Clin Chem Lab Med 2017; 55(10): 1564–1573

Open Access

Matthijs Oyaerta,*, An Boel, Julie Jacobs, Stefanie Van den Bremt, Maxime De Sloovere, Hilde Vanpoucke and Lieve Van Hoovels Analytical performance and diagnostic accuracy of six different faecal assays in inflammatory bowel disease

DOI 10.1515/cclm-2016-1012 cut-off at a fixed specificity of 75% ranged from 95.2% to Received November 4, 2016; accepted December 29, 2016; previously 100%. Introduction of multiple result intervals increased published online February 21, 2017 the clinical interpretation of all the assays. Abstract Conclusions: Analytical and diagnostic performance of the evaluated faecal calprotectin assays is good, but numeri- Background: We evaluated the analytical performance of cal values differ substantially between the assays necessi- six different faecal calprotectin immunoassays together tating the use of different clinical cut-offs. Introduction of with their diagnostic accuracy in the discrimination multiple result intervals aids in clinical decision-making. between functional and organic bowel disorders. Keywords: faecal calprotectin; immunoassays; inflamma- Methods: The faecal samples were obtained from inflam- tory bowel disease. matory bowel disease patients (n = 27) at the time of diag- nosis [Crohn’s disease (n = 15), ulcerosa (n = 12)], gastroenterologic disease control patients (n = 52) and rheumatologic disease control patients (n = 26). All indi- Introduction viduals included in the study underwent a concurrent ileocolonoscopy. Analytical performance (imprecision, Inflammatory bowel diseases (IBD) are chronic diseases accuracy, carry-over, correlation and agreement) and that result from the of the intestinal wall and diagnostic accuracy (sensitivity, specificity, likelihood include Crohn’s disease (CD) and (UC). In ratios) of the different assays were evaluated. IBD, activated leukocytes infiltrate the mucosa and appear Results: All methods demonstrated good analytical per- in faeces due to shedding in the intestinal lumen [1, 2]. formance, but within-run and total imprecision varied According to the World Gastroenterology Organization, depending on the assay methodology used. Using Passing the diagnosis of IBD relies on a combination of physical Bablok and Bland-Altman analyses, low quantitative examination, patient history as well as a number of diag- agreement was observed between the assays. All assays nostic tests including laboratory analyses, stool examina- showed excellent diagnostic accuracy, with areas under tion, endoscopy, biopsy and imaging studies [3]. the receiver operating characteristic curves (ROC) ranging Among the various (non-invasive) biomarkers that have from 0.974 to 0.998. The AUCs were not significantly differ- been proposed over the past few years, faecal calprotectin ent between assays (p > 0.05). Diagnostic sensitivity at the has gained an important role [4, 5]. Calprotectin is a calcium binding protein present in the cytoplasm, repre- senting over 60% of cytosolic proteins. Consequently, faecal aPresent address: Department of Laboratory Medicine, Ghent calprotectin concentration may be related to inflammation University Hospitals, Gent, Belgium. *Corresponding author: Matthijs Oyaert, Department of Laboratory of the bowel mucosa in IBD. Despite some analytical and Medicine, OLV Hospital Aalst, Moorselbaan 164, 9300 Aalst, pre-analytical drawbacks [6–10], encouraging data have Belgium, Phone: + 32 (0)53/72 47 91, Fax: + 32 (0)53/72 45 88, recently been published about the clinical efficacy of this E-mail: [email protected] biomarker for monitoring disease activity, response to treat- An Boel, Julie Jacobs, Stefanie Van den Bremt and Lieve Van ment and relapse [11, 12]. In addition, faecal calprotectin is Hoovels: Department of Laboratory Medicine, Onze-Lieve-Vrouw Hospital, Aalst, Belgium resistant to bacterial degradation, being stable in stool for Maxime De Sloovere and Hilde Vanpoucke: Department of up to 3 days at room temperature, which adds more benefits Laboratory Medicine, AZ Delta Hospital, Roeselare, Belgium to its use as a laboratory marker [13].

©2017, Matthijs Oyaert et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License. Oyaert et al.: Evaluation of faecal calprotectin assays 1565

Over the past years, different assays for the detec- tion of faecal calprotectin have been introduced includ- ing point-of-care methods [7, 14–21]. These methods have the advantage of being easy to use and allowing analy- sis as soon as the sample is received in the laboratory. 23.0 (13.0–41.9) 99.4 (44.1–364.2) In addition, results are rapidly available, which allows 141.7 (21.9–237.0) Orgentec Calprotectin Orgentec 2539.0 (890.2–5174.3) quick clinical decision making. However, as laboratories 1835.2 (1168.5–7275.0)

are being consolidated, more automated assays such as enzyme-linked immunoassays (ELISA) [7, 22, 23], chemi- luminescence immunoassays (CLIA), fluoro enzyme Euroimmun Euroimmun - the gastro from patients Including Calprotectin immunoassays (FEIA) and particle enhanced turbidimet- e ric immunoassays (PETIA) for the measurement of faecal 15.6 (3.3–42.5) 78.0 (33.2–218.0) calprotectin have been introduced [10]. 137.2 (14.6–205.1) 985.5 (542.6–1829.6) 1555.9 (659.5–1985.3) a

The aim of this study was to evaluate six different automated faecal calprotectin immunoassays for the Including oesophagitis, erosive gastritis, gastric ulcers, ulcers, gastric gastritis, erosive oesophagitis, Including diagnosis of IBD using the results of ileocolonoscopy as c the criterion standard. Analytical and diagnostic perfor- mances were compared for the different tests. < 20 ( 20–29.9) 71.9 (16.2–83.8) 105.9 ( < 20–222.8) Bühlmann fCAL Turbo fCAL Bühlmann 799.1 (483.7–1491.6) 1058.0 (708.1–1844.1)

