Contribution of genetic and environmental factors to obesity indices and their overlap

Abstract

Aims

Numerous indices have been developed to quantify obesity and the distribution of body fat; however, none are sufficient alone, and combined usage is complicated by their potential intercorrelation. This study aims to quantify genetic and environmental influences on anthropometric measures, 12 derived obesity indices and the extent of their overlap.

Materials and Methods

We used four anthropometric measurements (height, weight, waist and hip circumference) from the baseline of the multi‐generational Lifelines cohort study to calculate 12 indices of obesity and body fat distribution. Variance components attributable to genetic (h ²), shared (c ²) and unique environmental (e ²) factors along with pairwise phenotypic (r _P), genetic (r _G), shared (r _C), and unique environmental (r _E) correlations were estimated using ASReml software. Genetic and environmental contributions to the phenotypic correlations were also quantified.

Results

A total number of 152 298 adult individuals (females = 89 091, 58.4%) were included. Strong correlations were observed among most indices. (r _P, r _G, r _C, r _E > 0.8). A body shape index (ABSI) and hip index (HI) were weakly correlated with other indices, largely independent of body mass index (BMI) ($r_{\mathrm{P}}$ < 0.10), and had the highest $e^2$, accounting for 64.2% and 75.7% of their variance. Height showed the highest heritability ($h^2$ = 91.7%), whereas most other traits were moderately heritable ($h^2$ = 45%–55%).

Conclusion

The high correlation between the majority of obesity indices implies their redundancy. In contrast, ABSI and HI were relatively independent of BMI and other indices and showed the greatest influence from individual‐specific environmental factors, suggesting their potential utility as complementary tools in epidemiological research, clinical risk prediction, and monitoring of targeted interventions.

1. INTRODUCTION

The global prevalence of obesity has almost tripled between 1975 and 2016 and it is predicted that by 2030, over 1 billion people will be obese. ¹ , ² This trend will induce a major health and economic burden in the near future and significantly impact mortality and life expectancy. ³ Obesity is a heritable trait, with numerous genetic variants strongly influencing individual susceptibility in addition to environmental factors. ⁴ , ⁵

Extensive research emphasises the importance of body composition and fat mass distribution in overall health. ⁶ Visceral fat, rather than total body fat, is associated with cardiometabolic risk. ⁶ , ⁷ Moreover, fat accumulation at different sites is associated with different morbidities. For example, abdominal fat is associated with an increased risk of metabolic disorders, whereas gluteofemoral fat plays a protective role. ⁸

Quantitative indices are practical indicators of obesity, and play a central role in clinical and epidemiological research. ⁹ Nonetheless, their limitation to explain the overall phenotype, and consequently overall health should not be overlooked. For example, while body mass index (BMI) is the current standard measure in clinical practice as an index of general obesity, it cannot discriminate between fat and lean mass and is not an indicator of how body weight is distributed. ¹⁰ Similarly, waist‐to‐hip ratio (WHR) measures abdominal obesity and is considered an indicator of fat distribution, but it cannot differentiate the respective contributions of abdominal versus gluteofemoral size. ¹¹ Using a combination of indices to assess central obesity alongside overall weight status provides a more comprehensive evaluation of obesity in clinical practice. ¹² A recent Lancet Commission report recommends that confirmation of excess fat mass should be based on at least two anthropometric criteria, such as BMI combined with at least one other index. ¹³ However, some degree of overlap is expected between indices which complicates analyses and interpretations. ¹⁴ , ¹⁵ Indices might even be considered equivalent if their correlation is high. For example, it has been suggested that BMI, waist circumference (WC), and waist‐to‐height ratio (WHtR) can be used interchangeably to a moderate extent. ¹⁶ , ¹⁷ Novel indices, including A body shape index (ABSI) and hip index (HI), have therefore been introduced to capture central and gluteofemoral fat distribution independently of overall body size. These indices are derived using an approach analogous to that of BMI. They are based on the allometric (power‐law) relationship among height, weight, WC, and hip circumference (HC), which describes how the measurements scale non‐linearly relative to overall body size. ¹⁴ , ¹⁸ , ¹⁹ While BMI reflects relative weight adjusted for height, ABSI and HI represent relative waist and hip circumference measurements, respectively, in comparison to individuals of the same weight and height. ²⁰ ABSI and HI are predictors of many adverse health outcomes and have been associated with mortality, cardiometabolic disorders, metabolic syndrome and various types of cancer. ²¹ , ²² , ²³ , ²⁴

