A revolution is underway in cancer therapy and care. Now, based on identifying in a patient’s cancer those genetic alterations that drive cancer growth, use of cancer therapy is more targeted and cancer care is more personalised. Many genetic alterations are ‘passenger mutations’ in the oncogenic process, but other ‘driver mutations’ promote cancer growth and survival. These harmful genetic alterations usually result in the production of abnormal proteins such as V600-mutant BRAF in melanoma, or in the overproduction of normal proteins such as HER2 in breast cancer. In drug development, the abnormal or excess proteins are described as ‘druggable targets’, and drugs developed to selectively inhibit the function of these proteins are called ‘targeted therapies’. Since a driver mutation can be found in more than one different type of cancer, an approved and available targeted therapy can make the mutation ‘clinically actionable’. Examples of targeted therapies include the small-molecule drug, vemurafenib, for advanced V600-mutant BRAF melanoma, and the monoclonal antibody, trastuzumab (Herceptin®), for HER2-positive breast cancer. Although targeted therapies are generally considered less toxic than conventional cytotoxic chemotherapy, the toxicities may be problematic and dose limiting. However, careful clinical management of these toxicities can allow patients to continue to receive effective therapy.
The established modalities of cancer therapy are surgery, radiotherapy and chemotherapy. Immunotherapy can now be included as the fourth pillar of standard cancer therapy.1,2 Chemotherapy includes targeted therapies like small-molecule drugs and biopharmaceutical products. Whereas surgery and radiotherapy are local forms of treatment, chemotherapy and immunotherapy are systemic forms of treatment that may be used alone or with local therapy. Cytotoxic chemotherapy used alone can treat locally advanced or metastatic cancer. Chemotherapy with radiotherapy can augment the effects of radiotherapy and add to the cure of patients with locally advanced, unresectable cancers. Used pre-operatively, chemotherapy can make unresectable lesions operable or, after surgery, chemotherapy can reduce the chances of distant recurrence of a cancer. Targeted therapies for non-haematological malignancies have been approved in first, second, or third-line indications (table 1) on the basis of progression free survival and/or overall survival benefits in which control groups of patients received placebo, best supportive care or standard treatment.3–5 Hence, we expect that targeted therapy will be as versatile as cytotoxic chemotherapy in the therapeutic armamentarium against cancer.6
We define ‘targeted therapy’ operationally because the term has become so broad in scope and widely used. Targeted therapy for cancer can be defined as rationally designed therapy, which usually has a biological rationale and a predefined mechanism of action. Conversely, ‘non-targeted’ therapy such as conventional cytotoxic chemotherapy, has often been developed empirically, and the targets of drug action, such as molecules involved in DNA synthesis and replication, have been determined retrospectively.
The promise of personalised cancer medicine is to deliver the right drug to the right person at the right time. However, describing therapy as ‘targeted’ does not necessarily: (i) predict which tumour type will be most responsive; (ii) decide which patient will benefit from its use; nor (iii) determine whether toxicity either depends on tumour response or is less than that associated with cytotoxic drugs.
The scope of targeted therapy is broad and aims to encompass the complex biology intrinsic to most cancers. This complex biology is best described by the hallmarks of cancer (table 2).7 A cancer behaves as a chaotic organ comprising malignant tissue and supporting non-malignant tissues and exerting local (autocrine and paracrine) and distant (endocrine) effects. These tissues are not organised with the tight interlocking architecture typical of a normal organ.8 Rather, tissue such as tumour blood vessels is poorly and intermittently functional because the vessels are irregular, tortuous and leaky. Additional pathological features include tissue hypoxia and necrosis.9–12 Cancers attract and build a stroma of normal host cells,7 which contribute to such systemic inflammatory manifestations as cancer cachexia and illness behaviour, and to adverse prognosis and impaired metabolism of anti-cancer drugs.13,14
Table 1 shows that targeting approaches are manifold: typical points of intervention include ligand-receptor interactions and catalytic and allosteric sites of enzymes. Classes of drugs include small molecule inhibitors, competitive antagonists, and monoclonal antibodies (mAb) (figure 1). Strikingly, the kinome (the set of protein kinases in the cancer genome) representing mainly tyrosine kinases is the principal source of therapeutic targets.15 Interestingly, these kinase targets are found in malignant cells and in stromal or supporting cell types.
In spite of its manifestly complex biology, cancer is widely accepted to be a disease of somatic genetics.16 The presence of frequent genetic mutations in cancer cells and not in normal cells represents one of cancer’s differentiating features.17,18 Some of these mutations contribute to the malignant phenotype of uncontrolled growth, invasion and metastasis via gain of function of growth-promoting oncogenes or loss of function tumour suppressor genes. These mutations have been characterised as ‘driver mutations’, which in many cancer genomes exist among a sea of so-called ‘passenger mutations’. Driver mutations are often the targets of new cancer therapeutics. Stringent international cancer genomic efforts are underway both to understand the diversity of cancer genomes and the means by which driver mutations can be identified and validated.15,18,19
Nevertheless, two examples clearly demonstrate the value of finding the target and show that tumour genotype trumps clinical phenotype. Cetuximab is administered to metastatic colorectal cancer patients irrespective of the level of tumour EGFR expression. In a randomised control trial of cetuximab and best supportive care versus best supportive care alone, the survival benefit accrued only to patients whose tumours did not contain an activating KRAS mutation.20 In another randomised controlled trial of gefitinib versus chemotherapy in East Asian non or light smokers with metastatic pulmonary adenocarcinoma, patients whose tumours were EGFR mutation-positive derived the greatest progression free survival benefit from gefitinib, whereas those patients without a tumour mutation fared very poorly if they received gefitinib, rather than chemotherapy.21
Finding exactly the right target has been a protracted and difficult exercise, despite the indubitably curative effect of trastuzumab and chemotherapy as adjuvant treatment in breast cancer. In the registration study of trastuzumab for treatment of metastatic breast cancer patients, patients were randomised after HER2 positivity was determined by immunohistochemistry. However, a retrospective analysis of breast cancer samples from this study showed that the only patients who obtained a survival benefit had tumours with 3+ HER2 expression by immunohistochemistry or HER2 gene-amplification by fluorescence in situ hybridisation. Subsequently, in a consensus statement, the stringency of HER2 testing guidelines was increased so that better target definition might more accurately determine which patients could benefit.22,23 Moreover, after retrospective analyses of breast cancer samples from adjuvant samples, uncertainty remains as to whether patients with discordant HER2 results by immunohistochemistry and fluorescence in situ hybridisation benefit from adjuvant trastuzumab therapy.23
Neo-angiogenesis or new blood vessel formation is a prerequisite of tumour progression. Tumour cell production of vascular endothelial growth factor (VEGF) was identified as one of the key pro-angiogenic factors. VEGF has been targeted either by neutralising immuno-interactive molecules such as bevacizumab and aflibercept, or signalling blockade downstream of VEGF receptor(s) using a host of small-molecule tyrosine kinase inhibitors (TKIs) (table 1). Although early clinical successes offered universal promise, clinical enthusiasm for anti-angiogenic therapy has been tempered by a series of negative randomised controlled data in certain tumour types. For example, the US Food and Drug Administration (FDA) recently revoked its approval for use of bevacizumab in metastatic breast cancer after concluding that it was not proven to be safe and effective for this indication, a view subsequently supported by a Cochrane review.24 Bevacizumab added to oxaliplatin-based chemotherapy did not offer a survival advantage as adjuvant treatment for colon cancer and may be detrimental.25
In advanced non-small cell lung cancer (NSCLC), single-agent activity of VEGFR TKIs has been modest.26 Drug resistance arising from tumour cell elaboration of other pro-angiogenic factors, such as fibroblast growth factor (FGF) and platelet-derived growth factor (PDGF), indicates that specific inhibition of these signalling pathways should be incorporated in future multi-agent clinical trial designs.27
The clinical benefits of identifying an actionable target mutation are often evident. For example, the V600 BRAF mutation is the main driver of tumour growth in the ‘oncogene-addicted’ state of advanced melanoma (figure 2).28,29 V600 BRAF mutations occur in 50% of metastatic melanoma patients and the BRAF inhibitor, vemurafenib, significantly prolongs the overall survival to 13.6 months compared with 9.7 months with chemotherapy.30 Conversely, oncogene-addicted states are not so apparent in some cancers. For example, among the substantial tumour heterogeneity of 99 early-stage pancreatic cancer samples, an analysis of 16 putative driver gene mutations, some known and some novel, suggested that many different kinds of drugs may ultimately be needed for disease control.31
