Method Comparison of Dried Blood Spot and Plasma Apolipoprotein B Measurement on a Semi‐Automated Analyser: Toward Accessible Lipid Profiling in Resource‐Limited Settings

ABSTRACT

Apolipoprotein B (ApoB) is a key marker of atherogenic lipoprotein burden, but conventional plasma‐based testing requires venous sampling and centralized laboratory infrastructure. Dried blood spot (DBS) sampling offers a minimally invasive alternative suitable for decentralized settings. This study evaluated the analytical performance of a DBS‐based ApoB assay on the Chem7 semi‐automated analyser and compared it with the Abbott ARCHITECT ci4100 plasma reference method. DBS samples prepared from 50 de‐identified EDTA whole‐blood specimens were extracted in saline and analysed using an immunoturbidimetric ApoB assay on the Chem7 analyser with a correction factor of 2 applied for haematocrit dilution. Paired plasma specimens were analysed on the ARCHITECT ci4100. Method comparison included Passing–Bablok and Deming regression and Bland–Altman analysis. Potential outliers were assessed using Grubbs’ test (α = 0.05). Precision verification followed CLSI EP15‐A3. Corrected DBS ApoB showed strong correlation with plasma values (r = 0.97; R ² = 0.94). Passing–Bablok regression showed a slope of 0.872 and intercept of 14.29 mg/dL, consistent with Deming regression. Bland–Altman analysis demonstrated a mean bias of −1.5 mg/dL with acceptable limits of agreement. Eighty percent of results were within ±10% of plasma values. No statistically significant outliers were identified, and precision estimates (within‐day and between‐day coefficients of variations (CVs) of 4.6% and 9.2%) met CLSI criteria. DBS‐based ApoB measurement on the Chem7 analyser provides reliable, reproducible and clinically acceptable agreement with plasma testing, supporting its applicability in decentralized and resource‐limited settings.

1. Introduction

Cardiovascular disease (CVD) remains the leading global cause of mortality, accounting for nearly one‐third of all deaths worldwide [1]. Among lipid‐related biomarkers, apolipoprotein B (ApoB) has emerged as a robust and mechanistically relevant indicator of atherogenic risk [2]. Each ApoB‐containing lipoprotein particle—including LDL, VLDL and Lp(a)—carries one molecule of ApoB; therefore, total ApoB directly reflects the number of atherogenic particles rather than lipid content alone [3]. Contemporary clinical guidelines increasingly endorse ApoB as a superior alternative to LDL cholesterol for cardiovascular risk assessment, especially in patients with hypertriglyceridemia or familial hypercholesterolaemia (FH) [4, 5].

Despite its clinical relevance, ApoB testing remains largely confined to centralized laboratories, requiring venous blood samples and automated immunoassay platforms [6]. This dependency restricts its implementation in primary care, rural and resource‐limited settings, where phlebotomy, cold‐chain transport and advanced instrumentation are often unavailable.

Dried blood spot (DBS) sampling offers an innovative, minimally invasive alternative. A simple finger‐prick blood drop dried on filter paper enables convenient transport and storage at ambient temperature without cold‐chain logistics [7, 8, 9]. DBS‐based testing has been successfully adapted for infectious disease screening and selected biochemical assays [10, 11, 12, 13, 14]; however, translation to lipid and apolipoprotein measurement has remained challenging due to matrix effects, small sample volumes and variable elution efficiency [15, 16, 17, 18]. To date, no clinically evaluated or methodologically standardized DBS‐based ApoB assay has been reported.

In previous exploratory work, our laboratory attempted to adapt automated immunoturbidimetric platforms (Abbott ARCHITECT ci4100, Beckman AU480) for DBS eluates, but results were inconsistent, likely due to matrix interference and limited analytical sensitivity (unpublished data). To overcome these challenges, we developed and evaluated a DBS‐based ApoB assay using the Chem7 semi‐automated biochemistry analyser, a compact, low‐cost instrument capable of endpoint colorimetric and immunoturbidimetric measurements. By optimizing DBS extraction and applying a standardized haematocrit correction factor, we sought to achieve plasma‐equivalent ApoB quantification suitable for decentralized use.

