Candidate Biomarkers as Predictors of Future Kidney Disease and Cardiovascular Dysfunction in Adolescents with Type 2 Diabetes

Abstract

Aims:

Evaluate changes in circulating biomarkers as predictors of kidney disease, and cardiac/vascular dysfunction in participants from the Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study.

Methods:

Candidate biomarkers were assessed annually in 507 participants over a mean follow-up of 6.9±2.4 years. Moderate albuminuria was defined as urine albumin-to-creatinine ratio ≥30 mg/g and hyperfiltration as eGFR ≥135 mL/min/1.73 m² at two consecutive visits. Echocardiography (n=256) and pulse wave velocity (n=193) were evaluated twice, 5 years apart. Adjusted Cox proportional hazard models and logistic regression models were used to examine associations between biomarkers and outcomes.

Results:

At baseline, 35.7% were male, with a mean age 13.9 years, diabetes duration 7.8 months, and HbA1c 6.0%. Higher concentrations of E-selectin and proinsulin were associated with incident moderate albuminuria and hyperfiltration. Higher concentrations of FGF-23 were associated with lower risk of hyperfiltration and negatively correlated with eGFR. No candidate biomarkers predicted a decline in cardiac or vascular function.

Conclusions:

Circulating biomarkers of endothelial dysfunction and markers of β-cell dysfunction and insulin sensitivity could be used in a more personalized risk assessment of kidney disease in youth-onset type 2 diabetes. However, biomarkers studied have limited value in predicting cardiac dysfunction or vascular stiffness.

Graphical Abstract

1. Introduction

Participants in the Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study demonstrated a high burden of cardiovascular disease (CVD) risk factors and developed early-onset kidney disease and adverse cardiac and vascular structure and function.¹ Biomarkers beyond conventional risk factors that may predict the progression of renal and cardiovascular abnormalities are lacking in youth with type 2 diabetes.

The TODAY study has shown a few associations of candidate biomarkers with kidney disease, vascular stiffness (carotid-femoral pulse wave velocity), and echocardiographic outcome measures in cross-sectional analyses, but for most variables there was a lack of association.² We observed high visit-to-visit variability for troponin, brain natriuretic peptide (BNP), and tumor necrosis factor-alpha (TNF-α) with a high frequency of values outside the normal range. By the end of the study, TNF-α concentrations showed an association with left ventricular (LV) mass, blood pressure, and HbA1c.² Inflammatory markers previously evaluated in TODAY included vascular cell-adhesion molecule-1(VCAM-1), intercellular adhesion molecule-1 (ICAM-1), and endothelial-leukocyte adhesion molecule (E-selectin) and were found to be related to hyperglycemia and moderate albuminuria over 3 years in the study.³ Other biomarkers of CVD related to insulin resistance and hyperglycemia and found to predict CVD events and kidney disease in adult studies were also measured longitudinally in TODAY. They include adiponectin, proinsulin, copeptin, and fibroblast growth factor-23 (FGF-23).^4-7 Copeptin, co-secreted with arginine vasopressin and more stable in the circulation, is associated with decline in kidney function independent of traditional risk factors (body mass index [BMI] and HbA1c).^8,9 FGF-23, a bone-derived hormone and a regulator of phosphate and vitamin D homeostasis, is considered one of the most powerful predictors of adverse outcomes in patients with chronic kidney disease (CKD). Some but not all studies have shown a positive correlation between FGF-23 concentrations with arterial stiffness, calcification^10,11, and cardiovascular event rates¹² suggesting FGF23 may serve as a cardiovascular and renal biomarker.

In this analysis, we capitalized on the availability of these candidate biomarkers (troponin, BNP, TNF-α, VCAM-1, ICAM-1, E-selectin, FGF-23, copeptin, proinsulin, adiponectin, and insulin-like growth factor binding protein-1 [IGFBP-1]) shown to be predictors of micro- and macro-vascular disease in adults^4,7,8,13-16 and measured several times during the course of the TODAY study.^2,3 We evaluated these multiple biomarkers and changes thereof during the TODAY trial and its observational follow-up study (TODAY2) as predictors of the progression of kidney disease, and cardiac and vascular dysfunction, in youth-onset type 2 diabetes.

