Anticipating the Market Access Outlook for Drugs in R&D

Anticipating the Market Access Outlook for Drugs in R&D
Walter Brooks, Vice President, Equinox Group; David Godolphin, Vice President, Equinox Group; and Allan Miller, Vice President, Equinox Group
This article describes an analytical framework to evaluate the market access outlook for drugs in R&D. Based on the anticipated clinical attributes and planned price for a new drug, this method provides an objective basis for quantifying a drug’s clinical benefits and its impact on direct costs of treatment (both drug price and cost off-sets). The resulting ratio of benefit to cost is then compared to the corresponding ratios for drugs launched in recent years. This historical dataset shows there is a strong correlation between clinical benefit vs. cost and the ability to achieve market access. Comparing a new drug to those in the historical dataset reveals where the new drug’s benefit vs. cost ratio falls along the spectrum of good to poor market access. Finally, this modeling technique easily accommodates sensitivity analysis – useful when there is high uncertainty about the drug’s clinical attributes or planned price.
Keywords: Market Access, Pricing, Clinical Benefit vs. Cost

The market access new drugs ultimately achieve is affected by decisions made at all stages of R&D, including those undertaken in early development. Examples of such decisions include responses to questions such as:
  • In what patient segments should the drug be developed – only the most severely ill, or more broadly for all patients with the disease; patients with prognostic markers; patients at specific lines of therapy?
  • With which agents, if any, should the drug be combined?
  • Against what comparators should the drug be tested in clinical trials?
Most biopharmaceutical companies recognize the need to apply analytics to address these and related questions early and at each decision point in the R&D cycle. Traditional market research tools provide essential information from clinical experts. But those tools also have inherent weaknesses, which will be familiar to anyone who has watched their newly launched drug vastly underperform its commercial expectations. As valuable as primary research is, its weaknesses should be accounted for by adding other analytical methods to reduce the risk of mistaken commercial assessments, especially those driving important development decisions.

To ensure that a new drug is positioned for favorable market access, development decisions should be made with a clear notion of how they might affect the magnitude of clinical benefit in the target patient population. Will the clinical benefits warrant the desired price? Drugs that miss that goal are unlikely to achieve favorable market access, adequate patient share, or a revenue stream that pays back the R&D investment. This paper presents historical data showing that the balance between a drug’s clinical benefit and its price helps explain its ability to achieve market access and patient share. Having an historical perspective provides guidance about what will be required of drugs in R&D. The anticipated benefit/cost balance of a developmental drug can be compared to drugs in the historical data set to determine if its balance aligns better with recent historical successes or with failures. This paper suggests a practical solution to improve the analytics needed to inform these decisions in a time- and cost-effective way.
Primary Market Research is Crucial, But Its Inherent Weaknesses Must Be Addressed
Primary research results frequently contain “framing problems”–the introduction of bias arising from the manner in which information is provided and questions are posed. Framing problems are hard to avoid. For instance, if a series of questions puts great emphasis on dosing, the results are likely to artificially raise the importance of dosing in relation to efficacy and safety/tolerability. Moreover, questions posed in market research are typically affected by individuals with a stake in the outcome of the research (product team members), so the tendency is to elicit feedback that is favorable to the product in question.

Another weakness of primary research is in predicting patient share, specifically, the conversion of preference share to patient share, which is notoriously imprecise and usually at the heart of forecasts gone bad.

Finally, primary research focuses on the clinical attributes of drugs with little direct assessment of disease burden (mortality, morbidity, and costs). But disease burden is a crucial factor affecting payer policies. This limitation should be addressed directly.

We argue that the weaknesses described here explain why many new drugs miss their prelaunch commercial expectations. To address these weaknesses, we propose using an objective and comprehensive framework to compare a new drug to relevant competitors and measure its benefits and costs compared to those alternatives.
An Additional Analytical Perspective
There is a powerful source of information about the clinical benefits of a new drug–the hard clinical data upon which drug approvals are based. Such data has been collected under the supervision of regulatory bodies that insist on valid measurement of clinical advantages and disadvantages. Clinical trial data is an ideal source of information to measure the strengths and weaknesses of any drug. Other researchers have also shown that hard clinical data directly affects prescribing.1,2,3

To assess the magnitude of clinical improvement offered by a developmental drug, its anticipated clinical characteristics can be compared, attribute by attribute, to those of the standard of care (SOC). By exploiting clinical trial data we gain deeper and more precise insights into the strengths and weaknesses of the new drug relative to the SOC. This approach can also measure changes in disease burden–mortality, morbidity, and costs–precisely.
Measuring Clinical Improvement– Benchmarking Against the SOC
The first step in this method is to measure a drug’s benefits and costs in an objective and consistent way. We illustrate this concept using the metrics Equinox Group has developed for this purpose, but other holistic metrics can serve the same goal (the sidebar at the end of this article provides more details and observations about the design of a consistent system for measuring clinical benefit).

The clinical benefit of the new drug is its marginal improvement relative to the SOC. The starting point should be to determine how much unmet medical need remains after a typical patient has been treated with the SOC. The domains of medical need that should be included are shown in Table 1–note the inclusion of both drug attributes and elements of disease burden.

Table 1: Domains of Medical Need
Efficacycure, prophylactic success, symptom relief, slowing of progression, damage reversal, pharmacokinetics
Safety/Tolerability frequency and severity of each side effect, net of placebo, warnings and monitoring requirements
Convenience mode and frequency of dosing
Mortality age-adjusted excess risk of mortality
Morbidity pain, disability, hospitalization, quality of life, complications
Costs drug price, non-drug costs, lost work time

Unmet need in each domain should be quantified using data from clinical trials and peer-reviewed literature. For instance, it is possible to quantify mortality in virtually any disease. For the SOC’s efficacy, we use endpoint data collected in pivotal trials to measure factors such as impact on symptoms and disease progression.

These clinical data are mapped onto scales for each domain to reflect the extent to which remaining need is low or high. The scores for each domain are weighted (see sidebar at the end of article) to reflect the inherent importance of each domain. The sum of the weighted scores reflects overall unmet medical need for the SOC in the target population.

The process can then be repeated with data or assumptions for the candidate drug, using the same scales, to calculate its total unmet need score. Improvements in efficacy reduce disease burden. For instance, symptom relief might reduce pain scores or improve quality of life. Reduction in cardiovascular events should reduce mortality.

This approach provides objective and consistently derived scores for both the SOC and the new drug, so they can be validly compared. The new drug’s reduction in medical need relative to the standard of care is a measure of its clinical improvement.

We have conducted these analyses for hundreds of drugs and have data that show what constitutes low, medium, and high clinical improvement, based on historical observations of peak patient share achieved. We have validated this framework through a statistical model that predicts peak-year patient share as a function of clinical improvement and other causative variables. The correlation is strong (R-squared of 85%).*
Defining the Balance Between a New Drug’s Clinical Benefit and Its Cost (Price)
Figure 1

Figure 1 shows a simple framework to compare the balance between clinical benefit (X-axis) and cost impact (Y-axis) for a prescription drug. The greater a drug’s clinical benefit, the further to the right it is plotted. The higher the drug’s net cost impact, the lower it is plotted. The dotted line running through the origin represents “value equal to cost” or, put another way, the price at which all of the drug’s clinical value is clawed back by the developer. Drugs to the right of the dotted line offer more benefit than cost and drugs to the left cost more than the clinical benefit they deliver. Our analysis shows that there is a strong correlation between successful market access and being on the right side of the “value equal to cost” line (Figure 2).
Figure 2

Government regulators, academia, and medical societies have grappled with initiatives to measure the value of prescription drugs. Peter Neumann and Joshua Cohen at the Institute for Clinical Research and Health Policy Studies at Tufts Medical Center have written about the strengths and weaknesses of many of these approaches.4 Most of these initiatives focus on disease burden–the drug’s impact on mortality, morbidity, and cost. For instance, The National Institute of Clinical and Health and Care Excellence (NICE) and The Institute of Clinical and Economic Review (ICER) emphasize disease burden in their analyses. The quality-adjusted-life-year (QALY) is a well-established metric for this purpose, although it is not clear how it can be used to predict commercial performance and market access. Clearly, developers need to understand how their developmental drugs affect disease burden.

