Impact of FDG PET Imaging for Expanding Patient Eligibility and Measuring Treatment Response in a Genome-Driven Basket Trial of the Pan-HER Kinase Inhibitor, Neratinib

This article is featured in Highlights of This Issue, p. 7267

Defining tumor burden is an essential component of oncology clinical trials, both for determining trial eligibility and for measuring treatment response (1). Response Evaluation Criteria in Solid Tumors (RECIST) is a system for quantifying tumor burden that focuses primarily on tumor size (2, 3). RECIST has become the standard for oncology clinical trials because it is easily implemented, applicable in a wide array of solid tumors, and accepted as a surrogate endpoint for drug approval in rare patient populations (1).

Despite the utility of RECIST, it has several important limitations. Target lesions are often selected on the basis of being amenable to repeated unambiguous measurement rather than on a clinical determination of which are most likely to cause morbidity or mortality for the patient. Moreover, tumor size is recorded unidimensionally and may therefore imperfectly reflect therapy response (2–5). Additional, interrater reliability of RECIST assessment is highly dependent on selection of target lesions (6). Furthermore, patients with advanced cancers may never develop disease that is measurable by RECIST. Often, this is due to bone-predominant disease, as is common in prostate and breast cancers, as osseous lesions are not measurable using RECIST criteria. A significant subset of patients with these common cancers have historically been ineligible for enrollment into most clinical trials and consequently been denied access to potentially beneficial therapy. As the presence of osseous only disease may co-associate with molecular subtypes (7), trial populations defined by the presence of RECIST measurable disease may not be reflective of the genetic diversity of patients seen in the clinic. Overall, these limitations suggest that opportunities for alternative measures of tumor burden should be considered in specific situations.

Although RECIST focuses on tumor size, other tumor characteristics may be used to monitor treatment response. ¹⁸F-fluorodeoxyglucose (FDG) Positron-emission tomography (PET) can be used to monitor response to cancer treatments by measuring changes in tumor metabolism (8, 9). Indeed, in patients with lymphoma, FDG-PET has become the standard method for measuring tumor burden in both routine clinical practice and clinical trials (10, 11). For solid tumors, although routine clinical use of FDG-PET has grown, it has not been widely used for response assessment in clinical trials.

SUMMIT (NCT01953926) is an international, multicenter “basket” study evaluating the efficacy of the pan-HER kinase inhibitor neratinib in multiple indications, including patients with HER2-mutant solid tumors (12). As HER2 mutations are rare in solid tumors, with an estimated incidence of 3% in breast cancers (13) and similar low rates in other malignancies, SUMMIT was designed from the outset to enable enrollment and response evaluation on the basis of PET scans. To date, breast cancer represents the tumor type with the largest number of patients enrolled into SUMMIT. Patients with metastatic breast cancer often have bone-predominant disease providing the unique opportunity to evaluate the impact of incorporating PRC into the core eligibility and design of a precision medicine study targeting a rare genomic population (14).

Here, we report the utility of using PET Response Criteria (PRC) to supplement RECIST in the identification of patients eligible for trial accrual, and the impact of PET on assessing treatment response in a clinical trial focused on a rare molecularly defined population.

The design of the neratinib basket study (SUMMIT) has been previously described in detail (12). Given that breast cancer accounted for the largest patient population on SUMMIT, and to help ensure homogeneity of patient characteristics with the goal of permitting meaningful analysis of the data obtained using RECIST and PRC, this analysis was limited to patients with breast cancer. The patients with HER2-mutant breast cancer analyzed here were accrued from 15 medical centers worldwide. Accrual began on September 30, 2013, and the data cutoff for the analysis was June 22, 2018. Eligibility criteria for the breast cancer cohorts included male and female patients ages ≥18 years, Eastern Cooperative Oncology Group performance status of 0–2, adequate hepatic and kidney function, and left ventricular ejection fraction of ≥50%. Patients were eligible to enroll regardless of hormone receptor status.

Depending on date of enrollment and hormone receptor status, patients received neratinib 240 mg daily on a continuous basis, either as monotherapy or in combination with fulvestrant (500 mg intramuscularly once every 2 weeks for 4 weeks, then once every 4 weeks). Patients were treated until disease progression, unacceptable toxicity, or withdrawal of consent.

