Next Article in Journal
Analysis of the Effect of Demographic Variables on Lysosomal Enzyme Activities in the Missouri Newborn Screening Program
Previous Article in Journal
Nationwide Survey on Neonatal Critical Congenital Cardiopathies in Mexico: Data from 76 Public Health Service Hospital Units
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Assessment and Application of Acylcarnitines Summations as Auxiliary Quantization Indicator for Primary Carnitine Deficiency

Department of Pediatric Endocrinology and Genetic Metabolism, Xinhua Hospital, Shanghai Institute for Pediatric Research, Shanghai Jiaotong University School of Medicine, Shanghai 200092, China
*
Author to whom correspondence should be addressed.
These authors contributed equally to this work.
Int. J. Neonatal Screen. 2025, 11(2), 47; https://doi.org/10.3390/ijns11020047
Submission received: 11 April 2025 / Revised: 6 June 2025 / Accepted: 10 June 2025 / Published: 19 June 2025

Abstract

:
Background: Newborns are referred primary carnitine deficiency (PCD) when a low free carnitine (C0) concentration (<10 μmol/L) is detected, leading to high false-positive referrals. To improve the follow-up protocol for PCD, various acylcarnitines and the summations were comprehensively evaluated in the present study. Methods: A retrospective study was performed using samples due to low C0 concentration. Data were available for 72 patients with genetically confirmed PCD, whereafter C0 with the selected sum of (butyrylcarnitine (C4) + isovalerylcarnitine (C5)) was validated in an additional cohort study including about 80,000 samples. Results: In the discovery study, C4, acetylcarnitine (C2) and C5 exhibited significant discriminant power in distinguishing PCDs from NoPCDs. The area under the ROC curve (AUC) was 99.792% (C4), 98.715% (C2) and 98.620% (C5). The excellent performances in sensitivity, specificity, negative predictive value, positive predictive value (PPV) and accuracy indexes suggested that C4, C2 and C5 would be ideal auxiliary indicators in improving the diagnostic performance of C0 for PCD. Multivariate ROC curve-based exploratory analysis showed that C5, C4 and C2 were the most top-ranked features in differentiating PCDs from NoPCDs. AUC for C4 + C5 was the highest with a cutoff required for 100% sensitivity at 0.181 μmol/L. In the validation cohort, adding C4 + C5 in the NBS program could elevate PPV from 0.75% to 1.54%. Conclusions: Our work revealed that C4 + C5 summation should be used as the auxiliary quantization indicator to reduce false-positive results for PCD.

1. Introduction

Primary carnitine deficiency (PCD, MIM 212140) is an autosomal recessive disorder caused by mutations in the SLC22A5 gene, encoding the Organic Cation Transporter Novel 2 (OCTN2) protein [1,2,3,4]. OCTN2 can maintain intracellular free carnitine (C0) concentrations at normal levels by transporting carnitine into cells [5,6]. Carnitine is essential for transferring long-chain fatty acids from cytosol to mitochondria for subsequent beta oxidation [2,7,8]. Most PCD reported cases occur in infants and very young children. These patients often display an acute metabolic presentation with symptoms, such as hypoketotic hypoglycemia, elevated ammonia levels in blood, organic acidosis, and sudden death [2,7,9]. Onset is extremely rare in adulthood, and the clinical manifestations include progressive cardiomyopathy, myopathy, and encephalopathy due to hypoketotic hypoglycemia and hyperammonemia. The diagnosis and specific treatment of these age groups may be delayed for years [10,11]. The incidence of PCD varies in different countries, depending on the studied population, ranging from 1:300 in the Faroe Islands to 1:20,000–142,000 in Japan, Australia, or the USA [7,10,12]. Many large-scale Newborn Screening (NBS) studies have shown that PCD was the most prevalent fatty acid oxidation disorder in China, the incidence varying from 1:9000 to 1:34,000 [12,13,14,15,16,17].
Early diagnosis and timely supplementation of L-carnitine in the neonatal period are essential to achieve optimal clinical outcomes [8,18]. PCD is diagnosed based on levels of C0 and acylcarnitines, genetic mutation and clinical symptoms [10,12,19,20]. C0 and acylcarnitines are present at extremely decreased levels in PCD patients. However, transplacental transport of carnitine from mothers during pregnancy may lead to a low C0 level in neonates [12,21,22,23]. It is noteworthy that prematurity, malabsorption, malnutrition, and several inherited errors of metabolism, including organic acidemias and fatty acid oxidation defects, often lead to secondary carnitine deficiency [5,12,18,22]. These conditions represent pitfalls for the diagnosis and management of PCD neonates. Additionally, the currently used MS/MS-based markers for PCD diagnosis have undesirable performances, leading to high rates of false positives and occasional false negatives [13,16]. Follow-up of false positives is challenging, requiring a long period of confirmatory testing, including carnitine transporter activity measurement in cultured fibroblasts and/or SLC22A5 gene sequencing [15,23]. The prolonged uncertainty concerning the health of newborns during this extensive assessment can induce significant anxiety in the families regarded. A faster exclusion of PCD diagnosis may reduce harm from false-positive referrals. Consequently, several algorithms for PCD screening have been proposed, including different thresholds for C0 and other biomarkers, molecular sequencing of SLC22A5 and functional confirmation by carnitine uptake assay on skin fibroblasts [24].
At present, only a few countries apply secondary biomarkers in order to decrease false-positive rates for PCD [10,24]. Utilizing the sum of propionylcarnitine (C3) and palmitoylcarnitine (C16) on dried blood spot (DBS) < 2 µmol/L, in addition to a cutoff value of C0 < 10 µmol/L, allowed for a significant reduction in scree-positive cases and did not increase false-negative cases [25,26]. The positive predictive value (PPV%) increased from 7.1 to 13.0 [25]. Whilst acylcarnitines and their summations may be promising markers for the early differentiation of healthy newborns from PCD-affected newborns, information on acylcarnitines and their summations is limited. Furthermore, the optimization of newborn screening indicators and improvement in PPV are major challenges faced by newborn screening. According to the “Consensus on Primary carnitine deficiency Screening, Diagnosis and Treatment” in China, neonates with C0 < 10 µmol/L or C0 of 10~15 µmol/L, accompanied by a variety of acylcarnitine reductions, need to be recalled for review [27]. However, how to quantify multiple acylcarnitine levels is rarely reported in the literature.
In the present study, performances of free carnitine, specific acylcarnitines and their summations were comprehensively investigated on a larger scale to identify more effective NBS markers for PCD deficiency. This study demonstrates that the sum of butyrylcarnitine (C4) and isovalerylcarnitine (C5) could be used as an auxiliary diagnostic indicator in early discriminative tests for PCD.

