Adiposity, Sex, and Cardiovascular Disease Risk in Children With CKD: A Longitudinal Study of Youth Enrolled in the Chronic Kidney Disease in Children (CKiD) Study

Abstract

Rationale & Objective:

Traditional and nontraditional cardiovascular disease (CVD) risk factors are highly prevalent in children with chronic kidney disease (CKD). We examined the longitudinal association of adiposity with cardiac damage among children with CKD and explored whether this association was modified by sex.

Study Design:

Prospective cohort study.

Setting & Participants:

Children with mild-to-moderate CKD enrolled in the Chronic Kidney Disease in Children (CKiD) study at 49 pediatric nephrology centers across North America.

Exposure:

Age- and sex-specific body mass index (BMI) z-score.

Outcome:

Age- and sex-specific left ventricular mass index (LVMI) z-score and left ventricular hypertrophy (LVH).

Analytical Approach:

Longitudinal analyses using mixed effects models to estimate sex-specific associations of BMI z-scores with LVMI z-score and with LVH, accounting for repeated measurements over time.

Results:

Among 725 children with 2829 person-years of follow-up, median age was 11.0 years and median estimated glomerular filtration rate was 52.6 ml/min/1.73m². Nearly one-third of both boys and girls were overweight or obese, median LVMI z-score was 0.18 (IQR: −0.67, 1.08), and 11% had LVH. Greater BMI z-scores were independently associated with greater LVMI z-scores and greater odds of LVH. For each 1-unit higher BMI z-score, LVMI z-score was 0.24 (95% CI: 0.17, 0.31) SD higher in boys and 0.38 (95% CI: 0.29, 0.47) SD higher in girls [p_interaction=0.01]. For each 1-unit higher BMI z-score, the odds of LVH was 1.5-fold (95% CI: 1.1, 2.1) higher in boys and 3.1-fold (95% CI: 1.8, 4.4) higher in girls [p_interaction= 0.005].

Limitations:

Not all children had repeated measurements. LVH is a surrogate and not a hard cardiac outcome. The observational design limits causal inference.

Conclusions:

In children, adiposity is independently associated with the markers of cardiac damage, LVMI z-score, and LVH. This association is stronger among girls than boys. Pediatric overweight and obesity may therefore have a substantial impact on CV risk among children with CKD.

Introduction

Chronic kidney disease (CKD) remains a leading risk factor for cardiovascular disease (CVD) among children and young adults¹. Both traditional and nontraditional CVD risk factors are highly prevalent among youth with CKD, leading to a significant burden of intermediate cardiovascular (CV) outcomes such as left ventricular hypertrophy (LVH), increased arterial stiffness, and increased carotid intima media thickness^2–4. These pathological findings precede hard CV outcomes in adults and, when present in children with CKD, likely contribute to their elevated rates of CV morbidity and mortality^{5, 6}. Determining the modifiable CVD risk factors most associated with these intermediate outcomes in youth with CKD could therefore significantly aid prevention and treatment efforts for these at-risk children.

While overweight and obesity are also associated with increasing numbers of traditional CVD risk factors among children with mild-to-moderate CKD⁷, the impact of adiposity on target organ damage among children with CKD has not been well described. Previous studies within the Chronic Kidney Disease in Children (CKiD) population have described the association of various assessments of adiposity with left ventricular mass index (LVMI)^{8, 9}, but the independent association of adiposity as determined by the clinically used marker of BMI on LVH, a clinically defined form of target organ damage, has not been previously described in this population. Further, due to the expected change in LVM over time by age and height^{10, 11}, it can be challenging to interpret associations of risk factors with LVMI without accounting for these expected normative variations in LVM. We therefore sought to quantify the longitudinal association of adiposity, as determined by age-sex-specific BMI z-score, with both LVMI z-score and LVH among children with CKD. In addition, because of the known differential increase in BPs among adolescent boys compared to girls related to body size¹² and because earlier work within the CKiD study described a greater prevalence of LVH among girls compared to boys¹³ we a priori aimed to determine if the independent association of adiposity with target organ damage was modified by sex.

