ORIGINAL ARTICLE
Quantification of Patient-Specific Assay Interference in Different Formats of Enzyme-Linked Immunoassays for Therapeutic Monoclonal Antibodies Nicolai Grebenchtchikov, MSc,* Anneke J. Geurts-Moespot, BSc,* Linda Heijmen, MD, PhD,† Hanneke W. M. van Laarhoven, MD, PhD,‡ Carla M. L. van Herpen, MD, PhD,† Annemarie M. J. Thijs, MD,† Paul N. Span, PhD,§ and Fred C. G. J. Sweep, PhD*
Background: The use of therapeutic monoclonal antibodies for clinical purposes has significantly increased in recent years, and so has the need to monitor antibody concentrations. This may be achieved using the well-established enzyme linked immunoassay (ELISA) methods; however, these assays are subject to a variety of interferences.
Methods: In the present study, the authors have tested the ELISA methods for quantifying bevacizumab (BVZ) to investigate this interference. Three different ELISA methods were used and exhibited similar characteristics. Results: The detection limits of the ELISA methods varied from 0.05 to 0.07 ng/mL. To monitor assay performance, BVZ was measured in a control sample during each run. The BVZ concentration in the control sample was 15.4 mg/mL, the within-run imprecision (CV) and between-run CV were 4.3% and 10.4% (direct ELISA), 5.2% and 12.9% (indirect/Rabbit ELISA), and 3.9% and 9.1% (indirect/Chicken ELISA). The assays exhibited good precision and parallelism in serial dilutions of samples and a mean recovery of 98% (range, 78%–118%). Conclusions: The authors show that the degree of interference by using direct and indirect target immobilization depends heavily on the method of target immobilization on the surface of the ELISA plate, and is patient-specific. The results highlight pitfalls of potential relevance to sandwich-type assays, and an approach to rectify such problems. This approach will yield a valid assay protocol for the measurement of monoclonal therapeutic antibodies in case no target is available for direct immobilization. Key Words: ELISA, therapeutic monoclonal antibodies, interference, heterophilic antibodies (Ther Drug Monit 2014;36:765–770)
Received for publication July 5, 2013; accepted April 9, 2014. From the Departments of *Laboratory Medicine; †Medical Oncology, Radboud University Nijmegen Medical Centre; ‡Department of Medical Oncology, Academic Medical Center, University of Amsterdam; and §Department of Radiation Oncology, Radboud University Nijmegen Medical Centre, the Netherlands. The authors declare no conflict of interest. Correspondence: Paul N. Span, PhD, Department of Radiation Oncology 874, Radboud University Nijmegen Medical Centre, Geert Grooteplein 8, 6525 GA Nijmegen, the Netherlands (e-mail: [email protected]). Copyright © 2014 by Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
INTRODUCTION The use of therapeutic antibodies and their diagnostic counterparts in clinical practice has increased rapidly during the last decade.1–6 To support pharmacokinetic studies and to elucidate the underlying mechanism of drug action and interaction, an accurate, sensitive, and specific quantification of these agents is required. Enzyme linked immunoassay (ELISA) is the most widely adopted method for quantifying circulating antibodies in samples from patients treated with therapeutic monoclonal antibodies (MoAbs). However, these methods have been shown to generate falsely elevated results because of the cross-reactivity with endogenous immunoglobulins (IgG) or their fragments and by the possible presence of antimonoclonal or heterophilic antibodies.5 There are 2 ways of capturing the MoAb when using an ELISA for quantification, (1) by direct target immobilization, a carrier which the drug reacts with, or (2) by indirect target immobilization, where capture of the MoAb is achieved by a preimmobilized carrier. The nature of the carrier varies from target-specific antibodies to target-related receptors, and in some cases even a cell preparation with the target molecule exposed on the cell surface has been applied.7–11 Ideally, quantification of therapeutic antibodies could be done by an ELISA through direct immobilization because the vulnerability to interference is small. However, a specific carrier (such as vascular endothelial growth factor (VEGF) in a bevacizumab (BVZ) assay) is not always available, and an ELISA immobilizing the target indirectly would then have to be used. This type of assay is more vulnerable to interaction between the carrier and natural human antibodies and can result in false-positive signals. In the present study, the authors used BVZ ELISA methods to investigate and quantitate the degree of interference in both direct and indirect ELISA methods. BVZ (Avastin; Roche, Woerden, the Netherlands), a recombinant humanized MoAb against VEGF, is widely used in the treatment of several types of cancer, including colon, breast, lung, and renal cell cancer.12–21 For direct target immobilization, VEGF, which is commercially available, was immobilized to the wall of the microtiter plate. For the indirect target immobilization, we used polyclonal antibodies against VEGF, which earlier were raised in chicken and in rabbits for the development of a previously described VEGF ELISA.22 The aim of the present article, using BVZ as a test
765
Grebenchtchikov et al
drug, is to study the nature, frequency, and magnitude of analytical interference in blood samples of patients treated with therapeutic MoAb. The intent is that these results may be applied to other therapeutic antibodies where such interference may also be present.
Ther Drug Monit Volume 36, Number 6, December 2014
followed by incubation with tetramethylbenzidine solution for 15–20 minutes for color development. The reaction was stopped by addition of 100 mL 0.5 mol/L H2SO4. To study interfering signals, parallel ELISA methods were performed in the absence of VEGF. This format is called the VEGF-negative mode. All other conditions were kept the same.
MATERIAL AND METHODS Apparatus and Reagents Flat-bottomed low-radiation Microlon ELISA plates (art. no. 655092) were provided by Greiner Bio-One (Alphen a/d/Rijn, the Netherlands). The washing procedures were performed using a 96 PW plate washer (TecanGroup Ltd, Männedorf, Switzerland). Absorbance was measured with a Multiskan Ascent plate reader (Lab Systems, Oy, Helsinki, Finland) using a 450-nm filter. VEGF obtained from Genentech, Inc (San Francisco, CA) was used for coating in the direct ELISA and as capture in the indirect ELISA methods. As standard in the BVZ ELISA methods, Avastin was used. Prestained 3,30 ,5,50 -tetramethylbenzidine plus was provided by Kem-En-Tec (Taastrup, Denmark) and Mouse antihuman IgG (Fc) horseradish peroxidase was supplied by SouthernBiotech (Birmingham, United Kingdom). The compositions of the buffers are as follows: Coating buffer, 15 mmol/L Na2CO3 and 35 mmol/L NaHCO3, pH 9.6; Phosphate buffered saline (PBS), 140 mmol/L NaCl, 2.7 mmol/ L KCl, 1.5 mmol/L KH2PO4, and 8.1 mmol/L Na2HPO4, pH 7.4; Blocking buffer, 1% bovine serum albumin (BSA) in PBS; Washing buffer, 0.1% Tween-20 in PBS; and Dilution buffer, 1% BSA in washing buffer.