This study presents the first comprehensive comparison of DBS and plasma ApoB measurements on a semi‐automated analyser, evaluated against the Abbott ARCHITECT ci4100 reference method. Analytical performance was assessed following CLSI EP09‐A3 and EP15‐A3 guidelines [19, 20], incorporating Passing–Bablok and Deming regression, Bland–Altman agreement and Grubbs’ outlier testing. The findings demonstrate that ApoB quantification from DBS is feasible, accurate and reproducible, supporting its potential for deployment in low‐resource laboratory networks.

2. Materials and Methods

2.1. Sample Collection and Preparation

Residual EDTA whole‐blood specimens (n = 50) from routine lipid testing were de‐identified and used for DBS preparation. Approximately 50 µL of whole blood was spotted onto standardized filter paper cards and air‐dried overnight (≥4 h) at ambient temperature. Each DBS card was stored in a sealed plastic bag containing desiccant at 2°C–8°C until analysis.

Matched plasma from the same specimens was analysed using the Abbott ARCHITECT ci4100 analyser with the manufacturer's ApoB immunoturbidimetric assay, which quantifies ApoB based on turbidity generated by insoluble immune complexes formed between ApoB and anti‐ApoB antibodies. The assay was calibrated and quality‐controlled per manufacturer specifications and traceable to international ApoB reference standards. Results were expressed in mg/dL and served as the reference method for comparison.

DBS samples in this study were prepared from venous EDTA whole blood rather than capillary finger‐prick blood. This approach enabled direct pairing with matched plasma obtained from the same specimen for analytical method comparison. Although venous‐derived DBS is commonly used during early assay validation, potential differences between venous and capillary DBS, such as haematocrit distribution and microvascular composition, were not directly evaluated in this study.

2.2. DBS Extraction and Processing

Each 3.2 mm DBS punch was incubated at 2°C–8°C for 12–16 h in 125 µL of 0.9% saline to elute ApoB. After incubation, the eluate was separated from the filter matrix, clarified by brief centrifugation, and used directly for analysis. All extractions yielded uniform final volumes, ensuring analytical consistency across samples.

2.3. ApoB Assay on the Chem7 Analyser

DBS extracts were analysed on the Chem7 semi‐automated biochemistry analyser (Erba Mannheim, Czech Republic), a compact, bench‐top system capable of performing end‐point colorimetric and immunoturbidimetric assays.

For each assay, 200 µL of Reagent 1 (R1) was pipetted into a reaction tube, followed by 100 µL of DBS extract, gently mixed and incubated for 5 min at room temperature. Then, 50 µL of Reagent 2 (R2) was added, and the mixture was incubated for another 5 min. The increase in turbidity was measured at 340 nm, and the ApoB concentration was calculated against a five‐point serum‐based calibration curve generated using manufacturer‐supplied ApoB calibrators (Erba Mannheim, Czech Republic).

Each calibrator had a known ApoB concentration traceable to IFCC reference material, allowing quantification in mg/dL. The calibration curve was fitted linearly, with analyser‐generated slope and intercept values periodically verified in Microsoft Excel to confirm linearity and stability. Calibration parameters were maintained throughout the reagent lot's validity to ensure inter‐run consistency.

Because DBS eluates derive from whole blood (∼50% haematocrit), measured ApoB concentrations were approximately half of plasma levels. Accordingly, a correction factor of 2 was applied to all DBS values to obtain plasma‐equivalent concentrations. This empirically derived correction was confirmed in preliminary validation experiments, which demonstrated consistent proportionality between DBS and plasma results across the measurement range.

Potential turbidity or matrix interference was evaluated by duplicate measurement, baseline blanking and visual inspection of eluates; no abnormal background absorbance or drift was observed.

2.4. Plasma ApoB Reference Measurement

Paired plasma samples were analysed on the Abbott ARCHITECT ci4100 immunoturbidimetric ApoB assay as the reference method. This method is traceable to internationally recognized ApoB standards and is widely used in clinical lipid profiling. The plasma ApoB concentration (mg/dL) obtained from this system served as the comparative benchmark (gold standard) for evaluating DBS‐based results obtained from the Chem7 analyser.

2.5. Statistical Analysis

All analyses were conducted according to CLSI EP09‐A3 (Measurement Procedure Comparison and Bias Estimation Using Patient Samples) and EP15‐A3 (User Verification of Precision and Estimation of Bias) guidelines.

Agreement between DBS and plasma ApoB concentrations was assessed using Passing–Bablok regression (for proportional and constant bias) and Deming regression (for error in both axes). Bland–Altman analysis was used to estimate the mean bias and 95% limits of agreement (LoA), expressed both in absolute (mg/dL) and relative (%) terms.

Outliers were evaluated using the Grubbs’ test (α = 0.05) applied to the bias distribution (DBS − plasma). This test detects statistically significant extreme values in a normally distributed dataset based on their deviation from the group mean. Samples identified as potential outliers were further reviewed for possible analytical or pre‐analytical causes (e.g., spotting inconsistency, incomplete elution or pipetting variation) before inclusion in regression and agreement analyses.

2.6. Precision and Reproducibility Testing

Precision verification was conducted following CLSI EP15‐A3 guidelines to evaluate both within‐day and inter‐day performance of the DBS‐based ApoB assay. All 50 samples were analysed across four consecutive runs performed under standardized conditions using the same reagent lot, calibration parameters and instrument settings. For each sample, ApoB concentrations from the four runs were used to calculate within‐sample coefficients of variation (CV%) and determine analytical precision. The mean within‐day CV and inter‐day CV were subsequently derived to characterize repeatability and reproducibility, respectively. Between‐run variability was further assessed using one‐way analysis of variance (ANOVA) to verify the absence of statistically significant drift across runs.

2.7. Use of Artificial Intelligence (AI) Tools

Portions of this manuscript (including language editing, grammar refinement and assistance with figure/illustration generation) were supported using the large language model ChatGPT (OpenAI). All scientific content, data interpretation, experimental design and conclusions were developed and verified entirely by the authors. The authors take full responsibility for the accuracy and integrity of the manuscript. ChatGPT was not listed as an author and did not contribute to conceptual, analytical or decision‐making aspects of the study.

3. Results

3.1. Correlation Between DBS and Plasma ApoB

DBS ApoB results obtained from the Chem7 analyser (after applying a correction factor of 2) demonstrated a strong linear association with plasma ApoB values measured on the Abbott ARCHITECT ci4100. The overall Pearson correlation coefficient was r = 0.968 (p < 0.001; R ² = 0.937), confirming excellent comparability between methods (Figure 1).

FIGURE 1.

Method comparison between dried blood spot (DBS) and plasma apolipoprotein B (ApoB) measurements. Passing–Bablok and Deming regression analyses comparing CF2‐corrected DBS ApoB concentrations measured on the Chem7 semi‐automated analyser with paired plasma ApoB concentrations measured on the Abbott ARCHITECT ci4100 reference system (n = 50). The dashed line represents the line of identity (y = x). Both regression models demonstrate strong linear agreement with minor proportional and constant bias across the analytical range.

3.2. Method Comparison and Regression Analysis

Following the CLSI EP09‐A3 guideline, both Passing–Bablok and Deming regression analyses were performed to evaluate proportional and constant bias. Passing–Bablok regression yielded a slope of 0.872 (95% CI 0.811–0.933) and an intercept of 14.29 mg/dL (95% CI 7.62–20.34).

