Papers in Press. Published July 18, 2017 as doi: /clinchem

Transcription

1 Papers in Press. Published July 18, 2017 as doi: /clinchem The latest version is at Clinical Chemistry 63: (2017) Other Areas of Clinical Chemistry The EuBIVAS Project: Within- and Between-Subject Biological Variation Data for Serum Creatinine Using Enzymatic and Alkaline Picrate Methods and Implications for Monitoring Anna Carobene, 1,11* Irene Marino, 1 Abdurrahman Coşkun, 2,11 Mustafa Serteser, 2 Ibrahim Unsal, 2 Elena Guerra, 1 William A. Bartlett, 3,11 Sverre Sandberg, 4,5,11 Aasne Karine Aarsand, 4,11 Marit Sverresdotter Sylte, 4 Thomas Røraas, 5,11 Una Ørvim Sølvik, 6 Pilar Fernandez-Calle, 7,11 Jorge Díaz-Garzón, 7 Francesca Tosato, 8 Mario Plebani, 8 Niels Jonker, 9,11 Gerhard Barla, 9 and Ferruccio Ceriotti 10 on behalf of the European Biological Variation Study of the EFLM Working Group on Biological Variation BACKGROUND: The European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) European Biological Variation Study (EuBIVAS) has been established to deliver rigorously determined biological variation (BV) indices. EuBIVAS determined BV for serum creatinine using the enzymatic and alkaline picrate measurement methods. METHOD: In total, 91 healthy individuals (38 males, 53 females; age range, years) were bled for 10 consecutive weeks at 6 European laboratories. An equivalent protocol was followed at each center. Sera were stored at 80 C before analysis. Analyses for each patient were performed in duplicate within a single run on an ADVIA 2400 system (San Raffaele Hospital, Milan). The data were subjected to outlier and homogeneity analysis before performing CV-ANOVA to determine BV and analytical variation (CV A ) estimates with confidence intervals (CI). RESULTS: The within-subject BV estimates [CV I (95% CI)] were similar for enzymatic [4.4% ( )] and alkaline picrate [4.7% ( )] methods and lower than the estimate presently available online (CV I 5.9%). No significant male/female BV differences were found. Significant differences were observed in mean creatinine values between men and women and between Turkish individuals and those of other nationalities. Between-subject BV (CV G ) estimates, stratified accordingly, produced CV G values similar to historical BV data. CV A was 1.1% for the enzymatic and 4.4% for alkaline picrate methods, indicating that alkaline picrate methods fail to fulfill analytical performance specifications for imprecision (CV APS ). CONCLUSIONS: The serum creatinine CV I obtained by EuBIVAS specifies a more stringent CV APS than previously identified. The alkaline picrate method failed to meet this CV APS, raising questions regarding its future use American Association for Clinical Chemistry Serum creatinine measurement is one of the most requested tests in clinical laboratories (1). The measurement can be used for estimation of glomerular filtration rate (egfr), delivering an easily accessible and internationally accepted method for initial evaluation and monitoring of kidney function (2 3). Correct creatinine measurements are thus essential for appropriate classification of a patient as having chronic kidney disease (CKD) 12 of a particular stage; such an important clinical application underscores the requirement for an appropriate analytical performance specification (APS) for creat- 1 Servizio di Medicina di Laboratorio, Ospedale San Raffaele, Milan, Italy; 2 Acibadem University School of Medicine, Atasehir, Istanbul, Turkey; 3 Blood Sciences, Ninewells Hospital & Medical School, Scotland, UK; 4 Laboratory of Clinical Biochemistry, Haukeland University Hospital, Bergen, Norway; 5 Norwegian Quality Improvement of Laboratory Examinations (Noklus), Haraldsplass Deaconess Hospital, Bergen, Norway; 6 Department of Global Public Health and Primary Care, University of Bergen, Bergen, Norway; 7 Hospital Universitario La Paz, Madrid, Spain, and Quality Analytical Commission of Spanish Society of Clinical Chemistry (SEQC); 8 Department of Laboratory Medicine University Hospital, Padua, Italy; 9 Certe, Wilhelmina Ziekenhuis Assen, Europaweg-Zuid 1, 9401 RK Assen, the Netherlands; 10 Central Laboratory, Fondazione IRCCS Ca Granda, Ospedale Maggiore Policlinico, Milan, Italy; 11 Biological Variation Working Group, European Federation of Clinical Chemistry and Laboratory Medicine, com. * Address correspondence to this author at: San Raffaele Hospital, via Olgettina 60, 20132, Milan, Italy Fax ; carobene.anna@hsr.it Received April 18, 2017; accepted June 19, Previously published online at DOI: /clinchem American Association for Clinical Chemistry 12 Nonstandard abbreviations: CKD, chronic kidney disease, APS, analytical performance specification; B APS, analytical performance specification for bias; BV, biological variation; BVD, biological variation data; BV-DB, biological variation database on-line, updated 2014; BV- WG, EFLM working group on biological variation; CV A, analytical variation; CV APS, analytical performance specification for imprecision; CV I, within-subject biological variation; CV G, between-subjects biological variation; EFLM, European Federation of Clinical Chemistry and Laboratory Medicine; egfr, estimation of glomerular filtration rate; EuBIVAS, European Biological Variation Study; RCV, reference change values; SD A, analytical standard deviation. 1 Copyright (C) 2017 by The American Association for Clinical Chemistry

