Half-Century of Cancer Biomarkers: Lessons from the Past and Projections for the Future

A handful of cancer biomarkers are used widely in clinical practice, mainly for aiding in diagnosis and for monitoring therapy. These biomarkers include α-fetoprotein (AFP),4 carcinoembryonic antigen (CEA), prostate-specific antigen, and the carbohydrate antigens CA125, CA19.9, and CA15.3. Other, more specialized cancer biomarkers are used less frequently. Molecular/genomic markers will not be discussed here. An examination of the history of these biomarkers reveals that they were discovered in the mid-1960s and 1970s (AFP, CEA) or in the 1980s, thanks to the monoclonal antibody technology revolution. The success of these markers in clinical practice sparked interest to discover new cancer biomarkers that could be applied for population screening, early diagnosis, prediction of therapeutic response, monitoring therapy, etc. However, for almost 40 years, no major cancer biomarkers have been discovered and entered the clinic, despite spectacular advances in biology, medicine, genomics, proteomics, and other omic technologies. One wonders as to why this field has not progressed to expectations. The few biomarkers that obtained Food and Drug Administration approval in the last 40 years are mainly used for very specific and restricted clinical applications and are mainly genomic markers. The lack of new cancer biomarkers in the clinic is in stark contrast to the number of relevant publications, describing supposedly fantastic biomarkers for various malignancies. What happened to all these biomarkers and why have they not been clinically used? Writing on this issue in other forums, we identified 3 reasons for the failure of most cancer biomarkers to reach the clinic (1, 2). Scientific fraud (very rare). Discovery of cancer biomarkers that show statistical differences between the comparison groups but have poor characteristics such as sensitivity, specificity, and predictive value and are thus clinically useless. False discovery, which means discovery of cancer biomarkers that in the initial publication showed much promise, but failed on subsequent validation. The last category is relevant to the highly discussed issue of irreproducibility of scientific publications. It was recently realized that many scientific papers, even in the highest impact journals, fail to reproduce for various reasons, as described elsewhere (3). This issue was highlighted many years ago by us and others (4). The situation with cancer biomarkers is different when considering molecular changes such as mutations, copy number variations, deletions, insertions, etc. The advent of next-generation sequencing has revealed that a myriad of genomic changes may be related to cancer aggressiveness, progression, and response to therapy. While some of these molecular changes have clear clinical value, such as selecting targeted therapies, there is still an issue as to which molecular changes are clinically actionable (drivers) and which ones are bystanders (passengers). As more cancer genomes are being sequenced, the strongest molecular markers will likely guide future targeted therapies, something that is now known as “precision medicine.” Is there any hope that we will witness a renaissance of the classic serum circulating cancer biomarkers in the future? Or should we accept that the best performing biomarkers have already been discovered and there is nothing much to expect? Recently, we suggested that it may be possible to use the hundreds of published cancer biomarkers with poor clinical performance (e.g., low sensitivity) in isolated cases, thus introducing the concept of personalized cancer biomarkers. We speculated that although many new biomarkers exhibit very low sensitivity (at high specificity) for cancer diagnosis and monitoring, it may be possible that these biomarkers may serve a clinical purpose in small groups of patients for whom these markers are altered in the circulation (5). In one of our latest iterations of the idea of personalized biomarkers, we suggested that newly diagnosed patients may submit their serum to centralized laboratories that will screen for hundreds or thousands of biomarkers simultaneously, to identify some that may have clinical value for these specific patients (e.g., for monitoring success of therapy). In essence, our suggestion is similar to organ transplantation, whereby the donor and recipient are HLA-typed so that the most compatible organ can be selected, to avoid rejection. Our initial suggestion (5) had some limitations, especially the technology that could be used to screen quickly and relatively cheaply thousands of molecules simultaneously, in small sample volumes, to identify the ones that are probably most useful. However, recent technological developments, such as mass spectrometry or simultaneous multiplex ELISAs, may alleviate this problem. For example, it is now possible to quantify every human protein without the use of a specific reagent, such as an antibody, by using mass spectrometry and selected reaction monitoring assays. While this technology is still not sensitive enough for this application, future developments may make this a feasible approach. Additionally, there are now companies that expanded on the original Luminex multiparametric assay, to analysis of thousands of serum proteins simultaneously, using quantitative micro-ELISAs (e.g., see www.raybiotech.com). We envision, then, that the current stagnation with serological cancer biomarker discovery may be lifted by new approaches that will be based on personalized cancer biomarkers. These biomarkers could be identified after an initial screen of patient serum as described above. We caution that this approach needs experimental verification for its effectiveness. Validation of New Cancer Biomarkers: A Position Statement from the European Group on Tumor Markers Michael J. Duffy, Catharine M. Sturgeon, György Sölétormos, Vivian Barak, et al. Clin Chem. 2015;61:809–20 α-fetoprotein carcinoembryonic antigen.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.