Michelle K. Wilson

Outcomes and endpoints in trials of cancer treatment: the past, present, and future

Cancer treatment should allow patients to live better or longer lives, and ideally, both. Trial endpoints should show clinically meaningful improvements in patient survival or quality of life. Alternative endpoints such as progression-free survival, disease-free survival, and objective response rate have been used to identify benefit earlier, but their true validity as surrogate endpoints is controversial. In this Review we discuss the measurement, assessment, and benefits and limitations of trial endpoints in use for cancer treatment. Many stakeholders are affected, including regulatory agencies, industry partners, clinicians, and most importantly, patients. In an accompanying Review, reflections from individual stakeholders are incorporated into a discussion of what the future holds for clinical trial endpoints and design.

Outcomes and endpoints in cancer trials: bridging the divide

Cancer is not one disease. Outcomes and endpoints in trials should incorporate the therapeutic modality and cancer type because these factors affect clinician and patient expectations. In this Review, we discuss how to: define the importance of endpoints; make endpoints understandable to patients; improve the use of patient-reported outcomes; advance endpoints to parallel changes in trial design and therapeutic interventions; and integrate these improvements into trials and practice. Endpoints need to reflect benefit to patients, and show that changes in tumour size either in absolute terms (response and progression) or relative to control (progression) are clinically relevant. Improvements in trial design should be accompanied by improvements in available endpoints. Stakeholders need to come together to determine the best approach for research that ensures accountability and optimises the use of available resources.

A phase II study of single-agent RO4929097, a gamma-secretase inhibitor of Notch signaling, in patients with recurrent platinum-resistant epithelial ovarian cancer: A study of the Princess Margaret, Chicago and California phase II consortia

PURPOSEnA phase II study was performed to evaluate the efficacy and safety of single-agent RO4929097 (a gamma-secretase inhibitor) in patients with recurrent platinum-resistant ovarian cancer.nnnEXPERIMENTAL DESIGNnWomen with progressive platinum-resistant ovarian cancer treated with ≤2 chemotherapy regimens for recurrent disease were enrolled in this trial. Patients received oral RO4929097 at 20 mg once daily, 3 days on/4 days off each week in a three week cycle. The primary endpoint was progression-free survival (PFS) rate at the end of 4 cycles. Secondary objectives included assessment of the safety of RO4929097 and exploration of molecular correlates of outcome in archival tumor tissue and serum.nnnRESULTSnOf 45 patients enrolled, 40 were evaluable for response. Thirty-seven (82%) patients had high-grade ovarian cancer. No objective responses were observed. Fifteen patients (33%) had stable disease as their best response, with a median duration of 3.1 months. The median PFS for the whole group was 1.3 months (1.2-2.5). Treatment was generally well tolerated with 10% of patients discontinuing treatment due to an adverse event. In high grade serous ovarian cancer patients, the median PFS trended higher when the expression of intracellular Notch (NICD) protein by immunohistochemistry was high versus low (3.3 versus 1.3 months, p=0.09). No clear relationship between circulating angiogenic factors and PFS was found despite a suggestion of an improved outcome with higher baseline VEGFA levels.nnnCONCLUSIONSnRO4929097 has insufficient activity as a single-agent in platinum-resistant ovarian cancer to warrant further study as monotherapy. Future studies are needed to explore the potential for cohort enrichment using NICD expression.

Stage I granulosa cell tumours: A management conundrum? Results of long-term follow up

