Clinica Chimica Acta 438 (2015) 388–395

Contents lists available at ScienceDirect

Clinica Chimica Acta journal homepage: www.elsevier.com/locate/clinchim

Multicenter evaluation of a new progastrin-releasing peptide (ProGRP) immunoassay across Europe and China Catharina M. Korse a,⁎, Stefan Holdenrieder b, Xiu-yi Zhi c, Xiaotong Zhang d, Ling Qiu d, Andrea Geistanger e, Marcus-Rene Lisy e, Birgit Wehnl e, Daan van den Broek a, José M. Escudero f, Jens Standop b, Mu Hu c, Rafael Molina f a

The Netherlands Cancer Institute, Plesmanlaan 121, Amsterdam 1066CX, The Netherlands University Hospital Bonn and Center of Integrative Oncology Cologne/Bonn, Sigmund-Freud-Strasse, Bonn 25-53127, Germany c Xuanwu Hospital, Capital Medical University, No. 45 Changchun Street, Xicheng District, Beijing 100053, China d Peking Union Medical College Hospital (PUMCH), 1Shuaifuyuan, Dongcheng District, Beijing 100730, China e Roche Diagnostics GmbH, Nonnenwald 2, Penzberg 82377, Germany f Hospital Clinic, University of Barcelona, Carrer Villarroel 170, Barcelona 08036, Spain b

a r t i c l e

i n f o

Article history: Received 15 September 2014 Received in revised form 18 September 2014 Accepted 18 September 2014 Available online 28 September 2014 Keywords: Differential diagnosis Immunoassay Progastrin-releasing peptide ProGRP SCLC Stability

a b s t r a c t Background: We performed a multicenter evaluation of the Elecsys® progastrin-releasing peptide (ProGRP) immunoassay in Europe and China. Methods: The assay was evaluated at three European and two Chinese sites by imprecision, stability, method comparison and differentiation potential in lung cancer. Results: Intermediate imprecision across five analyte concentrations ranged from 2.2% to 6.0% coefficient of variation. Good stability for plasma and serum samples was shown for various storage conditions. There was excellent correlation between the Elecsys® and ARCHITECT assays in plasma (slope 1.02, intercept −2.72 pg/mL). The Elecsys® assay also showed good correlation between serum and plasma samples (slope 0.93, intercept 2.35 pg/mL; correlation coefficient 0.97). ProGRP differentiated small-cell and non-small-cell lung cancer (NSCLC; area under the curve 0.90, 95% CI 0.87–0.93; 78.3% sensitivity, 95% specificity; at 84 pg/mL), with no relevant effects of ethnicity, age, gender or smoking. Median ProGRP concentrations were low in benign diseases (38 pg/mL), other malignancies (40 pg/mL) or NSCLC (39 pg/mL), except chronic kidney disease above stage 3 (N 100 pg/mL). Conclusions: Increased stability of the Elecsys® ProGRP assay in serum and plasma offers clear benefits over existing assays. This first evaluation of a ProGRP assay in China demonstrated comparable differentiation potential among different ethnicities. © 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

1. Introduction Abbreviations: NSCLC, non-small-cell lung cancer; SCLC, small-cell lung cancer; NSE, neuron-specific enolase; ProGRP, progastrin-releasing peptide; NET, neuroendocrine tumors; MCT, medullary carcinoma of the thyroid; PUMCH, Peking Union Medical College Hospital; SST, serum separation tube; RST, rapid serum tube; CLSI, Clinical Laboratory and Standards Institute; UICC, Union for International Cancer Control; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; AUC, area under the curve; HSP, human sample pool. ⁎ Corresponding author at: The Netherlands Cancer Institute, PO Box 90203, Plesmanlaan 121, Amsterdam 1066CX, The Netherlands. Tel.: + 31 20 5122794; fax: +31 20 5122799. E-mail addresses: [email protected] (C.M. Korse), [email protected] (S. Holdenrieder), [email protected] (X. Zhi), [email protected] (X. Zhang), [email protected] (L. Qiu), [email protected] (A. Geistanger), [email protected] (M.-R. Lisy), [email protected] (B. Wehnl), [email protected] (D. van den Broek), [email protected] (J.M. Escudero), [email protected] (J. Standop), [email protected] (M. Hu), [email protected] (R. Molina).

Tumor markers have been extensively studied in patients with lung cancer as a means to differentiate between the two major subtypes of lung cancer—non-small-cell lung cancer (NSCLC) and small-cell lung cancer (SCLC)—and, thereby, improve diagnosis and treatment selection [1–5]. NSCLC accounts for around 80% of all new lung cancer cases, with SCLC making up the remaining 20%. SCLC differs from NSCLC in having neuroendocrine differentiation, a higher tumor growth rate and earlier development of metastasis [6,7], and as such requires a different treatment approach. Of the two subtypes, NSCLC is more likely to present at an early stage, when surgery offers the best chance of cure [8]. The early-stage diagnosis of SCLC is very rare, meaning surgery is uncommon, but SCLC is highly sensitive to radiotherapy and chemotherapy [7]. Patients with SCLC often relapse; however, and 5-year survival

http://dx.doi.org/10.1016/j.cca.2014.09.015 0009-8981/© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/3.0/).

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

rates have remained constant in recent years [9]. The differential diagnosis of lung cancer subtype at initial presentation is critical to ensure appropriate therapeutic intervention. Tumor biopsies are an essential component in the histologic differentiation of lung cancer. However, because of the submucosal location of many SCLCs, accurate tissue sampling can be difficult, and biopsies do not allow for early detection of the disease [10]. When diagnosed with limited-stage disease, approximately 20% of patients with SCLC can achieve long-term survival with aggressive chemotherapy and radiotherapy, versus just 5% when diagnosed at an advanced stage [10]. The analysis of tumor markers in plasma or serum samples offers clear benefits over histologic differentiation, including the potential for early detection of SCLC, and the chance to improve survival rates. Neuron-specific enolase (NSE) and progastrin-releasing peptide (ProGRP) have proved most beneficial as tumor markers in SCLC [11, 12]. Although NSE was historically the recommended tumor marker for SCLC [13], NSE also stains up to 80% of NSCLCs in tissue examinations and is elevated in the sera of 20–30% of patients with NSCLC [14]. Furthermore, NSE has low sensitivity, particularly in patients with disease confined to the hemithorax or ipsilateral mediastinum [15]. As NSE is present in platelets and erythrocytes, samples with hemolysis must be excluded and rapid storage of samples is essential [14]. ProGRP accurately discriminates between NSCLC and SCLC [16,17] and is rarely elevated in other malignant diseases or in benign conditions, except in patients with renal insufficiency, neuroendocrine tumors (NET) of the lung and medullary carcinoma of the thyroid (MCT) [16–22]. Evaluation of the first fully automated ProGRP ARCHITECT assay (Abbott Laboratories, Wiesbaden, Germany) was reported in 2009 [23]. Owing to the poor stability of ProGRP in serum on the ARCHITECT assay, which is believed to be due to thrombin-induced proteolysis, plasma samples are the recommended source material [24,25]. The Elecsys® ProGRP assay (Roche Diagnostics GmbH, Penzberg, Germany) is a new immunoassay designed to quantitatively determine levels of ProGRP in both human serum and plasma. As the two monoclonal antibodies in the Elecsys® ProGRP assay bind to epitopes in the ProGRP peptide that are relatively resistant to endoproteolytic cleavage [16,17,26] (Supplementary Fig. S1), serum samples as well as plasma samples can be used. Here we report on the technical and clinical performance of the Elecsys® ProGRP assay across a number of European and Chinese sites.

389

(Roche Diagnostics). Both the calibrator and the control contain recombinant ProGRP. 2.2. Sample sources, preparation and handling Samples were sourced from patients with previously untreated active disease, who had sufficient sample material available for analysis. No further demographic or histologic selection criteria were utilized. Lung cancer histologic types were classified according to the 1999 World Health Organization recommendations [27]. Differential diagnosis between SCLC and NSCLC was based on morphologic characteristics plus a positive CD56 and/or synaptophysin immunohistochemistry of the tumor. Lung cancer staging (TNM) was established according to international guidelines [28]. All sites collected samples and performed assay measurements. An additional German site (Institut für Klinische Pharmakologie GmbH, Kiel) contributed samples from apparently healthy individuals as a reference cohort. Owing to the low prevalence of SCLC in the patient population, differential diagnosis was based primarily on patient serum samples sourced from sample banks in Europe, whereas samples were collected prospectively in China (January to September 2013). All three European sites and PUMCH used serum for clinical evaluation. Xuanwu used K2-EDTA primary tubes for plasma collection. The stability experiment was performed in Bonn and Amsterdam using K2-EDTA and serum separation tubes (SSTs). In addition, Amsterdam used rapid serum tubes (RSTs) for serum sampling in this experiment. The interior wall of the RSTs was coated with thrombin to promote rapid clotting. All studies were performed on cobas® e411 and e601 analyzers. 2.3. Statistical analyses All analytical data were captured using the WinCAEv (Windowsbased computer aided evaluation) software program. Demographic and clinical data were collected using the MACRO software program. All statistical analyses using clinical information were performed in the biostatistics department at Roche Diagnostics Penzberg using SAS (Statistical Analysis Software, version 9.2) and R (version 2.13.2). Outliers identified by visual inspection of the primary data set were re-measured.

2. Materials and methods

2.4. Technical assessment

Between August 2012 and September 2013, the Elecsys® ProGRP assay was evaluated at three European investigational sites in Amsterdam, Barcelona and Bonn, and two Chinese sites in Beijing (Peking Union Medical College Hospital [PUMCH] and Peking Xuanwu Hospital). Ethical approval/waiver was obtained from each institution before clinical study work began. All investigational sites conducted the study in accordance with the Declaration of Helsinki (rev. Tokyo, Venice, Hong Kong and Fortaleza 2013) and International Conference on Harmonisation Good Clinical Practice guidelines. The objective of the study was to evaluate the performance of the Elecsys® ProGRP assay in terms of imprecision, stability, method comparison and differentiation potential for SCLC.

2.4.1. Interlaboratory survey Three EDTA plasma sample pools were prepared by Roche R&D (~30, ~200 and ~1500 pg/mL) and distributed to all five investigational sites to evaluate differences in recovery and day-to-day variability between the laboratories in these three concentration ranges. The two levels of PreciControl ProGRP were also used as sample material. Each sample was measured in single determination over 10 days in each laboratory. The percentage recovery per sample was calculated as the measured concentration/all laboratory median ×100.