Polyclonal Antibodies Polyclonal antibodies against VEGF were raised in rabbit and chicken, purified by affinity chromatography, and subsequently diluted with glycerol (1:1) and stored in aliquots at 2208C as previously described.23,24
ELISA Procedures The procedure for the direct ELISA started with exposing microtiter plates with VEGF in coating buffer (0.50 mg/mL) overnight at 48C. Next, the plates were blocked with 1% BSA in PBS (300 mL per well, 2 hours at 378C). For the indirect ELISA methods, microtiter plates were treated with coating antibody solution, that is, rabbit antiVEGF (3.7 mg/mL) or chicken anti-VEGF (6.6 mg/mL) in 0.05 mol/L sodium carbonate buffer (pH 9.6) overnight at 48C. Next, the plates were blocked with blocking buffer (300 mL per well, 2 hours at 378C), and subsequently incubated with VEGF (0.016 mg/mL in dilution buffer) for 2 hours at ambient temperature. All incubation steps were separated by washing steps (4 times, 300 mL of washing buffer per well). To all plates (direct and indirect ELISA methods), standards and unknowns in dilution buffer were added, and the plates were incubated overnight at 48C. All further incubations were performed at ambient temperature. The next day the plates were incubated with mouse anti-human IgG (Fc) horseradish peroxidase in dilution buffer (dilution 1:25,000), and
766
Samples EDTA plasma samples from 22 BVZ-naive patients (age, 48–74 years) with advanced colorectal carcinoma were collected before treatment (t0) and during treatment combined with chemotherapy and BVZ at a dose of 15 mg/ kg of body weight. Samples were taken 1 week after the first administration (t1) and after the third administration, that is, 8–9 weeks after the start of treatment (t2). Samples were collected in accordance with protocols approved by the relevant institutional review boards, and informed consent was obtained in accordance with the Declaration of Helsinki.
Statistics For all graphic and statistical analyses, GraphPad Prism 4.0 was applied. Method comparison was performed by the Deming regression.25
RESULTS Dose–Response Curves in Bevacizumab ELISA Methods Typical calibration curves for the direct ELISA and the 2 indirect ELISA methods using chicken and rabbit antibodies are shown in Figure 1A. Concentration of BVZ ranged in the standard curves from 0.6 to 40 ng/mL. The lower detection limit, defined as the minimum BVZ concentration evoking a response significantly different from that of the zero calibrator, of the direct assay was 0.05 ng/mL and of both indirect ELISA methods was 0.07 ng/mL. To each run, a control sample was added to monitor long-term performance of the assays. The control sample was prepared from a patient treated with BVZ; plasma was aliquoted in 50 mL portions in cryotubes and stored at 2808C. The mean BVZ concentration in the control sample was 15.4 mg/mL. In the direct ELISA, the within-run imprecision (CV) of the control sample was 4.3%, and the between-run CV was 10.4%. The within-run and between-run CVs in the indirect/Rabbit ELISA were 5.2% and 12.9% and 3.9% and 9.1% in the indirect/Chicken ELISA, respectively. Three samples, used in the linearity study, were diluted 4000, 8000, 16,000, 32,000, and 64,000 times in dilution buffer. In the direct ELISA for each sample, the dilution curve was close to linear, confirming parallelism between the calibrator and the samples. No trend was observed when the dilutions were calculated to the original volume. Six samples (endogenous BVZ concentration of 0–7.0 mg/mL) were spiked with 3 concentrations of BVZ (0.2, 0.95, and 3.7 ng/mL), and BVZ concentrations were measured. The recovery ranged from 78% to Ó 2014 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
Quantification of Patient-Specific Assay Interference
118% with a mean recovery of 98%. For the indirect/Rabbit and indirect/Chicken ELISA, similar results were obtained.
Bevacizumab in Patient Samples Patient serum samples were diluted 7250 times with dilution buffer. The BVZ results obtained for each ELISA before treatment (t0), 1 week after the first injection (t1), and after the third injection (t2) with BVZ are presented in Figure 1B. In the direct ELISA, BVZ values before treatment (t0) were below the detection limit of assay. After the first injection (t1), BVZ concentration was 69.8 6 19.5 mg/mL (mean 6 SD), and after the third injection (t2), BVZ concentration was 81.7 6 29.7 mg/mL. In the indirect/Rabbit ELISA, BVZ concentrations were 59.5 6 28.3 mg/mL, 146.3 6 40.5 mg/mL, and 163.6 6 52.8 mg/mL at t0, t1, and t2, respectively. In the indirect/Chicken ELISA, BVZ concentrations were 13.1 6 8.5 mg/mL at t0, 94.3 6 25.3 mg/mL at t1, and 108.9 6 37.8 mg/mL at t2.
VEGF-Negative Mode of ELISA Methods
FIGURE 1. Three formats (direct, indirect/Chicken, and indirect/Rabbit) of BVZ ELISA. A, BVZ calibration curves. B, Concentration of BVZ in samples collected at 3 different time points (t0, t1, and t2) for all 3 ELISA formats. C, Results obtained from VEGF-negative mode for all ELISA formats. Being actually of interfering nature, these results are presented as “apparent” BVZ concentration values, calculated with the help of “genuine” BVZ calibration curve from the same ELISA plate.
Ó 2014 Lippincott Williams & Wilkins
Parallel to the 3 ELISA methods, measurements of calibrators and patients’ samples were performed in the assay types in which VEGF was omitted. Any signal obtained in this format would be due to interference. These parallel measurements were performed simultaneously in the same plate as the corresponding BVZ samples. In none of the 3 VEGFnegative ELISA-modes, dose–response signals were observed for the calibrators (data not shown), indicating that BVZ standard by itself does not evoke any interference. Blood samples of patients treated with BVZ also showed no positive signals in VEGF-negative mode of the direct ELISA. In chicken and rabbit indirect VEGF-negative ELISA format, however, the patient samples produced positive signals of various concentrations independent of whether the sample was collected before or after administration of BVZ. In the indirect/Rabbit ELISA, false-positive signals are higher than in the indirect/ Chicken ELISA. To compare the interfering signals with the BVZ concentrations, transformation of the degree of interference into “apparent concentration” values is necessary. This can be achieved by measuring patient samples in both VEGFpositive and VEGF-negative formats on the same ELISA plate and expressing the level of interference as “interfering concentration” using the corresponding BVZ standard curve from VEGF-positive mode ELISA (Fig. 1C). In the VEGFnegative mode of the indirect/Rabbit ELISA, the mean 6 SD BVZ concentrations were 54.3 6 28.3 mg/mL, 55.1 6 26.3 mg/mL, and 58.9 6 29.9 mg/mL at t0, t1, and t2, respectively. In the VEGF-negative mode of the indirect/Chicken ELISA, the BVZ concentrations were 12.1 6 7.9 mg/mL at t0, 12.3 6 7.5 mg/mL at t1, and 13.4 6 9.0 mg/mL at t2. In Figure 2, the data obtained from both indirect modes of VEGF-negative ELISA methods are ranked per patient. As can be observed, the degree of interfering concentration is patient dependent but not depending on the time of sampling, and thus of the presence of BVZ itself in the sample. The interfering concentrations in the indirect/Rabbit ELISA are substantially higher than those obtained with the indirect/ Chicken ELISA format (Fig. 3).
767
Grebenchtchikov et al
Ther Drug Monit Volume 36, Number 6, December 2014
DISCUSSION
FIGURE 2. Patient-dependent character of interference as found from VEGF-negative mode of both indirect formats of ELISA. A, Results obtained in indirect/Chicken format. B, Results in indirect/Rabbit format.
Corrections of Indirect Bevacizumab ELISA Methods The results between the direct and indirect/Rabbit ELISA methods are plotted in Figure 3A. Agreement (R2) between the BVZ concentrations measured in the direct ELISA and the indirect/Rabbit ELISA is 0.698 (slope, 1.88 6 0.20; intercept 34.4 6 11.6). An identical plot was made for the direct and the indirect/Chicken ELISA (R2 = 0.952; slope, 1.24 6 0.05; intercept, 10.2 6 2.6). Correction for the interference was made by subtraction of the interfering concentration specific for each individual patient from the BVZ concentrations measured in the indirect/Rabbit ELISA. This correction improved the agreement with the direct ELISA values significantly, as can be seen in Figure 3C (R2 = 0.952; slope, 1.14 6 0.04, intercept 1.0 6 2.4). A similar approach was applied to correct the results obtained in the indirect/Chicken ELISA (Fig. 3D). After correction, the agreement becomes R2 = 0.981; slope, 1.15 6 0.03; intercept, 20.3 6 1.5. In Figure 4, the interferences obtained in the indirect Rabbit ELISA is plotted against the interferences in the indirect Chicken ELISA. No relation was found (R2 = 0.110).
768
Assessing concentrations of therapeutic MoAbs in human samples has a potential to be distorted by falsepositive signaling derived from native human IgGs. As shown in this article for BVZ, this type of interference is patientspecific and depends heavily on the method of target immobilization in the ELISA. In the direct BVZ ELISA, where the assay plate is coated directly with pure VEGF, no interference was experimentally found, while both indirect BVZ ELISA methods showed substantial interference. Although the assay based on direct immobilization, if possible, is always preferable, the results of an indirect immobilization assay can be successfully corrected by calculating the patient-specific interference. However, the use of obviously less vulnerable direct format of target immobilization is not always possible. For some therapeutic antibodies like trastuzumab, the target (Her2/neu, Erbb2) for instance is not sufficiently commercially available, leaving no choice but to use the indirect target immobilization approach instead. From our present results, it is clear that this type of assay must be designed along with an adequate test for potential interference. In several articles using ELISA methods with indirect target immobilization, the problem of interference was acknowledged and addressed,7–11,16 but in our view no solution of the problem has yet been presented. For example, Jamieson et al developed an ELISA for the determination of trastuzumab (Herceptin, Roche) in which capture of therapeutic antibody is achieved through preimmobilized formaldehydefixed SKBR3 Her2-positive breast cancer cells. The test for potential interference, however, was not performed in an accompanying ELISA setup but in a separate cytochemical staining, in which the behavior of SKBR3 cells was compared with that of Her2-negative MDA MB 231 cells. In both cases, plasma of only 1 healthy individual spiked with trastuzumab was used. Maple et al, used commercial preparation of Her2-containing cell lysate as a source of the target, indirectly immobilized by a mixture of 2 MoAbs directed against Her2. The initially developed trastuzumab ELISA was found to be susceptible to several interference problems, such as high background, incorrect recovery of the analyte, and the presence of falsepositive results. After several modifications, including the addition of mouse IgG and 10% PEG to the assay diluent, combined with initial exposure of the samples to human IgG-coated plate, the final ELISA had become apparently less vulnerable. As stated by the author, false-positive signals still remained. Our present experimental design resulted from our aim not only to detect interference, but also to “measure” it in the most accurate way. The latter could be achieved by comparison of the results of the 2 ELISA formats, both measuring the same therapeutic antibody, but one based on direct and another on indirect target immunization. The difference between the results, found by both ELISA methods combined with possible elevation of the values obtained in the indirect immobilization ELISA, provided information of the magnitude and frequency of interference within the studied group of patients. There were several reasons to choose BVZ, a humanized monoclonal antibody against VEGF, as the subject of our Ó 2014 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
Quantification of Patient-Specific Assay Interference
FIGURE 3. Correction procedure for the results obtained in both indirect formats of BVZ ELISA. A and B, Scatter plots for uncorrected results in indirect/Rabbit format (A) and indirect/Chicken format (B) as compared with the results in direct format. C and D, Scatter plots for the corrected results.
study, including (1) that it is widely used in treatment of different types of cancer, (2) its quantification is addressed in numerous articles,17,26,27 (3) the availability of 2 wellcharacterized and affinity-purified polyclonal anti-VEGF
FIGURE 4. Relative magnitudes and comparison of both interferences found in indirect formats of BVZ ELISA. “Chicken” interference as compared with the “rabbit” interference. Ó 2014 Lippincott Williams & Wilkins
antibodies in our laboratory. Because one antibody was raised in rabbit and the other in chicken, there was an opportunity to study a class-dependent nature of the interference. The presence of substantial interference in the form of false-positive signaling for both indirect BVZ ELISA methods as compared with that obtained by direct ELISA was shown. The main difficulty in precise quantification of such interference in terms of BVZ concentrations, however, is the absence of the corresponding calibrator or ELISA calibrator. Target-negative ELISA formats, that is, ELISA sandwich constructs placed on the same ELISA plate together with the “normal” assay and treated identically in all aspects, but omitting VEGF, can be used to calculate interference. This targetnegative ELISA format served as the “interference ELISA.” The necessary calibration curve for assessing the degree of interference in this case was taken from normal BVZ assay. That resulted in “measuring” observed interference in the same units as it was provided by calibration curve. Using the interference ELISA, we were able to demonstrate that the degree of interference was patient dependent and virtually unchanged on administration of BVZ to the patients. Interference is, however, dependent on the type of antibodies used as chicken and rabbit anti-VEGF antibodies that yielded different but consistent results. Such consistency leads to the conclusion that this interference is a characteristic for each particular individual, that it has a constant nature and that by measuring this interference, the “contaminated” results can be corrected to “clean” values. The fact that all the samples collected before the treatment was started (subgroup t0)
769
Ther Drug Monit Volume 36, Number 6, December 2014
Grebenchtchikov et al
were found to be negative regarding their BVZ content after correction, serves as an additional confirmation of the legitimacy of this procedure. A noticeable additional finding was made by comparing the interferences caused by the use of rabbit and chicken antiVEGF antibodies. The average magnitude of interference caused by rabbit anti-VEGF antibodies was about 74% of the average value of BVZ, whereas that caused by chicken antiVEGF antibodies was only about 17% of the average value of BVZ. Between themselves, these interferences were found to be completely unrelated. After corrections for interference, the agreement between chicken indirect and direct assay is slightly better than that between rabbit indirect and direct assay.
CONCLUSIONS The accurate quantification of therapeutic antibodies in the current ELISA methods is known to be affected by interference from other substances in patients’ blood samples in a clinically significant manner. The present article demonstrates a method for determining the degree of this interference and a means for correcting the error. REFERENCES 1. Pavlou AK, Belsey MJ. The therapeutic antibodies market to 2008. Eur J Pharm Biopharm. 2005;59:389–396. 2. Wang W, Wang EQ, Balthasar JP. Monoclonal antibody pharmacokinetics, and pharmacodynamics. Clin Pharmacol Ther. 2008;84:548–558. 3. Reichert JM. Antibody-based therapeutics to watch in 2011. MAbs. 2011; 3:76–99. 4. Reichert JM, Dhimolea E. Foundation review: the future of antibodies as cancer drugs. Drug Discov Today. 2012;17:954–963. 5. Hoofnagle AN, Wener MH. The fundamental flaws of immunoassays and potential solutions using tandem mass spectrometry. J Immunol Methods. 2009;347:3–11. 6. Buckwalter M, Dowell JA, Korth-Bradley J, et al. Pharmacokinetics of gemtuzumab ozogamicin as a single-agent treatment of pediatric patients with refractory or relapsed acute myeloid leukemia. J Clin Pharmacol. 2004;44:873–880. 7. Jamieson D, Cresti N, Verrill MW, et al. Development and validation of cell-based ELISA for the quantification of trastuzumab in human plasma. J Immunol Methods. 2009;345:106–111. 8. Maple L, Lathrop R, Bozich S, et al. Development and validation of ELISA for Herceptin detection in human serum. J Immunol Methods. 2004;295:169–182. 9. Ternant D, Mulleman D, Degenne D, et al. An enzyme-linked immunosorbent assay for therapeutic drug monitoring of Infliximab. Ther Drug Monit. 2006;28:169–174. 10. Hampson G, Ward TH, Cummings J, et al. Validation of an ELISA for the determination of rituximab pharmacokinetics in clinical trials subjects. J Immunol Methods. 2010;360:30–38.
770
11. Li J, Li X, Tan M, et al. Two doses of humanized anti-CD25 antibody in renal transplantation: a preliminary comparative study. MAbs. 2009;1:49–55. 12. Hurwitz H, Fehrenbacher L, Novotny W, et al. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med. 2004;350:2335–2342. 13. Kabbinavar FF, Schulz J, McCleod M, et al. Addition of bevacizumab to bolus fluorouracil and leucovorin in first-line metastatic colorectal cancer: results of a randomized phase II trial. J Clin Oncol. 2005;23: 3697–3705. 14. Kabbinavar FF, Hambleton J, Mass RD, et al. Combined analysis of efficacy: the addition of bevacizumab to fluorouracil/leucovorin improves survival for patients with metastatic colorectal cancer. J Clin Oncol. 2005;23:3706–3712. 15. Saltz LB, Clarke S, Díaz-Rubio E, et al. Bevacizumab in combination with oxaliplatin-based chemotherapy as first-line therapy in metastatic colorectal cancer: a randomized phase III study. J Clin Oncol. 2008;26: 2013–2009. 16. Tebbutt NC, Wilson K, Gebski VJ, et al. Capecitabine, bevacizumab, and mitomycin in first-line treatment of metastatic colorectal cancer: results of the Australasian Gastrointestinal Trials Group Randomized Phase III MAX Study. J Clin Oncol. 2010;28:3191–3198. 17. Ternant D, Cézé N, Lecomte T, et al. An enzyme-linked immunosorbent assay to study bevacizumab pharmacokinetics. Ther Drug Monit. 2010; 32:647–652. 18. Rini I, Garcia JA, Cooney MM, et al. A phase I study of sunitinib plus bevacizumab in advanced solid tumors. Clin Cancer Res. 2009; 15:6277–6283. 19. Lang I, Inbar MJ, Kahán Z, et al. Safety results from a phase III study (TURANDOT trial by CECOG) of first-line bevacizumab in combination with capecitabine or paclitaxel for HER-2-negative locally recurrent or metastatic breast cancer. Eur J Cancer 2012;48:3140–3149. 20. An SJ, Huang YS, Chen ZH, et al. Posttreatment plasma VEGF levels may be associated with the overall survival of patients with advanced non-small cell lung cancer treated with bevacizumab plus chemotherapy. Med Oncol. 2012;2:627–632. 21. Yang JC, Haworth L, Sherry RM, et al. A randomized trial of bevacizumab, an anti-vascular endothelial growth factor antibody, for metastatic renal cancer. N Engl J Med. 2003;349:427–434. 22. Span PN, Grebenchtchikov N, Geurts-Moespot J, et al. EORTC Receptor and Biomarker Study Group Report: a sandwich enzyme-linked immunosorbent assay for vascular endothelial growth factor in blood and tumor tissue extracts. Int J Biol Markers. 2000;15:184–191. 23. Grebenschikov N, Geurts-Moespot A, De Witte H, et al. A sensitive and robust assay for urokinase and tissue-type plasminogen activators (uPA and tPA) and their inhibitor type I (PAI-1) in breast tumor cytosols. Int J Biol Markers. 1997;12:6–14. 24. Grebenschikov N, Sweep F, Geurts A, et al. ELISA for complexes of urokinase-type and tissue-type plasminogen activators with their type-1 inhibitor (uPA-PAI-1 and tPA-PAI-1). Int J Cancer 1999;81:598–606. 25. Deming WE. Statistical Adjustment of Data. Wiley, NY: Dover Publications; 1943 (edition, 1985). ISBN 0-486-64685-8. 26. Krohne TU, Eter N, Holz FG, et al. Intraocular pharmacokinetics of bevacizumab after a single Intravitreal injection in humans. Am J Ophthalmol. 2008;146:508–512. 27. Kim MJ, Han ES, Kim JW, Kim RW, et al. Aqueous Humor concentration of bevacizumab after Subconjunctival injection in rabbit. J Ocul Pharmacol Ther. 2010;26:49–53.
Ó 2014 Lippincott Williams & Wilkins
Quantification of Patient-Specific Assay Interference in Different Formats of Enzyme-Linked Immunoassays for Therapeutic Monoclonal Antibodies Nicolai Grebenchtchikov, MSc,* Anneke J. Geurts-Moespot, BSc,* Linda Heijmen, MD, PhD,† Hanneke W. M. van Laarhoven, MD, PhD,‡ Carla M. L. van Herpen, MD, PhD,† Annemarie M. J. Thijs, MD,† Paul N. Span, PhD,§ and Fred C. G. J. Sweep, PhD*
Background: The use of therapeutic monoclonal antibodies for clinical purposes has significantly increased in recent years, and so has the need to monitor antibody concentrations. This may be achieved using the well-established enzyme linked immunoassay (ELISA) methods; however, these assays are subject to a variety of interferences.
Methods: In the present study, the authors have tested the ELISA methods for quantifying bevacizumab (BVZ) to investigate this interference. Three different ELISA methods were used and exhibited similar characteristics. Results: The detection limits of the ELISA methods varied from 0.05 to 0.07 ng/mL. To monitor assay performance, BVZ was measured in a control sample during each run. The BVZ concentration in the control sample was 15.4 mg/mL, the within-run imprecision (CV) and between-run CV were 4.3% and 10.4% (direct ELISA), 5.2% and 12.9% (indirect/Rabbit ELISA), and 3.9% and 9.1% (indirect/Chicken ELISA). The assays exhibited good precision and parallelism in serial dilutions of samples and a mean recovery of 98% (range, 78%–118%). Conclusions: The authors show that the degree of interference by using direct and indirect target immobilization depends heavily on the method of target immobilization on the surface of the ELISA plate, and is patient-specific. The results highlight pitfalls of potential relevance to sandwich-type assays, and an approach to rectify such problems. This approach will yield a valid assay protocol for the measurement of monoclonal therapeutic antibodies in case no target is available for direct immobilization. Key Words: ELISA, therapeutic monoclonal antibodies, interference, heterophilic antibodies (Ther Drug Monit 2014;36:765–770)
Received for publication July 5, 2013; accepted April 9, 2014. From the Departments of *Laboratory Medicine; †Medical Oncology, Radboud University Nijmegen Medical Centre; ‡Department of Medical Oncology, Academic Medical Center, University of Amsterdam; and §Department of Radiation Oncology, Radboud University Nijmegen Medical Centre, the Netherlands. The authors declare no conflict of interest. Correspondence: Paul N. Span, PhD, Department of Radiation Oncology 874, Radboud University Nijmegen Medical Centre, Geert Grooteplein 8, 6525 GA Nijmegen, the Netherlands (e-mail: [email protected]). Copyright © 2014 by Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
INTRODUCTION The use of therapeutic antibodies and their diagnostic counterparts in clinical practice has increased rapidly during the last decade.1–6 To support pharmacokinetic studies and to elucidate the underlying mechanism of drug action and interaction, an accurate, sensitive, and specific quantification of these agents is required. Enzyme linked immunoassay (ELISA) is the most widely adopted method for quantifying circulating antibodies in samples from patients treated with therapeutic monoclonal antibodies (MoAbs). However, these methods have been shown to generate falsely elevated results because of the cross-reactivity with endogenous immunoglobulins (IgG) or their fragments and by the possible presence of antimonoclonal or heterophilic antibodies.5 There are 2 ways of capturing the MoAb when using an ELISA for quantification, (1) by direct target immobilization, a carrier which the drug reacts with, or (2) by indirect target immobilization, where capture of the MoAb is achieved by a preimmobilized carrier. The nature of the carrier varies from target-specific antibodies to target-related receptors, and in some cases even a cell preparation with the target molecule exposed on the cell surface has been applied.7–11 Ideally, quantification of therapeutic antibodies could be done by an ELISA through direct immobilization because the vulnerability to interference is small. However, a specific carrier (such as vascular endothelial growth factor (VEGF) in a bevacizumab (BVZ) assay) is not always available, and an ELISA immobilizing the target indirectly would then have to be used. This type of assay is more vulnerable to interaction between the carrier and natural human antibodies and can result in false-positive signals. In the present study, the authors used BVZ ELISA methods to investigate and quantitate the degree of interference in both direct and indirect ELISA methods. BVZ (Avastin; Roche, Woerden, the Netherlands), a recombinant humanized MoAb against VEGF, is widely used in the treatment of several types of cancer, including colon, breast, lung, and renal cell cancer.12–21 For direct target immobilization, VEGF, which is commercially available, was immobilized to the wall of the microtiter plate. For the indirect target immobilization, we used polyclonal antibodies against VEGF, which earlier were raised in chicken and in rabbits for the development of a previously described VEGF ELISA.22 The aim of the present article, using BVZ as a test
765
Grebenchtchikov et al
drug, is to study the nature, frequency, and magnitude of analytical interference in blood samples of patients treated with therapeutic MoAb. The intent is that these results may be applied to other therapeutic antibodies where such interference may also be present.
Ther Drug Monit Volume 36, Number 6, December 2014
followed by incubation with tetramethylbenzidine solution for 15–20 minutes for color development. The reaction was stopped by addition of 100 mL 0.5 mol/L H2SO4. To study interfering signals, parallel ELISA methods were performed in the absence of VEGF. This format is called the VEGF-negative mode. All other conditions were kept the same.
MATERIAL AND METHODS Apparatus and Reagents Flat-bottomed low-radiation Microlon ELISA plates (art. no. 655092) were provided by Greiner Bio-One (Alphen a/d/Rijn, the Netherlands). The washing procedures were performed using a 96 PW plate washer (TecanGroup Ltd, Männedorf, Switzerland). Absorbance was measured with a Multiskan Ascent plate reader (Lab Systems, Oy, Helsinki, Finland) using a 450-nm filter. VEGF obtained from Genentech, Inc (San Francisco, CA) was used for coating in the direct ELISA and as capture in the indirect ELISA methods. As standard in the BVZ ELISA methods, Avastin was used. Prestained 3,30 ,5,50 -tetramethylbenzidine plus was provided by Kem-En-Tec (Taastrup, Denmark) and Mouse antihuman IgG (Fc) horseradish peroxidase was supplied by SouthernBiotech (Birmingham, United Kingdom). The compositions of the buffers are as follows: Coating buffer, 15 mmol/L Na2CO3 and 35 mmol/L NaHCO3, pH 9.6; Phosphate buffered saline (PBS), 140 mmol/L NaCl, 2.7 mmol/ L KCl, 1.5 mmol/L KH2PO4, and 8.1 mmol/L Na2HPO4, pH 7.4; Blocking buffer, 1% bovine serum albumin (BSA) in PBS; Washing buffer, 0.1% Tween-20 in PBS; and Dilution buffer, 1% BSA in washing buffer.
Polyclonal Antibodies Polyclonal antibodies against VEGF were raised in rabbit and chicken, purified by affinity chromatography, and subsequently diluted with glycerol (1:1) and stored in aliquots at 2208C as previously described.23,24
ELISA Procedures The procedure for the direct ELISA started with exposing microtiter plates with VEGF in coating buffer (0.50 mg/mL) overnight at 48C. Next, the plates were blocked with 1% BSA in PBS (300 mL per well, 2 hours at 378C). For the indirect ELISA methods, microtiter plates were treated with coating antibody solution, that is, rabbit antiVEGF (3.7 mg/mL) or chicken anti-VEGF (6.6 mg/mL) in 0.05 mol/L sodium carbonate buffer (pH 9.6) overnight at 48C. Next, the plates were blocked with blocking buffer (300 mL per well, 2 hours at 378C), and subsequently incubated with VEGF (0.016 mg/mL in dilution buffer) for 2 hours at ambient temperature. All incubation steps were separated by washing steps (4 times, 300 mL of washing buffer per well). To all plates (direct and indirect ELISA methods), standards and unknowns in dilution buffer were added, and the plates were incubated overnight at 48C. All further incubations were performed at ambient temperature. The next day the plates were incubated with mouse anti-human IgG (Fc) horseradish peroxidase in dilution buffer (dilution 1:25,000), and
766
Samples EDTA plasma samples from 22 BVZ-naive patients (age, 48–74 years) with advanced colorectal carcinoma were collected before treatment (t0) and during treatment combined with chemotherapy and BVZ at a dose of 15 mg/ kg of body weight. Samples were taken 1 week after the first administration (t1) and after the third administration, that is, 8–9 weeks after the start of treatment (t2). Samples were collected in accordance with protocols approved by the relevant institutional review boards, and informed consent was obtained in accordance with the Declaration of Helsinki.
Statistics For all graphic and statistical analyses, GraphPad Prism 4.0 was applied. Method comparison was performed by the Deming regression.25
RESULTS Dose–Response Curves in Bevacizumab ELISA Methods Typical calibration curves for the direct ELISA and the 2 indirect ELISA methods using chicken and rabbit antibodies are shown in Figure 1A. Concentration of BVZ ranged in the standard curves from 0.6 to 40 ng/mL. The lower detection limit, defined as the minimum BVZ concentration evoking a response significantly different from that of the zero calibrator, of the direct assay was 0.05 ng/mL and of both indirect ELISA methods was 0.07 ng/mL. To each run, a control sample was added to monitor long-term performance of the assays. The control sample was prepared from a patient treated with BVZ; plasma was aliquoted in 50 mL portions in cryotubes and stored at 2808C. The mean BVZ concentration in the control sample was 15.4 mg/mL. In the direct ELISA, the within-run imprecision (CV) of the control sample was 4.3%, and the between-run CV was 10.4%. The within-run and between-run CVs in the indirect/Rabbit ELISA were 5.2% and 12.9% and 3.9% and 9.1% in the indirect/Chicken ELISA, respectively. Three samples, used in the linearity study, were diluted 4000, 8000, 16,000, 32,000, and 64,000 times in dilution buffer. In the direct ELISA for each sample, the dilution curve was close to linear, confirming parallelism between the calibrator and the samples. No trend was observed when the dilutions were calculated to the original volume. Six samples (endogenous BVZ concentration of 0–7.0 mg/mL) were spiked with 3 concentrations of BVZ (0.2, 0.95, and 3.7 ng/mL), and BVZ concentrations were measured. The recovery ranged from 78% to Ó 2014 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
Quantification of Patient-Specific Assay Interference
118% with a mean recovery of 98%. For the indirect/Rabbit and indirect/Chicken ELISA, similar results were obtained.
Bevacizumab in Patient Samples Patient serum samples were diluted 7250 times with dilution buffer. The BVZ results obtained for each ELISA before treatment (t0), 1 week after the first injection (t1), and after the third injection (t2) with BVZ are presented in Figure 1B. In the direct ELISA, BVZ values before treatment (t0) were below the detection limit of assay. After the first injection (t1), BVZ concentration was 69.8 6 19.5 mg/mL (mean 6 SD), and after the third injection (t2), BVZ concentration was 81.7 6 29.7 mg/mL. In the indirect/Rabbit ELISA, BVZ concentrations were 59.5 6 28.3 mg/mL, 146.3 6 40.5 mg/mL, and 163.6 6 52.8 mg/mL at t0, t1, and t2, respectively. In the indirect/Chicken ELISA, BVZ concentrations were 13.1 6 8.5 mg/mL at t0, 94.3 6 25.3 mg/mL at t1, and 108.9 6 37.8 mg/mL at t2.
VEGF-Negative Mode of ELISA Methods
FIGURE 1. Three formats (direct, indirect/Chicken, and indirect/Rabbit) of BVZ ELISA. A, BVZ calibration curves. B, Concentration of BVZ in samples collected at 3 different time points (t0, t1, and t2) for all 3 ELISA formats. C, Results obtained from VEGF-negative mode for all ELISA formats. Being actually of interfering nature, these results are presented as “apparent” BVZ concentration values, calculated with the help of “genuine” BVZ calibration curve from the same ELISA plate.
Ó 2014 Lippincott Williams & Wilkins
Parallel to the 3 ELISA methods, measurements of calibrators and patients’ samples were performed in the assay types in which VEGF was omitted. Any signal obtained in this format would be due to interference. These parallel measurements were performed simultaneously in the same plate as the corresponding BVZ samples. In none of the 3 VEGFnegative ELISA-modes, dose–response signals were observed for the calibrators (data not shown), indicating that BVZ standard by itself does not evoke any interference. Blood samples of patients treated with BVZ also showed no positive signals in VEGF-negative mode of the direct ELISA. In chicken and rabbit indirect VEGF-negative ELISA format, however, the patient samples produced positive signals of various concentrations independent of whether the sample was collected before or after administration of BVZ. In the indirect/Rabbit ELISA, false-positive signals are higher than in the indirect/ Chicken ELISA. To compare the interfering signals with the BVZ concentrations, transformation of the degree of interference into “apparent concentration” values is necessary. This can be achieved by measuring patient samples in both VEGFpositive and VEGF-negative formats on the same ELISA plate and expressing the level of interference as “interfering concentration” using the corresponding BVZ standard curve from VEGF-positive mode ELISA (Fig. 1C). In the VEGFnegative mode of the indirect/Rabbit ELISA, the mean 6 SD BVZ concentrations were 54.3 6 28.3 mg/mL, 55.1 6 26.3 mg/mL, and 58.9 6 29.9 mg/mL at t0, t1, and t2, respectively. In the VEGF-negative mode of the indirect/Chicken ELISA, the BVZ concentrations were 12.1 6 7.9 mg/mL at t0, 12.3 6 7.5 mg/mL at t1, and 13.4 6 9.0 mg/mL at t2. In Figure 2, the data obtained from both indirect modes of VEGF-negative ELISA methods are ranked per patient. As can be observed, the degree of interfering concentration is patient dependent but not depending on the time of sampling, and thus of the presence of BVZ itself in the sample. The interfering concentrations in the indirect/Rabbit ELISA are substantially higher than those obtained with the indirect/ Chicken ELISA format (Fig. 3).
767
Grebenchtchikov et al
Ther Drug Monit Volume 36, Number 6, December 2014
DISCUSSION
FIGURE 2. Patient-dependent character of interference as found from VEGF-negative mode of both indirect formats of ELISA. A, Results obtained in indirect/Chicken format. B, Results in indirect/Rabbit format.
Corrections of Indirect Bevacizumab ELISA Methods The results between the direct and indirect/Rabbit ELISA methods are plotted in Figure 3A. Agreement (R2) between the BVZ concentrations measured in the direct ELISA and the indirect/Rabbit ELISA is 0.698 (slope, 1.88 6 0.20; intercept 34.4 6 11.6). An identical plot was made for the direct and the indirect/Chicken ELISA (R2 = 0.952; slope, 1.24 6 0.05; intercept, 10.2 6 2.6). Correction for the interference was made by subtraction of the interfering concentration specific for each individual patient from the BVZ concentrations measured in the indirect/Rabbit ELISA. This correction improved the agreement with the direct ELISA values significantly, as can be seen in Figure 3C (R2 = 0.952; slope, 1.14 6 0.04, intercept 1.0 6 2.4). A similar approach was applied to correct the results obtained in the indirect/Chicken ELISA (Fig. 3D). After correction, the agreement becomes R2 = 0.981; slope, 1.15 6 0.03; intercept, 20.3 6 1.5. In Figure 4, the interferences obtained in the indirect Rabbit ELISA is plotted against the interferences in the indirect Chicken ELISA. No relation was found (R2 = 0.110).
768
Assessing concentrations of therapeutic MoAbs in human samples has a potential to be distorted by falsepositive signaling derived from native human IgGs. As shown in this article for BVZ, this type of interference is patientspecific and depends heavily on the method of target immobilization in the ELISA. In the direct BVZ ELISA, where the assay plate is coated directly with pure VEGF, no interference was experimentally found, while both indirect BVZ ELISA methods showed substantial interference. Although the assay based on direct immobilization, if possible, is always preferable, the results of an indirect immobilization assay can be successfully corrected by calculating the patient-specific interference. However, the use of obviously less vulnerable direct format of target immobilization is not always possible. For some therapeutic antibodies like trastuzumab, the target (Her2/neu, Erbb2) for instance is not sufficiently commercially available, leaving no choice but to use the indirect target immobilization approach instead. From our present results, it is clear that this type of assay must be designed along with an adequate test for potential interference. In several articles using ELISA methods with indirect target immobilization, the problem of interference was acknowledged and addressed,7–11,16 but in our view no solution of the problem has yet been presented. For example, Jamieson et al developed an ELISA for the determination of trastuzumab (Herceptin, Roche) in which capture of therapeutic antibody is achieved through preimmobilized formaldehydefixed SKBR3 Her2-positive breast cancer cells. The test for potential interference, however, was not performed in an accompanying ELISA setup but in a separate cytochemical staining, in which the behavior of SKBR3 cells was compared with that of Her2-negative MDA MB 231 cells. In both cases, plasma of only 1 healthy individual spiked with trastuzumab was used. Maple et al, used commercial preparation of Her2-containing cell lysate as a source of the target, indirectly immobilized by a mixture of 2 MoAbs directed against Her2. The initially developed trastuzumab ELISA was found to be susceptible to several interference problems, such as high background, incorrect recovery of the analyte, and the presence of falsepositive results. After several modifications, including the addition of mouse IgG and 10% PEG to the assay diluent, combined with initial exposure of the samples to human IgG-coated plate, the final ELISA had become apparently less vulnerable. As stated by the author, false-positive signals still remained. Our present experimental design resulted from our aim not only to detect interference, but also to “measure” it in the most accurate way. The latter could be achieved by comparison of the results of the 2 ELISA formats, both measuring the same therapeutic antibody, but one based on direct and another on indirect target immunization. The difference between the results, found by both ELISA methods combined with possible elevation of the values obtained in the indirect immobilization ELISA, provided information of the magnitude and frequency of interference within the studied group of patients. There were several reasons to choose BVZ, a humanized monoclonal antibody against VEGF, as the subject of our Ó 2014 Lippincott Williams & Wilkins
Ther Drug Monit Volume 36, Number 6, December 2014
Quantification of Patient-Specific Assay Interference
FIGURE 3. Correction procedure for the results obtained in both indirect formats of BVZ ELISA. A and B, Scatter plots for uncorrected results in indirect/Rabbit format (A) and indirect/Chicken format (B) as compared with the results in direct format. C and D, Scatter plots for the corrected results.
study, including (1) that it is widely used in treatment of different types of cancer, (2) its quantification is addressed in numerous articles,17,26,27 (3) the availability of 2 wellcharacterized and affinity-purified polyclonal anti-VEGF
FIGURE 4. Relative magnitudes and comparison of both interferences found in indirect formats of BVZ ELISA. “Chicken” interference as compared with the “rabbit” interference. Ó 2014 Lippincott Williams & Wilkins
antibodies in our laboratory. Because one antibody was raised in rabbit and the other in chicken, there was an opportunity to study a class-dependent nature of the interference. The presence of substantial interference in the form of false-positive signaling for both indirect BVZ ELISA methods as compared with that obtained by direct ELISA was shown. The main difficulty in precise quantification of such interference in terms of BVZ concentrations, however, is the absence of the corresponding calibrator or ELISA calibrator. Target-negative ELISA formats, that is, ELISA sandwich constructs placed on the same ELISA plate together with the “normal” assay and treated identically in all aspects, but omitting VEGF, can be used to calculate interference. This targetnegative ELISA format served as the “interference ELISA.” The necessary calibration curve for assessing the degree of interference in this case was taken from normal BVZ assay. That resulted in “measuring” observed interference in the same units as it was provided by calibration curve. Using the interference ELISA, we were able to demonstrate that the degree of interference was patient dependent and virtually unchanged on administration of BVZ to the patients. Interference is, however, dependent on the type of antibodies used as chicken and rabbit anti-VEGF antibodies that yielded different but consistent results. Such consistency leads to the conclusion that this interference is a characteristic for each particular individual, that it has a constant nature and that by measuring this interference, the “contaminated” results can be corrected to “clean” values. The fact that all the samples collected before the treatment was started (subgroup t0)
769
Ther Drug Monit Volume 36, Number 6, December 2014
Grebenchtchikov et al
were found to be negative regarding their BVZ content after correction, serves as an additional confirmation of the legitimacy of this procedure. A noticeable additional finding was made by comparing the interferences caused by the use of rabbit and chicken antiVEGF antibodies. The average magnitude of interference caused by rabbit anti-VEGF antibodies was about 74% of the average value of BVZ, whereas that caused by chicken antiVEGF antibodies was only about 17% of the average value of BVZ. Between themselves, these interferences were found to be completely unrelated. After corrections for interference, the agreement between chicken indirect and direct assay is slightly better than that between rabbit indirect and direct assay.
CONCLUSIONS The accurate quantification of therapeutic antibodies in the current ELISA methods is known to be affected by interference from other substances in patients’ blood samples in a clinically significant manner. The present article demonstrates a method for determining the degree of this interference and a means for correcting the error. REFERENCES 1. Pavlou AK, Belsey MJ. The therapeutic antibodies market to 2008. Eur J Pharm Biopharm. 2005;59:389–396. 2. Wang W, Wang EQ, Balthasar JP. Monoclonal antibody pharmacokinetics, and pharmacodynamics. Clin Pharmacol Ther. 2008;84:548–558. 3. Reichert JM. Antibody-based therapeutics to watch in 2011. MAbs. 2011; 3:76–99. 4. Reichert JM, Dhimolea E. Foundation review: the future of antibodies as cancer drugs. Drug Discov Today. 2012;17:954–963. 5. Hoofnagle AN, Wener MH. The fundamental flaws of immunoassays and potential solutions using tandem mass spectrometry. J Immunol Methods. 2009;347:3–11. 6. Buckwalter M, Dowell JA, Korth-Bradley J, et al. Pharmacokinetics of gemtuzumab ozogamicin as a single-agent treatment of pediatric patients with refractory or relapsed acute myeloid leukemia. J Clin Pharmacol. 2004;44:873–880. 7. Jamieson D, Cresti N, Verrill MW, et al. Development and validation of cell-based ELISA for the quantification of trastuzumab in human plasma. J Immunol Methods. 2009;345:106–111. 8. Maple L, Lathrop R, Bozich S, et al. Development and validation of ELISA for Herceptin detection in human serum. J Immunol Methods. 2004;295:169–182. 9. Ternant D, Mulleman D, Degenne D, et al. An enzyme-linked immunosorbent assay for therapeutic drug monitoring of Infliximab. Ther Drug Monit. 2006;28:169–174. 10. Hampson G, Ward TH, Cummings J, et al. Validation of an ELISA for the determination of rituximab pharmacokinetics in clinical trials subjects. J Immunol Methods. 2010;360:30–38.
770
11. Li J, Li X, Tan M, et al. Two doses of humanized anti-CD25 antibody in renal transplantation: a preliminary comparative study. MAbs. 2009;1:49–55. 12. Hurwitz H, Fehrenbacher L, Novotny W, et al. Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer. N Engl J Med. 2004;350:2335–2342. 13. Kabbinavar FF, Schulz J, McCleod M, et al. Addition of bevacizumab to bolus fluorouracil and leucovorin in first-line metastatic colorectal cancer: results of a randomized phase II trial. J Clin Oncol. 2005;23: 3697–3705. 14. Kabbinavar FF, Hambleton J, Mass RD, et al. Combined analysis of efficacy: the addition of bevacizumab to fluorouracil/leucovorin improves survival for patients with metastatic colorectal cancer. J Clin Oncol. 2005;23:3706–3712. 15. Saltz LB, Clarke S, Díaz-Rubio E, et al. Bevacizumab in combination with oxaliplatin-based chemotherapy as first-line therapy in metastatic colorectal cancer: a randomized phase III study. J Clin Oncol. 2008;26: 2013–2009. 16. Tebbutt NC, Wilson K, Gebski VJ, et al. Capecitabine, bevacizumab, and mitomycin in first-line treatment of metastatic colorectal cancer: results of the Australasian Gastrointestinal Trials Group Randomized Phase III MAX Study. J Clin Oncol. 2010;28:3191–3198. 17. Ternant D, Cézé N, Lecomte T, et al. An enzyme-linked immunosorbent assay to study bevacizumab pharmacokinetics. Ther Drug Monit. 2010; 32:647–652. 18. Rini I, Garcia JA, Cooney MM, et al. A phase I study of sunitinib plus bevacizumab in advanced solid tumors. Clin Cancer Res. 2009; 15:6277–6283. 19. Lang I, Inbar MJ, Kahán Z, et al. Safety results from a phase III study (TURANDOT trial by CECOG) of first-line bevacizumab in combination with capecitabine or paclitaxel for HER-2-negative locally recurrent or metastatic breast cancer. Eur J Cancer 2012;48:3140–3149. 20. An SJ, Huang YS, Chen ZH, et al. Posttreatment plasma VEGF levels may be associated with the overall survival of patients with advanced non-small cell lung cancer treated with bevacizumab plus chemotherapy. Med Oncol. 2012;2:627–632. 21. Yang JC, Haworth L, Sherry RM, et al. A randomized trial of bevacizumab, an anti-vascular endothelial growth factor antibody, for metastatic renal cancer. N Engl J Med. 2003;349:427–434. 22. Span PN, Grebenchtchikov N, Geurts-Moespot J, et al. EORTC Receptor and Biomarker Study Group Report: a sandwich enzyme-linked immunosorbent assay for vascular endothelial growth factor in blood and tumor tissue extracts. Int J Biol Markers. 2000;15:184–191. 23. Grebenschikov N, Geurts-Moespot A, De Witte H, et al. A sensitive and robust assay for urokinase and tissue-type plasminogen activators (uPA and tPA) and their inhibitor type I (PAI-1) in breast tumor cytosols. Int J Biol Markers. 1997;12:6–14. 24. Grebenschikov N, Sweep F, Geurts A, et al. ELISA for complexes of urokinase-type and tissue-type plasminogen activators with their type-1 inhibitor (uPA-PAI-1 and tPA-PAI-1). Int J Cancer 1999;81:598–606. 25. Deming WE. Statistical Adjustment of Data. Wiley, NY: Dover Publications; 1943 (edition, 1985). ISBN 0-486-64685-8. 26. Krohne TU, Eter N, Holz FG, et al. Intraocular pharmacokinetics of bevacizumab after a single Intravitreal injection in humans. Am J Ophthalmol. 2008;146:508–512. 27. Kim MJ, Han ES, Kim JW, Kim RW, et al. Aqueous Humor concentration of bevacizumab after Subconjunctival injection in rabbit. J Ocul Pharmacol Ther. 2010;26:49–53.
Ó 2014 Lippincott Williams & Wilkins
