Intended for healthcare professionals


Conditional approval of medicines by the EMA

BMJ 2017; 357 doi: (Published 02 May 2017) Cite this as: BMJ 2017;357:j2062
  1. Rita Banzi, researcher1,
  2. Chiara Gerardi, researcher11,
  3. Vittorio Bertele’, researcher1,
  4. Silvio Garattini, director1
  1. 1IRCCS-Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italy
  1. Correspondence to: R Banzi rita.banzi{at}

The process fails to improve the evidence after early licensing

Like other expedited procedures for early marketing authorisation, conditional approval12 stems from the assumption that immediate availability of new drugs offers patients a benefit that outweighs the risks of limited clinical information. To some extent, this assumption applies to all licensed drugs, but early approval magnifies the need for “managing the unknown.”34

Earlier this year, the European Medicines Agency reported its experience so far with conditional approval of new medicines.5 The report is an important step towards greater transparency and contributes useful data to help reflections on two important questions: what is the minimum evidence acceptable at the time of early approval, and do postmarketing commitments improve the evidence in a meaningful way?

In response to the first question, between 2006 and 2016 the EMA granted conditional approval to 30 medicines on the basis of 58 pivotal studies. The regulator reports that over half of these studies were preliminary phase II studies (30, including phases I/II and IIb); only a third (21) were phase III trials,5 and many of these were short term or analysed before completion of the study.6 Half the studies (33) were open label, and more than a third (20) had an uncontrolled single arm design.5 Only two trials (of pixantrone and crizotinib) included an active control.5 Most studies chose response rate as the primary outcome, and reported few clinically meaningful outcomes or none.6

At best, this kind of clinical research allows informed betting on the beneficial effects and harms of a new medicine, and hardly supports a realistic view of its true risk-benefit profile.78 Conditional approvals share some parallels with the early termination of clinical trials, which are known to overestimate treatment benefits and underestimate harms.910

To answer the second question, the EMA required drug manufacturers to provide more data from clinical studies in nearly three quarters of specific postmarketing obligations imposed for drugs conditionally approved between 2006 and 2016 (77/107). A third of these requests (25) were for studies that were more preliminary than phase III or uncontrolled single arm studies, or both. Two thirds (56) were for open label studies.

The EMA asked for at least one comparative study (with active or placebo control) for eight drugs that did not have evidence from comparative trials at the time of the authorisation. But the manufacturers of five products with no comparative trials were not given this obligation. So clinicians and patients have no comparative data on these agents at all. Since the final aim of innovative treatments should be a longer and better life for patients, it is surprising that out of the 77 studies requested, only nine—all oncology studies, not necessarily randomised—reported overall survival as the primary outcome, and not one reported quality of life.5

According to the EMA report, the main reasons for allowing manufacturers extra time to complete their specific obligations (extending the deadline by more than a year in a tenth of cases), were slow recruitment or difficulties in activating clinical sites.5 After licensing, there can be opposition to starting new trials, or even finishing ongoing trials, from patients, clinical investigators, and even ethics committees. If a drug is licensed, particularly if its conditional approval is understood as breakthrough rather than incomplete,11 equipoise may be questioned and randomisation may seem inappropriate. But the weaker and biased study designs that can be used as alternatives to randomisation are unlikely to fill the evidence gaps that were evident at the time of approval.

Although the main findings reported by the EMA do not differ from those we reported in our 2015 analysis,6 the interpretation does. The EMA sees the glass half full. However, it is not clear from the report how far the fulfilment of the specific obligations helped to upgrade clinical knowledge and evidence. It is hard to believe that a few more trials—mostly phase II, open label, and reporting outcomes of questionable clinical importance, possibly against inappropriate comparators—can fill the knowledge gap at the time of conditional approval.

The EMA emphasises that conditional approval has brought forward authorisation of medicines by an average of four years.5 Any suggestion that early access to medicines without comprehensive data benefits patients should be supported by evidence. Even when manufacturers do eventually provide comprehensive clinical data, it can hardly be considered as ethically and clinically appropriate that patients and their doctors have been unaware of the benefit-risk profile of medicines they have been using for a long time—over seven years in some cases.

Regulatory processes should be designed first and foremost to establish the clinical value of new drugs, to satisfy the health needs of patients and the public. As early approvals, including conditional approvals, cannot serve this aim, they should be regarded as exceptions. When considering conditional approvals, the EMA should require better evidence of safety and efficacy at the time, and insist on even more convincing data before full approval.


  • Competing interests: We have read and understood BMJ policy on declaration of interests and have no relevant interests to declare.

  • Provenance and peer review: Not commissioned; externally peer reviewed.


View Abstract