2.1. Assay description The Elecsys® ProGRP assay is an electrochemiluminescence immunoassay that uses a biotinylated ProGRP-specific mouse monoclonal antibody and a ruthenium-labeled ProGRP-specific mouse monoclonal antibody to capture and detect ProGRP in human serum and plasma. The assay is calibrated using the ProGRP CalSet (Roche Diagnostics) and has been standardized against the ARCHITECT ProGRP assay. Quality control is performed using two levels of PreciControl ProGRP

2.4.2. Imprecision according to clinical laboratory and standards institute EP5-A2 Imprecision was assessed by means of the Clinical Laboratory and Standards Institute (CLSI) EP5-A2 guideline [29]. Three sample pools were prepared by each of the European sites with specified target concentration ranges of 7–60 pg/mL, 61–1000 pg/mL and 1001–5000 pg/mL. Approximately 27 mL of serum or plasma was required to prepare 84 × 300 μL aliquots. Samples were stored at −20 °C and used on the respective day of measurement. Repeatability and intermediate precision estimates were obtained by modeling the data according to the CLSI EP5-A2 variance component model. Between-day, between-run and repeatability variance components

390

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

were modeled. The intermediate precision estimate was given as the coefficient of variation (CV) based on the total variance estimate. 2.4.3. Stability of sample material Matched serum and plasma specimens were obtained from patients who provided written informed consent, with approval of the institutional review board. The concentration of ProGRP in the samples was ideally to exceed 100 pg/mL; however, some samples below this concentration were included as it proved difficult to obtain informed consent and samples. A minimum of 3 mL each of serum and plasma was required per patient to prepare nine 300 μL aliquots. Measurements were performed within 1 h of sampling for baseline determinations, and after 1, 2, 3 and 4 hours at room temperature and 3, 6, 24 and 48 hours at 2–8 °C. Measurements were first taken on the ARCHITECT instrument and then, within 30 minutes, on the cobas® instrument. Serum and plasma were tested within the same run. The percentage recovery was calculated as the actual concentration/baseline concentration ×100. 2.4.4. Method comparison Comparisons were obtained with the ARCHITECT ProGRP assay and the Fujirebio microtiter plate enzyme-linked immunosorbent ProGRP assay (Fujirebio Diagnostics, Japan). Ideally, plasma samples were to be used for the ARCHITECT instrument and serum samples for the Fujirebio assay. Deming regression with an error ratio of 1 was used for the estimation of the regression lines [30]. Confidence intervals (CIs) for intercept and slope were obtained via the bootstrap method [31]. 2.4.5. Matrix comparison Matched serum and plasma specimens were collected from routine samplings between Q3 2012 and Q1 2013 and frozen until batch-wise measurement in Q1 2013. Samples were collected over several weeks to attempt to cover the entire measuring range (3–5000 pg/mL). 3. Clinical assessment 3.1. Reference range determination For the ProGRP reference range calculation, healthy cohorts were compared from the five European and two Chinese sites, plus the cohort of apparently healthy individuals recruited from the Kiel site according to their responses to a health check questionnaire and results of basic clinical chemistry parameters, such as glucose, cholinesterase, creatinine, C-reactive protein and hemoglobin levels. Three specimen types were collected at the Kiel site: serum, K2-EDTA plasma and Li-Heparin plasma. Samples from Kiel were stored at − 80 °C in the sample bank at Roche Penzberg and measured in Bonn. Amsterdam, Barcelona and Bonn collected serum samples only. PUMCH in China collected serum and K3-EDTA plasma samples, and Xuanwu collected K2-EDTA plasma samples only. Quantiles were calculated with the type = 3 method in the R software. The CIs were nonparametric with regard to the method of Hahn and Meeker [32]. 3.2. Differential diagnosis The malignant cohorts of NSCLC and SCLC were the primary focus, with control cohorts of apparently healthy individuals, patients with benign lung diseases (i.e., chronic obstructive pulmonary disease, tuberculosis, pneumonia and asthma), patients with other benign diseases (i.e., acute and chronic inflammatory, liver, renal, autoimmune and metabolic diseases) and patients with other malignant diseases. For the malignant cohorts, clinical classification (cTNM) or pathologic classification (pTNM) as well as Union for International Cancer Control (UICC) staging was obligatory, with the number of metastases and metastatic sites recorded. The date of sampling, date of primary diagnosis,

