Personalized medicine and economic evaluation in oncology: all theory and no practice?

Abstract

The clinical definition of personalized medicine (PM) is closely related to that of pharmacogenomics. Ideally, PM could lead the pharmaceutical industry to differentiate products by subgroups of patients with the same pathology and find new gene targets for drug discovery. Here, we focus on the potential impact of PM on the design of clinical trials and economic evaluations limited to oncology (its first and main field of application). Then, we assess the European economic evaluations focused on trastuzumab and cetuximab, the two drugs usually mentioned as emblematic examples of targeted therapies. Clinical results of PM in oncology have not been as encouraging as hoped so far. Of course, economic evaluations on targeted therapies cannot help overcome the lack of clinical evidence for most of them. The two paradigmatic examples of cetuximab and trastuzumab indicate that the methodological implications on economic evaluations debated in the literature are more theoretical than practical.

Keywords:

• cetuximab
• economic evaluation
• oncology
• personalized medicine
• targeted therapy
• trastuzumab

Personalized medicine (PM) is a much discussed topic at present and serves as an aspiration to industry to support the development of innovative drugs that will be more effective, safer and, of course, command premium prices. But what is PM? What opportunities and challenges does it offer in clinical development and economic evaluation?

Defining PM

Increasing numbers of articles on PM have been published. Four dedicated journals were launched recently [1–4], a European Association was founded in 2009 [5] and a leading US academic institution ‘curates knowledge about the impact of genetic variation on drug response for clinicians and researchers’ [6].

Although a bibliographic attempt to support a unique definition of PM has been published [7], various competing definitions persist in the literature [8], using somewhat different nomenclature (e.g., ‘stratified’/’individualized’ instead of ‘personalized’). All definitions, however, share the use of some form of genetic testing aiming to identify and target specific characteristics of the patient being treated, to give ‘the right drug to the right patient’ to maximize the effectiveness and safety of the treatment [9]. Thus, ‘personalization’, based on molecular investigation, is conceptually different from ‘personalization’ tailored to the patient’s preferences and choices, which is a common feature of good quality healthcare services [10]. The clinical definition of PM is closely related to that of pharmacogenomics, that is, the effects of an individual’s genetic variability on drug response [11]. The desired result of pharmacogenomics testing is to maximize efficacy and/or safety by matching the best available drug or dose to an individual’s genomic profile. The EMA has issued various scientific guidance documents on pharmacogenomics since 2002 and appointed a specific working party that gives recommendations on this subject [12].

The first notion of the importance of pharmacogenetic variance dates back to the Second World War and referred to American soldiers who were more or less prone to malaria depending on their ethnicity [13]. However, the first identified use of the term PM in the context of ‘the right drug to the right patient’ was in an article on patients with cancer in 1999 [9]. Many publications followed [14], focusing on the advances that the human genome project and its technology spin-offs could bring in drugs discovery and development for the pharmaceutical industry [15]. Pharmacogenomics claimed that major new drug advances would come from understanding the genetic basis of diseases. Although there is still a chasm between identifying a genetic susceptibility and developing safe and effective medicines [16], the pharmaceutical industry increasingly recognized that ‘one-size-fits-all’ is not the only strategy, particularly when it is at least in part broken with many failures in late-stage development, and it had instead to consider the potential for ‘tailored’ medicines, which will cure patients according to their individual genetic makeup. This may have been fuelled by the very high prices that can be achieved for effective medicines in oncology. In marketing terms, research-oriented companies started switching from simply attempting to identify ‘block-busters’ to recognizing the value of ‘niche-busters’ [17].

Understanding the role of genes is the PM cornerstone. The exploration of genes involved in different pharmacological responses may potentially be used to identify responders, nonresponders and patients at higher risk of side effects, which qualitatively affect the balance of benefits and risks. Accordingly, PM has the potential to play an important role in several areas [18,19], particularly for a better definition of a disease and consequently the identification through genetic markers of subtypes of patients as specific targets. The basic idea with this kind of investigation is the possibility of diagnostic, prognostic and therapeutic strategies precisely tailored to each patient’s requirements [20]. Hence, the latter-day shift from the expression ‘personalized’ to ‘precision’ medicine announced by the US President as one of the main challenges that scientific research will have to face. Ideally, PM could lead the pharmaceutical industry to: differentiate products by subgroup of patients with the same pathology, and find new gene targets for drug discovery. Patients should benefit from better health through: the higher probability of the desired outcome with a drug, and the lower probability of untoward side effects [21].

Evaluating PM

Here, we focus on PM limited to oncology (its first and main field of application) [16,20], that is, prospective genotyping with the aim of identifying patients on the basis of their genetic profile who, when prescribed anticancer medicines, have a substantially better ratio of benefits to risks [22]. After a general background of PM, we review its potential impact on the design of clinical trials and economic evaluations, mainly referring to Europe as a setting. Then, we assess the ‘state of the art’ of the economic evaluations focused on the two drugs that are usually mentioned as emblematic examples of targeted therapies, that is, trastuzumab (TR) and cetuximab (CX), in their first indications (respectively, metastatic breast and colorectal cancers).

PM & clinical trials

Clinical trials on targeted therapies become more complicated, since the choice is no longer between a new medicine and an existing one but rather between ‘treat-all’ and ‘test-and-treat’ strategies. Some have argued that, before randomizing patients to the new drug, they should be randomized to the diagnostic test too, to prove its efficacy [23]. But, more realistically, this implies a two-stage process in which we evaluate first the screening test, and then seek to confirm the effectiveness of the treatment in the identified population. Ideally, epidemiological data should be available to estimate the prevalence of the biomarker and the predictive value of the test. In general, PM should result in the need for more and different data, to reflect accurately the impact of a diagnostic test on a patient care pathway [10]. These results should continue to be estimated from clinical trials, supported by well-designed prospective observational cohorts analyzed with appropriate methods and adjustments for confounding variables.

Another important issue is the timing of the diagnostic test, either together with the drug (e.g., TR) or after drug approval and thus requiring a post hoc retrospective analysis (e.g., CX). In this second case, if pharmacogenomic data were collected alongside a clinical trial, post hoc subanalyses could detect an interaction between drug response and a particular genotype in subgroups of patients, although such analyses would face challenges of multiplicity and false-positive findings. Finally, the results from companion tests are usually continuous rather than binary, so dichotomizing their scores (negative/positive) to set a cutoff for treating/not treating patients could be questionable [10].

PM & economic evaluations

In the recent pharmacoeconomic literature, it has often been noted that PM needs particular guidelines to assess the costs and benefits of alternatives correctly [23–25], again mainly due to the genetic tests, which drive targeted therapies and thus increase the number of potential alternatives. In principle, the cost of testing the whole population of patients should be posted only to the targeted therapy if the test was not routinely adopted before the therapy was introduced [26].

Economic evaluations on targeted therapies should require not only a more complex decision space but also a clear understanding on what tests will be done after the initial one and what treatment will be given after the tests [23]. In practice, even when a biomarker has been mentioned as a part of the product approval to target the eligible population by EMA, the technological procedure to be used will generally remain unspecified. To assess diagnostic performance (in terms of sensitivity and specificity) [23] and its costs, one should know what test will actually be used in clinical practice – a laboratory-made test or a specific test marketed by a diagnostic company (or the pharmaceutical company itself) with a self-certified CE mark – since the way patients are managed could differ widely among centers [26]. The choice of a particular test should also depend on the prevalence of the biomarker, the cost of testing and the payer’s strategy to treat all patients or not with stratified therapies depending on budgetary constraints. Finally, the different funding options in the healthcare systems should be taken into account when monetizing the costs of the targeted therapy and its ‘companion test’. For instance, the two main types of funding in the EU countries for TR in breast cancer are test and drug either through a hospital or a combination between the hospital and the third-party payer [27]. Finally, the number of biomarkers for targeting care is likely to keep on increasing, so many genomic tests could become ‘obsolete’ very fast [10], undermining the credibility of economic evaluations conducted according to long-term horizons.

Accordingly, the uncertainty of economic evaluations in PM is clear, both about their frame and the inputs concerning the two technologies (the drug and its companion test) [23], calling for thorough sensitivity analysis. This might imply reconsidering their cost–effectiveness later in the technological lifecycle, particularly if new biomarkers for targeting care become available in the meantime [10].

Two emblematic examples of targeted therapies

Two separate literature searches were conducted on the PubMed international database to select full economic evaluations conducted in Europe and published in English from January 2000 for TR and January 2008 for CX (post hoc evidence dates from that year) till May 2015. The search terms were ‘trastuzumab’ and ‘metastatic breast cancer’ or ‘cetuximab’ and ‘metastatic colorectal cancer’ combined with ‘cost’ or ‘cost–effectiveness’ or ‘economic evaluation’.

We retrieved 93 articles for TR: 83 were discarded because they did not include a full economic evaluation on TR as first-line innovative therapy in metastatic breast cancer¹. Since five full economic evaluations did not concern the European setting, we selected five articles eventually [28–32]. We retrieved 58 articles for CX and 47 were discarded². Since seven full economic evaluations were not conducted in European countries, we selected four studies finally [33–36].

We reviewed all nine selected studies according to a restricted checklist focused on the items relevant to a PM scenario: source of efficacy; number of alternatives; allocation and assessment of the companion test cost; proportion of companion test and targeted therapy costs on total costs; and most influential variables in sensitivity analysis.

The nine full EEs concerned six European jurisdictions (Table 1), six of them concluding in favor of targeted therapies [29–31,33–35] and the other three against [28,32,36]. While CX efficacy was entirely derived from post hoc analyses – in one study [34] adjusted through an expert panel – TR efficacy was assessed from randomized clinical trials in all but the two French studies [30,31], which instead used observational studies based on uncontrolled, small samples of patients [37]. All the clinical trials included had high proportions of cross-over patients.

Table 1. CX & TR studies reviewed according to our checklist focused on PM.

First author (setting year)	Efficacy sources	Multiple alternatives (>2)	Test cost fully allocated to the bioagent	Test and source of the unit cost	Test unit cost (% total costs)	Bioagent (% total costs)	Most influential variables in sensitivity analysis	Ref.
Hoyle et al. (UK 2013)	Post hoc analysis from RCT	Yes	Yes	KRAS Hospital price list	€205 (1)	CX (44−54)	CX cost, OS; PFS	[36]
Vijayaraghavan et al. (Germany 2012)	Post hoc analysis from RCT	Yes	No	KRAS Manufacturer	€280 (1)	CX (49−52)	CX cost, target population	[35]
Asseburg et al. (Germany 2011)	Post hoc analysis from RCT	No	–	–	–	CX (27)	Time horizon, OS, PFS	[34]
Blank et al. (Switzerland 2011)	Post hoc analysis from RCT	Yes	No	KRAS/BRAF Hospital price list	€394 (1.1)	CX (46)	CX cost, OS, PFS, QoL	[33]
Matter-Walstra et al. (Switzerland 2010)	RCT^†	No	–	–	–	TR (35)	TR cost, PFS, TTP	[32]
Perez-Ellis et al. (France 2009)	Monocenter retrospective study	No	–	–	–	TR (43)	TR cost, hospitalization cost	[31]
Poncet et al. (France 2008)	Multicenter prospective study	No	No	IHC Assumption	€64 (0.2)	TR (43)	–	[30]
Lidgren et al. (Sweden 2008)	RCT	Yes	No	IHC/FISH Hospital price list	€200/€548 (0.4–1.2)	TR (51–57)	TR cost, QoL	[29]
Norum et al. (Norway 2005)	RCT	No	Yes	IHC Manufacturer	€31.7 (0.3)	TR (89)	TR cost, OS	[28]

^†Mix of locally advanced and metastatic breast cancer patients.

BRAF: B-Raf proto-oncogene serine/threonine-protein kinase; CX: Cetuximab; FISH: Fluorescence in situ hybridization; IHC: Immunohistochemical assay; KRAS: Kirsten rat sarcoma oncogene homolog; OS: Overall survival; PFS: Progression-free survival; PM: Personalized medicine; QoL: Quality of life; RCT: Randomized controlled trial; TR: Trastuzumab; TTP: Time to progression.

Four studies had more than two alternatives through the inclusion of companion tests [29,33,35,36]. The cost of testing was not considered at all in three studies [31,32,34] and only two of the remaining six [28,36] fully attributed it to the targeted therapy alternative. The source of the companion tests’ unit cost was weak in half of these six studies [28,30,35], and the proportion of the test cost on the total was negligible (<1.5%) in all of them, while that of the targeted therapy was substantial in all nine studies. As a consequence, while the targeted therapy cost was an influential variable in seven of the eight studies that conducted a sensitivity analysis, this was never the case for the companion test cost, even though its unit cost was quite high (>€200) in four studies.

Policy implications

PM is a fascinating subject in principle; however, clinical results have not been as encouraging as hoped so far, particularly in oncology, which was expected to be its major field of application and where targeted therapies have not yet been able to replace classical chemotherapy [22]. Of course, there may be many reasons behind these disappointing results and some caution should be exercised about calling genes ‘cancer-related’ since not all somatic mutations are tumor driving and need to be seen in the broader context of tumor heterogeneity [16,22]. In general, although target prioritization is a major issue, targeted therapy discovery and development are still the main bottlenecks [22], and the growing number of companion tests might also be a marketing strategy followed by manufacturers to differentiate their bioagents [26]. Somewhat cynically, post hoc analyses might be useful to find a subset of patients as a ‘market niche’ for drugs that failed to show efficacy in clinical trials for approval.

Of course, economic evaluations on targeted therapies cannot help overcome the lack of clinical evidence for most of them and improve the still confusing PM ‘state of the art’ [24,25]. The two paradigmatic examples of CX and TR indicate that the methodological implications on economic evaluations debated in the literature [10,23] are more theoretical than practical, since the cost of companion tests was negligible in all the European studies reviewed, so this item could in fact have been excluded from calculations, according to economic evaluation theory [38]. From an economic point of view, the real ‘crux of the matter’ lies in targeted therapies, which are very expensive add-ons whose additional cost can hardly be compensated by ‘trade-offs’ with other healthcare items, particularly in end-of-life situations [37,39]. Public decision makers in European countries where economic evaluations are required for pricing and reimbursement procedures should be fully aware of these limits when coping with dossiers submitted by targeted therapy manufacturers.

Financial & competing interests disclosure

N Freemantle has received funding for research and consulting from a number of companies that may be currently developing personalized medicine strategies. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Notes

¹83 articles were excluded, being: focused on clinical issues or other subjects not related to our aim(50); reviews or comments (13); partial economic evaluations (11); and full economic evaluations not focused on TR for treatment in metastatic cancer (9).

²47 articles were excluded, being: focused on clinical issues or other subjects not related to our aim (27); reviews or comments (11); partial economic evaluations (8) and one full economic evaluation not focused on CX.