Materials and methods

Patient population 6.1 ( < 5–17.3) 60.8 (10.9–86.8) 36.6 (15.9–57.7)

The total population consisted of 105 patients (median age, 35 years; arthritis. (undifferentiated) spondylo-arthritis, Including d Diasorin Calprotectin Diasorin 278.0 (179.5–1520.0) range, 14–94 years; 44 males, 61 females) with suspicion of IBD for 560.5 (320.8–1025.3)

whom a stool sample for faecal calprotectin analysis was sent to the ® clinical laboratory. Of those 105 patients, 86 consecutive patients were recruited over a 6-month period at the OLV Hospital Aalst. Based on the endoscopic results (see further), these patients were categorised into 21 IBD patients, 39 gastroenterologic disease control < 16.1 ( 16.1) patients (GCG) and 26 rheumatologic disease control patients (RCG). 61.6 ( < 16.1–140.6) 37.9 ( < 16.1–109.0)

To enhance the diagnostic power of the likelihood calculations, 19 Flash QUANTA Inova 451.4 (291.7–2588.7) 619.5 (396.7–1136.4)

diagnostic stool samples from AZ Delta Roeselare were added from six IBD patients and 13 gastroenterologic disease control patients. A summary of the patient characteristics with the faecal calprotectin concentrations for the different assays is presented in Table 1. The main reasons for performing faecal calprotectin deter- mination were diarrhoea, mucous or bloody stools, weight loss 14.5 (5.4–30.3) EliA Calprotectin 2 EliA Calprotectin 64.0 (24.0–114.0) and abdominal pain and cramping. Diagnostic work-up included 98.0 (19.5–206.5) physical examination and case history, endoscopic and histologic 1020.0 (460.5–3168.0) 1167.0 (641.0–2734.0)

analysis, radiologic work-up and laboratory tests. To allow accu- b Age Age rate diagnosis, all included patients underwent ileocolonoscopy. ages. maximum and minimum indicate in brackets Values years. in median as presented is Age b

Patients who had previously been diagnosed with IBD or had not (range) 40 (17–59 29 (14–67) 53 (26–68) 36 (14–62) received ileocolonoscopy were excluded. Other exclusion criteria 35 (20–76)

included unclear diagnosis (e.g. indeterminate colitis), inability 7/5 4/11 10/5 to collect enough faecal sample and age younger than 14 years. Men/ 13/27 10/13 women

Senior gastroenterologists performed all endoscopies and findings n 40 23 15 15 were documented in a computer-based database. The final diagno- 12

sis of IBD (i.e. CD and UC) was independently made by a patholo-

gist or gastroenterologist who was blinded for calprotectin results. d Faecal calprotectin was determined on a single stool sample c in all patients. Samples were transported to the laboratory at room e temperature and stored at – 20 °C prior to extraction. For all assays, Data are presented as median (IQR). median as presented are Data logic and rheumatologic disease control group with a final diagnosis of IBS. IBS, . bowel irritable IBS, IBS. of diagnosis a final with group control disease rheumatologic and logic diverticulitis, microscopic colitis, , hyperplastic polyps, adenomatous polyps. polyps. adenomatous polyps, hyperplastic cancer, colorectal colitis, microscopic diverticulitis, a Final Diagnosis Final g/g) obtained with the different methods evaluated. methods the different with μ g/g) obtained ( concentrations calprotectin faecal and characteristics the patient of 1: Overview Table IBD disease Crohn’s IBS Rheumato Colitis ulcerosa Colitis all analyses were performed within the same freeze-thaw cycle of Non-IBD Gastro 1566 Oyaert et al.: Evaluation of faecal calprotectin assays the faeces sample, except for the Inova assay, for which the samples 24 64 96 100 200 underwent one additional freeze-thaw cycle [17]. 100 The study was performed with full respect for individuals’ rights Number of of Number

to confidentiality and in accordance with procedures supervised by per kit tests

Local Authorities responsible for Ethical Research (Belgian registra- tion number of ethical approval B126201526847). ) 2 Origin not reported. reported. not Origin c Linearity Linearity (R 0.996 NR NR 0.990 NR 0.980

Measurement of faecal calprotectin 5.0 5.2 3.8 6.5 LOQ 20.0 16.1

Faecal calprotectin extraction: To prevent pre-analytical variation b b [6, 10], all samples were weighted and extracted with the Smart μ g/g Prep Extraction Device (Roche Diagnostics, Mannheim, Germany) in

accordance with the manufacturer’s instructions but using the dif- Measuring range, range, 5.2–1000.0 3.8–6000.0 6.5–2100.0 5.0–8000.0 16.1–3500.0

ferent company specific faecal calprotectin buffers. Extractions were 20.0–8000.0

performed in different batches, immediately followed by calprotectin 50 50 50 50 50 analyses. 50 μ g/g

Faecal calprotectin tests: Measurement of faecal calprotectin was Proposed performed using six automated immunoassays: Thermo Fisher cut-off,

EliA Calprotectin 2 assay on the Phadia 250 (Thermo Fisher Scien- tific, Uppsala, Sweden); Diasorin Calprotectin assay on the Liaison (Diasorin S.P.A., Saluggia, Italy), Inova QUANTA Flash Calprotec- Measuring Measuring principle ELISA CLIA ELISA FEIA PETIA tin (research use only) on the Inova BIO-FLASH instrument (Inova CLIA

Diagnostics, San Diego, CA, USA); Bühlmann fCAL Turbo (Bühlmann Laboratories AG, Schönenbuch, Switzerland) on the Roche Cobas c501 (Roche Diagnostics, Mannheim, Germany), Euroimmun Calpro- tectin assay (Euroimmun; Lübeck, Germany), on an automated ELISA No. of of No. calibrators NR 3 6 6 6 2

instrument (QUANTA-Lyser 2; Inova) and Orgentec Calprotectin assay a on the Alegria (Orgentec Diagnostika, Mainz, Germany). All assays 1 : 50 1 : 50 1 : 50 1 : 75 1 : 50 were performed following manufacturer’s instructions. An overview 1 : 50 of the different methods is presented in Table 2. dilution

Comparison of immunoassays extraction Faecal

Analytical performance: The imprecision was assessed by dupli- cate testing of at least five aliquots of extracted patient sample and internal quality control samples (one sample with a high calprotec- c tin value and one with a low calprotectin value), according to the CLSI EP5-A2 guideline [24]. To determine the amount of carry-over, a sample with a high calprotectin concentration and one with a low Monoclonal Monoclonal, native Monoclonal, Monoclonal, native Monoclonal, Detection mouse Monoclonal, Polyclonal avian Polyclonal concentration (i.e. blank extraction buffer) were analysed in the recombinant Monoclonal,

sequence HHBBB, where H is the sample with the high concentration

and B is the blank [25]. Accuracy was evaluated from eQC samples be obtained. can calprotectin μ g/g faecal 8000.0 to up concentrations dilution, 1 : 4 (Bühlmann) and 1 : 10 (Diasorin) After b (n = 4, 2 positive, 2 negative) obtained from the INSTAND® external quality control program for faecal makers. For analytical method comparison, Bland-Altman plots, Passing-Bablok regression analysis and Pearson correlation coefficients were determined for all assays. c

Diagnostic performance: Based on the results of the consecu- Polyclonal, native Polyclonal, Monoclonal, native Monoclonal, Capture native Monoclonal, NA tive sample collection, receiver operating characteristics (ROC) recombinant Monoclonal, Polyclonal

analyses were carried out to determine the diagnostic performance

of each test. Deduced from the results of the ROC-analyses, cut- ® off values with the highest sensitivity and specificity have been calculated for diagnosing IBD. In addition, using the results of the whole patient sample population (n = 105), the calculation of positive likelihood ratio’s (LR) for IBD was performed for four result intervals for each assay ( < upper limit of normal (ULN), Using the Roche Smart Prep Extraction Device. Device. Extraction Prep Smart the Roche Using Manufacturer a Faecal calprotectin assay characteristics. assay calprotectin 2: Faecal Table FEIA, fluoro-enzyme immunoassay; CLIA, chemiluminescence immunoassay; PETIA, particle enhanced turbidimetric immunoassay; ELISA, enzyme-linked immunosorbent assay; LOQ, limit of of limit LOQ, assay; immunosorbent enzyme-linked ELISA, immunoassay; turbidimetric PETIA, particle enhanced immunoassay; CLIA, chemiluminescence immunoassay; FEIA, fluoro-enzyme reported. NR, not NA,applicable; not quantification; Inova Quanta Flash Quanta Inova Calprotectin Euroimmun EliA Calprotectin 2 EliA Calprotectin Bühlmann fCAL Turbo fCAL Bühlmann 1–3 × ULN, 3–10 × ULN and > 10 × ULN). Comparisons were made Calprotectin Diasorin Calprotectin Orgentec Oyaert et al.: Evaluation of faecal calprotectin assays 1567 using non-parametric tests (Kruskall-Wallis for unpaired samples). method were significantly lower (p < 0.05) compared A p-value < 0.05 was considered statistically significant. All statis- to the concentrations obtained with the Euroimmun, tical analyses were performed using MEDCALC® (software version Orgentec and Phadia assays. The highest values were 15.6.1, Ostend, ­Belgium). measured with the Orgentec assay, which were found significantly higher (p < 0.05) compared to the other assays (Figure 1). Results For the total population (i.e. disease + disease control group), significantly lower (p < 0.05) concentrations were obtained with the Diasorin assay compared to the Euro- Analytical performance immun, fCAL Turbo, Orgentec and Phadia assay. The Orgentec assay measured significantly higher (p < 0.05) Imprecision, carry-over and accuracy compared to the Diasorin and Inova assay. For the other assays, no significant differences in faecal calprotectin Within-run imprecision varied between 0.6% and 19.7%. concentration were found. Total imprecision results ranged from 1.5% to 23.3% Using Pearson correlation, we assessed linear agree- (Table 3). The lowest imprecision was obtained with the ment among the different assays. The correlation coef- CLIA methods and the highest with the ELISA assays. ficients (r) for all possible correlations are presented No significant carry-over was detected (< 0.03% for all in Table 4. The lowest correlation coefficients were methods). The results of the eQC-samples were qualita- observed between the Euroimmun assay and the Inova, tively correctly interpreted (i.e. positive or negative) by Orgentec, fCAL Turbo and Diasorin assays (ranging from all the assays using the manufacturer’s cut-offs (Table 2). 0.60 to 0.75). For all other assays, we observed a very However, important differences in the numerical values good correlation (mean r, 0.89; median, 0.90; range, were obtained with the different assays. 0.81–0.95) (Table 4). A high correlation does not mean that two methods have a good agreement. We there- fore performed Passing-Bablok regression and Bland- Method comparison Altman analyses (Table 4). These analyses showed low agreement between assays. Additionally, Bland-Altman The range of the faecal calprotectin concentrations and analysis, which was performed by plotting the differ- median faecal calprotectin concentrations in diagnostic ence between observations against the mean, revealed IBD samples varied substantially between the measure- that the difference between assays increased for increas- ment methods (Table 1). Quantitative data distribution ing means (heteroscedasticity). The best agreement was showed that calprotectin concentrations in diagnostic observed between the results obtained with the Inova samples of IBD patients measured with the Diasorin and Diasorin methods.

Table 3: Intra-run and total imprecision for faecal calprotectin determinations with the tested assays.

Intra-run imprecisiona Total imprecisiona

Kit-controlsb Patient samplesc Kit-controlsb Patient samplesc

Low High Low High Low High Low High

EliA Calprotectin 2 9.1 8.2 5.7 8.0 10.0 9.1 5.7 8.0 Diasorin Calprotectin 3.3 4.3 2.5 2.1 3.3 7.2 4.7 2.8 Inova QUANTA Flash® 3.6 0.9 1.7 0.6 3.7 1.9 1.7 3.7 Bühlmann fCAL Turbo 5.6 1.4 4.1 2.5 5.6 1.5 8.4 2.5 Orgentec Calprotectind – – 13.2 9.7 – – 13.2 9.7 Euroimmun Calprotectin 10.6 19.7 16.2 7.7 10.6 23.3 16.2 18.1 aThe results are presented as %CV (coefficient of variation). bLow and high kit-controls ranged from 20 to 60 μg/g and 130–270 μg/g. Kit- controls for the EliA and Inova Quanta Flash assays were analysed 10 times in duplo, for the Diasorin, fCAL Turbo and Euroimmun assays, five in duplo determinations were performed. cLow and high patient-sample controls ranged from 15 to 60 μg/g and 100 to 450 μg/g, ­respectively. The patient-sample controls were measured five times in duplo for all assays. dAs internal kit-controls are reported as positive/ negative, no data for intra-run and total imprecision for kit-controls are available for this assay. 1568 Oyaert et al.: Evaluation of faecal calprotectin assays

10,000

8000

6000

4000

Faecal calprotectin concentration, µ g/g 2000

0

Bühlmann fCAL Turbo Diasorin Calprotectin EliA Calprotectin 2 Euroimmun Calprotectin Inova QUANTA Flash Orgentec Calprotectin

Figure 1: Box-Whisker plots representing faecal calprotectin concentrations among the different assays for the IBD populations. The values falling outside the Box-Whisker plot are outliers.

Diagnostic performance of the faecal was used to calculate post-test probabilities as a function ­calprotectin assays of pre-test probability. By defining result intervals, IBD was ruled out in 51 Among the study population, the prevalence of IBD was (64.6%), 60 (75.9%), 48 (67.6%), 48 (60.8%), 45 (57.0%) 25.7% (27/105). Sensitivities and specificities were calcu- and 45 (57.0%) of the non-IBD patients with the Phadia, lated for each faecal calprotectin method (Table 5). The Diasorin, Inova, Bühlmann, Euroimmun and Orgentec areas under the curves (AUC) were 0.974 (0.915–0.996), assay (< 50 μg/g), respectively. Furthermore, calprotectin 0.993 (0.946–1.000), 0.978 (0.916–0.998), 0.988 (0.936– values between 1 and 3 times and 3 and 10 times the ULN 1.000), 0.991 (0.942–1.000), and 0.998 (0.954–1.000) for have no clinical usefulness and cannot be used to exclude the Phadia, Diasorin, Inova, Bühlmann, Euroimmun and or to confirm the diagnosis of IBD, except for the 3–10 Orgentec assay, respectively. The AUCs were not statisti- times ULN for the Diasorin assay, where the diagnosis of cally significantly different between the assays (p > 0.05). IBD was confirmed in seven of the IBD patients (25.9%) At the cut-off value proposed by the manufacturer (i.e. [LR = 10.1 (95% CI 2.2–45.7)]. The diagnosis was confirmed 50 μg/g for all assays), sensitivity for all assays was 100%, in 19 (70.4%), 13 (48.1%), 14 (56.0%), 20 (74.1%), 21 (77.8%) but the specificity was much lower (Table 5). No signifi- and 26 (96.3%) of the IBD patients when using a cut-off of cant differences in specificity were obtained between the > 500 μg/g faecal calprotectin with the Phadia, Diasorin, assays, except for the Phadia and Inova assay (p < 0.05). Inova, Bühlmann, Euroimmun and Orgentec assays, Based on the ROC curve analysis, cut-offs at fixed specific- respectively. ity of 75% and 95% differed between the assays, ranging from 41.3 to 142.9 μg/g and 110.0 to 411.9 μg/g faecal cal- protectin, respectively (Table 5). Sensitivities at these cut-off levels ranged from 95.2% to 100% and 84.2% to Discussion 100%, but no significant differences in sensitivity were obtained between the assays. In this study, we evaluated and compared six current avail- Calculation of likelihoods and associated LRs for able faecal calprotectin assays and investigated whether chosen result intervals clearly illustrated that the LR quantitative results of different assays were comparable. increases significantly when the faecal calprotectin con- Generally, all evaluated faecal calprotectin assays centration was higher than the cut-off value (Table 6). demonstrated good analytical performance. Within- Consequently, diagnostic information is lost when results run and total imprecision were found to be higher for are interpreted as positive/negative only. Bayes theorem the ELISA/FEIA techniques (Euroimmun, Orgentec and Oyaert et al.: Evaluation of faecal calprotectin assays 1569 0.81 (0.74–0.87) 1.72 (1.53–2.02) Comparison method Comparison Orgentec Calprotectin Orgentec − 0.67 (− 7.36 to 5.22) − 0.67 (− 7.36 to 640.1 (− 3934.8 to 5215.1) 640.1 (− 3934.8 to

0.90 (0.86–0.93) 5.83 (1.87–8.10) 0.37 (0.33–0.50) 0.87 (0.81–0.91) 4.00 (1.43–6.63) 0.80 (0.66–0.96) Turbo fCAL Bühlmann − 67.6 (− 1389.1 to 1253.8) − 67.6 (− 1389.1 to − 707.8 (− 4924.3 to 3508.7) − 707.8 (− 4924.3 to

0.62 (0.53–0.71) 0.70 (0.59–0.79) 1.35 (1.11–1.65) 0.60 (0.46–0.70) 0.82 (0.75–0.88) 1.07 (1.00–1.22) 1.27 (− 3.06 to 2.93) 1.27 (− 3.06 to 1.68 (− 0.72 to 5.84) 1.68 (− 0.72 to –6.77 (–13.2 to 1.49) –6.77 (–13.2 to Calprotectin Euroimmun –7.6 (–1861.7 to 1846.7) –7.6 (–1861.7 to − 75.2 (− 1518.4 to 1368.0) − 75.2 (− 1518.4 to − 715.3 (− 6131.2 to 4700.5) − 715.3 (− 6131.2 to

1.87 (0.19–2.70) 0.24 (0.21–0.29) 0.74 (0.64–0.82) 0.42 (0.37–0.45) 0.95 (0.93–0.97) 0.56 (0.51–0.63) 0.85 (0.79–0.90) 0.90 (0.86–0.93) 0.45 (0.41–0.48) 1.12 (0.20–1.29) 1.15 (− 0.07 to 1.71) 1.15 (− 0.07 to Calprotectin Diasorin –2.57 (–3.77 to 0.02) –2.57 (–3.77 to –189.0 (–1367.4 to 989.3) –189.0 (–1367.4 to − 256.7 (− 1572.1 to 1058.8) − 256.7 (− 1572.1 to − 896.8 (− 6019.3 to 4225.6) − 896.8 (− 6019.3 to − 181.5 (− 1567.7 to 1204.7) − 181.5 (− 1567.7 to

® 3.20 (0.85–5.50) 0.29 (0.25–0.36) 0.95 (0.92–0.96) 1.25 (1.09–1.39) 0.75 (0.64–0.82) 0.49 (0.40–0.64) 0.90 (0.86–0.94) 1.73 (0.82–2.81) 0.63 (0.52–0.72) 0.85 (0.78–0.90) 0.95 (0.92–0.97) 0.58 (0.48–0.66) 3.40 (− 0.59 to 5.40) 3.40 (− 0.59 to 1.64 (− 0.03 to 5.27) 1.64 (− 0.03 to 6.69) 5.54 (− 1.41 to Inova QUANTA Flash QUANTA Inova 79.3 (− 518.1 to 676.7) 79.3 (− 518.1 to − 209.1 (− 1313.3 to 895.1) − 209.1 (− 1313.3 to –136.6 (–1437.8 to 1164.5) –136.6 (–1437.8 to − 887.0 (− 6124.0 to 4350.4) − 887.0 (− 6124.0 to − 129.0 (− 1588.6 to 1330.7) − 129.0 (− 1588.6 to

Bias Intercept in brackets. indicated are intervals The 95% confidence Bias Intercept Slope 2 EliA Calprotectin Diasorin Calprotectin Diasorin r Bias Intercept Slope Calprotectin Euroimmun r Bias Intercept Slope Turbo fCAL Bühlmann r Bias Intercept Slope Calprotectin Orgentec r r Slope g/g]), and Passing-Bablok Regression analysis. Regression Passing-Bablok [ μ g/g]), and differences (mean analysis (r), Bland-Altman correlation Pearson of results ELISAs: calprotectin 4: Faecal Table 1570 Oyaert et al.: Evaluation of faecal calprotectin assays

50 Table 6: Likelihood ratios for various result intervals for the differ- 477 ent test. 411.9 142.9 Assay 0.0025

Method Likelihood ratios (95% CI) 100 (83.9–100) 100 (83.9–100) 100 (83.9–100) 100 (83.9–100) 98.5 (91.7–100)

58.6 (45.6–70.6) Negative 1-3xULN 3-10xULN > 10xULN 0.998 (0.954–1.000)

Orgentec Calprotectin Orgentec EliA 0.00 0.41 1.4 55 Calprotectin 2 (0.00–0.44) (0.10–1.70) (0.6–3.5) (8–391)

50 Diasorin 0.00 1.2 10.1 + ∞ 125 371

345.4 Calprotectin (0.00–0.38) (0.6–2.7) (2.2–45.7) (5 to + ∞) 0.0070 Inova QUANTA 0.00 0.43 2.8 + ∞ Flash® (0.00–0.46) (0.10–1.78) (1.3–6.3) (5 to + ∞) 100 (83.9–100) 100 (83.9–100) Bühlmann fCAL 0.00 0.14 1.9 59 95.2 (76.2–99.9) 58.4 (45.6–70.6) 96.9 (89.3–99.6) 95.2 (76.2–99.9) Turbo (0.00–0.47) (0.02–1.03) (0.8–4.9) (8–410)

0.991 (0.942–1.000) Euroimmun 0.00 0.21 0.76 + ∞ Calprotectin Euroimmun Calprotectin (0.00–0.50) (0.03–1.50) (0.30–1.80) (7.6 to + ∞) Orgentec 0.00 0.00 0.19 19 50 285

71.9 Calprotectin (0.00–0.50) (0.00–1.67) (0.03–1.39) (7–49) 284.5 0.0109 The values are presented as likelihood ratios with the 95% con- fidence interval (in parentheses); ULN, upper limit of normal; CI, 100 (83.9–100) 100 (83.9–100) confidence interval. 95.2 (76.2–99.9) 66.2 (53.4–77.4) 95.4 (87.1–99.0) 95.2 (76.2–99.9) 0.988 (0.936–1.000) Turbo fCAL Bühlmann

Phadia) compared to the CLIA/PETIA assays (Diasorin, 50 111 41.3

110.0 Inova and Bühlmann fCAL Turbo). It should be men- 0.0062 tioned that the protocol for Euroimmun calprotectin ELISA was applied on Inova QUANTA-Lyser 2, which is 100 (83.9–100) 100 (83.9–100) 100 (83.9–100) 100 (83.9–100) 78.5 (66.5–87.7) 96.9 (89.3–99.6) a non-validated instrument for this method. This could contribute to the higher imprecision results found for this 0.993 (0.946–1.000) Calprotectin Diasorin

assay. ® 50

115 Qualitative correlations between the methods from the 61.6 208.7

0.0129 different manufacturers were found to be good, but quan- titative agreement was poor, which means that the result of one method cannot be replaced by the result of another 100 (82.4–100) 100 (82.4–100) 100 (82.4–100) 84.2 (60.4–96.6) 72.9 (59.7–83.6) 86.4 (75.3–92.9) (Table 4). Therefore, faecal calprotectin concentrations 0.978 (0.916–0.998) Inova QUANTA Flash QUANTA Inova are not interchangeable. These results are in line with a

study from the UK National External Quality Assessment 50

376 Service that revealed up to 3.8-fold differences between 250.0 102.0

0.0164 methods from different manufacturers [6]. This suggests that, among other possible reasons, the used in the different assays are directed against different com- 100 (83.9–100) 100 (94.5–100) 85.7 (63.7–97.0) 95.2 (76.2–99.9) 66.2 (53.4–77.4) 85.7 (63.7–97.0)

EliA Calprotectin 2 EliA Calprotectin plexes of the faecal calprotectin protein. In addition, the 0.974 (0.915–0.996) methods evaluated use different antibodies (monoclonal

vs. polyclonal) of different origins (recombinant vs. s. native) with different immunoassay techniques (ELISA vs. PETIA vs. CLIA vs. FEIA) (Table 2). With the introduc- tion of more assays, efforts to harmonise faecal calprotec- tin assays would be of particular interest [26]. Our study and other studies identify the need for both a standard- ised calprotectin reference material against which these assays can be calibrated, as well as an in-house clinical evaluation. By comparing faecal calprotectin concentra- tions with endoscopy and histology, the diagnostic accu- 86 patients) were selected for ROC curve analysis. CI, confidence interval; SE, standard error. standard SE, interval; confidence CI, analysis. curve ROC for selected were (n = 86 patients) population the consecutive Only Sensitivity (95% CI) Sensitivity g/g) at fixed specificity of 95% of specificity fixed ( μ g/g) at Cut-off Sensitivity (95% CI) Sensitivity g/g) at fixed specificity of 75% of specificity fixed ( μ g/g) at Cut-off Specificity (95% CI) Specificity Sensitivity (95% CI) Sensitivity Manufacturer’s cut-off, μ g/g cut-off, Manufacturer’s Specificity (95% CI) Specificity Sensitivity (95% CI) Sensitivity SE AUC g/g) at maximal maximal ( μ g/g) at cut-off ROC’s specificity and sensitivity Sensitivities and specificities of the faecal calprotectin test at the the cut-off with maximal sensitivity and specificity, the manufacturer’s cut-off, and cut-off with fixed specificity of of specificity fixed with cut-off and cut-off, the manufacturer’s specificity, and sensitivity maximal with the cut-off at test calprotectin the faecal of specificities and 5: Sensitivities Table 95%. 75% and racy of the assays can be verified and if necessary, be used Oyaert et al.: Evaluation of faecal calprotectin assays 1571 to establish local cut-off values according to the intended less favourable disease course. Therefore, faecal calpro- use of the assay (i.e. diagnostic vs. disease monitoring). tectin analysis is part of the clinical diagnostic work-up Along with non-standardisation between the differ- of SpA patients [28]. In our study, 26 consecutive rheuma- ent faecal calprotectin assays, pre-analytical issues of tologic patients (24.8%) were included. However, a recal- faecal calprotectin analyses can contribute to the (often culation of the diagnostic performance of the different large) variation in faecal calprotectin concentrations assays for IBD by excluding this control group revealed between different assays [6, 7, 10, 26]. In this study, all no significant differences in optimal cut-off values and faecal samples where appropriately homogenised and performance characteristics of the different tests (data weighed before extraction. Extraction was performed not shown). using the same extraction device (i.e. the Roche Smart A drawback of the use of a single cut-off value is Prep Extraction device) for all assays. By using this device, that this diagnostic information is lost when results are we excluded variation in faecal calprotectin concentra- interpreted as positive/negative only. For example, a cal- tions due to liquid or hard faecal samples, which can be protectin result higher than 477 μg/g (Orgentec) does not found when using other extraction devices [6]. Moreover, mean the patient has IBD; neither does a result lower than by adding the appropriate amount of extraction buffer 477 μg/g rules out the diagnosis of IBD, but the probabil- based on the weighed faecal sample, we are convinced ity is different. To tackle this interpretative ambiguity, we that the initial dilution was correct. These pre-analytical defined result intervals and, thus, take full advantage of confounders need to be taken into account when labo- the quantitative test data. In addition, LRs allow us to ratories are planning to determine faecal calprotectin integrate (through Bayesian calculations) a physician’s concentrations. judgment about pre-test probability of the disease in the Overall, the diagnostic performance of the six evalu- patients’ cohort under estimation. For all assays, we were ated faecal calprotectin assays was excellent, with only able to define intervals with associated LRs that increase minor differences in sensitivity and specificity. The diag- when the results deviate quantitatively more from normal. nostic accuracy at the cut-off with the highest sum of Our approach results in a clear gain of diagnostic power sensitivity and specificity was high and not statistically of individual laboratory tests as compared with classic different between the assays evaluated, signifying that dichotomous interpretation [29]. the diagnostic performance of the various assays was Interpretation of laboratory results must be done in comparable. Cut-off values calculated from ROC analy- the clinical context and in function of the pre-test prob- ses for detection of IBD were considerably higher for the ability [29]. For example, in our study, we determined Orgentec assay (477 μg/g) than for the other assays, com- that in patients with diarrhoea, mucous or bloody stools, pared to the recommended cut-offs by all manufacturers weight loss and abdominal pain and cramping, the pre- (all 50 μg/g faeces). The sensitivity as well as specificity test probability for IBD is 25.7%. The post-test probabil- of all assays was comparable when using the optimal ity for IBD in these patients will be 31.0%, 77.4%, 49.3%, cut-offs. On the contrary, we found that all assays had 39.4%, 19.6% and 19.6% for patients with a faecal cal- a sensitivity of 100% when the cut-off of the manufac- protectin concentration between 3 and 10 times the ULN turer was used (i.e. 50 μg/g) but specificity at this cut-off for the Phadia, Diasorin, Inova, Bühlmann, Euroimmun value differed between the assays (ranging from 58.4% to and Orgentec assay, respectively. The post-test probabil- 78.5%). This, together with the fact that there is not yet ity for patients with a faecal calprotectin concentration a standardised method for faecal calprotectin measure- greater than 10-fold the ULN will be 94.9%, 99.4%, 93.1%, ment, indicates that recommended cut-off values would 95.1%, 99.6% and 86.8% for the Phadia, Diasorin, Inova, have to be determined depending on the clinical setting Bühlmann, Euroimmun and Orgentec assay, respectively. they are used in [10]. These findings suggest that patients with a high pre-test There is a close relationship between spondylar- probability and levels of faecal calprotectin greater than thropathies (SpA) and IBD. Clinically, 5%–10% of all 10-fold the cut-off value have a high probability for having patients with ankylosing spondylitis have concurrent IBD, aiding in clinical decision making. However, these IBD. Even in the absence of relevant gastrointestinal findings depend on the clinical setting and the assay used symptoms, microscopic gut inflammation was observed for faecal calprotectin determination and therefore need in more than 40% of SpA patients [27]. The presence to be confirmed in larger cohorts of patients. of (microscopic) gut inflammation appears relevant The strength of our study lies in the fact that we for prognosis and therapeutic decision making as SpA included six different assays, of which the Inova QUANTA patients with a chronic type of gut inflammation have a Flash® and EliA Calprotectin 2 are evaluated for the first 1572 Oyaert et al.: Evaluation of faecal calprotectin assays time. In addition, we defined two well-described disease 4. Van Rheenen PF, Van de Vijver E, Fdler V. Faecal ­calprotectin control populations, thereby reflecting the real clinical for screening of patients with suspected ­inflammatory bowel disease: diagnostic meta-analysis. Br Med J 2010; use of this test. However, our study also has two impor- 341:c3369. tant limitations. Compared to other studies, we included a 5. Menees SB, Powel C, Kurlander J, Goel A, Chey WD. A meta-anal- relative low number of samples. Second, the sensitivity of ysis of the utility of C-reactive protein, erythrocyte sedimenta- all assays at their recommended cut-off is 100%, which is tion rate, fecal calprotectin, and fecal lactoferrin to exclude in contrast to other studies. Including more IBD patients inflammatory bowel disease in adults with IBS. Am J Gastroen- of which at least some have faecal calprotectin concentra- terol 2015;110:444–54. 6. Whitehead SJ, Franch J, Brookes MJ, Ford C, Gama R. tions lower than 50 μg/g would make the test population Between-assay variability of faecal calprotectin enzyme-linked more suitable for comparison. The only differences we immunosorbent assay kits. Ann Clin Biochem 2013;50:53–61. found are in specificity, but there are definitely also differ- 7. Oyaert M, Trouvé C, Baert F, De Smet D, Langlois M, ences in sensitivity between the assays. ­Vanpoucke H. Comparison of two immunoassays for measure- In conclusion, we demonstrated good analytical per- ment of faecal calprotectin in detection of inflammatory bowel disease: (pre)-analytical and diagnostic performance character- formance of all assays tested. Variation in the faecal cal- istics. Clin Chem Lab Med 2014;52:391–7. protectin concentrations between the different methods 8. Dumoulin EN, Van Biervliet S, Langlois MR, Delanghe JR. Pro- does not allow to use these assays interchangeably. Diag- teolysis is a confounding factor in the interpretation of faecal nostic performance is high but depends on the clinical calprotectin. Clin Chem Lab Med 2015;53:65–71. setting in which the test is used. Use of LRs for different 9. Kolho KL, Turner D, Veereman-Wauters G, Sladek M, de result intervals aids in the clinical interpretation of faecal ­Ridder L, Shaoul R, et al. Rapid test for fecal calprotectin levels in children with Crohn’s disease. J Pediatr Gastroenterol Nutr calprotectin results, thereby improving the clinical inter- 2012;55:436–9. pretation for the diagnosis of IBD as compared with con- 10. Delefortrie Q, Schatt P, Grimmelprez A, Gohy P, Deltour P, ventional dichotomous test interpretation. Collard G, et al. Comparison of the Liaison® Calprotectin kit with a well-established point-of-care test (Quantum Blue – ® Acknowledgments: We thank Thermo Fisher Scientific, Bühlmann-Alere in terms of analytical performances and ability to detect relapses amongst a Crohn population in follow-up. Clin Diasorin, Inova, Bühlmann, Euroimmun and Launch Biochem 2016;49:268–73. Diagnostics for the donation of the assays. We are grate- 11. Basso D, Zambon CF, Plebani M. Inflammatory bowel diseases: ful to the laboratory technicians for their most valuable where from pathogenesis to laboratory testing. Clin Chem Lab efforts. Med 2014;52:471–81. Author contributions: All the authors have accepted 12. Tonutti E, Agostinis P, Bizzaro N. Inflammatory Bowel Diseases: responsibility for the entire content of this submitted where are we and where should we go. Clin Chem Lab Med 2014;52:463–5. manuscript and approved submission. 13. Lasson A, Strid H, Ohman L, Isaksson S, Olsson M, Rydström B, Research funding: None declared. et al. Fecal calprotectin one year after ileocaecal resection for Employment or leadership: None declared. crohn’s disease- a comparison with findings at oleocolonos- Honorarium: None declared. copy. J Crohns Colitis 2014;8:789–95. Competing interests: The funding organization(s) played 14. Hessels J, Douw G, Yildrim DD, Meerman G, van Herwaarden MA, van den Bergh FA. Evaluation of Prevent ID and Quantum no role in the study design; in the collection, analysis, and Blue rapid tests for fecal calprotectin. Clin Chem Lab Med interpretation of data; in the writing of the report; or in the 2012;50:1079–82. decision to submit the report for publication. 15. Dolci A, Panteghini M. Comparative study of a new quantitative rapid test with an established ELISA method for faecal calprotec- tin. Clin Chim Acta 2012;413:350–1. 16. Wassell J, Wallage M, Brewer E. Evaluation of the quantum References blue rapid test for faecal calprotectin. Ann Clin Biochem 2012;49:55–8. 1. Silicia B, Saro C. Inflammatory bowel diseases: a disease(s) of 17. Coorevits L, Baert FJ, Vanpoucke HJ. Faecal calprotectin: compar- modern times? Is incidence still increasing. World J Gastroenterol ative study of the Quantum Blue rapid test and an established 2008;14:5491–8. ELISA method. Clin Chem Lab Med 2013;51:825–31. 2. Soon I, Molodecky N, Rabi D, Ghali W, Barkema H, Kaplan G, et al. 18. Benahmed NA, Manéné D, Barbot-Trystram L, Kapel N. Evalua- Increasing incidence and prevalence of the inflammatory bowel tion of Calfast® immunochromatographic quantitative assay for diseases with time, based on systematic review. Gastroenterol the measurement of calprotectin in faeces. Clin Chem Lab Med 2012;143:46–54. 2014;52:e143–5. 3. World Gastroenterology Organisation (WGO). World Gastroen- 19. Vestergaard T, Nielsen SL, Dahlerup JF, Hornung N. Fecal cal- terology Organisation Global Guideline. Inflammatory Bowel protectin: assessment of a rapid test. Scand J Clin Lab Invest disease, a global perspective. Munich, Germany: WGO, 2009. 2008;68:343–7. Oyaert et al.: Evaluation of faecal calprotectin assays 1573

