Progression-free survival as a surrogate marker of overall survival

Is it the good question?


  • Bernard Escudier MD

    Corresponding author
    1. Department of Medical Oncology, Institut Gustave Roussy, Villejuif, France
    • Department of Medical Oncology, Institut Gustave-Roussy, 114 rue Edouard Vaillant, 94805 Villejuif, France

    Search for more papers by this author
    • Fax: (011) 33142115211

  • See original referenced article on pages 2637–42, this issue.


The use of progression-free survival as a surrogate marker of efficacy in cancer treatment has been the subject of much discussion in recent years. However, it should not be considered a surrogate marker for overall survival.

The use of progression-free survival (PFS) as a surrogate marker of efficacy in cancer treatment has been the subject of much discussion in recent years. A similar debate occurred in the 1990s regarding the use of response rates as a surrogate for overall survival (OS), but it is now accepted that such a correlation is weak.1, 2 There are several reasons for the use of response rates or PFS as surrogate markers of OS in cancer.

OS is the gold standard marker of efficacy for any cancer treatment and it is self-evident that any therapeutic intervention that can prolong survival should be considered to be effective. However, if second or subsequent lines of treatment are significantly active, then these can obscure the observed benefit on survival afforded by the new effective intervention.

In many cancers, OS benefits can take time to become evident but getting evidence that a new treatment is active as soon as possible is obviously very important for the oncology community in general and, of course, the patients in particular. Every effort should be made to obtain early evidence of efficacy or indeed the opposite.

PFS is now considered to be a reproducible endpoint, especially with the independent radiological review that is now frequently required by both the US Food and Drug Administration (FDA) and European Medicines Agency (EMA). It is interesting to note that in many large phase 3 trials, there is little difference noted between independent, centrally reviewed PFS and investigator-assessed PFS. This supports the validity of PFS as an endpoint.

PFS has now been formally accepted as a valid primary endpoint by the FDA in colon cancer, after published evidence that PFS was correlated with OS in patients with this disease.3, 4

To my knowledge, in metastatic renal cell carcinoma (mRCC), the statement that PFS is correlated with OS has never been made. However, all the recently approved drugs in the treatment of mRCC, apart from temsirolimus, were approved based on PFS benefit, including sorafenib, sunitinib, bevacizumab, everolimus, and, recently, pazopanib.

In a recent issue of Cancer, Heng et al5 concluded that PFS is a predictor of OS in patients with mRCC who are treated with the new targeted agents. Their conclusion is drawn from a retrospective analysis of 1158 mRCC patients treated with targeted agents at 12 institutions. There is no doubt that in their cohort patients with a short PFS were found to have a shorter OS compared with patients with a long PFS, but whether it was a treatment effect on PFS that prolonged OS in their series of patients was not addressed.

There are several weakness to the study by Heng et al,5 some of which are highlighted by the authors themselves. First, only 70% of the patients received targeted agents as first-line treatment and therefore were really eligible for the analysis performed. Second, PFS was not independently reviewed and outside of clinical studies, which was the case for the majority of the patients, the treating clinicians are often biased and may overestimate PFS in patients with slow-growing disease. Consequently, such patients may have a longer than expected OS. Finally, only 35% of the patients in the study by Heng et al received additional (ie, subsequent) therapy,5 which is different from the current real-life situation, now that so many drugs have been approved.

The question of whether PFS is a good surrogate marker for OS in patients with mRCC has been quite well answered by the phase 3 trials that have led to the approval of the novel compounds currently under investigation. None of them demonstrated a statistically significant improvement in OS in an intent-to-treat analysis. Thus, based on these large studies, it should be concluded that PFS is not a good surrogate marker of OS. The real question is whether PFS is a good relative surrogate marker of the efficacy of one drug compared with another. The answer to this question in my view is yes, and the change in OS in mRCC from the late 1990s to the present day is the best evidence of this positive answer. Thanks to the efficacy of the many new drugs used in the treatment of mRCC, it is unlikely that a single drug will improve OS, except perhaps as third- or fourth-line treatment. This reality has been recognized by regulatory and funding authorities in many countries when they have given approval for the use of the currently available targeted agents for the treatment of mRCC.

We should accept that PFS is a much better surrogate marker of the true efficacy of a drug than OS, particularly in diseases in which multiple lines of treatment are given. This view begs the question: is the order in which patients receive active drugs important? Unfortunately, this important question was not addressed by Heng et al.5


The author made no disclosures.