Effect of Carnitine Supplementation in Pediatric Patients with Left Ventricular Dysfunction

Carnitine is an essential amino acid involved in transporting fatty acids across the mitochondrial membrane. Fatty acids are a primary source of energy for the myocardium. Studies in adults demonstrated decreased carnitine levels in the ischemic myocardium, but subsequent exogenous carnitine supplementation showed improvement of myocardial metabolism and left ventricular function. However, only limited data regarding carnitine are available in pediatrics. A single-center retrospective, paired data study was conducted. Patients < 18 years, left ventricular ejection fraction (LVEF) < 55% by echocardiography, and had received at least 7 days of oral or intravenous carnitine supplementation between January 2018 and March 2021 are included in the study. Several endpoints and covariates were collected for each patient: before, one week after, one month after, and 6 months after carnitine supplementation. Univariate analysis consisted of an analysis of variance (ANOVA), followed by an analysis of covariance (ANCOVA) to model LVEF while adjusting for other variables. 44 patients included in the final analyses. LVEF significantly improved from 50.5 to 56.6% (p < 0.01). When LVEF was adjusted for other interventions (mechanical ventilation, afterload reduction, diuretic therapy, spironolactone), the estimated means demonstrated a significant increase from 45.7 to 58.0% (p < 0.01). Free carnitine level increased significantly (p = 0.03), and N-terminal-pro-brain natriuretic peptide (p = 0.03), creatinine (p < 0.01), and lactate (p < 0.01) all significantly decreased over the study period. Carnitine supplementation in pediatric patients with left ventricular systolic dysfunction may be associated with an increase in LVEF and improvement in laboratory markers of myocardial stress and cardiac output.


Introduction
Carnitine is an essential amino acid that is synthesized endogenously in the liver, kidney, and brain from lysine or methionine. Carnitine plays an important role in transporting fatty acids across the mitochondrial membrane [1]. Fatty acid serves as the primary source of energy for cardiac muscle and other vital organs [2]. In adults there are studies that have demonstrated significantly lower myocardial carnitine levels during periods of ischemia. These studies also noted that exogenous carnitine supplementation was associated with improvement in myocardial metabolism and left ventricular function [3,4]. In the pediatric population, however, there are limited data regarding the effect of carnitine on myocardial function.
Loomba et al. conducted a systematic review and metaanalysis study of studies regarding carnitine supplementation in the pediatric population and found only 6 studies with total of 144 patients [5]. They found a statistically significant increase in LVEF and shortening fraction after carnitine supplementation.
The primary aim of this study was to characterize LVEF at various time points before and after carnitine supplementation in pediatric patients with left ventricular dysfunction. Secondary aims included characterizing other echocardiographic, laboratory, and hemodynamic indices at similar time points.

Study Design
This was a single-center retrospective study following patients longitudinally. Data were collected for predefined variables at each time point for comparison between time points of the same patient. Mediation analysis was used to try to isolate the effect of the intervention.
This study was approved by the local institutional review board and is in concordance with the Helsinki Declaration.

Patient Identification
The following inclusion criteria were used for this study: (1) pediatric patients (under 18 years of age); (2) carnitine supplementation initiated while admitted to Advocate Children's Hospital in Oak Lawn, IL; (3) left ventricular dysfunction defined by ejection fraction (EF) < 55% documented by echocardiography; (4) available echocardiographic images from which an ejection fraction could be estimated; and (5) received at least 7 days of carnitine either orally or intravenously. Patients who received carnitine supplementation initiated while having normal LVEF (≥ 55%) and those in whom carnitine supplementation was initiated while on mechanical circulatory support were excluded from this study.

Variables of Interest
LVEF was the primary endpoint of this study. Other echocardiographic findings (shortening fraction, strain), laboratory findings (N-terminal-pro-brain natriuretic peptide, troponin, creatinine, blood urea nitrogen, aspartate aminotransferase, alanine aminotransferase, and lactate), and hemodynamics (heart rate, systolic blood pressure, diastolic blood pressure, cerebral near-infrared spectroscopy, and renal nearinfrared spectroscopy) were collected for secondary endpoints. Other covariables, felt to influence LVEF, were also collected. These included the dose of carnitine (daily dose per kilogram), intravenous carnitine (yes/no), mechanical ventilation (yes/no), vasoinotrope score, afterload reduction medications (yes/no), diuretics (yes/no), spironolactone (yes/ no), and total and free carnitine levels. Data collection time points included immediately prior to carnitine supplementation, one week after carnitine supplementation (± 2 days), one month after carnitine supplementation (± 1 week), and six months after carnitine supplementation (± 1 month).