histology results, demographic data and smoking habits (current smoker, past smoker or never smoker) were also documented. Creatinine was measured in parallel to ProGRP, and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [33,34] was used to calculate estimated glomerular filtration rate (eGFR) to clarify the effect of renal insufficiency on ProGRP concentrations. A cutoff of ≥30 mL/min/1.73 m2 (CKD stage 3) was chosen because ProGRP levels were found to be significantly increased in patients with a CKD stage N 3 (Supplementary Fig. S2). Statistical analysis was based on receiver operating characteristic (ROC) curves and area under the curve (AUC) calculations, with CIs based on the DeLong method [35]. Sensitivities at 95% specificity, and specificity at 95% sensitivity, were also calculated. 4. Results 4.1. Technical assessment The CV of the interlaboratory survey pooled samples ranged from 1.2% to 4.9%. The recovery of the ~ 30 pg/mL and ~ 200 pg/mL pooled samples ranged from 94% to 106%. The recovery of the ~ 1500 pg/mL sample varied more widely between laboratories, with an approximate 20% difference in recovery between the Bonn site (104%) and the Amsterdam site (85%). The intermediate imprecision for the three European human serum pool samples and two control samples ranged from 2.2% (95% CI 1.8–2.7) to 6.0% (95% CI 4.7–8.4) CV (Supplementary Table S1), which is comparable with data reported for the ARCHITECT ProGRP assay (2.2–5.7% CV for serum and plasma) [23]. Within-run imprecision ranged from 1.1% (95% CI 0.9–1.4) to 3.0% (95% CI 2.5–3.8) CV. Sample stability was evaluated in nine samples from Amsterdam and 10 samples from Bonn, with ProGRP concentrations ranging from 18 to 4259 pg/mL. Comparable recovery on the cobas® assay was obtained with plasma and serum samples when stored at room temperature over 4 hours (median recovery at 4 hours: 98% in plasma, 102% in SST and 99% in RST samples), or at 2–8 °C over 48 hours (median recovery at 48 hours: 97% in plasma, 101% in SST and 101% in RST samples) (Fig. 1). Good ProGRP stability was also demonstrated in plasma and serum samples stored at − 20 °C for 12 weeks (Supplementary Fig. S3). Conversely, with the ARCHITECT assay, ProGRP in serum was not well recovered over time (median recovery at 4 hours: 90% in SST and 69% in RST samples), and results for plasma and serum samples were not comparable, as previously reported [24]. The correlation between the ARCHITECT and the cobas® assays using plasma samples up to ProGRP concentrations of 500 pg/mL is shown in Fig. 2a. The slope and intercept across all sites combined was 1.02 (95% CI 0.96–1.08) and −2.72 pg/mL (95% CI −5.22–0.66), respectively, with a correlation coefficient of 0.96 (Supplementary Table S2). Similar findings were reported for all sites across the entire measuring range of 3–5000 pg/mL (data not shown). Across all sites, the relationship between serum and plasma samples on the cobas® assay was given by a slope of 0.93 (95% CI 0.89–0.98) and an intercept of 2.35 pg/mL (95% CI −0.21–4.60), with a correlation coefficient of 0.97 for ProGRP concentrations up to 500 pg/mL (Fig. 2b). The slope for the matrix comparison at the Barcelona site was significantly higher than the other sites. The reason for this is unknown but is not considered to be clinically relevant and highlights the variability of the assay in the field. Comparable results were obtained for the correlation between the Fujirebio and the cobas® assays using serum samples up to ProGRP concentrations of 500 pg/mL (Fig. 2c). Across all sites, the slope was 1.33 (95% CI 1.15–1.47) and the intercept was − 4.18 pg/mL (95% CI −10.50–2.95), with a correlation coefficient of 0.84. 4.2. Clinical assessment Sample numbers from each patient cohort, and for each site, are shown in Table 1. The reference range cohort consisted of 1085

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

391

Room temperature

2–8°C

Plasma (K2-EDTA) samples (n = 19)

B 120

120

100

100

Recovery (%)

Recovery (%)

A

80 60

80 60

cobas® ARCHITECT

40 0

1

2 3 Time (hours)

cobas® ARCHITECT

40 4

0

3

6 24 Time (hours)

48

SST samples (n = 19)

D 120

120

100

100

Recovery (%)

Recovery (%)

C

80 60

80 60

cobas® ARCHITECT

40 0

1

2 3 Time (hours)

cobas® ARCHITECT

40 4

0

3

6 24 Time (hours)

48

RST samples (n = 9)

F 120

120

100

100

Recovery (%)

Recovery (%)

E

80 60

80 60

cobas® ARCHITECT

40 0

1

2 3 Time (hours)

cobas® ARCHITECT

40 4

0

3

6 24 Time (hours)

48

Fig. 1. Median ProGRP recovery from patient samples taken in Amsterdam (n = 9) and Bonn (n = 10), incubated at room temperature over 4 hours (A, C, E) or at 2–8 °C over 48 hours (B, D, F) and measured on the cobas® or ARCHITECT instruments.

apparently healthy individuals with a mean age of 48 years (standard deviation [SD] 16 years). The NSCLC cohort included 852 patients with a mean age of 64 years (SD 12 years) and the SCLC cohort consisted of 207 patients with a mean age of 62 years (SD 11 years). In the reference range cohort, 95th percentile ProGRP concentrations were 68 pg/mL (95% CI 63.7–74.5) in Li-Heparin plasma, 60 pg/mL (95% CI 55.8–65.3) in EDTA plasma and 66 pg/mL (95% CI 62.4–72.6) in serum samples (Fig. 3). These data were comparable with the reference range values on the ARCHITECT assay (63 pg/mL

serum; 65 pg/mL EDTA plasma) [36] but higher than those reported on the Fujirebio assay (43 pg/mL in serum) [37]. No significant difference in median ProGRP concentration was observed between the European and Chinese sites. ProGRP showed good differential diagnosis between SCLC and NSCLC, with an AUC of 0.90 (95% CI 0.87–0.93) with 78.3% sensitivity at 95% specificity in the NSCLC cohort. The distribution of ProGRP is shown in Fig. 4. No clinically relevant between-site effects were observed within the NSCLC and SCLC cohorts across Europe and China,

