Dual X-Ray Absorptiometry Leg Length in Youth

Abstract

Introduction:

Leg length has been found to be a predictor of the risk cardiovascular disease, metabolic disease and type 2 diabetes. Traditional measures of leg-length are conducted by subtracting seated leg length from standing height. We examined whether dual X-ray absorptiometry (DXA) may provide another accurate method to determine leg length. The data were obtained from a study of adolescents. The aims were to examine the relationship between DXA-derived leg length (DXA-LL) and traditionally derived leg length (T-LL) and determine whether the two types of measures yielded similar associations with measures of cardiometabolic health and bone density.

Methods:

140 adolescents either conceived and reared in their birth families in the United States (BR birth reared, n = 74, 39 female) or internationally adopted prior to age 5 from orphanage-like institutions (PI previously institutionalized, n = 66, 44 female) had their leg length measured using traditional methods and then using whole body DXA scan. DXA-LL was determined using custom regions of interest.

Results:

DXA-LL and T-LL measures were nearly perfectly correlated (r (138) =0.96, p<0.001). However, DXA-LL estimates were less than T-LL estimates (77.75±6.04 vs 78.96±6.00cm, p<0.001). The BR group had longer DXA-LL than the PI group (80.67 cm, 74.84 cm) after controlling for age and sex. Controlling for age and sex, longer leg length using DXA-LL was associated with lower glucose, cholesterol, triglycerides, and HDL, and denser bone density. Similar results were obtained with T-LL.

Conclusion:

DXA-LL is an accurate and reliable leg length measurement tool.

INTRODUCTION

Leg length development in children and adolescents is an area of interest due to its bearing on growth trajectory into adulthood¹, and its association with childhood nutritional status and the development of cardiometabolic health^2–8. It has been used as a proxy for growth in childhood because of the proportionally rapid growth of the legs as compared to the trunk or head^1–8. Individual differences in leg length rates may reflect the sufficiency of nutritional resources and social resources available in childhood to support healthy growth and development^2–8. Consistent with the Developmental Origins of Health and Disease (DOHaD) hypothesis⁹, physical and social resources early in life may lay the groundwork for health across the lifespan^9–11. Supporting this, shorter leg lengths in adolescence have been correlated with risk cardiovascular disease (CVD)^{3, 11–13}, type 2 diabetes mellitus (T2DM)^14–16, behavioral disorders¹⁵, and all-cause mortality^{3–4, 8, 16–22}. While the DoHAD hypothesis focuses on the first few years of life, in most of the populations studied, nutritional and social adversity has existed throughout the child’s life. Determining whether threats to growth at any one point in development matter, or the cumulative impact is what is associated with greater cardiovascular risk, is made difficult. To understand the implications of early life nutritional and social deprivation we identified a population of adolescents whose nutritional and social adversity was largely confined to the earliest years of life and compared them to a population of youth who had experienced little to no deprivation during development.

Studies of leg length have traditionally involved subtracting seated height measure from standing height^{9, 14, 23–25}. Recently, dual X-ray absorptiometry (DXA) has been used to measure limb length^23–25. This method has the advantage of also providing information on body composition and bone health (e.g., bone mineral density (BMD))^23–27 However; to integrate the literature emerging using DXA-derived leg length (DXA-LL) with the older literature, it is necessary to determine the degree of agreement between the two types of measures.

Data using traditional leg length (T-LL) and DXA-LL was obtained to compare adolescents and young adults who had been adopted internationally from institutional care (previously institutionalized, PI) with similar-aged youth who were conceived and raised in birth families (birth reared, BR) of similar economic and educational levels as the families adopting internationally. This selection strategy was intended to broaden the distribution of leg length to include shorter leg lengths, as the nutritional and social deprivation in institutional care typically results in growth delays. Institutional care for is a complex and extensively researched form of childhood adversity⁶, often involving inadequate provision of basic needs (e.g., nutrition, cognitive stimulation, absence of stable attachment^{6, 28}). Institutional settings also increase risk of disease, intestinal parasites, and stress, all of which can result in micronutrient deficiencies and height stunting^{3–6, 28–30}. For PI adoptees in the United States, most of their infancy is typically spent institutional care, with their deprivation ending with adoption into well-resourced families mostly before age three⁶. After adoption, PI youth experience rapid catch-up growth and early onset puberty³⁸, though typically fail to achieve similar height as their BR peers⁴. Previous work has shown children adopted from such orphanage-like institutions have a higher likelihood of experiencing stunted growth and developing CVD and T2DM^3–6. However, no studies have examined leg length differences because of early institutional care and how that may be associated with cardiometabolic health by adolescence and early adulthood.