Methods

Study Population.

Between 2005 and 2014, a total of 891 children with mild to moderate CKD were enrolled in the CKiD study, a multicenter, prospective cohort study conducted at 49 pediatric nephrology centers across North America. The study design and conduct were approved by the institutional review boards of each participating center as well as an observational study monitoring board appointed by the National Institute of Diabetes and Digestive and Kidney Diseases. Informed consent was obtained from each participating family. After an initial baseline visit, study participants attended annual follow up visits with echocardiography obtained biennially. Our analysis included participants with echocardiography data and complete data on the exposure of interest and all covariates for at least one CKiD visit.

Echocardiography.

At the first annual follow-up visit and at every other subsequent follow-up visit (i.e., once every two years), participants underwent standardized protocol echocardiography at their local sites. The recorded images were read centrally by a study cardiologist. Triplicate measurements were taken and averaged to obtain left ventricular mass (LVM) according to American Society of Echocardiography criteria^{14, 15} using 2D M-mode.

Primary Outcomes.

The LVMI was calculated as LVM/height^2.7, where LVM is in grams and height is in meters^{14, 16}; then age-sex specific LVMI z-scores based on published norms¹⁰ were calculated and used in analyses. Left ventricular hypertrophy (LVH) was defined as LVMI ≥ age-sex specific 95^th percentile¹⁰.

Primary Exposure.

Weight and height were measured twice at each annual visit, with the values averaged. These average weight and height measurements were used to calculate body mass index (BMI) and to calculate age- and sex-specific BMI z-scores¹⁷ for those ≤ 20 years of age. The age- and sex-specific BMI z-score from the current visit (visit with echocardiography) and previous visit (visit immediately prior to echocardiography) were averaged. This average BMI z-score, time-updated at each visit, was used in all analyses as the primary exposure.

Covariates.

The following variables were included in the multivariable longitudinal analyses: years from baseline visit to current visit with echocardiography (time-updated at each visit); baseline age; years with CKD at the time of the baseline visit; sex; race (African American vs. non-African American); CKD etiology (glomerular disease vs. non-glomerular disease); average age-, sex-, and height-specific systolic (SBP) and diastolic blood pressure (DBP) z-scores from current and previous visit (measured via manual auscultation and time-updated at each visit); average serum calcium-phosphorous product from current and previous visit (time-updated at each visit); and average estimated glomerular filtration rate (GFR) from current and previous visit¹⁸ (time-updated at each visit).

Statistical Analyses.

We conducted two longitudinal analyses using mixed effects models to determine the association of LVMI z-score and LVH with age- and sex- specific BMI z-scores over time, while taking into account repeated measurements over follow-up. A linear mixed effects model was used for continuous LVMI z-scores and a logistic mixed effects model was used for the dichotomous outcome of LVH. The linear model included a random intercept and a correlated random coefficient for years from baseline, while the logistic model had only a random intercept. These mixed effects models were used to estimate the relationship between current LVMI z-score and current LVH with the average BMI z-score from the current visit (visit with LVMI z-score and LVH) and previous visit (visit immediately prior to LVMI z-score and LVH). Both regression models were adjusted for years from baseline visit to current visit with echocardiography and the other previously defined covariates either at baseline or time-updated average measures from the current and previous visits as noted above. Because we hypothesized that sex would modify the association of adiposity on cardiac structure (LVMI z-score and LVH), an interaction term of sex with BMI z-score was included in each model.

Finally, to allow for informative drop out, i.e., initiation of kidney replacement therapy (KRT), we conducted a joint analysis of the LVMI z-score and time to KRT as a sensitivity analysis¹⁹. In addition to the linear mixed model described previously, the joint analysis model included a Weibull survival submodel. The covariates for the survival submodel were CKD etiology (glomerular vs non-glomerular), urinary protein-creatinine ratio (UPCR), and baseline eGFR. The survival submodel shared the random intercept and slope of the mixed model, which had their own regression coefficients in the survival submodel. The stjm command in Stata 15 was used for the joint analysis model. A two- sided p-value <0.05 was considered statistically significant.

