Endpoints in cancer trials
A cancer trial is measured on how long patients live or on endpoints that arrive sooner and stand in for survival. Overall survival is the reference endpoint and needs long follow-up. Progression and response endpoints arrive sooner but depend on how progression and response are defined and assessed, and response criteria based on tumour size fit targeted therapies poorly. Drugs aimed at a molecular abnormality add pharmacodynamic and molecular endpoints, a phase I trial is measured on safety, and a surrogate endpoint supports a claim only as far as it has been shown to predict clinical benefit.
Overall survival
Most products in cancer development aim to cure or at least extend life rather than to provide purely palliative care, so cancer trials lend themselves to the most solid endpoint available: survival. Overall survival is the gold-standard efficacy endpoint, typically expressed as the proportion of patients alive at one or more time points, as survival over time in a Kaplan–Meier analysis, and as a mean or median duration. It is thoroughly objective and clinically relevant, and regulators prefer it.
Where overall survival is defined in advance, the data are often not mature by the time the report has to be written. For example, if most patients in both groups are still alive, it may be difficult to establish any treatment effect on long-term survival. Conversely, an effect on short-term survival, whether an increase or a decrease, may not have the same effect on long-term survival. Mature survival data and updated survival and efficacy analyses may have to be provided as report addenda or revisions after the first report.
Follow-up in this setting can last almost indefinitely. It may range from simple survival follow-up at regular intervals until the patient has died, to full data collection for patients who continue on trial treatment after the defined treatment period has ended. Extended follow-up is usually incorporated in cancer studies to evaluate the overall survival benefit.
Comparison of survival between groups may also be confounded by a protocol that allows crossover from the control arm to the experimental arm after treatment failure. Once patients leave a trial they may receive a wide range of further lines of therapy for their underlying disease. Data on second- and further-line treatments are ideally collected during follow-up, to evaluate whether such treatments have influenced the efficacy results.
Progression-free survival and time to progression
Reliable overall survival results can require a long follow-up, so other endpoints such as time to progression and tumour response rate are also used. Time to progression, or progression-free survival, is a commonly used endpoint in place of survival itself. Patients undergo regular clinical or radiological investigation, or both, to find indicators of progression such as further growth of the primary tumour or new metastases. Progression-free survival gives results sooner, which is an advantage commercially and also ethically, in that a treatment that may have significant benefits can reach patients sooner.
Where softer efficacy endpoints such as progression-free survival are used, it is common for a blinded independent committee to assess patient data such as X-rays and computed tomography scans to determine progression. The committee serves to overcome investigator bias and any lack of investigator blinding. Toxicity, which may cause characteristic side effects, presents challenges for blinding, and investigators, and sometimes patients, are usually not blinded to trial treatment. Where both investigator and independent assessments are available, it is important to present both. Investigator assessments, whatever the treatment, tend to be more optimistic than independent ones.
Choosing between progression-free and overall survival
Phase III cancer trials typically take either progression-free survival or overall survival as their primary endpoint, and the choice is not straightforward.
Besides arriving sooner, progression-free survival can be strongly argued for on clinical relevance: when patients are in remission and enjoying good quality of life, the time they spend in remission before the disease returns is highly clinically relevant, perhaps even more so than overall survival. It can also be argued that there is little benefit in prolonging the time between the return of incurable cancer and death, whereas prolonging the time spent in good health until the disease returns is self-evidently of great benefit. In the design of one phase III trial of a dendritic-cell therapy for ovarian cancer, that argument won, and progression-free survival was chosen as the primary endpoint.
Progression-free survival has become the preferred endpoint in phase II oncology studies, because the delay before analysis is substantially shorter than for overall survival and interference from salvage therapies is limited. In cancers where life expectancy is short or salvage therapies are ineffective, overall survival may be the most appropriate endpoint. Phase III studies also often use progression-free survival as the primary endpoint, to show the treatment's advantage in delaying disease progression or stabilising the disease. A gain in progression-free survival does not always correlate with a gain in overall survival, however, because of interference from the salvage treatment given after the experimental treatment fails.
In January 2026 the European Medicines Agency recommended extending the indication of Zynyz (retifanlimab), in combination with carboplatin and paclitaxel, to adults with squamous cell carcinoma of the anal canal that cannot be removed by surgery and is metastatic or locally recurrent, on the results of a randomised, double-blind trial in 308 such adults. Half of the patients given Zynyz with chemotherapy lived for 9.3 months or longer without their cancer worsening, against 7.4 months with placebo and chemotherapy. Half of the patients given Zynyz lived for at least 32.8 months, against 22.2 months with placebo, but this difference in survival was not statistically significant, meaning the difference between the groups could have happened by chance. Although there is uncertainty about the extent to which Zynyz prolongs life, the agency concluded that these findings, considered alongside the data on how long patients lived without their cancer worsening, indicate that the medicine has a beneficial effect.