Statistical Analysis
Normalcy of distribution of data were assessed using skewness and kurtosis. For continuous variables, data were reported as mean and standard deviation if normally distributed and as median and range if not normally distributed. Descriptive data were reported as absolute frequency and percent. Unadjusted, univariate analyses for continuous variables were compared across time points using a repeated measures analysis of variance analysis. Unadjusted, univariate analyses for descriptive variables were conducted using a Fisher's test.
As change in LVEF was of primary interest, an adjusted analysis was conducted using a repeated measures analysis of covariance. LVEF was the dependent variable in this model with need for mechanical ventilation, afterload reduction, diuretic therapy, and spironolactone therapy entered as independent variables. A Bonferroni correction was used. Mean estimates for LVEF were then tabulated for each time point.
Next, a linear regression analysis was conducted with the absolute ejection fraction as the dependent variable. Need for mechanical ventilation, afterload reduction, diuretic therapy, and spironolactone therapy were entered as independent variables. Each data point was treated as a separate case and the time point at which that data were collected was also entered as an independent variable. Time was entered in weeks with baseline being 0, 1 week being 1 week, 1 month being 4 weeks, and 6 months being 24 weeks. The purpose of this particular analysis was to determine the association of time with ejection fraction.
Finally, a continuous variable was created to represent the total change in LVEF between the baseline and 6-month follow-up data (LVEF at 6 months minus LVEF at baseline). This was then used as the dependent variable in a paired linear regression analysis with all baseline and 6-month followup variables with a p-value of less than 0.20 by univariate analyses entered as independent variables. The regression was conducted using backward elimination with a likelihood ratio method of variable selection. This strategy allowed for both a priori selection and stepwise selection to help result in a robust, reproducible model.

Mediation Analysis
The first regression, a paired regression, modeled the ejection fraction using the selected variables of interest. The data here were analyzed at a patient level with each patient having an ejection fraction at multiple times points. Ultimately, the effects of the following variables could not be estimated using this regression: time and carnitine supplementation.
The second regression, a multivariable linear regression, modeled the ejection fraction using the selected variables of interest as well. The data here, however, were analyzed with each time point being a separate "subject." This allowed for the time point to be entered as an independent variable in the regression. This was not possible in the first regression. Ultimately, the effects of the following variables could not be estimated using this regression: carnitine supplementation.
The importance of conducting both regressions was that the second regression allowed for quantification of the effect of time. Thus, the adjusted effect on ejection fraction could further be adjusted using the results of the second regression to help adjust for the effect of time. This was truly a mediation analysis in which the effect of carnitine supplementation on ejection fraction is modeled with time as a mediator. The multiple regression approach of mediation analysis was used for this study with the individual regressions having been detailed above for clarity and transparency. Further mathematical details of such an analysis is beyond the scope of this paper but can be found elsewhere, specifically for time series analysis [6]. The use of such an approach for observational studies is previously described [7]. The use of such an approach specifically for studies with longitudinal data with an endogenous exposure is described by Bind and colleagues, while others have described the approach with observational studies in general [8][9][10][11]. A recently published pediatric cardiology study by Savla and colleagues used causal-mediated analysis as well [12]. This approach has benefits over using a historic cohort for comparison as other uncaptured variables may differ within the two cohorts which may make the use of historic data less accurate. This two regression approach allows for the estimation of the change in ejection fraction over time in the cohort of interest itself.
All statistical analyses were conducted using SPSS Version 23.0. A p-value of 0.05 was considered statistically significant. Any use of "significant," "significantly," or "significance" in this manuscript refers to statistical significance unless otherwise explicitly stated.