UNLABELLEDnOptimal management of women with early stage granulosa cell tumours (GCT) presents a management conundrum - they have excellent prognosis but a third will relapse. Advances uncovering the molecular characteristics of GCT have not been matched by improvements in our understanding and treatment.nnnMETHODSnStage I GCT patients referred to Auckland City Hospital (1955-2012) and Princess Margaret Cancer Centre (1992-2012) were identified. Baseline characteristics, histopathology and outcomes were recorded retrospectively.nnnRESULTSnOne hundred and sixty stage I GCT patients were identified with a median age of 49 years. Median follow-up was 7.0 years (range 0.1-44.2 years). Fifty-one patients (32%) relapsed with a median time to relapse (TTR) of 12.0 years (1.3-17.7 years) - 20 initial relapses occurred 10 years post-diagnosis. Higher relapse rates (43% vs. 24% p=0.02) and shorter TTR (10.2 vs. 16.2 years p=0.007) were seen with stage Ic versus stage Ia disease. Cyst rupture was associated with increased relapse (p=0.03). Surgery was the main therapeutic modality at relapse. Eighty six percent of patients received non-surgical management at least once post-relapse. Clinical benefit rate was 43% with chemotherapy, 61% with hormonal therapy and 86% with radiation. Five- and 10-year overall survival (OS) were 98.5 and 91.6%, respectively. Median OS was similar in patients with (24.3 years) and without relapse (22.3 years).nnnCONCLUSIONnSurgery remains fundamental at diagnosis and relapse. Caution should be exercised in recommending adjuvant chemotherapy at initial diagnosis given median OS was greater than 20 years even with relapse. Hormonal therapy at relapse appears encouraging but needs further assessment. Novel treatment strategies need exploration with international collaboration essential for this.

Genotype-matched treatment for patients with advanced type I epithelial ovarian cancer (EOC)

BACKGROUNDnGenomic alterations that activate the MAPK signaling pathway frequently occur in Type I Epithelial Ovarian Cancers (EOCs). We evaluated therapeutic response outcomes in patients with type I EOC treated with genotype-matched therapy on clinical trials enrolled in a prospective molecular profiling program.nnnMATERIAL AND METHODSnFormalin fixed paraffin embedded tumor tissues were prospectively screened for genomic alterations using MALDI-ToF mass-spectrometry platform or targeted sequencing using the Illumina MiSeq TruSeq Amplicon Cancer Panel. Treatment outcomes on genotype-matched trials were retrospectively reviewed using RECIST version 1.1 and Gynecological Cancer Intergroup CA125 related-response criteria RESULTS: 55 patients with type I EOC underwent molecular profiling, 41 (75%) low grade serous (LGS), 9 (16%) clear cell (CC), and 5 (9%) mucinous (MC) histologies. Thirty-five patients (64%) were found to have ≥1 somatic mutations: 23 KRAS, 6 NRAS, 5 PIK3CA, 2 PTEN, 1 BRAF, 1 AKT, 1 TP53, and 1 CTNNB1. Fifteen patients were subsequently enrolled in genotype-matched phase I or II trials, including 14 patients with KRAS/NRAS mutations treated with MEK inhibitor targeted combinations. Among 14 RECIST evaluable patients, there were 7 partial responses (PR), 7 stable disease (SD) and 1 disease progression (PD). CA125 responses were observed in 10/10 evaluable KRAS/NRAS mutant patients treated with MEK inhibitor combinations CONCLUSIONS: Genotyping and targeted sequencing of Type I EOCs frequently identifies actionable mutations. Matched treatment with MEK-based combination therapy in KRAS and/or NRAS mutant type I EOC patients is an active therapeutic strategy.

Estimation of expectedness: predictive accuracy of standard therapy outcomes in randomized phase 3 studies in epithelial ovarian cancer.

The anticipated clinical outcome of the standard/control arm is an important parameter in the design of randomized phase 3 (RP3) trials to properly calculate sample size, power, and study duration. Changing patterns of care or variation in the study population enrolled may lead to a deviation from the initially anticipated outcome. The authors hypothesized that recent changes in patterns of care in epithelial ovarian cancer (EOC) have led to challenges in correctly estimating the outcome of control groups.