In this study, analysed data from the large multi‐generational Lifelines cohort study (n > 150 000) to estimate the pedigree‐based narrow‐sense heritability ($h^2$) of anthropometric measurements and several commonly used indices, as well as the proportion of variance due to shared (or common) environmental ($c^2$) and unique environmental ($e^2$) factors. We also quantified the phenotypic ($r_{\mathrm{P}}$), additive genetic ($r_{\mathrm{G}}$), shared environmental ($r_{\mathrm{C}}$) and unique environmental ($r_{\mathrm{E}}$) correlations between each pair of indices to explore the extent to which they overlap. Subsequently, we partitioned each $r_{\mathrm{P}}$ into its underlying genetic and environmental components in order to assess the relative contribution of each factor to the overall correlation. Our findings provide insights into the contribution of genetic and environmental factors to four anthropometric traits (height, weight, waist and hip circumference) and a comprehensive selection of 12 calculated indices of obesity and body fat distribution, as well as to their degree of overlap.

2. METHODS

2.1. Study population

This study used available data from the Lifelines Cohort Study and Biobank. Lifelines is a multi‐generational, population‐based study that includes more than 167 000 individuals from the northern Netherlands. The overall design and cohort profile of the Lifelines study has been described before. ²⁵ , ²⁶ In short, eligible participants were invited to join this cohort study by their general practitioners. Other family members including their parents, children and partners were also invited to participate in order to gather a multi‐generation family dataset. Biological samples, questionnaire data and measurements were collected at baseline (2006–2013) and during follow‐up assessments every 5 years, which will continue for at least 30 years. The Lifelines study was approved by the ethics committee of the University Medical Center Groningen and conducted according to the principles of the Declaration of Helsinki. Written informed consent has been obtained from all participants prior to enrollment. Only adult participants above 18 years of age were included in this analysis (N ~ 152 000). Household composition of the individuals was inferred based on registered postal codes. Participants residing at the same postal address were assigned a common household ID, as described in a previous report. ²⁷ This resulted in 29 744 families of up to four generations (of size ≥2) and average family size of 3.72. The largest family contained 169 members. In this study, “genetic” refers to additive genetic effects, and “shared environment” refers to the contribution of household to trait variation, unless otherwise specified.

2.2. Anthropometric measurements and indices

Anthropometric measurements were taken by trained personnel according to a standard protocol. Data from the baseline assessment in Lifelines were used. Participants with extreme measurements of weight, height, WC, and HC, defined as values exceeding 5 standard deviations from the population mean, were excluded from subsequent analyses. Anthropometric indices were then calculated as defined in Table 1. ABSI values were multiplied by 1000 to improve data representation and analytic properties because the original values were very small (<0.1).

TABLE 1.

Formulas used to calculate anthropometric indices.

No.	Abbreviation	Index	Equation
1	BMI	Body mass index 28	$\frac{W}{H^2}$
2	TMI	Tri‐ponderal mass index 29	$\frac{W}{H^3}$
3	WHtR	Waist‐to‐height ratio 30	$\frac{\mathit{WC}}{H}$
4	WHR	Waist‐to‐hip ratio 31	$\frac{\mathit{WC}}{\mathit{HC}}$
5	BAI	BAI body adiposity index 32	$\frac{\mathit{HC}}{\sqrt{H^3}}-18$
6	BRI	Body roundness index 33	$364.2-365.5\times \sqrt{1-\frac{{\left(\frac{\mathit{WC}}{2\pi }\right)}^2}{{\left(\frac{H}{2}\right)}^2}}$
7	AVI	Abdominal volume index 34	$\frac{2\times {\mathit{WC}}^2+0.7{\left(\mathit{WC}-\mathit{HC}\right)}^2}{1000}$
8	ABSI	A body shape index 14	$\mathit{WC}\times W^{\frac{-2}{3}}\times H^{\frac{5}{6}}$
9	HI	Hip index 18	$\mathit{HC}\times W^{-0.482}\times H^{0.310}$
10	CI	Conicity index 35	$\frac{\mathit{WC}}{0.109\times \sqrt{\frac{W}{H}}}$
11	eTBF	Estimated total body fat 36 , 37	$100\times \frac{\left(-Z+A-B\right)}{C}$ $\mathrm{A}=\left(4.15*\mathrm{WC}*39.3701\right)$ $\mathrm{B}=\left(0.082*\mathrm{W}*2.20462\right)$ $\mathrm{C}=\left(\mathrm{W}*2.20462\right)$ $\mathrm{Z}=98.42\left(\mathrm{men}\right)$ $\mathrm{Z}=76.76\left(\text{women}\right)$
12	RFM	Relative fat mass 38	$64-\left(20\times \frac{H}{\mathit{WC}}\right)+\left(12\times S\right)$ $\mathrm{S}=0\left(\mathrm{men}\right)$ $\mathrm{S}=1\left(\text{women}\right)$