The study complied with the International Ethical Guidelines for Biomedical Research Involving Human Subjects, Good Clinical Practice guidelines, the Declaration of Helsinki, and local laws and was approved by the institutional review boards of all institutions. Written informed consent was obtained from all patients.

Tumor measurements were performed every 8 weeks using computed tomography (CT), MRI, and/or FDG-PET/CT. The FDG PET/CT procedure was aligned across centers, following published uniform procedural guidelines (15), including quality control of FDG, instrumentation specifications, and patient preparation. Trial eligibility and subsequent measurement of treatment response was primarily performed according to RECIST version 1.1 (3). In patients whose disease was not measurable by RECIST, disease could be defined as trial eligible, and treatment response followed, using PRC, a modified version of PERCIST (8), as previously described (Supplementary Appendix Box SA1; ref. 16). Briefly, in PRC, the18F-fluorodeoxyglucose (FDG) avidity of each lesion is calculated as: Standardized uptake values (SUV) SUV_{corrected for background} = SUV_{max lesion} – SUV_{max liver background}. Values < 0 were treated as 0. This allows the FDG avidity of the lesion to be considered as the excess avidity above background. Complete metabolic response was defined as all lesions reduced to, at, or below SUV_{max liver background}. Partial metabolic response and progressive metabolic disease were defined as 30% decrease or increase in the sum of SUVs, respectively. This modification was utilized to deal with the issue of multi-scanner sites, where the PERCIST requirement of liver background reproducibility of 20% on all follow-up scans may cause scans to be deemed ineligible for analysis.

In Amendment 3 (finalized March 17, 2015) of the protocol, patients were optionally permitted, at their own and their treating physician's discretion, to be followed by both RECIST and PRC; in such cases, RECIST was used as the primary criteria for determining response. In Protocol Amendment 4 (finalized May 20, 2016), PET assessment was mandatory for all patients with breast cancer. Patients with no baseline PET/CT scan were followed using RECIST only. Four patients had no liver background standard uptake values (SUV) reported and their responses were calculated without this information.

For patient demographics, medians and ranges were used to summarize continuous variables and percentages were used to summarize categorical variables.

The primary aim of this analysis was to determine the incremental number of patients eligible for accrual and measurement of treatment response as a result of the use of PET criteria when disease was not measurable using RECIST.

In patients followed by both RECIST and PRC, concordance in overall response was evaluated as a secondary aim. For this analysis, response was defined as a complete response (CR) or partial response (PR), and non-response as stable disease (SD) or progressive disease (PD). Overall response included patients with CR and PR. Patients non-evaluable for response were not included in the analysis. The 95% confidence interval for concordance of response (CR/PR) and non-response (SD/PD) was calculated using the Clopper exact method.

A total of 81 patients with breast cancer and somatic HER2 mutations were enrolled in SUMMIT as of June 22, 2018. Age of patients ranged from 37 to 87 years (mean 60 years). All were metastatic with a median number of prior therapies in the metastatic setting of 3 (range 0 to11 therapies) and a median time from first metastasis of 2.3 years (range 0.1 to 11.5 years). Patient demographics are summarized in Table 1. Four of 81 patients (5%) did not have post-baseline RECIST or PRC assessments; thus 77 patients (95%) were evaluable for response using RECIST, PRC, or both (Fig. 1). Of the 77 efficacy evaluable patients, 63 (82%) were evaluable by RECIST and 43 (56%) were evaluable by PRC; 34 (44%) were followed using RECIST only, 14 (18%) were followed using PRC only, and 29 patients (38%) were followed using both RECIST and PRC.

Responses according to RECIST and PRC are summarized in Table 2. At the data cutoff, 64 patients (83%) had discontinued treatment, 61 (79%) as a result of disease progression; 13 patients (17%) remained on therapy (Table 3).

Thirty-four of the 63 patients (54%) followed by RECIST did not have a PET/CT scan at baseline or follow-up and were followed using RECIST alone. The overall response rate (ORR) in these 34 patients was 24%. The best overall response (BOR) in these patients was CR in 2 patients, PR in 6, and SD in 12 patients; the remaining 14 patients had PD as best response.