Deming regression produced consistent results, with a slope of 0.85 (95% CI 0.80–0.90) and an intercept of 17.4 mg/dL (95% CI 10.2–24.6). The implied x‐intercept (≈−16 mg/dL) indicated a small positive bias at low concentrations, whereas the slope below unity reflected a minor proportional underestimation at higher concentrations.

Both regression models were statistically concordant, confirming proportional agreement with modest constant bias (Figure 1).

3.3. Agreement and Bias Evaluation

Agreement between DBS‐Chem7 (CF2) and plasma ApoB was assessed by Bland–Altman analysis in accordance with CLSI EP09‐A3. The absolute Bland–Altman plot showed a mean bias of +1.48 mg/dL (95% CI −1.29 to +4.26) with 95% LoA of −18.12 to +21.09 mg/dL.

The relative Bland–Altman plot, expressed as percent difference, showed a mean bias of +4.01% (95% relative LoA −20.98% to +28.99%), reflecting analytical and biological variability typical of immunoturbidimetric assays (Figures 2 and 3).

FIGURE 2.

Bland–Altman analysis of absolute bias between DBS and plasma ApoB measurements. Bland–Altman plot showing the absolute difference (DBS − plasma, mg/dL) plotted against the mean ApoB concentration for paired samples (n = 50). The solid horizontal line represents the mean bias (+1.48 mg/dL), and the dashed lines indicate the 95% limits of agreement (−18.12 to +21.09 mg/dL). Most data points fall within the limits of agreement, indicating clinically acceptable concordance. ApoB, apolipoprotein B; DBS, dried blood spot.

FIGURE 3.

Bland–Altman analysis of relative bias between DBS and plasma ApoB measurements. Relative Bland–Altman plot expressing percent difference [(DBS − plasma)/plasma × 100] versus the mean ApoB concentration for paired samples (n = 50). The solid line indicates the mean relative bias (+4.01%), with dashed lines representing the 95% limits of agreement (−20.98% to +28.99%). The distribution demonstrates acceptable relative agreement across the measurement range.

Although the confidence interval for the mean bias included zero, the narrow dispersion confirmed clinically acceptable agreement across the measurement range.

All regression and Bland–Altman analyses were performed using the complete dataset (n = 50). Sensitivity analysis excluding two practical outliers (≈±29 mg/dL) produced comparable regression slopes and bias estimates, confirming their inclusion did not distort overall agreement.

3.4. Stratified Analysis Across Concentration Range

To evaluate potential concentration‐dependent effects, results were stratified at 118 mg/dL, corresponding to the dataset median and mid‐range of the analytical interval (30–180 mg/dL).

Range	n	r	Mean bias (mg/dL)
≤118 mg/dL	28	0.907	+6.09
>118 mg/dL	22	0.836	−4.37

No significant trend in proportional bias was observed between the two strata, confirming consistent performance across the assay's working range.

3.5. Outlier and Interference Assessment

Outlier detection was performed using the Grubbs’ test (α = 0.05) applied to the bias distribution (DBS—plasma). The mean bias was +2.86 ± 10.64 mg/dL (n = 50). The calculated Grubbs statistic (G = 3.33) exceeded the critical threshold (Gcrit = 2.69), identifying one statistically significant outlier (Sample 20; plasma = 65 mg/dL, DBS = 103.3 mg/dL, bias = +38.3 mg/dL). A second sample (Sample 19; plasma = 180 mg/dL, DBS = 156.2 mg/dL, bias = −23.8 mg/dL) showed large deviation but was below the significance threshold (G = 2.51 < Gcrit) (Table S1).

Both samples were examined for potential analytical and pre‐analytical causes, including spotting volume, elution completeness and pipetting accuracy. No spotting irregularities, incomplete elution or reagent handling deviations were observed, suggesting that the deviations likely represent normal analytical variability at concentration extremes rather than procedural error.

