Genomic Correlates of Response to Everolimus in Aggressive Radioiodine-refractory Thyroid Cancer: A Phase II Study Free

Of an estimated 64,300 new cases of thyroid cancer diagnosed in 2016, 90% or more were differentiated thyroid cancer (DTC) that comprises both papillary and follicular subtypes, as well as the oncocytic variant of follicular thyroid cancer (FTC): Hürthle cell thyroid carcinoma (HCTC; ref. 1). A subset of DTC patients will eventually progress to radioiodine refractory (RAIR) status with 10-year survival rates approaching 10% (2). Beyond DTC, medullary thyroid cancer (MTC) is a neuroendocrine tumor arising from the parafollicular C-cells of the thyroid gland occasionally linked to heritable cancer syndromes (3). Poorly differentiated TC (PDTC) or anaplastic thyroid cancer (ATC) are rare subtypes with aggressive clinical behavior and unfavorable outcomes (4).

Multitargeted tyrosine kinase inhibitors (TKI) have proved beneficial in the treatment of RAIR DTC and advanced MTC. Both sorafenib and lenvatinib have demonstrated improvements in progression-free survival (PFS) compared with placebo (10.8 vs. 5.8 months and 18 vs. 3.6 months, respectively) in RAIR DTC (5, 6). Similarly, vandetanib and cabozantinib have shown a PFS benefit in MTC (7, 8). However, available treatments for patients with ATC have limited efficacy.

Current multitargeted TKIs target the VEGF receptor (VEGFR) that increases endothelial cell proliferation and promotes angiogenesis. Frequent somatic gene alterations in thyroid cancer involve BRAF (namely the V600E mutation), RAS, or RET that activate the MAPK, but are also linked to PI3K/mTOR/Akt pathways (9). Downstream of RAS, alterations in PI3K/mTOR/Akt have gained interest in thyroid cancer. Upon activation by a receptor tyrosine kinase (RTK), PI3K leads to the subsequent activation of Akt in a RAS-dependent manner, activating mTOR. mTOR promotes cell proliferation and alters apoptosis. Constitutive activation of the pathway can result from activating mutations in PI3K/mTOR/Akt or with loss of PTEN (10). In a group of unselected, early-stage thyroid cancer patients, these alterations appear to be rare (11). Similarly, alterations in RET can lead to PI3K/mTOR/Akt activation (12). These findings suggest that targeting mTOR may be a useful therapeutic strategy in thyroid cancer.

We sought to evaluate the clinical efficacy of everolimus, a rapamycin analogue which allosterically inhibits mTOR and prevents assembly of the mTOR complex 1 (mTORC1), in patients with aggressive, incurable DTC and explore efficacy in MTC and ATC. We also present genomic sequencing data to identify a molecular basis for treatment response.

Patients were eligible if they were 18 years of age or older with measurable disease, and had a diagnosis of locoregionally advanced or metastatic, incurable thyroid cancer with radiographic evidence of progression by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 within the previous 6 months. DTC patients had to have radioactive iodine refractory (RAIR) disease. Patients could have received up to one prior TKI. No documented disease progression was required for symptomatic MTC or ATC patients. Additional eligibility criteria included: Eastern Cooperative Oncology Group (ECOG) performance status of 2 or less and adequate organ function. Patients with untreated brain metastases were excluded. The study was approved by the institutional review board at each participating site and was conducted in compliance with the Declaration of Helsinki and local regulatory requirements.

In this prospective open-label, nonrandomized, single-arm phase II study, patients were recruited at four centers throughout the United States to receive everolimus 10 mg by mouth daily. The institutional review board at each participating site approved the study and all patients provided written informed consent. Treatment was administered until disease progression, unacceptable adverse event, withdrawal, physician discretion, or death, whichever occurred first. Two dose reductions were permitted for toxicity: everolimus to 5 mg daily and 5 mg every other day. Patients were permitted to reescalate to the prior dose if symptoms were not attributed to treatment. Patients were followed for response with CT or MRI until first disease progression and for survival for 5 years according to the following schedule: at baseline, every 2 months for the first 24 months, and every 3 months beyond 24 months through 5 years. Adverse events were classified and graded according to National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 3.0. The primary endpoint of the study was PFS, with response rate and overall survival (OS) as key secondary endpoints.

