Kidney function in severely obese adolescents undergoing bariatric surgery.

OBJECTIVE: Determine objective measures of kidney function and analyze factors associated with kidney dysfunction in severely obese adolescents undergoing weight loss surgery were described.
METHODS: Cross-sectional data from 242 adolescent participants in the Teen-Longitudinal Assessment of Bariatric Surgery (Teen-LABS) study before weight loss surgery were analyzed. Kidney status was assessed by measuring urine albumin creatinine ratio to determine microalbuminuria and by calculating serum cystatin C-based estimated glomerular filtration rate (eGFR) to assess kidney function.
RESULTS: Mean age and median body mass index (BMI) were 17.1 years and 50.5 kg/m(2) , respectively; 76% were females and 65% were non-Hispanic white race. Fourteen percent of the cohort had microalbuminuria, and 3% had macroalbuminuria; 3% had eGFR < 60 ml/min/1.73 m(2) , and 7.1% had eGFR > 150 ml/min/1.73 m(2) . In adjusted analyses, female gender and increasing ferritin levels were significantly associated with the presence of microalbuminuria/macroalbuminuria. Increasing BMI and homeostasis model assessment of insulin resistance values were significantly associated with lower eGFR.
CONCLUSIONS: A significant number of severely obese adolescents undergoing weight loss surgery have evidence of early kidney dysfunction. Longitudinal studies following weight loss surgery in these individuals are needed to determine whether these kidney abnormalities are reversible following weight loss therapy.

Introduction

The most recent data suggest that the overall prevalence of obesity among U.S. youth might have reached its peak.[1] In contrast, the prevalence of severe obesity, defined as an absolute BMI >35 kg/m2 or > 120th percent of the 95th percentile, is increasing and now affects 4–6% of U.S. children and adolescents.[2,3] Recently, the American Heart Association issued a scientific statement on associated risk factors and treatment approaches for severely obese children.[4] The statement specifically focused on immediate and long-term risks including cardiovascular disease, metabolic complications, obstructive sleep apnea, nonalcoholic liver disease, musculoskeletal and behavior problems. Notably, it did not address the issue of obesity associated kidney dysfunction.

Obesity, and particularly severe obesity, has important pathophysiologic consequences for the kidney. Obesity-associated focal segmental glomerular sclerosis has been well described in adolescents and adults.[5-7] It is also well-documented that obesity during adolescence is associated with a higher prevalence of chronic kidney disease (CKD) in adulthood.[8-10]

The Teen-Longitudinal Assessment of Bariatric Surgery (Teen-LABS) study aims to answer questions related to the safety and effectiveness of bariatric surgery in adolescence. As part of the baseline evaluation of this cohort of 242 patients, detailed clinical phenotyping included kidney function data, specifically urinary albumin and serum cystatin C-based glomerular filtration rate (eGFR). In this report, the baseline, preoperative kidney status of the Teen-LABS cohort was analyzed to describe the prevalence and factors associated with kidney abnormalities in this carefully studied population of severely obese adolescents.

Methods and procedures

Teen-LABS was designed as an ancillary study to the Longitudinal Assessment of Bariatric Surgery (LABS, NCT00465829). The standardized methodology developed for the LABS-2 study was modified for this adolescent cohort.[11] Briefly, 277 consecutive adolescents (age ≤19 years) undergoing bariatric surgery at each of five Teen-LABS centers between March 2007 and February 2012 were offered enrollment. However, 13 declined participation and 22 did not undergo the operation by the end of the enrollment period, leaving a final cohort of 242 subjects. The study protocol, assent/consent forms, data and safety monitoring plans were approved by the IRBs of each institution and by an independent Data and Safety Monitoring Board prior to study initiation. Pre-operative data were collected within 30 days of operation by trained study personnel. A Teen-LABS-trained clinical research coordinator or surgical investigator followed standardized procedures to determine the presence or absence of co-morbid conditions using medical records, physical exam, patient interview, and laboratory values. Detailed descriptions of study definitions, including hypertension, dyslipidemia, and diabetes have been described elsewhere.[11] All laboratory assays were performed at the central laboratory, the Northwest Lipid Research Laboratories at the University of Washington, Seattle, WA. Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated as (fasting glucose (mg/dL) × insulin (uU/ml)) / 405. Kidney function was assessed by calculating cystatin C-based eGFR, where eGFR = 77.24 × (Cystatin-C)−1.2623 according to the Larsson formula as recommended by the assay manufacturer (Dade Behring, Deerfield, Illinois).[12] Microalbuminuria was defined as having a urine Albumin to Creatinine Ratio (ACR) ≥30 mg/gm and <300 mg/gm; macroalbuminuria was defined as ACR ≥300 mg/gm. Abnormal kidney function was defined based on Kidney/Dialysis Outcome Quality Initiative (KDOQI) guidelines.[13]