The goals of the study were to examine: (1) the relations between T-LL and DXA-LL; (2) leg-length differences between PI and BR youth; and (3) the relation between DXA-LL and cardiometabolic markers, hear rate variability (HRV) metrics, and measures of body composition, most notably BMD. The sample size was too small to examine whether leg-length was more highly associated with cardiometabolic measures for the PI youth than BR youth.

METHODS

Subjects and Recruitment

The parent study involved a sample of 179 children and adolescents (102 females, 76 males). The current study sample included 140 children and adolescents, composed of 66 PI (44 female, 22 male) and 74 BR (39 female, 35 male). The difference in sample size was due to deciding to measure leg-length part way through data collection, and thus only then including T-LL along with coding the DXA data for leg length. Recruitment of the PI group was conducted using the Minnesota International Adoption Project Registry which consisted of a pool of families who had adopted children internationally. BR youth in the eligible age were randomly selected from the available registry pool. At screening, parents answered brief questions to place their child into early/middle/late pubertal development, and recruitment continued until we had enrolled a balance of males and females across ages and pubertal development. Despite efforts, the registry under-represents minoritized populations and lower income households. While minoritized populations and lower income households are currently under-represented in the registry, the resultant sample of BR is otherwise well matched for family composition, socioeconomic status, and parent race/ethnicity demographics, if not child race/ethnicity demographics. The inclusion criteria were children and adolescents between the ages of 12–21 years, and for adoptees, who had been adopted internationally into a Minnesotan family before the age of 5 years old (M=15.12 months). Exclusion criteria included a gestational age of less than 37 weeks, and the presence of congenital and/or chromosomal disorders (cerebral palsy, Turner syndrome, Downs syndrome, fetal alcohol syndrome (FAS), intellectual disabilities, Fragile X). This study was conducted in accordance with the Declaration of Helsinki of 1975 and approved by the Institutional Review Board of the University of Minnesota, Twin Cities (protocol STUDY00009969, approved on 6 September 2020). Written assent (participants under 18) and consent (participants 18 year and older and guardians of youth under 18 years) was obtained.

Measurement and Protocol

Participants visited the University of Minnesota Center of Pediatric Obesity Medicine for laboratory-based anthropometric and clinical assessments at Time 1 and returned two years later (Time 2) to repeat the measurements. For this study, anthropometric and clinical data were taken from a participant’s most recent visit that included both T- LL and DXA-LL.

Weight and height were obtained using a digital scale (Scale-Tronix 5002, Welch Allyn), and wall-mounted stadiometer (Model S100, Ayrton Corporation, Prior Lake, MN, United States of America), respectively. Standing height was measured by standing against the wall. Seated height was measured as participants sat on a 50-centimeter (cm) box pushed directly against the wall stadiometer. T-LL was determined by subtracting the seated height from standing height. Body mass index (BMI) was calculated as body weight in kilograms (kg) divided by height in meters squared (m²)³¹. Pubertal stage was determined through self-reporting of Tanner stage using previously explained methodologies³².

Body composition was measured in the supine position with legs internally rotated using standard DXA procedures (iDXA; General Electric Medical Systems, Madison, WI) with enCore^™ software (platform version, 16.0, General Electric Medical Systems). DXA provided measures of lean mass, fat mass, and BMD. DXA-LL was measured by creating a custom region of interest (ROI) by drawing a line vertically from the pubic symphysis to the calcaneus. BMD was also determined in the ROI as well as each leg. The Shapiro-Wilk test and density plots showed DXA-LL was not normally distributed (W = 0.98, p = 0.04). However, for the purpose of analysis, DXA-LL was not transformed.

Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were obtained once after ten minutes of rest in the supine position using a SphygmoCor XCEL system (AtCor Medical, Sydney, Australia). Pulse wave velocity (PWV) assessments were conducted using a SphygmoCor XCEL system (AtCor Medical, Sydney, Australia) to determine measures of arterial health. Following blood pressure measures, HRV was recorded in the supine position using a 3-lead electrocardiogram (ECG) (SphygmoCor MM3 AtCor Medical, Sydney, Australia) for five minutes.

Fasting blood samples were drawn by a trained phlebotomist for analysis of total cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), glucose, insulin, and triglycerides. Homeostatic model of insulin resistance (HOMA-IR) was calculated by taking fasting insulin (mU/L) times fasting glucose (mg/dL)/405²⁹. All blood measurements were analyzed at the University of Minnesota Fairview Hospital, a CLIA-certified laboratory.