Note: Except for HI for which HC and H are in cm, BAI for which HC is in cm, BRI for which WC and H are in cm, and AVI for which HC and WC are in cm.

Abbreviations: H, height (m); HC, Hip circumference (m); W, weight (kg); WC, waist circumference (m).

2.3. Statistical analysis

Baseline characteristics and anthropometric features of the participants are presented as means, with standard deviations in parentheses. Two‐sample independent t tests were used to compare the measurements and indices between male and female groups. The ASReml‐R package (v4.2.0.257) was used to perform univariate and bivariate quantitative genetic analyses. ³⁹ This software uses the residual maximum likelihood (REML) method for variance decomposition and provides unbiased estimates of variance and covariance components. This method separates the total variability of a trait into components attributed to genetic and environmental differences among individuals. Using the mixed‐model framework in ASReml allows the estimation of these components using the differential degrees of relatedness between individuals in the pedigrees. Additionally, phenotypic, genetic, and environmental correlations describe how two traits co‐vary for different reasons. Genetic correlation measures the extent to which the same genetic factors influence both traits, environmental correlations reflect shared and unique environmental effects, and the phenotypic correlation represents their overall observable covariation resulting from genetic and environmental contributions.

Household composition of the individuals was included in univariate and bivariate models as a proxy for the influence of shared environment. We used the Bonferroni method to adjust the p values, with a two‐sided significance threshold set at α = 0.05. The R programming language (v4.4) was used for data preparation and analysis. ⁴⁰ We used ggplot2 (v3.5.1) ⁴¹ and corrplot (v0.92) ⁴² packages for visualising the results.

2.4. Univariate analysis

We estimated the proportion of variance in each trait, attributable to genetic factors ($h^2$; narrow‐sense heritability), shared environment ($c^2;\text{household}$), and unique environment ($e^2$; residual variance) using the ASReml‐R software. These were calculated as $h^2=V_A/V_P,{\mathrm{c}}^2=V_C/V_P$, and $e^2=V_E/V_P$, where $V_A,V_C$, and $V_E$ represent the genetic, shared environmental, and unique environmental variances, respectively, and $V_P$ is the total phenotypic variance.

Age, age² and sex of the subjects were included in the mixed model analysis as fixed effect variables. The genetic relatedness matrix and household information were included as random effect variables. Lifestyle factors (e.g., smoking) were not included as covariates, allowing unbiased estimates of the overall genetic and environmental variance components. The significance of variance components ($h^2>0,c^2>0$) was assessed using a likelihood‐ratio test and by comparing the full model to a sub‐model where the respective component was constrained to be zero. To account for multiple testing, p value <0.0016 (0.05/32) was considered significant (16 phenotypes × 2 variance components = 32).

2.5. Bivariate analysis

The genetic $\left(r_{\mathrm{G}}\right)$, shared environmental $\left(r_{\mathrm{C}}\right)$, and unique environmental $\left(r_{\mathrm{E}}\right)$ correlations between pairs of traits were estimated using bivariate models in ASReml‐R software. In this model, the genetic correlation is computed as $r_{\mathrm{G}}=\frac{{\sigma }_{A_x{A}_y}}{\sqrt{{\sigma }_{A_x}^2{\sigma }_{A_y}^2}},$ where ${\sigma }_{A_x{A}_y}$ is the genetic covariance between traits x and y and ${\sigma }_{A_x}^2$ and ${\sigma }_{A_y}^2$ are the genetic variances for x and y, respectively. A $r_{\mathrm{G}}=0$ would indicate that the two studied traits are influenced by different genetic factors and $\mid r_{\mathrm{G}}\mid =1$ indicates complete sharing of genetic factors. To test whether the genetic correlation was significantly different from zero ($\mid r_{\mathrm{G}}\mid$ >0), the full model was compared to a sub‐model in which the genetic covariance between traits was constrained to be zero, using a likelihood‐ratio test. Other bivariate correlations $\left(r_{\mathrm{C}},r_{\mathrm{E}}\right)$, were estimated and interpreted using the same methods.

