Making Investments in Medical Technology: Time to Get Real About Real Options

Abstract

The article by Grutters et al. on using real options analysis to investigate the adoption of medical technology in The Netherlands, published in this issue of The Oncologist, is reviewed.

Physicians enjoy training and experience in making critical decisions under high levels of uncertainty, matched perhaps only by members of the investment community. Despite this comfort level, the medical profession has not routinely applied robust analytical frameworks to inform capital allocation decisions. Physicians participate in debates about whether health systems should invest in new technological capabilities, but their role is often limited to commenting on value from a purely clinical perspective. Payers, investors, administrators, regulators, and politicians increasingly drive decisions about whether or not to allocate resources toward the development and adoption of novel medical technologies. This decision-making process can quickly devolve into each interest group protecting its own “turf,” resulting in suboptimal decisions that may ultimately fail to help patients.

The culmination of years of suboptimal decision making may finally have reached a tipping point. Health care costs have grown at a staggering rate, despite the recent financial crisis and broader economic slowdown. Health care expenditures now comprise >17% of the U.S. gross domestic product and are the primary drivers of long-term fiscal budget deficits [1]. Across most advanced economies, budget realities are placing a greater burden on medical professionals to justify the value of the treatments they provide, or wish to provide in the future. It is in this context that we welcome attempts by teams of physicians and health economists to apply analytical tools from the financial community when assessing the potential value of emerging medical technologies.

Real options analysis (ROA) is a powerful analytical approach that investors and managers are using with increasing frequency to help inform capital budgeting decisions, especially in instances where high levels of uncertainty exist. First coined by Steward Myers of the Massachusetts Institute of Technology in 1977 [2], ROA has its roots in financial options contracts that provide holders with the right but not the obligation to buy or sell an underlying security by a certain point in the future. Similarly, real options give decision makers the right but not the obligation to take an action in the future. The duration of the option period, the level of uncertainty underlying the outcome, and the decision makers' flexibility in adapting to new information all determine the value of the options, both in finance and in “real life.” The greater the uncertainty and the more flexibility a decision maker has to change course, the greater the value of the option, particularly if it is long lived (Fig. 1).

Figure 1.

Factors that influence the value of real options.

Take, for instance, the decision of whether or not to build a $500 million factory to assemble a new electric vehicle. Absent managerial flexibility, there are only two options: either build the factory or do not (Fig. 2). What if, however, a manager is not fully committed to a plan of action and there is a way to acquire additional information and remove uncertainty before making a final investment decision? It may be possible, for example, to first build a $20 million pilot plant to help determine the feasibility of producing new electric vehicles in a cost-effective manner, and perhaps also to wait to see how government policy concerning the tax treatment of electric vehicles may evolve. If the manager learns that the electric car cannot be profitably produced, the only cost is the $20 million spent on a pilot plant, not the $500 million required to construct a full production facility. Quantifying the economic value of these “real-life” options is the essence of ROA.

Figure 2.

The addition of real options analysis (ROA) to the decision-making process.

In their article “When wait for more evidence, Real options analysis in proton therapy,” Grutters et al. [3] used ROA to investigate the adoption of medical technology in The Netherlands. In this particular case, the authors examined the use of proton therapy for the treatment of patients with inoperable stage I non-small cell lung cancer (NSCLC), compared with stereotactic body radiotherapy (SBRT) with photons for the same indication. The question is an appropriate application of this powerful analytical approach, because (a) there is uncertainty concerning the relative efficacy of the two treatment options and (b) proton therapy requires a large, upfront investment of €95 million ($127 million) [4] that is difficult to reverse should the value of the technology fail to provide the expected benefit.

The authors compared two options: whether it is better to adopt proton therapy before receiving the results of a clinical trial (“adopt and trial”) or to wait until the clinical trial results are known (“delay and trial”) before moving from SBRT to the adoption of proton therapy. To answer this question, they compared the health benefits foregone by not having proton therapy available before the trial with the unnecessary cost of investing in proton therapy equipment if there is, in fact, no clinical benefit. The investigators concluded that the best decision is the adopt and trial option and suggested beginning the 5-year process of constructing a proton facility in The Netherlands before additional clinical trial data are collected. Concurrent with facility construction, the team suggested sending an unselected cohort of inoperable stage I NSCLC patients to receive proton therapy abroad where appropriate facilities exist today. The investigators plan to collect survival data from the traveling patient cohort and determine whether or not there is indeed a clinical benefit. If confirmed, proton therapy will be offered in The Netherlands to patients with inoperable stage I NSCLC. Their analysis concluded that 200 patients is the optimal size for the confirmatory observational trial.

We applaud the authors for attempting to use ROA to answer the question of whether or not to adopt a costly medical technology. They have also taken efforts to make their work comprehensible to a primarily clinical audience. However, their analysis has two major limitations that may call into question the applicability of their conclusions.

First, the authors' core conclusion (that it is best to adopt proton therapy now and run a concurrent observational trial, because the health benefits foregone by delaying adoption are greater than the potential costs of investing in an unnecessary technology) relies heavily on findings from their previous work, which in turn is based on a small number of nondefinitive observational trials. In their previously published cost-effectiveness analysis, the authors reported using a meta-analysis of nonrandomized proton studies published after 2004 to find that the application of proton therapy in patients with inoperable stage I NSCLC will lead to an improvement of 0.21 quality-adjusted life years (QALYs) relative to SBRT at an incremental cost of €10,962 ($14,689) per patient [5]. A review of their meta-analysis indicates, however, that only two proton therapy trials were published after 2004 (one of them was retrospective and neither was randomized), involving a total of only 58 patients [6]. The precision of their input is therefore questionable in the absence of controlled trials, yet it drives the core finding in this paper.

Specifically, the fact that the average expected benefit of 0.21 QALYs (the Dutch would value 0.21 QALYs at €16,800 [$22,512]) is greater than the expected incremental cost of proton therapy of €10,862 ($14,555) influences to a large extent the authors' adopt and trial conclusion. In our view, the authors should have focused more on the likelihood that the findings of their meta-analysis would in fact be borne out in clinical study. The authors could have, for example, solved for the range of probabilities of randomized trial success wherein the adopt and trial option is still the optimal health policy decision. If we knew with near certainty that proton therapy would improve survival rates over those seen with SBRT, there would be no need to run a trial at all. The answer would be simple: adopt now, as long as the costs are reasonable. On the other hand, if we had little basis to assess how proton therapy performed relative to SBRT, the uncertainty of adopting without additional information would be too high; we would have to delay adoption before additional trial data could be collected. ROA is most useful in the “gray area,” where there is a reasonable chance that a treatment modality is superior but additional information would be helpful. In this gray area, the benefit of gathering new information is high, but there is also a potential societal cost to withholding a treatment that may prove superior to what is currently available.

The second major limitation of the study is the confirmatory trial design. For a trial to be worth its cost, it must enhance our ability to discern the relative clinical utility and cost-effectiveness of existing treatments. Here, the authors' choice of a confirmatory single-arm, observational trial design that is prone to patient selection and other biases is unlikely to provide any definitive conclusions. For example, each patient with inoperable NSCLC who gets enrolled in their trial would, by definition, need to be well enough to travel to another country to receive treatment. That assumption alone may bias the outcome of the single-arm design. The simulations that the authors employ do in fact allow for the possibility that any given observational trial may show that protons are no better than SBRT, but because the authors set the estimated mortality rates lower for proton therapy, with enough simulations, proton therapy will always be found to have a survival advantage under their design. We share the view with others that performing a randomized trial would indeed be ethical [7], and is the best way to determine whether proton therapy does lead to an improvement in health outcomes in this setting. Layering powerful analytical tools over dubious outcomes data will not move the field forward. Instead, it will only obfuscate the clinical information that is ultimately most important to the patient: is the treatment in question definitively better?

Despite this article's limitations, we hope that additional groups explore the use of ROA to resolve critical areas of uncertainty throughout the cancer field. In addition to other clinical settings, regulatory policy and drug development seem especially suitable to ROA because policy makers and drug sponsors must frequently trade off the benefit of gathering new information from the cost of withholding a treatment that may prove superior to what is currently available. This dynamic is present when considering the accelerated approval mechanism for drugs and biologics to treat serious or life-threatening illnesses based on clinical data from earlier-stage, developmental trials. The U.S. Food and Drug Administration grants these accelerated approvals conditionally based on surrogates of clinical benefit and requires confirmatory trials while reserving the right to withdraw the product from the market should the sponsor fail to complete the trial or fail to show a clinical benefit in the confirmatory trial. ROA could conceivably help the agency and its advisory committees to quantify the societal costs and benefits of accelerated versus regular approval for any new drug or biologic licensing application. Industry sponsors may also apply ROA in evaluating various development paths for their investigational agents; such an analysis may be particularly helpful in quantifying the economic value of one or more biomarker-based strategies to help select patients most likely to receive a clinical benefit based on particular molecular subsets. We suspect that, in most instances, the economic value of the biomarker-based development approach (where available) would be shown to greatly exceed the “treat all” approach.

Going forward, we encourage decision makers to employ ROA to a range of areas with the following guiding framework: (a) What uncertainty exists today? (b) What options exist for removing uncertainty? and (c) What are the costs of waiting to remove that uncertainty versus acting now? This process is more complex than merely plugging numbers into a formula. Thoughtful assessment is critical, and balancing capital allocation decisions with clinical utility and the value society places on health gains will require increased collaboration among medical practitioners, economists, administrators, and payers. In finance, the stakes can be high, but in no field are the stakes as high as in medicine. Money is fungible, but patient health is decidedly not.

See the accompanying article on pages 1752–1761 of this issue.

Author Contributions

Conception/Design: Thomas G. Roberts Jr., Philip D. Dreyfuss

Data analysis and interpretation: Thomas G. Roberts Jr., Philip D. Dreyfuss

Manuscript writing: Thomas G. Roberts Jr., Philip D. Dreyfuss

Final approval of manuscript: Thomas G. Roberts Jr., Philip D. Dreyfuss