392

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

Cobas® ProGRP plasma (pg/mL)

A

500

Identity

400

300

200 Deming regression fit: Amsterdam (n = 147, r = 0.98) Barcelona (n = 107, r = 0.96) Bonn (n = 148, r = 0.94) PUMCH (n = 187, r = 0.98) Xuanwu (n = 214, r = 0.98) All sites (n = 803, r = 0.96)

100

0 0

100

200

300

400

500

ARCHITECT ProGRP plasma (pg/mL)

Cobas ® ProGRP plasma (pg/mL)

B

500

Identity

400

300

5. Discussion

200 Deming regression fit:

100

Amsterdam (n = 147, r = 1.00) Barcelona (n = 107, r = 0.95) Bonn (n = 148, r = 0.99) PUMCH (n = 171, r = 0.99) All sites (n = 573, r = 0.97)

0 0

100

200

300

400

500

Cobas® ProGRP serum (pg/mL)

Cobas ® ProGRP plasma (pg/mL)

C

although ProGRP concentrations in the SCLC cohort tended to be higher in the Chinese sites (data not shown). No clinically relevant effect of age, gender or smoking habits on ProGRP concentrations was noted (data not shown). A clinical differentiation cutoff for separating SCLC (n = 207) from NSCLC (n = 852) was calculated as 84 pg/mL (95% CI 76.9–98.8), based on 95% specificity in the NSCLC cohort. ProGRP values differed in samples from patients with limited versus extensive SCLC (median 351 pg/mL vs. 757 pg/mL, respectively; p = 0.0023) (Supplementary Fig. S4). At a fixed specificity of 95% in the NSCLC cohort, the sensitivity for limited-stage SCLC versus NSCLC was 71.7% (AUC of 85.6; 95% CI 80.3–90.9) and for extensive-stage SCLC was 83.5% (AUC of 93.0; 95% CI 89.8–96.3). In patients with SCLC, a correlation between UICC disease stage and ProGRP levels was apparent, although the sample sizes of patients with stage I and II SCLC were small (data not shown). ProGRP concentrations across all sites were low in patients with benign diseases when excluding renal disease (median 38 pg/mL), and in patients with other malignancies when excluding renal disease, MCT or NET of the lung (median 40 pg/mL). ProGRP showed even greater differential diagnosis between SCLC and NSCLC in patients at the two Chinese sites (AUC 0.94 [95% CI 0.89–0.99]; 83.8% sensitivity at 95% specificity) compared with the three European sites (AUC 0.89 [95% CI 0.85–0.92]; 76.9% sensitivity at 95% specificity in the NSCLC cohort), although this was not statistically significant, possibly owing to the smaller sample size of the Chinese cohort (Fig. 5).

500

Identity

400

300

200

Deming regression fit:

100

Amsterdam (n = 147, r = 0.97) Barcelona (n = 109, r = 0.90) Bonn (n =143, r = 0.82) All sites (n = 399, r = 0.84)

0 0

100

200

300

400

500

Fujirebio ProGRP serum (pg/mL) Fig. 2. (A) Method comparison: ARCHITECT versus cobas® using plasma samples up to a ProGRP concentration of 500 pg/mL; (B) Matrix comparison: serum versus plasma samples on cobas®; (C) Method comparison: Fujirebio versus cobas® using serum samples.

This multicenter evaluation of the Elecsys® ProGRP assay showed good imprecision, stability and specificity across a number of European and Chinese sites. Intermediate imprecision values ranged from 2.2% to 6.0% CV and were comparable with data reported on the ARCHITECT ProGRP assay [23]. Within-run imprecision ranged from 1.1% to 3.0% CV. Of note, the Elecsys® ProGRP assay performed equivalently in EDTA plasma and serum samples, with recovery maintained across a range of storage conditions. Archival samples are not always believed to be as reliable as fresh samples, but our data demonstrated that ProGRP remained stable in both archival and fresh tissues. In contrast, the recovery of ProGRP in serum decreased over time with the ARCHITECT assay, and differing results were obtained with plasma and serum samples, as previously reported [24]. Stability issues reported with serum samples analyzed on the ARCHITECT assay were thought to be partially due to thrombin within the serum samples, and plasma was identified as the preferred source material [24]. The greater stability of the Elecsys® ProGRP assay in serum, relative to the ARCHITECT assay, most likely relates to the antibody-binding sites on the ProGRP peptide. The two monoclonal antibodies in the Elecsys® ProGRP assay bind to epitopes that are relatively resistant to endoproteolytic cleavage [16,17,26], whereas the two monoclonal capture antibodies in the ARCHITECT assay bind directly next to the thrombin-cleavage site (Supplementary Fig. S1). The ARCHITECT and cobas® assays showed good correlation in plasma across all sites up to ProGRP concentrations of 500 pg/mL (slope 1.02, intercept − 2.72 pg/mL) and across the entire measuring range. Good correlation between serum and plasma was also demonstrated on the cobas® assay, up to ProGRP concentrations of 500 pg/mL (slope 0.93, intercept 2.35 pg/mL), and across the entire measuring range. There was also good correlation between the Fujirebio and cobas® assays in serum samples across all sites up to ProGRP concentrations of 500 pg/mL (slope 1.33, intercept − 4.18 pg/mL). Notably, samples from laboratories in Bonn and Barcelona did not evenly cover the measuring range, with most samples in the region below 500 pg/mL ProGRP. The difference between the ARCHITECT and cobas® assays over the whole measuring range is smaller than that reported for the correlation between the ARCHITECT and Fujirebio