Given that phenotypic covariance and variances are sums of their respective genetic, shared environmental and unique environmental components, the phenotypic correlation $\left(r_{\mathrm{P}}\right)$ can be expressed as $r_{\mathrm{P}}=\frac{{\sigma }_{A_x{A}_y}+{\sigma }_{C_x{C}_y}+{\sigma }_{E_x{E}_y}}{\sqrt{\left({\sigma }_{A_x}^2+{\sigma }_{C_x}^2+{\sigma }_{E_x}^2\right)*\left({\sigma }_{A_y}^2+{\sigma }_{C_y}^2+{\sigma }_{E_y}^2\right)}}$. Besides calculating the phenotypic correlation, this formula allowed us to also determine the proportions of this correlation attributable to the three components. Here, ${\sigma }_{A_x{A}_y},{\sigma }_{C_x{C}_y},{\sigma }_{E_x{E}_y}$ are the genetic, shared and unique environmental covariances between traits x and y, respectively, and, ${\sigma }_{A_x}^2,{\sigma }_{C_x}^2,{\sigma }_{E_x}^2$ are the genetic, shared and unique environmental variances for trait x, respectively. The same pattern applies for trait y. Similar to the univariate mixed model analyses, age, age² and sex of the subjects were included as fixed effect variables, and the genetic relatedness matrix and household information were included as random effect variables. To account for multiple testing, p value <0.00014 (0.05/360) was considered significant (120 bivariate comparisons × 3 correlations = 360).

3. RESULTS

3.1. Overall results

A total number of 152 298 adult participants with a mean age of 44.63 (13.13) years were included. Baseline characteristics of the population, overall and stratified by sex, are shown in Table 2. Among the anthropometric measurements, male participants had significantly higher mean height (182.54 vs. 169.29 cm), mean weight (87.83 vs. 73.93 kg), and mean WC (95.07 vs. 86.58 cm), whereas female participants had a significantly higher mean HC (99.95 vs. 98.82 cm). In the calculated indices, the mean of BMI, WHtR, WHR, BRI, AVI, ABSI, and CI were significantly higher in males. The higher HC in females was reflected in elevated BAI and HI indices compared to the males. Indices used for estimating body fat (eTBF and RFM) were also higher in the female population.

TABLE 2.

Baseline characteristics of the participants.

	Total (N = 152 298)	Men (N = 63 207)	Women (N = 89 091)
Anthropometric measurements
Age	44.63 (13.13)	45.28 (13.20)	44.17 (13.07)
Height (cm)	174.79 (9.41)	182.54 (7.06)	169.29 (6.57)
Weight (kg)	79.70 (15.16)	87.83 (13.23)	73.93 (13.72)
HC (cm)	99.48 (9.49)	98.82 (7.47)	99.95 (10.67)
WC (cm)	90.10 (12.40)	95.07 (10.82)	86.58 (12.25)
Obesity indices
BMI	26.04 (4.28)	26.35 (3.64)	25.81 (4.67)
TMI	14.95 (2.66)	14.47 (2.14)	15.29 (2.93)
WHtR	0.52 (0.07)	0.52 (0.06)	0.51 (0.08)
WHR	0.91 (0.08)	0.96 (0.07)	0.87 (0.07)
BAI	25.27 (5.32)	22.14 (3.34)	27.48 (5.35)
BRI	3.76 (1.43)	3.84 (1.25)	3.70 (1.54)
AVI	16.66 (4.57)	18.35 (4.23)	15.45 (4.42)
ABSI	77.74 (5.15)	79.58 (4.16)	76.43 (5.38)
HI	60.08 (3.54)	57.60 (2.46)	61.84 (3.10)
CI	1.23 (0.09)	1.26 (0.08)	1.20 (0.09)
eTBF	26.73 (8.54)	20.71 (6.04)	30.99 (7.41)
RFM	31.57 (7.50)	25.12 (4.53)	36.15 (5.59)

Note: Data are expressed as mean (standard deviation). All differences between male and female groups were statistically significant (p value < 0.0001).

