Comprehensive method comparisons: getting more from the data

It is an established procedure in laboratories to verify measurement methods by comparison of results from patient samples (split sample). The information retrieved and evaluated from a comparison is often limited to a scatter plot, a regression analysis and a differential plot. However, many more features and performance criteria are available from the same data. Access to a comprehensive set of information can also be used to harmonize results by recalibration and estimate any risk in applying alternate methods or surrogates to reference methods. Common principles for comparison of results are explained. Well known and new relations between commonly used methods and their limitations are discussed. A new application of the empirical folded density functions is introduced, and the use of a simplified error grid in the evaluation of a comparison is explored. Parametric and nonparametric pairwise significance tests and a convenient option of partitioning and limiting the dataset, ordinary linear and Deming regression functions, including uncertainties of the constants are provided in an Excel spreadsheet that is submitted as a downloadable supplemental file. The information harboured in a split sample comparison of methods is explored in detail. Procedures cover the main content of CLSI recommendations EP9, EP21 and EP27. Several additional practical procedures are described in a submitted single spreadsheet program.

[1]  T Pettersson,et al.  A method to estimate the uncertainty of measurements in a conglomerate of instruments/laboratories , 2005, Scandinavian journal of clinical and laboratory investigation.

[2]  R. B. Payne.  Calculation of slope by Deming's method from least-squares regression coefficients and imprecision. , 1984, Clinical chemistry.

[3]  Paul De Bièvre.  Does understanding metrological traceability in analytical measurement require going “back to basics”? , 2014, Accreditation and Quality Assurance.

[4]  J S Krouwer,et al.  A simple, graphical method to evaluate laboratory assays. , 1995, European journal of clinical chemistry and clinical biochemistry : journal of the Forum of European Clinical Chemistry Societies.

[5]  D. Cox,et al.  Evaluating Clinical Accuracy of Systems for Self-Monitoring of Blood Glucose , 1987, Diabetes Care.

[6]  J. Krouwer.  Why Bland–Altman plots should use X, not (Y+X)/2 when X is a reference method , 2008, Statistics in medicine.

[7]  Katherine L. Monti,et al.  Folded Empirical Distribution Function Curves—Mountain Plots , 1995 .

[8]  Werner Wosniok,et al.  Comparison of ordinary linear regression, orthogonal regression, standardized principal component analysis, Deming and Passing-Bablok approach for method validation in laboratory medicine , 2013 .

[9]  B H Ginsberg,et al.  A new consensus error grid to evaluate the clinical significance of inaccuracies in the measurement of blood glucose. , 2000, Diabetes care.

[10]  C A Parvin,et al.  A direct comparison of two slope-estimation techniques used in method-comparison studies. , 1984, Clinical chemistry.

[11]  W. Bablok,et al.  A New Biometrical Procedure for Testing the Equality of Measurements from Two Different Analytical Methods. Application of linear regression procedures for method comparison studies in Clinical Chemistry, Part I , 1983, Journal of clinical chemistry and clinical biochemistry. Zeitschrift fur klinische Chemie und klinische Biochemie.

[12]  J M Bland,et al.  Statistical methods for assessing agreement between two methods of clinical measurement , 1986 .

[13]  S Eksborg,et al.  Evaluation of method-comparison data. , 1981, Clinical chemistry.

[14]  D. Altman,et al.  STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT , 1986, The Lancet.