Imprecision of tumour biomarker measurements on Roche Modular E170 platform fulfills desirable goals derived from biological variation

Introduction

Monitoring of assay imprecision is a relevant quality indicator in clinical laboratories, as only assays reaching acceptable analytical performances should rather be appropriately used in clinical practice. To objectively define performance, imprecision goals (expressed as CV [coefficient of variation]) are usually derived from biological variation data. ¹ Particularly, the desirable goal (DG, %) for analytical variation should be less than or equal to one-half of the within-subject biological variation of the analyte (≤0.5 CV_I) and optimal goal (OG, %) should be ≤0.25 CV_I. ² Information on the assay CV can be obtained from several sources. Manufacturers of assay diagnostic kits report on the package insert information on the imprecision that is primarily based on within- and between-run analyses, often portraying better precision that in clinical practice. Some data on assay imprecision are available in the scientific literature, but may be biased due to study design. Factors potentially affecting imprecision include sample type or the number of performed determinations. Only results from daily use of assays, through the implementation of an appropriate Internal Quality Control (IQC) programme based on the determination of a commutable material having analyte concentrations close to the decision limit, can provide optimal estimation of the assay imprecision that may significantly differ from that obtained from previously mentioned sources. ^3,4 Using a fresh–frozen serum pool, we report here the evaluation of the imprecision of eight widely used and clinically relevant tumour biomarker measurements performed in our laboratory during 2008. Particularly, non-commutability of the control materials could be an issue, because immunoassays may perform differently with an artificial matrix (e.g. lyophilized samples) than with native patient samples. ^5,6

Materials and methods

Measurements were carried out on the Modular E170 platform (Roche Diagnostics, Basel, Switzerland) in accordance with the manufacturer's instructions. Estimation of CVs was obtained using an IQC program based on the analyte measurement every working day in a human frozen serum pool, prepared by mixing selected fresh leftover human samples to reach biomarker concentrations around the recommended clinical cut-offs. Testing a fresh–frozen human serum pool permits a realistic estimate of CVs, avoiding any possible influence by modified sample matrix, in which commercial control materials are usually prepared. After preparation, the pool was filtered and centrifuged to remove particulate debris and fibrin strands, and 0.4-mL aliquots were stored at −20°C until they were thawed for measurement. The same pool was used for the long-term study. For each biomarker, the monthly CV (n = 12) and the cumulative mean (average yearly CV), which together account for most potential sources of laboratory variation (i.e. calibrations, reagent lots, ambient temperature, personnel, changes in analyser software), were calculated and compared with corresponding DG and OG. ⁷ The CVs were calculated and evaluated on a monthly basis to monitor the systematic variation of the assays and enable detection of increased random error at a sufficiently early stage, as part of the IQC program of the laboratory. For CYFRA 21.1, due to the low number of runs (n = 21) performed within the year 2008, yearly CV was calculated using single IQC results from each run. Critical difference (CD), i.e. the difference that must be exceeded before a change in two consecutive test results, performed on the same system in an individual, is statistically significant, was also derived for the eight tumour markers using the equation: 2.77 (CV_I ² + CV_A ²)^1/2, where the multiplier 2.77 derives from the z(2)^1/2 formula, where z is the bidirectional z-score, indicating how many standard deviations an observation is above or below the mean, for the selected probability of 95.5%; ¹ the specific CV_I value was obtained from reference ⁷ and represents the within-subject biological variation; the CV_A was the respective yearly CV representing the analytical variation. As CVs were tested at only one concentration level, the corresponding CD was expressed in concentration value.

Results

As shown in Table 1, all assays had a mean yearly CV markedly lower than the DG, except for CA15.3 that showed a mean CV equal to the stringent DG (3.1%). Assay CVs for prostate-specific antigen (PSA), CA125, CA19.9 and CYFRA 21.1 were all within optimal limits.

Discussion

For tumour biomarkers reaching and maintaining DG over time is crucial because they are measured for disease monitoring using data produced by serial testing and evaluated by the corresponding CD, which depends on both analytical CV and CV_I. ⁸ In two sequential tumour marker concentrations, only values exceeding the CD could be considered as an informative change about clinical course of malignancy, i.e. response to therapy and progression or recurrence of disease. Our study shows that in routine practice condition, including any potential source of variation for laboratory measurements and over a clinically and analytically relevant time-span, the imprecision of the tumour biomarker measurements on Modular E170 analytical system fulfils DG derived from biological variation, minimally affecting total variation of results.

The excellent CVs of the tumour biomarker assays evaluated here confirm the analytical suitability of these methods for use in oncology. The influence of analytical variation on calculated CD values is small and likely to have minimal clinical impact.

DECLARATIONS

Competing interests: None.

Funding: No fund was received for the study.

Ethical approval: Not applicable.

Guarantor: AD.

Contributorship: AD and MP conceived the study. LS and RM were involved in protocol development. AD wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript.