Cumulative Effect of Metabolic Risk Factors on Left Ventricular Geometry in Those With versus Without Early-Onset Type 2 Diabetes or Prediabetes – The Coronary Artery Risk Development in Young Adults (CARDIA) Study

Abstract

Aims

We investigated metabolic risk factors (RFs) that accumulated over 20 years related to left ventricular mass index (LVMI), relative wall thickness (RWT), and LV remodeling patterns in participants with versus without early-onset type 2 diabetes (T2D) or pre-diabetes (pre-DM).

Methods

287 early-onset T2D/pre-DM individuals vs. 565 sociodemographic-matched euglycemic individuals were selected from the Coronary Artery Risk Development in Young Adults (CARDIA) study, Years 0–25. We used the area under the growth curve (AUC) derived from quadratic random-effects models of ≥4 repeated measures of RFs (fasting glucose [FG], insulin, triglycerides [TG], LDL-c, HDL-c, total-c, blood pressure, and BMI) to estimate the cumulative burden, and their associations with LV outcomes.

Results

One SD greater AUC of log (TG) (per 0.48) and HDL-c (per 13.5 mg/dL) were associated with RWT (β 0.21, and −0.2) in the early-onset T2D/pre-DM group but not in euglycemia group (β 0.01 and 0.05, p-interactions 0.02 and 0.03). In both early-onset T2D/pre-DM and euglycemic groups, greater AUCs of log (FG) (per 0.17) and log (insulin) (per 0.43) were associated with higher RWT (β ranges 0.12–0.24). Greater AUCs of SBP (per 10 mmHg) and DBP (per 7.3 mmHg) were associated with higher RWT and LVMI, irrespective of glycemic status (β ranges 0.17 to 0.28). Cumulative TG (OR 3.4, 95%CI 1.8–6.3), HDL-c (0.23, 0.09–0.59), and total-c (1.9, 1.1–3.1) and FG (2.2, 1.25–3.9) were statistically associated with concentric hypertrophy in the T2D/pre-DM group only.

Conclusions

Sustained hyperglycemia and hyperinsulinemia are associated with RWT, with those with early T2D/pre-DM at excess risk, potentially due to their higher levels of glucose and insulin. Dyslipidemia was associated with LV structural abnormalities in those early-onset T2D/pre-DM.

Introduction

The prevalence and incidence of early-onset type 2 diabetes (T2D) and pre-diabetes (pre-DM) (onset before 40 years) have been increasing in the past three decades, paralleling the rising obesity prevalence in young adults worldwide. ¹ Early-onset T2D is characterized by a more aggressive disease course and is associated with a greater relative risk of cardiovascular disease (CVD), including ischemic heart disease and heart failure, compared to later-onset T2D. ^{2, 3} Increases in left ventricular (LV) mass and wall thickness are early manifestations of myocardial involvement in T2D. Emerging evidence from middle-aged cohorts has demonstrated the association between T2D or pre-DM and alterations in LV geometry. ^{4, 5 6} However, only a limited number of studies have evaluated the longitudinal effect of early-onset T2D/pre-DM on LV geometric changes. ^6–8 Importantly, it remains unclear to what extent metabolic risk factors commonly clustered with hyperglycemia and insulin resistance confer differential associations with LV geometric changes.

Characterizing the development of cardiac remodeling associated with early-onset T2D/pre-DM is critical for early prevention to mitigate future cardiovascular events in this high-risk population. Quantifying the relative contributions of modifiable risk factors in addition to hyperglycemia and insulin resistance, including lipid dysregulation, hypertension, and adiposity, would help delineate mechanisms for cardiovascular risk that are disparate or common in young adults with or without T2D/pre-DM. Such findings can inform precise, potentially more intensified risk factor management among individuals with early-onset T2D to reduce the risk of cardiovascular complications later in life.

The study’s main objective was to investigate how metabolic risk factors accumulated over 20 years among individuals with versus without early-onset T2D/pre-DM relate to LV geometry alternations. We hypothesized that the cumulative burden of metabolic risk factors differentially associated with LV structural abnormalities, including higher LV mass index (LVMI) and relative wall thickness (RWT), as well as LV remodeling patterns (i.e., eccentric hypertrophy, concentric hypertrophy, and concentric remodeling) in young adults with early-onset T2D/pre-DM. We further hypothesized that the cumulative effect of metabolic risk factors on LV structural changes is more detrimental in people with early-onset T2D/pre-DM than in those with euglycemia.

Methods

Data source:

We used data from the Coronary Artery Risk Development in Young Adults (CARDIA) study. CARDIA is a longitudinal cohort study that began in 1985 and 1986 with an enrollment of 5,115 healthy young adults, balanced on sex, age (18 to 24 and 25 to 30 years), race (Black and White participants), and education (up to/through and more than high school), at 4 U.S. field centers: Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California. A total of 9 in-person follow-up examinations have occurred. CARDIA participants have undergone in-person examinations at baseline (Year 0: Y0) and at Y2, Y5, Y7, Y10, Y15, Y20, Y25, Y30, and Y35. Retention rates among surviving participants at each in-person examination were 91%, 86%, 81%, 79%, 74%, 72%, 72%, 71%, and 67% (during the COVID-19 pandemic), respectively. Contact is maintained with participants via telephone, mail, or email every 6 months, with annual interim medical history ascertainment. The institutional review board at each of the study sides approved the study protocols, and written informed consent was obtained from all participants. We used data from baseline (year 0) and years 7, 10, 15, 20, and 25. We included participants who completed echocardiographic examinations in year 25 (N=3,265). Among these participants, we identified those with early-onset T2D/pre-DM at or before Y10 (N=329, definition in Measures). We omitted early-onset T2D/pre-DM cases after Y10 for two reasons. First, the mean age of all CARDIA participants at Y10 was 40 years, which resembles the upper age limit for early onset T2D/pre-DM. ¹⁸ Additionally, a minimum of 4 measures of risk factors were needed for the quadratic random-effects models (see Analysis). We identified a comparison group that remained euglycemic from year 0 to year 25 (N=1,612) and further age-, sex-, and race-matched (ratio 1 to 2, N=658) to early-onset T2D/pre-DM participants at the examination year where the early-onset T2D/pre-DM status occurred. We excluded women who were pregnant at any year, had been using anti-diabetic medications, and participants with less than four measurements of risk factors across examination years prior to age 40 years. ^{9, 10} The final sample included 287 early-onset T2D/pre-DM individuals and 565 matched individuals with euglycemia. A flow chart for sample selection is in Figure 1.

Figure 1.

Sample Selection

Glycemic status measures:

We defined early-onset T2D by self-reported diabetes, fasting glucose (FG) ≥126 mg/dL (years 0, 7, 10), or 2-hour oral glucose tolerance test (OGTT) ≥200 mg/dL (year 10) at or before examination year 10 (maximum age was40). Pre-DM status among those not using antidiabetic medication was based on FG 100–125 mg/dL or OGTT 140–199 mg/dL at or before year 10. Individuals with euglycemia were those with normal glycemic status based on available FG or OGTT from year 0 to year 25.

Doppler echocardiographic assessment for outcomes:

2D-guided M-mode echocardiography was performed with an Artida cardiac ultrasound scanner using standardized protocols across all field centers. The echocardiography protocol has previously been published and followed existing American Society of Echocardiography guidelines for study acquisition and measurement. ¹¹ Quality control and image analysis were performed at a core reading center (Johns Hopkins University, Baltimore, MD). ¹² LVMI was measured using the Devereux formula and indexed to the body surface area. RWT was calculated by dividing the sum of the posterior wall and interventricular septal thickness by the internal left ventricular diameter. ¹³ Hypertrophy was defined by LVMI ≥115 g/m² in men or ≥95 g/m² in women. LV geometry patterns defined by LVMI and RWT suggest varied prognoses. ¹⁴ The proportion of different LV geometry patterns affected by diabetes varied across studies. ^{15, 16} We categorized three LV geometry patterns, including concentric remodeling (RWT>0.42 and no hypertrophy), concentric hypertrophy (RWT>0.42 and hypertrophy), and eccentric hypertrophy (RWT ≤0.42 and hypertrophy) ^{13, 17} to examine their prevalence and associations with the cumulative burden of metabolic risk factors.

Risk factors:

Glucose was assayed at baseline with the hexokinase UV method by American Bio-Science Laboratories (Van Nuys, CA) and by hexokinase coupled to glucose-6-phosphate dehydrogenase (Merck Millipore, Billerica, MA) at years 7, 10, 15, 20, and 25. The insulin measurements were performed with the use of a radioimmunoassay (Linco Research, St. Charles, MO) at baseline and years 7, 10, 15, and 20, and an Elecsys sandwich immunoassay (Roche Diagnostics Corporation, Indianapolis, IN) was performed at year 25.^{18, 19} Total cholesterol and triglycerides (TG) were measured enzymatically, high-density lipoprotein cholesterol (HDL-c) was determined by precipitation with dextran sulfate magnesium chloride, and low-density lipoprotein cholesterol (LDL-c) was calculated using the Friedewald equation.²⁰ Sitting resting heart rate measurements were made at all examinations., Blood pressure (BP) was measured on the right arm of seated participants at three 1-min intervals using a Hawksley random zero sphygmomanometer after a 5-miniute rest (W.A. Baum Company, Copaigue, NY) at baseline through year 15. At years 20 and 25, BP was measured using a standard automated BP measurement monitor (IntelliSense Blood Pressure Monitor, model HEM-907XL; Omron) and standardized to the sphygmomanometer measures.²¹ Body mass index (BMI) was calculated as weight divided by height (kg/m²). Demographics and smoking status were measured by standard questionnaires across all CARDIA examination visits.²²

Analysis:

We used the area under the curve (AUC) between Y10 and Y25 that was derived from quadratic random-effects models of ≥4 repeated measures of metabolic risk factors^{23, 24}, including FG, fasting insulin, TG, LDL-c, HDL-c, total-c, systolic BP (SBP), diastolic BP (DBP), and BMI to estimate the cumulative risk factors burden through adulthood. The AUC, computed from longitudinal growth curve models, uses multiple measurements of risk factors throughout life to reduce within-person variability. AUC adjusts for the uneven age distribution and captures the underlying or average risk factors, leading to improved estimates of risk factor correlation and outcome prediction. ²³ This method has been widely adopted for estimating risk factor burden in life-course cohorts. ^23–26 Because the growth curves of the risk factors may be cubic models, a minimum of four measures were needed.²⁴ This linear mixed-effect model computes maximum likelihood estimates of curve parameters for all participants. The model selection was based on Akaike’s Information Criterion (AIC). We selected the most parsimonious model using p values (<0.05) of the independent variable (age). We included age (centered to mean age) and its higher-order terms one by one for model building. We calculated the AUCs using an integral calculus formula based on the fixed and random-effect parameters of the model during the follow-up period for each subject, dividing them by follow-up years to reflect the varying intervals of follow-up between participants.^23–26 For example, the growth curve model for FG is specified as:

$$\text{F}{\text{G}}_{\text{i}}=\left(\text{β}0+\text{b}{0}_{\text{i}}\right)+\left(\text{β}1+\text{b}1\text{i}\right)\text{age}+\left(\text{β}2+\text{b}2\text{i}\right)\text{age}2+\left(\text{β}3+\text{b}3\text{i}\right)\text{age}3+\text{εi},$$

where $\text{β}=\left(\text{β}0,\text{β}1,\text{β}2,\text{β}3\right)$ is a vector of fixed-effect parameters, $\text{bi}=\left(\text{b}0\text{i},\text{b}1\text{i},\text{b}2\text{i},\text{b}3\text{i}\right)$ is a vector of random-effect parameters, and $\text{εi}$ is an unknown error for individual i. The detailed model-building process and model-fitting parameters are presented in Supplemental Table 3.

We performed linear regressions to examine the associations of the total AUC of each risk factor with LVMI or RWT, adjusting for sex, race, field center, mid-life (Year 25) age, maximum education achieved, smoking, obesity (not for AUC of BMI to avoid overfitting), medications for hypertension and lipids and resting heart rate. Covariates were selected based on prior knowledge regarding their associations with both exposure (metabolic risk factors) and outcomes (LV remodeling). ^23–26 We examined the association between the total ACU of each risk factor and the risk of concentric hypertrophy, eccentric hypertrophy, and concentric remodeling, respectively, using logistic regressions accounting for the same covariates as in the linear regression. Previous research suggests sex or racial differences in LV geometric alternations are directly or indirectly associated with diabetes. ^{6 7 8 27} Therefore, we assessed the cumulative effect of early-onset T2D in relation to LV geometric outcomes by sex and race. We tested interaction terms of sex or race with AUC of risk factors followed by stratified analysis in each subgroup.

Results

Among individuals with early-onset T2D/pre-DM in CARDIA, the mean age of diagnosis was 32 (SD=4.8) years, 45% were non-Hispanic Black participants, 64% were women, 44% had college or higher education attainment (vs. 55% in the euglycemia group, p=0.002), and 31% were current smokers (vs. 26% in euglycemia group, p=0.14) (Table 1). At year 25, the mean age of both early-onset T2D/pre-DM and euglycemia groups was approximately 51 years, and around 20% were current smokers. The prevalence of hypertrophy and RWT >0.42 was significantly higher in the early-onset DM or pre-DM group than in the euglycemic group (35.2% vs. 20.6%, p=0.001 and 13.4% vs.4%, p<.0001, respectively) at year 25. Resting heart rate was also significantly higher in the diabetic group than in the euglycemic group (69.5 vs 63.9 beats per minute [BPM], p<0.0001). Across all risk factors, the early-onset T2D/pre-DM had higher levels than the euglycemic counterparts cross-sectionally at baseline (year of diagnosis) and at year 25. Total AUCs of almost all risk factors were significantly higher among those with early-onset T2D/pre-DM compared to those with euglycemia (Table 1). The prevalence of concentric hypertrophy, concentric remodeling, and eccentric hypertrophy was significantly higher in the diabetic group compared with the euglycemic group (7.4% [18/244] vs. 1.7% [8/478] P<0.0001 6.7%, [16/244] vs. 1.7% [8/478] P<0.0005, and 27.9 % [68/244] vs. 22.4% [107/478] P=0.10 respectively) (Figure 2).

Table 1.

Characteristics of Participants With Early-Onset DM/PreDM Versus Euglycemia

	Early-Onset DM/PreDM N=287	Matched-Euglycemia N=565	P-value
Year of diagnosis/matched euglycemia
Age, year, mean SD	32.4 (4.8)	32.2 (4.8)	0.46
Non-Hispanic Blacks, n %	128 (45)	253 (45)	0.96
Women, n %	184 (64)	361 (64)	0.95
≥ College education, n %	127 (44)	315 (55)	0.0016
Current smoker, n %	76 (31)	114 (26)	0.14
Metabolic risk factors, mean SD
Fasting glucose (FG), mg/dL, median IQR	101 (23) (87, 110)	83 (10) (78, 88)	<0.0001
Log (FG), mean SD	4.6 (0.2)	4.4 (0.2)	<0.0001
Fasting insulin, mIU/L, median IQR	18.5 (16.0) (12, 28)	10 (5) (8, 13)	<0.0001
Log (insulin), mean SD	2.9 (0.7)	2.4 (0.4)	<0.0001
BMI, kg/m2, mean SD	31 (7.9)	25.6 (5.4)	<0.0001
HDL-c, mg/dL, mean SD	46.7 (13.4)	54.3 (13.1)	<0.0001
LDL-c, mg/dL, mean SD	110 (31.2)	106 (31)	0.07
Triglycerides, mg/dL, median IQR	87 (89) (57, 146)	59 (38) (43, 81)	<0.0001
Log-transformed triglycerides	4.6 (0.7)	4.1 (0.5)	<0.0001
Systolic blood pressure, mmHg, mean SD	112 (14)	107 (11)	<0.0001
Diastolic blood pressure, mmHg	72 (14)	67 (9.5)	<0.0001
Anti-hypertensive medication, n %	8 (2.8)	5 (0.9)	0.004
Lipid-lowering medication, n %	1, (0.4)	0 (0)	0.34
Year 25
Age, year, mean SD	50.7 (3.6)	50.5 (3.6)	0.5
Current smoker, n %	52 (22)	85 (18)	0.05
Metabolic risk factors, mean SD
Fasting glucose (FG), mg/dL, median IQR	110 (52) (94, 146)	89 (9) (85, 94)	<0.0001
Log (FG), mean SD	4.8 (0.4)	4.5 (0.1)	<0.0001
Fasting insulin, mIU/L, median IQR	12.9 (11.8) (8, 19.8)	7.5 (6.3) (4.7, 10.9)	<0.0001
Log (insulin), mean SD	2.5, 0.7	2.0, 0.6	<0.0001
BMI, kg/m2, mean SD	33.9 (8.4)	29.1 (6.7)	<0.0001
HDL-c, mg/dL, mean SD	53.7 (17.4)	61.4 (18.6)	<0.0001
LDL-c, mg/dL, mean SD	105.0 (34)	112.3 (31.3)	0.0026
Triglycerides, mg/dL, median IQR	111 (87) (79, 166)	83 (56) (60, 116)	<0.0001
Log-transformed triglycerides	4.8 (0.6)	4.5 (0.5)	<0.0001
Systolic blood pressure, mmHg, mean SD	123.1 (15.9)	118.7 (15.1)	<0.0001
Diastolic blood pressure, mmHg	76.9 (10.3)	74.2 (11.2)	0.0004
Left ventricular mass, mean SD	184.3 (59.2)	159.7 (46.9)	<0.0001
Left ventricular mass index (LVMI), median IQR	92.8 (30.5) (79.5, 110.0)	85.2 (30.1) (71.7, 101.8)	0.0001
LVMI≥115 g/m2 in men or ≥95 g/m2 in women, n %	84 (35.3)	117 (24)	0.0014
Relative wall thickness (RWT), median IQR	0.34 (0.08) (0.30, 0.38)	0.32 (0.07) (0.28, 0.35)	<0.0001
RWT>0.42, n %	36 (13.3)	22 (4)	<0.0001
LV ejection fraction <50%, n %	6 (2.3)	2 (0.4)	0.0107
Resting heart rate, beats/min, mean SD	69.3 (11.9)	63.9 (10.6)	<0.0001
Anti-hypertensive medication, n %	144 (50.5)	107 (19.0)	<0.0001
Lipid-lowering medication, n %	100 (35.7)	53 (9.4)	<0.0001
Anti-diabetic medications, n %	110 (38.7)	0 (0)	<0.0001
Total AUC of risk factors, mean, SD
Log (FG)	4.7 (0.2)	4.4 (0.1)	<0.0001
Log (insulin)	2.8 (0.4)	2.3 (0.3)	<0.0001
BMI, kg/m2	32.6 (7.7)	26.4 (5.4)	<0.0001
HDL-c, mg/dL	49.2 (12.9)	56.1 (13)	<0.0001
LDL-c, mg/dL	109.2 (24.9)	107.3 (24.1)	0.6
Log-transformed triglycerides	4.6 (0.6)	4.4 (0.4)	<0.0001
Systolic blood pressure, mmHg	116.5 (11.4)	109.3 (8.9)	<0.0001
Diastolic blood pressure, mmHg	75.1 (7.5)	69.9 (6.7)	<0.0001

Figure 2.

Prevalence of LV Remodeling Patterns at CARDIA Year 25, 2010–2011

Table 2 shows results from the linear regressions for the association between AUC of risk factors and RWT and LVMI. A higher positive β coefficient indicates a stronger association with a risk of LV outcomes (for HDL-c, a greater negative β coefficient indicates an association with a higher risk of LV outcomes). In the fully adjusted model (model 2), one standard deviation (SD) higher total AUC of log (FG) (per 0.17) was significantly associated with higher RWT (β coefficient 0.26, 95% confidence interval 0.11–0.41) in participants with early-onset T2D/pre-DM. One SD increment of AUC of log (insulin) (per 0.43) was significantly associated with higher RWT in both early-onset T2D/pre-DM (0.24, 0.07–0.41) and euglycemia groups (0.12, 0.004–0.23). One SD higher AUC of log (TG) (per 0.47) was significantly associated with higher RWT (0.21, 0.08–0.34) in the early-onset T2D/pre-DM group but not in the euglycemia group (p-interaction 0.02), indicative the modifying effect of early-onset T2D/pre-DM status. We found a significant inverse association between the AUC of HDL-c (per 13.5 mg/dL) and RWT (−0.2, −0.37 - −0.02) in the early-onset T2D/pre-DM group but not in the euglycemia group (p-interaction 0.03). In sex-stratified analysis, the increased AUC of HDL-c was associated with increased LVMI in hyperglycemic men but not in hyperglycemic women (−0.33, −0.65- −0.01 and −0.01, −0.22– 0.2, in men and women with early DM/pre-DM, P-interaction 0.3) (Supplemental Table 1). The AUCs of SBP (per 10 mmHg) and DBP (per 7.3 mmHg) were significantly positively associated with LVMI and RWT (β coefficients ranging from 0.17 to 0.28) in both early-onset T2D/pre-DM and euglycemia groups. One SD increment of AUC of BMI (per 6.9 kg/m²) β coefficients ~0.2) was associated with higher LVMI in participants with early-onset T2D/pre-DM and euglycemia. However, BMI was not associated with RWT in either group. In the sex and race-stratified analyses (Supplemental Table 1), we found a one-SD increment of AUC of BMI (per 6.9 kg/m²) was more strongly associated with LVMI in men than in women in both early-onset T2D/pre-DM (0.51, 0.09–0.94 in men and 0.01, −0.23–0.24 in women, p-interaction 0.2) and euglycemia groups (0.39, 0.04–0.74 in men and 0.12, −0.03–0.27 in women, p-interaction 0.1). Greater AUC of BP was more strongly associated with higher LVMI in Black participants with euglycemia than White participants with euglycemia (p-interaction<.05). No significant sex or race differences were observed in the AUC of risk factors and RWT (Supplemental Table 1). In a separate analysis of the associations between cumulative risk factors and LV mass, we had findings similar to those of the LVMI (results not shown).

Table 2.

. Adjusted β Coefficients (95% CI) of One SD Increase in AUCs of Risk Factors and LVMI and RTW in Participants with or Without Early-Onset Diabetes or Pre-Diabetes

		RWT (β)			LVMI (β)
		Early T2D/PreDM	Euglycemia	P #	Early T2D/PreDM	Euglycemia	P #
Log (FG), per 0.17	Model 1	0.14 (0.01, 0.27)*	0.19 (−0.07, 0.44)	0.8	0.08 (−0.05, 0.21)	0.04 (−0.23, 0.30)	0.8
	Model 2	0.26 (0.11, 0.41)*	0.16 (−0.1, 0.42)	0.5	0.13 (−0.03, 0.29)	−0.1 (−0.37, 0.16)	0.1
Log (Insulin), per 0.43	Model 1	0.17 (0.03, 0.31)*	0.13 (0.03, 0.23)*	0.6	0.22 (0.07, 0.36)*	0.18 (0.07, 0.29)	0.7
	Model 2	0.24 (0.07, 0.41)*	0.12 (0.004, 0.23)*	0.2	0.11 (−0.07, 0.29)	0.06 (−0.05, 0.18)	0.6
Log (TG), per 0.47	Model 1	0.21 (0.08, 0.33)*	0.06 (−0.05, 0.16)	0.06	0.15 (0.02, 0.28)*	0.13 (0.03, 0.24)*	0.8
	Model 2	0.21 (0.08, 0.34)*	0.01 (−0.1, 0.12)	0.02	0.12 (−0.02, 0.25)	0.03 (−0.08, 0.14)	0.3
HDL-c, per 13.5 mg/dL	Model 1	−0.15 (−0.31, 0.00)	−0.01 (−0.09, 0.08)	0.07	−0.16 (−0.31, −0.01)*	−0.09 (−0.18, 0.01)	0.4
	Model 2	−0.20 (−0.37, −0.02)*	0.009 (−0.08, 0.1)	0.03	−0.10 (−0.27, 0.07)	−0.02 (−0.11, 0.07)	0.4
LDL-c, per 23.9 mg/dL	Model 1	0.05 (−0.09, 0.18)	0.05 (−0.03, 0.12)	0.9	−0.01 (−0.14, 0.13)	0.01 (−0.07, 0.09)	0.8
	Model 2	0.03 (−0.11, 0.17)	0.03 (−0.05, 0.11)	0.9	−0.04 (−0.18, 0.10)	−0.03 (−0.11, 0.05)	0.9
Total-c, per 25.7 mg/dL	Model 1	0.12 (−0.01, 0.25)	0.04 (−0.03, 0.12)	0.3	0.03 (−0.11, 0.16)	0.00 (−0.08, 0.08)	0.7
	Model 2	0.10 (−0.03, 0.23)	0.028 (−0.05, 0.11)	0.3	0.02 (−0.12, 0.15)	−0.03 (−0.1, 0.05)	0.5
SBP, per 10 mmHg	Model 1	0.18 (0.05, 0.31)*	0.20 (0.11, 0.30)**	0.8	0.27 (0.14, 0.40)**	0.34 (0.25, 0.44)**	0.3
	Model 2	0.18 (0.03, 0.33)*	0.19 (0.09, 0.29)**	0.9	0.23 (0.08, 0.37)*	0.28 (0.19, 0.38)**	0.5
DBP, per 7.3 mmHg	Model 1	0.21 (0.07, 0.35)*	0.20 (0.11, 0.29)**	0.9	0.26 (0.12, 0.40)**	0.25 (0.16, 0.34)**	0.9
	Model 2	0.21 (0.05, 0.38)*	0.18 (0.09, 0.28)*	0.8	0.18 (0.02, 0.35)*	0.17 (0.08, 0.27)*	0.9
BMI, per 6.9 kg/m 2	Model 1	0.01 (−0.13, 0.14)	0.08 (−0.01, 0.18)	0.3	0.19 (0.05, 0.32)*	0.23 (0.14, 0.33)**	0.6
	Model 2	0.08 (−0.12, 0.27)	0.06 (−0.07, 0.19)	0.8	0.09 (−0.10, 0.29)	0.15 (0.02, 0.28)*	0.6

Model 1. Adjusted for mid-life age, sex, and race.

Model 2. Additionally adjusted for mid-life smoking status, obesity status, and medications for diabetes, hypertension and lipidemia, and heart rate.

^*<.05;

^**<.005

^#P for the interaction of AUC of risk factor and glycemic status

Table 3 presents results from the adjusted logistic regression models that examined the cumulative effects of risk factors on the risk of LV remodeling patterns. Odds ratios (OR) >1 indicates a positive association between AUC of risk factors and LV remodeling patterns. ORs <1 indicated negative associations between the AUC of risk factors and the outcomes. In the fully adjusted models (model 2), we found that one-SD increase of AUC of log (FG) (per 0.17) (OR 2.22, 95% CI 1.25–3.93), log (TG) (per 0.47) (3.4, 1.8–6.4), and total-c (1.85, 1.1–3.07) were statistically significantly associated with higher risk of concentric hypertrophy in early-onset T2D/pre-DM group. Greater AUC of HDL-c (per 13.5 mg/dL) (0.23, 0.09–0.59) was inversely associated with the risk of concentric hypertrophy in the early-onset T2D/pre-DM group. For most variables, the cumulative risk factor burden did not appear to affect the risk of eccentric hypertrophy or concentric remodeling in the fully adjusted models. Greater AUC of SBP (per 10 mmHg) (1.64, 1.22–2.22) and DBP (per 7.3 mmHg) (1.38, 1.04–1.82) was associated with a higher risk of eccentric hypertrophy in the euglycemia group. The associations between BP and the risk of eccentric hypertrophy or concentric remodeling did not reach statistical significance in the early-onset T2D/Pre-DM group. In the sex-stratified analysis (Supplemental Table 2), we found greater AUC of log(FG) was more strongly associated with the risk of concentric hypertrophy in women than in men with early-onset T2D/pre-DM (3.56,1.27–10 in women vs. 1.34, 0.5–3.52 in men, p-interaction=0.2).

Table 3.

Adjusted Odds Ratios (95% CI) of Each Unit Increase in AUCs of Risk Factors and LV Remodeling Patterns in Participants with or Without Early-Onset Diabetes or Pre-Diabetes

		Concentric LVH (OR)			Concentric Remodeling (OR)			Eccentric LVH (OR)
		Early T2D/PreDM	Euglycemia	P #	Early T2D/PreDM	Euglycemia	P #	Early T2D/PreDM	Euglycemia	P #
Log (FG), per 0.17	Model 1	1.38 (1.26, 1.51)**	4.84 (2.56, 9.17) **	<.005	1.23 (1.12, 1.35)**	0.78 (0.52, 1.2)	<.05	0.95 (0.89, 1.02)	0.73 (0.62, 0.85) *	<.05
	Model 2	2.22 (1.25, 3.93)*	3.3 (0.14, 78)	0.8	1.56 (0.86, 2.84)	0.7 (0.1, 4.6)	0.4	0.95 (0.63, 1.44)	0.53 (0.25, 1.1)	0.2
Log (Insulin), per 0.43	Model 1	1.38 (1.21, 1.57)**	2.30 (1.88, 2.81) **	<.005	0.98 (0.87, 1.11)	1.6 (1.3, 1.9) **	<.005	1.37 (1.27, 1.48)**	1.6 (1.5, 1.7) **	<.005
	Model 2	1.47 (0.72, 3.02)	1.81 (0.58, 5.5)	0.7	1.37 (0.62, 3.03)	1.6 (0.56, 4.81)	0.8	1.06 (0.69, 1.62)	1.26 (0.9, 1.78)	0.1
Log (TG), per 0.47	Model 1	2.54 (2.28, 2.84)**	1.95 (1.6, 2.4) **	<.05	1.10 (0.99, 1.21)	1.12 (0.91, 1.4)	0.8	0.87 (0.81, 0.93)**	1.4 (1.30, 1.48) **	<.005
	Model 2	3.40 (1.81, 6.36)**	1.53 (0.55, 4.26)	0.2	1.01 (0.59, 1.73)	1.04 (0.37, 2.9)	0.9	0.77 (0.54, 1.10)	1.18 (0.85, 1.63)	0.08
HDL-c, per 13.5 mg/dL	Model 1	0.36 (0.30, 0.43)**	0.25 (0.19, 0.3) **	<.05	1.10 (0.96, 1.27)	0.7 (0.6, 0.8) **	<.005	0.87 (0.81, 0.94)**	0.82 (0.78, 0.9) **	0.2
	Model 2	0.23 (0.09, 0.59)*	0.38 (0.12, 1.18)	0.5	1.06 (0.52, 2.16)	0.7 (0.27, 1.8)	0.5	1.03 (0.69, 1.55)	0.96 (0.72, 1.27)	0.7
LDL-c, per 23.9 mg/dL	Model 1	1.25 (1.11, 1.41)**	1.57 (1.36, 1.8) **	<.05	1.15 (1.02, 1.30)*	1.04 (0.9, 1.2)	0.3	0.86 (0.81, 0.92)**	1.05 (0.99, 1.10)	<.005
	Model 2	1.45 (0.79, 2.66)	1.76 (0.83, 3.76)	0.7	1.25 (0.70, 2.24)	1.02 (0.5, 2.1)	0.7	0.75 (0.54, 1.06)	0.94 (0.74, 1.19)	0.3
Total-c, per 25.7 mg/dL	Model 1	1.68 (1.51, 1.85)**	1.21 (1.05, 1.40) *	<.005	1.14 (1.02, 1.27)*	0.94 (0.81, 1.1)	<.05	0.76 (0.71, 0.81)**	1.02 (0.97, 1.07)	<.005
	Model 2	1.85 (1.11, 3.07)*	1.3 (0.63, 2.66)	0.3	1.19 (0.69, 2.03)	0.92 (0.45, 1.9)	0.6	0.69 (0.49, 0.98)*	0.95 (0.75, 1.19)	0.1
SBP, per 10 mmHg	Model 1	1.75 (1.60, 1.92)**	3.05 (2.59, 3.6) **	<.005	0.87 (0.77, 0.99)*	2.1 (1.8, 2.5) **	<.005	1.25 (1.18, 1.33)**	1.8 (1.71, 1.9) **	<.005
	Model 2	2.00 (1.17, 3.40)*	2.26 (0.98, 5.22)	0.8	0.96 (0.45, 2.07)	2.3 (0.91, 5.6)	0.2	1.20 (0.86, 1.68)	1.64 (1.22, 2.22)*	0.2
DBP, per 7.3 mmHg	Model 1	1.43 (1.28, 1.60)**	2.48 (2.10, 2.93) **	<.005	0.92 (0.81, 1.05)	2.0 (1.7, 2.4) **	<.005	1.37 (1.28, 1.47)**	1.59 (1.5, 1.7) **	<.005
	Model 2	1.42 (0.76, 2.65)	1.75 (0.74, 4.2)	0.8	0.88 (0.42, 1.81)	2.0 (0.89, 4.9)	0.1	1.28 (0.88, 1.87)	1.38 (1.04, 1.82)*	0.8
BMI, per 6.9 kg/m 2	Model 1	0.92 (0.82, 1.03)	2.14 (1.83, 2.49) **	<.005	0.95 (0.84, 1.08)	1.05 (0.84, 1.3)	0.4	1.62 (1.51, 1.73)**	1.45 (1.37, 1.5) **	<.05
	Model 2	0.81 (0.38, 1.70)	1.58 (0.59, 4.25)	0.2	1.15 (0.47, 2.80)	0.79 (0.2, 3.17)	0.7	1.50 (0.97, 2.33)	1.05 (0.72, 1.55)	0.2

Model 1. Adjusted for mid-life age, sex, and race.

Model 2. Additionally adjusted for adjusted for mid-life smoking status, obesity status, medications for hypertension and lipidemia, and resting heart rate.

^*<.05

^**<.005

^#P for the interaction of AUC of risk factor and glycemic status

Discussion

A body of research has shown positive associations of diabetes with LVMI and RWT. In the Framingham Heart Study, participants with T2D experienced a steeper increase in LV mass over time compared to those without T2D. ⁵ The Treatment Options for Type 2 Diabetes Mellitus in Adolescents and Youth (TODAY) study observed a significant adverse change in cardiac structure among those with T2D from youth compared to their obese or non-obese euglycemia counterparts.⁸ A prior CARDIA study reported that a longer duration of diabetes exposure beginning in young adulthood was associated with adverse LV remodeling in middle age. ⁷ Despite the evidence supporting the link between diabetes and LV remodeling, the mechanism for the adverse impact of early-onset T2D/pre-DM on LV structure has not been fully elucidated. It is unclear to what extent the metabolic risk factors clustering with hyperglycemia and insulin resistance are differentially associated with LV geometric abnormalities. Our study filled the research gap by examining the cumulative effect of a series of common risk factors that are associated with LV geometric alternations accounting for mid-life confounding factors such as medications for diabetes, hypertension, lipids, and obesity status. Our principal findings include the significant longitudinal association between hyperglycemia and insulin resistance with RWT but not with LVMI in both early-onset T2D/pre-DM and euglycemic groups, the association of dyslipidemia, particularly cumulative increase of TG and decrease of HDL-c and LV structural abnormalities among young participants with T2D/pre-DM, and a stronger association of cumulative risk factors with concentric hypertrophy than with eccentric hypertrophy or concentric remodeling in participants with and without early-onset T2D/pre-DM.

We found that cumulative exposure to FG was a significant predictor of increased RWT in both glycemic groups, with statistical significance shown in the early-onset T2D/pre-DM group. Our finding was in accordance with previous population studies that reported glucose and insulin concentrations were correlated with RWT but not with LVMI. ^{15, 28, 29} Additionally, research has shown that insulin resistance determined by hyperinsulinemia euglycemic clamp or impaired glucose tolerance by an OGTT is related to thick LV walls rather than to LV hypertrophy. ^30–32 The reasons for this difference are not fully understood but may be related to insulin’s trophic effect. Insulin has been regarded as a trophic factor responsible for the development of cardiovascular hypertrophy. Animal studies have shown that chronic moderate hyperinsulinemia, while maintaining control of hormones with effects opposing insulin, resulted in a pronounced hypertrophy of cardiac ventricles.³³ In view of the results from our study, the trophic effect might, in humans, mainly influence ventricular wall thickness, leaving cavity dimensions potentially unaffected.

Among many modifiable risk factors, we found cumulative lipid dysregulation, characterized by higher TG and lower HDL-c from young adulthood to mid-adulthood, is associated with LV structural abnormalities, particularly in participants with early-onset T2D/pre-DM. High TG and low HDL-c is a classic lipid phenotype in T2D. Elevated TG is considered the dominant lipid abnormality in insulin resistance and plays a pivotal role in determining the characteristic lipid profile of diabetic dyslipidemia. ³⁴ Elevated TG concentrations are the result of increased production and decreased clearance of TG-rich lipoproteins in both fasting and non-fasting states. ³⁴ An increase in TG-rich lipoproteins is associated with a reduction in HDL and an increase in small dense LDL concentrations. Diabetic dyslipidemia is significantly associated with the risk of CVD. ³⁴ Apart from indirectly affecting the function of the heart by promoting the development of atherosclerosis, diabetic dyslipidemia affects the cardiac electrophysiological response of the heart directly, which may be related to the gradual accumulation of cardiac lipids, and consequent systemic oxidative stress, proinflammatory state, and mitochondrial dysfunction. ³⁵ However, the mechanism underlying the direct effects of diabetic dyslipidemia on the heart is not fully understood. Our results pointed to the need for additional studies to substantiate the link between diabetic dyslipidemia and LV structural abnormalities in those with early-onset T2D/pre-DM. Further research is also needed to validate whether intensified lipid-control is beneficial in young adults with T2D.

Additionally, our study highlighted the significant role of longitudinal elevated BP in greater LVMI and RWT, irrespective of T2D/pre-DM status. Chronic hypertension and the resulting increased hemodynamic load are a major risk factor for cardiac remodeling. ^{24, 36, 37} Our finding extends the evidence that long-term elevation of BP was associated with adverse LV geometric changes independent of diabetes status. We also found the cumulative BMI was significantly associated with increased LVMI even accounting for the middle-age obesity status. Excess adiposity affects heart size through hemodynamic, metabolic, and inflammatory alternations, leading to LV enlargement and LVH. ²⁴ Our result again noted the importance of weight management from earlier time regardless of the diabetes status.

Another important finding of our study is that cumulative metabolic risk factors were associated with concentric hypertrophy among all included participants, with statistically significant associations mostly observed in the early-onset T2D/pre-DM group. Among the common modifiable risk factors, unfavorable lipid changes, cumulative increased FG and SBP are the most significant longitudinal risk factors to concentric hypertrophy at mid-life among those with early-onset T2D/pre-DM. Our finding aligns with a body of studies that have demonstrated the association between diabetes and increased concentric LV geometry. ^{15, 28, 29} In contrast, cumulative metabolic risk factors on eccentric hypertrophy or concentric remodeling were less pronounced in our study. The pathophysiological mechanism to explain the stronger link between early onset T2D/pre-DM and concentric hypertrophy than other geometric remodeling patterns is not well understood and needs to be delineated in further studies.

Our study demonstrated that cumulative exposure to metabolic risk factors was associated with increased RWT and LVMI in all participants. Individuals with early-onset T2D/pre were particularly at risk of adverse LV structural changes due to chronic hyperglycemia, insulin resistance, and diabetic dyslipidemia. While the exact mechanisms underlying these associations remain fully elucidated, our findings suggest the importance of aggressive management of modifiable risk factors for those with early-onset T2D/pre-DM to prevent adverse LV remodeling. Given the increasing prevalence of early-onset T2D, additional research is needed to assess the impact of lifestyle interventions or pharmacotherapy on LV remodeling in this patient population.

Strengths of the study include the repeated measures of risk factors from young adulthood to mid-adulthood, which enables assessment of the longitudinal impact of these risk factors on LV structural outcomes. Large sample size and serial screening of multiple measures of glycemic status allowed us to identify sex- and race-balanced early-onset T2D/pre-DM subjects and matched euglycemia comparison group. However, it’s essential to acknowledge the limitations of our study. We excluded around 38 early-onset T2D cases (12%) after Y10 (e.g., had a diagnosis or elevated glucose before 40 years old at Y15), because a minimum of 4 risk factor measures was needed for the quadratic random-effects models to examine the association of AUCs of risk factors with echocardiographic outcomes at Y25. In a separate analysis, including early-onset T2D cases at Y15 and using an extended follow-up to Y30, we found nearly identical AUCs to our primary results. Our selection method may potentially introduce bias. However, the bias may be minimal given the small number of exclusion cases and our focus on the demographics-matched case-control design. Furthermore, insulin use was not available across cycles of CARDIA, so we were not able to account for the insulin effect on the associations. Additionally, because of the limited number of individuals in sex and race subgroups, the stratified results should be interpreted with caution, particularly results from logistic regressions for the association between AUCs of risk factors and LV remodeling patterns in the early-onset T2D/pre-DM group. Further, while T2D predominated in the CARDIA study, a small fraction of participants may have developed T1D, especially in the early-onset group. The CARDIA study did not collect the autoimmune markers to determine T1D; it is impossible to discriminate between T1D and T2D within the CARDIA data.

Conclusion

Based on this population-based bi-racial young adult cohort, we characterized the association between cumulative modifiable risk factors and LV remodeling through adulthood among individuals with and without early-onset T2D/pre-DM. Our findings suggest that sustained hyperglycemia and hyperinsulinemia are associated with RWT in participants with or without early-onset T2D/pre-DM at mid-life. However, those with early T2D/pre-DM were at excess risk potentially due to their higher levels of glucose and insulin. In addition, we demonstrated that dyslipidemia is a risk factor for LV structural abnormalities, particularly in early-onset T2D/pre-DM. Further research is warranted to elucidate the mechanistic links between diabetic dyslipidemia and LV geometric alterations and to explore potential targeted management strategies in young adults with T2D/pre-DM. Longitudinal studies with larger sample sizes are needed to confirm our findings and investigate the effectiveness of early intervention strategies in mitigating cardiovascular risk in this population.