Results

Our analysis included 725 children who contributed a total of 1483 study visits over 2829 person-years (median follow-up from study entry was 3.3 [interquartile range: 1.2, 5.2] years). Specifically, 286 (40%) of the 725 children contributed one visit, 228 (31%) contributed two visits, 127 (18%) contributed three visits, 60 (8%) contributed four visits, and 24 (3%) contributed five visits.

Table 1 describes the clinical and demographic characteristics of the 725 children (38% female, 21% African American) with mild to moderate CKD who were included in the analyses. At the first visit with echocardiography and BMI measured concurrently, the median age was 11.0 years, the median length of time with CKD was 8.0 years, and the median eGFR was 52.6 (IQR: 39.5, 67.2) ml/min/1.73m². Ten percent of the study population had an average BP in the hypertensive range. Nearly one third of the study population was overweight or obese: 13% (n=93) had a BMI in the overweight category (defined as a BMI in the 85^th-<95^th age- and sex-specific percentile range) and 16% (n=119) had a BMI in the obese category (BMI ≥ 95^th age- and sex-specific percentile). There was no significant difference in the prevalence of overweight or obesity by sex (32% of females were overweight/obese compared to 28% of males, p=0.2). Median LVMI was 30.5 (IQR: 25.6, 36.8) g/m^2.7 and median LVMI z-score was 0.18 (IQR: −0.67, 1.08). Eleven percent of the study population had LVH.

Table 1.

Characteristics of the 725 Participants of the CKiD Study at the First Visit with Echocardiography and BMI Measured Concurrently.

Characteristic	Value
Age- and sex-specific BMI z-scorea	0.44 (−0.33, 1.23)
Overweight: BMI in 85th–<95th percentile rangea	13%
Obese: BMIa ≥ 95th percentile	16%
Age, years	11.0 (7.5, 14.7)
CKD Duration, years	8.0 (3.8, 12.3)
Female sex	38%
African American race	21%
Glomerular CKD diagnosis	29%
Age-, sex-, and height-specific SBP z-scorea	0.25 (−0.42, 0.85)
Age-, sex-, and height-specific DBP z-scorea	0.40 (−0.14, 0.94)
Hypertensiona,b	10%
Calcium-Phosphorus producta, mg2/dL2	42.8 (38.5, 47.0)
Estimated glomerular filtration ratea,c, ml/min/1.73m2	52.6 (39.5, 67.2)
LVMI, g/m2.7	30.5 (25.6, 36.8)
Age- and sex-specific LVMI z-score	0.18 (−0.67, 1.08)
Left Ventricular Hypertrophy (LVMI ≥ 95th percentile)	11%
Initiated KRT	31%

Values for continuous variables given as median (interquartile range).

^aAverage of current visit and previous visits.

^bSystolic blood pressure (SBP) or diastolic blood pressure (DBP) ≥ 95^th percentile for children < 13 years old; SBP > 130 or DBP ≥ 80 for children > 13 years old.

^cBased on estimating equations from Schwartz et al.¹⁸

Figure 1 shows the relationship of LVMI z-score and age-sex-specific BMI z-score by sex across 1483 study visits, with data contributed by 725 participants. Each standard deviation (SD) increase in BMI was associated with a significantly greater SD increase in LVMI for females versus males (0.15, p=0.01). Specifically, each SD increase in BMI among females was associated with a 0.40 SD increase in LVMI compared to the 0.25 SD increase in LVMI for a SD increase in BMI in males.

Figure 1.

Scatter plot of age-sex specific Left Ventricular Mass Index (LVMI) z-score and age-sex specific BMI z-score at 1483 visits contributed by 725 participants of the Chronic Kidney Disease in Children (CKiD) study. The average BMI z-score of the current visit and previous visit are shown. Regression lines are shown for males and females and were obtained from a linear mixed model with LVMI z-score as the dependent variable and male sex, BMI z-score, and male sex x BMI z-score interaction term as the independent variables.

Figure 2 presents the prevalence of LVH across 1483 study visits by sex in a total of 725 participants for three BMI categories: normal (BMI < 85^th age- and sex-specific percentile), overweight, and obese. In girls, the prevalence of LVH was 8.4% among those with a normal BMI, 16.0% among those who were overweight, and 33.7% among those who were obese. Among boys, the relationship between BMI and LVH was not nearly as pronounced, with the prevalence in each category of BMI being close to 7.0%, which was similar to the overall prevalence of LVH across all visits among boys.

Figure 2.

Prevalence of LVH by sex and category of Body Mass Index at 1483 visits contributed by 725 participants of the Chronic Kidney Disease in Children (CKiD) study. LVH was defined as LVMI ≥ age-sex specific 95^th percentile. The average BMI of the current visit and previous visit were used.

Each additional year from baseline was associated with both a decrease in LVMI z-score (−0.10 [95% CI: −0.12, −0.08]; unadjusted analysis) and a lower odds of LVH (−20% [95% CI: −11%, −30%] per year).

The results of the multivariable longitudinal analyses are presented in Table 2. The association of BMI z-score with LVMI z-score was significantly different by sex (p=0.01). Specifically, each 1-SD greater BMI z-score was associated with a 0.38 (95% CI: 0.29, 0.47) SD greater LVMI among girls, which was significantly greater (p=0.02) than the 0.24 (95% CI: 0.17, 0.31) SD greater LVMI among boys. The association of BMI z-score with LVH was also significantly different (p=0.005) in girls and boys. Specifically, for each 1-SD greater BMI z-score, girls had a 3.1 (95% CI: 1.8, 4.4) times greater odds of LVH, while boys had a 1.5 (95% CI: 1.1, 2.1) times greater odds of LVH.

Table 2.

Longitudinal Association of Body Mass Index with measures of Cardiac Structure among 725 Children with Mild-Moderate Chronic Kidney Disease at a total of 1,483 visits.

	LVMI z-score		LVH*
Exposure	Estimate (95% CI)	P	Odds ratio 95% CI	P
Age-Sex specific BMI z-scorea, per 1 SD increase for Males	0.24 (0.17, 0.31)	p<0.001	1.5 (1.1, 2.1)	p=0.01
Age-Sex specific BMI z-scorea, per 1 SD increase for Females	0.38 (0.29, 0.47)	p<0.001	3.1 (1.8, 4.4)	p<0.001
Covariates
Years from baseline, per 1 year increase	−0.08 (−0.11, −0.06)	p<0.001	0.8 (0.7, 0.9)	p=0.001
Baseline Age, per 1 year increase	−0.002 (−0.02, 0.02)	p=0.8	1.0 (0.9, 1.1)	p=0.8
Baseline CKD Duration, per 5 year increase	0.001 (−0.0004, 0.002)	p=0.2	1.002 (0.998, 1.007)	p=0.3
Female (vs. Male)	0.14 (0.004, 0.28)	p=0.04	1.4 (0.7, 2.8)	p=0.3
African American (vs. non-African American)	0.29 (0.12, 0.45)	p=0.001	2.2 (1.2, 4.2)	p=0.01
Glomerular diagnosis (vs. non-Glomerular)	0.16 (0.001, 0.33)	p=0.05	1.8 (0.9, 3.4)	p=0.09
Age-Sex-Height specific SBP z-scorea, per 1 SD increase	0.18 (0.11, 0.25)	p<0.001	2.0 (1.4, 2.8)	p<0.001
Age-Sex-Height specific DBP z-scorea, per 1 SD increase	0.03 (−0.05, 0.11)	p=0.5	1.1 (0.7, 1.5)	p=0.8
log(Calcium* Phosphorous)a, mg2/dL2	0.38 (−0.02, 0.79)	p=0.06	2.1 (0.3, 13.2)	p=0.4
Estimated glomerular filtration ratea, b, ml/min/1.73m2, per 10 unit decrease	0.12 (0.08, 0.15)	p<0.001	1.5 (1.3, 1.8)	p<0.001

^*LVMI ≥ 95th percentile

^aAverage of current visit and previous visits.

^bBased on estimating equations from Schwartz GJ et al., Kidney Int 2012; 82:445–453.

BMI – body mass index; CKD – chronic kidney disease; DBP – diastolic blood pressure; SBP – systolic blood pressure; LVH – left ventricular hypertrophy; LVMI – left ventricular mass index; SD – standard deviation

For the LVMI z-score outcome, we conducted a joint analysis with time to KRT as a sensitivity analysis. The results of this joint model did not differ from the longitudinal model that did not take into account informative censoring. Therefore, we report the simpler longitudinal model in Table 2 and the joint analysis model in the supplementary material (Table S1). We did not conduct the joint analysis for the categorical variable LVH.

Discussion

In this longitudinal analysis of children with mild-to-moderate CKD, we demonstrate that adiposity is more greatly associated with cardiac target organ damage among girls than boys, independent of other known traditional and non-traditional co-morbid CVD risk factors. Specifically, compared to boys, we found that girls had a greater increase in LVMI z-score and a greater odds of LVH for the same degree of BMI z-score increase.

Adiposity is a known CVD risk factor²⁰, and is independently associated with LVH among children with non-CKD-related hypertension²¹ and among children without hypertension or CKD^{22, 23}. Among children with primary hypertension, adiposity is more strongly associated with both LVH and LVH regression than blood pressure (BP)^{22, 24–26}. In a pre-post study of morbidly obese adolescents who underwent bariatric surgery, an LVMI decrease of 12 g/m^2.7 was observed.²⁷

Knowing that 1) adiposity is as prevalent among children with CKD as within the general US population^{28, 29}; 2) CVD risk factors track from childhood to adulthood^{30, 31}, with childhood BMI a strong predictor of adult LVH³²; 3) LVH is a known risk factor for cardiac arrhythmias, diastolic dysfunction, and heart failure³³; 4) co-morbid obesity not only impairs the ability to effectively treat hypertensive LVH, but can also lead to a greater likelihood of CV events^34–36; and 5) CVD statistics in adults also differ by sex³⁷, our findings suggest both an urgent need for enhanced obesity prevention and treatment in pediatric CKD and a potential role for targeted interventions.

Adiposity contributes to LVH both directly and indirectly via its effect on BP, and sex-differences in BP trends among children and adults has been well-described. Rosner et al. has shown that beginning in adolescence, the observed age related increase in BP is significantly greater in boys compared to girls, regardless of race¹². Adult studies have shown that female sex has a protective effect on hypertension prevalence and CVD morbidity and mortality until menopause. After menopause, female sex is associated with a greater risk of hypertension and a similar risk for CV events as males³⁸. These peri-pubertal and peri-menopausal differences suggest a hormonal cause for CVD risk.

In fact, the existing literature does support a causal role of endogenous sex-hormones on BP and cardiac target organ damage. Estrogen has been shown to have antiproliferative effects on cardiac fibroblasts and vascular smooth muscle cells^39–41 and to inhibit the development of LVH⁴². Testosterone, on the other hand, increases vascular smooth muscle cell proliferation⁴³ and promotes LVH^{42, 44}. Notably, these hormonal levels are affected by obesity: in males, obesity lowers free testosterone (effectively decreasing its pro-CVD promoting effects⁴⁵) and in pre-menopausal women, obesity decreases estrogen levels (effectively decreasing its CVD-protective effects⁴⁶).

While compelling, other more traditional BP-related hormones that are likewise influenced by degree of adiposity may also explain our findings. Under normal circumstances, adipose cells secrete a multitude of adipokines and hormones, including angiotensinogen⁴⁷. With obesity these hormones are upregulated, and with progressive obesity there is increasing inflammation and progressive loss of metabolic control leading to a dysregulated state⁴⁸. When adipose tissue becomes dysfunctional, the RAAS system is activated⁴⁹ leading to an increase in angiotensin II and aldosterone. Animal models suggest that these BP-related hormones may exert sex-dependent effects on CVD risk factors. Specifically, angiotensin II is not only related to increased cardiac hypertrophy in male rats, but angiotensin-converting enzyme levels are also greater in male mice compared to their female counterparts⁵⁰. When one considers that long-term angiotensin II blockade has been shown to prevent cardiac remodeling in these same animal models⁵¹, one must consider the potential for a sex-based targeted therapeutic approach to CVD, particularly when there is co-morbid obesity.

The differential impact of obesity on LVH by sex that we describe in this work has also been described in the adult literature. Over the last several decades, it has been increasingly evident that women have different risks for CVD than men and present with less “classic” symptoms when experiencing a CV event. Adult women, particularly obese adult women who are hypertensive, are also more likely to have LVH³⁴. It is thought that the increased prevalence of LVH among obese adult women is related to an increase in overall fat mass (vs. lean mass) compared to men of the same BMI⁵², and to an increase in visceral adiposity versus subcutaneous adiposity compared to men^{52, 53}. In fact, in a recent study of Italian adults, while the overall CVD risk profiles between men and women with LVH were not different, women with obesity had a much greater likelihood of having LVH than men with obesity (comparing obese to non-obese individuals, ORs for LVH were 4.23 [95% CI: 3.60, 4.97] and 2.87 [95% CI: 2.49, 3.29] in women and men, respectively)⁵⁴. Perhaps the most striking finding from the Italian study was that the “protective” effect of female sex on CV events (HR, 0.65; 95% CI: 0.44, 0.96; p=0.03) disappeared when LVH was present (HR, 0.94; 95% Cl: 0.69, 1.30; p = 0.7).

Interestingly, despite the work in this area among adults there is not much published regarding the potential differential impact of obesity on LVH by sex among children. In a school-based study of 676 healthy Korean adolescents 12–16 years of age⁵⁵, the authors found that adipose mass as assessed by bioelectric impedance analysis correlated with LVM in females whereas fat free mass associated with LVM in males, consistent with findings among adults in the Strong Heart Study⁵².

Our study suggests that pediatric overweight and obesity may have a substantial impact on CV risk among children with CKD. The impact of intervening to improve and normalize BMI among youth, particularly females, on future risk of CVD deserves study. Pediatric nephrologists may benefit from partnering with renal dieticians and other members of their multidisciplinary teams to target weight loss/maintenance more aggressively when presented with seemingly resistant CKD-related pediatric LVH.

There are several limitations to our study which deserve mention. In our study, many children had repeated measurements; however 40% of participants only contributed data from one visit, raising the possibility for bias related to differential loss to follow up. Fortunately, sensitivity analysis allowing for informative drop out suggests this is likely minimal. It is also worth noting that while LVH is a clinical biomarker for CVD risk, itis a surrogate and not a hard cardiac outcome. In addition, our study was observational which does not allow us to infer causality.

We also note multiple strengths. The large population of children studied with mild-to-moderate CKD had both a variety of etiologies for their kidney disease and a wide range of BMIs. These children were followed longitudinally, in a national cohort study, with standardized anthropometrics, and BP and echocardiographic measurements. As a pediatric population, those studied were free of many of the usual co-morbidities that confound studies in adults (smoking, alcohol use, and diabetes as examples) and did not have an extended period of time exposed to other traditional CV risk factors (including dyslipidemia, hypertension, and kidney failure). Despite the lower burden of CVD risk compared to adults, we present adjusted results to describe the independent association of adiposity on cardiac target organ damage, specifically examining the effect of sex on this relationship, which has not been previously done.

Future studies that focus on enhanced risk stratification strategies and targeted treatment approaches among this at-risk population are needed. Determining the effect of tailored interventions on children with CKD and co-morbid obesity will provide more insight into these sex-related differences and establish the effect of weight loss on CV risk.

Supplementary Material

1

Table S1: Longitudinal association of BMI and LVMI z-scores among 714 children with mild-to-moderate CKD at a total of 1,426 visits, accounting for informative censoring.