Statistical Analyses

Descriptive statistics were calculated to summarize subject characteristics. Frequencies and percentages are reported for categorical measures. Means and standard deviations or medians and interquartile ranges were calculated for continuous variables. Scatterplots and Pearson correlation coefficients were also generated to describe select variables. Among the 16 variables evaluated in the statistical models, 1.6% of data values were missing. Most subjects (85.5%) had complete data, while 14.5% were missing at least one value. Missing values ranged from 1.7% (n=4) for lab values (transferrin, ferritin, serum albumin, HOMA-IR, hsCRP, LDL, HDL, triglycerides) to 6.6% (n=16) for the microalbuminuria outcome. Multivariate imputation by chained equations was performed to address these missing data. IVEware software (version 0.2) in SAS (version 9.3) was used to generate 20 imputed data sets for use in multivariable modeling analyses. A single, confirmed outlier insulin value (1314.2 uU/mL) was replaced with multiply imputed values for these analyses. To evaluate predictors of elevated ACR (i.e., ≥ 30 mg/gm), we calculated crude and adjusted prevalence ratios (PR) and 95% confidence intervals by fitting modified Poisson regression models with robust error estimates (SAS Proc GENMOD). Multivariable linear regression (SAS Proc GLM) models were used to evaluate predictors of cystatin C-based eGFR. SAS Proc MiAnalyze was used to generate all estimates from the multiply imputed datasets. All aforementioned descriptive and clinical variables were considered for inclusion in the final models. Multicollinearity diagnostics were reviewed to confirm appropriateness of these variables. All reported p-values are two-sided and considered statistically significant at ≤ 0.05.

Results

Demographic and clinical characteristics of the cohort are presented in Table 1. Of the 242 subjects, 75.6% were female and 64.9% were non-Hispanic white race. Mean age at surgery was 17.1 years, while the median body mass index (BMI) was 50.5 kg/m2 (range: 34.0 to 87.7). Seventy-four percent had dyslipidemia, 13.6% were diagnosed with diabetes, and 4.3% had a history of kidney stones. Forty-five percent were hypertensive. Forty-six (19%) subjects were taking anti-hypertensive medications at baseline. Twenty-four (9.9%) subjects were taking angiotensin converting enzyme inhibitors (ACEs) and 2 (0.8%) were taking angiotensin receptor blockers (ARBs) to control blood pressure.

Table 1

Demographic and Clinical Characteristics

	N=242
Age at surgery (years) – x̄ (SD)	17.1 (1.56)
Female – n (%)	183 (75.6%)
Race/Ethnicity – n (%)
Non-Hispanic White	157 (64.9%)
Non-Hispanic Black	54 (22.3%)
Hispanic	17 (7.0%)
Non-Hispanic Other races	14 (5.8%)
Body Mass Index (kg/m2) – median (Q1,Q3)	50.5 (45.2, 58.3)
Cystatin C-based eGFR (ml/min/1.73m−2) – x̄ (SD)*	107.6 (26.77)
Urine Albumin Creatinine Ratio – median (Q1,Q3)†	6.70 (3.93, 16.18)
Microalbuminuria – n (%)‡	32 (14.2%)
Macroalbuminuria – n (%)‡	7 (3.1%)
Hypertension – n (%)	109 (45.0%)
Diabetes – n (%)	33 (13.6%)
LDL (mg/dL) – x̄ (SD)*	93.0 (25.99)
HDL (mg/dL) – x̄ (SD)*	37.6 (9.05)
Triglycerides (mg/dL) – median (Q1,Q3)*	113.0 (82.0, 162.0)
Dyslipidemia – n (%)	180 (74.4%)
History of Kidney Stones – n (%)+	10 (4.3%)
Transferrin (mg/dL) – x̄ (SD)*	271.6 (38.06)
Ferritin (µg/L) – median (Q1,Q3)*	37.0 (23.0, 66.0)
Serum albumin (g/L) – x̄ (SD)*	4.1 (0.34)
HOMA-IR – median (Q1,Q3)*	5.91 (3.64, 9.84)
hsCRP (mg/dL) – median (Q1,Q3)*	0.63 (0.30, 1.17)
HbA1c – median (Q1,Q3)++	5.2 (5.0, 5.5)

n = 4 missing.

n = 12 missing.

n = 16 missing.

n = 10 missing.

n = 11 missing.

