The EMA final report on adaptive pathways pilot describes MAPPs as an iterative development plan that will initially target a well-defined group of patients that is likely to benefit most from the new treatment (Davis et al., 2016, EMA⁴). This is supposed to be followed by iterative phases of evidence gathering and progressive licensing adaptations, concerning both the authorized indication and the potential further therapeutic uses of the medicine, to expand its use to a wider patient population as more data becomes available. Our understanding is that this means initial marketing authorization is to be granted for promising populations based on phase II (or earlier) clinical data. It implies granting marketing authorization and, ultimately, reimbursement, based principally on surrogate endpoints. We are aware that whilst phase II trials do not always measure surrogate endpoints, any clinical endpoint data may be immature. We understand this to mean that speeding up access to new medicines via adaptive pathways is to be achieved by putting aside traditional phase III clinical trials in favor of post-marketing evidence generation. This requirement is problematic since only a very limited subset of surrogate endpoints have been validated to date and relying on unvalidated surrogate endpoints seriously increases uncertainty and is generally advised against (EPHA⁸, Mangiapane and Velasco Garrido, 2009; Velasco Garrido and Mangiapane, 2009; Svensson et al., 2013; Henshall et al., 2014; Naci et al., 2015; Prasad et al., 2015; IQWiG¹¹). Examples of concerns include extrapolating surrogate markers in patients with advanced solid tumors such as disease free survival to overall survival (Tuma, 2009; Kantarjian et al., 2013; Cortazar et al., 2014; Prasad et al., 2015).

Accepting such clinical and economic uncertainties can only be justified for new medicines meant for patients with high unmet medical need, especially those with rare diseases and if patient-relevant outcomes cannot be explored within a reasonable amount of time. However, “unmet medical need” is currently not precisely defined, although attempts have been made when assessing priority areas for new medicines (Kaplan et al., 2013; Banzi et al., 2015). For regulators and payers to be aligned, there needs to be a common agreed position for unmet medical need. This definition should be based upon a public health perspective and provide a balanced, level playing field, and the same level of risk governance for different disease conditions.

MAPPs suggest an initial marketing authorization for a restricted patient population upon demonstration of a positive benefit/risk balance. However, one has to consider that even for seriously ill patients, being treated with a new medicine with poorly characterised benefits, and which might only cause harm, should not be part of everyday care. Routine treatment, unlike testing in clinical trials, should be guided by known outcomes and not by limited or lack of knowledge.

The extent of uncertainty which patients would have to accept if new medicines were to be authorised earlier is illustrated by available data. The combined success rate at phase III and submissions is only approximately 50%. Two thirds of terminations of new medicines are due to lack of efficacy, more than 20% due to safety issues (Arrowsmith, 2011). This suggests that preliminary data from earlier phases of development typically overestimate the potential benefits of a new treatment, and do not allow a robust assessment of risks and benefits. There are also a number of examples of new medicines that were given marketing authorization on the basis of the regular data package but were proven to be ineffective or unsafe when used in routine clinical care (Eichler et al., 2011; Godman et al., 2015). The latter examples have been used to demonstrate the possibility of foregoing some of the current requirements for marketing authorization in exchange for better investigation after approval in a wider more co-morbid population likely to receive the new medicine. However, in view of the failure rates in phase III trials, including withdrawals due to safety concerns, this scenario is incomplete. If promising data after phase II were sufficient for marketing authorization, a larger number of ineffective or unsafe medicines - often at higher prices than current standards - would have been approved. To our knowledge, no adequate measures to safeguard against this have been proposed for MAPPs.

According to the EMA report on MAPPs, adaptive pathways seek to balance timely access for patients who are likely to benefit most from the new medicine with the need to provide adequate evolving information on the benefits and risks of the new medicine itself⁴. Publications arising from the EMA suggest that the EMA is of the opinion that earlier access to new medicines is necessary and that adaptive pathways are a key way to achieve this goal. We have concerns with this belief. This is because the number of new medicines that have been approved by EMA has been quite stable in recent years with no discernible decline indicating problems in the approval process¹². Consequently, we see little need to change this. Pharmaceutical companies should as a matter of course already be seeking ways to optimize development cycles. Lastly, the potential to speed up market access for new medicines when there is true unmet medical need already exists. This includes compassionate use programs or named patient programs⁷ (Baird et al., 2014). Consequently, there should be objective and scientifically valid justifications for new measures such as adaptive pathways to ensure protection of patients⁸ (Davis et al., 2016; Hawkes, 2016a). As a result, the current rationale for MAPPs does not appear compelling.