2. Materials and Methods

The TODAY clinical trial (2004-2011) was designed to evaluate the effects of three treatment arms (metformin alone, metformin + rosiglitazone, and metformin + lifestyle) on time to failure to maintain glycemic control (HbA1c ≥8% for 6 months or inability to wean from temporary insulin started for acute metabolic decompensation). Detailed methods have been previously published.¹⁷ Briefly, 699 participants with recent- onset type 2 diabetes, ages 10-17 years, were enrolled among 15 participating diabetes centers across the United States. Study eligibility criteria included negative diabetes autoantibodies (glutamic acid decarboxylase-65 and tyrosine phosphatase), measurable C-peptide, BMI ≥85^th percentile, and <2 years’ duration of type 2 diabetes. Participants were followed for an average of 3.86 years during the treatment phase. The TODAY follow-up study (TODAY2: 2011-2020) was designed to provide longitudinal follow-up data on the original TODAY cohort. Type 2 diabetes treatment during the TODAY follow-up study was not randomized, but was guided by existing best practice recommendations. Of the 699 TODAY participants enrolled in the cohort, 22 participants with monogenic diabetes mutations were excluded for this analysis. All participants provided informed consent and minor children confirmed assent according to local guidelines.

2.1. Evaluations and Candidate Biomarkers

All study visits included a detailed medical history and a physical examination with the same procedures used to measure height, weight, waist circumference, and blood pressure. Fasting laboratory studies and HbA1c were also obtained at every visit and analyzed centrally at the TODAY Central Biochemistry Laboratory (Northwest Lipid Metabolism and Diabetes Research Laboratories, University of Washington, Seattle WA). Blood samples were collected yearly in EDTA tubes, centrifuged, frozen immediately upon sample processing, and stored in 24/7 monitored −80°C freezers. Analyses for biomarkers were performed immediately after samples were thawed. Serum BNP and plasma copeptin concentrations were measured by ELISA (RayBioTech Inc, Peachtree Corners, GA, and Phoenix Pharmaceuticals Inc, Burlingame, CA, respectively), with intra-assay and inter-assay coefficient of variations (CV) of 10% and 12%, and 10% and 10%, respectively. Analysis of FGF-23 concentrations was performed using a two-site ELISA based on a biotinylated monoclonal antibody for capture and an affinity-purified polyclonal antibody conjugated to horseradish peroxidase for detection; the assay sensitivity was 1 pg/mL and the intra- and inter-assay CVs were consistently <5% and <10%, respectively (Immutopics Quidel Corp., San Diego, CA). Troponin, TNF-α, E-selectin, IGFBP-1, ICAM-1, and VCAM-1 assays were performed using a Multiplex protein arrays system using magnetic beads (R&D Systems, Minneapolis, MN), with intra-assay and inter-assay CVs of 6.5% and 11%, 7.2% and 12.5%, 4.0% and 8.0%, 4.5% and 10.8%, 4.1% and 13.9%, and 3.7% and 11.3%, respectively. Proinsulin was measured by radioimmunoassay (Millipore Inc, Burlington MA), with assay sensitivity of 2 pM and inter-assay CVs for high and low proinsulin level samples of 7.5% and 10%, respectively. Analyses of plasma total adiponectin and high-molecular weight (HMW) adiponectin were performed as previously described.¹⁸

Candidate biomarkers were assessed annually through study year 10 (mean participant follow-up 6.9±2.4 years), except for adiponectin and HMW adiponectin which were only evaluated through study year 3. Among the 677 participants without monogenic diabetes mutations, 507 participants with CVD biomarker data available at baseline were included in the present analysis (Appendix Figure 1).

2.2. Kidney Disease Outcome Measures

Moderate albuminuria was defined as a urine albumin-to-creatinine ratio (ACR) ≥30 mg/g at two consecutive visits or as an elevated ACR and on reported therapy. The Full Age Spectrum (FAS) combined serum creatinine and cystatin C equation, which has been validated in children and adults¹⁹, was used to calculate estimated glomerular filtration rate (eGFR). Hyperfiltration was defined as eGFR (FAS combined equation) ≥135 mL/min/1.73 m² at two consecutive visits.²⁰

2.3. Echocardiographic Ventricular Function Outcomes

Transthoracic echocardiography was performed during the last year of the TODAY clinical trial (2010-2011) and 5 years later during the observational follow-up, TODAY2 (2015-2016). The exams were read centrally at the Echocardiography Reading Centers (Johns Hopkins University, and A. I. duPont Hospital for Children).^2,21 Studies were performed in accordance with recommendations of the American Society of Echocardiography.²² In these analyses, we included 256 participants with echocardiographic measurements available at both time points and with CVD biomarker data available within 1 year of the initial echocardiogram (Appendix Figure 1). These participants were slightly younger (13.7 vs 14.2 years; p=0.0007), with a lower BMI (34.0 vs. 35.8 kg/m²; p=0.0005) and a shorter diabetes duration (7.0 vs. 8.3 months; p=0.008) at randomization compared to the 421 participants not included in the analysis. Sex, race-ethnicity, and baseline HbA1c were comparable between the two groups. The two primary echocardiographic outcomes were defined as a decline in ejection fraction ≥10% over 5 years and an increase in the lateral ratio of the early transmitral flow velocity to the early diastolic tissue velocity (E/Em ratio) >1.0 cm/sec over 5 years²³. We also evaluated the presence of an ejection fraction <52% or the presence of a lateral E/Em ratio >13.0 cm/s at the time of the follow-up echocardiogram.²² Left ventricular (LV) global strain was measured during the second examination.²⁴ A cutoff of −17% (quartile of the distribution closest to 0) was used to determine abnormal values for speckle-tracking 4D and 2D longitudinal strains.

2.4. Pulse Wave Velocity Outcomes

Pulse wave velocity (PWV), a measure of vascular stiffness, was assessed twice during the observational follow-up, TODAY2, once in 2013-2014 and 5 years later in 2018-2019. These were also evaluated centrally at the Vascular Reading Center (Cincinnati Children’s Hospital Medical Center). In these analyses, we included 193 participants with PWV measurements available at both time points and with CVD biomarker data available within 1 year of the initial PWV assessment (Appendix Figure 1). These participants were more likely to be Hispanic (51.6% vs. 39.4%; p=0.02), slightly younger (13.6 vs 14.2 years; p=0.0002), and had a lower BMI (33.7 vs. 35.7 kg/m²; p=0.0015) at randomization compared to the 484 participants not included in the analysis. Sex, baseline diabetes duration, and HbA1c were comparable between the two groups. The primary carotid-femoral PWV outcome was defined as an increase in PWV ≥1.0 m/sec over 5 years. Prior work has shown that a 1 m/sec increase in carotid-femoral PWV was associated with a 12 % increased risk of CVD as well as with a 9% increased risk of CVD mortality, and as such was deemed to be a clinically meaningful change²⁵.

2.5. Statistical Analyses

Owing to the zero-inflated distributions, abnormal risk categories were applied to three of the biomarkers (BNP, ≥100 pg/mL; troponin, ≥0.01 pg/mL; and TNFα, >5.6 pg/mL) based on current consensus and as previously reported.² All other biomarkers (FGF-23, copeptin, E-selectin, ICAM-1, VCAM-1, IGFBP-1, proinsulin, adiponectin, and HMW adiponectin) were evaluated continuously. Due to skewed distributions, biomarkers were log-transformed prior to testing to approximate normality. As adiponectin and HMW adiponectin data were not obtained at visits near the first echocardiogram or first PWV assessment, associations for these biomarkers were only evaluated using the baseline value.

For kidney disease outcomes, candidate biomarkers were evaluated as baseline (B) or time-dependent covariates (current value [C] and updated mean [U]) in Cox proportional hazards regression models up to study year 10, censoring outcomes after year 10 and excluding participants who had the renal outcomes at study entry. The updated mean (computed by weighting each value by the interval between measurements to account for the varying frequencies of measurement during the study) was used to reflect the cumulative exposure from baseline up to and including the biomarker value at each visit throughout the study. For the echocardiography and PWV outcomes, candidate biomarkers were evaluated at baseline, within 1 year of the first outcome assessment, and as a summary value between randomization and the first outcome assessment. Summary values for continuous biomarkers were defined by the mean value over all visits between randomization and within 1 year of the first outcome assessment. For categorical biomarkers, summary values were defined by reaching abnormal risk thresholds (i.e., ≥100 pg/mL for BNP) at ≥75% of the visits between randomization and within 1 year of the first outcome assessment. There were, on average, 5 visits (range 2-7) between randomization and the first echocardiography visit and, on average, 7 visits (range 4-10) between randomization and the first PWV visit. Candidate biomarkers were evaluated in logistic regression models. All models were adjusted for sex, race/ethnicity, treatment group, age, BMI, and HbA1c.

The coefficients of variation, expressed in percent, are presented and calculated as the standard deviation divided by the mean of the biomarker value assessed between randomization and within 1 year of the first outcome assessment. Given multiple comparisons, a p-value <0.01 was considered statistically significant.

3. Results

Among the 507 participants with CVD biomarker data at baseline, 35.7% were male, with mean (SD) age 13.9 (2.0) years, duration of diabetes 7.8 (5.9) months, BMI 34.8 (7.6) kg/m², and HbA1c 6.0 (0.7) % (Appendix Table 1). Baseline characteristics were similar between the 507 participants and the subsets with echocardiographic and PWV measurements (Appendix Table 1).

3.1. Kidney Disease

Among the subset of 507 participants, the baseline prevalence of moderate albuminuria was 8.1% with a 10-year cumulative incidence of 42.2% and median time to moderate albuminuria of 6 years. Similarly, for hyperfiltration, the baseline prevalence was 12.6% and the 10-year cumulative incidence was 49.0% with a median time to hyperfiltration of 4 years. Ten participants had both moderately increased albuminuria and hyperfiltration at baseline, and up to 54 participants during follow-up. Higher current E-selectin was associated with higher risk of moderate albuminuria at the time of assessment (HR=1.07 95% CI 1.03-1.12) (Table 1). A higher current or updated mean value of proinsulin was associated with an increased risk of moderate albuminuria (HR=1.04 95% CI 1.03-1.06 and HR=1.05, 95% CI 1.03-1.08, respectively) and hyperfiltration (HR=1.03, 95% CI 1.01-1.05, and HR=1.03 95% CI 1.01-1.05, respectively). Similarly, higher proinsulin to insulin ratio (current and updated means) were associated with higher risk for hyperfiltration. A higher proinsulin to adiponectin ratio at baseline was associated with moderate albuminuria (HR=1.03, 95% CI 1.01-1.05). Finally, higher baseline and updated mean values of FGF-23 were associated with a decreased risk of hyperfiltration (HR=0.94, 95% CI 0.90-0.98 and HR=0.93, 95% CI 0.89-0.98, respectively) (Table 1). However, in correlation analyses, there were negative relationships between updated mean values of eGFR with baseline FGF-23 (r=−0.12, p<0.0001; Figure 1A) and with updated mean concentrations of FGF-23 (r=−0.18, p<0.0001; Figure 1B). This inverse relationship persisted after adjusting for time-varying BMI and HbA1c. Over time, the average change in eGFR per 1 unit increase in log FGF-23 was −7.29 (SE 0.7) ml/min/1.73 m², p<0.001. There were no statistically significant associations for other biomarkers with moderate albuminuria or hyperfiltration.

Table 1.

Multivariable Cox proportional hazard models predicting risk of albuminuria (UACR ≥30 mg/g; N=114 events) and hyperfiltration (≥135 mL/min/1.73 m²; N=150 events) with the biomarkers as fixed or time-dependent covariate.