Cohort Information
A total of 44 patients were included in the final analyses. There were 25 males and 19 females in the group. Neonates were defined as less than 1 month of age, infants between 1 month and 1 year of age, and children between 1 and 18 years of age. There were 27 total patients with primary diagnosis of congenital heart disease (CHD) ( Table 1). Table 2 demonstrates the breakdowns of 27 patients with primary diagnosis of CHD. Mean age at time of carnitine initiation was 28 months (2 years). There were a large range of age, however, with neonates to 17-year olds in the cohort. Mean weight at first dose was 14.6 kg. The daily dose of carnitine on initiation was 65.4 mg/kg/day (Table 3).

Concomitant Interventions
There was attrition throughout the follow-up due to varying follow-up status. At the baseline time point there were 44 patients, at the 1-week time point there were 33, at the 1-month follow-up there were 24, and at 6-month followup there were 17. Interventions, other than carnitine, were divided into five large groups: mechanical ventilation, vasoactive medications (Epinephrine, Norepinephrine, Dopamine, Dobutamine, Milrinone), afterload reduction, diuretic therapy, and Spironolactone. Spironolactone was placed in a separate group as it is not used for diuretic effect in this population at our institution. Tables 4 and 5 summarize the proportion of patients who received various interventions at each time point. Mechanical ventilation was used less at 6 months after carnitine initiation when compared to earlier time points (p < 0.01). Afterload reduction was increasingly used over the study period (p < 0.01). Diuretic and Spironolactone use did not significantly change over the study period (Table 4). Vasoactive support is described as vasoinotrope score in Table 5. Vasoactive medications were also decreasingly used throughout the study period (p = 0.04). Table 6 summarizes the changes in echocardiographic indices at the various time points. LVEF significantly improved over the study period, increasing from 50.5 to 56.6% (p < 0.01). When ejection fraction was adjusted for Left ventricular shortening fraction did not demonstrate significant change, increasing from 28.5 to 29.8% over the study period (p = 0.10).

Echocardiographic Changes After Carnitine Initiation
Left ventricular global longitudinal strain improved over the study period, increasing from − 9.3 to − 17.2%, although did not reach statistical significance (p = 0.10).

Laboratory Marker Changes After Carnitine Initiation
Free carnitine level significantly increased over the study period compared to baseline (p = 0.03). N-terminal-probrain natriuretic peptide (p = 0.03), creatinine (p < 0.01), and lactate (p < 0.01) all significantly decreased over the study period (Table 5).

Hemodynamic Changes After Carnitine Change
There was no statistical significance on heart rate or blood pressure (both systolic and diastolic) after carnitine supplementation (p = 0.42, 0.72, and 0.46, respectively).
There were no significant changes in near-infrared spectroscopy values.

Regression Analyses
The regression analysis was done to model ejection fraction with time point included as an independent variable demonstrated that time was not significantly associated with ejection fraction (beta-coefficient 2.5, p-value 0.08). This demonstrates that over the total study time period that there was likely a total change of 7.5 attributable to time if time had come out to be statistically significant. Although it was not significant, even if the overall change in ejection fraction is still adjusted for this that results in a 4.7% increase in ejection fraction over the study time period from baseline to 6-month follow-up that is isolated to carnitine or unaccounted factors.
Paired regression analysis done to model change in ejection fraction from baseline to the 6-month time point demonstrated that a lower baseline ejection fraction was associated with a greater increase in ejection fraction. For every 1 lower the ejection fraction (%) at baseline there was 0.6 greater increase in the ejection fraction over the 6-month time period. The other independent variables were not found to be independently associated with the change in ejection fraction.