Earlier approval, which is already in existence, has for instance increased the number of black-box warnings and withdrawals in the US7 (Frank et al., 2014). In addition Prescrire, an independent drug information journal, recently assessed 22 medicines “approved conditionally” in the EU since 2006. They believed 27% were “not acceptable,” such as medicines without evident benefit but with potential or real disadvantages, 28% as having “judgment reserved,” i.e., assessments reserved until better data becomes available, 9% as “nothing new,” 18% as “possibly helpful” and only 18% as “offering an advantage”¹³. There have also been concerns with five out of the six orphan medicines receiving marketing authorisation in 2004 (Joppi et al., 2016).

Existing ways to speed up approval for marketing authorization (like conditional or exceptional approval) already pose problems for conducting health technology assessments (HTAs) and implementing measures such as valuing new medicines based on the data presented. Earlier access means inherently less available data at the time of approval. This lack of information impedes assessments of the therapeutic value of new medicines and thereby decisions on value-based pricing, exacerbated by concerns with defining terms such as innovation (Annemans et al., 2012). This has important implications for ever-increasing requested prices, especially new medicines for cancer and for orphan diseases (Cohen and Felix, 2014; Kelly and Smith, 2014; Godman et al., 2015; Howard et al., 2015). While up to now marketing authorization decisions and information are generally used as the baseline for reimbursement decisions, this position will need to be reconsidered in the framework for HTA for new medicines subject to the MAPPs process. This will create considerable challenges in already resource constrained healthcare systems.

EMA noted that adaptive licensing is already happening within the current framework as part of conditional approvals or initial approval as last-line treatment with clinical trials for earlier stages of disease to be completed at a later point in time⁷ (Banzi et al., 2015; Cerreta, 2015). Currently, these approaches are limited to special cases such as life-threatening or seriously debilitating diseases³. However, there are major concerns among payers if adaptive licensing becomes the preferred approach in the future (Eichler et al., 2015; Davis et al., 2016; Hawkes, 2016a,b). These concerns are heightened by the opinion of some that not only immediately life-threatening diseases have an unmet medical need, but in the end all diseases have a “treatment window of opportunity.” “The urgency of access to promising treatments is independent of disease dynamics: for every year without access, the window will shut on one annual cohort of patients, whether the window is short or long” (Eichler et al., 2015). This would open the possibility of adaptive pathways being considered for all new medicines across populations. Due to the inherent uncertainty of earlier access, greater risks to patients are expected to arise from this uncertainty. Payers and their advisers are concerned that this paradigm shift in the regulatory process could result in a breach of the legislative decision on limiting conditional approvals, confirmed in the context of the pharmacovigilance legislation in 2010. Even if MAPPs were limited to special cases, payers would need to actively consider updates to the framework to optimize the managed entry of new medicines in areas of high unmet need. This would create considerable challenges within countries. Consequently, earlier access even under the current legislative environment harbors risks and should only be considered under exceptional circumstances.

Instigating MAPPs also means that more evidence is to be gathered after marketing authorization. This is of concern since it is often difficult to recruit patients for post authorization studies⁷. Additional concerns with this approach have been highlighted in a recent publication (Banzi et al., 2015). The authors demonstrated that the obligations imposed by the EMA for conditional approvals are typically fulfilled only with delay and often incompletely, if ever, by pharmaceutical companies. It has been said that ensuring the fulfillment of obligations like post-approval evidence generation is one of the core issues to be addressed in adaptive pathways, and that these obligations will be legally binding. However, if there are already delays and concerns with such schemes, it is difficult to see how this will change especially with more medicines subject to this route.

Consequently, again we believe MAPPs should only apply to areas where there is an agreement between all concerned parties (including payers) that there is a true considerable unmet medical need that may be ameliorated without creating new problems and that all the uncertainties with the new medicine have been fully communicated and concerns with existing schemes addressed (Hawkes, 2016a).

Stratified medicines, especially fragmentation of populations and better understanding of diseases, have been described as drivers and enablers of adaptive pathways (Eichler et al., 2015). Stratified medicine undoubtedly results in smaller treatment populations. Due to these smaller populations, RCTs are seen by some as unfeasible. We strongly dispute this in most situations given the appreciable number of patients that exist across Europe within the current definition for orphan diseases (Garattini, 2012), and pharmaceutical research and development being planned globally. On this scale, most potentially targeted populations are big enough to support randomized trials as shown in a recent study demonstrating randomisation in studies for new medicines for orphan diseases (Picavet et al., 2013). This is further supported by publications suggesting that a number of medicines for orphan diseases have now reached blockbuster status (Kesselheim et al., 2012).

Additionally, stratification promises to provide a population where the chances of increased effectiveness in terms of reducing the numbers needed to treat, or alternatively, increasing the numbers needed to harm, are improved (Godman et al., 2013). This should further improve the chances of showing clinically relevant differences in relatively small trials.

Stratified medicine combined with better understanding of diseases is supposed to open opportunities for extrapolation of results from one population to another.

In this context, basket trials of patients with different diseases harboring the same genomic aberration that is supposed to be causative to their diseases have been exemplified as modern trial designs fitting adaptive pathways (Redig and Janne, 2015). Payers are wary of this especially if companies use small populations to drive up requested prices¹⁴ (Godman et al., 2013). Even when for example a specific cancer mutation has been shown to be causative in some entities, response in other cancer entities harboring the same mutation was not predictable (Hyman et al., 2015). This clearly demonstrates the limits to the extent of possible extrapolations during approval.

We have explained our concern that early availability of new, untested medicines, among the general population, which typically exhibits greater co-morbidities and higher age than the well-controlled population entered into phase II and III clinical trials, may jeopardize patient safety (Joppi et al., 2013; Malmstrom et al., 2013). In an environment where medicines are approved for small, well defined populations early on and evidence of their potential benefits and risks for a wider population is to be generated after this approval, restricting the use of these medicines is essential. This is why EMA's report emphasizes the need for prescription control⁴. However, it is left very unclear how this is to be accomplished.

Limiting the use of medicines to approved populations is challenging (Davis et al., 2016). Payers can limit reimbursement for use of medicines outside approved populations; however, enforcement of these restrictions can be difficult in a number of European countries without specific regulations in place (Godman et al., 2010). Any added costs and use of resources due to unwarranted use of medicines outside of approved populations will be a burden to payers especially if requested prices are higher than current standards. Several member states also currently do not provide a legal framework to limit doctors' therapeutic freedom, even if reimbursement may be excluded. However, this is not universal.

Authorization of the use of medicines to approved populations is not a valid reason for restricting their off-label use where this is warranted on clinical grounds. Other mechanisms exist, such as devolved budgets, physician education, 100% co-payment for prescribing outside of agreed indications and payback mechanisms. However, these can be labor-intensive and costly to administer (Adamski et al., 2010; Godman et al., 2012a, 2014; Ferrario and Kanavos, 2013; Malmstrom et al., 2013). EMA's report on Adaptive Licensing acknowledges the necessity of clear indications and we emphasize this point in this paper⁴. Vague or ambiguous definitions of indications are not helpful, e.g. when a new medicine is licensed for “patients for whom the [standard of care] is not appropriate.”

MAPPs imply that medicines approved in limited populations due to promising results of phase II trials are already being further researched whilst being prescribed in clinical care. It is important to repeat that this would further blur the boundary between research and routine use, and that patients would need to be informed about and accept this change (Husereau et al., 2014). It is also important to point out that clinical trials in populations for whom a medicine is already approved face certain obstacles. Such patients, unlike volunteers in randomised trials, may well not accept randomisation and blinding. Non-randomized, observational trials are more vulnerable to biases and they need to be designed even more meticulously than RCTs to avoid false results. Even then, the validity of observational trials is limited, compared to equally well designed, prospectively planned, randomized trials. We would expect to see more new medicines with unclear benefit-risk-assessments as MAPPs progresses. Currently, it is unclear how more data collected under adaptive pathways can compensate the inherent flaws with such approaches.

MAPPs aim to replace the so called “single magic moment” of approval by progressive management (Eichler et al., 2012; EMA¹⁵). From a payers' point of view, there may well be instances in which adaptive pathways are reasonable and necessary. However, there are concerns that high prices eventually granted for a targeted group of patients will be exploited when the numbers of indications are expanded, with considerable budgetary implications. One of the examples is ivacaftor, which already causes concerns to payers with its high requested price and comparatively limited patient benefits (Eichler et al., 2015; Godman et al., 2015). We are also already seeing a number of orphan medicines achieving “blockbuster” status as a result of such exploitation (Kesselheim et al., 2012). Differential pricing by indication is difficult for payers to cope with especially given limited routine availability of patient level data among countries and regions in Europe, MEAs difficult to administer and the appreciable use of external reference pricing among European countries (Adamski et al., 2010; Leopold et al., 2012; Ferrario and Kanavos, 2013).

In a recent publication, the authors expressed the hope that MAPPs might help to address the strain that high-priced medicines put on healthcare systems (Eichler et al., 2016). This can be disputed especially as the prices requested by pharmaceutical companies for their new medicines rarely mirror their development costs (Experts in Chronic Myeloid Leukemia, 2013). In addition, ever increasing prices are being charged for new medicines despite only a small minority of them considered as innovative by independent authorities (Kelly and Smith, 2014; Godman et al., 2015; Howard et al., 2015). The majority of new medicines are seen as similar or with marginal improvements over existing standards (Godman et al., 2015; Prescrire Editorial, 2016). This is especially the case for new cancer medicines despite similar development costs (Kantarjian et al., 2013). However, cancer medicines are not an isolated case. The prices of new medicines for hepatitis C have stimulated debates over reimbursed prices especially where there are appreciable differences between the costs of manufacture and requested prices (Phelan and Cook, 2014; de Bruijn et al., 2016), especially when the latter threaten health care budgets and their viability (Brennan and Shrank, 2014).

MEAs are an attempt to achieve risk sharing and flexible approaches to reimbursement and are perceived as an integral part of adaptive pathways that already involve payers. Currently, MEAs are not legally feasible in all EU Member States. Consequently, the need for this would result in an adjustment of national legislations. In addition, MEAs are predominantly in the control of pharmaceutical companies. As a result, they control which Member States are involved as well as the different conditions in Member States, depending on the Member State's power of negotiation (Ferrario and Kanavos, 2013). This leads to a distorted market and significantly different levels of access and affordability of new medicines for patients across Europe. From a public health or societal perspective, this is unacceptable. Apart from this, performance-based agreements under MEAs bear high transaction and administrative costs and are not easily implemented (Adamski et al., 2010; Ferrario and Kanavos, 2013, 2015). Adaptive pathways are not designed to lower these hurdles, which further underlines the belief among payers that they need to be low in number and limited to very special cases. Consequently, adaptive pathways cannot be the preferred approach for new medicines in the future.

There also are concerns regarding who should fund any additional costs to healthcare systems arising from higher prices for a new medicine versus existing standards during the conditional approval process as well as fund the data collection process. This is because increasing the amount of data generated post-licensing will result in a shift in costs. In the current environment, pre-marketing clinical trials are funded by pharmaceutical companies alone. If these costs are shifted to payers or the public, pharmaceutical companies need to come to an agreement with health authorities and health insurance companies on how the increased costs to healthcare systems will be compensated. Such costs include introducing comprehensive IT systems where currently it is problematic to collect comprehensive patient level data for the entire population.

A central task in adaptive pathways will be to ensure that exit strategies are in place for new medicines not fulfilling expectations. No application should be complete without a full description of what will happen if patients do not respond to treatment. “No response” should be clearly defined in terms of the hoped-for outcomes and the consequences for all stakeholders. Revoking a marketing authorization has been difficult in the past and we do not expect it to be any easier in the future. This is no different for medicines approved with a conditional approval which do not manage to proceed to a standard approval. We have observed protests from doctors and patients likewise, independent of whether a medicine has been withdrawn or considered for withdrawal due to safety issues, lack of proven benefit or economic reasons (Raftery, 2010; Simoens et al., 2013). It has been indicated that rules for withdrawal could be part of MEAs (Eichler et al., 2015). In our experience, revoking a reimbursement decision can be even more difficult than revoking a marketing authorization; however the degree of difficulty depends on the local legal and cultural environment. Consequently for MAPPs to be workable, conditions for suspending marketing authorization of a new medicine should be included as part of any adaptive licensing process and decision. These should cover situations where there is a worsening of the risk-benefit-balance as more data becomes available and/ or where a company fails to generate the necessary data within defined time frames to continue to support requested prices. This process needs to be well-grounded in European law.

In the context of adaptive pathways, a prohibition of product liability law suits during the initial marketing period has been proposed (Eichler et al., 2012). We see this proposal as unacceptable and not in line with European legislation. It also contradicts the official position of EMA that the current legislative framework is not to be changed. One cannot on the one hand open the way to earlier commercialisation and on the other hand transfer any risks arising from the new medicine to patients and the public domain. This liability prohibition exploits the special vulnerability of ill persons. This is even more questionable in the context of discussions about whether adaptively licensed products require even longer periods of intellectual property protection⁶. Earlier commercialisation, longer periods of exclusivity paired with reduced liability for damages gravely unbalance the concept of adaptive pathways.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.