Biomarker		UACR ≥30 mg/g			Hyperfiltration ≥135 ml/min/1.73m2
		Hazard Ratio	95% CI	P-value	Hazard Ratio	95% CI	P-value
Troponin ≥ 0.01 ng/mL	B	1.15	(0.78-1.71)	0.48	1.05	(0.76-1.46)	0.75
	C	1.05	(0.71-1.57)	0.79	1.05	(0.75-1.47)	0.79
BNP ≥ 100 pg/mL	B	1.00	(0.64-1.56)	0.99	0.73	(0.47-1.13)	0.15
	C	0.82	(0.51-1.30)	0.39	0.72	(0.46-1.12)	0.14
TNFα ≥ 5.6 pg/mL	B	0.70	(0.41-1.19)	0.19	1.17	(0.77-1.76)	0.46
	C	0.77	(0.46-1.27)	0.30	1.04	(0.69-1.54)	0.86
Copeptin (ng/mL; per 10% increase)	B	1.00	(0.95-1.04)	0.95	1.01	(0.97-1.06)	0.59
	C	1.00	(0.95-1.05)	0.99	1.02	(0.98-1.06)	0.33
	U	0.98	(0.93-1.04)	0.53	1.02	(0.97-1.08)	0.46
E-selectin (ng/mL; per 10% increase)	B	1.01	(0.97-1.06)	0.61	1.01	(0.97-1.04)	0.66
	C	1.07	(1.03-1.12)	0.0007	1.04	(1.00-1.07)	0.04
	U	1.05	(0.99-1.10)	0.06	1.01	(0.98-1.05)	0.46
IGFBP-1 (pg/mL; per 10% increase)	B	0.98	(0.96-0.99)	0.02	0.99	(0.97-1.01)	0.09
	C	1.00	(0.98-1.02)	0.97	0.99	(0.98-1.01)	0.23
	U	0.98	(0.96-1.00)	0.07	0.98	(0.97-1.00)	0.01
ICAM-1 (ng/mL; per 10% increase)	B	1.00	(0.99-1.02)	0.61	1.01	(0.99-1.03)	0.36
	C	1.01	(0.99-1.03)	0.54	1.01	(0.99-1.03)	0.29
	U	1.01	(0.95-1.08)	0.74	1.01	(0.99-1.03)	0.23
VCAM-1 (ng/mL; per 10% increase)	B	1.02	(0.96-1.08)	0.60	1.00	(0.95-1.04)	0.86
	C	1.00	(0.95-1.06)	0.96	1.00	(0.95-1.05)	0.99
	U	1.01	(0.95-1.08)	0.74	0.99	(0.94-1.05)	0.72
FGF-23 (pg/mL; per 10% increase)	B	0.99	(0.95-1.04)	0.76	0.94	(0.90-0.98)	0.005
	C	1.02	(0.98-1.07)	0.38	0.98	(0.95-1.01)	0.17
	U	1.00	(0.95-1.06)	0.93	0.93	(0.89-0.98)	0.004
Proinsulin (pM; per 10% increase)	B	1.03	(1.00-1.05)	0.05	1.02	(0.99-1.04)	0.14
	C	1.04	(1.03-1.06)	<0.0001	1.03	(1.01-1.05)	0.002
	U	1.05	(1.03-1.08)	<0.0001	1.03	(1.01-1.05)	0.005
Proinsulin/insulin (per 10% increase)	B	1.01	(0.98-1.04)	0.58	1.02	(0.99-1.04)	0.18
	C	1.02	(0.99-1.04)	0.11	1.03	(1.01-1.05)	0.0006
	U	1.03	(0.99-1.06)	0.10	1.03	(1.01-1.06)	0.0009
Adiponectin (ng/mL; per 10% increase)	B	0.96	(0.92-1.00)	0.04	0.98	(0.95-1.02)	0.28
Proinsulin/adiponectin (per 10% increase)	B	1.03	(1.01-1.05)	0.005	1.02	(1.00-1.03)	0.04
HMWA (ng/mL; per 10% increase)	B	0.99	(0.96-1.01)	0.30	1.00	(0.98-1.02)	0.97

B = baseline biomarker value (fixed); C = current value of the biomarker at the time of the event/censoring (time-dependent); U = updated mean value of the biomarker up to the time of the event/censoring (time-dependent). Of the 507 TODAY participants with biomarker data available at baseline, 114 and 150 developed moderate albuminuria and hyperfiltration during the first 10 years in the study, respectively. Separate Cox proportional hazard regression models adjusted for sex, age at baseline, race-ethnicity, randomized treatment group, BMI (baseline for B models, updated mean for C and U models), and HbA1c (baseline for B models, updated mean for C and U models). The hazard ratios (95% CI) and corresponding P-values are presented from each model. Results with P-values <0.01 are bolded. Log-transformed values of the biomarkers are entered in the models for each continuous variable.

