Vol. 18 No 2 | Winter 2016
Realising the promise of precision oncology
Prof Andrew V Biankin

This article is 8 years old and may no longer reflect current clinical practice.

There is now compelling evidence that the molecular heterogeneity of cancer leads to disparate molecular phenotypes with variable disease outcomes and responses to therapy in histologically indistinguishable cancers.1 Moreover, therapy induces selection pressures in the heterogenenous tumour environment, leading to either the rapid emergence of resistant clones and/or the acquisition of mutations that confer resistance and alter the molecular phenotype of the tumour. Knowledge of the molecular phenotype has the potential to improve therapeutic selection and, hence, the early delivery of the optimal therapeutic regimen, which would improve overall outcomes and minimise treatment-related morbidity and cost by avoiding ineffective therapies. Although these advances are creating substantial opportunities for improved treatment outcomes, we are faced with significant challenges in implementing precision oncology strategies.2

We have made some spectacular advances in some cancer types – or, rather, subgroups of some cancer types. Targeted therapeutics include: anti-oestrogen therapy in oestrogen receptor-positive breast cancer; anti-human epidermal growth receptor 2 (HER2) therapy in HER2 amplified breast and gastric cancer; anti-epidermal growth factor receptor (EGFR) therapy in EGFR mutant lung cancer; imatinib therapy that targets BCR-ABL fusion positive CML and KIT mutant gastrointestinal stromal tumours (GIST); BRAF inhibitors in BRAFV600E mutant melanoma; cetuximab for KRAS wild type colon cancer; and crizotinib for EML-ALK fusion-positive lung cancer. However, we are yet to experience significant improvements in outcome for most cancer types.3 Although clinical trials of targeted therapies in unselected patients have shown positive signals in several cancer types, the improvements have generally been made in small increments, leading to a lack of adoption and relatively few approvals (for example, erlotinib in pancreatic cancer).

Large-scale genomic sequencing efforts, such as the International Cancer Genome Consortium (ICGC)4 and The Cancer Genome Atlas (TCGA)5, are unveiling marked heterogeneity and diversity in previously indistinguishable cancers and, as a consequence, it is perhaps not surprising that therapies designed to target specific molecular aberrations have not had broad success. The appreciation of this diversity, with potential gains in smaller and smaller subgroups, is putting enormous pressure on therapeutic development strategies and health systems to modify processes, from individual patient care through to regulatory approval mechanisms.

In order to advance precision oncology strategies, there is now the need to molecularly characterise both the patient and the tumour they develop in a meaningful way and to match the hypothesised right treatment to the right patient.6 In principle, this molecular marker selection of potential responders based on the known mechanism of a specific drug appears a relatively straightforward approach; however, our ever-increasing appreciation of the complexity and diversity of cancer makes this challenging to implement in drug development, let alone in routine healthcare.

Progress to date has been made where the proportion of a responsive subgroup of patients within a traditional clinical trial has been substantial enough to generate a detectable signal. For example, the proportion of BRAFV600E mutant melanoma approaches 40 per cent. In contrast, studies such as the ICGC and TCGA are showing that most mutated genes occur at a prevalence of less than five per cent. As a consequence, clinical trials would not detect a signal even if all five per cent of these patients had an excellent response to the treatment. In response, the drug development paradigm is shifting towards a need to select patients based on a marker that reflects the drug’s mechanism of action.

This presents significant logistical hurdles, particularly in ‘finding’ patients that satisfy the molecular criteria for a specific study. Take, for example, a molecular subtype that exists at two per cent (an average proportion in many cases), then with attrition owing to assay failure and for clinical reasons (each of 15 per cent) one would need to screen 78 patients to find one eligible for the study. This attracts a significant cost to drug development, but more importantly is a poor experience for an individual patient and their clinician, since 98 per cent of patients are effectively told that they are not eligible for the study and need to move on. This further delays therapy as there is a need for additional assays in order to satisfy recruitment criteria for other trials.

These challenges have seen the emergence of molecular screening programs and infrastructure in many countries. These include the US National Cancer Institute’s NCI-MATCH program and cancer-specific programs such as Pancreatic Cancer Action Network‘s Know Your Tumor program, also in the US, and SpectaCOLOR and SpectaLUNG in Europe. These programs screen patients for molecular markers and then allocate them to clinical trials. This emerging model of finding the trial for the patient, rather than the traditional model of finding the patient for the trial is gaining traction with many stakeholders.