Discussion
This study demonstrated carnitine may be associated with an increase in LVEF in pediatric patients with left ventricular systolic dysfunction. Carnitine was also associated with improvement in markers of myocardial stress in N-terminalpro-brain natriuretic peptide, markers of aerobic metabolism in serum lactate, and markers of kidney function in creatinine over a 6-month study period. As LVEF was the primary focus of this study, these values were adjusted for other concomitant interventions and a significant increase remained. Adjusted LVEF increased from 45.7% immediately prior to carnitine initiation to 58.0%, representing an absolute increase of 12.3% in LVEF over 6 -months. Previous studies have demonstrated similar increases in LVEF. Pooled analyses by Loomba et al. demonstrated that in 144 pediatric patients across six studies LVEF significantly increased by 3.68%. These same pooled analyses also demonstrated a significant increase in left ventricular shortening fraction [5].
The original studies included in the aforementioned pooled analyses included three with cardiomyopathy patients. Of these, only two, that by Wang et al. as well as that by Kotby et al. quantified ventricular function by ejection fraction, both noting improvements in ejection fraction and shortening fraction associated with carnitine. Both studies also demonstrated improvement in clinical symptoms associated with carnitine [13,14]. Thus, the current study is among a small number of studies focusing on delineating the effect of carnitine supplementation on LVEF in children with left ventricular systolic dysfunction.
Cardiomyocytes rely on β-oxidation, the aerobic breakdown of fat within the mitochondria, to produce energy [13]. Levocarnitine, or L-carnitine, is a cofactor involved in the transport of fatty acids across the inner mitochondrial membrane [15]. L-carnitine plays an integral role in ATP production and assists in removing acylcarnitine derivatives from the mitochondria [16]. Defects in the carnitine shuttle can impair mitochondrial energy production. The myocardium cannot synthesize carnitine and thus relies on the liver and kidney as well as dietary sources to provide the necessary carnitine to transport fatty acids across the mitochondrial membrane. Neonates and infants in particular have decreased biosynthetic capacity and are at risk of developing carnitine deficiency, particularly when they are not receiving enteral nutrition [17,18].
In the process of β-oxidation, the fatty acid is activated into fatty acyl-CoA by coenzyme A. The fatty acyl-CoA cannot cross the inner mitochondrial membrane without carnitine which acts as a cofactor. Carnitine acyltransferase I enables the formation of an acylcarnitine molecule which can be transported across the inner mitochondrial membrane by Carnitine acyltranslocase. In the mitochondrial matrix, carnitine acyltransferase II transfers the fatty acylcarnitine molecule back to CoA forming fatty acyl-CoA. The fatty acyl-CoA can then undergo β-oxidation [19].
Through the aforementioned process, carnitine modulates the transfer of fatty acids into the mitochondrial matrix. Once this transfer is complete, Carnitine can be relocated to the cytosol by carnitine acyltranslocase. Carnitine acyltransferase II may complex carnitine with acyl-CoA in the mitochondrial matrix to form an ester. The ester may be removed from the mitochondria by translocase providing a pathway to remove acyl derivatives [20].
Carnitine is an especially attractive option for intervention in this patient population as it has few adverse effects. Most adverse effects are gastrointestinal symptoms, such as reflux and diarrhea [19]. Those with known seizure disorders may have increased frequency of seizures potentially related to carnitine and thus carnitine must be used with caution in this patient population [21].
This study is not without limitations. The most apparent limitation of this study is the lack of a control group in this retrospective study. This was not a controlled study with a non-carnitine arm. While the paired statistical analyses in this study utilize the patient's baseline levels as controls there remains the possibility that the ejection fraction would have improved with time without carnitine. While the effects of other medications can be estimated and adjusted for in these analyses as not all patients received them at any single point, this cannot be done for carnitine as inclusion into the study was based on receiving carnitine. Thus, time and carnitine are essentially combined as a variable in the paired regression analyses without the real ability to discern what degree of the change was from either component in the paired analyses. A statistical approach known as mediation analysis was utilized to account for this. That is why two separate regressions were utilized. A regression analysis was conducted in an unpaired fashion to help determine the independent association of time and ejection fraction and this demonstrated no significant association between time and ejection fraction. Although, even one is assumed and the effect of time adjusted for there is still a net benefit that appears secondary to carnitine. Another recently published pediatric study has also shown improvement in echocardiographic parameters with carnitine supplementation, perhaps further supporting the potential impact of carnitine on cardiac function [18]. There is attrition through the time points. The nature of this is multifactorial. Most importantly some of these patients were treated prior to 6 months of follow-up. Finally, due to the limited sample size, regression analyses accounting for specific drug doses of medications other than carnitine. These were simply treated as binary variables in the regression analyses.
Despite the limitations, these data are additive to the literature as they are able to quantify the effect of carnitine and the other medications over the same time period even though the effect of time itself cannot be quantified. Additionally, the current study demonstrates safety of carnitine in pediatric patients with left ventricular dysfunction.

Conclusion
Carnitine supplementation in pediatric patients with left ventricular systolic dysfunction may be associated with increase in LVEF and improvement in laboratory markers of myocardial stress and cardiac output.