2. Material and Methods

2.1. Research Subjects

The main objective of this study was to investigate the optimal auxiliary diagnostic indicators and significant changes in the acylcarnitines profiles of patients with PCD. A total of 72 PCD patients were recruited from 2003 to 2024. Detailed results of genomic confirmatory tests are listed in Supplementary Table S1. Infants with normal C0 levels and no clinical symptoms suggestive of metabolic disease were enrolled as the NonPCD group (n = 80). The utilities of C0 and various acylcarnitines and their summations in DBSs as markers for PCD were evaluated by comparing the data of the two groups.
To further validate the practicality of the selected auxiliary diagnostic indicators, an additional NBS cohort including 85,404 newborns was collected. Among them, 400 infants primarily displayed low C0 levels. Among them, 376 newborns were successfully recalled. Eventually, 3 newborns were diagnosed with PCD, yielding a PPV of 0.75% (3/400).

2.2. NBS Test for PCD

DBSs were collected from the heel pricks 72–120 h after birth. DBS samples were sent by cold chain transportation after natural air drying to Newborn Screening Center for tandem mass analysis. Both derivatized and underivatized methods were used for analyzing levels of amino acids, free carnitine, and acylcarnitines (API4000/API4500, Applied Biosystems, Foster City, CA, USA) [17,28]. The C0 cutoff value in our Newborn Screening Center was 10–60 µmol/L. Newborns in the initial screening with positive results (C0 < 10 µmol/L) were recalled and reexamined. The suspected PCDs were referred for confirmatory SLC22A5 gene mutation analysis.

2.3. Confirmatory Tests for PCD

Genetic analysis for PCDs was conducted according to the methods described in previous research with minor modifications after receiving informed consent [4,17]. Genomic DNA was extracted from peripheral blood leukocytes. All exons and flanking intron regions containing SLC22A5 were amplified using a polymerase chain reaction and analyzed on an automated DNA sequencer (Applied Biosystems, ABI3700). The SLC22A5 gene mutations were analyzed using Sanger sequencing. The mutations were identified using a normal human SLC22A5 sequence as a reference (NM_003060.4). Patients were diagnosed with PCD based on C0 levels, genetic mutations, and clinical symptoms. Newborns with two variants in the SLC22A5 gene were diagnosed with PCD. The suspected positives carrying one variant or not, performing genetic testing but with persistently lower C0 values until therapy, were diagnosed with PCD [17].

2.4. Statistics

The sample distribution was determined using the Kolmogorov–Smirnov normality test. For statistical comparisons, Student’s t-test or Mann–Whitney U-test followed by the post hoc tests (Dunnett’s test) was utilized for the normal or non-normal distributed data, respectively.
Receiver operating characteristic (ROC) curves were generated to evaluate the sensitivity and specificity of free carnitine and various acylcarnitines and their summations as markers for PCD. Univariate and multivariate ROC curve analyses were performed in MetaboAnalyst 6.0. Multivariate ROC curves were performed based on three multivariate algorithms: support vector machines (SVMs), partial least squares discriminant analysis (PLS-DA), and random forests (RFs). ROC curves were generated by Monte-Carlo cross-validation (MCCV) using balanced sub-sampling. In each MCCV, two-thirds of the samples (2/3) were used to evaluate the feature importance. The most important features were then used to build classification models. The established models were then validated on the samples that were left out (1/3). The procedure was conducted multiple times to generate the confidence interval and performance of each model.
The area under the curve (AUC), fold changes (FC), p-values and false-discovery rate (FDR) were used to evaluate the significance of the respective acylcarnitines as potential diagnostic biomarkers. All the p-values were corrected by the false-discovery rate using the Benjamin–Hochberg method. The adjusted p-value (q-value) of 0.05 was taken as the significance level.

3. Results

3.1. Diagnostic Performances of Currently Used Free Carnitine and Acylcarnitines for PCD

Supplementary Table S2 shows levels of C0 and individual acylcarnitine [acetylcarnitine (C2), C3, C4, C5, caproylcarnitine (C6), caprylylcarnitine (C8), actinylcarnitine (C10), lauroylcarnitine (C12), myristoylcarnitine (C14), C16, octadecylcarnitine (C18)] in DBSs in the present study. In the PCD (n = 72) versus the NoPCD (n = 80) population, the C0 median concentration [interquartile range (IQR)] was 5.839 [4.392–7.201] and 26.418 [21.973–32.944] μmol/L, respectively (Supplementary Table S2).
The discriminant performances of C0 and individual acylcarnitines were investigated using ROC analysis (Table 1). AUC with the optimal cutoff was highest for C0 (100.000%), followed by C4 (99.792%), C2 (98.715%) and C5 (98.620%). This analysis evidenced that the currently used C0 is an optimal primary marker for PCD. The sensitivity, specificity, negative predictive value (NPV), PPV and accuracy indexes in Table 1 suggested that C4, C2 and C5 might be ideal auxiliary indicators in improving diagnostic performance of C0 for PCD.
Figure 1 shows the ROC curves and corresponding box plots of C0, C4, C2 and C5 in DBSs in the PCDs and NoPCDs. All of these markers could clearly distinguish PCDs from NoPCDs. Supplementary Figure S1 presents ROC curves of the other specific acylcarnitines. Cluster heatmaps and fold changes of free carnitine and specific acylcarnitines (Figure 1E,F) show that C0, C4, C2 and C5 had the most significant changes (q value < 10−25). C3 and C16 displayed lower levels in patients, with the q value < 10−20 and < 10−15, respectively.

3.2. Machine Learning-Assisted Identification of Candidate Acylcarnitines for PCD Diagnosis

Multivariate ROC curve-based exploratory analysis was performed for identifying the optimal auxiliary diagnostic markers for PCD. ROC curve analyses were performed based on three multivariate algorithms: SVM, PLS-DA, and RF. Figure 2 shows the average important features obtained from multivariate ROC curve-based exploratory analysis. C5, C4 and C2 were identified as the most top-ranked features in all three methods. The AUC for the diagnosis of PCD in the studied cohort ranged from 0.999 to 1, with confidence intervals (CI) of 0.996–1, 0.997–1 and 0.999–1. To mine features having the most optimal performance and suitability for PCD diagnosis, summations of the top five features (C4, C5, C2, C14, C3) were selected and evaluated in discriminating PCDs from NoPCDs.

