Intended for healthcare professionals


Inappropriate use of progression-free survival in cancer drug approvals

BMJ 2020; 368 doi: (Published 10 March 2020) Cite this as: BMJ 2020;368:m770
  1. Huseyin Naci, assistant professor in health policy1,
  2. Courtney Davis, reader2
  1. 1Department of Health Policy, London School of Economics and Political Science, London, UK
  2. 2Department of Global Health and Social Medicine, King’s College London, London, UK
  1. Correspondence to: H Naci: h.naci{at}

New drugs should be judged on overall survival or quality of life, preferably both

The aim of cancer drugs is to prolong life or improve its quality. Yet only a third of cancer drugs entering the market in Europe and the US have evidence of overall benefits in survival or quality of life.1 Instead, regulators now routinely approve new cancer drugs on the basis of surrogate endpoints that measure disease progression or tumour shrinkage.

This was not always the case. Historically, regulators had little appetite to accept surrogate measures for approving cancer drugs. In the mid-1980s, the US Food and Drug Administration required (with some exceptions) that clinical trials to support new cancer drug approvals show an improvement in patient survival, symptoms, or quality of life. This requirement was based on evidence that the correlation between tumour response and clinical benefit was weak.2

Nevertheless, in 1996 the Clinton administration’s “reinventing government” initiative—which aimed to cut red tape for the benefit of business—announced that the FDA would begin to accept trials using “non-validated” surrogate endpoints such as tumour shrinkage for accelerated approval of new cancer drugs. It was expected that evidence of improved survival or quality of life would be demonstrated later, once drugs were on the market and being prescribed to patients.3

Since then, progression-free survival, defined as time to death or disease progression, has replaced tumour response as the preferred surrogate endpoint in oncology drug trials.4 Both endpoints were developed to identify early signals of drug activity in phase II studies. They were not intended to measure patient benefit.5 Yet progression-free survival is now the most commonly used efficacy endpoint in trials of treatments for advanced or metastatic solid tumours.6 Moreover, European medicines regulators now claim that progression-free survival is not a surrogate for patient relevant outcomes but reflects an intrinsic clinical benefit if a sufficient effect is observed.7

Patient benefit

Does this shift in regulatory standards and research practice reflect an evolution in the scientific community’s understanding of the relation between progression-free survival and patient benefit? A review of the literature suggests not. Although regulators initially hoped evidence would emerge to support the use of progression-free survival as a valid surrogate for patient benefit, that evidence has failed to materialise. Numerous quantitative and qualitative studies reported contradictory findings,8 but large meta-analyses have found that progression-free survival is not a valid surrogate for overall survival or quality of life, even in areas where it was previously thought to be useful.910

In a recent example, authors of a meta-analysis using individual participant data from 17 randomised trials concluded that their study “did not establish progression-free survival as a surrogate endpoint for overall survival in first line treatment of ovarian cancer.”11 Findings from the recent BELLINI trial suggest that cancer drugs that improve progression-free survival may even have detrimental effects on overall survival.12 Taken together, both experience and evidence raise substantial doubts about the value of progression-free survival in the approval process for cancer drugs.

Regulators point to patient demand as justification. However, a recent systematic review questions whether patients value this endpoint.13 Studies examining this question differed in how they defined progression-free survival to patients, reflecting the inherent complexity of this endpoint and the challenge in interpreting changes, even for clinicians and researchers. Regulators claim that “science and standards are here to serve patients and not the other way around.”7 But patients are not well served when regulatory standards disregard science.

Regulatory standards matter. After approval, companies have little or no incentive to evaluate the clinical benefit of their products. Data on overall survival or quality of life rarely emerge, even years after market entry.14 There is a clear need for effective therapies in many advanced cancers. But allowing potentially harmful drugs onto the market without reliable evidence of improvement in the outcomes most important to patients is not a positive development. Especially concerning is the dangerous precedent this practice sets for the regulatory use of non-validated surrogates in other areas. A growing number of oncologists have called for new cancer drugs to be approved on the basis of their overall survival or quality of life benefits, and preferably both.15161718 We agree.



View Abstract