Abbreviations: ABSI, A body shape index; AVI, abdominal volume index; BAI, body adiposity index; BMI, body mass index; BRI, body roundness index; CI, conicity index; eTBF, estimated total body fat; HC, hip circumference; HI, hip index; RFM, relative fat mass; TMI, tri‐ponderal mass index; WC, waist circumference; WHR, waist‐to‐hip ratio; WHtR, waist‐to‐height ratio.

3.2. Results from univariate analysis

Figure 1 illustrates the percentage of variance in each trait attributable to genetic, shared environmental, and unique environmental factors. Corresponding data are available from Table S1. Height and weight had the highest heritability among all measurements and indices with estimates of 91.7% and 56.1%, respectively. Heritability estimates for other measurements including WC and HC were 43.4% and 41.9%, respectively, and ranged between 30% and 54% for the calculated indices, except for HI which had the lowest estimate at 18.0%. The estimated heritability for BMI was 50.7%.

FIGURE 1.

The percentage of observed variance in anthropometric measurements and indices explained by genetic, shared and unique environmental factors. Analyses were adjusted for age, age² and sex of the subjects. All $h^2$ and $c^2$ estimates were significantly different from 0 (p value <0.0001). ABSI, A body shape index; AVI, abdominal volume index; BAI, body adiposity index; BMI, body mass index; BRI, body roundness index; CI, conicity index; $c^2$, proportion of variance due to shared environment; $e^2$, proportion of variance due to unique environment; eTBF, estimated total body fat; $h^2$ indicates the heritability estimate; HC, hip circumference; HI, hip index; Ht, height; RFM, relative fat mass; TMI, tri‐ponderal mass index; WC, waist circumference; WHR, waist‐to‐hip ratio; WHtR, waist‐to‐height ratio; Wt, weight.

Shared environment had a minor impact on height, ABSI and HI ($c^2$ < 10%), and explained in the range of 10%–20% of the variance of other measurements and indices. On the other hand, individual differences in ABSI and HI were predominantly influenced by the individuals' unique environment ($e^2$ = 64.2% and 75.7%, respectively). This finding is in a sharp contrast with height for which the influence of this factor was very small ($e^2$ = 0.3%).

3.3. Results from bivariate analysis

Estimated phenotypic (r _P), genetic (r _G), shared (r _C) and unique environmental (r _E) correlations between the trait pairs are presented in Figure 2. Corresponding separate values are provided in Figures S1–S4.

FIGURE 2.

Estimated correlations between anthropometric measurements and indices. Analyses were adjusted for age, age² and sex of the subjects. All $r_{\mathrm{G}},r_{\mathrm{C}},r_{\mathrm{E}}$ estimates were significantly different from 0 (p value < 0.0001). ABSI, A body shape index; AVI, abdominal volume index; BAI, body adiposity index; BMI, body mass index; BRI, body roundness index; CI, conicity index; eTBF, estimated total body fat; HC, hip circumference; HI, hip index; Ht, height; RFM, relative fat mass; $r_{\mathrm{P}},r_{\mathrm{G}},r_{\mathrm{C}},r_{\mathrm{E}}$ indicate phenotypic, genetic, shared and unique environmental correlations, respectively; TMI, tri‐ponderal mass index; WC, waist circumference; WHR, waist‐to‐hip ratio; WHtR, waist‐to‐height ratio; Wt, weight.

Height showed a positive correlation with the measured traits across all factors, and the strongest association was observed with weight ($r_{\mathrm{P}}$ = 0.35, $r_{\mathrm{G}}$ = 0.45). However, it generally showed a negative or weak correlation with the calculated obesity indices. Other anthropometric measurements (weight, WC, and HC) had strong positive bivariate correlations ($r_{\mathrm{P}}$ > 0.8) and also showed high correlations with most of the calculated indices, especially BMI ($r_{\mathrm{P}},r_{\mathrm{G}}$ > 0.8). ABSI and HI were exceptions in this regard. Height and weight had very low phenotypic and genetic correlations with both ($r_{\mathrm{P}},r_{\mathrm{G}}$ < 0.15). Additionally, there was very little correlation between ABSI and HC and between HI and WC ($r_{\mathrm{P}},r_{\mathrm{G}}$ ≤ 0.15).

