Education And Debate

Clinical endpoints in trials of drugs for cancer: time for a rethink?

BMJ 2002; 324 doi: (Published 08 June 2002) Cite this as: BMJ 2002;324:1389
  1. P P Koopmans, Specialist in internal diseases, clinical pharmacology, and infectious diseases (p.koopmans{at}
  1. Department General Internal Medicine, University Medical Center St Radboud, Nijmegen, Netherlands

    There is a continual debate between health authorities, doctors, researchers, the pharmaceutical industry, and regulatory authorities over the question of when a new drug should be admitted to the market and when it should be prescribed or reimbursed.1 Divergent interests often obscure the scientific debate, but it is in the interest of all parties that valuable and effective drugs should be licensed without undue delay. 2 3 Yet it may take many years (mostly 10-15) from the time when a clinical development plan for a drug (mapping out how to investigate it) is established to its being prescribed for patients. This is frustrating for all drugs but is particularly so for drugs for fatal diseases such as cancer. I argue here that the process can be improved and focus particularly on three issues: the endpoints, side effects, and more collaboration between the parties, and especially the participation of patients.

    Summary points

    The introduction to market of drugs for lethal diseases such as cancer can be improved

    More attention should be given to clinical symptoms and quality of life and less to survival

    Surrogate endpoints and biomarkers should be used as support for proof of effectiveness

    Clinical endpoints such as survival can be investigated after the drug's introduction

    Licensing authorities, pharmaceutical companies, and patients should agree in advance the relevant endpoints and desired effect sizes


    The major points at issue in drugs for cancer are the desired endpoint and the degree of change required over existing treatment.4 Licensing authorities require that drugs have a clinically meaningful effect. In the case of cancer drugs this is often translated as meaning an increase in survival compared with standard treatment. In potentially lethal diseases, however, waiting to assess these clinical endpoints (for example, the effect on five year survival) poses a dilemma as the introduction of a potential valuable drug may be significantly delayed. Kessler and Feiden have argued for a more rapid drug approval process for lethal diseases,2 but in many conditions, including drugs for cancer, the authorities insist on clinical endpoints. On the other hand, the need for endpoints that can be evaluated in a shorter time is widely acknowledged.

    Double standards

    Such endpoints are likely to be surrogate endpoints such as biomarkers of a disease.5 Most licensing authorities accept surrogate endpoints only if the clinical relevance of the surrogate has been well validated. In cancer this is regrettably often not the case.

    On the other hand, changes in biomarkers have been accepted as proof of effectiveness before effects on clinical endpoints were shown. Examples are drugs for hypertension, obesity (orlistat), and hypercholesterolaemia. Highly active retroviral treatment for HIV infection and interferon alfa for viral hepatitis or chronic myeloid leukemia were also introduced to market predominantly because of their effects on biomarkers (viral load, CD4 counts, Philadelphia chromosome, etc) and certainly without survival data. The same holds for the enzyme based therapies for rare storage conditions like Gauchez and Fabry's disease, which were accepted mainly on the basis of pharmacodynamic effects.

    In these cases the assumption was that the biomarkers (blood pressure, cholesterol, body weight, or viral load of HIV, hepatitis C or B virus) indicated disease activity and that influencing them would be beneficial. Examples of diseases for which the assumption turned out to be true are hypertension, HIV infection, and cholesterolaemia, though for others it remains unclear (such as viral hepatitis, obesity).

    Thus a double standard seems to be operating in that well validated surrogates are insisted on for cancer drugs but not for others. Why not make the same assumption for cancer and abandon the demand to show a survival benefit before licensing? Instead the effect of the treatment on survival could be investigated after marketing.

    The arguments for this are, firstly, that in this way more uniform assessment criteria for the benefit : risk ratio of drugs for lethal diseases can be achieved. Secondly, interpreting changes in survival, particularly in clinical trials, is difficult. Many other factors, not related to the disease, may influence survival, such as side effects, confounding factors in clinical trials, and differences between trial populations and those in real clinical practice.6

    Tumour response as a surrogate

    What data then should be available when a drug is introduced on to the market? Beneficial effects based solely on biomarkers may still be too premature. Although scientific data on the meaning of biomarkers for predicting outcome in malignancies are growing, 79 clinical validation is still poor. Nevertheless, changes in biomarkers for malignancies may support an assessment of benefit, particularly in the presence of equivocal clinical data.