Twenty-nine of the 63 patients with breast cancer (46%) with RECIST-measurable disease had a PET/CT scan at baseline and at least one follow-up PET/CT scan. All 29 patients had measurable disease by PRC, allowing secondary evaluation of treatment response by PRC and a comparison of the primary BOR determined by RECIST and by PRC. Responses in patients with disease that was measurable using RECIST and PRC are shown in Fig. 2. In total, 25 of 29 (86%; 95% CI, 68%–96%) of patients were concordant for response versus non-response as measured by RECIST and PRC at the individual patient level. This included 2 patients with CR, 6 with PR, 7 with SD, and 3 with PD using both RECIST and PRC criteria, as well as 7 patients with a PR by RECIST and a CR by PRC. The remaining four patients (14%) were discordant with regards to response versus non-response. In 3 patients, BOR was SD by RECIST but PR by PRC. Conversely, one patient had a BOR that was better by RECIST than by PRC (PR vs. SD; Fig. 3, patient 29). Of note, patients 2, 4, and 10 demonstrated progression in non-target lesions, and patient 23 demonstrated a RECIST PR due to a non-target lesion.

An additional 14 patients were eligible for the SUMMIT trial due to the use of PRC to measure response, a 22% (14/63) increase over the 63 patients evaluable by RECIST. Of the 14 patients whose treatment response could be assessed by PRC alone, 11 had FDG-avid bone-only disease that was not apparent on CT, two had FDG-avid bone and liver disease that was not apparent on CT, and one had FDG-avid bone and nodal disease that was not apparent on CT (Fig. 3). PRC allowed effective measurement of tumor response in the absence of RECIST-measurable disease in all 14 patients. The ORR in these 14 patients was 64%, with 5 achieving a CR, 4 PR, 2 SD, and 3 with PD as BOR.

HER2 mutations have been identified in the tumors of 3% to 6% of patients with breast cancer (13, 17, 18), making it challenging to accrue to clinical trials targeting this rare molecularly defined patient population. Here, we leveraged an ongoing multicenter basket trial of the HER kinase inhibitor neratinib (SUMMIT) to determine the clinical utility of using PRC to expand the population of eligible patients. RECIST was the primary method used to quantitate measurable disease and define trial eligibility; however, utilization of PRC as a secondary determinant of measurable disease for study eligibility led to a 22% increase in accrual of patients with HER2-mutant breast cancer. Any response criteria under consideration as a supplement to RECIST must provide a clear advantage over this established and widely used response criteria (2). Our data suggest that PRC can help accelerate accrual of a rare genetically defined patient population such as those with HER2-mutant breast cancer. In addition, use of PRC can allow patients with non-RECIST measurable disease access to therapy on clinical trials that otherwise would have been denied. These data provide strong rationale for including PRC in future clinical trials involving rare tumors or rare genomically defined subpopulations of more common cancers.

The potential utility of using PRC in clinical trials is not limited to accelerating patient accrual or broadening patient eligibility. Patients with RECIST non-measurable disease, such as those with metastatic breast cancer and bone-only metastases, are frequently seen in the clinic yet have been historically excluded from enrollment and/or efficacy assessment in clinical trials. In a pooled analysis of data from 13 prospective clinical trials of patients with metastatic breast cancer (n = 10,521), 12.5% of patients in these trials had bone-only metastases (19), and thus were non-measurable by RECIST. A quantitative criterion that uses PET to assess metabolic response may allow recruitment into clinical trials of patients more representative of those seen in the clinic. This could be particularly applicable to estrogen receptor positive cancers, which demonstrate a tendency to metastasize to bone. In addition, tumor response in bone has been shown to be better monitored by FDG PET/CT than CT and MR (20, 21). The use of PRC in this respect in similar to prior attempts to quantify disease in bone-predominant breast cancer with circulating tumor cells (CTC; refs. 22–25). Of note, CTCs are known to correlate with FDG PET (23, 25).

