Originally published as JCO Early Release 10.1200/JCO.2009.22.4337 on May 4 2009

This Article Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Google Scholar Articles by Duncan, M. W. PubMed PubMed Citation Articles by Duncan, M. W. Related Articles Related Article Social Bookmarking                            What's this?

Place for Biochemical Markers in Early-Stage Lung Cancer Detection?

Department of Medicine, Division of Endocrinology, Metabolism and Diabetes, University of Colorado Denver, Aurora, CO

A recent publication by Yee and colleagues¹ provides an opportunity to reflect on how we discover blood-borne biomarkers, validate them, and establish criteria for their use in routine clinical practice. As of yet, no blood test has made a significant impact on lung cancer diagnosis, but that will change. In the meantime, Yee et al would likely argue that we demand too much of our candidate markers, and while no marker has transformed diagnosis, one or more offering suboptimal specificity or sensitivity could be part of our existing armamentarium. It is an intriguing idea.

The work by Yee et al¹ is notable for several reasons. First, the group adopted an unconventional discovery strategy. Typically, cancer-specific differences are measured in the systemic circulation and the analytic challenge is to define quantitative differences that distinguish cases from controls in the presence of a "complex and variable" chemical background. It is a demanding task that often requires impractically large sample sets to account for patient heterogeneity and the mediocre precision of our discovery tools. Alternatively, Yee et al used each subject as their own control, and determined elevated proteins in the pulmonary venous effluent draining the tumor vascular bed relative to matched systemic arterial blood. Because the chemical background is essentially constant, the task is reduced to determining differences between two samples drawn from the same individual pre- and postpassage through the affected organ.

Initially they paired pulmonary venous-radial arterial blood samples from 16 lung cancer patients and identified two candidate biomarkers by surface-enhanced laser desorption ionization mass spectrometry—connective tissue-activating peptide III/neutrophil activating protein-2 (CTAP III/NAP-2) and haptoglobin. Both were higher in venous versus arterial blood.

Given that extraordinarily encouraging findings can arise through chance or bias alone, verification is a critical next step in the biomarker development pipeline. They verified their initial findings in 64-paired venous-arterial blood samples by immunoassay—a more practical, precise, and cost-effective analytic approach. In subsequent validation studies, elevated levels of CTAP III/NAP-2 and haptoglobin were confirmed in 28 pre- and postsurgical resection peripheral blood samples, two blinded sets of plasma from 149 participants in a lung cancer screening study (49 lung cancers and 100 controls), and in 266 participants from the National Heart, Lung, and Blood Institute Lung Health Study (45 lung cancers and 221 matched controls).

CTAP III/NAP-2 levels were elevated in lung cancer and in the plasma of smokers who subsequently developed lung cancer, and in some instances levels predated clinical diagnosis by up to 29 months. Levels decreased after tumor resection (P =.01), except in a subset of patients who later developed recurrent disease. It seems that lung cancer could be detected at the preinvasive or early invasive stage because 47% of the validation cohort had stage 0/IA lung cancer.

Findings for haptoglobin were less compelling: significantly elevated levels were found in lung cancer patients, but levels did not change significantly after surgery.

Alone, the two markers offered less than optimal sensitivity and specificity. Yee and colleagues therefore suggest that they be used together with clinical characteristics, such as age, lung function, and smoking status, and, in this manner, they report that they can predict the presence of lung cancer with a positive predictive value of 63% and a negative predictive value of 89%. The incremental benefit offered by these blood biomarkers is uncertain, but is now being assessed.

Yee and colleagues offer a well-reasoned approach to biomarker discovery and make a promising start to validating their candidate markers. Surprisingly, only two candidates were identified and investigated in detail. Further studies of these in additional cohorts are required, but perhaps more importantly, replication of these findings in independent laboratories is mandatory.

This study highlights the obvious—whether our tools are genomic, epigenomic, or proteomic methods, it is not trivial to discover and validate diagnostic biochemical markers. A slew of imprecise discovery tools have been used in poor or underpowered studies, and validation has often been ignored, leaving us awash in unsubstantiated candidates. By contrast, Yee and colleagues employ an alternative to the usual needle-in-a-haystack approach to discovery and make a solid attempt at validating their markers in independent cohorts, but more needs to be done. Their candidates are not optimal stand-alone markers of early-stage disease, and were therefore incorporated into a multimodal lung cancer risk prediction model. They offer modest additional value to the diagnostic matrix, but their full impact will only be established over time.

It is also important to put these findings into a clinical context. High specificity is the primary requirement of any biochemical diagnostic test for lung cancer, especially given the poor selectivity of imaging approaches. Yee et als receiver operating characteristics curve incorporates age, sex, forced expiratory volume in 1 second, haptoglobin, and CTAP III/NAP-2 values into a multimodal lung cancer risk prediction model. Based on a high-risk population, at a specificity of 90%, the predictive sensitivity of the model is approximately 45%. In other words, at what might be considered the minimum acceptable specificity (90%; ie, where 10% of patients would be incorrectly classified or overtreated), less than half of all cases would be detected. On this basis, the clinical utility of the approach as it currently stands is questionable.

It is easy to be critical of any diagnostic biomarker study. No single study can realistically be expected to discover, validate, and accurately define the clinical utility of a candidate marker. Major methodological or study design enhancements (such as those adopted here) may facilitate discovery, but there is no substitute or shortcut for extensive real-world testing of candidate biomarkers. The passage from discovery to routine application is painfully protracted, and several decades of experience with established markers, such as carcinoembryonic antigen, demonstrate that independent testing in multiple centers, and across disparate populations, is the only way to establish the appropriate context in which to use a biochemical marker.

Yee et al would have us question our expectations of biochemical markers. Tests based on a single modality may not deliver the exacting performance criteria we currently demand, even if several analytes are measured simultaneously. That seems true for these two candidates, and will likely hold for any biochemical approach based on just one to two markers. The marked heterogeneity of lung cancer makes it highly unlikely that the full spectrum of this disease can be reflected in a couple of proteins. Incorporating biochemical, clinical, imaging, and other data into a diagnostic algorithm that weights each factor and maximizes clinical utility may be one way around this.

Yee and colleagues have provoked thought and made a good start in this complex area. Hopefully, others will adopt similarly innovative, yet more comprehensive approaches to discovery, unearth additional diagnostic markers, and then cooperatively subject their candidates to rigorous testing. If we hold to realistic expectations and are prudent in our approach to these studies, the ultimate payoff will be substantial.

AUTHOR'S DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

REFERENCE

1. Yee J, Sadar MD, Sin D, et al: Connective tissue-activating peptide III: A novel blood biomarker for early lung cancer detection. J Clin Oncol 27:2787–2792, 2009.[Abstract/Free Full Text]

CiteULike    Complore    Connotea    Del.icio.us    Digg    Facebook    Reddit    Technorati    Twitter    What's this?

Related Article

• Connective Tissue-Activating Peptide III: A Novel Blood Biomarker for Early Lung Cancer Detection
  John Yee, Marianne D. Sadar, Don D. Sin, Michael Kuzyk, Li Xing, Jennifer Kondra, Annette McWilliams, S.F. Paul Man, and Stephen Lam
  JCO 2009 27: 2787-2792 [Abstract] [Full Text]

This Article Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Email this article to a colleague Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Save to my personal folders Download to citation manager Rights & Permissions Google Scholar Articles by Duncan, M. W. PubMed PubMed Citation Articles by Duncan, M. W. Related Articles Related Article Social Bookmarking                            What's this?