A recent success using molecular selection of patients for treatment is evident in ovarian cancer. The standard of care for advanced disease is a platinum-taxane combination. Although the precise mechanism of action of these therapeutics is not known, they are thought to target dividing cells through damaging DNA (platinums) and cell division (taxanes).

Large-scale genomic studies have revealed that a significant proportion of ovarian cancers harbour defects in genes involved in the DNA damage response, supporting the efficacy of platinum agents in this subgroup. Based on these data, a recent study (Ariel2) used rucaparib, an inhibitor of the enzyme poly ADP ribose polymerase (PARP), which is important in the DNA damage response.

This approach exploited a therapeutic strategy called synthetic lethality. Synthetic lethality means that in normal cells the drug has minimal effect, yet in cells that have specific abnormalities, the drug combines with this defect and is effective at killing these cells. Tumours that had defects in DNA damage response (specifically homologous recombination, in most cases caused by either inherited or acquired mutations in BRCA or related genes) made these cells dependent on PARP for repairing their DNA. Inhibiting the enzyme with rucaparib increased DNA damage, killing the cells.

The Ariel2 study used a new strategy to identify women with recurrent ovarian cancer who might benefit from rucaparib treatment. In addition to mutations in BRCA1 and BRCA2, which occur in about 20 per cent of cases of high-grade ovarian cancer, Ariel2 investigated whether a molecular signature of defective DNA damage repair could be used as a predictive biomarker of sensitivity to rucaparib. The signature was based on the ability to use next-generation genomic sequencing to detect scars in tumour DNA caused by imperfect repair. The Ariel2 investigators hypothesised that tumours with large amounts of scarring would respond better than those with low levels of scarring.

More than 200 women were enrolled in the study and the results, presented at the American Society of Clinical Oncology Annual Meeting in 2015, confirmed the hypothesis. Overall, response rates were highest in tumours with mutations in BRCA1 or BRCA2, with over 80 per cent of patients responding to treatment. However, the main focus of the study was on those patients with no mutations in BRCA1 or BRCA2. Response rates were 45 per cent in those with high levels of DNA scarring; double that seen in patients with low levels of scarring (21 per cent). This is the first time that a predictive signature of response to any treatment has been successfully applied in ovarian cancer. Further validation of this DNA scarring signature is still required.

More recently, there have been dramatic successes using immune therapies to inhibit tumour mechanisms that evade immune destruction. These include checkpoint inhibitors that inhibit PD-1 and other mechanisms. Similar to other targeted agents, immune therapies will also require selection markers, such as the mutation load within tumour cells, where high mutation burdens, such as those seen in mismatch repair deficient colon cancer and melanoma, are associated with responsiveness.7

Is summary, we have made some significant advances in implementing precision oncology strategies for some cancer types (and some molecular subtypes) yet there are significant challenges ahead if we are to fully realise the promise of personalised medicine.


  1. Biankin AV, Hudson TJ. Somatic variation and cancer: therapies lost in the mix. Human Genetics. 2011; 130:79-91.
  2. Biankin AV, Piantadosi S, Hollingsworth SJ. Patient-centric trials for therapeutic development in precision oncology. Nature. 2015; 526:361-70.
  3. Chin L, Andersen JN, Futreal PA. Cancer genomics: from discovery science to personalized medicine. Nat Med. 2011; 17:297-303.
  4. Hudson TJ, Anderson W, Aretz A et al. International network of cancer genome projects. Nature. 2010; 464:993-998.
  5. Leiserson MD, Vandin F, Wu HT et al. Pan-cancer network analysis identifies combinations of rare somatic mutations across pathways and protein complexes. Nat Genet. 2015;47:106-14.
  6. Biankin AV, Piantadosi S, Hollingsworth SJ. Patient-centric trials for therapeutic development in precision oncology. Nature. 2015; 526:361-70.
  7. Le DT, Uram JN, Wang H et al. PD-1 Blockade in Tumors with Mismatch-Repair Deficiency. N Engl J Med. 2015; 372:2509-20

Leave a Reply

Your email address will not be published. Required fields are marked *