These clinical data emphasise the importance of prompt and correct identification of actionable tumour mutations, and are reinforced by recent phase 1 clinical trial data. Non-randomised patients whose tumours had actionable mutations, demonstrated significant improvements in tumour response rates, time-to-treatment failure (TTF) and survival if they received a matched targeted therapy (when available), compared to consecutive patients who were not given a matched targeted therapy. Indeed, in a multivariate analysis of patients with one molecular aberration, matched therapy was an independent factor predicting response and TTF.32
In 2012, the Australian Government agreed for the first time, through the Medical Benefits Schedule, to reimburse the cost of genetic tests for determining use of targeted cancer therapies. The approved items were KRAS mutation testing for use of cetuximab in KRAS wild type cases of colorectal cancer, EGFR mutation testing for use of gefitinib in cases of non-small cell lung cancer with activating mutations of EGFR, and in situ hybridisation of HER2 for cases of HER2 gene amplification in breast cancer. Other approvals are anticipated, including BRAF mutation testing for use of vemurafenib in cases of unresectable stage 3 or 4 melanoma with V600 mutations of the BRAF gene. This progress strengthens the rationale for tumour genotyping at the time that a routine histopathological diagnosis is made. However, considerable technical, drug access and reimbursement challenges remain for the delivery of truly personalised cancer medicine.33
The prerequisites of successful targeted therapy are that the target be present in the cancer and that the target limits the ability to cure the cancer.34 Targeted therapy is ineffective unless the tumour contains the target. For example, in the absence of the activating V600 mutation of the BRAF gene, metastatic melanoma patients did not respond at all to vemurafenib and died soon after.35 Hypoxia must be cure limiting for hypoxia targeting drugs combined with radiotherapy to be effective in head and neck cancer. Thus, hypoxia should be present initially and during radiotherapy and if not, then as is suspected with better prognosis human papilloma virus-related head and neck cancer, hypoxia may not be cure limiting.34
Approved targeted therapies (table 1) interdict cellular signalling pathways at different points along the signalling cascade (figure 1). For example, the antigen-binding fragment of a therapeutic mAb may neutralise the ligand (e.g. bevacizumab and VEGF), engage the receptor preventing dimerization associated with signalling (e.g. trastuzumab and HER2), blockade the receptor (eg. cetuximab and EGFR), or modulate the receptor from the cell surface (e.g. cetuximab and EGFR).36 The opposite or Fc end of a therapeutic mAb may engage elements of the host immune system to induce complement-dependent cytotoxicity, antibody-dependent cellular cytotoxicity and cellular immunity, although the contribution of these mechanisms to clinical anti-tumour effects remains uncertain.33 Similarly, small-molecule kinase or other inhibitors act at the receptor itself or at nodes in the signalling pathway where critical adaptor molecules are phosphorylated (figure 1).
Sometimes a drug does not hit the target hard enough. In spite of its specificity for the BRAF kinase, the multi-targeted kinase inhibitor, sorafenib, has low selectivity for V600-mutant BRAF.37 Sorafenib failed to exhibit significant anti-melanoma activity in patients with BRAF-mutant melanoma.38,39 In contrast, dose escalation of the highly selective BRAF inhibitor, vemurafenib, achieved sufficient signalling pathway shutdown to produce marked tumour regressions.29,35
It is almost inevitable that once a kinase inhibitor drug is applied, tumour adaptation via signalling redundancy will result in therapeutic failure (figure 1). In drug-resistant cells, autocrine, paracrine, or endocrine secretion of ligands for receptor tyrosine kinases, which transduce parallel signalling pathways, circumvents targeted kinase inhibition and enables tumour cell survival.40 Alternatively, other genetic aberrations may constitutively activate the redundant PI3K/AKT pathway. For example, metastatic colorectal cancer (CRC) patients, who harboured KRAS wild type tumours and who were given anti-EGFR mAb therapy, suffered worse outcomes if their tumours also contained PIK3CA exon 20 mutations,41 or PTEN loss.42 In a metastatic melanoma patient treated for seven months with vemurafenib, a single progressing lesion contained two sub-clones, both V600E BRAF-mutant, but one with an activating NRAS mutation.19 The NRAS mutation in effect overrides mutant BRAF signalling.
Genomic studies of primary and metastatic lesions show that tumours of the same type evolve differently over time between patients, but also between metastases within the same patient.43,44 This tumour heterogeneity is at the root of primary and acquired drug resistance and of the differential responses that a particular tumour type makes to the same treatment. Importantly, acquired drug resistance may derive from mutational or non-mutational mechanisms or both.40,45
Tumour heterogeneity limits the ability of genomic approaches to capture therapeutically relevant information because tumour sampling: (i) is invasive and sometimes not feasible or adequate because of practical, clinical and logistical factors; (ii) may not be contemporaneous to the clinically significant disease process; (iii) may derive from an uninformative part of a tumour deposit, which does not contain clinically significant driver mutations.46 Furthermore, some genomics technologies may not be sufficiently sensitive to detect low frequency primary resistance mutations or secondary mutations arise under the selection pressure of treatment.4
It also clear that the same mutation in a different tumour may produce an unexpected response to treatment. For example, selective targeting of V600E-mutant BRAF with vemurafenib produces responses in most melanoma patients, but not in colorectal cancer patients. BRAF-mutant colorectal carcinoma cells express EGFR unlike BRAF-mutant melanoma cells. Consequently, applying vemurafenib to BRAF-mutant colorectal carcinoma cells immediately results in activation of EGFR and Ras and produces BRAF inhibitor resistance, which can be overcome by concurrent anti-EGFR therapy.47
Following the paradigm established by combination cancer chemotherapy and highly active anti-retroviral therapy that suppresses HIV escape mutants, the obvious implication of tumour heterogeneity and tumour adaptation to therapy is that targeted therapies should be combined and/or incorporated in multimodality therapy.
As figure 3 shows, single targeting of an oncogenic signalling pathway can result in ‘oncogene bypass’ as a resistance mechanism.48,49 In vitro studies have shown that pharmacological blockade of these alternative pathways can restore drug sensitivity to resistant BRAF-mutant melanoma cells.50 However, this example of parallel pathway targeting risks greater toxicity, as was observed in renal cell carcinoma patients with the multi-targeted tyrosine kinase inhibitor (TKI), sunitinib, versus the more selective pazopanib.51 Another successful strategy is vertical pathway targeting (figure 1).5 When inhibitors of BRAF and MEK were combined in patients with BRAF-mutant metastatic melanoma, progression free survival was extended by approximately 50% compared to BRAF inhibitor alone.52 In an unprecedented example, this combination therapy was less toxic than each individual therapy because BRAF inhibitor-induced paradoxical MAPK pathway activation was blunted.53
Therapeutic index is important in the evaluation of any new therapy. The relatively narrow therapeutic index of many cytotoxic drugs may encourage the perception that targeted therapies are safer because ‘targeting’ could imply that their anti-cancer activity is more discriminating. However, targeted therapies have been associated with life-threatening, catastrophic and fatal adverse events. A meta-analysis of randomised clinical trials comparing the mTOR inhibitors, everolimus and temsirolimus, to controls, indicated that mTOR inhibitors doubled the risk of death, with infection being identified as a significant cause.54 Similarly, another meta-analysis of randomised clinical trials of VEGFR TKIs showed twice the risk of death from active therapy.55 The anti-VEGF mAb, bevacizumab, has been responsible for fatal haemorrhage and gastrointestinal perforation.56 The mAb, ipilimumab, blocks the T-cell surface molecule, CTLA-4, thus removing the brake on expansion of lymphocytes that subsequently target melanoma as well as normal tissues. In the first phase 3 trial, ipilimumab was associated with a 2% rate of death from autoimmune pathology.2 However, after implementation of improved toxicity management in the subsequent phase 3 trial, no ipilimumab-related deaths were reported.1
Even though serious toxicities of targeted therapies tend to occur in less than 10% of patients, chronic, low-grade toxicities also hamper quality of life for some patients. For example, although sorafenib extends overall survival among patients with advanced hepatocellular carcinoma, the typical best response of stable disease must be reconciled with frequent drug-related toxicities of fatigue, anorexia, weight loss, diarrhoea and hand-foot syndrome.57
As the same signalling pathways are inhibited in normal tissues as in tumour tissue, many targeted therapy toxicities are on-target and correlate positively with anti-tumour efficacy. Indeed, hypertension is an on-target or mechanism-related toxicity of the VEGFR-targeted agent, sunitinib, and serves as a biomarker of tumour response in metastatic gastrointestinal stomal tumour and renal cell carcinoma.58,59 Acneiform rash is a very common toxicity of EGFR inhibitor therapy, and may be severe enough to warrant treatment discontinuation in up to 10% of patients. Given the evidence that rash may be associated with improved tumour control or survival,60,61 dose interruption or reduction, and appropriate supportive measures may be sufficient to manage this EGFR inhibitor-related toxicity while treatment continues.62 The question of equipoise between toxicity and perceived or actual clinical benefit will be more important as the duration of targeted therapy increases, because of adjuvant therapy or maintenance therapy programs with consequently increased risks of cumulative toxicity.63
VEGFR-targeted kinases such as sorafenib, sunitnib, pazopanib and axitinib are particularly effective single agents in advanced cases of the clear cell variant of renal cell carcinoma, because the disease depends on VEGF-driven angiogenesis.64 In the COMPARZ phase 3 study of non-inferiority design, pazopanib was compared with sunitinib as first-line treatment in metastatic renal cell carcinoma patients. The progression free survival was similar, but pazopanib-treated patients experienced fewer troublesome side-effects and an increased quality of life. Some of the most troublesome side-effects, such as fatigue and hand foot syndrome, occurred more frequently with sunitinib.51 As a less selective TKI than pazopanib, sunitinib’s inhibition of the critical haematopoietic growth factor receptors, KIT and FLT3 among others, may contribute to its poorer tolerability,65 Consequently, this example illustrates that toxicity associated with broader inhibition of redundant pathways is generally greater and thus represents off-target toxicity since the intended target is VEGFR.
Aside from the evident clinical benefits of disease control, the shrinkage or stabilisation of disease afforded by targeted therapies may provide a ‘therapeutic platform’ for concomitant use of emerging and promising immunotherapies, which have a slower onset of action but more durable effects than some targeted therapies.
Despite application of the same cancer treatment, intra-patient and inter-patient tumour heterogeneity help to explain the wide disparity in patient outcomes. Tumour heterogeneity also highlights the shortcomings of current genomic technologies in identifying the cancer treatment targets that limit cure. Graphically, a biopsy sample taken from a single tumour mass at a single time point is unlikely to represent the diverse tumour genomic landscape and may not reliably guide the choice of therapy. Hence, other emerging clinical investigative modalities such as functional cancer imaging, which can apprehend tumour heterogeneity in the whole patient, may complement cancer genomic information to improve the accuracy of personalised cancer medicine.66,67
Future targeted therapy approaches are likely to include more mAb armed with potent toxins or radionuclides,68,69 cell and gene therapeutics,70 and novel therapeutic molecules such as small interfering RNA,71 and stapled peptides that may make protein-protein interaction targets pharmaceutically tractable.72
Although unselected patients seen in routine clinical oncology practice may differ significantly from patients enrolled in the pivotal control clinical trials,63 prompt and sure-footed management of toxicities continues to be essential to maintaining patients on an effective dose and schedule of a targeted therapy. These toxicities also indicate the ongoing need to identify those biomarkers allied with toxicity and tumour response. As cancer incidence tends to increase as populations age worldwide, targeted therapies will be increasingly used in older people who often receive polypharmacy, with an enhanced risk of drug-drug interactions and drug-related toxicities.73 Therefore, work on providing electronic point of care services may help to mitigate the risks of polypharmacy. Ultimately, however, the targeted therapies boom will oblige clinical oncology professionals to obtain new skills to control cancer.74
MPB received funding support from NHMRC Project APP1010386, Therapeutic Innovation Australia, and the Australian and New Zealand Melanoma Trials Group.