2 inine assays (4). The First Strategic Conference of the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) held in Milan in 2014 identified 3 models to define the APS: the effect on clinical outcome, the biological variation (BV), and the state of the art (5, 6). The EFLM Task and Finish Group, focusing on the relative utility of these models, proposed that for creatinine measurement, a clinical outcomes model should be applied (7). However, studies defining the relationship between clinical outcomes and APSs are scarce or nonexistent, and high-quality estimates of BV data are therefore still the best available approach for setting APS for creatinine at this time. An understanding of the significance of a change in sequential measurements of creatinine within subjects is required to adequately monitor disease progression. Using a so-called reference change value (RCV), BV data (BVD) can be used to evaluate changes in serial measurements of the same individual (8). Assessment of significant change is important for both monitoring the progression of CKD and identifying acute kidney injury. BVD can also be used to define the utility of the reference intervals using the index of individuality (9). Currently, summarized BV estimates for creatinine are available and accessible online from a database (BV-DB) hosted on the Westgard website (10), which was updated by the Spanish Society of Laboratory Medicine in 2014 (11). In recent years, concerns have been raised about the quality of the data available in published studies of BV. This it has led to questions being raised regarding the reliability of some of the headline BVD included in the BV-DB (12). Factors impacting on the reliability and hence on the applicability of the existing data to current clinical practice are manifold. For example, estimates of BV may be derived from publications employing oldergeneration analytical methods, with performance characteristics that are very different from current methods. In other cases, BVD are derived from studies with deficiencies in experimental design, where areas of concern include the preanalytical and analytical phases of the experimental protocols employed and the data analysis (12). Overall, many published studies have failed to address fundamental issues identified in the protocol proposed in the seminal paper on the subject published by Fraser and Harris (13) and have only partially met the requirements of the checklist recently published by the EFLM Working Group on BV (BV-WG) (14). To address some of the current concerns about the quality of existing BVD and to address the need to deliver new data using latest-generation analytical methods, the BV-WG has designed and initiated the European Biological Variation Study (EuBIVAS) (15) to obtain highquality BV indices for the highest possible number of measurands, including creatinine. In this setting, 2 approaches are available for creatinine measurement in routine laboratories. Historically, colorimetric assays based on alkaline picrate (Jaffe) have been the mainstay for creatinine measurement, but more recently enzymatic methods have become available. The enzymatic methods represent a generation of assays with improved analytical specificity and precision that are being increasingly used within clinical laboratories. This paper aims to use EuBIVAS creatinine data to (a) provide rigorous BV estimates for creatinine, (b) investigate the possible influence of analytical methodology used for creatinine measurement on BV estimates, (c) set APSs for creatinine assays, and (d) examine the potential impact of the creatinine method used when monitoring patients. Materials and Methods The protocol used to collect and to store data on the serum samples, the health status, and the inclusion/exclusion criteria of enrolled individuals is described in a recent paper (15). SAMPLE COLLECTION AND HANDLING Briefly, EuBIVAS involved 6 European laboratories that enrolled 91 healthy volunteers (38 males and 53 females; age range, years) [see Table 1 in the Data Supplement that accompanies the online version of this article at All involved laboratories followed the same protocol for the preanalytical phase (15). All volunteers compiled an enrolment questionnaire to verify their health status and to collect information regarding their lifestyle (15, 16). Further exclusion criteria were verified by some laboratory tests done during the first collection (15). For most eligible subjects, fasting blood samples were drawn for 10 consecutive weeks (April June 2015). Eleven participants completed only 9 collections, 3 participants completed 8 collections, and 2 participants completed only 7 collections. Moreover, from the information gathered using the short questionnaire completed by each participant at each phlebotomy visit and as a consequence of the biochemical analyses also performed at the time of each visit, 29 samples from 26 individuals were discarded (15). Serum samples from all the participants, were obtained by the participating laboratories. Following centrifugation at 3000g for 10 min at room temperature, the samples were aliquoted and sent frozen in dry ice to the San Raffaele Hospital in Milan and stored in a freezer at 80 C until analysis (15). The EuBIVAS protocol was approved by the Institutional Ethical Review Board of San Raffaele Hospital in agreement with the World Medical Association Declaration of Helsinki and by the Ethical Board/Regional Ethics Committee of each center. 2 Clinical Chemistry 63:9 (2017)

3 EuBIVAS: Biological Variation Data for Serum Creatinine Table 1. BV data of creatinine enzymatic and alkaline picrate methods. Number of subjects Total number of results Mean number of samples/ subjects Mean number of replicates/ subjects Creatinine, Mean value (95% CI), mg/dl Creatinine, Mean value (95% CI), μmol/l CV A (95% CI), % CV I (95% CI), % CV G (95% CI) % Online BV database (10) CV I % CV G % ENZYMATIC METHOD All subjects ( ) 70.7 ( ) 1.1 ( ) a 4.4 ( ) 17.1 ( ) Males ( ) 79.5 ( ) 4.2 ( ) 13.8 ( ) Females ( ) 64.5 ( ) 4.6 ( ) 13.4 ( ) All subjects (without ( ) 72.3 ( ) 14.8 ( ) Turkish subjects) Males (without Turkish ( ) 80.8 ( ) 12.8 ( ) people) Females (without ( ) 66.4 ( ) 10.2 ( ) Turkish people) ALKALINE PICRATE All subjects ( ) 65.6 ( ) 4.4 ( ) 4.7 ( ) 19.0 ( ) Males ( ) 75.0 ( ) 4.1 ( ) 16.5 ( ) Females ( ) 59.0 ( ) 5.0 ( ) 13.9 ( ) All subjects (without ( ) 67.1 ( ) 17.8 ( ) Turkish people) Males (without Turkish ( ) 75.9 ( ) 16.7 ( ) people) Females (without ( ) 60.6 ( ) 12.1 ( ) Turkish people) Note: Data are given for all subjects, for males, for females, and with and without Turkish subjects. a The CVI and CV G values in bold were used to calculate APS in Table 2. Clinical Chemistry 63:9 (2017) 3

4 ANALYTICAL METHODS Measurements were performed on an ADVIA 2400 Clinical Chemistry System (Siemens Healthineers) in San Raffaele Hospital, using Siemens reagents (Creatinine Enzymatic, code , and Creatinine Jaffe, code ), Siemens calibrator (Chemistry Calibrator ADVIA), and Bio-Rad control materials (Liquid Assayed Multiqual level 1, level 2). In the enzymatic method, creatinine is converted by creatinine deiminase to ammonia and N-methylhydantoin. The ammonia is subsequently assayed by use of oxoglutarate and glutamate dehydrogenase. The reduction in absorbance at 340 nm, caused by the oxidation of NADPH, is proportional to the ammonia concentration released by creatinine. The compensated Jaffe creatinine method is a kinetic reaction based on the colorimetric reaction of creatinine with picric acid. To compensate for nonspecific reactions caused by pseudo-creatinine chromogens, mainly due to proteins, the results are automatically corrected by subtracting 0.30 mg/dl (26.5 mol/l) (factor calculated by Siemens). All samples from the same individual were analyzed in duplicate within a single run. Bio-Rad control materials, used for internal quality control (IQC), were also measured in duplicate in each analytical run. Eleven runs were performed in total. DATA ANALYSIS Data analysis was performed using the CV-ANOVA, an ANOVA method, where data is first transformed using the CV-transformation (17). CV-ANOVA was adopted for the analysis as this straightforward, nonparametric procedure has been shown to be a robust, largely distribution-free procedure for estimating CV A and CV I in 3-level nested random models (17). Before analysis, outlier identification and removal were performed on the transformed data for replicates and samples to assure homogeneity. The homogeneity of analytical CV (CV A ) (between-replicates) was verified using the Bartlett test (18) and the homogeneity of within-subject biological variation (CV I ) was verified using the Cochran test (19). The steady-state situation was verified by a visual check for a systematic common change in creatinine concentration during the whole length of the study, whereas larger individual systematic changes were detected by the homogeneity test of CV I. To calculate the between-subject biological variation (CV G ) estimate, the data were natural-log transformed. The Shapiro Wilk test was used to verify the normality (20). The Dixon Reed criterion (21) was used to detect any outliers between subjects. A visual inspection of subjects values ordered by country was made to evaluate differences among subgroups from different countries (Figs. 1A and 2A). Any differences between female subgroups (pre- and postmenopause) and between males and females were first examined by visual inspection of the values ordered by sex and age (Figs. 1B and 2B). When doubt arose, CV- ANOVA of subgroups was performed. The data sets from males and females in each subgroup were also analyzed separately. The significance of the differences in the CV I and CV G between subgroups was determined by the overlapping of their 95% confidence intervals (CI). The differences between mean creatinine concentrations in the male and female subgroups were examined using the Student t-test. If the mean values were found to be significantly different, the lower value of the 2 CV G estimates was used to calculate APS. Where no significant difference between the female and male mean value was present, the CV G estimate based on results for all subjects, both female and male, was used. CV I estimates based on data from all individuals were always used to calculate APS. BV estimates were used to calculate APS for the analytical imprecision (CV APS ), analytical bias (B APS ), and total error (TE APS ), by applying the following formulas: CV APS 1 2 CV I ;B APS 0.25(CV I 2 CV G 2 ) 0.5 ; and TE APS CV APS (1.65 B APS ) (22). The RCV were estimated using the formulas below: 2 SD A,log 2 SD I,log log e (CV A 2 1) log e (CV I 2 1) 2 SD combined,log SD A,log 2 SD I,log RCV% 100% (exp( Z 2 SD combined,log ) 1) where Z used was 1.96 at the probability level of significant change set at 95%. Data analysis was performed using Excel 2010 and XLSTAT (statistical software for Excel) (23). Results and Discussion Details on gender, age, and body mass index (BMI) of the enrolled individuals per country are shown in Table 1 in the online Data Supplement. Three percent of participants were regular smokers; the alcohol assumption was low/moderate; and the participants, with the exception of those from the Turkish cohort, declared a rather high level of physical activity. The analysis revealed no outliers in the creatinine results obtained with the enzymatic method, but for the Jaffe method, 10 results from 6 different individuals were removed because of analytical heterogeneity and 1 sam- 4 Clinical Chemistry 63:9 (2017)

5 EuBIVAS: Biological Variation Data for Serum Creatinine Fig. 1. Biological variation of creatinine, enzymatic method. Median value (circle for Turkish individuals, hyphen for other countries) (minimum maximum) for each individual ordered by country (A). Range (minimum maximum) and median value (circle for Turkish individuals, hyphen for other countries) for each individual ordered by sex and age (B). Continuous lines point out the 5th and 95th percentiles and the median value (+ CI) for the female subgroup; dashed lines point out the 5th and 95th percentiles and the median value (+ CI) for the male subgroup (calculated excluding Turkish people). ple (2 replicates) was removed because of intraindividual heterogeneity. In total, for Jaffe, 12 results out of 1716 (0.7%) were removed. No individuals were recognized as outliers for either method. Creatinine concentrations were higher in males than in females for both methods (P 0.001) (Table 1, Figs. 1B and 2B). Based on visual inspection (Figs. 1A and 2A), a higher number of individuals in the Turkish co- Clinical Chemistry 63:9 (2017) 5

6 Fig. 2. Biological variation of creatinine, Jaffe method. Median (circles for Turkish individuals, hyphens for other countries) (minimum maximum) for each individual ordered by country. To convert creatinine concentrations from mg/dl to μmol/l, multiply by 88.4 (A). Range (minimum maximum) and median value (circles for Turkish individuals, hyphens for other countries) for each individual ordered by sex and age (B). Continuous lines point out 5th and 95th percentiles and the median value (+ CI) for the female subgroup; dashed lines point out the 5 th and 95 th percentiles and the median value (+ CI) for the male subgroup (calculated excluding Turkish people). 6 Clinical Chemistry 63:9 (2017)

7 EuBIVAS: Biological Variation Data for Serum Creatinine Table 2. Analytical performance specification (APS), for imprecision, bias, and total error derived from BVD reported in Table 1 for the 2 methods. APS derived from BVD in Table 1 Imprecision % Bias % Total error % RCV a 95% (95% CI) Method CV APS (95% CI) B APS (95% CI) TE APS (95% CI) Crea-E b 2.2 ( ) 2.8 ( ) 6.4 ( ) 12.7 ( ) Crea-J 2.4 ( ) 3.2 ( ) 7.1 ( ) 17.8 ( ) a Reference change value (RCV) calculated as a 2-tailed value at level of probability of significant change set at 0.95 RCV = [2.77 (CV 2 A +CV 2 I ) 1/2 ]. b Crea-E, creatinine, enzymatic method; Crea-J, creatinine, alkaline picrate method. hort had creatinine values lower than 0.68 mg/dl (60 mol/l) compared with individuals in other countries, and there was a significant difference between the mean values of the Turkish cohort and all other participants for females (P 0.005) but not for males (P 0.13). The results from the Turkish cohort were therefore excluded from the estimation of CV G. No significant time trend for creatinine concentration in each subgroup considered was found (see creatinine trend.xls the online Supplemental Data). There were no significant differences in the CV I estimates (calculated based on results for all participants) and CV G estimates (after exclusion of the Turkish cohort) obtained with the 2 different creatinine methods or when separated into male and female subgroups (Table 1). The CV A estimates calculated by ANOVA on the results of duplicate measurements of the Bio-Rad control materials were: for enzymatic method, 2.4% ( ) [level 1; mean, 0.83 mg/dl ( ) (73 mol/l; 73 74)] and 1.5% ( ) [level 2; mean, 1.87 mg/dl ( ) (165 mol/l; )]; and for Jaffe method, 6.1% ( ) [level 1; mean 0.68 mg/dl ( ) (60 mol/l; 59 62)] and 2.9% ( ) [level 2; mean 1.75 mg/dl ( ) (155 mol/l; )]. CV A calculated by CV-ANOVA on the results of duplicate measurements of samples was 1.1% and 4.4%, for the enzymatic and Jaffe methods, respectively. The EuBIVAS project enabled the production of BVD for creatinine using 2 different analytical principles, the alkaline picrate method and a newer-generation enzymatic method. The EuBIVAS protocol followed a stringent experimental design (15) powered to deliver estimates of BV with a high degree of confidence for a well-characterized multinational cohort of subjects. The study delivered data that enabled comparison of BVD obtained using contemporary analytical methods with data in the BV-DB (10), as well as an assessment of the impact of new-generation enzymatic methods as compared with older methods upon estimates of BVD. The latter is particularly important as these new methods are increasingly being used. Any changes in estimates of CV I will have impact on the interpretation of the significance of change in serial creatinine results within a patient and thus on the ability to identify disease progression at any point in time. COMPARISON BETWEEN EUBIVAS ESTIMATES OF BV AND THOSE PRESENTED IN BV-DB CV I estimates, respectively, 4.44% ( ) and 4.69% ( ) for the enzymatic and Jaffe methods, were lower than those presented in the BV-DB (5.95%) (11), but direct comparisons were difficult as the BV-DB does not provide estimates of uncertainty. The finding of lower EuBIVAS BV estimates of creatinine is likely multifactorial. The BV-DB data are presented without CI, making direct comparison difficult. The comparison is further confounded by the fact that the estimates displayed in the BV-DB are median values derived from an analysis of 28 published studies (10 11) of varying quality spanning a substantial historical time frame and consequently a range of analytical methods. Similarly, APS values obtained in our study, and consequently also RCV values, were lower than those presented in the BV-DB (Table 2) (11). The APSs for imprecision obtained for the enzymatic and Jaffe methods were 2.20% ( ) and 2.35% ( ), respectively, instead of the currently used 2.98%, and the APS values for the bias were 2.8% ( ) and 3.2% ( ) instead of the currently used 3.96%. The EuBIVAS took into account all criteria presented in the checklist recently published for production of BV studies (14). Furthermore, it followed a strict preanalytical and analytical protocol (15) and used a new method (CV-ANOVA) to calculate CV I, which was demonstrated to have advantages over the traditional approach of nested ANOVA (17). The strictness of the approach to the design and delivery of the EuBIVAS project, which has obtained results from a very large cohort of participants, enables delivery of BV estimates (CV I and CV G ) that may be considered more definitive than those previously published. Clinical Chemistry 63:9 (2017) 7

8 Graphical and statistical analyses indicated the need to stratify data on the basis of sex to produce valid estimates of CV G for clinical applications. It also proved necessary to exclude the Turkish population from the subanalysis, as they had significantly lower mean creatinine concentrations in comparison to other nationalities in the study. When comparing the CV G values before stratification and outlier analysis, these were similar to those presented in the BV-DB. The observation that the mean creatinine concentration was lower in Turkish participants is not surprising. This has been reported previously (24), being explained as resulting from differences in diet (low meat consumption) and different physical activity. This was also confirmed by the lifestyle information provided by the enrolled subjects in EuBIVAS (16). INFLUENCE OF THE METHOD OF CREATININE MEASUREMENT UPON ESTIMATES OF BV AND CV A AND IMPLICATIONS FOR MONITORING The analytical imprecision of the enzymatic method for creatinine measurement was significantly lower than that for the Jaffe method, and only the enzymatic method met the APS-I, defined as half the CV I (Table 2). Reagents for both methods were provided by the same manufacturer and represent the latest iteration of the technologies. It has been identified that clinical outcomes should be considered to be the most appropriate method for assessing APS for creatinine (7), but because studies to enable this approach are not available, the use of high-quality estimates of BVD may provide the best available approach for setting APS for creatinine at the present time. The APS values shown in Table 2 were obtained after exclusion of the Turkish subjects, and considering CV G values by not excluding the Turkish subjects, the bias and total error specifications would be affected, increasing their values respectively for the enzymatic and Jaffe methods B APS 3.5% ( ) and 3.7% ( ), respectively; TE APS 7.1% ( ) and 7.6% ( ), respectively. The greater imprecision of the Jaffe method has clinical disadvantages by delivering higher RCVs when combined with the slightly higher estimates of CV I in this study. Clinicians applying RCVs as a tool to monitor their patients will need information on the method by which the creatinine measurement is performed and its analytical variation. The CV and RCV values derived for the Siemens methods in this study are highly likely to be applicable to the results of other manufacturers methods and are therefore generalizable if the analytical specificity and, for RCV, the imprecision of those methods are comparable. There is published evidence to suggest that such data are transferable (25 27). The RCVs presented in Table 2 are derived from the formula using a Z value to obtain a 2-tailed estimate of significant change. These are used to assess the significance of bidirectional change between consecutive samples. However, in monitoring the progression of CKD, the clinical question that is of importance is whether an increase in consecutive creatinine results is significant. In that case, the RCV formula should be used with application of a Z value to calculate a 1-tailed change value. This then might best be described as the critical difference between consecutive results. From this study, a 1-tailed value for percentage unidirectional change in creatinine concentration was determined to be 10.6% ( ) for the enzymatic method and 15.0% ( ) for the Jaffe method. Fraser reported that if CV A and CV I are known, then the RCV formula can be rearranged to make Z as the unknown: Z change/[2 1/2 (CV A 2 CV I 2 ) 1/2 ] (8). This value for Z can then be converted to a 1-tailed probability, enabling a plot of percentage increase between consecutive results against probability to be constructed. Plots of this type demonstrate the impact of the method used upon the probability of change, as demonstrated in Fig. 3. When using the Jaffe method to monitor patients, a greater percentage change between consecutive creatinine results is required to signal a statistically significant change at any given probability. In practice, for a patient with a creatinine of 1.13 mg/dl (100 mol/l), an increase to 1.28 mg/dl (113 mol/l) could be explained by analytical and biological variation if the measurement was performed with the enzymatic method with the same analytical imprecision obtained in this study (1-sided z-score, 0.025). However, if the analysis was performed using the Jaffe method, an increase to 1.33 mg/dl (118 mol/l) could be explained by analytical and biological variation. This will imply challenges when patients are being monitored when moving between different healthcare institutions if each location uses different technologies (28). The introduction of egfr to identify and monitor CKD has been recommended and adopted internationally (2, 3). The equations demonstrate a nonlinear relationship between creatinine concentration and egfr. The confidence limits around the estimate of egfr in a patient and any single time point will be a function of the population variances inherent in the population from which the equation was derived and the biological variation of the individual patient s creatinine estimate and the CV A of the creatinine method. The nonlinear relationship between creatinine concentration and egfr also means that there is a nonlinear relationship between the probability of change between estimates of egfr and the starting egfr. That is to say, a decrease of 4 ml/min/ 1.73 m 2 in egfr will be less significant for a patient with an initial egfr of 90 ml/min/1.73 m 2 than one with an initial egfr of 30 ml/min/1.73 m 2. An estimate of RCV for egfr in terms of percent change may therefore 8 Clinical Chemistry 63:9 (2017)

9 EuBIVAS: Biological Variation Data for Serum Creatinine Fig. 3. Probability of a unidirectional change between 2 consecutive creatinine results obtained from the same analytical method. The figure illustrates the probability plots of percentage unidirectional change between consecutive creatinine results obtained by 1 of the 2 methods (enzymatic or Jaffe). The plots are derived from a simulation using the formula Z = change/[2 1/2 2 (CV A+ CV 2 I ) 1/2 ] as described in the text. The resultant Z value is converted to a 1-tailed probability, which is then plotted against the percentage change used to generate Z. The plot also enables one to identify the 1-tailed RCV for each of the methods at P = 0.05 (10.6% enzymatic, 15.0% Jaffe) and P = 0.01 (14.9% enzymatic, 21.2% Jaffe) where the respective curves cross the probability lines set at 0.95 and To convert creatinine concentrations in mg/dl to μmol/l, multiply by be more appropriate to use for monitoring CKD progression. The variability of egfr within an individual will be influenced by the biological variability in serum creatinine and the analytical performance of the assay used to deliver the measurand. The difference between the CV I estimates obtained by the 2 different creatinine methods was not significant. It is often claimed that the analytical method does not affect estimates of BV, but a few head-to-head comparisons exist for different-generation analytical principles, where differences in analytical specificity may deliver a challenge. The reduced specificity of the Jaffe method may have a greater impact on variability between subjects at lower creatinine concentrations than on that between subjects with significantly higher creatinine concentrations (25 28). The impact of the effect in CKD in which creatinine concentrations are much higher may be less problematic in clinical practice. It is interesting that in this study, the differences observed in mean creatinine concentrations between the Turkish population and the other nationalities measured using the enzymatic method were not present in the Jaffe data set. The data for both the Jaffe and enzymatic methods were stratified to exclude the Turkish population to identify the magnitude of the effect. STUDY LIMITATIONS The analyses were performed using only 1 manufacturer s Jaffe and enzymatic reagents. Reagents from different manufacturers may perform differently, especially regarding the effect of interfering substances (29), but it is unlikely that this will affect the estimates of biological variation. Although all the analyses of the same subject were performed in the same run and the within-run analytical variability was statistically eliminated, different individuals were measured on different days. However, the IQC was stable during this period, and so the effect on the between-subject variation is probably negligible. Conclusions In this study, estimates of CV I and CV G for creatinine were obtained using sera from a multinational cohort consisting of healthy individuals from different Euro- Clinical Chemistry 63:9 (2017) 9

10 pean countries. The CV I estimates obtained from the creatinine data sets from individuals from 5 different countries (Italy, Norway, Spain, the Netherlands, and Turkey) were homogeneous. The Turkish subgroup did demonstrate a nationality-specific mean for creatinine as measured by the enzymatic method. The 2 different methods, enzymatic and Jaffe, gave similar BV estimates. These data confirm that the BV estimates obtained herein are widely applicable and that these may be used to determine APS at an international level in the absence of suitable clinical outcome studies. The Jaffe and enzymatic methods delivered different RCVs in this study. Such differences may confound clinical interpretation of serial results in practice if consecutive results are delivered from different analytical systems, and these differences will affect the confidence limits surrounding estimates of egfr. References Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article. Authors Disclosures or Potential Conflicts of Interest: Upon manuscript submission, all authors completed the author disclosure form. Disclosures and/or potential conflicts of interest: Employment or Leadership: None declared. Consultant or Advisory Role: None declared. Stock Ownership: None declared. Honoraria: None declared. Research Funding: Siemens Healthineers donated all reagents, control materials and calibrators used for the creatinine measurements. Expert Testimony: None declared. Patents: None declared. Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, and final approval of manuscript. Acknowledgment: We would like to thank Siemens Healthineers, who allowed us to work on a dedicated Siemens ADVIA 2400 and donated all reagents, control materials, and calibrators used for the creatinine measurements. We would also like to thank all the participants for their blood donations. 