorbidity, the fat distribution of participants with complex comorbidity was more closely related to comorbidity risk, and this trend was not affected by sex. The results of this study were unprecedented because it effectively filled the deficiency in previous studies that relied solely on BMI to determine the risk of complex comorbidity.

A large number of studies have shown a strong correlation between obesity and various types of comorbidities, with the most widely reported comorbidities being cardiovascular, metabolic, and respiratory diseases (29–31). Albert et al. showed that obesity can cause a variety of hemodynamic changes, which may lead to cardiac morphological changes and ventricular dysfunction (32). Although this theory has been widely confirmed, we cannot ignore the latest research on the obesity paradox in cardiovascular disease. The mortality of patients with any kind of heart failure has decreased as BMI has increased (33). This contradictory phenomenon prompts us to focus our research on body composition. We found that increasing the Android fat ratio, visceral fat ratio, and V/S would increase the risk of CCVD and even the mortality of special causes. However, the Gynoid fat ratio and subcutaneous fat ratio played considerable protective roles. This conclusion has not been explored in depth before and it may be a reasonable explanation for the “obesity paradox”. The increase in BMI will not benefit all obese people. Participants will not benefit from a rise in BMI induced by Android and visceral fat. Only the increase in BMI caused by Gynoid and subcutaneous fat may achieve the effect of the “obesity paradox”. Similarly, while obesity is associated with the occurrence of MD such as diabetes and gout (34, 35), the risk of MD caused by fat in different regions was not the same.

The most reasonable explanation for this phenomenon is fat heterogeneity. The Android fat and visceral fat are composed of white adipose tissues (WAT), which contribute to metabolism and chronic inflammation in vivo, while triglycerides accumulation in WAT cells in obese people triggers WAT cells remodeling, proliferation, and hypertrophy. The ERK and p38 MAPK pathways are activated by adipocytokines secretion, resulting in increased CCL2 expression in adipocytes. This in turn triggers pro-inflammatory macrophage aggregation, Treg cell reduction, and IL-6 and TNF-α secretion increases, leading to systemic inflammation, insulin resistance, oxidative stress, and a series of metabolic reactions (36, 37). As for Gynoid fat, estrogen induction increases the anti-lipolytic α2-adrenergic receptors in the gluteal-femoral subcutaneous fat depot, causing fat to accumulate in the Gynoid area; hence, Gynoid fat distribution is closely related to estrogen levels (38). Estrogen has been widely recognized as an important factor in regulating obesity and metabolic balance in the body (39). Tran’s study showed that estrogen-regulated multiple calcium-dependent activities in cardiovascular tissues via influencing calcium signaling mechanism components (40). Alternatively, by activating eNOS and increasing NO production, as well as activating cardioprotective signaling cascades including Akt and MAP kinases, cardiac and endothelial cells are protected against apoptosis and necrosis, alleviating pathological myocardial hypertrophy (41). Estrogen also plays an important role in metabolic pathways. Animal experiments have shown that estrogen can increase insulin content and glucose-stimulated insulin secretion in isolated mouse islets, and maintain glucose homeostasis, while its deficiency will disturb oxidative stress and endoplasmic reticulum function, resulting in a complete disorder of insulin function and in vivo metabolism. Therefore, the accumulation of Gynoid fat caused by increased estrogen significantly reduces the risk of comorbidity in obese people.

Interestingly, the increase in total abdominal fat did not appear to affect the risk of any type of comorbidity in this study, which differs slightly from previous reports indicating total abdominal fat was an independent risk factor for cardiovascular disease (42). Visceral fat is primarily responsible for the risk of total abdominal fat, while subcutaneous fat has a protective effect. Therefore, we specially analyzed the role of V/S in comorbidity. It has been proved that V/S can be used to assess the risk of comorbidity, but merely judging total abdominal fat cannot be very effective in guiding clinical work. Individual evaluation of visceral fat, abdominal subcutaneous fat and V/S is necessary to meet the requirements of precision nutrition and precision medicine.