20. Burri E, Manz M, Rothen C, Rossi L, Beglinger C, Lehmann FS. 24. CLSI. Evaluation of Precision Performance of Quantitative Meas- Monoclonal antibody testing for fecal calprotectin is superior to urement Methods; Approved Guideline – Second Edition. CLSI polyclonal testing of fecal calprotectin and lactoferrin to identify Document EP05-A2. Wayne, PA: Clinical Laboratory Standards organic intestinal disease in patients with abdominal discom- Institute, 2004. fort. Clin Chim Acta 2013;416:41–7. 25. Haeckel R. Recommendations for definition and determination 21. Sydora M, Sydora B, Fedorak R. Validation of a point-of-care of carry-over effects. J Automat Chem 1988;10:181–3. desk top device to quantitate fecal calprotectin and distinguish 26. Sun S, Cavey T, Peltier L, Bendavid C, Bouguen G. Letter: wide inflammatory bowel disease from irritable bowel syndrome. variation in faecal calprotectin values according to the assay. J Crohns Colitis 2012;6:207–14. Aliment Pharmacol Ther 2016;43:166–80. 22. Labaere D, Smismans A, Van Olmen A, Christiaens P, D’Haens G, 27. Cypers H, Varkas G, Beeckman S, Debusschere K, Vogl T, Roth J, Moons V, et al. Comparison of six different calprotectin assays et al. Elevated calprotectin levels reveal bowel inflammation in for the assessment of inflammatory bowel disease. United Euro- spondyloarthritis. Ann Rheum Dis 2015;75:1357–62. pean Gastroenterol J 2014;2:30–7. 28. Wendling D, Vuitton L, Koch S, Prati C. Spondyloarthritis and the 23. Nilsen T, Sunde K, Hansson L, Havelka AM, Larsson A. A novel gut: a new look. Joint Bone Spine 2015;82:77–9. turbidimetric immunoassay for fecal calprotectin optimized for 29. Bossuyt X. Clinical performance characteristics of a laboratory routine chemistry analyzers. J Clin Lab Anal 2016. DOI:10.1002/ test. A practical approach in the autoimmune laboratory. Auto- jcla.22061. immun Rev 2009;8:543–8.