Figure 1.

Negative correlation between updated mean eGFR and baseline FGF-23 (Panel A; Spearman r = −0.12, p-value <0.0001) and updated mean FGF-23 (Panel B, Spearman r = −0.18, p-value <0.0001), across all study visits combined.

3.2. Echocardiography

Among the subset of 256 participants, 10.5% of participants had a decline in ejection fraction of ≥10% over 5 years and only four participants had ejection fraction <52% at the time of the follow-up echocardiogram. For diastolic function, 36.7% of participants had an increase in lateral E/Em ratio of >1.0 cm/sec over 5 years and only one participant had a lateral E-Em ratio >13 cm/sec at the time of the follow-up echocardiogram. There were no statistically significant associations with any biomarkers for the two primary outcomes (Appendix Table 2). There were no statistically significant relationships between biomarkers and speckle-tracking 4D and 2D longitudinal strain measures (Table 2). Lower baseline IGFBP-1 related to abnormality in 2D longitudinal strain (≥−17%), but the association was borderline at the 0.01 alpha level of statistical significance.

Table 2.

Speckle-tracking 4D and 2D longitudinal strain (value from the follow-up echocardiogram), using −17% as the cutoff for abnormal values (quartile of the distribution closest to 0), in relation to BNP ≥100 pg/mL and IGFBP-1.

median (IQR) or %	4D Longitudinal Strain		P-value	2D Longitudinal Strain		P-value
	Normal <−17% (n=185; 75.5%)	Abnormal ≥−17% (n=60; 24.5%)		Normal <−17% (n=192; 84.2%)	Abnormal ≥−17% (n=36; 15.8%)
BNP ≥100 pg/mL †
Baseline	16.8	20.0	0.94	17.2	22.2	0.74
Within one year of 1st echo	15.7	16.7	0.99	14.6	27.8	0.04
Summary measure	13.5	15.0	0.72	12.0	27.7	0.04
IGFBP-1 (pg/mL) §
Baseline	2.7 (1.5-6.0)	2.4 (1.2-3.8)	0.05	2.7 (1.7-5.8)	2.1 (1.0-3.2)	0.01
Within one year of 1st echo	4.6 (2.0-8.9)	3.8 (1.9-7.0)	0.03	4.6 (2.1-9.6)	3.8 (1.9-6.3)	0.02
Summary measure	4.0 (2.2-8.0)	3.9 (2.2-6.8)	0.20	4.2 (2.3-8.7)	3.4 (2.2-6.9)	0.06

^†Categorical variables: % of participants with the biomarker above the threshold value are shown; P-values are derived from multivariable logistic regression models adjusted for sex, age at baseline, race-ethnicity, randomized treatment group, BMI, and HbA1c. Separate models are presented for the baseline value of the biomarker, the biomarker value within one year of the first echocardiogram assessment, and the summary measure (defined as the % of participants with values equal to or above the threshold on 75% or more of the visits between randomization and within 1 year of the first echocardiogram assessment). Results with P-values <0.01 are bolded.

^§Continuous variables: Median (IQR) values are shown; P-values are derived from multivariable logistic regression models adjusted for sex, age at baseline, race-ethnicity, randomized treatment group, BMI, and HbA1c. Separate models are presented for the baseline value of the biomarker, the biomarker value within 1 year of the first echocardiogram assessment, and the summary measure (defined as the median (IQR) of values across all visits between randomization and within 1 year of the first echocardiogram assessment). Adiponectin and HMW adiponectin were only available at the baseline time point. Log-transformed values of the biomarkers are entered in the models for each continuous variable.

3.3. Pulse Wave Velocity

Among the subset of 193 participants, 58.5% of participants had an increase in carotid-femoral PWV ≥1.0 m/sec over 5 years. There were no statistically significant associations between any biomarkers with PWV whether we examined biomarkers at baseline, within 1 year of the PWV, or when considering a summary measure of biomarkers above the threshold value on 75% or more of the visits. Of note, some of the biomarkers were highly variable over time (Appendix Table 3).