Calculated obesity and body fat distribution indices were generally highly correlated with one another ($r_{\mathrm{P}},r_{\mathrm{G}}$ > 0.8), with the exception of ABSI and HI, which displayed relatively low correlations with the rest, particularly with BMI ($r_{\mathrm{P}},r_{\mathrm{G}}$ < 0.1). As expected, ABSI showed the highest correlation with indices directly related to WC, such as WHR ($r_{\mathrm{P}}$ = 0.70, $r_{\mathrm{G}}$ = 0.71), and HI showed moderate correlations with indices based on HC, such as BAI ($r_{\mathrm{P}}$ = 0.53, $r_{\mathrm{G}}$ = 0.33). We further observed near perfect correlations between certain indices. For example, all pairwise correlation coefficients ($r_{\mathrm{P}},r_{\mathrm{G}},r_{\mathrm{C}},r_{\mathrm{E}}$) between WHtR, BRI, and RFM were above 0.95. Similarly high correlations were found between WC and AVI. It was also observed that $r_{\mathrm{C}}$ between the indices was generally higher than $r_{\mathrm{G}}$. This is particularly evident for ABSI, which, despite having a relatively low $c^2$ of 5.7% and $h^2$ of 30.1%, had high $r_{\mathrm{C}}$ with other traits ($r_{\mathrm{C}}$ > 0.65).

To better understand the contributions of genetic and environmental factors to pairwise $r_{\mathrm{P}}$, we partitioned each correlation into genetic (A), shared (C), and unique environmental (E) components. Figure 3 displays a schematic representation of these components. Corresponding values are provided in Table S2. The phenotypic correlations between height and other traits were almost entirely attributable to the genetic component. This is consistent with the high heritability of height ($h^2$ = 91.7%), and indicates that the observed phenotypic relationships between height and other traits are primarily a result of shared genetic factors rather than environmental components. For other pairwise phenotypic correlations, the genetic component remained the most significant contributor, though to a lesser extent, followed by unique and shared environmental components. However, given the low heritability of ABSI and HI ($h^2$ = 30.1%, 18.0%), pairwise phenotypic correlations involving these traits were predominantly influenced by the unique environmental component.

FIGURE 3.

Components of phenotypic correlations attributable to genetic (A), shared (C), and unique environmental (E) effects. ABSI, A body shape index; AVI, abdominal volume index; BAI, body adiposity index; BMI, body mass index; BRI, body roundness index; CI, conicity index; eTBF, estimated total body fat; HC, hip circumference; HI, hip index; Ht, height; $r_{\mathrm{P}}$, phenotypic correlation; RFM, relative fat mass; TMI, tri‐ponderal mass index; WC, waist circumference; WHR, waist‐to‐hip ratio; WHtR, waist‐to‐height ratio; Wt, weight.

4. DISCUSSION

In this study, we analysed 4 anthropometric measurements (height, weight, WC, HC) and 12 indices for obesity and fat mass distribution (BMI, TMI, WHR, WHtR, BAI, BRI, AVI, ABSI, HI, CI, eTBF and RFM) using comprehensive data from the family‐based Lifelines Cohort Study to investigate the genetic and environmental factors underlying their variability and interrelationships. We observed strong phenotypic and genetic correlations among the indices, with some showing near perfect correlations. In contrast, ABSI and HI were largely independent from the others, especially from BMI. Height had the highest heritability ($h^2$ = 91.7%), while most other traits had moderate heritability estimates at around 0.50% (±0.05%).

The independence of the two newer indices (ABSI and HI) from BMI, as intended by design, ¹⁴ , ¹⁸ was confirmed in our analysis ($r_{\mathrm{P}}$ = 0.06 and 0.09, respectively). The phenotypic correlation between these two indices was also small ($r_{\mathrm{P}}$ = 0.17). In contrast, WC‐AVI and WHtR‐BRI‐RFM showed almost perfect correlations across the four different correlation types ($r_{\mathrm{P}},r_{\mathrm{G}},r_{\mathrm{C}},r_{\mathrm{E}}$ > 0.95), suggesting they reflect the same underlying phenotype.

As the most commonly used obesity index in clinical practice, BMI had negligible correlations with ABSI and HI ($r_{\mathrm{P}},r_{\mathrm{G}}$ < 0.1), moderate correlations with WHR, CI, eTBF ($r_{\mathrm{P}},r_{\mathrm{G}}$ = 0.4–0.6) and strong correlations with other indices ($r_{\mathrm{P}},r_{\mathrm{G}}$ > 0.8). Consistent with previous reports, we observed strong phenotypic correlations between BMI and WC ($r_{\mathrm{P}}$ = 0.86), BMI and WHtR ($r_{\mathrm{P}}$ = 0.88), WC and WHtR ($r_{\mathrm{P}}$ = 0.95) and a lower correlation between BMI and WHR ($r_{\mathrm{P}}$ = 0.38). ⁴³ A recent report recommends using at least one additional anthropometric index alongside BMI for confirmation of excess body fat. ¹³ Based on our findings, WHR, ABSI and HI may be preferable alternatives due to lower correlation compared to WHtR and WC.

Another commonly used measure for abdominal obesity, WC, displayed moderate to high correlations with most calculated indices ($r_{\mathrm{P}},r_{\mathrm{G}}$ > 0.65), moderate correlations with ABSI ($r_{\mathrm{P}},r_{\mathrm{G}}$ ~ 0.50), and low correlations with HI ($r_{\mathrm{P}},r_{\mathrm{G}}$ < 0.15). WHR also showed high correlations with several indices ($r_{\mathrm{P}},r_{\mathrm{G}}$ = 0.6–0.9), moderate correlations with BMI ($r_{\mathrm{P}}$ = 0.38, $r_{\mathrm{G}}$ = 0.49), but had lower relationships with both BAI and HI ($\mid r_{\mathrm{P}},r_{\mathrm{G}}\mid$ < 0.4). These correlation patterns should be considered when BMI, WC, and WHR are used in combination with other indices.

By partitioning the phenotypic correlation between anthropometric and calculated measures, our findings showed that genetic factors are generally more influential than (shared or unique) environmental factors in the overall phenotypic relationship. This influence is particularly evident for height. Given its very high heritability of >90%, nearly all phenotypic correlations involving this variable could be explained to large extent by the genetic component. In contrast, ABSI and HI, exhibited much lower heritability ($h^2$ = 30.1% and 18.0%, respectively), and therefore, relationships involving these traits appeared to be predominantly caused by unique environmental influences.

Our results showed that height had the highest heritability estimate among all anthropometric measurements and indices ($h^2$ = 91.7%), which is consistent with previous reports of 86% heritability in twin studies, ⁴⁴ and a narrow‐sense heritability estimate of approximately 80%. ⁴⁵ Lower heritability estimates have been reported for body weight, falling in the range of 65%–70%. ⁴⁴ , ⁴⁶ For other anthropometric measurements, including WC and HC, reported estimates are typically in the range of 40%–80%, depending on the study design. ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ Our results align with these findings, as genetic factors explained 56.1%, 43.4%, and 41.9% of the variation in body weight, WC, and HC, respectively. The contribution of shared environmental factors to these three measurements ranged from 14% to 20%. HC was more influenced by unique environmental factors ($e^2$ = 43.5%) than other measurements.

In our results, $h^2$ for BMI was estimated at 50.7% in line with previous reports that have shown a range of 47%–90% in twin studies and 24%–81% in family‐based studies. ⁵¹ Additionally, $h^2$ for WHtR and WHR were 44.6% and 30.1%, respectively, in our analysis. Previous twin and family studies have reported similar values in the range of 30%–60%. ⁴⁷ , ⁵⁰ , ⁵² , ⁵³ , ⁵⁴ A previous genome‐wide association study (GWAS) reported SNP‐based heritability estimates of 14% in men and 21% in women for ABSI, and 14% in men and 17% in women for HI. ⁴³ Total narrow‐sense heritability had not been previously reported, but in our study was estimated at 30.1% for ABSI and 18.0% for HI. The similarity between SNP‐based and narrow‐sense heritability of HI suggests that the genetic architecture of this phenotype is likely influenced predominantly by common genetic variants. As far as we know, heritability estimates for the other indices (TMI, BAI, BRI, AVI, RFM, eTBF and CI) have not been previously reported. In our study, they ranged between 32.1% and 54.4%. The effect of shared environment on the studied anthropometric indices generally ranged from 10% to 20%, but this effect was lower for ABSI and HI ($c^2$ ~ 6%). In contrast, these indices were predominantly influenced by unique environmental factors, which accounted for 64.2% and 75.7% of their variances, respectively. This suggests that they could be considered as sensitive indicators of individual‐specific lifestyle exposures, such as physical activity and nutrition, and be useful measures for assessing changes in body fat distribution in response to targeted interventions.