Abstract P3-09-05: Clinical outcome of patients with advanced triple negative breast cancer with germline and somatic variants in homologous recombination gene

Background: Variants in homologous recombination (HR) genes other than BRCA1/2 may cause a BRCA -like phenotype triple negative breast cancer (TNBC), which includes the sensitivity to platinums and DNA repair inhibitors. Evaluation of HR proficiency may influence the clinical management of TNBC. Our aim was to evaluate germline and somatic HR gene variants in advanced TNBC patients (pts) and clinical outcome. Methods: Our cohort included advanced TNBC pts unselected for family history or age at diagnosis, enrolled in an institutional molecular screening program (NCT01505400). DNA from matched blood and FFPE tumor samples was assessed using a lab developed next generation sequencing Hereditary Cancer Panel (NGS-HCP) that includes all exons of 52 cancer predisposition genes, with 20 HR genes (Illumina MiSeq/NextSeq, germline coverage 100x, somatic coverage 500x). Medical records were reviewed for clinical outcome, pathology and prior germline BRCA1/2 testing results. All pts consented for research on banked samples and return of pathogenic germline variants was optional. Log rank test was used to determine time from surgery with curative intent to relapse (TTR) and overall survival from diagnosis to death (OS) differences based on presence of HR variants. Results: We included 32 pts who consented for return of pathogenic germline variants and had sufficient DNA for NGS-HCP analysis. Median age at diagnosis was 45 years (range 21-80). Initial stages at diagnosis were: I (12.5%), II (62.5%), III (19%) and IV (6%). Germline HR variants were detected in 17 pts (53%) with a median number of variants per patient of 1 (range 0-6). Five pts had likely pathogenic or pathogenic variants in HR genes: BRCA1 (2), BRCA2 (1) FANCC (1) and FANCC + BML (1). Another patient had a BRCA1 pathogenic variant previously detected by Multiplex Ligation-dependent Probe Amplification but was not detected by NGS-HCP. 26 variants of unknown significance (VUS) were identified in 13 HR genes, including FANCA (6), FANCF (3) and BRCA1 (3). Only one patient had a somatic HR variant in FANCA not found in the germline. 30 pts (94%) had somatic TP53 variants. Sporadic somatic BRCA1/2 variants were not seen. BRCA1/2 variants present in the tumor were equivalent to those detected in blood of BRCA1/2 carriers. Median (m) TTR was 17 months (range 1-119) and mOS was 49 months (range 8-123). Presence of likely pathogenic or pathogenic germline variants was not associated with TTR (p=0.78) and OS (p=0.23). Presence of germline VUS, likely pathogenic or pathogenic variants also did not correlate with TTR (p=0.72) and OS (p=0.47) Conclusions: In our cohort of pts with advanced TNBC, 12% had germline pathogenic variants in BRCA1/2 , similar to the previously reported rate in early stage TNBC pts. Prevalence of likely pathogenic or pathogenic variants in non- BRCA HR genes was 6%. The presence of germline variants in HR genes was not associated with clinical outcome, however, the number of patients included was small and we had limited power to detect survival differences. Background: Variants in homologous recombination (HR) genes other than BRCA1/2 may cause a BRCA -like phenotype triple negative breast cancer (TNBC), which includes the sensitivity to platinums and DNA repair inhibitors. Evaluation of HR proficiency may influence the clinical management of TNBC. Our aim was to evaluate germline and somatic HR gene variants in advanced TNBC patients (pts) and clinical outcome. Methods: Our cohort included advanced TNBC pts unselected for family history or age at diagnosis, enrolled in an institutional molecular screening program (NCT01505400). DNA from matched blood and FFPE tumor samples was assessed using a lab developed next generation sequencing Hereditary Cancer Panel (NGS-HCP) that includes all exons of 52 cancer predisposition genes, with 20 HR genes (Illumina MiSeq/NextSeq, germline coverage 100x, somatic coverage 500x). Medical records were reviewed for clinical outcome, pathology and prior germline BRCA1/2 testing results. All pts consented for research on banked samples and return of pathogenic germline variants was optional. Log rank test was used to determine time from surgery with curative intent to relapse (TTR) and overall survival from diagnosis to death (OS) differences based on presence of HR variants. Results: We included 32 pts who consented for return of pathogenic germline variants and had sufficient DNA for NGS-HCP analysis. Median age at diagnosis was 45 years (range 21-80). Initial stages at diagnosis were: I (12.5%), II (62.5%), III (19%) and IV (6%). Germline HR variants were detected in 17 pts (53%) with a median number of variants per patient of 1 (range 0-6). Five pts had likely pathogenic or pathogenic variants in HR genes: BRCA1 (2), BRCA2 (1) FANCC (1) and FANCC + BML (1). Another patient had a BRCA1 pathogenic variant previously detected by Multiplex Ligation-dependent Probe Amplification but was not detected by NGS-HCP. 26 variants of unknown significance (VUS) were identified in 13 HR genes, including FANCA (6), FANCF (3) and BRCA1 (3). Only one patient had a somatic HR variant in FANCA not found in the germline. 30 pts (94%) had somatic TP53 variants. Sporadic somatic BRCA1/2 variants were not seen. BRCA1/2 variants present in the tumor were equivalent to those detected in blood of BRCA1/2 carriers. Median (m) TTR was 17 months (range 1-119) and mOS was 49 months (range 8-123). Presence of likely pathogenic or pathogenic germline variants was not associated with TTR (p=0.78) and OS (p=0.23). Presence of germline VUS, likely pathogenic or pathogenic variants also did not correlate with TTR (p=0.72) and OS (p=0.47) Conclusions: In our cohort of pts with advanced TNBC, 12% had germline pathogenic variants in BRCA1/2 , similar to the previously reported rate in early stage TNBC pts. Prevalence of likely pathogenic or pathogenic variants in non- BRCA HR genes was 6%. The presence of germline variants in HR genes was not associated with clinical outcome, however, the number of patients included was small and we had limited power to detect survival differences. Citation Format: Stjepanovic N, Kim RH, Wilson M, Mandilaras V, Berman H, Amir E, Cescon D, Elser C, Randall Armel S, McCuaig J, Volenik A, Demsky R, Chow H, Misyura M, Wang L, Oza AM, Kamel-Reid S, Stockley T, Bedard PL. Clinical outcome of patients with advanced triple negative breast cancer with germline and somatic variants in homologous recombination gene [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P3-09-05.