4. Discussion

In this study, we found that higher concentrations of E-selectin associated with incident albuminuria, and biomarkers of insulin resistance (lower IGFBP-1, adiponectin) and β-cell dysfunction (proinsulin) were associated with incident moderate albuminuria and hyperfiltration in adolescents with type 2 diabetes. Higher concentrations of FGF-23 correlated with lower eGFR. However, no candidate biomarkers predicted worsening arterial stiffness or decline in left ventricular systolic or diastolic function.

4.1. Kidney Disease

Type 2 diabetes is associated with a proinflammatory state. The cell-adhesion molecules ICAM-1 and VCAM-1 and E-selectin (CD62E) are expressed on endothelial and/or smooth muscle cells after their activation by cytokines such as interleukins (IL-6) and TNF-alpha, and then promote the adhesion of monocytes and other inflammatory cells. These circulating inflammatory markers are interrelated, increase during the progression of diabetes,²⁶ and are variably associated with the risk for kidney disease in adults, including cohorts with type 2 diabetes.^15,26-28 In this study, higher concentrations of E-selectin were associated with higher risk of moderate albuminuria. Consistent with our findings, the longitudinal development of increasing urinary albumin excretion was significantly and independently determined by baseline concentrations of E-selectin in adults.²⁶ In individuals with type 1 diabetes form the DCCT/EDIC, E-selectin predicted nephropathy better than other circulating biomarkers or conventional risk factors.¹⁹ Interestingly, expression of E-selectin was found to be significantly increased in the glomeruli and interstitia of patients with nephropathy as compared with those with other glomerular diseases.²⁹

FGF-23 is a bone-derived phosphaturic hormone that regulates phosphate and vitamin D homeostasis.³⁰ FGF-23 progressively rises with declining kidney function to maintain phosphate homeostasis.³¹ FGF-23 is independently associated with left ventricular mass index and left ventricular hypertrophy in adult patients with and without chronic kidney disease.^22,32 In our study, we observed an inverse relationship between FGF-23 and eGFR, consistent with these findings in adults. However, higher FGF-23 predicted less increase in hyperfiltration in our analyses that accounted for important covariates including glycemia and blood pressure. It is unclear if this reflects an early compensatory mechanism to counter other changes such as elevated phosphate in the diet or to RAAS activation in our youth who had a high prevalence of hypertension.^30,33 More detailed investigations of mineral metabolism are needed to better understand these relationships. In this young population without advanced kidney disease, we also did not find a significant relationship between FGF-23 and arterial stiffness or cardiac structure and function.

A major pathophysiological alteration of type 2 diabetes is insulin resistance, which has been implicated in the pathogenesis of kidney disease, including in the preclinical stage in children.³⁴ Proinsulin is a marker for pancreatic beta-cell dysfunction, and elevated concentrations have been related to insulin resistance in type 2 diabetes^35,36 and to CVD risk factors³⁷. Adiponectin is positively related to insulin sensitivity and predicted treatment response in TODAY with lower rise in adiponectin in response to therapy in those who failed to maintain glycemic control.¹⁸ Low IGFBP-1 is associated with insulin resistance and dysglycemia and predicts CVD risk.^16,38 Proinsulin and the proinsulin:adiponectin ratio were associated with greater risk for kidney disease in the current analyses, with an opposite relationship observed for IGFBP-1. This is consistent with our previous reports of a role of insulin resistance in addition to dysglycemia in the progression of nephropathy in TODAY.²⁰

4.2. Cardiac Function and Vascular Stiffness

Biomarkers, including troponin, showed no relationship to future deterioration of vascular stiffness, or systolic or diastolic function in adolescents with type 2 diabetes. In adults, troponin and BNP have been associated with adverse outcomes in patients with heart failure ³⁹ and older adults with diabetes.⁴⁰ We have demonstrated substantial visit-to-visit variability in these cardiac biomarkers during the TODAY study with many participants having isolated, but not persistently, elevated values.²