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

393

Table 1 Sample numbers for each cohort from each site. Cohort Healthy

SCLC

NSCLC

Benign lung disease

Other benign disease

Thereof renal Other malignant disease

Thereof MCT Thereof NET of the lung

N Mean age, years (SD) Male/female, n N Mean age, years (SD) Male/female, n Limited/extensive disease, n N Mean age, years (SD) Male/female, n Adenocarcinoma, n Squamous-cell carcinoma, n NSCLC NOS, n Large-cell carcinoma, n N Mean age, years (SD) Male/female, n N Mean age, years (SD) Male/female, n N N Mean age, years (SD) Male/female, n N N

Kiel

Amsterdam

Barcelona

Bonn

PUMCH

Xuanwu

All

698 49 (17) 336/362 – – – – – – – – – – – – – – – – – – – – – – –

100 51 (13) 50/50 67 61 (10) 38/29 38/29 397 62 (12) 243/154 144 130 106 17 – – – – – – – 116 59 (13) 58/58 11 13

100 51 (7) 33/67 68 65 (12) 53/15 26/42 316 67 (11) 229/87 183 91 32 10 28 61 (13) 16/12 72 47(21) 34/38 27 100 68 (13) 50/50 4 8

41 38 (12) 4/37 35 62 (10) 23/12 14/21 44 62 (10) 29/15 15 12 16 1 11 61 (9) 3/8 74 57 (17) 29/45 6 119 63 (12) 47/72 – –

71 39 (15) 31/40 22 62 (13) 17/5 9/13 42 60 (10) 25/17 28 10 4 – 17 44 (15) 7/10 16 44 (18) 9/7 5 17 51 (14) 12/5 – 1

75 38 (12) 24/51 15 60 (8) 12/3 5/10 53 65 (11) 28/25 28 20 4 1 41 57 (14) 24/17 15 62 (12) 9/6 1 15 65 (11) 13/2 – –

1085 48 (16) 478/607 207 62 (11) 143/64 92/115 852 64 (12) 554/298 398 263 162 29 97 56 (15) 50/47 177 52 (19) 81/96 39 367 63 (13) 180/187 15 22

NOS: not otherwise specified; MCT: medullary carcinoma of the thyroid; NET: neuroendocrine tumors.

ProGRP assays (slope 0.93, Passing–Bablok regression; correlation coefficient 0.99) [36]. The ProGRP reference range measured in apparently healthy individuals on the Elecsys® ProGRP assay gave comparable results to the reference range on the ARCHITECT assay [36]. Samples from Barcelona and Bonn had slightly lower ProGRP concentrations than the other sites. ProGRP showed clear differential diagnosis between SCLC and NSCLC in serum, with 78% sensitivity at 95% specificity in the NSCLC cohort, as previously noted with the ARCHITECT ProGRP assay [16,17,

21,36]. No clinically relevant differences between sites were noted in the NSCLC and SCLC cohorts. In addition, no clinically relevant effects of age, gender or smoking habits were observed. This is the first evaluation of a ProGRP assay specifically in Chinese patients and demonstrates the reproducibility of results among different ethnicities. Although the number of patients with early-stage SCLC was small, ProGRP concentrations were higher in patients with extensive versus limited SCLC, as previously reported [25,38]. A correlation between tumor size (according to UICC stage) and ProGRP concentration was also noted. As expected, ProGRP concentrations were low in patients Healthy

n = 698

n = 100

n = 100

n = 41

n = 71

n = 698

n = 698

n = 71

n = 75

68.3 40.3

70.5 38.6

55.7 33.4

64.2 33.8

70.5 47.4

68.0 41.4

59.5 35.5

60.0 41.3

50.9 37.8

150 95%−q. median

ProGRP (pg/mL)

100

50 median li.hep. median serum median plasma

0 Kiel Amsterdam Barcelona Bonn serum serum serum serum

PUMCH Kiel Kiel PUMCH Xuanwu serum Li-Heparin K2-EDTA K3-EDTA K2-EDTA plasma plasma plasma plasma

Fig. 3. Reference range distribution of ProGRP in serum and plasma samples from healthy individuals for all sites. The yellow crosses represent the mean.

Healthy

Benign

SCLC/NSCLC

n = 21

n=1

n = 15

n=0

n = 299

n = 31

n = 33

n=6

n = 113

n = 25

n = 39

n = 58

n = 757

n = 95

n = 170

n = 37

n = 939

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395

n = 146

394

Other malignant

120000

ProGRP (pg/mL)

20000 5000

1000

200 100 50 20 10 China Europe

3 Healthy

SCLC

NSCLC

Benign lung

Other benign

Renal Other diseases malignant

MCT

NET

Fig. 4. ProGRP concentrations in different clinical cohorts in China and Europe. The yellow crosses represent the mean.

with NSCLC, benign lung disease, other benign diseases or other malignancies (excluding renal disease, MCT and NET of the lung). Because elevated ProGRP levels are only found in NET originating from the lung or from an unknown primary site [22], ProGRP may be a useful diagnostic tool in locating the primary site of unknown NETs. ProGRP showed good differential diagnosis between SCLC and NSCLC in serum from patients with eGFR ≥ 30 mL/min/1.73 m2 (CKD stage 3) at all European and Chinese sites, but care should be taken in interpreting ProGRP results in patients with eGFR b30 mL/min/1.73 m2. The clinical utility of ProGRP as a biomarker to distinguish SCLC from other lung cancers has been established in a number of studies [12,16, 20,21,39–42]. In a meta-analysis of 5146 patients enrolled in 11 clinical trials, the sensitivity and specificity of ProGRP in diagnosing SCLC was 0.716 (95% CI 0.688–0.743) and 0.921 (95% CI 0.909–0.932), respectively [40]. The clinical performance of the Elecsys® ProGRP assay is fully inline with these results. Data from this multicenter evaluation of the Elecsys® ProGRP assay demonstrate that ProGRP is a specific tumor

ProGRP SCLC vs. NSCLC

Conflicts of interest Catharina M. Korse, Xiu-yi Zhi, Xiaotong Zhang, Ling Qiu, Daan van den Broek, José M. Escudero, Jens Standop, Mu Hu and Rafael Molina have no conflict of interest to declare. Stefan Holdenrieder has received honoraria for lectures from Roche. Andrea Geistanger, Birgit Wehnl and Marcus-Rene Lisy are employees of Roche Diagnostics. Acknowledgments

1.0

Sensitivity

marker for SCLC, which supports differential diagnosis in lung cancer. The increased stability of the Elecsys® ProGRP assay in serum is a clear benefit for routine use in clinical laboratories, because serum is a preferred sample material in tumor-marker testing. An investigation into the use of ProGRP in treatment monitoring is currently ongoing, as well as measurement and analysis of other tumor markers (CEA, NSE, CYFRA 21–1) in the same collectives. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.cca.2014.09.015.

0.8

We acknowledge biostatistical support from Christine Engel. Support for third-party writing assistance for this manuscript was provided by Roche Diagnostics.

0.6

References

0.4

0.2

Europe AUC: 0.89 (0.85–0.92) China AUC: 0.94 (0.89–0.99)

0.0 0.0

0.2

0.4

0.6

0.8

1.0

1−Specificity Fig. 5. ROC curves of ProGRP (eGFR ≥30 mL/min/1.73 m2) for SCLC versus NSCLC for European and Chinese sites. Europe SCLC (n = 170), NSCLC (n = 757); China SCLC (n = 37), NSCLC (n = 95).

[1] Hensing T, Chawla A, Batra R, Salgia R. A personalized treatment for lung cancer: molecular pathways, targeted therapies, and genomic characterization. Adv Exp Med Biol 2014;799:85–117. [2] Pass HI, Beer DG, Joseph S, Massion P. Biomarkers and molecular testing for early detection, diagnosis, and therapeutic prediction of lung cancer. Thorac Surg Clin 2013;23:211–24. [3] Rosell R, Bivona TG, Karachaliou N. Genetics and biomarkers in personalisation of lung cancer treatment. Lancet 2013;382:720–31. [4] Moran C. Importance of molecular features of non-small cell lung cancer for choice of treatment. Am J Pathol 2011;178:1940–8. [5] Kulesza P, Ramchandran K, Patel JD. Emerging concepts in the pathology and molecular biology of advanced non-small cell lung cancer. Am J Clin Pathol 2011; 136:228–38. [6] Nisman B, Biran H, Ramu N, Heching N, Barak V, Peretz T. The diagnostic and prognostic value of ProGRP in lung cancer. Anticancer Res 2009;29:4827–32. [7] Kalemkerian GP, Akerley W, Bogner P, et al. Small cell lung cancer. J Natl Compr Canc Netw 2013;11:78–98.

C.M. Korse et al. / Clinica Chimica Acta 438 (2015) 388–395 [8] Zarogoulidis K, Zarogoulidis P, Darwiche K, et al. Treatment of non-small cell lung cancer (NSCLC). J Thorac Dis 2013;5:S389–96. [9] Chan BA, Coward JI. Chemotherapy advances in small-cell lung cancer. J Thorac Dis 2013;5:S565–78. [10] Harmsma M, Schutte B, Ramaekers FC. Serum markers in small cell lung cancer: opportunities for improvement. Biochim Biophys Acta 1836;2013:255–72. [11] Molina R, Auge JM, Filella X, et al. Pro-gastrin-releasing peptide (proGRP) in patients with benign and malignant diseases: comparison with CEA, SCC, CYFRA 21–1 and NSE in patients with lung cancer. Anticancer Res 2005;25:1773–8. [12] Molina R, Holdenrieder S, Auge JM, Schalhorn A, Hatz R, Stieber P. Diagnostic relevance of circulating biomarkers in patients with lung cancer. Cancer Biomark 2010;6:163–78. [13] European Group on Tumour Markers. Tumour markers in lung cancer: EGTM recommendations. Anticancer Res 1999;19:2817–9. [14] Slodkowska J, Zych J, Szturmowicz M, Demkow U, Rowinska-Zakrzewska E, Roszkowski-Sliz K. Neuroendocrine phenotype of non-small cell lung carcinoma: immunohistological evaluation and biochemical study. Int J Biol Markers 2005;20: 217–26. [15] Miyake Y, Kodama T, Yamaguchi K. Pro-gastrin-releasing peptide (31–98) is a specific tumor marker in patients with small cell lung carcinoma. Cancer Res 1994;54:2136–40. [16] Molina R, Augé JM, Bosch X, et al. Usefulness of serum tumor markers, including progastrin-releasing peptide in patients with lung cancer: correlation with histology. Tumour Biol 2009;30:121–9. [17] Molina R. ProGRP: a new biomarker for small cell lung cancer. Eur J Clin Med Oncol 2009;1:25–32. [18] Nakahama H, Tanaka Y, Fujita Y, Fujii M, Sugita M. CYFRA 21–1 and ProGRP, tumor markers of lung cancer, are elevated in chronic renal failure patients. Respirology 1998;3:207–10. [19] Inaji H, Komoike Y, Motomura K, et al. Demonstration and diagnostic significance of pro-gastrin-releasing peptide in medullary thyroid carcinoma. Oncology 2000;59: 122–5. [20] Molina R, Augé JM, Alicarte J, Filella X, Viñolas N, Ballesta AM. Pro-gastrin-releasing peptide in patients with benign and malignant disease. Tumour Biol 2004;25:56–61. [21] Molina R, Bosch X, Augé JM, et al. Utility of serum tumor markers as an aid in the differential diagnosis of patients with clinical suspicion of cancer and in patients with cancer of unknown primary site. Tumour Biol 2012;33:463–74. [22] Korse CM, Taal BG, Bonfrer JM, Vincent A, van Velthuysen ML, Baas P. An elevated progastrin-releasing peptide level in patients with well-differentiated neuroendocrine tumours indicates a primary tumour in the lung and predicts a shorter survival. Ann Oncol 2011;22:2625–30. [23] Yoshimura T, Fujita K, Kinukawa H, et al. Development and analytical performance evaluation of an automated chemiluminescent immunoassay for pro-gastrin releasing peptide (ProGRP). Clin Chem Lab Med 2009;47:1557–63. [24] Yoshimura T, Fujita K, Kawakami S, et al. Stability of pro-gastrin-releasing peptide in serum versus plasma. Tumour Biol 2008;29:224–30.

395

[25] Kim HR, Oh IJ, Shin MG, et al. Plasma proGRP concentration is sensitive and specific for discriminating small cell lung cancer from nonmalignant conditions or non-small cell lung cancer. J Korean Med Sci 2011;26:625–30. [26] Nordlund MS, Warren DJ, Laerdahl JK, Paus E. Studies on multiple forms of proGRP in serum from small cell lung cancer patients. Tumour Biol 2009;30:265–75. [27] World Health Organization. Histological typing of lung tumors. 3rd ed. Springer Verlag, Geneva: WHO; 1999. [28] Travis WD, Brambilla E, Noguchi M, et al. International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society international multidisciplinary classification of lung adenocarcinoma. J Thorac Oncol 2011;6: 244–85. [29] CLSI. Evaluation of Precision Performance of Quantitative Measurement Methods: Approved Guideline. CLSI document EP05-A2 2nd ed. Wayne, . PA: Clinical and Laboratory Standards Institute; 2004. [30] Linnet K. Evaluation of regression procedures for methods comparison studies. Clin Chem 1993;39:424–32. [31] Carpenter J, Bithell J. Bootstrap confidence intervals: when, which, what? A practical guide for medical statisticians. Stat Med 2000;19:1141–64. [32] Hahn GJ, Meeker WQ. Statistical Intervals: A Guide for Practitioners. New York: Wiley; 1991. [33] Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150:604–12. [34] Matsushita K, Mahmoodi BK, Woodward M, et al. Comparison of risk prediction using the CKD-EPI equation and the MDRD study equation for estimated glomerular filtration rate. JAMA 2012;307:1941–51. [35] DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837–45. [36] Abbott ARCHITECT ProGRP package insert, 2009. [37] Fujirebio ProGRP package insert, 2011. [38] Biran H, Perelman M, Nisman B, et al. Circulating ProGRP and tissue thyroid transcription factor 1 (TTF1) in small-cell lung carcinoma (SCLC) limited versus extensive stages. J Clin Oncol 2014;32(Suppl.):e18555 (Abst). [39] Ono A, Naito T, Ito I, et al. Correlations between serial pro-gastrin-releasing peptide and neuron-specific enolase levels, and the radiological response to treatment and survival of patients with small-cell lung cancer. Lung Cancer 2012;76:439–44. [40] Yang HJ, Gu Y, Chen C, Xu C, Bao YX. Diagnostic value of pro-gastrin-releasing peptide for small cell lung cancer: a meta-analysis. Clin Chem Lab Med 2011;49: 1039–46. [41] Tang JH, Zhang XL, Zhang ZH, et al. Diagnostic value of tumor marker pro-gastrinreleasing peptide in patients with small cell lung cancer: a systematic review. Chin Med J (Engl) 2011;124:1563–8. [42] Holdenrieder S, von Pawel J, Dankelmann E, et al. Nucleosomes, ProGRP, NSE, CYFRA 21–1, and CEA in monitoring first-line chemotherapy of small cell lung cancer. Clin Cancer Res 2008;14:7813–21.