Focusing exclusively on disease burden, as most of these initiatives do, can miss other benefits arising from new therapies; these factors affect the patient’s experience and therefore influence market success. For instance, the once-a-day oral dosing of Gilenya (fingolimod) offers an important advantage over the subcutaneous injections of beta-interferons in the treatment of relapsing multiple sclerosis. That dosing advantage contributed substantially to Gilenya’s rapid adoption. The safety/tolerability advantages of Stelara (ustekinumab) over the TNF-alpha inhibitors in the treatment of psoriasis does not affect disease burden, but it is a significant benefit for patients and was the main contributor to Stelara’s commercial success. QALYs do not fully capture these benefits. Any method to accurately measure the “value” of a new drug must take into account both product attributes (efficacy, safety/tolerability, dosing) and disease burden.
Measuring a New Drug’s Reduction in Unmet Need
A model containing all of the domains included in Table 1 reflects both disease burden and other clinical benefits (e.g., safety/tolerability and dosing) that QALYs miss. To illustrate the concept, Figure 3 compares Gilenya to Rebif in relapsing multiple sclerosis. Gilenya reduces medical need by 9.3%, and the waterfall chart shows the sources of its advantages and disadvantages of Rebif for each domain of medical need.
Figure 3: Comparing Gilenya to Rebif in RR-MS

Using a 0-5 scale that reflects how high unmet need is given treatment with each drug (see side-bar, at the end of article), the total unmet need score is 2.76 for Rebif and 2.50 for Gilenya. Gilenya reduces medical need; that is, it offers clinical improvement over Rebif. The overall improvement of Gilenya is 9.3%, and the sources and magnitudes of Gilenya’s advantages appear in the graphic – modest improvements in efficacy and safety/tolerability, with a more substantial benefit in dosing (convenience). The slightly higher price for Gilenya is a disadvantage, reflected in the direct cost domain (where drug costs and cost offsets are captured).

Because this analysis measures how much of the overall improvement comes from each domain of medical need, we can separate out the impact of cost (driven mainly by drug price) from other drivers. “Clinical Benefit” is defined as the net impact of all non-direct-cost factors combined – i.e., improvements in mortality, morbidity, efficacy, safety/tolerability, dosing, and indirect cost. Gilenya’s “clinical benefit" (compared to Rebif) is 9.8%:
  • Efficacy 2.1%
  • Safety/tolerability 2.1%
  • Convenience (dosing) 5.2%
  • Mortality 0%
  • Morbidity 0.5%
  • Indirect costs -0.1%
The “cost” value for Gilenya (driven by its higher price) is -0.5%, as shown in Figure 3. These two values are used to map Gilenya onto the benefit vs. cost graph, shown here in Figure 4. It shows that Gilenya’s coordinates place it well to the right of “value equal to cost” line, in the favorable zone. Gilenya had rapid market penetration and has achieved annual sales of more than $3 billion, indicating that market access was favorable.
Figure 4: Mapping Gilenya’s Clinical Benefit and Cost Into the Pricing and Market Access Graph

Adding Observations to the “Benefit vs. Cost” Graph
Figure 5 shows the results for 40 drugs launched in recent years analyzed through this method. In addition to mapping benefit vs. cost, we have colored the points representing each drug to indicate whether it has achieved high patient share (green), low patient share (red), or patient share somewhere in between or not yet clear (grey). Naturally, the ability to achieve strong patient share requires that market access be favorable. Drugs to the right of the “benefit equals cost” line have tended to achieve strong patient share, while those to the left have not.

Figure 5: Clinical Benefit vs. Cost for 40 Recent Launches

Here we comment on results for several agents shown above:
  • Harvoni – compared to the Incivek regimen in the treatment of HCV genotype 1. Despite the controversy surrounding Harvoni’s high price, this analysis shows that its clinical benefits are well worth that price. Much of the benefit is attributable to the elimination of pegylated interferon and ribavirin from the regimen, resulting in large improvements in safety/tolerability and dosing. The improvement in efficacy also confers gains in mortality and morbidity. The net “clinical benefit” measured here is over 48% – very high by historical standards. The net drug price impact (an increase of $12,000 for a course of treatment) is small relative to the benefits. FDA approved in October 2014; Harvoni generated revenue of $13.9 billion in 2015, indicating that market access was not problematic.
  • Xalkori – compared to carboplatin + paclitaxel in first line ALK+ non-small-cell lung cancer. Xalkori is an example of a branded agent competing successfully against inexpensive generics. Xalkori’s revenue is not high by old “blockbuster” standards due to the small target patient population. But because its clinical benefit is so high (substantial improvements in efficacy and mortality, with added improvements in safety/tolerability and dosing), Xalkori quickly went on to achieve overwhelming patient share in this small patient population.
  • Pradaxa – compared to warfarin in stroke prevention in atrial fibrillation (SPAF). As the first of the novel oral anticoagulants indicated for SPAF, Pradaxa offered strong clinical benefits at a price premium to generic warfarin, mainly from much improved efficacy. Pradaxa achieved rapid revenue growth despite its high price before competition from other novel oral agents slowed its growth.
  • Brintellix – compared to escitalopram in major depressive disorders. Brintellix offers modest improvements in efficacy and safety/tolerability, but the magnitude of these advantages are low, earning a clinical benefit value of only 1.7%. But the branded price against generic escitalopram results in a cost disadvantage of -3.7%. Brintellix has struggled to achieve patient share.
  • Anoro Ellipta – compared to Spiriva in chronic obstructive pulmonary disorder. Anoro Ellipta offers improvements in efficacy and safety/tolerability over Spiriva, but the magnitude of those improvements is modest at 1.5%. It is priced at a significant discount to Spiriva, providing a cost advantage at 1.1% (note its placement above the X axis). Despite being on the favorable side of the Benefit-Cost line, since its approval in December of 2013, Anoro Ellipta has had modest sales. In recent years, drugs with overall clinical improvement at levels this modest tend to struggle to achieve differentiation.
  • Entresto regimen – compared to a traditional heart failure regimen. Entresto (replacing the traditional ACEi or ARB) offers a moderate improvement in efficacy, which confers benefits in mortality and morbidity, and with modest disadvantages in safety/tolerability and dosing. The net clinical benefit is 5.3%. The price disadvantage is -1.1%. This places Entresto on the favorable side of the benefit-cost line. Approved in July 2015, Entresto has underperformed Wall Street predictions. This is another example of a new agent with relatively modest clinical improvement entering a market dominated by many effective and safe generic alternatives. Our analysis suggests that Entresto’s sales will improve, but will fall well short of the $5 billion predicted by Wall Street analysts.
Applying This Method to Assess a New Drug Program
To analyze a new drug’s market access outlook through this lens requires characterizing the extent to which the current SOC meets the medical need in the target population (see the domains of need listed in Table 1). The analysis of the SOC creates a baseline against which the new drug is compared. For the SOC, information on product attributes (efficacy, safety/tolerability, and dosing) can be taken directly from its package insert or clinical trial data published in peer-reviewed journals.

Data describing the SOC’s efficacy endpoint values are translated into a numerical score that reflects how efficacious the SOC is. The new drug’s efficacy score is based on endpoint values it is expected to achieve and calculated using the same scales and weights that were used for the SOC. The same process is applied to safety/tolerability and dosing.

To illustrate the concepts, imagine a team is evaluating a drug targeted for relapsing multiple sclerosis. There is high uncertainty about its efficacy. Management wants to know how the market access outlook is affected by assumed efficacy improvements. The uncertain range in efficacy is bounded by the Low Case and High Case profiles (Table 2).
Table 2: Clinical Attributes of MS Drug in Two Profiles vs. Tecfidera
Relapsing Multiple Sclerosis Current SOC: Tecfidera (Dimethyl fumarate) Low Case High Case
Relative Reduction in Annual Relapse Rates (net of placebo) 49% 55% 60%
Relative Reduction in % Progressing at 2 years (net of placebo) 38% 38% 50%
Damage Reversal (net of placebo) 0% 0% 20%
Price $54,750/year $49,274/year $60,225/year
Oral BID Safety/tolerability and dosing equal to Tecfidera

The team decides that Tecfidera (dimethyl fumarate) is the appropriate SOC to benchmark against their new drug’s profiles. Table 2 shows key clinical attributes for Tecfidera and the assumed efficacy for the Low Case and High Case profiles. Note that the low profile is expected to offer only a modest improvement in relapse rates. The high profile is expected to have greater reduction in relapses, improved reduction in progression, and damage reversal. Furthermore, the team thinks that the High Case profile will command a premium price relative to the SOC, and the Low Case profile will require a lower price.

In addition to differences in clinical attributes, the model measures how the new drug’s improved efficacy will reduce disease burden; for instance, how symptom reduction diminishes pain or improves quality of life. Objective data to measure disease burden (mortality, morbidity, and costs) is also obtained from peer reviewed literature.

Each of the subdomains of morbidity listed in Table 1 is analyzed and quantified. For instance, reduction in the progression of MS will yield improvements in disability, an important benefit. Table 3 shows the disability scores for Tecfidera, and the Low and High cases. The High Case reduces disability because it offers slower disease progression and reverses damage compared to Tecfidera. We assume that these efficacy improvements shift the distribution of disability from the most severe level (ADL-impaired) to lower severities (activities limited, minor disability, no disability). The lower total chronic disability score for the High Case reflects improvement (reduced medical need), and is included as part of the total clinical benefit attributable to that profile.

Table 3: Disability Scores for the SOC, Low and High Cases
Tecfidera Low Case High Case
Chronic Disability Score 2.17 2.17 1.93
  Minor (% of patients) 46.5% 46.5% 41.1%
  Activities Limited (% of patients) 16.1% 16.1% 13.2%
  ADL-Impaired (% of patients) 1.4% 1.4% 1.1%

Similar analyses are conducted for each element of disease burden to calculate scores that reflect the total unmet medical need for Tecfidera and the Low and High Case profiles. Both profiles are compared to Tecfidera (as in the “Drivers” analysis shown in Figure 3) to calculate total clinical benefit and the cost impact of each profile, relative to Tecfidera. The outputs of that analysis are plotted on the clinical benefit vs. cost graph to identify launched drugs that have similar benefit-to-cost ratios; drugs with similar ratios represent plausible market access analogs.

In this instance, the Low Case profile has a benefit to cost ratio that is similar to Anoro Ellipta, a drug that has struggled commercially (Figure 6). Alternatively, should the new agent’s efficacy reflect the values of the High Case, its balance of benefit to cost more closely matches Gilenya, a drug that has achieved relatively favorable market access. Knowing where a drug is likely to fall in this space provides the team with information to anticipate how challenging market access is likely to be.
Figure 6: Mapping Low and High Case Profiles in Market Access Grid
Limitations of This Approach
The agents we have evaluated through this method treat a wide range of disorders, including those treated by primary care doctors and by specialists (e.g., oncology and immunology). While we have analyzed agents targeted to rare and ultra-rare disorders, to date we have too few observations to say anything definitive. Neither does this approach take into account how promotion affects patient share.
Conclusion: Anticipating the Market Access Outlook for Drugs in R&D
Crucial decisions made at all stages of R&D affect the ultimate pricing and market access outlook for a new drug. Analytics to assess the effect of alternative development strategies on market access should be applied early and refreshed at each decision point in the R&D cycle.

Traditional primary market research provides valuable insights to inform these decisions, but it has serious weaknesses, most typically leading to inflated forecasts. Primary research collects expert opinion; as valuable as that opinion is, it is nonetheless opinion and therefore a subjective basis of measurement. Second, primary research focuses on the clinical attributes of drugs, with little direct assessment of disease burden. But disease burden is a crucial factor affecting payer policies. Despite the important insights arising from primary research, its weaknesses explain why many new drugs miss prelaunch commercial expectations.

To gain a more accurate view of the market access potential and commercial outlook for developmental drugs, biopharmaceutical companies should supplement traditional tools with alternative analytics that address traditional tools’ weaknesses. We propose adding a method that explicitly measures improvement in both clinical attributes and disease burden to provide a robust comparison of a drug’s clinical benefit against its cost as seen by payers. Such an approach is grounded in objective clinical data, providing a valid basis for evaluating that critical trade-off. Arrived at independently from primary research techniques, this can confirm or challenge the findings from traditional methods. The redundancy of two independent approaches strengthens crucial developmental decisions.

This approach can be applied to new drug programs as soon as there is a hypothesis for clinical attributes and target price. Uncertain clinical attributes can be modeled in multiple scenarios to determine thresholds for commercial success (e.g., what level of efficacy will be required to achieve an adequate patient share, given anticipated safety issues?), and can inform decisions even at early phases of R&D, when clinical uncertainty is high.

Before important development decisions are taken, such as which patient segments to pursue, which comparator to select for clinical trials, which drugs to be combined with, overall go/no-go and related decisions, the alternatives should be modeled and their effects assessed in a framework like the one described here. Applying this analytical lens to each developmental program can maximize the value of each asset and that of the overall portfolio.

* The causative variables explaining peak-year patient share are: 1) level of clinical innovation offered by the new drug, 2) the number of competitors, 3) the price of the new drug, 4) the size of the population (the last two affecting payer budgets). To test robustness we performed “Leave-One-Out Cross Validation”.

Design of a Consistent System for Measuring Clinical Benefit
To design a system that measures unmet need and clinical benefit using objective clinical data, a few specific characteristics are essential. We describe here how we have implemented those characteristics in the model Equinox Group has developed and applied for more than 20 years. In that model, the contribution of each domain to total need depends on 1) the objectively measured level of need for a patient treated with the standard of care, and 2) the intrinsic importance of the domain. We multiply these two values; the product drives the final contribution to medical need of each domain in a given disease. The total unmet need score for the disease equals the sum of those weighted scores for all domains.

The Level of Unmet Need in a Domain
To measure the level of need requires clinical data. We translate that data into a score on a 0-5 scale that indicates a spectrum of low (0) to high (5) unmet need. For example, in relapsing multiple sclerosis (MS), Gilenya reduced relapse rates by 55%, translating into an unmet need score of 2.25 (55% of the distance from 5 to 0) in symptom relief efficacy. The use of objective data on consistently applied scales minimizes subjectivity and inter-rater variability in the scoring method.

The Importance of Each Domain of Unmet Need
The intrinsic importance of each domain is captured in the weights. For instance, the weights reflect the observation that efficacy has higher intrinsic importance than side effects, and side effects have higher importance than dosing. The score for each domain, when multiplied by its corresponding weight, reflects the contribution of that domain to unmet need in that disease. It is the contribution to unmet need that matches intuitive notions of the importance of a domain in a particular disease.

The weights do not differ between diseases. This may seem counter intuitive, but consider the following: myocardial infarction inflicts high mortality, and psoriasis inflicts no mortality. In both cases, the respective mortality scores are multiplied by the fixed mortality weight (20%) to arrive at the contribution to unmet need. Due to the high mortality score in MI (4.5 on the 0 – 5 score), it has a high contribution to need, whereas in psoriasis, the mortality score is 0, and therefore mortality’s contribution to medical need is 0 x 20% = 0.

About the Authors
Walter Brooks is a founder and vice president of Equinox Group. He has specialized in pharmaceutical new product development strategy and R&D resource allocation for more than 25 years. Using elements from decision analysis, microeconomics, statistics and probability theory, and market research, Mr. Brooks has contributed to the creation of analytical methods to estimate the range of market potential for a new drug based on alternative assumptions about its clinical performance. Mr. Brooks holds a B.A. in economics and an M.S. in policy analysis from the State University of New York at Stony Brook.

David Godolphin, a vice president of Equinox Group, helped found the company in 1995. At Equinox, he has led more than 400 studies for U.S. and European clients, ranging from commercial outlooks for individual products to reviews of therapeutic-area portfolios and strategies. He has a special interest in forecasting techniques that reflect factors such as unmet need performance, compliance, reimbursement, and the strengths of emerging competitors. Mr. Godolphin received his A.B. from Harvard College.

Allan Miller, a Vice President at Equinox Group, joined the company in 2007 as a Consultant after nearly 20 years of senior management experience in both commercial and scientific roles at several biotechnology and pharmaceutical firms. At Equinox Group he has focused on oncology and on the development of sophisticated simulation-based forecast models. Dr. Miller holds a M.A. in Natural Sciences (Biochemistry), from the University of Cambridge and a Ph.D. in Molecular Biology from the Laboratory of Molecular Biology at Cambridge. He conducted post-doctoral research and taught at Harvard University.


1 Azoulay P. (2002). Do pharmaceutical sales respond to scientific evidence? Journal of Economics and Management Strategy, 11(4), 551-594.

2 Ashish S, Kappe E, Stremersch S. (2014). The commercial contribution of clinical studies for pharmaceutical drugs. International Journal of Research in Marketing, 31, 65-77.

3 Kappe E, Stremersch S. (2016). Drug detailing and doctors’ prescription decisions: the role of information content in the face of competitive entry. Marketing Science, 35(6), 915-933.

4 Neumann P, Cohen J. (2015) Measuring the value of prescription drugs. New England Journal of Medicine, 373;27, 2595-2597.