3.3. Diagnostic Accuracy Evaluation of Acylcarnitines Summations for PCD

Supplementary Figure S2 shows box plots of acylcarnitines summations in DBSs. Supplementary Table S3 shows individual acylcarnitines summation levels (C4 + C5, C4 + C2, C4 + C14, C4 + C3, C5 + C2, C5 + C14, C5 + C3, C2 + C14, C2 + C3, C14 + C3, C4 + C5 + C2, C4 + C5 + C14, C4 + C5 + C3, C5 + C2 + C14, C5 + C2 + C3, C2 + C14 + C3, C4 + C5 + C2 + C14, C5 + C2 + C14 + C3, C4 + C5 + C2 + C14 + C3, C3 + C16) in DBSs from PCDs and NoPCDs in the present study. Additionally, C3 + C16, which is the current auxiliary indicator for PCD NBS, was also incorporated. According to the ROC analysis (Table 2), C4 + C5 had the highest AUC (99.983%), followed by C4 + C5 + C14 (99.948%), C4 + C14 (99.887%), and C5 + C14 (98.958%). Regarding sensitivity, C4 + C5, C4 + C5 + C14, and C4 + C14 exhibited sensitivity of 100.000%, followed by C5 + C14 with sensitivity of 97.222%. For specificity, C4 + C5 had the highest value of 98.750%, followed by C4 + C5 + C14, C4 + C14 and C5 + C14 with specificities of 97.500%, 97.500% and 92.500%, respectively. Positive predictive values of C4 + C5, C4 + C5 + C14, C4 + C14 and C5 + C14 were 98.673%, 97.297%, 97.297% and 92.105%, respectively. The negative predictive values were 100.000% for C4 + C5, C4 + C5 + C14, and C4 + C14. C4 + C5 had the highest accuracy of 99.342%.
The ROC curves and the corresponding box plots of C4 + C5, C4 + C5 + C14, C4 + C14 and C5 + C14 are shown in Figure 3. ROC curves of the other various acylcarnitines summations are shown in Supplementary Figure S3. The notches of box plots for C4 + C5 had the least overlap (Figure 3A), indicating the best classification power for PCDs and NoPCDs. Figure 3E,F show a heatmap and the fold changes of the various acylcarnitines summations in PCDs and NoPCDs. C4 + C5, C4 + C5 + C14 and C4 + C14 exhibited the most significant changes (q value < 10−25). C3 + C16 showed a q value < 10−20.
To further verify the utility of C4 + C5 in NBS for PCD, ROC analysis between false-positive subjects from NBS data and the true-positive patients was performed. Univariate ROC curve analysis revealed that AUC was highest for C4 (89.608%), followed by C18 (87.993%), C5 (87.420%), C16 (85.873%) and C0 (85.457%) (Table 3 and Figure 4A). Multivariate exploratory ROC analysis also showed that C5, C4, C16 and C18 were the top-ranked features (Figure 4A). Therefore, the performances of C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 between false-positive and true-positive subjects were evaluated (Table 4 and Figure 4B). The ROC curves and the corresponding box plots are shown in Figure 4B. According to the ROC analysis (Table 4), C4 + C5 and C4 + C5 + C18 had AUCs of 91.052% and 91.054%, respectively, followed by C4 + C5 + C16 (87.777%) and C18 + C16 (86.770%). Regarding sensitivity, C4 + C5 + C16 exhibited the highest value (88.295%), followed by C4 + C5 (82.952%), C4 + C5 + C18 (78.880%), and C18 + C16 (58.015%). For specificity, C18 + C16 had the highest value of 97.222%, followed by C4 + C5 + C18, C4 + C5 and C4 + C5 + C16 with specificities of 86.111%, 84.722% and 68.056%, respectively. Positive predictive values of C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 were 96.736%, 96.875%, 93.784% and 99.130%, respectively. The negative predictive values were fairly low (<60.000%) for the four indicators. C4 + C5 and C4 + C5 + C16 had accuracies of 83.226% and 85.161%, respectively, followed by C4 + C5 + C18 (80.000%) and C18 + C16 (64.086%). C4 + C5 exhibited pretty good performances in AUC, sensitivity, specificity, positive predictive value and accuracy.
ROC analysis between false-positive subjects and those with normal C0 values was also performed for C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 (Table 4 and Figure 5A). C4 + C5 had the highest AUC of 88.034%, followed by C4 + C5 + C18 (77.861%), C4 + C5 + C16 (63.386%), and C18 + C16 (62.056%), respectively. This result indicated that C4 + C5 was the best indicator in differentiating false-positive subjects from normal ones. Additionally, ROC analysis between true-positive patients and those with normal C0 values for C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 also showed that the C4 + C5 exhibited the best performances in discriminating true-positive patients from normal ones (Table 4 and Figure 5B). AUC with the optimal cutoff was highest for C4 + C5 (99.983%), followed by C4 + C5 + C18 (98.880%), C4 + C5 + C16 (94.045%), and C18 + C16 (92.561%). The sensitivity, specificity, NPV, PPV, and accuracy indexes for C4 + C5 were 100.000%, 98.750%, 100.000%, 98.630%, and 99.342%, respectively (Table 4). This analysis evidenced that the screened C4 + C5 was a credible auxiliary diagnostic indicator for PCDs.
Table 5 shows the diagnostic performances of the C4 + C5, C4 + C5 + C14, C4 + C14 and C5 + C14 summations using optimal cutoffs determined by ROC analysis for PCD. The C4 + C5 summation yielded a false-positive result in only one patient. The C4 + C5 + C14 and C4 + C14 summations yielded false-positive results in two patients. There were no false-negative results in the C4 + C5, C4 + C5 + C14, C4 + C14 summation.

3.4. Validation and Application of Free Carnitine and C4 + C5 Summation in PCD NBS

Based on the above results, in this part, an additional NBS cohort program in Xinhua Hospital from 2021 to 2024 was utilized to assess the performance of the C0 and C4 + C5 summation in discriminating PCDs from NoPCDs. As shown in Table 6, the additional C4 + C5 summation (<0.181 μmol/L) utilization could elevate PPV from 0.75% to 1.54%, significantly reducing the false positives recalled (256 to 123). Among the 400 screening positives for PCD, up to 205 presented C4 + C5 summation above 0.181 µM. No one was a PCD patient in this population.

4. Discussion

PCD screening in neonates has been widely implemented across the world [2,29,30,31]. However, NBS for PCD is well known to have an unsatisfactory PPV, ranging from 1.6% to 4.7% [31,32]. If the C0 concentration is low at the initial screening, a second DBS sample is collected, and newborns are only referred when the low C0 concentration persists [17]. However, this workflow comes at the cost of timeliness. Initial DBSs are usually obtained 72–170 h after birth, and, in the event that a second test is requested, there is a further delay of 2–3 days before a newborn is referred. Additionally, false-positive outcomes in NBS could cause unnecessary anxieties for parents [33,34]. Hence, markers that have the best performances in sensitivity and specificity should be clarified to significantly reduce the false positives in PCD screening. In the present study, the levels of specific acylcarnitines and C0 were first compared in the 152 DBSs. In addition to C0, C4, C2 and C5 were also the most significantly downregulated markers (q < 10−25). Furthermore, the ROC analysis also showed that decreased C4, C2 and C5 levels exhibited the highest discrimination value (AUC > 0.980). Multivariate exploratory ROC analysis based on three algorithms (SVM, RF and PLS-DA) was performed to further screen the most reasonable acylcarnitines signatures, which could be utilized as auxiliary quantitative diagnostic indicators for PCD. The ROC analysis showed that the top three acylcarnitines (C4, C2 and C5, Figure 2) presented the highest discrimination value (AUC = 0.999–1). Next, C4 + C5, C4 + C2, C2 + C5 and C4 + C2 + C5 summations were evaluated in the 72 PCDs. Decreased C4 + C5 levels exhibited the highest discrimination value (AUC > 0. 999).
Considering the clinical symptoms of PCD and the accessibility of effective medical management, both false-positive and false-negative results should be prevented. Table 5 summarizes the performance of the C4 + C5 with the optimal cutoff applied to 72 PCDs and 80 NoPCDs. C4 + C5 yielded only one false-positive and no false-negative results. Furthermore, the performance of the C4 + C5 biomarker was evaluated in an additional NBS cohort including 85,404 infants. When using C0 < 10 µmol/L with C4 + C5 < 0.181 µmol/L, the PPV increased to 1.54% (3/195). It is worth noting that 92 cases with mildly decreased C0 (9.0~10.0 µmol/L) and C4 + C5 > 0.181 µmol/L were identified, and no abnormalities were found in 67 of them upon repeat testing (Table 6). Those with mildly decreased C0 but C4 + C5 > 0.181 µmol/L are less likely to be PCD patients. Therefore, in the screening system, adding the screening criteria C4 + C5 < 0.181 µmol/L can improve the positive predictive value, reduce the false-positive rate, and potentially help differentiate secondary carnitine deficiency, thereby improving the performance of newborn screening.
We should also pay attention to false negatives in PCD neonatal screening [35]. In the NBS cohort study, two PCD cases had a C0 value of 8.525 µmol/L and 8.937 µmol/L at the initial screening, which is very easy to miss. However, the corresponding C4 + C5 values were 0.121 µmol/L and 0.104 µmol/L, respectively. Our previous research on “Genomic Sequencing as a First-Tier Screening Test and Outcomes of Newborn Screening” also confirmed one PCD patient whose C0 and C4 + C5 values were 8.330 µmol/L and 0.120 µmol/L, respectively [36]. Therefore, it is important to pay attention to the recall of full-term neonates with a C0 critical value and C4 + C5 < 0.181 µmol/L to reduce the rate of missed diagnosis. Even if the neonatal screening is negative, if the neonate is clinically suspected of PCD, it still needs to be differentiated using tandem mass spectrometry. To further validate the performance of C4 + C5 < 0.181 µmol/L as an additional screening indicator, retrospective analysis was conducted on newborns recruited for PCD screening from 2003 to 2024 in Xinhua Hospital. One false-negative case (male) was reported, delivered at 38 weeks of gestation, weighing 3650 g at birth in 2016. Biochemical NBS screening showed that the C0 level was 11.637 µmol/L, not diagnosed with PCD in 2016. However, in 2019, this case exhibited clinical symptoms and SLC22A5 gene mutation analysis showed two variants (c.760C > T; c.1400C > G). And the initial screening mass data in 2016 showed that C4 + C5 was 0.111 µmol/L. This false-negative case would not be missed if the sum of C4 + C5 was added to the NBS program. Based on the above results, the sum of C4 + C5 could be used as an effective auxiliary diagnostic indicator for PCDs with the lowest increase in false-positive and false-negative rates.
The added benefit of acylcarnitine summation in NBS has been reported previously in Guangzhou Newborn Screening Center, China. This research demonstrated that adding the quantitative biomarker C3 + C16 < 2 µmol/L into the newborn screening program could improve the PCD screen performance. However, in the present study population, using C3 + C16 < 2 µmol/L yielded 17 false positives and 5 false negatives (Table 5). Although we utilized the optimal cutoff of C3 + C16 determined by ROC analysis (1.839 µmol/L), there were up to 12 false positives and 6 false negatives, respectively. The performance is obviously lower than C4 + C5.
We acknowledge the limitations of using the C4 + C5 summation as an additional discriminative tool for PCD. In the Genomic Sequencing Program, with 29,989 infants recruited from eight NBS centers throughout China, one PCD patient was not detected by genetic NBS but revealed via tandem mass spectrometry. This patient exhibited a C0 value of 6.65 µmol/L. The sum of C4 + C5 was 0.190 µmol/L, slightly higher than the optimal cutoff (0.181 µmol/L) determined in the present study. Larger studies are necessary to validate the C4 + C5 threshold. The other limitation of the present study was the lack of reference values for the NBS program in PCD, except the currently used C0 level. The sum of (C4 + C5)’s performance must be further validated using nationwide NBS data.

5. Conclusions

In this study, we performed a comprehensive evaluation of the diagnostic performances of specific acylcarnitine levels and various acylcarnitine summations using two PCD cohorts. The utilization of C0 < 10 µmol/L with C4 + C5 < 0.181 µmol/L can effectively discriminate between true- and false-positive referrals for PCD, reducing time to diagnosis and mitigating the negative effects of a false-positive referral. The performance of C0 plus C4 + C5 will be further evaluated in the retrospective and prospective nationwide studies for PCD. The optimal and appropriate reference ranges can be established, improving differential PCD diagnosis with significant reductions in the false positivity and false negativity rates.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/ijns11020047/s1, Figure S1: Performance of acylcarnitines in identifying patients with PCD; Figure S2: Box plots of acylcarnitines summations in dried blood specimens; Figure S3: ROCs of various acylcarnitines summations in dried blood specimens to diagnose PCDs; Table S1: Genetic variants of the PCD patients in the study cohort; Table S2: Levels of free carnitine and specific acylcarnitines in dried blood specimens (μmol/L); Table S3: Levels of various acylcarnitines summations in dried blood specimens (μmol/L).

Author Contributions

H.Z. (Haijuan Zhi): Conceptualization, supervision, validation, data curation, formal analysis, investigation, methodology, visualization, writing—original draft. S.C.: Project administration, formal analysis, visualization, writing—review and editing, investigation. T.C.: Methodology, data curation. L.L.: Methodology, data curation. W.Q.: Visualization, validation, writing—review and editing. H.Z. (Huiwen Zhang): Writing—review and editing. X.G.: Writing—review and editing. L.H.: Conceptualization, investigation, supervision. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by grants from the Scientific research Project Plan of Shanghai Municipal Health Commission (No. 202140346) and the National Key Research and Development Program of China (No. 2016YFC0901505).

Institutional Review Board Statement

The study was approved by the Ethics Committee of Xinhua Hospital Affiliated to Shanghai Jiaotong University School of Medicine (Approval No. XHEC-D-2025-016, Approval Date: date 10 February) and conducted in accordance with the principles of the Declaration of Helsinki.

Informed Consent Statement

Written informed consent was obtained from participants or their parents.

Data Availability Statement

The datasets used and analyzed in this study are available from the corresponding author on reasonable request.

Acknowledgments

We are grateful for the invaluable contributions of all the patients. We also express our sincere gratitude to all of the pediatricians who provided associated clinical data and samples. These were essential in successfully accomplishing this research.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Nez, J.I.; Tamai, I.; Oku, A. Primary systemic carnitine deficiency is caused by mutations in a gene encoding sodium ion-dependent carnitine transporte. Nat. Genet. 1999, 21, 91–94. [Google Scholar] [CrossRef] [PubMed]
  2. Zhou, J.; Li, G.; Zeng, Y.; Qiu, X.; Zhao, P.; Huang, T.; Wang, X.; Luo, J.; Lin, N.; Xu, L. Screening primary carnitine deficiency in 10 million Chinese newborns: A systematic review and meta-analysis. Orphanet J. Rare Dis. 2024, 19, 248. [Google Scholar] [CrossRef] [PubMed]
  3. Crefcoeur, L.L.; Visser, G.; Ferdinandusse, S.; Wijburg, F.A.; Langeveld, M.; Sjouke, B. Clinical characteristics of primary carnitine deficiency: A structured review using a case-by-case approach. J. Inherit. Metab. Dis. 2022, 45, 386–405. [Google Scholar] [CrossRef] [PubMed]
  4. Han, L.; Wang, F.; Wang, Y.; Ye, J.; Qiu, W.; Zhang, H.; Gao, X.; Gong, Z.; Gu, X. Analysis of genetic mutations in Chinese patients with systemic primary carnitine deficiency. Eur. J. Med. Genet. 2014, 57, 571–575. [Google Scholar] [CrossRef]
  5. Koleske, M.L.; McInnes, G.; Brown, J.E.H. Functional genomics of OCTN2 variants informs protein-specific variant effect predictor for Carnitine Transporter Deficiency. Proc. Natl. Acad. Sci. USA 2022, 119, e2210247119. [Google Scholar] [CrossRef]
  6. Koepsell, H.; Daws, L.C. Organic Cation Transporters in Health and Disease. Pharmacol. Rev. 2019, 72, 253–319. [Google Scholar]
  7. Loos, M.; Klampe, B.; Schulze, T.; Yin, X.; Theofilatos, K.; Ulmer, B.M.; Schulz, C.; Behrens, C.S.; van Bergen, T.D.; Adami, E.; et al. Human model of primary carnitine deficiency cardiomyopathy reveals ferroptosis as a novel mechanism. Stem Cell Rep. 2023, 18, 2123–2137. [Google Scholar] [CrossRef]
  8. Jolfayi, A.G.; Naderi, N.; Ghasemi, S.; Salmanipour, A.; Adimi, S.; Maleki, M.; Kalayinia, S. A novel pathogenic variant in the carnitine transporter gene, SLC22A5, in association with metabolic carnitine deficiency and cardiomyopathy features. BMC Cardiovasc. Disord. 2024, 24, 1. [Google Scholar] [CrossRef]
  9. Zhou, Y.; Liu, Y.; Shen, Y.; Xu, F.; Xu, F.; Huang, H.; Duan, J. A report of a pedigree with compound heterozygous mutations in the SLC22A5 gene. Front. Pediatr. 2023, 11, 985720. [Google Scholar] [CrossRef]
  10. Lefevre, C.R.; Labarthe, F.; Dufour, D.; Moreau, C.; Faoucher, M.; Rollier, P.; Arnoux, J.B.; Tardieu, M.; Damaj, L.; Bendavid, C.; et al. Newborn Screening of Primary Carnitine Deficiency: An Overview of Worldwide Practices and Pitfalls to Define an Algorithm before Expansion of Newborn Screening in France. Int. J. Neonatal. Screen. 2023, 9, 6. [Google Scholar] [CrossRef]
  11. Guerrero Cervera, B.; Donoso Trenado, V.; Fernandez-Sellers, C.; Vila Clerigues, N.; Correcher Medina, P.; Zorio, E. Cardiomyopathy associated with primary carnitine deficiency in adults: Lack of complete response in patients with delayed treatment initiation. Rev. Esp. Cardiol. (Engl. Ed.) 2024, 77, 265–267. [Google Scholar] [CrossRef] [PubMed]
  12. Ji, X.; Ge, Y.; Ni, Q.; Xu, S.; Xiong, Z.; Yang, L.; Hu, L.; Cao, Y.; Lu, Y.; Wei, Q.; et al. Primary carnitine deficiency: Estimation of prevalence in Chinese population and insights into newborn screening. Front. Genet. 2023, 14, 1304458. [Google Scholar] [CrossRef]
  13. Lin, Y.; Zhang, W.; Huang, C.; Lin, C.; Lin, W.; Peng, W.; Fu, Q.; Chen, D. Increased detection of primary carnitine deficiency through second-tier newborn genetic screening. Orphanet J. Rare Dis. 2021, 16, 149. [Google Scholar] [CrossRef]
  14. Lin, Y.; Lin, B.; Chen, Y.; Zheng, Z.; Fu, Q.; Lin, W.; Zhang, W. Biochemical and genetic characteristics of patients with primary carnitine deficiency identified through newborn screening. Orphanet J. Rare Dis. 2021, 16, 503. [Google Scholar] [CrossRef]
  15. Yang, X.; Li, Q.; Wang, F.; Yan, L.; Zhuang, D.; Qiu, H.; Li, H.; Chen, L. Newborn Screening and Genetic Analysis Identify Six Novel Genetic Variants for Primary Carnitine Deficiency in Ningbo Area, China. Front. Genet. 2021, 12, 686137. [Google Scholar] [CrossRef]
  16. Lin, Y.; Lin, C.; Lin, B.; Zheng, Z.; Lin, W.; Chen, Y.; Chen, D.; Peng, W. Newborn screening for fatty acid oxidation disorders in a southern Chinese population. Heliyon 2024, 10, e23671. [Google Scholar] [CrossRef] [PubMed]
  17. Chang, S.; Yang, Y.; Xu, F.; Ji, W.; Zhan, X.; Gao, X.; Chen, T.; Qiu, W.; Zhang, H.; Liang, L.; et al. Clinical, biochemical, and molecular genetic characteristics of patients with primary carnitine deficiency identified by newborn screening in Shanghai, China. Front. Genet. 2022, 13, 1062715. [Google Scholar] [CrossRef] [PubMed]
  18. Lin, Y.; Zheng, Z.; Lin, W.; Peng, W. Incorporating Next-Generation Sequencing as a Second-Tier Test for Primary Carnitine Deficiency. Mol. Genet. Genomic Med. 2024, 12, e70003. [Google Scholar] [CrossRef]
  19. Chen, Y.; Lin, Q.; Zeng, Y.; Qiu, X.; Liu, G.; Zhu, W. Gene spectrum and clinical traits of 10 patients with primary carnitine deficiency. Mol. Genet. Genomic Med. 2021, 9, e1583. [Google Scholar] [CrossRef]
  20. Wilcken, B.; Wiley, V.; Hammond, J. Screening Newborns for Inborn Errors of Metabolism by Tandem Mass Spectrometry. N. Engl. J. Med. 2003, 348, 2304–2312. [Google Scholar] [CrossRef]
  21. Crefcoeur, L.; Ferdinandusse, S.; van der Crabben, S.N.; Dekkers, E.; Fuchs, S.A.; Huidekoper, H.; Janssen, M.; Langendonk, J.; Maase, R.; de Sain, M.; et al. Newborn screening for primary carnitine deficiency: Who will benefit?—A retrospective cohort study. J. Med. Genet. 2023, 60, 1177–1185. [Google Scholar] [CrossRef] [PubMed]
  22. Lin, W.; Wang, K.; Zheng, Z.; Chen, Y.; Fu, C.; Lin, Y.; Chen, D. Newborn screening for primary carnitine deficiency in Quanzhou, China. Clin. Chim. Acta 2021, 512, 166–171. [Google Scholar] [CrossRef] [PubMed]
  23. Lin, Y.; Xu, H.; Zhou, D.; Hu, Z.; Zhang, C.; Hu, L.; Zhang, Y.; Zhu, L.; Lu, B.; Zhang, T.; et al. Screening 3.4 million newborns for primary carnitine deficiency in Zhejiang Province, China. Clin. Chim. Acta 2020, 507, 199–204. [Google Scholar] [CrossRef] [PubMed]
  24. Crefcoeur, L.L.; Heiner-Fokkema, M.R.; Maase, R.E.; Visser, G.; de Sain-van der Velden, M.G.M. Assessment of carnitine excretion and its ratio to plasma free carnitine as a biomarker for primary carnitine deficiency in newborns. JIMD Rep. 2022, 64, 57–64. [Google Scholar] [CrossRef]
  25. Tang, C.; Tan, M.; Xie, T.; Tang, F.; Liu, S.; Wei, Q.; Liu, J.; Huang, Y. Screening for neonatal inherited metabolic disorders by tandem mass spectrometry in Guangzhou. J. Zhejiang Univ. (Med. Sci.) 2021, 50, 463–471. [Google Scholar] [CrossRef]
  26. Huang, Y.L.; Tang, C.F.; Liu, S.C.; Sheng, H.Y.; Tang, F.; Jiang, X.; Zheng, R.D.; Mei, H.F.; Liu, L. Newborn screening for primary carnitine deficiency and variant spectrum of SLC22A5 gene in Guangzhou. Chin. J. Pediatr. 2020, 58, 476–481. [Google Scholar]
  27. Yang, R. Consensus on screening and diagnosis of primary carnitine deficiency. Natl. Med. J. China 2019, 99, 88–92. [Google Scholar]
  28. Wang, Z. Expert consensus on tandem mass spectrometry screening techniques for neonatal diseases. Chin. J. Lab. Med. 2019, 42, 89–97. [Google Scholar]
  29. McCandless, S.E.; Wright, E.J. Mandatory newborn screening in the United States: History, current status, and existential challenges. Birth Defects Res. 2020, 112, 350–366. [Google Scholar] [CrossRef]
  30. Therrell, B.L.; Padilla, C.D. Newborn screening in the developing countries. Curr. Opin. Pediatr. 2018, 30, 734–739. [Google Scholar] [CrossRef]
  31. Gallant, N.M.; Leydiker, K.; Wilnai, Y.; Lee, C.; Lorey, F.; Feuchtbaum, L.; Tang, H.; Carter, J.; Enns, G.M.; Packman, S.; et al. Biochemical characteristics of newborns with carnitine transporter defect identified by newborn screening in California. Mol. Genet. Metab. 2017, 122, 76–84. [Google Scholar] [CrossRef] [PubMed]
  32. Schiergens, K.A.; Weiss, K.J.; Röschinger, W.; Lotz-Havla, A.S.; Schmitt, J.; Dalla Pozza, R.; Ulrich, S.; Odenwald, B.; Kreuder, J.; Maier, E.M. Newborn screening for carnitine transporter defect in Bavaria and the long-term follow-up of the identified newborns and mothers: Assessing the benefit and possible harm based on 19 ½ years of experience. Mol. Genet. Metab. Rep. 2021, 28, 100776. [Google Scholar] [CrossRef] [PubMed]
  33. Tajima, G.; Aisaki, J.; Hara, K.; Tsumura, M.; Kagawa, R.; Sakura, F.; Sasai, H.; Yuasa, M.; Shigematsu, Y.; Okada, S. Using the C14:1/Medium-Chain Acylcarnitine Ratio Instead of C14:1 to Reduce False-Positive Results for Very-Long-Chain Acyl-CoA Dehydrogenase Deficiency in Newborn Screening in Japan. Int. J. Neonatal Screen. 2024, 10, 15. [Google Scholar] [CrossRef] [PubMed]
  34. Liu, N.; Xiao, J.; Gijavanekar, C.; Pappan, K.L.; Glinton, K.E.; Shayota, B.J.; Kennedy, A.D.; Sun, Q.; Sutton, V.R.; Elsea, S.H. Comparison of Untargeted Metabolomic Profiling vs Traditional Metabolic Screening to Identify Inborn Errors of Metabolism. JAMA Netw. Open 2021, 4, e2114155. [Google Scholar] [CrossRef]
  35. Lee, N.-C.; Tang, N.L.-S.; Chien, Y.-H.; Chen, C.-A.; Lin, S.-J.; Chiu, P.-C.; Huang, A.-C.; Hwu, W.-L. Diagnoses of newborns and mothers with carnitine uptake defects through newborn screening. Mol. Genet. Metab. 2010, 100, 46–50. [Google Scholar] [CrossRef]
  36. Chen, T.; Fan, C.; Huang, Y.; Feng, J.; Zhang, Y.; Miao, J.; Wang, X.; Li, Y.; Huang, C.; Jin, W.; et al. Genomic Sequencing as a First-Tier Screening Test and Outcomes of Newborn Screening. JAMA Netw. Open 2023, 6, e2331162. [Google Scholar] [CrossRef]
Figure 1. ROC curves and corresponding box plots of (A) C0, (B) C4, (C) C2, and (D) C5. The notch indicates the 95% confidence interval around the median of each group. The optimal cutoff is indicated with a horizontal red line on the boxplot. The mean concentration of each group is indicated with a yellow diamond. Each dots represents one individual sample. (E) Cluster heatmap and (F) fold changes of free carnitine and various acylcarnitines in dried blood specimens from PCDs and NoPCDs. PCD, primary carnitine deficiency; ROC, receiver operator characteristic; AUC, area under the curve; C0, carnitine; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Figure 1. ROC curves and corresponding box plots of (A) C0, (B) C4, (C) C2, and (D) C5. The notch indicates the 95% confidence interval around the median of each group. The optimal cutoff is indicated with a horizontal red line on the boxplot. The mean concentration of each group is indicated with a yellow diamond. Each dots represents one individual sample. (E) Cluster heatmap and (F) fold changes of free carnitine and various acylcarnitines in dried blood specimens from PCDs and NoPCDs. PCD, primary carnitine deficiency; ROC, receiver operator characteristic; AUC, area under the curve; C0, carnitine; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Ijns 11 00047 g001
Figure 2. Average important features obtained from multivariate ROC curve-based exploratory analysis and the corresponding ROC curves of C5, C4 and C2. (A) linear surport vector machine-Linear SVM. (B) random forest-RF. (C) partial least square-discriminant analysis-PLS-DA. AUC, area under the curve; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Figure 2. Average important features obtained from multivariate ROC curve-based exploratory analysis and the corresponding ROC curves of C5, C4 and C2. (A) linear surport vector machine-Linear SVM. (B) random forest-RF. (C) partial least square-discriminant analysis-PLS-DA. AUC, area under the curve; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Ijns 11 00047 g002