Statistical Analysis

All statistical analyses were performed using RStudio (version 4.0.1) and SPSS (28.0.0.0 190). Normality was determined via density plot visualization and the Shapiro-Wilk Normality Test. We used Steiger’s test of correlational matrixes comparisons⁴⁸ and paired t-tests to examine the difference between DXA-LL and T-LL. Coefficient of variation percentage (CV%) was used to assessed reliability, while Bland–Altman plots evaluated agreement between leg length measurement methods. Simple correlations compared DXA-LL and T-LL values for significant differences. To increase the range of values for cardiometabolic analysis, PI and BR youth were combined. Hierarchical linear regression models examined associations between DXA-LL and cardiometabolic risk factors, arterial stiffness, HRV, and BMD, adjusting for age and sex.

RESULTS

Demographic characteristics of participants are provided in Table 1. The PI youth were slightly older than the BR youth and of a broader racial distribution.

Table 1.

Descriptive statistics between Previously Institutionalized and Birth Reared subjects. Categorical variables presented as frequency (% within columns).

	Total (n=140)	Previously Institutionalized (n=66)	Birth Reared (n=74)	p-value
Age (years)	17.2 ± 2.3	18.1 ± 2.4	16.4 ± 1.9	<0.01
Tanner Stage, n(%)				0.06
I	0(0%)	0 (0%)	0(0%)
II-IV	93 (77%)	49 (80%)	44 (73%)
V	28 (23%)	12 (20%)	16(27%)
Median Family Income	100–150K	100–150K	100–155K	0.68
Ethnicity/Race, n(%)				<0.01
Asian	33(19%)	31(38%)	2(6%)
Black	11(10%)	10(22%)	1(2%)
Indigenous to the Americas	12(7%)	11(19%)	1(0%)
More than one race	13(11%)	5(9%)	8(12%)
White	70(53%)	9(12%)	61(80%)

Traditional and DXA-derived Leg Length

T-LL and DXA-LL differed significantly (p <0.001), though the two measurement types were highly correlated (r (138) = 0.96, p<0.001). A Bland-Altman analysis showed agreement between the two measurement techniques (Bias = −1.23+0.165, limits of agreement = −4.47 – 2.01; Figure 1). DXA-LL was found to be an accurate and reliable measurement, with a CV% of 7.8 while T-LL had a CV% of 7.6.

Figure 1:

Bland-Altman plot of Traditional and DXA-derived Leg Length Measures

Difference determined via DXA-LL – T-LL.

Group Differences in Leg Length using DXA-LL

Controlling for age and sex, the PI youth had shorter legs than the BR youth (p <0.001). BR leg lengths ranged from 71.0 to 95.0 cm, while PI ranged from 66.5 to 84.3 cm. 24% percent of PI leg lengths fell below the shortest BR leg length, and 30% of BR leg lengths exceeded the longest PI leg length. Combining groups widened the overall range of leg lengths.

Relations between DXA-LL and Cardiometabolic Measures (see Table 2)

Table 2.

Regression of cardiometabolic variables of interest on DXA-leg length after controlling for age and sex

Independent Variables of Interest	β	T	p-value
Glucose Regulation
Glucose (mg/dL)	−0.24	−2.29	0.02 *
Insulin (mg/dL)	−0.178	−1.61	0.11
HOMA-IR	−0.19	−1.73	0.09
Lipid Measures
Total Cholesterol (mg/dL)	−0.29	−2.12	0.04 *
Triglyceride (mg/dL)	−0.23	−2.08	0.04 *
HDL (mg/dL)	−0.23	−2.15	0.03 *
LDL (mg/dL)	−0.17	−1.48	0.14
Cardiovascular health
Heart Rate (bpm)	0.35	0.33	0.74
RMSSD (ms)	−0.06	−0.59	0.56
LF:HF	0.001	0.01	0.99
PWV (m/s)	0.13	1.30	0.19
Blood Pressure
SBP (mmHg)	0.06	0.55	0.58
DBP (mmHg)	0.05	0.47	0.64
Bone Mineral Density
Leg BMD (g/cm2)	0.39	4.22	<0.01 *

^*Indicates p-value ≤0.05

Abbreviations: milligrams per deciliter (mg/dL), high density lipoprotein (HDL), low density lipoprotein (LDL), beats per minute (bpm), root mean squared standard deviation of N-N intervals (RMSSD), milliseconds (ms), low frequency: high frequency ratio, pulse wave velocity (PWV), meters per second (m/s), systolic blood pressure (SBP), millimeters of Mercury (mmHg), diastolic blood pressure (DBP), bone mineral density (BMD), grams per centimeter squared (g/cm²)

^**It should be noted that age was significantly and positively correlated with PWV (ms), HOMA-IR and Leg BMD

Total cholesterol, triglyceride, HDL, and glucose were significantly associated with DXA-LL (p = 0.02–0.04). There were no associations between any metrics of HRV or PWV and DXA-LL (p>0.05). SBP and DBP were not associated with DXA-LL (p = 0.58–0.64).