Sequencing was performed on DNA extracted from pretreatment, fresh-frozen paraffin-embedded (FFPE) tumor samples. While biopsy just prior to study enrollment was encouraged, archival tumor samples were also permitted. Briefly, total gDNA concentration was determined and samples with sufficient starting material were taken to library construction. Libraries were quantified using qPCR, pooled in equal mass, and captured using the OncoPanel_v2 bait set. Oncopanel_v2 consists of 504 genes and 15 intronic regions with known or potential importance in cancer. Captured libraries underwent paired end 100 (2 × 100 nt) sequencing on a Hiseq 2500 (Illumina Inc.). Read pairs were aligned and demultiplexed using Picard tools. Mutation analysis for single-nucleotide variants was performed using MuTect v1.1.4 in paired mode for tumors with matched germline or single mode for samples without matched germline, and annotated using Oncotator (13). The alignments were further refined using the GATK tool for localized realignment around indel sites. All samples met a minimum requirement that 80% of targets had a minimum coverage of 30×. Structural rearrangements were not reported.

The primary objective was to assess PFS in patients with DTC. Thirty-three patients with DTC provided 97% power to rule out a 12% PFS rate and target a 35% PFS rate (two-stage design used with type-1 error rate of 10%, assuming ≥2 alive and progression-free at 6 months among the first 18 eligible patients accrued in the first stage). Ten patients with MTC and 7 patients with ATC were also to be accrued for a total overall accrual of 50 patients. PFS was defined as time from study entry to disease progression or death, whichever occurred first, and otherwise censored at date last known progression-free. Secondary endpoints included response rate (defined according to modified RECIST criteria), overall survival, and assessment of adverse events. Exact binomial confidence intervals were calculated for response. PFS and OS were estimated using the Kaplan–Meier method. Fisher exact test (categorical variables) and Wilcoxon rank-sum test (continuous variables) were used to assess for differences between patients with and without genomic data, with respect to baseline characteristics.

Between July 2009 and August 2013, 50 patients were enrolled (n = 33 DTC including 13 HCTCs, n = 10 MTC, n = 7 ATC). The median age at registration was 63 years (range 26–82), the majority were male (66%, 33/50), and PTC was the most common histologic subtype (28%, 14/50). Forty-four percent (22/50) of patients received prior TKI (Table 1).

At the time of data cutoff, 46 patients had discontinued protocol treatment [median time on treatment was 10.1 months (range < 1–29.8) with disease progression as the most common reason for stopping]. There were 43 PFS events and 23 patients had died. There were no complete responses (CR), but 3 (6%) partial responses (PR) were observed and 37 patients (74%) experienced stable disease (SD), for a clinical benefit rate (CBR = CR+PR+SD) of 80% (Table 2; Supplementary Table S2). At a median follow-up of 31.8 months, median PFS was 12.5 months (95% CI, 7.2–17.5) with 22.2% of patients alive and progression-free at 2 years. Median OS was 32.7 months (95% CI, 18.7–61.3+) for the entire cohort with 60.7% of patients alive at 2 years.

For the DTC cohort, there was 1 PR (3%) and 27 patients (82%) had SD for a CBR of 84.8%. Median PFS was 12.9 months (95% CI, 7.3–18.5) with a 2-year PFS of 23.6%. Median OS was not reached and 2-year OS was 73.5% (Fig. 1). Three patients continued on treatment at 45.1, 47.1, and 57.6 months and 6 remain in follow-up. Among patients with HCTC, median PFS was 11.7 months and 2-year PFS was 16.8%. Serial thyroglobulin (Tg) levels on treatment were available in 23 DTC patients and levels declined in 74% (17/23) of cases. Median PFS was not significantly different for patients who had previously received treatment with a TKI versus those who had not [10.6 (95% CI, 3.3–18.6) vs. 15.2 (95% CI, 9.2–20.6) months, respectively]. Among 10 MTC patients, there was 1 PR (10%) and 8 (80%) had SD. Median PFS was 13.1 months with a 2-year PFS of 22.5%. Median OS in this group was 21.4 months. Among 7 ATC patients, one achieved a near-complete response, found to have a nonsense mutation of the tumor suppressor gene TSC2 (14). Another had ongoing disease stability for 26 months without evidence of progression prior to death (from congestive heart failure), while four progressed within 3 months of study entry.