Thirty-nine (17.3%) subjects had elevated ACR, of which 32 (14.1%) had microalbuminuria (ACR range: 31–283 mg/gm) and 7 (3.1%) had macroalbuminuria (ACR range: 489–1758mg/gm). Based on crude analyses, increasing levels of ferritin, HOMA-IR, and HbA1c were significantly associated with the presence of elevated ACR (each p < 0.05; Table 2). The fully adjusted model demonstrated that increasing ferritin value (PR 1.07, 95% CI 1.03–1.10) and female gender (PR 2.34, 95% CI 1.02 – 5.34) were significantly associated with elevated ACR.

Table 2

Crude and adjusted associations with elevated albumin/creatinine ratio

	Elevated albumin/creatinine ratio
	Crude		Adjusted
	PR (95% CI)	p-value	PR (95% CI)	p-value
Age at surgery (years)	1.02 (0.84, 1.23)	0.87
Sex
Female	1.83 (0.81, 4.14)	0.15	2.34 (1.02,5.34)	0.04
Male	<Ref>		<Ref>
Race/ethnicity
NH White	<Ref>
NH Black	1.55 (0.77, 3.09)	0.22
Hispanic	2.28 (0.89, 5.82)	0.09
NH Other	1.84 (0.62, 5.51)	0.27
BMI (per 10 units)	1.12 (0.81, 1.54)	0.49
Hypertension
Yes	1.19 (0.65, 2.19)	0.58
No	<Ref>
Diabetes
Yes	1.70 (0.80, 3.60)	0.17
No	<Ref>
Dyslipidemia
Yes	1.19 (0.53, 2.64)	0.68
No	<Ref>
Kidney stones
Yes	0.74 (0.12, 4.42)	0.74
No	<Ref>
Transferrin (mg/dL)	1.00 (0.99, 1.01)	0.83
Ferritin (µg/L)/per 10units	1.06 (1.02, 1.10)	< 0.01	1.07 (1.03,1.10)	< 0.01
Serum Albumin (g/L)	0.50 (0.21, 1.20)	0.12
HOMA-IR	1.04 (1.00, 1.07)	0.04
hsCRP (mg/dL)	1.12 (0.99, 1.27)	0.07
HbA1c	1.32 (1.08, 1.61)	< 0.01

PR = Prevalence ratio. CI = Confidence interval.

Elevated albumin/ creatinine ratio (ACR): microalbuminuria with ACR ≥ 30 mg/gm and <300 mg/gm and macroalbuminuria with ACR ≥300 mg/gm.

The mean cystatin C-based eGFR was 107.6 ml/min/1.73m2. Most of the Teen-LABS participants (68.5%) had a normal eGFR (90–150 ml/min/1.73m2) However, 51 (21.4%) had an eGFR of 60–90 ml/min/1.73m2; 7 (3.0%) had an eGFR <60 ml/min/1.73m2; and 17 (7.1%) had an eGFR >150 ml/min/1.73m2. Percentile distribution of eGFR from the Teen-LABS cohort and from published NHANES data is shown in Figure 1. There were more subjects in low percentile and high percentile brackets in the Teen-LABS cohort than in the NHANES participants.[14]

Figure 1

The percentile distribution of cystatin C-based eGFR from the Teen-LABS cohort and from published NHANES data (reference 14)

Cystatin C was measured using the same methodology in both Teen-LABS and NHANES (Dade Behring, Deerfield, Illinois).[39]

Table 3 displays crude and adjusted associations between clinical features and eGFR. Crude results indicate female gender, BMI, transferrin, serum albumin, and HOMA-IR were associated with eGFR (each p < 0.05). However, after adjustment, increasing BMI and HOMA-IR values were the only factors that were significantly associated with decreased eGFR (each p < 0.05). Analyses indicated that for each 10-unit increase in BMI, eGFR decreased by nearly 8 mL/min/1.73m2, while for each 1-unit increase in HOMA-IR, eGFR decreased by only 0.57 mL/min/1.73m2. Figures 2 and 3 show the inverse relationships of BMI and HOMA-IR with eGFR.

Table 3

Crude and adjusted associations with cystatin c-based eGFR

	Cystatin c-based eGFR
	Crude		Adjusted
	β (95% CI)	p-value	β (95% CI)	p-value
Age at surgery (years)	−0.04 (−2.26, 2.18)	0.97
Sex
Female	8.80 (0.93, 16.66)	0.03	6.87 (−1.09, 14.82)	0.09
Male	<Ref>
Race/ethnicity
NH White	<Ref>
NH Black	−0.91 (−9.35, 7.52)	0.83
Hispanic	−2.16 (−15.64, 11.32)	0.75
NH Other	11.70 (−3.01, 26.41)	0.12
BMI (per 10 units)	−9.29 (−12.87, −5.70)	< 0.01	−7.90 (−11.60,−4.20)	< 0.01
Hypertension
Yes	−6.10 (−12.95, 0.74)	0.08
No	<Ref>
Diabetes
Yes	2.03 (−7.98, 12.03)	0.69
No	<Ref>
Dyslipidemia
Yes	−4.20 (−12.45, 4.04)	0.32
No	<Ref>
Kidney stones
Yes	−8.66 (−25.61, 8.29)	0.32
No	<Ref>
Transferrin (mg/dL)	0.13 (0.04, 0.21)	< 0.01	0.08 (−0.004, 0.173)	0.06
Ferritin (µg/L) (per 10 units)	−0.32 (−0.99, 0.34)	0.34
Serum Albumin (g/L)	10.55 (0.55, 20.56)	0.04
HOMA-IR	−0.72 (−1.26, −0.19)	< 0.01	−0.57 (−1.09,−0.06)	0.03
hsCRP (mg/dL)	−0.30 (−2.72, 2.12)	0.81
HbA1c	−0.20 (−3.88, 3.48)	0.91

CI = Confidence interval.

Figure 2

Scatterplot of BMI and eGFR, Teen-LABS subjects

Figure 3

Scatterplot of HOMA-IR and eGFR, Teen-LABS subjects

Discussion

This study represents the first attempt to characterize kidney status in a large cohort of severely obese adolescents. The most important findings were the high proportion of severely obese adolescents with microalbuminuria/macroalbuminuria and the significant association of increasing BMI with declining GFR.

The percent of Teen-LABS participants with microalbuminuria/macroalbuminuria was higher than anticipated (17.3%) based on previous reports. Previous studies analyzing NHANES data in obese and non-obese adolescents reported microalbuminuria prevalence values ranging from 8.9 to 10.4%.[14-16] Furthermore, Nguyen et al. reported that microalbuminuria was less prevalent in overweight (0.3%) compared with normal weight (8.7%) adolescents, likely due to orthostatic proteinuria in lean children.[16] In their study, despite overall lower prevalence of microalbuminuria among obese adolescents, there was a strong association between microalbuminuria and cardiovascular risk factors, suggesting that proteinuria may be predictive of future kidney and cardiovascular disease. In this context, our results indicate that in addition to a higher prevalence of microalbuminuria, the frequency of macroalbuminuria (3.1%) was more than twice as high than the 1.3% which was reported from the NHANES unselected pediatric population.[14] These findings are worrisome and suggest that severely obese adolescents are at increased risk for future CKD.

Our data show that obese subjects with elevated ferritin were more likely to have elevated ACR. Ferritin has well-known associations with many cardiovascular risk factors including inflammation, obesity, insulin resistance, and the metabolic syndrome.[17-19] Furthermore, Hsu et al. recently demonstrated an independent association of ferritin with microalbuminuria in a cohort of adults with diabetes.[20] The similar association observed in our cohort is consistent with the concept that obesity is a chronic, mild-inflammatory process with alterations of iron metabolism. Iron causes cellular damage by the formation of reactive oxygen species[21] and promotes oxidative stress and inflammation in both murine adipocytes and human aortic endothelial cells.[22,23] Glomerular endothelial injury mediated by iron-induced oxidative stress may therefore contribute to the development of microalbuminuria and proteinuria in severely obese patients. Multivariate analysis also revealed an association of elevated ACR with female gender. Epidemiologic studies of NHANES data have demonstrated increased prevalence of microalbuminuria among female adolescents and young adults.[15,16] However, Nguyen et al. reported that this gender discrepancy was not observed in obese adolescents.[16] The relevance of this finding in our cohort of severely obese females warrants additional investigation.

When we compared kidney function among the severely obese adolescents in this cohort and children participating in the NHANES survey[14], we observed a consistent trend of lower eGFR among the bottom 10% and higher eGFR among the upper quartile of GFR distribution in severely obese adolescents (Figure 1). The presence of a higher eGFR may represent an increased prevalence of hyperfiltration in morbidly obese adolescents. Hyperfiltration is a proposed mechanism of early glomerular injury occurring in a number of conditions, including diabetes, hypertension, and obesity.[24] Furthermore, hyperfiltration precedes the development of proteinuria in patients with diabetes and hypertension.[25,26] Prospective studies of obese adolescents will be needed to evaluate if obese adolescents with hyperfiltration represent a group with early glomerular injury at risk for the development of future microalbuminuria and CKD.

One of the strongest relationships we found in adjusted analyses was that higher BMI was associated with lower eGFR in the Teen-LABS participants. This is in contrast to some adult studies that have reported an association of higher BMI with hyperfiltration.[26-28] These studies, however, were conducted in research participants with a broad BMI distribution, including lean and overweight (mean BMI 25–28 kg/m2), whereas the current cohort of exclusively severely obese youth (median BMI of 50.5 kg/m2) is a uniquely homogeneous study group. It has been postulated that hyperfiltration associated with obesity precedes a subsequent decline in GFR[24]. It is possible that as obesity progresses from mild to severe in the relatively short timeframe of childhood, nephron damage occurs and the cumulative effect is a reduced GFR in those with highest BMI values at an earlier time-point. This study was not designed to address this question but the results are intriguing and warrant further inquiry. The association of increased BMI with decreased GFR in our study population is concerning and suggests an increased risk for CKD in adolescents with severe obesity. Our results are also supported by several studies that have identified obesity as an independent risk factor for progression to CKD.[29-31]

In our study, we found a weak but statistically significant negative association between eGFR and HOMA-IR, suggesting that those with the greatest insulin resistance had the greatest kidney dysfunction. Other studies examining this relationship have shown inconsistent results.[32-35] Nerpin et al[32] analyzed data from a community-based cohort of 1070 elderly men and demonstrated a direct association between insulin resistance and eGFR after adjusting for demographic, metabolic and cardiovascular risks. The study also reported that higher insulin resistance was associated with a higher risk of impaired kidney function during a 7-year follow up period. In contrast, in a large study of 145,865 adults, Park et al[35] found no significant differences in the HOMA-IR values among eGFR groups when groups were stratified by the metabolic syndrome components. The association of both BMI and abnormal glucose metabolism with impaired kidney function in this cohort further emphasizes the complexity of the metabolic dysfunction that should be expected in these severely obese individuals, and provides the rationale for further data collection and analysis of longitudinal trends in kidney function over time, since both BMI and insulin resistance are significantly impacted by bariatric surgery.

Our study has a few limitations that deserve consideration. We used serum cystatin-C to estimate kidney function and did not have a direct measurement of GFR. Various creatinine-based estimations of kidney function have demonstrated significant variability in performance among obese patients.[36] While some feel that cystatin C-based formulas may provide a more sensitive and accurate quantitation of kidney function irrespective of body composition[14,37], others argue that cystatin C might underestimate eGFR in both normal and obese populations.[38] Therefore we cannot exclude that the performance of cystatin C-based eGFR was affected by body composition. However, median eGFR values in our study were comparable to those obtained in healthy NHANES subjects[14], which suggests that there was not a systematic bias of cystatin-C based GFR estimation in the Teen-LABS cohort. It is important to note that cystatin C determination in our study and in NHANES utilized the same methodology.[39] Another limitation was the evaluation of proteinuria using random samples (rather than first morning void), which precluded the identification of patients with orthostatic proteinuria. However, previous studies using a random sample in children showed lower prevalence of microalbuminuria.[16,40] Therefore we suspect this had little effect on the study results. It is also possible that we did not fully control for all confounding variables affecting the measured outcomes (e.g. birth weight, duration of obesity, maternal diabetes) since Teen-LABS study does not collect this information.

Conclusion

A significant number of severely obese adolescents undergoing bariatric surgery demonstrate evidence of significant kidney abnormalities. This comorbidity has previously not been recognized in this population. Even more concerning, for those with impaired kidney function, progressive deterioration is predicted in the absence of effective weight management.[10] Longitudinal analyses of kidney function are necessary to determine if kidney abnormalities can be reversed in obese adolescents following weight loss.