Discrepancies in drug approvals: A global dilemma

We read with great interest the article by Ezeife et al that was published recently in Cancer. The delay in drug approvals at an international level is challenging for both the patient and physician. Regulatory agencies are responsible for deciding drug approval based on the balance between safety and efficacy data. Different regulatory agencies govern this process around the world, with potentially different criteria underpinning the decision, yet all with the goal of improving access to effective therapies. The approval of a drug in one country does not necessarily ensure approval globally, resulting in variability in drug access and care. The article by Ezeife et al identified a median 3-month difference in the time required for drug approval by Health Canada (HC) compared with the US Food and Drug administration (FDA) (9 months vs 12 months; P<.0006). This timeframe has improved compared with earlier studies, with HC approval in the late 1990s taking a median of 518 days compared with 369 days for the FDA. However, these calculations do not incorporate differences in submission timing between regulatory agencies. Demonstrating this, Ezeife et al found a significant difference between time to drug availability based on formulary in Alberta, Canada from the initial FDA and HC approvals of 30.1 months and 16.7 months, respectively. Assuming a 3-month difference in the time required for drug approval, one could assume the most significant delay is the difference in the actual submission date. An analysis of all drug submissions to HC found a mean delay from first submission in a foreign jurisdiction (United States or European Union) to submission in Canada of 540 days. In the United States and the European Union, the majority of new drugs were submitted within 3 months after their first submission to any of the 3 jurisdictions, in contrast to Canada, in which 70% were submitted after 3 months and, more concerning, 40% were submitted after 1 year. These delays are likely to be longer in smaller markets, with even later submission dates noted in Australia. Furthermore, Ezeife et al demonstrated longer times for approval of targeted therapies by HC compared with the FDA (13.0 months vs 6.7 months). In patients with some diseases, this time delay can be a matter of life or death. Crizotinib was granted accelerated approval for use in patients with anaplastic lymphoma kinase (ALK)positive non-small cell lung cancer by the FDA on August 26, 2011 and by HC on April 25, 2012. The median overall survival (OS) in patients with ALK-positive non-small cell lung cancer was significantly improved with crizotinib (1-year OS rate of 70% vs 44%; hazard ratio, 0.36 [P5.004]). The median OS in the crizotinibnaive patients was only 6 months, thereby emphasizing the importance this 8-month difference in approval can have for the individual patient. In this example, there was no significant difference in the submission dates between the FDA and HC (March 30, 2011 for the FDA and June 8, 2011 for HC), but this was not the case for other agencies. Europe and Australia had even longer delays, with approval only granted in October 2012 and September 2013, respectively. Reasons for this delay have been explored, with the affordability of simultaneous submissions and the desire to obtain approval for first-in-class drugs identified as key considerations. The cost of a drug submission can make simultaneous submissions simply unaffordable for some companies, leading to prioritization based on market size. Applications from larger companies to HC were 20% more likely to occur within 6 months after submission to the FDA or the European Medicines Agency (EMA). This is only part of the puzzle because submission delays between Canada and the other jurisdictions cannot be explained simply by the difference observed between large and small company submission times. Furthermore, even if a drug is approved by different agencies, the indications may not be the same. Olaparib, a poly (ADP-ribose) polymerase (PARP) inhibitor, is a perfect example of this. On December 18, 2014, olaparib was approved by the EMA as the first maintenance therapy for patients with platinum-sensitive, recurrent, BRCA-mutated serous ovarian cancer. In contrast, the next day, December 19, the FDA approved olaparib as monotherapy for patients with germline BRCA-mutated advanced ovarian cancer who have been treated with 3 prior lines of chemotherapy, resulting in different patient populations. To the best of our knowledge, this drug has not yet been submitted to HC. The oncology community faces challenges in addressing these inequalities at a global level. Is there potential to explore harmonization or centralize the regulatory submission process between multiple jurisdictions, thereby allowing concurrent assessment of a single application by multiple individual regulatory agencies? This could potentially alleviate some of the delays, and may be an initial step for the multidimensional problem. The core evidence base necessary for decision-making would appear relatively uniform to an informed bystander but,

Non-target progression – The fine line between objectivity and subjectivity

We have read with great interest the article by Litie`reet al. ‘The components of progression as explanatoryvariables for overall survival in the Response EvaluationCriteria in Solid Tumours 1.1 database’ publishedrecently in the European Journal of Cancer [1]. We wereintrigued by the sub-categorisation of progressive dis-ease into (i) progression of targeted lesions; (ii) occur-rence of a new lesion and (iii) non-targeted progressivedisease and the potential implications of these on clinicalpractice and research.Response Evaluation Criteria in Solid Tumours(RECIST) were designed to provide an objective, uni-form and reliable method to reflect changes in tumourburden in clinical trials between international institu-tions [2–4]. Non-target lesions by definition are not eas-ily quantifiable and are said to be a qualitative measure.Progression based on non-target lesions in the setting ofpatients with baseline measurable disease was expectedto be a rare event but as demonstrated by Litie`reet al., is much more common than first thought [1].Intheir series, non-target lesion progression occurred in25%, 27% and 34% of patients with breast, lung andcolorectal cancer respectively [1].By RECIST 1.1, to achieve unequivocal progressionof a non-target lesion there must be: ‘an overall levelof substantial worsening in non-target disease that isof a magnitude that even in the presence of stable dis-ease or partial response of target lesions the treatingphysician would feel it important to change therapy’[3]. Progression in this setting has to be ‘unequivocal’but who defines what is ‘unequivocal’ and how does thisequate to objective assessments in international trials? Ifa quarter of patients may be removed from trials basedon these criteria as suggested by Litie`re et al. this needsbetter clarification particularly as these rates may behigher in other disease sites such as ovarian cancer,given the pattern of disease.In this setting, how do we ensure uniform assessmentacross sites with the current definition that creates a win-dow for bias? As benefits in trials become smaller, thesubjectivity of this evaluation may have a significantimpact on the overall study result. Close interaction withthe radiologist and clinician can help this decision pro-cess. In addition how can unequivocal progression ofnon-targeted lesions be defined in the setting of a centralreview? Central review is designed to be blinded to clin-ical information in order to standardise trial results.