1. National Kidney Foundation. K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Kidney Disease Outcome Quality Initiative. Am J Kidney Dis 2002;39:S1 S Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Ann Intern Med 1999;130: Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, Feldman HI, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150: Panteghini M. Estimating glomerular filtration rate from serum creatinine measurements: analytical issues and standardization programs. RIMeL/IJLaM 2007;3 Suppl: Sandberg S, Fraser FG, Horvath AR, Jansen R, Jones G, Oosterhuis W, et al. Defining analytical performance specifications: consensus statement from the 1st strategic conference of the European Federation of Clinical Chemistry and Laboratory Medicine. Clin Chem Lab Med 2015;53: Panteghini M, Sandberg S. Defining analytical performance specifications 15 years after the Stockholm conference. Clin Chem Lab Med 2015;53: Ceriotti F, Fernandez-Calle P, Klee GG, Nordin G, Sandberg S, Streichert T, et al. Criteria for assigning laboratory measurands to models for analytical performance specifications defined in the 1st EFLM strategic conference. Clin Chem Lab Med 2017;55: Fraser CG. Reference change values: the way forward in monitoring. Ann Clin Biochem 2009;46: Fraser CG, Hyltoft Petersen P. Analytical performance characteristics should be judges against objective quality specifications. Clin Chem 1999;45: Minchinela J, Ricós C, Perich C, Fernández-Calle P, Alvarez V, Domenech M, et al. Biological variation database and quality specifications for imprecision, bias and total error (desirable and minimum). The 2014 update. update.htm (Accessed November 23, 2016). 11. Perich C, Minchinela J, Ricos C, Fernández-Calle P, Doménech MV, Simóne M, et al. Biological variation database: structure and criteria used for generation and update. Clin Chem Lab Med 2015;53: Carobene A. Reliability of biological variation data available in an online database: need for improvement. Clin Chem Lab Med 2015;53: Fraser CG, Harris EK. Generation and application of data on biological variation in clinical chemistry. Crit Rev Clin Lab Sci 1989;27: Bartlett WA, Braga F, Carobene A, Coşkun A, Prusa R, Fernandez-Calle P, et al. A checklist for critical appraisal of studies of biological variation. Clin Chem Lab Med 2015;53: Carobene A, Strollo M, Jonker N, Barla G, Bartlett WA, Sandberg S, et al. Biological Variation Working Group, European Federation of Clinical Chemistry and Laboratory Medicine. Sample collections from healthy volunteers for biological variation estimates update: a new project undertaken by the Working Group on Biological Variation established by the European Federation of Clinical Chemistry and Laboratory Medicine. Clin Chem Lab Med 2016;54: Carobene A, Røraas T, Soelvik U, Sylte MS, Sandberg S, Guerra E, et al. Biological variation estimates obtained from 91 healthy subjects for nine enzymes in serum. On behalf of the European Biological Variation study (EuBIVAS) of the EFLM Working Group on Biological Variation. Clin Chem 2017;63: Røraas T, Støve B, Petersen PH, Sandberg S. Biological variation: the effect of different distributions on estimated within-person variation and reference change values. Clin Chem 2016;62: SnedecorGW, CochranWG. Statisticalmethods. 8thEd. Iowa City (IA): Iowa State University Press; Cochran WG. The distribution of the largest of a set of estimated variances as a fraction of their total. Ann Hum Genet 1941;11: Shapiro SS, Wilk MB. An analysis of variance test for normality (complete samples). Biometrika 1965;52: Dixon WJ. Processing data for outliers. Biometrics 1953;9: Fraser CG, Petersen PH. Quality goals in external quality assessment are best based on biology. Scand J Clin Lab Invest Suppl 1993;212: XLSTAT: Statistical Data Analysis Software (Accessed September 1, 2016). 24. Ozarda Y, Ichihara K, Aslan D, Aybek H, Ari Z, Taneli F et al. A multicenter nationwide reference intervals study for common biochemical analytes in Turkey using Abbott analyzers. Clin Chem Lab Med 2014;52: Boutten A, Bargnoux AS, Carlier MC, Delanaye P, Rozet E, Delatour V, et al. Enzymatic but not compensated Jaffe methods reach the desirable specifications of NK- DEP at normal levels of creatinine. Results of the French multicentric evaluation. Clin Chim Acta 2013;419: Jassam N, Weikamp C, Thomas A, Secchiero S, Sciacovelli L, Plebani M, et al. Post-standardization of routine creatinine assays: are they suitable for clinical applications. Ann Clin Biochem 2017;54: Carobene A, Ceriotti F, Infusino I, Frusciante E, Panteghini M. Evaluation of the impact of standardization process on the quality of serum creatinine determination in Italian laboratories. Clin Chim Acta 2014; 427: Panteghini M. On behalf of the IFCC scientific division. Enzymatic assays for creatinine: time for action. Clin Chem Lab Med 2008;46: Myers GL, Miller WG, Coresh J, Fleming J, Greenberg N, Greene T, et al. Recommendations for improving serum creatinine measurement: a report from the laboratory working group of the National Kidney Disease Education Program. Clin Chem 2006;52: Clinical Chemistry 63:9 (2017)