If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding authors. Addresses: Section of Gastroenterology and Hepatology, Department of Health Promotion, Mother and Child Care, Internal Medicine and Medical Specialties PROMISE, University of Palermo, Piazza delle Cliniche n 2, 90127, Palermo, Italy
Validated surrogate endpoints for overall survival (OS) are important for expediting the clinical study and drug-development processes. Herein, we aimed to validate objective response as an independent predictor of OS in individuals with unresectable hepatocellular carcinoma (HCC) receiving systemic anti-angiogenic therapy.
When evaluating efficacy in oncology trials, overall survival (OS) provides an unequivocal measure of true clinical benefit. OS, defined as the time from random assignment to the date of death from any cause, is a precise, objectively measured, and easy-to-interpret clinical outcome. It is considered the most reliable and clinically meaningful endpoint to evaluate drug efficacy in oncology trials.
However, the interpretation of OS can be affected by post-progression survival and/or subsequent cancer treatments, potentially confounding survival analysis (treatment crossover or second-line therapies), which has led to the use of several intermediate parameters based on radiology findings that are supposed to be surrogates of survival benefit.
These include objective response rate (ORR), progression-free survival (PFS) and time-to-progression, which have been extensively used in oncology trials despite their controversial value in predicting survival benefit, irrespective of cancer type. Indeed, a major rate of response with long term duration has led to accelerated approvals in the field of hepatocellular carcinoma (HCC). However, their relevance remains debated,
The most commonly evaluated surrogate endpoints in the field of HCC are generally related to radiological tumour response, and they might be classified as time-independent (e.g. ORR) or time-dependent (e.g. PFS) endpoints, whether they are able to capture the time to the event or not.
suggesting that, at the individual level, a patient treated with sorafenib or lenvatinib whose tumour responds or whose time to disease progression is longer lives longer. Median OS was 21.6 months for responders compared to 11.9 months for non-responders (hazard ratio 0.61; 95% CI 0.49-0.76; p <0.001). OS benefit significantly persisted in responders at 2-month and 6-month landmarks. Interestingly, when using modified RECIST criteria (both for investigator or independent radiologic review) but not RECIST criteria, non-responder patients with stable disease had a median survival closer to responders. One is thus prompted to ask, what is the true driver of survival outcome?
is relevant, particularly considering that, for a long time, tumour “shrinkage” (objective response) was postulated to be clinically relevant and to translate into enhanced survival, but adequate data were absent. Accordingly, the current data represent a major piece of information. However, the study focuses on lenvatinib and sorafenib in a single trial data set. Would it also be the case under other agents or in a separate population at a different evolutionary time point? Furthermore, should the decision to transition from early phase evaluation to phase III trials be based on evaluation of response?
It is important to note that tumour burden measurements by any system were intended simply to describe what happens to tumours during therapy—not to infer a meaningful benefit from those changes. Reduction in tumour burden is a marker of activity, but activity is not to be ruled out in the absence of response as previously exemplified by sorafenib.
Moreover, a significantly higher ORR does not always automatically translate into a proportional survival benefit. For example, the ORR of durvalumab plus tremelimumab was 4 times higher than that of sorafenib,
reaching a rate of 20% for the combination, but this significant ORR difference translated into a median OS difference of only 3 months between the two treatments. The same discrepancy was observed in the trial comparing the efficacy of lenvatinib vs. lenvatinib plus pembrolizumab. ORR increased from 17.5 to 26.1 but OS was not significantly different.
In the context of sequential treatments, time-to-event outcomes such as PFS are proposed to have great clinical utility owing to the fact that PFS better captures timing of transition to second-line therapy after progression.
Its interpretation cannot be confounded by post-progression survival and treatment crossover but the surrogacy of PFS has not been validated. Also, in this case, it is possible to observe differences between the radiological endpoint and the survival benefit, as for COSMIC-312 trial
on the other hand, the high toxicity led to a significantly reduced survival compared to sorafenib while maintaining the PFS substantially unchanged. Therefore, the importance of validating PFS at an individual level with respect to OS is clear.
explored the surrogate value of PFS and found a moderate to strong individual-level association between PFS and OS, that does not suffice to endorse PFS as a robust endpoint for future trials. Indeed, it is important to consider that the clinical value of PFS could be weakened by the fact that it does not consider the pattern of progression,
which has been proven to be an important (and extensively validated) survival predictor. Moreover, different classes of drugs could have different impacts on the pattern of progression and consequently on survival. Finally, the so-called pseudo-progression phenomenon, associated with immune checkpoint inhibitors (ICIs), may further reduce the value of PFS, as response may be registered later during follow-up. Therefore, since the evaluation of the benefit associated with ICIs is increasingly complex, traditional radiological response criteria, such as modified RECIST or RECIST, may no longer be appropriate. The accuracy of specifically designed radiological response criteria for ICIs, such as immune-RECIST, remains to be extensively explored in the field of HCC
as ICIs are now the backbone of systemic HCC treatment. Moreover, from a methodological point of view, the late pattern of response associated with ICIs, with a proportion of patients obtaining a long-term survival benefit, may violate the assumption of proportional hazards, making median OS and hazard ratio a potentially flawed measure to assess treatment benefit. In this line, milestone analysis or restricted mean survival time could overcome the issue of non-proportional hazard ratios and they should be routinely reported in ICI trials for HCC.
Other fundamental issues affecting evaluation of radiological response to treatment are reliability due to interobserver variability and use of several radiologic criteria across trials. Particularly in the setting of systemic therapies, the advantage of one system over the other is still to be validated and it might be more useful to use the one that requires the fewest measurements.
These comments highlight that caution should be exercised when evaluating benefit in terms of radiology-based endpoints which do not translate into clear OS benefits; this is particularly relevant if there is no adequate information on the withdrawal rates for adverse events (AEs), duration of treatment, safety measures, pattern of progression, and further treatment lines after progression.
AEs are usually presented as a frequency at the end of follow-up (as time-independent variables), without describing their occurrence over time. Thus, treatments that improve survival may appear to cause a disproportionate rate of AEs, as data accumulate over a longer follow-up time than for alternative treatments that confer less survival benefit. Moreover, treatment toxicity may overestimate PFS if a substantial proportion of patients dropout for AEs before radiological progression could be documented, introducing an informative censoring bias.
All these issues should be considered in future HCC clinical trials. Lack of inclusion of hepatic decompensation as an outcome (which could play a role as a competing risk)
partially address this area of need, evaluating the fine balance between antitumor drug efficacy and hepatic toxicity, considering death due to hepatic failure as a competing risk. However, hepatic decompensation, type of decompensation during follow-up, and decompensation-free survival, are not clearly assessed. A careful evaluation of the treatment failure (due to severe AEs that also consider hepatic decompensation) would avoid misleading conclusions resulting from uneven dropout rates, and failure to tolerate a drug is as much a sign of ineffective treatment as tumour progression.
In summary, demonstrating associations between the surrogate endpoints (ORR and PFS) and OS at the individual level is a necessary but not sufficient condition to conclude that ORR and PFS can replace OS for the assessment of clinical benefit in oncology trials.
While use of intermediate surrogate endpoints provides rapid assessments of experimental interventions, timeliness should not be achieved at the expense of a misleading risk-benefit profile. Moreover, the highest level of validation of surrogacy requires an analysis of correlation in terms of treatment effects between arms based on data from randomised-controlled trials (trial-level association). Also, surrogacy must be validated for all pharmacological approaches and lines of treatment. In the end, reliable blood- or tissue-based biomarkers that can accurately predict clinical benefit (i.e. improved survival) are an urgent unmet need. While waiting for such treatment-specific tools, the prediction of benefit in an individual patient or the assessment of activity in early phase trials will remain an educated guess far from precision oncology.
The authors received no financial support to produce this manuscript.
Conflict of interest
Giuseppe Cabibbo has served as a consultant or on advisory boards for Bayer, Eisai, Ipsen, AstraZeneca, Roche. Jordi Bruix has consulted for Arqule, Bayer-Shering Pharma, Novartis, BMS, BTG- Biocompatibles, Eisai, Kowa, Terumo, Gilead, Bio-Alliance, Roche, AbbVie, MSD, Sirtex, Ipsen, Astra-Medimmune, Incyte, Quirem, Adaptimmune, Lilly, Basilea, Nerviano, Sanofi; and received research/educational grants from Bayer, and lecture fees from Bayer-Shering Pharma, BTG-Biocompatibles, Eisai, Terumo, Sirtex, Ipsen.
Please refer to the accompanying ICMJE disclosure forms for further details.
Interpretation of data and drafting of the manuscript (all authors); critical revision of the manuscript for important intellectual content (all authors). All authors approve final version of the manuscript.
The following are the supplementary data to this article: