Triglycerides, the primary constituents of body fat in humans, are crucial biomarkers for cardiovascular health and metabolic disorders [1, 2]. Accurately assessing serum triglyceride concentrations is essential for epidemiological studies, public health monitoring, and individual patient management [3]. Interpreting triglyceride data requires the consideration of various factors to ensure accuracy and relevance [2].

Total glyceride concentrations constitute the sum of monoglyceride, diglyceride, triglyceride, and free glycerol concentrations [4–9]. For triglyceride measurements, the entity measured is glycerol, which is released following the hydrolysis of tri-, di-, and monoglycerides. The presence of free glycerol is considered an interference [7–10]. In clinical practice, confusion often arises regarding the use of triglycerides versus total glycerides as clinical biomarkers [7, 10]. Analytically, the glycerol non-blanking method, which measures total glycerides, has been used most commonly [7, 10]. In contrast, the glycerol blanking method, which eliminates free glycerol, results in lower triglyceride concentrations than with the glycerol non-blanking method. This phenomenon has also been observed in external quality-assurance programs, including the United States Centers for Disease Control and Prevention (CDC) Lipid Standardization Program (LSP) and the Korean Association of External Quality Assessment Service (KAEQAS) [7–9]. Given that the analytical method for measuring triglycerides in the Korea National Health and Nutrition Examination Survey (KNHANES) changed from glycerol blanking (2005–2021) to glycerol non-blanking (2022), a conversion equation to adjust the differences in triglyceride concentrations is necessary.

In Korea, KNHANES has been instrumental in understanding population health trends and shaping public health policy [11, 12]. Since 1998, KNHANES has monitored triglyceride concentrations, utilizing various analytical methods that reflect technological advances and shifts in scientific understanding [5, 6, 12, 13]. The glycerol blanking method, employed in the KNHANES from 2005 to 2021, offers the advantage of reduced interference from free glycerol. However, comparing its calibrator traceability with the CDC LSP, which assigns target values using methods without glycerol blanking, reveals substantial differences [3, 6, 7, 14, 15]. This shift in analytical methods mirrors broader international trends toward standardizing and enhancing the accuracy and comparability of lipid measurements [6, 7, 14]. Understanding the implications of methodological changes is crucial for interpreting historical data and ensuring continuity in longitudinal studies within KNHANES, which is representative of the Korean population [4, 5].

We developed a conversion equation to harmonize triglyceride 2005–2021 KNHANES data (glycerol blanking methods) with 2022 KNHANES data (glycerol non-blanking methods), enabling adjustment of the 2005–2021 data and illustrating the potential impacts of these methodological changes on research data. The need for such analyses is emphasized by the growing global focus on comparable and reliable health data, which is vital for international health comparisons and monitoring population health trends over time [15, 16].

Differences in the analytical methods used in the KNHANES from 2005 to 2022 are summarized in Table 1. All procedures and protocols of the KNHANES were approved by the Institutional Review Board of the Korea Disease Control and Prevention Agency (KDCA; approval number 2018-01-03-4C-A), with a waiver of informed consent granted under this approval. This study was conducted using anonymized data for the research task named “Quality Control of the Clinical Laboratory for the Korea National Health and Nutrition Examination Survey (KNHANES),” which covers the 2019–2021 (8th) and 2022–2024 (9th) phases. The task, abbreviated as the “Quality Control for KNHANES” task, was designed to improve the quality and reliability of the KNHANES and was conducted under the oversight of the KDCA. The findings are publicly accessible through the Policy Research Information Service & Management database [17]. To develop a conversion equation for triglyceride measurements, we utilized data from a method-comparison study (Quality Control for KNHANES). The comparison study involved comparing data from Seegene Medical Foundation (Neodin), which performed KNHANES triglyceride tests using the glycerol blanking method from 2008 to 2021, and GC Labs, which has been conducting triglyceride tests using the glycerol non-blanking method since 2022.

