Characterizing Lipoprotein(a) in Children With New-Onset Diabetes and Implications for Cardiovascular Risk Assessment

Abstract

Context

Children with diabetes mellitus (DM) have an increased risk for cardiovascular disease (CVD), a risk potentially exacerbated by elevated lipoprotein(a) (Lp(a)). While other cholesterol parameters are screened in this population, Lp(a) is often overlooked despite being an independent CVD risk factor. Lp(a) levels are historically believed to not change over an individual's life and are genetically determined, but newer literature suggests variation.

Objective

This study investigated Lp(a) levels and their relationship with glycated hemoglobin A_1c (HbA_1c) in children with incident diabetes mellitus (DM).

Methods

Children and adolescents aged 5 to 18 years with incident DM had baseline Lp(a) and lipid profiles. Repeat Lp(a) and HbA_1c were obtained 3 months later. Descriptive statistics (frequencies, proportions, means, medians) and nonparametric tests (Spearman correlation, Wilcoxon rank-sum/Kruskal-Wallis) were used. Statistical significance was set at P less than .05.

Results

Seventy-six children were included for evaluation: 76% with type 1% and 23% type 2 DM. Baseline median (Q1-Q3) Lp(a) was 43.3 nmol/L (13-73.7 nmol/L), 17 of which were elevated (≥75 nmol/L). Of the 22 participants with follow-up, 8 were abnormal: A total of 4 whose baseline Lp(a) were abnormal remained so and 4 with normal levels became abnormal. A positive correlation was found between 3-month Lp(a) values and HbA_1c (P = .004).

Conclusion

Children with DM have abnormal Lp(a) levels at a prevalence of approximately 20%, so this should be considered in CVD risk stratification. Further, observed Lp(a) fluctuations suggest value in serial Lp(a) assessments due to nongenetic influences. Without Lp(a) quantification, CVD risk characterization in children with DM may be inaccurate and should be considered for a comprehensive assessment.

Individuals with diabetes mellitus (DM) have a 2- to 4-fold increased risk of cardiovascular disease (CVD) compared to the general population [1]. As knowledge about CVD risk factors expands, it underscores the necessity of improving both the management of underlying DM and lifestyle interventions from an early age to prevent the onset and progression of CVD.

While dyslipidemia, characterized by high low-density lipoprotein (LDL) cholesterol and low high-density lipoprotein (HDL) cholesterol, and suboptimal glycemic control (as reflected by elevated glycated hemoglobin A_1c [HbA_1c]), are well established to contribute to CVD risk, lipoprotein(a) (Lp(a)) is an independent risk factor largely overlooked [1-3]. Established guidelines exist for screening and managing LDL and HDL in pediatric and DM populations, but similar guidance for Lp(a) is lacking [4, 5].

Lp(a) levels are thought to be predominantly genetically determined, established early in childhood, and stable into adulthood [2]. However, a growing body of literature suggests that levels can change and that nongenetic factors influence an individual's Lp(a) level both positively and negatively [6, 7].

The goal of this pilot feasibility study was to characterize Lp(a) levels in DM, a population inherently at risk for CVD, as a potential proposed adjunct parameter in evaluating CVD risk. This study will contribute to understanding the prevalence of elevated Lp(a) in this at-risk population, assess the effect of standard DM care on Lp(a) levels, and evaluate its potential utility in CVD risk stratification and dyslipidemia management for children with DM.

Materials and Methods

Children and adolescents aged 5 to 18 years with incident type 1 or type 2 DM referred between January 9, 2024, and March 6, 2025, to the pediatric endocrine division at a single-center tertiary hospital were eligible for inclusion. Exclusion criteria were children younger than 5 years or adolescents older than 18 years, diagnosis of forms of diabetes other than type 1 and type 2, and failure to obtain initial bloodwork. A diagnosis of type 2 DM was differentiated from type 1 DM based on negative autoantibodies and/or clinical characteristics consistent with insulin resistance, such as acanthosis nigricans and obesity. Baseline Lp(a) and lipid profiles (total cholesterol [total C], HDL, LDL) were obtained at diagnosis (unless presenting in diabetic ketoacidosis [DKA], in which case samples were collected at the first clinic follow-up visit after DKA resolution within ∼1 month). Demographic and clinical data were collected at diagnosis. All participants with initial Lp(a) levels were eligible to then have repeat Lp(a) and HbA_1c measurements taken approximately 3 months after baseline. Lp(a) levels greater than or equal to 75 nmol/L were considered elevated based on consensus data that less than 75 nmol/L represents low risk and above enters intermediate and high-risk range [8]. The lower limit of Lp(a) quantification was 9.0 nmol/L and the upper limit of HbA_1c quantification was 14.0% due to assay limitations. All samples were processed in a single laboratory with the same assay techniques. Laboratory assays measured as follows: Lp(a) via an immunoturbidimetric assay (RRID:AB 3698038), HbA_1c via Tina-quant Hemoglobin A1cDx Gen.3 (RRID:AB 2909459) or via a point-of-care machine using a boronate affinity assay, and lipid profile via enzyme-linked immunoassay with LDL calculated via the Sampson/National Institute of Health equation.

Descriptive statistics (frequencies and proportions for categorical data, as well as means ± SDs or medians [Q1-Q3]) were computed. The Spearman rank correlation was used to assess correlations between continuous variables and the Wilcoxon rank-sum test or Kruskal-Wallis test to compare distributions of continuous variables with categorical variables of interest; these nonparametric tests were used when the normality assumptions were not met. For all analyses, results yielding P values less than .05 were considered statistically significant. All analyses were conducted using SAS version 9.4 (SAS Institute Inc). Data were graphed using GraphPad Prism 10.4.1 and Microsoft Excel.

Results

Demographic and Clinical Characteristics

Seventy-six children with new-onset type 1 or type 2 DM (n = 58 [76.3%] type 1 and n = 18 [23.7%] type 2) were included for initial evaluation. Of these, 37 (48.7%) were male and 39 (51.3%) were female. Racial/ethnic distribution was White (n = 24, 31.6%), African American (n = 18, 23.7%), Asian (n = 12, 15.8%), Hispanic/Latino (n = 8, 10.5%), and unknown race (n = 14, 18.4%). Mean ± SD age was 12.0 ± 3.7 years. Diabetes autoantibody positivity was as follows: glutamic acid decarboxylase (GAD; n = 48, 63.2%), zinc transporter (Zinc Trans; n = 48, 63.2%), islet antigen 2 antibody (IA2; n = 31, 40.8%), islet cell antigen (ICA; n = 29, 38.2%), and insulin antibody (IAb; n = 26, 34.2%). Thirteen participants (17.1%) had no detectable autoantibodies. Median (Q1-Q3) HbA_1c was 11.4% (9.7%-13.5%) (Table 1). Seven individuals had 1 positive autoantibody (5 with GAD, 1 with Zinc Trans, 1 with IAb) but were diagnosed as type 2 DM clinically, and 2 individuals were autoantibody negative but were diagnosed with type 1 DM clinically based on age, presentation, and body habitus. Sixty-seven individuals were treated with insulin therapy while 2 were treated with metformin monotherapy and 7 were recommended lifestyle interventions.

Table 1.

Characteristics of incident diabetes

Characteristics*
Sex
Female	39 (51.3%)
Male	37 (48.7%)
Type of diabetes
Type 1	58 (76.3%)
Type 2	18 (23.7%)
Race
White	24 (31.6%)
AA	18 (23.7%)
Asian	12 (15.8%)
Hisp/Latino	8 (10.5%)
Unknown	14 (18.4%)
Positive antibodies
GAD	48 (63.2%)
Zinc Trans	48 (63.2%)
IA2	31 (40.8%)
ICA	29 (38.2%)
IAb	26 (34.2%)
None	13 (17.1%)
Lp(a), nmol/L	43.3 (13-73.7)
HbA1c, %	11.4 (9.7-13.5)
Lipid profile, mg/dL
Total C	172 (152-207)
LDL	97 (67-120)
HDL	49 (41-61)

Categorical variables are presented as frequencies (%) and continuous variables as median (Q1-Q3).

Abbreviations: AA, African American; GAD, glutamic acid decarboxylase; HDL, high-density lipoprotein; HbA_1c, glycated hemoglobin A_1c; Hisp, Hispanic; IAb, insulin antibody; ICA, islet cell antibodies; LDL, low-density lipoprotein; Lp(a), lipoprotein(a); Total C, total cholesterol; Zinc Trans, zinc transporter.

^*Total n = 76.

Baseline Data

Median (Q1-Q3) baseline total C, LDL, and HDL were 172 mg/dL (152-207 mg/dL), 97 mg/dL (67-120 mg/dL), and 49 mg/dL (41-61 mg/dL), respectively. Median initial Lp(a) was 43.3 nmol/L (13-73.7) (Fig. 1). Seventeen participants (22.4%) presented with an elevated Lp(a) (≥75 nmol/L), 10 (59%) of whom had type 1 DM. A very weak, negligible correlation was observed between initial Lp(a) and HbA_1c (r = 0.04; P = .76); though technically positive, the strength of the relationship between Lp(a) and HbA_1c is minimal. Initial Lp(a) showed a weak but statistically significant positive correlation with total cholesterol (r = 0.25; P = .04), but not with LDL or HDL. No significant correlations were found between baseline demographic characteristics, DM type, autoantibody status, or initial Lp(a) levels.

Figure 1.

Distribution of A, initial lipoprotein(a) (Lp(a)); B, initial total cholesterol (total C); C, initial low-density lipoprotein (LDL); and D, initial high-density lipoprotein (HDL).

Follow-Up Data

Twenty-two children continued on to follow-up assessments at a median (Q1-Q3) of 113 days (94-125 days) after baseline. Of these, 45% (n = 10) were female and 95% (n = 21) had type 1 DM. Median follow-up Lp(a) was 52.7 nmol/L (14.9-86.5 nmol/L) and median follow-up HbA1c was 6.9% (6.3%-8.4%), with a mean decrease in HbA_1c of 4.3% (range, 0%-7.9%). A moderate, statistically significant positive correlation was observed between follow-up Lp(a) and HbA_1c (r = 0.59; P = .004). The change (Δ) in Lp(a) ranged from −167.8 nmol/L to +50 nmol/L (mean absolute Δ = 22.8 nmol/L). Percentage change in Lp(a) ranged from −61.1% to +115.6% (mean absolute percentage Δ = 12.1%) (Fig. 2). Of the 22 follow-up Lp(a) values, 8 (36%) were 75 nmol/L or greater. Of these, 4 had decreased from initially high/abnormal levels but remained abnormal (mean decrease = 55.8 nmol/L), while 4 had increased to abnormal from initially normal levels (mean increase = 35.2 nmol/L). Of the 22 follow-up Lp(a) values, a change was seen in 19 individuals, whereas 3 remained unchanged at 9 nmol/L (Fig. 3).

Figure 2.

Percentage change (δ) of A, glycated hemoglobin A_1c (HgbA_1c) and B, lipoprotein(a) (Lp(a)) with C, absolute δ of Lp(a).

Figure 3.

Comparison of baseline to follow-up lipoprotein(a) (Lp(a)) values.

Discussion

Lp(a), first described in 1963, has garnered increasing attention for its role in CVD development [9]. This plasma lipoprotein comprises an LDL-like core bound to apolipoprotein(a) and apolipoprotein B100, featuring a variable number of kringle domains [4, 6, 10]. Its encoding gene likely evolved from the plasminogen gene, resulting in structural and functional similarities [4]. This homology, along with its similarity to LDL, confers both proatherogenic and prothrombotic properties, contributing to CVD [4].

Elevated Lp(a) levels, estimated to occur in 20% of the general population, can vary up to 1000-fold between individuals [2, 11]. While Lp(a)'s precise biological function remains unclear, elevated levels are well established as a contributor to CVD, including stroke, ischemia, and atherosclerosis, independent of traditional lipid parameters [2, 4, 6, 10, 11]. Some studies suggest Lp(a) may be up to 6 times more atherogenic than LDL on a per-particle basis [12]. Individuals with both type 1 and type 2 DM experience a 2- to 4-fold increased baseline CVD risk, disproportionate to their level of dysglycemia [1]. This suggests the influence of factors beyond standardly measured parameters, with Lp(a) being a potential candidate.

This pilot study aimed to characterize baseline Lp(a) levels in children with new-onset DM and to explore whether DM treatment influences these putatively genetically determined levels. Clear pediatric and DM guidelines exist for universal cholesterol screening specifying targets for LDL and interventions for its elevation. For instance, in type 1 DM, lipid profiles are checked shortly after diagnosis, at age 9 to 11 years, and subsequently every 3 years, aiming for an LDL of 100 mg/dL or less. Statin therapy is initiated if 6 months of lifestyle intervention fails to lower LDL below 160 mg/dL [5]. However, no such guidelines exist for Lp(a). Non-US guidelines suggest a single lifetime Lp(a) measurement, given its presumed 90% genetic determination, attainment of adult levels by age 5 years, and stability despite medical status or lifestyle interventions [2, 6]. The National Lipid Association recommends targeted Lp(a) screening in youth with a family history of premature CVD, hypercholesterolemia, or elevated Lp(a), but these recommendations omit the at-risk pediatric DM population [6]. The lack of cohesive Lp(a) screening guidelines for children in the United States likely stems from limited knowledge and the absence of specific therapies for elevated levels [2, 4].

This study demonstrates that children with DM exhibit elevated Lp(a) at a prevalence similar to the general population (20% vs 22.4% found here) [2]. There was also a positive correlation found between HbA_1c and Lp(a), unexpected if Lp(a) were solely genetically determined. Alsaeid et al [3] similarly found higher mean Lp(a) levels in children with DM compared to matched controls, with 38.24% of those with poor glycemic control exhibiting elevated Lp(a) vs 12.5% with optimal control. Other studies corroborate this positive correlation between HbA₁c and Lp(a), suggesting an interplay with DM itself [13]. While not observed in this study, another study [14] demonstrated a significant correlation between percent reductions in HbA_1c and Lp(a). Similar findings were noted in 12 patients with diabetes who experienced concurrent decreases in HbA_1c and Lp(a) over 3 weeks [15]. One proposed explanation is that insulin suppresses apolipoprotein(a) synthesis, leading to higher Lp(a) levels in individuals with type 1 DM—who are insulin deficient— which subsequently decrease with insulin therapy [16]. Lp(a) levels are primarily determined by synthesis rather than clearance, so this can potentially lead to elevated levels during prolonged insulin deficiency in type 1 DM that are difficult to normalize after diagnosis and treatment [17].

Newer evidence suggests that Lp(a) levels can change with age and are influenced by nongenetic factors. These include factors that can decrease levels (hypothyroidism, growth hormone, chronic kidney disease, menopause) and those that can increase them (hyperthyroidism, hormone replacement therapy, acute illness, aging, liver disease) [6, 7, 9, 18, 19]. Dietary studies have shown that replacing saturated fatty acids with other macronutrients, while beneficial for LDL management, can increase Lp(a), highlighting that broad dietary guidelines for cholesterol management might reduce one CVD risk factor while inadvertently enhancing another [18, 20]. Additionally, Lp(a) has been found to be an acute phase reactant and thus may be under the influence of inflammation and metabolic derangement [21]. It is for this reason that in this study Lp(a) was not obtained when an individual was in DKA.

There were observed Lp(a) fluctuations over time in the cohort presented here, highlighting the ability for values to change and potential value of serial Lp(a) assessments. This is relevant for individuals near risk stratification cutoffs, as these data demonstrate the potential for risk category changes with 4 individuals having Lp(a) values that changed from normal to abnormal. There is likely a degree of interassay and intra-assay variability based on coefficients of variation of the assay itself that influences an individual's measured Lp(a). Additionally, modest intraindividual Lp(a) variability, often overlooked with single measurements, likely occurs. A trial by Marcovina et al [22] using an antisense oligonucleotide to lower Lp(a) demonstrated greater than 25% variation in serial Lp(a) values in 40% of placebo-treated individuals. This variability could affect CVD risk stratification, especially in individuals with additional risk factors like DM, when Lp(a) levels are near cutoff values. Multiple studies support this: A study of 11 669 adults with repeat Lp(a) measurements showed 51.2% with borderline initial levels changed risk stratification on retesting, Deshotels et al found 58.1% of 4734 adults with borderline-high initial Lp(a) progressed to high Lp(a) over 15 years, Harb et al observed risk category changes in 53% of 609 individuals, and Barkas et al reported risk reclassification in 60% of 1975 patients [23-26]. This has led the National Lipid Association to recommend considering not only a one-time measurement of Lp(a) but perhaps a repeated measurement if initial tests are near a risk threshold or if significant clinical changes occur [27].

Pediatric data on Lp(a) variation are more limited. One study of 3000 children in a lipid clinic revealed substantial (70%) intraindividual Lp(a) variation with repeated measurements, with most children showing at least a 20% difference between values [6]. Lp(a) increased by 22% in children not receiving lipid-lowering therapy, 43% in those on statins, and 9% in those on ezetimibe [6]. This aligns with evidence suggesting that LDL-lowering therapies, particularly statins, may counterintuitively increase Lp(a) [6]. This raises concerns about focusing solely on LDL reduction in pediatric DM, especially if it elevates CVD risk through increased Lp(a). Further, no correlation was found in these data between Lp(a) and LDL, suggesting that LDL measurement alone provides an incomplete dyslipidemia risk assessment.

As evidence mounts for the potential to modify Lp(a) levels throughout life by manipulating homeostatic factors, new avenues for lifestyle and pharmacotherapeutic interventions may emerge. While not currently available, pharmacotherapy studies targeting elevated Lp(a) are ongoing [12, 28]. While Lp(a) screening in adults helps characterize CVD risk, it is important to recognize that the risk begins accumulating in childhood, particularly as Lp(a) reaches full expression by age 5 years [2]. This is especially critical given the clinically silent nature of early CVD development in children [29].

The primary limitation of this hypothesis-generating pilot feasibility study is the small sample size for follow-up evaluation and the heterogeneity of the study population regarding demographic information and type of diabetes, limiting overall generalizability of the findings. Based on the preliminary findings of this pilot study, a larger study will continue this evaluation to include a longer time span between samples and a larger cohort to assist with validating generalizability. Further studies could include multicenter populations to assess this. Other helpful data may include markers of inflammation, initial insulin and C-peptide levels, and insulin dose requirements. Also beneficial would be correlating with other markers of CVD, such as intima-media thickness or arterial stiffness. The inability to quantify Lp(a) below 9 nmol/L and HbA_1c above 14% necessitates caution when analyzing these as continuous variables. Lastly, individuals who presented in DKA had evaluations conducted after metabolic stabilization and thus did have brief exposure to treatment of their DM prior to baseline levels being drawn, which may influence Lp(a) levels.

These data support that Lp(a) levels fluctuate in individuals and are therefore likely subject to nongenetic and modifiable influences, resulting in possible changes to an individual's CVD risk potential. This study aims to help in the development of Lp(a) monitoring guidelines in pediatric DM, ultimately improving CVD risk assessment and potentially shifting the current CVD risk management paradigm in DM. Knowledge of a child's Lp(a) status may enhance risk stratification, inform lifestyle interventions and dyslipidemia management, and potentially influence LDL target values until specific Lp(a)-lowering therapies become available. Without Lp(a) quantification, CVD risk assessment in children, especially those with DM, may be incomplete and its inclusion should be considered for a more comprehensive evaluation.

Acknowledgments

During the preparation of this work the authors used artificial intelligence–assisted technologies to edit the readability of the manuscript. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Abbreviations

CVD

cardiovascular disease

DKA

diabetic ketoacidosis

DM

diabetes mellitus

GAD

glutamic acid decarboxylase

HDL

high-density lipoprotein

HbA_1c

glycated hemoglobin A_1c

IAb

insulin antibody

ICA

islet cell antibodies

LDL

low-density lipoprotein

Lp(a)

lipoprotein(a)

total C

total cholesterol

Zinc Trans

zinc transporter

Funding

Funding for laboratory testing was provided by the Division of Pediatric Endocrinology at Cohen Children's Medical Center.

Disclosures

A.K., R.S., and P.S. have no disclosures to declare.

Data Availability

Some or all data sets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.