CHROMATOGRAM-BASED PREDICTORS OF HbA1c REPORTINGFAILUREANDPOSITIVEBIASONBIO-RADHPLC IN SAMPLES WITH ABNORMAL HAEMOGLOBIN PEAKS

Abstract

Background:Abnormal haemoglobin peaks are frequently observed on Bio-Rad cation-exchangeHPLC for glycated haemoglobin (HbA1c), particularly in regions with a high prevalence ofhaemoglobinopathies. These peaks may cause assay failure or clinically relevant bias, but routinedata comparing Bio-Rad HPLC with immunoassay methods under such conditions are limited.
Objectives:TodeterminethefrequencyofHbA1creportingfailureonBio-RadD-10/VariantHPLCin the presence of abnormal peaks, quantify bias versus a Roche Cobas immunoturbidimetric assay,and identify chromatographic predictors of failure and large method differences.
Methods:In this retrospective laboratory study, we analysed 156 unique EDTA blood samples inwhich Bio-Rad D-10 or Variant chromatograms displayed at least one abnormal haemoglobin peakand a paired HbA1c result was available from a Roche Cobas analyser. Abnormal-peak type, areaunder the curve (AUC, % of total haemoglobin), retention times of HbA1c and the abnormal peak,and the distance between peaks were recorded. Bio-Rad results were classified as failures whenHbA1cwasreportedas0%or>20%despiteanumericCobasresult.Insampleswithnumericresultson both methods, we calculated the difference (ΔHbA1c = Bio-Rad − Cobas), the absolutedifference (|ΔHbA1c|), correlation and Bland–Altman limits of agreement. Logistic regression wasused to identify predictors of Bio-Rad failure.
Results:Variant-window peaks were the most frequent abnormality (107/156; 68.6%). Based onCobasHbA1c,61patients(39.1%)werenormoglycaemic,37(23.7%)hadprediabetesand58(37.2%)haddiabetes.Bio-Radfailedin26/156chromatograms(16.7%),reportingHbA1cas0%in23and>20%in3cases.FailurewasmorefrequentontheVariantthanontheD-10(Variant14/34,41.2%vsD-1012/122,9.8%).In the130analyzablepairs,meanHbA1cwas7.20%by Bio-Radand6.62% byCobas,withameanbiasof0.58%andlimitsofagreementfrom-1.1to2.2%.Overall,28.5%ofpaired results differed by≥1.0 percentage point. Abnormal-peak AUC showed a modest positivecorrelation with |ΔHbA1c| (r=0.19), whereas greater peak-distance proportion was associated withsmallerdifferences(r=-0.21).Inmultivariableanalysis,each1%increaseinAUCincreasedtheoddsof Bio-Rad failure (odds ratio 1.12; 95% CI 1.08–1.17), whereas each 0.1-min increase in peakdistance reduced the odds (odds ratio 0.34; 95% CI 0.19–0.60).
Conclusions:Insampleswithabnormalhaemoglobinpeaks,Bio-RadHPLCfrequentlyyieldsnon-reportable or positively biased HbA1c values when compared with a Roche Cobasimmunoturbidimetric assay. Large variant-window or unknown peaks close to the HbA1c windoware the main determinants of failure and bias. Simple chromatogram-based rules using peak area andretention-time separation may help laboratories flag high-risk chromatograms and trigger reflextesting on an alternative HbA1c method.