This study involved 98 fresh leftover serum samples from 98 individuals undergoing routine clinical diagnosis for serum triglycerides at an outpatient clinic or during a healthcare check-up. On January 6, 2022, the specimens were collected at the Asan Medical Center Laboratory, the affiliated institution of the principal investigator for the Quality Control for KNHANES task. Specimens were collected to cover the analytical measurement interval shared by both methods (9–885 mg/dL, 0.10–10.0 mmol/L) and included specimens distributed across low-to-high concentrations according to the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III criteria (normal concentrations, <150 mg/dL [1.70 mmol/L]; very high concentrations, ≥500 mg/dL [5.65 mmol/L]) [1]. The specimens were prepared and refrigerated at the Asan Medical Center Laboratory on the same day of collection. Each sample was divided into two aliquots to provide identical testing materials for comparative analysis. The aliquots were concurrently sent to Seegene Medical Foundation and GC Labs (1 h transport time) on the specimen collection date under refrigeration, where the triglycerides were analyzed immediately. At Seegene Medical Foundation, triglyceride analysis was conducted using the glycerol blanking method with Qualigent TG Reagent Kits (Sekisui, Japan; lot 148RBS) and a Qualigent N calibrator (Sekisui; lot 119RES) on a Labospect 008AS analyzer (Hitachi, Japan), and the method was designated as the X method. Conversely, GC Labs performed measurements using a non-glycerol blanking method with TRIGL Reagent Kits (Roche, China; lot 599110) and a certified functional activity standard calibrator (Roche, Germany; lot 539840) on a Cobas 8000 c702 analyzer (Roche, Germany), where this method was designated as the Y method. Both laboratories analyzed the triglycerides in duplicate to ensure the reliability and reproducibility of the data.

Fig. 1 presents the outline of our study. Data from 98 fresh, refrigerated serum specimens demonstrated significantly higher triglyceride concentrations with the glycerol non-blanking method (2022) than with the glycerol blanking method (2005–2021), with an average difference of 10.7 mg/dL (0.12 mmol/L, 95% CI: 8.9–12.6 mg/dL [0.10–0.14 mmol/L]) and an average percent difference of 10.0% (95% CI: 8.3–11.7). At the 200 mg/dL (2.26 mmol/L) triglyceride concentration (the cutoff for hypertriglyceridemia according to NCEP ATP III criteria and the KNHANES annual prevalence-assessment standards), the estimated and percent differences were 9.9 mg/dL (0.11 mmol/L) and 5.0%, respectively. The estimated differences at the cutoffs for normal (150 mg/dL [1.70 mmol/L]) and very high (500 mg/dL [5.65 mmol/L]) concentrations were 10.4 mg/dL (0.12 mmol/L, 6.9%) and 6.9 mg/dL (0.08 mmol/L, 1.4%), respectively.

Passing–Bablok regression analysis indicated a high correlation between both methods with a correlation coefficient (r=0.997). This regression analysis established the relationship between both methods with the equation y (mg/dL)=11.94 (mg/dL)+0.99x, where x represents triglyceride measurements using the glycerol blanking method at Seegene Medical Foundation, and y denotes corresponding values from the non-glycerol blanking method at GC Labs (Fig. 2). The 95% CI of the slope includes 1.0 (0.9772–1.013), but the 95% CI of the intercept does not include 0.0 (9.50–14.76), signifying significant differences between the analytical methods.

The conversion equation y (mg/dL; glycerol non-blanking, 2022)=11.94 (mg/dL)+0.99x (glycerol blanking, 2005–2021) was applied to 2019–2021 data from the KNHANES (N= 16,015) to demonstrate data conversion. The mean triglyceride concentrations measured in the KNHANES (2019–2021; N= 16,015) shifted from 123.7 mg/dL (1.40 mmol/L, 95% CI: 122.2–125.1 mg/dL [1.38–1.41 mmol/L]) to 134.3 mg/dL (1.52 mmol/L, 95% CI: 132.9–135.8 mg/dL [1.50–1.53 mmol/L]) after conversion. Fig. 3 displays Kernel density plots of serum triglyceride concentrations from the KNHANES. Before conversion, the KNHANES data from 2019–2021 and 2022 (blue and green lines, respectively) showed similar distributions. After applying the conversion equation y (mg/dL)=11.94 (mg/dL)+0.99x to data from 2019–2021, the triglyceride concentrations shifted to the right (red line), resulting in an increased prevalence of hypertriglyceridemia (≥200 mg/dL [2.26 mmol/L]) from 11.9% to 13.8% (unweighted) during 2019–2021.