Figure 3. ROC curves and box plots of (A) C4 + C5, (B) C4 + C5 + C14, (C) C4 + C14, and (D) C5 + C14. The notch indicates the 95% confidence interval around the median of each group. The optimal cutoff is indicated with a horizontal red line on the boxplot. The mean concentration of each group is indicated with a yellow diamond. (E) cluster heatmap and (F) fold changes of specific various acylcarnitines summations in dried blood specimens from PCDs and NoPCDs. PCD, primary carnitine deficiency; ROC, receiver operator characteristic; AUC, area under the curve; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine.
Figure 3. ROC curves and box plots of (A) C4 + C5, (B) C4 + C5 + C14, (C) C4 + C14, and (D) C5 + C14. The notch indicates the 95% confidence interval around the median of each group. The optimal cutoff is indicated with a horizontal red line on the boxplot. The mean concentration of each group is indicated with a yellow diamond. (E) cluster heatmap and (F) fold changes of specific various acylcarnitines summations in dried blood specimens from PCDs and NoPCDs. PCD, primary carnitine deficiency; ROC, receiver operator characteristic; AUC, area under the curve; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine.
Ijns 11 00047 g003
Figure 4. ROC curves and corresponding box plots between false-positive subjects and true-positive patients. (A) Univariate ROC curve analysis and multivariate ROC curve-based exploratory analysis. (B) C4 + C5, C4 + C5 + C18, C4 + C14 + C16 and C18 + C16. Each dots represents one individual sample. C4, butyrylcarnitine; C5, isovalerylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Figure 4. ROC curves and corresponding box plots between false-positive subjects and true-positive patients. (A) Univariate ROC curve analysis and multivariate ROC curve-based exploratory analysis. (B) C4 + C5, C4 + C5 + C18, C4 + C14 + C16 and C18 + C16. Each dots represents one individual sample. C4, butyrylcarnitine; C5, isovalerylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Ijns 11 00047 g004
Figure 5. ROC analysis between false-positive subjects and those with normal C0 values (A). ROC analysis between true-positive patients and those with normal C0 values (B). Each dots represents one individual sample. C4, butyrylcarnitine; C5, isovalerylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Figure 5. ROC analysis between false-positive subjects and those with normal C0 values (A). ROC analysis between true-positive patients and those with normal C0 values (B). Each dots represents one individual sample. C4, butyrylcarnitine; C5, isovalerylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Ijns 11 00047 g005
Table 1. ROC analysis of free carnitine and specific acylcarnitines levels in dried blood specimens.
Table 1. ROC analysis of free carnitine and specific acylcarnitines levels in dried blood specimens.
CharacteristicFree Carnitine and Acylcarnitine
C0C2C3C4C5C6C8C10C12C14C16C18
AUC (%)100.000 98.715 96.753 99.792 98.620 86.623 92.274 95.200 88.168 95.738 91.319 95.095
Best Cutoff Value (μmol/L)11.854 10.973 0.882 0.111 0.063 0.043 0.035 0.042 0.046 0.077 0.583 0.305
Sensitivity (%)100.000 98.611 95.833 98.611 94.444 76.389 75.000 81.944 83.333 91.667 72.222 83.333
Specificity (%)100.000 95.000 85.000 98.750 100.000 82.500 97.500 96.250 78.750 87.500 95.000 91.250
Negative Predictive Value (%)100.000 98.701 95.775 98.750 95.238 79.518 81.250 85.556 84.000 92.105 79.167 85.882
Positive Predictive Value (%)100.000 94.667 85.185 98.611 100.000 79.710 96.429 95.161 77.922 86.842 92.857 89.552
Accuracy (%)100.000 96.711 90.132 98.684 97.368 79.605 86.842 89.474 80.921 89.474 84.211 87.500
AUC, area under the curve; C0, carnitine; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Table 2. ROC analysis of specific various acylcarnitines summations in dried blood specimens.
Table 2. ROC analysis of specific various acylcarnitines summations in dried blood specimens.
Acylcarnitines
Summations
AUC (%)Best Cutoff Value
(μmol/L)
Sensitivity
(%)
Specificity
(%)
Negative Predictive Value
(%)
Positive Predictive Value
(%)
Accuracy
(%)
C4 + C599.983 0.181 100.000 98.750 100.000 98.630 99.342
C4 + C5 + C1499.948 0.281 100.000 97.500 100.000 97.297 98.684
C4 + C1499.887 0.201 100.000 97.500 100.000 97.297 98.684
C5 + C1498.958 0.151 97.222 92.500 97.368 92.105 94.737
C4 + C5 + C2 + C14 + C398.872 11.792 98.611 96.250 98.718 95.946 97.368
C5 + C2 + C398.854 11.598 98.611 96.250 98.718 95.946 97.368
C4 + C5 + C2 + C1498.837 11.201 98.611 95.000 98.701 94.667 96.711
C5 + C2 + C14 + C398.837 11.691 98.611 96.250 98.718 95.946 97.368
C2 + C398.819 11.502 98.611 96.250 98.718 95.946 97.368
C4 + C5 + C298.811 11.117 98.611 95.000 98.701 94.667 96.711
C2 + C14 + C398.811 11.595 98.611 96.250 98.718 95.946 97.368
C5 + C298.750 11.019 98.611 95.000 98.701 94.667 96.711
C5 + C2 + C1498.750 11.099 98.611 95.000 98.701 94.667 96.711
C4 + C298.733 11.070 98.611 95.000 98.701 94.667 96.711
C2 + C1498.733 11.053 98.611 95.000 98.701 94.667 96.711
C4 + C5 + C397.986 0.764 84.722 100.000 87.912 100.000 92.763
C4 + C397.457 0.984 95.833 87.500 95.890 87.342 91.447
C5 + C397.274 0.937 95.833 88.750 95.946 88.462 92.105
C14 + C397.083 0.961 95.833 86.250 95.833 86.250 90.789
C3 + C1696.215 1.839 91.667 85.000 91.892 84.615 88.158
ROC, receiver operator characteristic; AUC, area under the curve; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine.
Table 3. ROC analysis of free carnitine and specific acylcarnitines levels in dried blood specimens from true-positive and false-positive subjects.
Table 3. ROC analysis of free carnitine and specific acylcarnitines levels in dried blood specimens from true-positive and false-positive subjects.
CharacteristicFree Carnitine and Acylcarnitine
C4C18C5C16C0C14C3C2C10C12C8C6
AUC (%)89.608 87.993 87.420 85.873 85.457 85.314 80.928 74.046 73.420 66.622 63.965 60.888
Best Cutoff Value (μmol/L)0.086 0.224 0.051 1.068 6.934 0.074 0.574 7.763 0.047 0.052 0.035 0.061
Sensitivity (%)73.791 88.550 81.679 61.069 85.751 64.122 68.957 63.359 46.819 47.837 48.346 23.664
Specificity (%)90.278 69.444 80.556 93.056 73.611 91.667 80.556 80.556 90.278 88.889 75.000 94.444
Negative Predictive Value (%)38.690 52.632 44.615 30.455 48.624 31.884 32.222 28.713 23.723 23.792 21.012 18.478
Positive Predictive Value (%)97.643 94.054 95.821 97.959 94.663 97.674 95.088 94.677 96.335 95.918 91.346 95.876
Accuracy (%)76.344 85.591 81.505 66.022 83.871 68.387 70.753 66.022 53.548 54.194 52.473 34.624
AUC, area under the curve; C0, carnitine; C2, acetylcarnitine; C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C6, caproylcarnitine; C8, caprylylcarnitine; C10, actinylcarnitine, C12, lauroylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Table 4. ROC analysis of C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 in dried blood specimens from true-positive (TP), false-positive (FP) and normal subjects.
Table 4. ROC analysis of C4 + C5, C4 + C5 + C18, C4 + C5 + C16 and C18 + C16 in dried blood specimens from true-positive (TP), false-positive (FP) and normal subjects.
Indicator: C4 + C5Indicator: C4 + C5 + C18
CharacteristicTP vs. FPFP vs. NormalTP vs. NormalCharacteristicTP vs. FPFP vs. NormalTP vs. Normal
AUC (%)91.052 88.034 99.983 AUC (%)91.054 77.861 98.880
Best Cutoff Value (μmol/L)0.132 0.233 0.181 Best Cutoff Value (μmol/L)0.441 0.841 0.573
Sensitivity (%)82.952 90.000 100.000 Sensitivity (%)78.880 66.250 92.500
Specificity (%)84.722 75.064 98.750 Specificity (%)86.111 81.170 97.222
Negative Predictive Value (%)47.656 97.360 100.000 Negative Predictive Value (%)42.759 92.197 92.105
Positive Predictive Value (%)96.736 42.353 98.630 Positive Predictive Value (%)96.875 41.732 97.368
Accuracy (%)83.22677.590 99.342 Accuracy (%)80.000 78.647 94.737
Indicator: C4 + C5 + C16Indicator: C18 + C16
CharacteristicTP vs. FPFP vs. NormalTP vs. NormalCharacteristicTP vs. FPFP vs. NormalTP vs. Normal
AUC (%)87.777 63.386 94.045 AUC (%)86.770 62.056 92.561
Best Cutoff Value (μmol/L)0.617 2.760 0.737 Best Cutoff Value (μmol/L)1.539 3.301 0.922
Sensitivity (%)88.295 28.750 98.750 Sensitivity (%)58.015 27.500 92.500
Specificity (%)68.056 90.840 72.222 Specificity (%)97.222 91.603 76.389
Negative Predictive Value (%)51.579 86.232 98.113 Negative Predictive Value (%)29.787 86.124 90.164
Positive Predictive Value (%)93.784 38.983 79.798 Positive Predictive Value (%)99.130 40.000 81.319
Accuracy (%)85.161 80.338 86.184 Accuracy (%)64.086 80.761 84.868
C4, butyrylcarnitine; C5, isovalerylcarnitine; C16, palmitoylcarnitine; C18, octadecylcarnitine.
Table 5. Assesment of the C4 + C5, C4 + C5 + C14, C4 + C14 and C5 + C14 summations using optimal cutoffs determined by ROC analysis.
Table 5. Assesment of the C4 + C5, C4 + C5 + C14, C4 + C14 and C5 + C14 summations using optimal cutoffs determined by ROC analysis.
Indicator: C4 + C5 < 0.181 μmol/LIndicator: C4 + C5 + C14 < 0.281 μmol/L
PCDData AvailablePositiveNegativeSensitivityPCDData AvailablePositiveNegativeSensitivity
727201727201
NoPCDData AvailablePositiveNegativeSpecificityNoPCDData AvailablePositiveNegativeSpecificity
801790.988802780.975
PCD + NoPCDPositive predictive valueNegative predictive valuePCD + NoPCDPositive predictive valueNegative predictive value
0.98610.9731
Indicator: C4 + C14 < 0.201 μmol/LIndicator: C5 + C14 < 0.151 μmol/L
PCDData AvailablePositiveNegativeSensitivityPCDData AvailablePositiveNegativeSensitivity
727201727020.972
NoPCDData AvailablePositiveNegativeSpecificityNoPCDData AvailablePositiveNegativeSpecificity
802780.975806740.925
PCD + NoPCDPositive predictive valueNegative predictive valuePCD + NoPCDPositive predictive valueNegative predictive value
0.97310.9230.974
Indicator: C3 + C16 < 1.839 μmol/LIndicator: C3 + C16 < 2.000 μmol/L
PCDData AvailablePositiveNegativeSensitivityPCDData AvailablePositiveNegativeSensitivity
726660.917726750.931
NoPCDData AvailablePositiveNegativeSpecificityNoPCDData AvailablePositiveNegativeSpecificity
8012680.8508017630.788
PCD + NoPCDPositive predictive valueNegative predictive valuePCD + NoPCDPositive predictive valueNegative predictive value
0.8460.9190.7980.926
C3, propionylcarnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine; C14, myristoylcarnitine; C16, palmitoylcarnitine.
Table 6. Application of the C0 level and C4 + C5 summation using optimal cutoffs determined by ROC analysis in NBS from 2021 to 2024.
Table 6. Application of the C0 level and C4 + C5 summation using optimal cutoffs determined by ROC analysis in NBS from 2021 to 2024.
Indicators (μmol/L)Screening Positive
(N)
Recall
(N)
Recall Negative
(N)
Recall Positive
(N)
Confirmed
(N)
Positive Predictive Value
(%)
C0 < 1040037625612030.75
C0 < 10 & C4 + C5 < 0.1811951851236231.54
C0 < 10 & C4 + C5 > 0.181205188133550NA
9 < C0 < 10 & C4 + C5 > 0.181928867210NA
8.5 < C0 < 10 & C4 + C5 < 0.181837658182NA
ROC, receiver operator characteristic; C0, carnitine; C4, butyrylcarnitine; C5, isovalerylcarnitine.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhi, H.; Chang, S.; Chen, T.; Liang, L.; Qiu, W.; Zhang, H.; Gu, X.; Han, L. Assessment and Application of Acylcarnitines Summations as Auxiliary Quantization Indicator for Primary Carnitine Deficiency. Int. J. Neonatal Screen. 2025, 11, 47. https://doi.org/10.3390/ijns11020047

AMA Style

Zhi H, Chang S, Chen T, Liang L, Qiu W, Zhang H, Gu X, Han L. Assessment and Application of Acylcarnitines Summations as Auxiliary Quantization Indicator for Primary Carnitine Deficiency. International Journal of Neonatal Screening. 2025; 11(2):47. https://doi.org/10.3390/ijns11020047

Chicago/Turabian Style

Zhi, Haijuan, Siyu Chang, Ting Chen, Lili Liang, Wenjuan Qiu, Huiwen Zhang, Xuefan Gu, and Lianshu Han. 2025. "Assessment and Application of Acylcarnitines Summations as Auxiliary Quantization Indicator for Primary Carnitine Deficiency" International Journal of Neonatal Screening 11, no. 2: 47. https://doi.org/10.3390/ijns11020047

APA Style

Zhi, H., Chang, S., Chen, T., Liang, L., Qiu, W., Zhang, H., Gu, X., & Han, L. (2025). Assessment and Application of Acylcarnitines Summations as Auxiliary Quantization Indicator for Primary Carnitine Deficiency. International Journal of Neonatal Screening, 11(2), 47. https://doi.org/10.3390/ijns11020047

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop