| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© The Author 2007. Published by Oxford University Press.
EDITORIALS |
Serum Proteomic Classifier for Predicting Response to Epidermal Growth Factor Receptor Inhibitor Therapy: Have We Built a Better Mousetrap?
Affiliations of authors: Department of Pathology (MST), Division of Applied Molecular Oncology (MST, GL), and Division of Medical Oncology and Hematology (FAS), University Health Network, Princess Margaret Hospital and Ontario Cancer Institute, Toronto, ON, Canada; Department of Laboratory Medicine and Pathobiology (MST), Department of Medicine (GL, FAS), University of Toronto, Toronto, ON, Canada
Correspondence to: Ming-Sound Tsao, MD, FRCPC, 610 University Ave, Rm 7-613, Toronto, ON, Canada M5G 2M9 (e-mail: Ming.Tsao{at}uhn.on.ca).
The ability to predict treatment response and clinical (survival) outcomes is one of the two "holy grails" of cancer biomarker development, the other being early detection or risk stratification by population screening (1). In advanced non–small-cell lung cancer (NSCLC), small-molecule tyrosine kinase inhibitors (TKIs) of the epidermal growth factor receptor (EGFR) represent a breakthrough for targeted therapy, yet only a small proportion of patients (especially among non–East Asians) appear to benefit from these expensive agents (2,3). Previous studies have shown that EGFR immunohistochemistry, tyrosine kinase domain mutations, and/or EGFR gene copy number by fluorescent in situ hybridization (FISH) are biomarkers that are potentially useful to select patients more likely to benefit from EGFR TKIs (4–6). Aside from the incomplete validation of these markers for routine clinical implementation, mutation and FISH analyses are limited by tissue availability and, at times, technical feasibility (5). Therefore, a strongly predictive but simple blood-based test has great potential for becoming the ultimate biomarker for selecting patients who are likely to benefit from EGFR TKI therapy.
In this issue of the Journal, Taguchi et al. (7) report a proteomic algorithm (signature) that classifies patients according to their likely outcomes after EGFR TKI therapy. This algorithm was generated by matrix-assisted laser desorption ionization (MALDI) mass spectroscopy (MS) analysis of pretreatment sera or plasma of patients. The authors used a training set of 139 patients treated with second- or greater-line gefitinib who were recruited from three separate institutions in Europe or Asia to discover eight MALDI signature peaks that form the predictive algorithm. This classifier was then validated in two independent cohorts of gefitinib- or erlotinib-treated patients from Italy and the United States. In fact, this study marks the first large-scale multi-institutional evaluation of an MS-based protein signature in the serum or plasma of NSCLC patients for predicting EGFR TKI–related clinical outcomes. The study showed that, despite variability in the source or nature of samples studied and in the spectra generated by different operators, the data preprocessing and normalization procedures were able to generate MALDI MS features that were reproducible in two independent laboratories. Furthermore, the clinical outcome classifier (which classified patients into good, poor, or undefined classes) was validated in two independent cohorts of patients receiving first- or greater-line gefitinib or erlotinib therapy. More important, the authors claim that the algorithm is predictive and not prognostic, in that its classification capability failed in three separate cohorts of patients who were not treated with EGFR TKIs. The authors conclude that their eight-peak MALDI MS algorithm might assist in the pretreatment selection of appropriate subgroups of NSCLC patients for treatment with EGFR TKIs.
The results of Taguchi et al. (7) need to be considered in the context of the limitations identified in previous studies. The first reports on the ability of serum MS proteomics to diagnose ovarian (8) and prostate cancers (9) with impressive sensitivity and accuracy generated great initial excitement and promise in cancer research (10). However, the reproducibility of the data and diagnostic profiles was questioned subsequently, leading to a general skepticism about the reliability of MS-based serum proteomics as cancer biomarkers (11,12). Based on these concerns, minimal requirements for confirmation and validation of such markers have been proposed (12). Taguchi et al. (7) should be lauded for their efforts in assessing the reproducibility of their technique and results by performing independent tests on duplicate samples at two different institutions. In fact, they also reported substantial sample- and nonsample-dependent variability, which they were apparently able to overcome by a new preprocessing and data normalization method applied in their study. It will take additional validation to prove that the authors’ assays and results are reproducible in a clinical practice setting, that is, in other laboratories (including commercial or community settings) using different equipment and personnel.
In this report, Taguchi et al. (7) did not provide the identities of the eight signature peaks used for the algorithm. In general, the lack of such information has also been a shortcoming of previous MS-based cancer biomarker reports. Some studies have identified a few such peaks as highly abundant serum proteins that represent acute-phase proteins that are released by the liver and other organs (11). Diamandis (12) has indicated that it is unlikely that such abundant proteins will succeed as cancer biomarkers. Therefore, knowing the protein identity of the signature peaks would better anchor their relevance to cancer-related biologic processes rather than to general acute-phase reactions. In addition, protein identification could allow future validation studies to be designed around enhanced or alternative technologies and protocols that improve the chances of translating these novel findings into a form in which they can be used by the clinical oncologic community. We eagerly await future reports by Taguchi et al. on the identity of their signature peak proteins.
Lastly, Taguchi et al. (7) assert that their algorithm or classifier is predictive but not prognostic (13). This conclusion is based on the inability of the algorithm to classify outcomes in patients in three independent cohorts of patients who were not treated by EGFR TKIs (control sets). However, due to the nature of using observational cohorts, the EGFR TKIs–treated and –nontreated patient cohorts [table 1 of Taguchi et al. (7)] had prominent differences in tumor stage and histology distributions, in addition to differences in the proportion of nonsmokers. Furthermore, the analysis of non–EGFR TKI–treated patients was underpowered to conclude that there was no survival benefit at all, as shown by the wide confidence intervals surrounding the hazard ratios, all of which were, interestingly, below 1. Lacking a comparison of the EGFR TKI–treated patients with control non–EGFR TKI–treated patient cohorts of similar clinical characteristics, it seems premature to conclude that the proteomic classifier is truly predictive for EGFR TKI therapy, rather than simply prognostic. Nevertheless, the fact that the algorithm could be validated in both gefitnib- and erlotinib-treated patients, even after adjustment for clinical variables, argues for at least a prognostic role of the proteomic algorithm. This is already a major finding because no such serum proteomic signature currently exists for EGFR TKI–treated NSCLC patients.
This study has opened important doors to proteomic analysis in NSCLC, but it will be up to large, adequately powered prospective studies—ideally in the context of a randomized treatment study involving EGFR TKIs in one arm—to confirm these results. In addition, one has to consider the potential ability of multiple markers, including those that are based on gene expression, tumor, and germline genetics for the most accurate way to assess treatment response. Although additional and extensive validation of the algorithm reported by Taguchi et al. (7) is necessary to confirm its predictive nature, this study represents an important milestone in the development of serum-based biomarkers for predicting NSCLC outcomes.
REFERENCES
(1) Liu G, Zhou W, Wang Z, McLeod H. Incorporating molecular oncology into prognosis. In: Prognostic factors in cancer—Gospodarowicz MK, O'Sulliban B, Sobin LS, eds. (2006) 3rd ed. Hoboken (NJ): Wiley and Sons. 314–22.
(2) Kato T, Nishio K. Clinical aspects of epidermal growth factor receptor inhibitors: benefit and risk. Respirology (2006) 11:693–8.[CrossRef][ISI][Medline]
(3) Calvo E, Baselga J. Ethnic differences in response to epidermal growth factor receptor tyrosine kinase inhibitors. J Clin Oncol (2006) 24:2158–63.
(4) Tsao MS, Sakurada A, Cutz JC, Zhu CQ, Kamel-Reid S, Squire J, et al. Erlotinib in lung cancer—molecular and clinical predictors of outcome. N Engl J Med (2005) 353:133–44.
(5) Hirsch FR, Varella-Garcia M, Bunn PA Jr, Franklin WA, Dziadziuszko R, Thatcher N, et al. Molecular predictors of outcome with gefitinib in a phase III placebo-controlled study in advanced non-small-cell lung cancer. J Clin Oncol (2006) 24:5034–42.
(6) Bunn PA Jr, Dziadziuszko R, Varella-Garcia M, Franklin WA, Witta SE, Kelly K, et al. Biological markers for non-small cell lung cancer patient selection for epidermal growth factor receptor tyrosine kinase inhibitor therapy. Clin Cancer Res (2006) 12:3652–6.
(7) Taguchi F, Solomon B, Gregorc V, Roder H, Gray R, Kasahara K, et al. Mass spectrometry to classify non–small-cell lung cancer patients for clinical outcome after treatment with epidermal growth factor receptor tyrosine kinase inhibitors: a multicohort cross-institutional study. J Natl Cancer Inst (2007) 99:838–46.
(8) Petricoin EF, Ardekani AM, Hitt BA, Levine PJ, Fusaro VA, Steinberg SM, et al. Use of proteomic patterns in serum to identify ovarian cancer. Lancet (2002) 359:572–7.[CrossRef][ISI][Medline]
(9) Petricoin EF 3d, Ornstein DK, Paweletz CP, Ardekani A, Hackett PS, Hitt BA, et al. Serum proteomic patterns for detection of prostate cancer. J Natl Cancer Inst (2002) 94:1576–8.
(10) Powell K. Proteomics delivers on promise of cancer biomarkers. Nat Med (2003) 9:980.[ISI][Medline]
(11) Diamandis EP. Analysis of serum proteomic patterns for early cancer diagnosis: drawing attention to potential problems. J Natl Cancer Inst (2004) 96:353–6.
(12) Diamandis EP. Serum proteomic profiling by matrix-assisted laser desorption-ionization time-of-flight mass spectrometry for cancer diagnosis: next steps. Cancer Res (2006) 66:5540–1.
(13) Shepherd FA, Tsao MS. Unraveling the mystery of prognostic and predictive factors in epidermal growth factor receptor therapy. J Clin Oncol (2006) 24:1219–20.
Related Articles in JNCI
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2007 99: 838-846.
J Natl Cancer Inst 2007 99: 825.
J Natl Cancer Inst 2007 99: 825.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||