The importance of the initial response to cancer treatment in predicting longer overall survival

KEYWORDS:

• Depth of response
• CNS
• targeted therapy
• overall survival
• NSCLC

1. Introduction

As cancer therapies have become more biomarker-based and successful, traditional clinical endpoints such as overall survival (OS) and progression-free survival (PFS) have become increasingly challenging to use as the primary endpoints for clinical trials due to their prolonged time to maturity. PFS and OS are now often collected as secondary outcomes in early phase studies and are the primary outcome in confirmatory studies done after drug approval while surrogate markers of efficacy, that result more quickly, have become commonplace as primary endpoints. The most routinely used surrogate endpoint is objective response rate (ORR). This endpoint may also have limits in distinguishing highly effective therapy as it is difficult to discriminate oncogene-targeted therapies that have reached the upper limit of efficacy. For example, in the head-to-head comparison of first-line crizotinib to alectinib, the ORR was similar, 75.5 versus 82.9 (p = 0.09), while the PFS was significantly different at 10.4 months versus 25.7 months (p < 0.001) [1]. Similarly, in the FLAURA trial, the ORR in patients with EGFR activating mutations was nearly identical at 80% in the osimertinib cohort and 76% in the standard-of-care cohort; however, the PFS and duration of response both demonstrated significant differences between the treatment cohorts [2]. Therefore, with the rapidly changing landscape of oncology therapeutics it is important to continue to develop and validate additional outcome measures that provide early insight into clinical benefit.

Determination of treatment efficacy requires systematic evaluation of disease sites before and during treatment. Staging systems for cancer are one of the basic elements for determining disease status and prognosis. Additionally, staging guidelines are continuously updated to be the most accurate reflection of patient prognosis. By design, within the Tumor, Node, Metastasis (TNM) staging system, the worst prognosis corresponds with patients who have the most advanced disease. If we can maximally reduce tumor burden by interventions such as systemic treatment, surgery, or radiation, we can potentially alter the natural course of disease and improve survival. Given this, it is critical to obtain maximal tumor reduction of the primary disease sites with treatment.

Investigators of hematologic malignancies have been utilizing maximal response to systemic therapy as a measure of treatment efficacy for many years [3–5]. Our group, and others, have more recently begun to explore the utility of reporting depth of response (DepOR) to therapy as a surrogate outcome to predict longer overall survival in non-small cell lung cancer (NSCLC) and other solid tumors. The value of DepOR lies in the fact that compared with ORR it maintains the granular patient-to-patient view of magnitude of response to treatment instead of dichotomizing patients into responders (≥30% shrinkage of target lesions) versus non-responders based on RECIST criteria [6]. We propose that by utilizing DepOR instead of ORR we incorporate a valid surrogate marker while retaining more detailed patient response data.

2. DepOR across solid tumor types

In our recent publication, we demonstrated that both PFS and OS are significantly associated with the DepOR in metastatic, NSCLC in response to either targeted therapy (in the setting of ALK fusions) or PD-1 inhibitor treatment after platinum doublet [7]. In another study of advanced, ALK positive NSCLC, brigatinib, another ALK tyrosine kinase inhibitor (TKI), also demonstrated a significant association between DepOR and PFS in most response quartiles [8]. Soria et al. also recently reported on DepOR as a secondary outcome measure in the FLAURA trial and found that DepOR was significantly greater in the osimertinib arm compared with standard of care (−54.7% vs. −48.5%; p = 0.003) [2]. Also in lung cancer, Morgensztern et al. identified a significant association between DepOR and survival in patients treated with platinum doublets [9]. Conversely, Wu et al. did not find a significant association in EGFR-mutated patients treated with first-line EGFR TKI both in individual patients using a Spearman rank correlation and in a study level analysis in which the patient population was split into two cohorts, those with greater than 50% or less than 50% DepOR [10]. These findings illustrate both retrospectively and prospectively the value of using DepOR as a treatment endpoint as well as the challenge of using one endpoint across different treatment modalities and different cancer molecular subtypes.

As noted above, in the study by Wu et al. in EGFR-mutated NSCLC, there was not a significant correlation between DepOR and survival observed with an EGFR TKI. Interestingly, the team did note that CNS progression was associated with OS. This illustrates that the association between DepOR and survival is dependent on tumor exposure to treatment. This is particularly pertinent in diseases like NSCLC where the CNS is a common location for metastatic spread. Despite similar mechanisms of action, not all therapies have effective CNS penetration. We speculate that the association between DepOR and PFS and OS may be more robust when reliably CNS penetrant treatments are utilized or when comparison between two agents with similar CNS activity is performed. Further, cytotoxic chemotherapy has unreliable CNS penetration while immunotherapy has some early reports of CNS activity [11].