Excluding these two points slightly adjusted the mean bias from +2.86 to −0.8 mg/dL, without materially affecting the regression slope, correlation (r = 0.97) or Bland–Altman LoA. Therefore, both samples were retained in the final analysis for completeness and transparency.

Potential turbidity or matrix interference was evaluated through replicate testing, blank‐baseline verification and visual inspection of DBS eluates. No abnormal background absorbance, drift or haemolytic interference was observed, confirming analytical robustness of the DBS‐based ApoB assay.

3.6. Precision and Reproducibility

Four‐run reproducibility testing was performed across all 50 DBS samples using CF2‐corrected ApoB results on consecutive days. The within‐sample CV across runs averaged 9.7% (median 6.1%; range 0.27%–48.06%). One‐way ANOVA (p = 0.545) and the non‐parametric Friedman test (p = 0.261) showed no statistically significant differences among the four runs, indicating stable inter‐day performance without systematic drift. These findings are consistent with acceptable precision for immunoturbidimetric assays in alignment with CLSI EP15‐A3.

3.7. Clinical Agreement Summary

Eighty percent (40/50) of DBS‐Chem7 results were within ±10% of paired plasma values, satisfying a practical threshold for clinical interchangeability. The remaining 20% deviated beyond this limit but remained within the expected analytical variability for immunoturbidimetric assays.

Collectively, these results confirm that the DBS‐based ApoB method on the Chem7 analyser provides high correlation, acceptable precision and clinically negligible bias relative to the plasma reference method, supporting its applicability for decentralized lipid risk assessment.

4. Discussion

This study demonstrates a novel, practical approach for quantifying ApoB from DBSs using the semi‐automated Chem7 clinical chemistry analyser. To our knowledge, this is the first successful DBS‐based ApoB assay evaluated on a low‐cost, bench‐top platform suitable for decentralized testing. The method exhibited excellent correlation with the plasma reference (Abbott ARCHITECT ci4100), with r = 0.97, R ² = 0.94 and minimal bias (+1.48 mg/dL) within the clinically acceptable range. Both Passing–Bablok and Deming regression confirmed strong proportional agreement, whereas Bland–Altman analysis indicated tight absolute LoA (−18.1 to +21.1 mg/dL). The method's reproducibility (inter‐assay CV ∼8%–10%) compares favourably with conventional immunoturbidimetric assays.

4.1. Analytical Performance and Outlier Behaviour

Eighty percent of DBS results were within ±10% of their plasma counterparts, underscoring reliable comparability for most samples. Two specimens exhibited large absolute deviations (≈±29 mg/dL). Grubbs’ test (α = 0.05) identified only one statistically significant outlier, and sensitivity analysis showed that excluding both did not materially alter regression parameters or bias estimates. Follow‐up review found no procedural anomalies, suggesting that outlier variability stemmed from expected matrix effects at extreme concentrations rather than analytical error [24, 25]. Together, these results affirm the analytical robustness of the DBS–Chem7 approach.

4.2. Haematocrit Correction and Matrix Considerations

A fixed correction factor of 2× was applied to account for the plasma fraction of whole blood (∼50% haematocrit), effectively harmonizing DBS and plasma ApoB concentrations. This simple adjustment enabled quantitative alignment across the clinical range (30–180 mg/dL). Although haematocrit variability may influence DBS recovery, the observed mean bias (<2% of average ApoB) remains analytically and clinically insignificant under NCEP guidelines [21]. Future studies incorporating individualized haematocrit or haemoglobin measurements, and exploring ApoB/Hb (mg/g Hb) normalization, may further refine precision in populations with anaemia, polycythemia or pediatric physiology.

4.3. Comparison With Previous Studies

Earlier DBS‐based lipid or apolipoprotein methods—typically using ELISA, radial immunodiffusion or nephelometry—were limited by low throughput, sample‐volume constraints and poor transferability to clinical analysers [22, 23, 26, 27, 28, 29]. Attempts to adapt fully automated platforms (e.g., Abbott ARCHITECT, Beckman AU480) to DBS matrices failed to achieve reliable quantification, likely due to inadequate elution efficiency or high background absorbance [26, 27, 28]. The present study demonstrates that, with optimized extraction and matrix‐specific calibration, the Chem7 analyser can overcome these barriers and deliver reproducible, clinically relevant ApoB data from DBS samples.

4.4. Clinical Implications

A validated DBS ApoB assay has significant translational potential for resource‐limited and decentralized healthcare settings. Finger‐prick sampling can eliminate venipuncture requirements, reduce cold‐chain dependency and enable field collection for lipid screening, epidemiological surveys or longitudinal follow‐up in cardiovascular‐risk programs. DBS‐based ApoB measurement could enhance FH detection in pediatric or rural populations, support population‐level atherogenic risk assessment and facilitate large‐scale public health monitoring where conventional plasma testing is impractical. Given the growing evidence that ApoB outperforms LDL‐C as an indicator of atherogenic burden, this method could broaden access to superior cardiovascular diagnostics [30, 31, 32, 33].

4.5. Limitations and Future Work

An important limitation of this study is that DBS samples were generated from venous EDTA whole blood rather than capillary finger‐prick blood. Although many analytes demonstrate comparable performance between venous and capillary DBS, this equivalence cannot be assumed without direct empirical evaluation. Capillary blood may differ in haematocrit distribution, microvascular composition and potential interstitial fluid admixture, all of which could influence DBS recovery and analyte quantification. Therefore, although the current results demonstrate strong analytical alignment between venous‐derived DBS and plasma ApoB measurements, additional validation using capillary DBS is required before full clinical interchangeability can be established.

This validation was conducted on a modest sample set (n = 50) from a single laboratory cohort. The fixed haematocrit correction, though effective here, may not fully account for physiological extremes. Additional studies should assess linearity, limit of detection, limit of quantification and extended DBS stability (≥1–3 months, under variable storage conditions). This comprehensive evaluation provided a robust estimate of within‐run and between‐run variability, consistent with CLSI EP15‐A3 recommendations. Automation of DBS punching and extraction or direct‐elution techniques could improve throughput for routine or epidemiologic use.

5. Conclusion

In summary, this study establishes a proof‐of‐concept for DBS‐based ApoB quantification using venous‐derived DBS on a semi‐automated analyser. The assay demonstrated high correlation, minimal bias and acceptable precision relative to the plasma reference method. Although the present findings support the analytical feasibility of DBS‐based ApoB measurement, full clinical interchangeability cannot yet be assumed, particularly for capillary finger‐prick DBS, which was not directly evaluated in this study. Nevertheless, the observed analytical performance supports the potential utility of this approach for population screening, risk stratification and decentralized testing applications. Further large‐scale validation incorporating capillary DBS, individualized haematocrit correction, extended stability assessment and clinical outcome correlation will be essential to establish its role as a scalable tool for global cardiovascular risk assessment.

Author Contributions

Bronwyn Pitampersad: investigation, data curation, methodology, formal analysis, writing – original draft. Savathree Madurai: conceptualization, methodology, supervision, writing – review and editing. Bhavani Manivannan: methodology, formal analysis, validation, statistical analysis, writing – review and editing. Kayla Pillay: investigation, data curation, project administration. Sharana Mahomed: conceptualization, supervision, resources, writing – review and editing.

Funding

The authors have nothing to report.

Ethics Statement

This study involved analytical method comparison using anonymized, residual clinical specimens. In accordance with the institutional policy and national guidelines, formal ethics committee approval was not required because no identifiable patient information was collected or used.

Consent

Patient consent was waived as the study utilized de‐identified leftover samples, and no personal health information was accessed.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: ansa70067‐sup‐0001‐SuppMat.docx.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request. Individual participant‐level data are not publicly available due to privacy and confidentiality restrictions.