    In colorectal cancer tumour response probably can be used as a surrogate endpoint for survival. Reproduced with permission from Buyse et al10

    Would an intermediate or a pharmacodynamic parameter (such as tumour response) be sufficient? Recently, for instance, Buyse et al reported the results of a meta-analysis which showed that in colorectal cancer tumour response probably can be used as surrogate endpoint for survival. 10 11 This result cannot simply be extrapolated to other malignancies because of the heterogeneous nature and pathogenesis of tumours, yet it seems likely that tumour response in most tumour types will indicate clinical benefit.

    Data on the change in tumour size should, however, always be accompanied by data on clinical symptoms. The simple argument for this statement is that patients mostly seek the help of a doctor because of their symptoms and because they want them alleviated. Often the question “how long will I live?” comes afterwards.

    The optimum size of the effect

    Another unresolved point of debate is the desired size of the effect and its durability. For instance the European Medicines Evaluation Agency has developed several guidelines for investigating drugs for cancer, but they give no clear guidance on the desired size of the effect required (

    Even in non-lethal conditions the definition of the optimal effect size is often difficult. In lethal diseases, such as cancer, the optimal response rate or time to progression is also often not well substantiated with scientific data, although an improvement of less than 25% is generally considered not very relevant. Yet for an individual patient even a very short period without symptoms can be of value despite serious side effects. Also a short period can make sense since theoretically “time can be bought” till possible better treatments are found.

    Side effects

    The serious side effects of cytotoxic drugs generally make assessing the benefit : risk ratio complex. Yet the existence of side effects should not preclude licensing of a drug. Most of the usual side effects of cytotoxic drugs are now manageable, for example, with strong antiemetic or antimicrobial drugs. In addition doctors or patients may estimate their gravity in a different way from health authorities.12 We now have better ways of assessing quality of life and patients' preferences and measuring treatment benefit.1215

    Moreover, serious side effects often become apparent only after introduction to the market—as happened with antiretroviral treatment, 16 17 cisapride, and cerivastatin. Thus the fact that few side effects occur during clinical trials does not guarantee safety in general practice or even long term safety.

    Involvement of patients in drug approval

    How can we get the best balance between preventing ineffective drugs coming to market and ensuring that valuable treatments are identified and brought quickly to market? Although collaboration between health authorities and industry has been criticised,18 there are advantages in the licensing authorities trying to achieve agreement with pharmaceutical companies on the requirements for licensing before a costly development plan for a drug is set up. In particular, they could agree on the nature of the primary measure of efficacy, the way it is measured, and, most important, the desired size of the response. If possible the choice of endpoint and size of the effect should be based on scientific data showing its clinical relevance.

    Representatives of the industry, the drug licensing authorities, academia, and also patients should participate in these discussions, in particular when scientific data are lacking. The opinion of patients on the measure of efficacy and the desired size of the effect or the tolerability of side effects can give us more insight into what is relevant and what not. 13 14 Such early discussion may also help avoid arguments afterwards if the outcome of clinical trials is disappointing. Like the Food and Drug Administration in the United States, the European Medicines Evaluation Agency offers the opportunity for pharmaceutical companies to obtain scientific advice on the development plan for drugs. The agency has plans to extend and improve these activities.


    The process for approving drugs for lethal cancers could be improved by paying more attention to clinical symptoms supported by biomarkers, surrogate endpoints, and quality of life data. Assessment of survival data can be postponed to the postmarketing period. In particular, all the parties concerned, including patients, should agree on the assessment criteria before clinical trials are started. These measures should shorten the time to introduce cancer drugs that might make a difference to patients.


    The view of the author does not necessarily reflect the view of the MEB of the Netherlands.


    • Competing interests None.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.
    13. 13.
    14. 14.
    15. 15.
    16. 16.
    17. 17.
    18. 18.
    View Abstract

    Sign in

    Log in through your institution