A wide variety of indices have been suggested in the literature to measure obesity. However, no single index is sufficient to fully capture its complexity ⁵⁵ , ⁵⁶ and integrating multiple anthropometric indices is essential for a comprehensive assessment. This approach becomes even more critical with the increasing use of weight‐loss drugs because it allows a broader evaluation of body composition and treatment outcomes beyond weight reduction alone. Weight adjusted for height (i.e., BMI) and waist and hip circumferences adjusted for both weight and height (i.e., ABSI and HI) produce measures that are approximately mutually independent, and each captures a distinct aspect of body fat distribution. Using these indices in combination could provide a more reliable characterisation of body composition, thereby enhancing the precision of studies examining the effects of adiposity on health outcomes. Their joint use has been suggested for constructing an anthropometric risk indicator (ARI) to characterise overall obesity. ¹⁹

Considering the vast number of indices and their importance in defining and capturing various aspects of obesity, their correlations, interchangeability, and complementary use have not received proper attention, and existing reports have mainly focused on a few indices (e.g., BMI, WHR, WC). ¹⁶ , ⁵⁷ To the best of our knowledge, this study presents the most comprehensive report to date on the contribution of genetic and environment to anthropometric indices, as well as their intercorrelations. To obtain accurate estimates of these contributions and correlations, we analysed extensive pedigree data from the large, multi‐generational, population‐based Lifelines cohort study (N ~ 152 000). Another strength is that we used continuous values instead of converting them to categorical phenotypes (e.g., overweight vs. underweight), because the optimal thresholds for certain indices (such as ABSI and HI) are not clearly defined. Additionally, using continuous variables offers greater accuracy for correlation analysis. ⁵⁸ All our results were highly significant, even though we applied the Bonferroni correction method to ensure the robustness of our findings. This multiple testing correction method is highly conservative particularly in case of correlated variables like was the case in this study. One limitation of this report is the lack of sex‐stratified analyses due to its family‐based design. Implementing such analyses would require the estimation of sex‐specific variance components, which is difficult to specify in large pedigree data. Another limitation is that, although the Lifelines cohort is broadly representative of the adult population in the north of the Netherlands, ⁵⁹ our results may not generalise to other ethnicities.

5. CONCLUSION

Our findings indicate a strong correlation between most of the commonly used anthropometric indices, whereas the recently proposed ABSI and HI measures appear relatively independent, particularly from BMI. This could imply an important complementary role for these two indices in epidemiological and clinical studies. Furthermore, they might also be valuable for monitoring changes in response to personalised interventions. Combining BMI with WHR, ABSI, and HI could provide a more informative and comprehensive assessment of obesity than using BMI with other indices or relying on a single index alone. Furthermore, we observed that certain indices (AVI, BRI, and RFM) are redundant, and simpler indices (WC and WHtR) could be utilised instead.

AUTHOR CONTRIBUTIONS

Ahmad Vaez, Ilja Maria Nolte, Harold Snieder: designed the research; Alireza Ani: conducted the statistical analyses and wrote the initial manuscript; Ahmad Vaez, Ilja Maria Nolte, Harold Snieder: critically reviewed the manuscript; and all authors: commented on previous versions of the manuscript, read and approved the final manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

Supporting information

Figure S1. Flowchart of individuals included in the study.

Figure S2. Pairwise phenotypic correlations (r _P) between anthropometric measurements and indices.

Figure S3. Pairwise genetic correlations (r _G) between anthropometric measurements and indices.

Figure S4. Pairwise shared environmental correlations (r _C) between anthropometric measurements and indices.

Figure S5. Pairwise unique environmental correlations (r _E) between anthropometric measurements and indices.

Table S1. Percentages of observed variance in anthropometric measurements and indices explained by genetic, shared and unique environmental factors.

Table S2. Components of phenotypic correlation attributable to genetic, shared and unique environmental factors.

DATA AVAILABILITY STATEMENT

Data may be obtained from a third party and are not publicly available. Researchers can apply to use the Lifelines data used in this study. More information about how to request Lifelines data and the conditions of use can be found on their website (https://www.lifelines-biobank.com/researchers/working-with-us).