Successful accrual of patients into a trial is not sufficient to warrant usage of a method to measure disease status in trials of novel therapeutics; effectiveness in measuring tumor response is also critical. In this limited series of patients, PRC identified responders and non-responders to treatment with neratinib and was able to categorize patients for CR, PR, SD, and PD, analogous to RECIST. Considering all responses (CR + PR), concordance in the 29 patients evaluated by both criteria was 86%. Although additional studies will be needed to further support the utilization of FDG-PET to measure tumor response, an area of continued research and refinement, these preliminary results are promising and suggest efficacy data generated using PRC in patients with RECIST non-measurable disease can be useful in guiding drug development decisions (8, 11).

We did note that BOR was more favorable when measured by PRC as compared with RECIST in a subset of patients. As demonstrated in Fig. 2, 3 patients with a BOR of SD by RECIST had a PR by PRC. In addition, 7 patients with a BOR of PR by RECIST had a CR by PRC. Several of the discordant results were a result of a greater percentage decrease in FDG-PET measurements of target lesions compared with the percentage of change in the size of lesions. Furthermore, residual lesions on CT that would preclude a RECIST CR may decrease to background on FDG-PET, resulting in an apparently better response according to PRC than by RECIST. Although this difference is noteworthy and should be recognized when comparing RECIST and PRC results, delineating between partial versus complete response has generally not been considered clinically meaningful in studies evaluating agents used in the advanced/metastatic solid tumors where cure is not currently achievable.

PRC was used as a modification of the well-known PERCIST criteria (8). Several other response criteria using FDG PET have been utilized, including EORTC, iPERCIST, and others (26, 27). We used this modification due to the issue of multi-scanner sites where patients may not be scanned on the same scanner at all time points, which could cause to PERCIST requirement of liver backgrounds within 20% on all scans to cause patients and scans to become ineligible for analysis.

Some limitations of this analysis warrant consideration. It was not possible to formally assess differences in time to progression (TTP) in patients followed using both RECIST and PRC because not all patients with PD according to RECIST at a specific time point underwent PET/CT at that time. Consequently, RECIST- and PRC-determined TTP could not be calculated for all patients. Performing RECIST and PRC evaluations at all time points would have allowed a comparison of TTP according to RECIST and PRC. Inclusion of only patients with breast cancer, a disease where bone predominant disease is relatively common, was another limitation of this analysis; the incremental utility of PRC may be attenuated in studies enriched for patients with other tumor types. All methods of measuring tumor response have limitations. For PRC and other methods based on FDG PET, low grade malignancies may not express sufficient FDG-avidity to be appreciated on FDG PET, although for this trial HER2-mutant breast cancers are normally FDG-avid. Biologic and technical factors of FDG PET are known to affect SUV values (8, 15).

Although RECIST has become the standard for defining response in clinical trials of solid tumors, it has important and well-recognized limitations. Many patients with advanced cancers, including those with breast, prostate and ovarian cancer, often spend the entire course of their illness and eventually die of their disease, without ever developing RECIST measurable lesions. As such, clinical trial populations defined by a requirement for RECIST measurable disease may not be entirely reflective of the broader population of patients. Moreover, genome-driven studies are increasing targeting rare genomic subpopulations in oncology. The ability to accrue and learn from all patients harboring these rare and highly relevant genomic biomarkers, without artificially limiting enrollment to patients with RECIST measurable disease, will be increasingly critical to the success of these studies and the field in general. Using PRC to supplement RECIST in the SUMMIT trial provided an advantage in assessing patient eligibility and measuring treatment response in patients in rare HER2-mutant breast cancers. Use of PRC can allow patients with non-RECIST measurable disease access to therapy on clinical trials that otherwise would have been denied. Responses were concordant between PRC and RECIST in 86% of patients. The ability to offer trial assess for patients with RECIST non-measurable disease, as well as the ability to facilitate more rapid accrual of patients with a rare biomarker, suggest PRC could be used as a supplement to RECIST in clinical trials of rare tumors, a proposal which should be evaluated in further studies.

PET response criteria can facilitate more rapid accrual of patients to clinical trials of rare biomarkers and allows patients with non-RECIST measurable disease access to therapy on these trials. This provides a rationale for including FDG PET measurements in future clinical trials involving rare tumors or rare genomically defined subpopulations of more common cancers.

G.A. Ulaner is an employee/paid consultant for Sanofi, and reports receiving commercial research grants from Puma. C. Saura is an advisory board member/unpaid consultant for Puma Biotechnology. S.A. Piha-Paul reports receiving other commercial research support from Abbvie, Aminex Therapeutics, BioMarin Pharmaceutical, Boehringer-Ingelheim, Cerulean Pharma, Chugai, Curis, Five Prime Therapeutics, Genmab A/S, GlaxoSmithKline, Helix BioPharma, Incyte, Jacobi Pharmaceuticals, Medimmune, Medication, Merck Sharp and Dohme, NewLink Genetics Corp/Blue Link Pharmaceuticals, NewLink Genetics/Blue Link, Novartis, Pieris, Pfizer, Principia, Puma, Rapt, Seattle Genetics, Taiho, Tesaro, TransThera Bio, and XuanZhu Biopharma, I. Mayer is an employee/paid consultant for Eli-Lilly, Novartis, GlaxoSmithKline, Immunomedics, Macrogenics, Seattle Genetics, AstraZeneca, and reports receiving commercial research grants from Genentech, Pfizer, and Novartis. D. Quinn is an employee/paid consultant for AstraZeneca, Genentech/Roche, Pfizer, and Novartis. K. Jhaveri is an advisory board member/unpaid consultant for Novartis, Pfizer, AstraZeneca, Genentech, Taiho Oncology, Juno Therapeutics, SpecSpectrum Pharmaceuticals, ADC Therapeutics, and Synthon. M. Dujka and R. Bryce are employee/paid consultants for and hold ownership interest (including patents) in Puma Biotechnology. F. Meric-Bernstam is an employee/paid consultant for Genetech, Pieris Pharmaceuticals, Samsung Bioepis, Aduro, OrigMed, Debiopharm Group, Zencor, Jackson Laboratory, Zymeworks, Inflection Bioscience, Darwin Health, Spectrum, Mersana, and Seattle Genetics, reports receiving commercial research grants from Novartis; AstraZeneca; Taiho Pharmaceutical; Genentech; Calithera Biosciences; Debiopharm Group; Bayer; Aileron, reports receiving speakers bureau honoraria from Chugai, and is an advisory board member/unpaid consultant for Taiho, Genentech, Debiopharm Group, Pfizer. D. Solit is an employee/paid consultant for Pfizer, Loxo Oncology, Illumina, Vivideon Therapeutics, and Lilly Oncology. D.M. Hyman is an employee/paid consultant for Chugai, Boehringer Ingelheim, AstraZeneca, Pfizer, Bayer, Genentech/Roche, Found, reports receiving commercial research grants from AstraZeneca, Puma, LOXO Oncology, and Bayer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G.A. Ulaner, C. Saura, D. Quinn, R. Bryce, F. Meric-Bernstam, D.B. Solit, D.M. Hyman

Development of methodology: G.A. Ulaner, R. Bryce, D.B. Solit

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.A. Ulaner, C. Saura, S.A. Piha-Paul, I. Mayer, D. Quinn, B. Stone, G. Mann, D.B. Solit, D.M. Hyman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.A. Ulaner, S.A. Piha-Paul, D. Quinn, K. Jhaveri, S. Shahin, G. Mann, R. Bryce, D.B. Solit, D.M. Hyman

Writing, review, and/or revision of the manuscript: G.A. Ulaner, C. Saura, S.A. Piha-Paul, I. Mayer, D. Quinn, K. Jhaveri, G. Mann, M. Dujka, R. Bryce, F. Meric-Bernstam, D.B. Solit, D.M. Hyman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.A. Ulaner, B. Stone, M. Dujka, D.B. Solit, D.M. Hyman

Study supervision: G.A. Ulaner, D.M. Hyman

This research was funded in part through the NIH/NCI Cancer Center Support Grant P30 CA008748 and Cycle for Survival. This work was supported by Puma Biotechnology. The authors thank the patients and families for participating in this study. Editorial support was provided by Lee Miller.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.