Translating metastasis-related biomarkers to the clinic
Successes and failures of preclinical models in metastasis research
- Date : 28 Feb 2013
- Topic : Translational research
Among the various hallmarks of cancer, metastasis formation is the most direct threat to survival from cancer. In the context of metastatic disease, from a clinical standpoint, the only major advance in the past 20 years has been the development of bone metastasis-targeting bisphosphonates and RANK-ligand inhibitors. Preclinical models have been used primarily to decipher different steps of the metastatic cascade. Numerous molecular processes operate in these model systems, but none of these has been successfully translated to the clinic. This crossover failure of research to the clinic might be due to several issues on the clinical side, such as the possible redundancy between molecular pathways, a lack of efficient drugs, or poor clinical trial designs.
In the perspectives article, published online 5 February 2013 in the Nature Reviews Clinical Oncology, a group of researchers led by Dr François-Clément Bidard of the Department of Medical Oncology, Institut Curie and Université Paris Descartes, France discuss some of the potential avenues of research including the use of adequate statistical methods and well-annotated cohorts in biomarker discovery; an objective assessment of the level of evidence provided by each biomarker; the development of robust molecular or cellular surrogates of metastasis in patients; and original designs for clinical trials.
Everyday clinical oncology still relies on the assumption that metastatic risk is estimated by primary tumour size and regional lymph-node status, among other criteria. A very limited integration between biological and clinical data is a major issue. This perspectives article offers some suggestions for improving the translation of biological results into the everyday clinical routine. Starting the discussion on biomarkers of metastasis onset, the authors further cover various issues related to transcriptome, genome, microRNA, biomarkers of metastatic site, surrogates of residual disease, disseminated tumour cells (DTCs), circulating tumour cells (CTCs), circulating tumour DNA (ctDNA), and microarrays, and discuss about new tools, biological and clinical validity, better preclinical models in order to move forward in improving metastasis research. In this news we focus on their view on anti-metastasis drug development. The full article from the Nature Reviews Clinical Oncology you can read through the ESMO Scientific Journal Access programme.
The entire history of adjuvant treatment has been predicated on drugs that are not specific to the metastatic process. Nonetheless, such studies have helped to distinguish between prognostic and predictive biomarkers. Moreover, the adjuvant setting has several characteristics: first, most patients are generally cured by the loco-regional treatment (mostly surgery) alone, so are 'overtreated' when receiving adjuvant treatments; second, current adjuvant treatments are effective, although incomplete; third, the optimal treatment length is unknown; and fourth, the metastatic outcome is a mix of prognostic factors, predictive factors for the response to adjuvant treatments and, as already discussed, the length of follow-up. Bypassing these obstacles might require changes in current drug development strategies.
For a potential antimetastatic treatment (directed against cancer cell adhesion, migration, homing), it is conceivable to use the classic framework of phase I trial design in metastatic patients to determine the toxicity and maximum-tolerated dose. An association of the investigated drug with the currently used adjuvant treatment should be also studied in phase I trials, as new drugs are likely to be used together with the current standard adjuvant treatments. The overall toxicity of antimetastasis drugs should be kept as low as possible, as the effect on patients without metastases (who represent ultimately the population for which antimetastatic treatments are designed) is directly associated with treatment toxicity. Depending on the drug mechanism of action, indirect signs of drug activity might be investigated during phase I trial, using dynamic tumour or blood biomarkers related to metastasis, such as CTC.
Next, the classic phase II design needs to be revisited when considering an antimetastasis drug. Phase II studies in oncology are generally conducted in patients with metastatic disease and the antitumour treatment efficacy is assessed by the RECIST criteria, that is, changes in the diameters of tumour metastases. However, no clear efficacy on the cancer burden is expected from an 'antimetastasis' drug in patients that have disseminated disease. Most non-metastatic patients do not have metastatic relapse after the primary treatment; therefore, it would be pointless to expect any significant increase in the metastasis-free survival for most cancer types in a phase II study. Two strategies could then be used: enrichment might select only those patients at high risk of early metastatic relapse. However, a randomised trial would be necessary to demonstrate the efficacy of the tested drug and would require hundreds of patients. Alternatively, a second option is to design phase II studies relying on biological surrogates of metastasis relapse that are quickly assessed, such as DTC or CTC, as well as ctDNA levels or serum protein levels, either for patient selection or as a study end point. However, using these biomarkers for assessing drug efficacy requires reproducible and trustable tests. Another suggested approach might be to include metastatic patients with a limited tumour burden, using the time to a new metastasis as an end point.
For prognostic or predictive biomarkers with unknown characteristics, and particularly for the metastasis-related biomarkers that deliver continuous values with no defined cutoff (such as any score, protein expression level, and number of CTC), using an adaptive design for a phase II study is a convenient way to investigate both the treatment activity and the clinically relevant threshold of the used biomarker. A simple example of such a trial in the metastasis-related field is the upcoming CirCe T-DM1 trial. This trial was designed at the ECCO-AACR-EORTC-ESMO Workshop on Methods in Clinical Cancer Research (“Flims Workshop”) in 2011 and will test the efficacy of T-DM1, an anti-HER2 treatment in patients with metastatic breast cancer who have a HER2-negative primary tumour, but with HER2-amplified CTC. The second phase of the CirCe T-DM1 study will include patients with a number of HER2-amplified CTC optimised towards treatment efficacy according to the first phase result. This study design should be able to report both treatment efficacy and the best biomarker cutoff, with good statistical power. This kind of design could be adapted to any other continuous biomarker. Finally, for the drugs that might act directly on the primary tumour (and not on disseminating individual cells), other phase II designs could be set up in non-metastatic patients during a short time course (such as 2–3 weeks) before primary tumour treatment, which represents a harmless 'window of opportunity' to study the biological changes induced by the treatment (such as, epithelial–mesenchymal transition reversal, and redifferentiation).
For regulating agencies to approve a metastasis-targeted drug, phase III trials would include a classic randomised design in non-metastatic patients with a long follow-up, to determine metastasis-free survival and overall survival. Patients might be also selected, or stratified for the randomisation, on the basis of any relevant biomarker and/or clinical estimate of the metastatic risk. An example is the CEREBEL trial, which compared lapatinib and capecitabine to trastuzumab and capecitabine in HER2-positive metastatic breast cancer patients with no brain metastasis at inclusion. Despite its negative results, this trial was the first 'metastasis-prevention' trial to be conducted in a large cohort, and paves the way to future similar trials. An adaptive trial design might be also considered here if several biomarkers are used in relation to several treatments, such as in the BATTLE trial. In this adaptive trial, patients were assigned to different treatments based on the results of biomarker analysis. This technique of adaptive allocation allowed patients to be paired with the therapies that were most likely to benefit them, according to the treatment efficacy observed in the previously included patients.
When considering phase III trials, clinical researchers have to consider some of the issues raised by the failure of the previous phase III trials, especially since biological processes are often compensated for by biological redundancy, and blocking only one target is likely to be insufficient. All of these considerations should be taken into account in any future 'antimetastasis' drug development.
Dr Bidard received a fellowship from the Fondation pour la Recherche Nuovo-Soldati, and Thiery JP is supported by core grants from A*STAR and School of Medicine, National University of Singapore. The authors declared no competing interests.