In adolescents with obesity, FGF-23 concentrations have been associated with abnormal cardiac structure, particularly in males.^6,41 Copeptin is elevated in adults with insulin resistance and type 2 diabetes, and is associated with hypertension, metabolic syndrome and inflammatory markers⁴², heart disease, and mortality risk.⁷ In the Atherosclerosis Risk In Communities (ARIC) study, ICAM-1 and E-selectin independently predicted the risk of carotid-artery atherosclerosis and coronary heart disease.¹³ The lack of predictive value of biomarkers for worsening arterial stiffness and cardiac changes in our study is likely due to the homogeneity of our study population and the absence of more severe abnormalities of systolic or diastolic function in the cohort. In addition, the relationship between these endothelial dysfunction markers and heart disease may take longer to be established. In the CARDIA study, the associations of E-selectin and ICAM-1, obtained in young adulthood (age 18 to 30), associated with adverse cardiac function (worse LV global longitudinal strain, an index of heart failure with preserved ejection fraction) 15 to 23 years later.⁴³ Of note, neither E-selectin nor ICAM-1 associated with measures of LV diastolic function in that study.⁴³

We chose to examine clinically relevant measures of cardiac dysfunction (increase in lateral E/Em ratio >1.0 cm/sec over 5 years; decline of 10% in ejection fraction over 5 years) in relationship to longitudinal changes in biomarkers while adjusting for important covariates of sex, race-ethnicity, BMI, and glycemia. Collectively, these results suggest caution in interpreting cardiac biomarker concentrations based on single values or in those without established heart disease as studies in adults generally find these values predictive in those with prevalent cardiac dysfunction.

The evaluation of biomarkers longitudinally in a pediatric cohort of youth-onset type 2 diabetes is novel and contributes to the limited literature about complications of type 2 diabetes in youth. An important strength of the TODAY study is the long-term longitudinal follow-up (over 10 years) and the repeat measurements of both biomarkers and study endpoints. Most studies evaluating the relationship of biomarkers and outcomes are cross-sectional or have much shorter follow-up. Additionally, the value of biomarkers in predicting outcomes was carefully assessed by determining clinically relevant outcome measures and adjusting for important confounders such as adiposity, glycemic control, treatment group, age, sex, and race-ethnicity. Another strength is the quality-control procedures that were in place at the various TODAY study central laboratories over the course of the study.

One limitation of this study is the relatively homogeneous nature of the cohort, being severely obese and having type 2 diabetes. This may limit the range of expression of certain biomarkers. However, comparable cohorts in age, race-ethnicity, and other demographics are not available in pediatrics to contribute a reference or comparative population. The availability of adiponectin and HMW adiponectin only through year 3 in TODAY is a limitation related to their use in our analyses. Hyperfiltration is reported to occur in up to 50% of individuals with type 2 diabetes and to precede the decline in eGFR ^32,36. Nevertheless, it remains a controversial marker of early kidney dysfunction. The relatively small number of our study population with longitudinal vascular and cardiac measures, compared with adult studies, may have limited the power to detect significant relationships.

5. Conclusions

In summary, with regard to diabetes-related kidney disease and its largely microvascular pathophysiology, our findings suggest that circulating biomarkers of endothelial dysfunction (namely E-selectin), and markers of β-cell dysfunction (proinsulin) and insulin sensitivity (adiponectin, IGFBP-1), could potentially be incorporated into a more personalized risk assessment of future kidney disease. Additional studies with more prolonged follow-up are needed to refine this approach. However, circulating biomarkers of cardiovascular dysfunction, found to be useful in detection of cardiovascular events in adults, have limited value when measured early in the first decade of diabetes onset in predicting risk of cardiac dysfunction in youth with type 2 diabetes with relatively preserved cardiac function and in the absence of significant cardiac injury.

HIGHLIGHTS.

1. Youth with type 2 diabetes have a high burden of kidney and cardiovascular disease

2. Biomarkers of endothelial dysfunction, β-cell dysfunction, insulin sensitivity

3. Higher E-selectin and proinsulin concentrations associated with kidney disease

4. Biomarkers for personalized risk assessments in youth-onset type 2 diabetes

Data Sharing:

Data collected for the TODAY study are available to the public through the NIDDK Central Repository (https://repository.niddk.nih.gov/studies/today/).