Relations Between DXA-LL and Bone Density

Lower body BMD was significantly associated with DXA-leg length (p<0.01; Table 2).

DISCUSSION

There was a significant difference between traditionally derived and DXA-LL in terms of the absolute length of leg. Although DXA-LL and T-LL differed by ~2 cm, similarly to prior work²⁴, the methods agreed closely, achieving our first aim. Our findings support prior work showing limb measurements can be reproduced from DXA scans²⁶.

Nevertheless, the leg length measures share predictive utility. Previous studies report associations between T-LL and cardiac health. In this study, we found that DXA-LL was also associated with cardiometabolic markers. In addition, it was associated with BMD as is T-LL. Notably there are some significant advantages to the use of DXA-LL measures. Specifically, there are studies that have collected DXA data to assess body composition but have not included T-LL that can now be re-analyzed to assess leg-length.

The second aim was to examine group differences between BR and PI youth in leg-length. The results showed BR subjects to have longer legs than their PI counterparts. This was expected as PI youth consistently exhibit shorter height-for-age^{4, 6}. Future studies should assess whether leg length or torso length drives group height differences. This would help focus attention on nutritional and social deprivation experiences in childhood as a function of early institutional care.

Our third aim was to examine associations between DXA measured leg length and markers of cardiometabolic health, HRV, arterial function, and BMD. Prior studies have used DXA-LL to determine the practicality of these measures²². Research has also examined adolescent limb length as a proxy for growth stunting^4–5. Similarly, BR/PI differences have also been reported for SBP and HDL^4–6. Unlike earlier reports⁵, we found no association between DXA-LL and HOMA-IR, possibly due to lifestyle differences between BR and PI subjects. Thus, the relationships among leg length, insulin sensitivity, and T2DM risk shown previously^16–17 were not replicated here. Still, DXA-LL was associated with cholesterol and HDL, echoing studies that connect normal growth with CVD risk in youth¹².

To our knowledge, no previous studies have associated HRV to leg length directly but have associated it with CVD and T2DM, separately^{28, 35–39}. Our study found no significant relationships between various metrics of HRV and DXA-LL. These findings were contrary to prior research^{28, 35–39} which found significant relationships with several HRV measures CVD/T2DM outcomes. These differences could be due to the specifics of study populations, with prior work focusing solely on white patients with Type 1 Diabetes³⁵ and adults. These studies have also included participants who experienced significant adversity throughout childhood, rather than adversity limited to the first years of life^35–39.

This was also the first study to compare measures of arterial stiffness and leg length. Prior studies report an inverse link between adult height and CVD³, but our findings do not align. A possible explanation is age confounding, as arterial stiffness increases with age; however, we included age as a covariate, and it remained strongly correlated with arterial stiffness (Table 2, footnote). Additionally, previous research found connections between psychological and socioeconomic stressors and increased risk for CVD,^{4–5, 29, 40} which may play a role in the lack of concordance in our findings.

To our knowledge, no previous studies have examined the relation between leg length and bone health measurements. Regional BMD has a strong association with DXA-leg length, suggesting that as children and adolescents continue to grow, the bone composition measurements continue to improve.

Strengths of this study include DXA-based bone composition measures, standardized leg-length assessment, and evaluation of vascular structure and function. A limitation is the difference in ethnicities between the PI and BR subjects. However, because the primary aim of this study was to determine difference in measurement techniques, each subject is compared to themselves and act as their own control. Another limitation is the reliance on HOMA-IR for insulin resistance: while reliable³³, it is not the gold standard, which may explain discrepancies with prior findings on insulin resistance and T2DM^4–6. Nevertheless, euglycemic clamp studies in large pediatric would be prohibitively expensive and time-consuming.

This study suggests that DXA-LL may be utilized accurately and reliably in place of traditional-leg length measurements. This study found minor associations between DXA-LL and cardiometabolic and bone composition markers in BR and PI youth. Future research is needed to clarify how these associations change with age and maturation.