During the KNHANES (2018–2022), each laboratory participated in CDC LSP surveys, and the results were available for triglyceride measurements. Fig. 4 displays the percent bias estimation for each round of CDC LSP survey materials from the KNHANES from 2018 to 2022. As the target values for triglycerides in the CDC LSP were assigned to total glycerides, the KNHANES triglyceride data for 2018–2021 (which eliminated free glycerol from total glycerides using the glycerol blanking method) differed significantly from the target values of the CDC LSP (exceeding −5% in terms of the bias limit, within –15% for the total allowable error limit). Conversely, the glycerol non-blanking methods used in the KNHANES during 2022 showed acceptable results, with biases within ±5.0% according to NCEP ATP III criteria for bias estimations.

The historical evolution of triglyceride measurement methods and assay standardization reflects advancements in technology, improved scientific understanding, and the pursuit of accuracy and reliability [3]. Standardizing analytical methods, including proficiency testing with reference materials and expanding global clinical research, has ensured consistent and comparable triglyceride measurements across diverse populations and various settings [3, 6, 16, 21–24].

We developed a conversion equation, y (triglyceride concentration without glycerol blanking; mg/dL)=11.94 (mg/dL)+0.99x (triglyceride concentration with glycerol blanking; mg/dL), which demonstrated a strong correlation (r=0.997) and an average bias of 10.0%. In mmol/L, the equation was y=0.13 (mmol/L)+ 0.99x. This finding aligns with previous data showing that y (triglyceride concentration without glycerol blanking; mmol/L)= 0.1623 (mmol/L)+1.012x (triglyceride with glycerol blanking; mmol/L), with a conversion factor of mmol/L×88.5=mg/dL (corresponding intercept y: 14.36 mg/dL) [3]. Prior studies have reported an average bias of 16.7% higher triglyceride concentrations with glycerol non-blanking methods than with glycerol blanking methods, with an estimated bias of 8.2% at a triglyceride concentration of 203 mg/dL (2.29 mmol/L) [3]. These results are consistent with findings from the KAEQAS lipid proficiency testing program (2016–2018) [25], which indicated a total glyceride concentration (glycerol non-blanking, mg/dL) of 6.87 (mg/dL)+1.01×triglyceride concentration with glycerol blanking (mg/dL) for target-value comparison [total glyceride (mmol/L)=0.08 mmol/L+1.01×triglyceride with glycerol blanking (mmol/L)]. When comparing the average values of all participants, the proficiency testing material showed a total glyceride concentration (glycerol non-blanking, mg/dL) of 8.04 (mg/dL)+ 0.99× the triglyceride concentration with glycerol blanking (mg/dL) [total glycerides (mmol/L)=0.09 mmol/L+0.99×the triglyceride with glycerol blanking (mmol/L)] [25]. Our current findings confirm these observations, indicating a generally greater negative bias in specimens with lower triglyceride concentrations [7, 9, 14, 26]. Free glycerol levels can increase due to factors such as exercise, liver disease, diabetes, hemodialysis, stress, and glycerol-containing medications, whereas decreases may occur in lipid metabolism disorders or treatments affecting lipolysis. Substances such as vitamin C, acetaminophen, N-acetyl cysteine, and bilirubin can cause negative bias in glycerol non-blanking methods, and variability in sample handling can further contribute to inaccuracies [2, 3, 9, 14, 26, 27]. Caution is required when applying the conversion equation in individuals with altered free glycerol concentrations. Future studies should evaluate the impact of analytical method changes on these subgroups [9, 27]. However, quantifying free glycerol to determine true triglyceride concentration is typically of limited importance in most clinical scenarios, as the potential error due to free glycerol is <9 mg/dL (0.10 mmol/L) in 99% of individuals [9].

Standardization of analytical methods is essential to ensure accuracy and consistency with reference measurement procedures (RMPs) for analytes across various measurement systems, laboratories, and time frames [22–24]. The primary RMP and standard reference material (SRM), namely the National Institute of Standards and Technology (NIST) SRM 1595 Trialmitin (a pure chemical SRM certified in 1983 and updated to NIST SRM 1595a in March 2024 for measuring triglycerides) were established by the NIST [28]. The US CDC developed the secondary RMP and secondary SRM in 1963, initially using chemical methods to reduce free glycerol, which were cumbersome for routine clinical applications [7, 9, 14]. These methods were replaced by isotope dilution mass spectrometry (IDMS) in 2012 to measure total glycerides, thereby improving standardization of routine clinical serum-glyceride measurements considering that most clinical laboratories now use glycerol non-blanking methods [9, 14]. During RMP development, NIST SRM 1595 was used as a calibrator, and NIST SRM 1951b (comprising a frozen human serum matrix with certified concentrations of total glycerides [I and II] and triglycerides [I]), along with additional frozen serum samples, were utilized for precision and bias estimations. However, a direct comparison between older enzymatic chromotropic acid methods and the newly developed CDC IDMS method was not performed [7].

The National Cholesterol Education Program (NCEP) recommends the CDC reference method as the standard, and NCEP ATP III-derived guidelines were based on CDC-developed secondary RMPs [5, 7, 9]. CDC LSP certification is valuable for estimating bias in terms of target concentrations and precision [7, 15, 29]. Using CDC LSP is crucial for comparing epidemiological data, as demonstrated by Finland’s National Public Health Institute, which ensures accuracy through robust quality-assurance methods [14, 15, 23, 24, 29, 30]. In Korea, the National Health Insurance Service (NHIS) generates triglyceride data through its health check-up program for most residents, with data generated by hundreds of hospitals using diverse analytical methods [31]. According to a 2016–2018 KAEQAS report, two-thirds of laboratories participating in the accuracy-based lipid proficiency testing program used glycerol non-blanking methods [23, 25]. During the second survey performed in 2022, 76.4% of survey participants used glycerol non-blanking methods for triglyceride measurements (approximately half of them used analytical methods by Roche) [25]. This proportion is comparable with data from CDC LSP-certified participants [29].

We evaluated the impact of changing analytical methods on triglyceride measurements within the KNHANES framework. Even minor systematic biases may have significant clinical and epidemiological implications. However, discrepancies in test values that do not affect classification thresholds may not alter prevalence rates [9, 14]. Considering that most clinical laboratories globally use glycerol non-blanking methods and that clinicians interpret triglyceride results without detailed knowledge of the analytical methods used, researchers relying on KNHANES data must recognize these changes. Therefore, applying the conversion equation judiciously is essential to preserve research integrity [5].

Our study has limitations, including differences in population characteristics, an uneven distribution of triglyceride concentrations, and a lack of data on reagent and calibrator lots. Comparisons were restricted to a single lot of reagents and calibrators for each analytical method, which may not account for potential variations between different lots employed. Variations in sample sizes between KNHANES data and the conversion equation study, which used a relatively small sample size, might have influenced the observed differences and could have affected the generalizability of our findings, raising concerns about potential overcorrections when applying the conversion equation. Further studies must be conducted to evaluate population-specific prevalences, demographic differences, and their underlying causes and impacts, as statistical correction methods can vary depending on the specific research objectives and population characteristics. Furthermore, cross-validation must also be performed to assess the impact of transitioning from glycerol blanking to non-blanking methods on cardiovascular disease risk monitoring in epidemiological studies.

In conclusion, we developed conversion equations to align historical data with current triglyceride measurement methods, ensuring robust and valid conclusions in longitudinal studies. Our findings contribute to the methodological literature on triglyceride measurements and offer a model for other epidemiological surveys undergoing similar transitions. Global standardization of biochemical triglyceride measurements in health surveys improves comparability and provides critical insights for guiding future methodological decisions in similar health examination surveys worldwide.