Tumour response criteria
Treatment activity is often measured as the change in the dimensions of the tumour lesions on imaging, by computed tomography or magnetic resonance imaging, assessed against response criteria. These criteria evolved from the World Health Organization classification to version 1.0 of the Response Evaluation Criteria in Solid Tumors (RECIST), which was then upgraded to version 1.1. The RECIST classification is used to assess the tumour response required to evaluate new treatments in oncology. Response to treatment may be assessed against common criteria such as RECIST or the World Health Organization response criteria, or disease-specific response criteria may be defined. The frequency of each response category will be compared between groups and across studies, so response definitions should be standardised within a development programme.
Most phase II oncology studies assess treatment activity on the response rate, either the best response obtained or the response at a given time.
Response criteria for targeted therapies
Response criteria based on tumour dimensions are not ideal for evaluating targeted therapies. With inhibitors of angiogenesis, for example, the size of the target lesions can be unchanged while their density and texture change substantially, particularly through intra-tumour necrosis.
Because the tumour response to an immunotherapy depends on the individual patient's immune function, it does not follow the pattern seen with traditional chemotherapy. With immunotherapies there is often an initial increase in tumour size, possibly in response to inflammation, before eventual shrinkage. In studies of checkpoint inhibitors, an initial temporary increase in tumour size, called pseudoprogression, can be seen, most probably due to lymphocyte infiltration and tumour swelling. Under RECIST version 1.1 an increase of at least 20% in the sum of the tumour lesion diameters would be classified as disease progression, and with a treatment other than an immunotherapy it would lead to a change of treatment strategy. With an immunotherapy, such pseudoprogression may instead be a clear indication of a treatment response.
The anti-tumour response to an immunotherapy is also often delayed compared with that to conventional cytotoxic therapy. For example, melanoma patients treated with ipilimumab continued to respond beyond 24 weeks of treatment, whereas tumour response to chemotherapy usually occurs early in treatment.
To evaluate tumour response with targeted therapies, the established classifications need to be revised. For immunotherapy trials, RECIST as used to assess outcomes such as progression-free survival and objective response rate needs to be adapted. Several new classifications have been proposed: the Choi criteria for tyrosine kinase inhibitors, and irRECIST, iRECIST and immune-related response criteria for immunotherapies. None of them has international consensus, and each needs validation before it can be widely used.
Most immunotherapy trials use the RECIST classification to evaluate tumour response for primary endpoints such as progression-free survival and objective response rate. These traditional endpoints are considered acceptable for regulatory approval. Some of them have been correlated with overall survival and are considered surrogate endpoints. Immunotherapy-specific endpoints, such as immune-related progression-free survival assessed by one of the new classifications, are often included as secondary endpoints. A central review of the imagery used to assess response is also often included. The aim is to evaluate the new classifications for assessing tumour response and to validate their endpoints as surrogates for overall survival.
Molecular and pharmacodynamic endpoints
New cancer drugs are increasingly targeted at specific molecular abnormalities of tumours. Where a treatment is targeted at a specific disease mechanism, biological endpoints may also be assessed. Specific pharmacodynamic endpoints may be evaluated alongside the genetic characteristics of the population or a detailed analysis of tumour characteristics, such as expression of particular genes or the presence of particular mutations. These measurements may be compared between treatment groups, used to define subgroups, or, particularly later in development, used as inclusion criteria.
Phase I endpoints
In the traditional sequence of oncology trials, phase I trials identify the maximum tolerated dose and evaluate the toxicity, pharmacodynamics and pharmacokinetics of the new drug. In an anticancer phase I trial the main endpoints (dose-limiting toxicities, the maximum tolerated dose and the recommended dose for phase II) are all safety variables. The main aim of the trial is to find the maximum tolerated dose and the recommended dose for phase II, through a dose-escalation design whose cohorts are based on the toxicity found with dose increments.
Dose-limiting toxicities are adverse events serious and frequent enough to prevent further administration of treatment, or to prevent dose escalation, and a dose-finding trial typically evaluates them. Dose-limiting toxicities vary between indications and treatments. The protocol should define them carefully, together with the number of such events, or of patients experiencing them. Because a single excess patient with a dose-limiting toxicity can prevent dose escalation, the study population should closely reflect the intended treatment population.
Surrogate endpoints
In the past, new cancer drugs were typically approved on tumour assessment outcomes, which are considered surrogate endpoints, such as objective response rate, duration of response, progression-free survival and time to progression. For several decades regulators across regions have unanimously held that a confirmatory phase III oncology trial should show direct evidence of clinical benefit from the investigational drug, that is longer survival and better quality of life, and that this evidence is intended as the basis for marketing approval. Overall survival and a selected patient-reported outcome such as health-related quality of life have therefore been the gold standard for assessing the efficacy of new cancer therapies.
Surrogate or biological endpoints such as tumour markers may also be assessed in a cancer trial. Both the European Medicines Agency and the Food and Drug Administration have published guidelines on acceptable endpoints in cancer trials.
With the emergence of real-world evidence, better knowledge of the omics of cancer and new transformative therapies that changed the natural history of certain cancers, a discussion arose about reassessing the endpoints in oncology trials for marketing approval. Regulators, while holding to the importance of clinically meaningful endpoints, acknowledged the potential benefits of new treatments based on surrogate endpoints and the need to make them rapidly available to patients with serious or life-threatening cancers. Evidence must, however, be available to justify the ability of a surrogate to predict clinical benefit, for instance how likely it is that patients with a given cancer whose progression is delayed will also live longer. An increasing number of oncology drugs were approved on surrogate endpoints, up to half of them through accelerated approval, with objective response rate and progression-free survival the most common primary endpoints. A company whose drug obtains accelerated approval is required to continue providing post-marketing efficacy data to verify the anticipated clinical benefit.
The counterarguments against overusing surrogate endpoints for marketing approval are equally extensive. Valid evidence that a surrogate predicts long-term overall survival or quality of life is generally lacking and, where it exists, is restricted to a specific tumour type such as advanced colorectal and ovarian cancer. As a result, most oncology drugs approved on surrogate endpoints did not prove clinical benefit. Of identified FDA approvals on surrogate endpoints from 2008 to 2012, 86% either failed to verify long-term overall survival or reported no such data at all. Of FDA approvals from 1992 to 2019, 58% reported no post-marketing efficacy data at all, and for new approvals more than half had no or poor correlation between the surrogate and overall survival. Of EMA approvals from 2009 to 2013, 49% did not show benefit on overall survival or quality of life.
To validate a surrogate endpoint, an Institute of Medicine committee proposed a three-step evaluation.
- Analytical validation: assess whether the surrogate itself can be accurately measured.
- Qualification: assess whether the investigational drug affects the surrogate and the clinical endpoint in a like manner.
- Utilisation: assess the context of the proposed use of the surrogate.
Distinguishing a validated surrogate, with robust estimated net effects of a drug on a clinically meaningful endpoint, from a mere correlate without any established evidence of clinical benefit is painstaking but crucial.
Advantages and disadvantages of the main approval endpoints
Each clinically meaningful and surrogate endpoint has its own advantages and disadvantages, which the FDA's guidance on clinical trial endpoints for the approval of cancer drugs and biologics summarises for the important ones.
| Endpoint | Advantages | Disadvantages |
|---|---|---|
| Overall survival | Easily and precisely measured; generally based on objective and quantitative assessment | May be affected by switch-over of control to treatment or by subsequent therapies; needs longer follow-up; includes non-cancer deaths |
| Symptom endpoints (patient-reported outcomes) | Generally assessed earlier and with a smaller sample size than survival studies | Blinding is important for assessing the endpoint; potentially subject to assessment bias, particularly in open-label studies; lack of validated instruments in many disease areas; definitions vary among studies; balanced timing of assessments among treatment arms is critical |
| Disease-free survival or event-free survival | Generally assessed earlier and with a smaller sample size than survival studies; generally based on objective and quantitative assessment | Potentially subject to assessment bias, particularly in open-label studies; definitions vary among studies; balanced timing of assessments among treatment arms is critical; includes non-cancer deaths |
| Objective response rate | Generally assessed earlier and with a smaller sample size than survival studies; effect on the tumour attributable to the drug, not natural history; generally based on objective and quantitative assessment | Definitions vary among studies; frequent radiological or other assessments; may not always correlate with survival |
| Complete response | Generally assessed earlier and with a smaller sample size than survival studies; effect on the tumour attributable to the drug, not natural history; generally based on objective and quantitative assessment | Definitions vary among studies; frequent radiological or other assessments; may not always correlate with survival |
| Progression-free survival or time to progression | Generally assessed earlier and with a smaller sample size than survival studies; measurement of stable disease included; generally based on objective and quantitative assessment | Potentially subject to assessment bias, particularly in open-label studies; definitions vary among studies; frequent radiological or other assessments; balanced timing of assessments among treatment arms is critical; may not always correlate with survival |