In metastatic colorectal cancer (mCRC), the values of DepOR and early tumor shrinkage (ETS) have been retrospectively evaluated. In the CRYSTAL and OPUS studies, mCRC patients with wildtype KRAS were treated with chemotherapy plus cetuximab or chemotherapy alone [12,13]. In post hoc analyses, patients treated with combination therapy had a significantly higher median DepOR and ETS compared with patients treated with chemotherapy alone [14,15]. In a prospective study of 54 Japanese patients with mCRC (KRAS wildtype) treated with FOLFIRI plus cetuximab, Tsuji et al. demonstrated that DepOR was associated with OS and post-progression survival (PPS) on a study level. They also determined that the association between DepOR and OS and PPS in patients with the largest lesions (≥5.05 mm) was driving this response [16]. In another example, Cremolini et al. demonstrated that DepOR and ETS were significantly associated with survival in mCRC patients treated with FOLFOXIRI + bevacizumab or FOLFIRI + bevacizumab and that the patients treated with FOLFOXIRI + bevacizumab had a larger DepOR and better survival, and DepOR was equivalent to RECIST in predicting survival [17].

In gastric cancer, one of the challenges of treatment has been to identify patients who will respond to trastuzumab-based first-line therapy within the HER2 positive cohort. One group measured DepOR and ETS in a cohort of 61 patients to investigate the predictive value of DepOR and ETS on long-term outcomes and define discriminatory cutoff values to guide treatment. They found that on a trial level DepOR was predictive of PFS, OS, and PPS and while ETS demonstrated a trend of correlation with these three measures the association was not significant, though the sample size was small [18].

The above studies clearly demonstrate that DepOR is a valuable surrogate measure of treatment response across several solid tumor types. ETS has also been demonstrated to be a potentially valuable measure though prospective investigations in larger cohorts are needed. ETS could be the most efficient readout of clinical activity for treatment, if validated, as it would be helpful for rapid go/no-go determinations in multi-arm clinical trials. Measurement of ETS at a prespecified time point and subsequent correlation with DepOR also avoids the issues of survival bias, though other biases may emerge that need to be accounted for (e.g. guarantee-time bias). Additionally, as in the FLAURA study described above, duration of response was also reported. Duration of response is a measure of time from documented response to progression and thus is a more refined view on response to disease. Though not evaluated, we speculate that this measure, like PFS, would also be correlated with DepOR. However, in the setting of increasingly effective first-line therapies this outcome measure is subject to similar limitations as PFS and will still suffer from extended read-out times.

3. DepOR with different therapeutic mechanisms of action

With the growing list of cancer treatment options, it is important to consider the validity of DepOR not only across tumor types but also across different therapeutic mechanisms of action. Tumor response to treatment can be via direct mechanisms such as by cytotoxic chemotherapies that cause cell death by damaging DNA or preventing formation of nucleic acids. Alternatively, targeted therapies are a large subcategory with numerous mechanisms of action. In the studies discussed here, kinase inhibitors are small molecule inhibitors that block growth and development signaling pathways, leading to cell death. Immunotherapies are a third broad therapeutic category that work less directly. Anti-cytotoxic T lymphocyte associated protein 4 (CTLA-4) or anti-programmed cell death-1 antibodies (PD-1) function by enhancing immune activation and T-cell proliferation. In addition, unlike chemotherapy and most targeted therapies, this treatment may not immediately translate to tumor cell death and the treatment effect may persist, even after cessation of treatment. Finally, biologic agents like cetuximab and bevacizumab work by activating antibody-dependent cellular cytotoxicity and complement mediated cytotoxicity in addition to direct antiproliferative effects. Each of these mechanisms result in different tempos of cell death. Given this, it is important to consider whether DepOR metric can be used to evaluate treatment response in therapies with different mechanisms of action.

In lung cancer, targeted therapy, chemotherapy, and immunotherapy all demonstrated significant correlations to PFS and OS. Interestingly, the interval improvements in PFS and OS we observed with increasing tumor shrinkage across quartiles in the targeted therapy trials were not seen in the immunotherapy trial cohorts. This may reflect the different mechanisms of action of immunotherapy compared with targeted therapy. In colon and gastric cancers, the treatments utilized combinations of chemotherapy and biologics and despite different analyses, the same pattern of interval improvement with increasing DepOR is evident. Therefore, these trials demonstrate that the utility of DepOR is maintained across different treatment mechanisms of action; however it may be that DepOR needs to be measured differently depending on the mechanism of action of the drug being studied, for example, using different cut-points or quantification of DepOR for targeted therapy versus immunotherapy.

In summary, markedly improved tumor response rates for oncogene-targeted therapies have allowed the exploration of additional tumor response criteria. Indeed, as therapies and therapeutic combinations become more effective, the field might aspire to measuring complete responses, which currently are rare. Furthermore, as the array of treatments for solid tumors continues to expand and the diverse mechanisms of action are characterized, we must continue to consider the utility of different outcome measures. This will allow a dynamic process of drug development and approvals that will evolve rapidly to help increase patient access to effective therapies. DepOR and ETS are newer surrogate endpoints that should be incorporated into clinical trial design to continue their prospective validation.

Declaration of interest

C McCoach discloses honoraria from Takeda. R Doebele discloses the following: advisory board, consulting, or honorarium from AstraZeneca, Ignyta, Ariad, Takeda,Guardant Health, Spectrum Pharmaceutical. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.