Forty-seven patients had at least one treatment-related adverse event (TRAE). TRAEs were predominantly grade 1 or 2 (50%) of which mucositis, acneiform rash, fatigue, and cough were most common. Anemia, thrombocytopenia and transaminitis were the most frequent grade 1 or 2 laboratory abnormalities. Grade 3 TRAEs occurred in 42% (21/50) of patients and the most common were fatigue, stomatitis, and infections. One patient experienced grade 4 hypercholesterolemia (Table 3). Dose modifications were necessary in 32 patients (62%) with the primary reason being toxicity.

Genomic profiling data was available for 38 patients (76%). There were no differences with respect to baseline characteristics between those patients with and without genomic profiling data. Over 1,400 unique, nonsynonymous somatic variants were identified (range: 1–60 per sample). A total of 235 unique indels were identified that affected the coding sequence. A total of 766 (766/1406, 54.5%) of these variants have been reported in the Catalog of Somatic Mutations in Cancer (COSMIC) database. The mutational profile, best clinical response to everolimus, and outcomes of sequenced patients are summarized in Fig. 2. Mutations in NOTCH2, ATM, and RET were frequent, with 42.1% of sequenced patients having somatic alterations in PI3K/mTOR/Akt.

Six of 23 sequenced DTC patients (26.1%) had a BRAF^V600E mutation (one with HCTC), while PI3K/mTOR/Akt alterations were frequent in those with PTC histology (6/10, 60%) most commonly with TSC2 missense mutations. Figure 3 shows PFS estimates by key mutational status for all histologies in the genomic cohort. DTC patients with only a BRAF mutation (n = 3) demonstrated a longer median PFS compared with all sequenced DTC patients (43.4 vs. 9.2 months). Three sequenced DTC patients (27.3%) had coexisting alterations in BRAF and either TP53 or PI3K/mTOR/Akt, and those with coexisting BRAF+TP53 mutations had a shorter median PFS compared with non-TP53, BRAF-mutated DTC patients (6.8 vs. 43.4 months). Few sequenced FTC patients harbored a PIK3CA alteration. The mutational profile of 10 patients with HCTC was bland with regards to BRAF or PI3K alterations compared with other histologies, although two patients had a TSC2 mutation whose disease progressed at 11.7 and 20.5 months. Median PFS for all sequenced HCTC patients was 12.6 months. There were three TP53-mutated HCTC patients: disease progressed at 2.0 and 11.7 months after enrolling in two patients, and one who died without radiographic progression at 13.5 months.

All but one patient in the sequenced MTC group was RET-mutated. Similar to the HCTC subgroup, coexisting PI3K/mTOR/Akt alterations were uncommon and there were no BRAF mutations. Median PFS for sequenced MTC patients was 14.5 months and was similar in those with only a RET mutation versus those with RET and either PI3K/mTOR/Akt or TP53 alterations (21.4 vs. 10 months). All sequenced MTC patients had evidence of SD with the exception of 1 patient whose best response was a PR who remained progression-free at 55.2 months, and that patient had a RET and NF1 truncating mutation.

The median PFS of those PDTC/ATC patients who were sequenced was 2.8 months (median OS, 8.9 months), the lowest of any subgroup. Two sequenced patients (33.3%) with ATC were BRAF-mutated, while 1 had an NRAS mutation, suggesting dedifferentiation from PTC or FTC histology, respectively. Among PDTC/ATC patients with PI3K/mTOR/Akt mutations, median PFS was 15.2 months. One patient had a TSC truncating mutation and coexisting TP53 mutation but had a best response of PR and remained progression-free until 17.9 months. Another patient had an NF1+TP53 mutation and had SD as best response and was progression-free at 26 months until death at 29.9 months.

The widespread availability of TKIs and their ability to stabilize disease have improved outcomes in thyroid cancer, but therapy eventually fails and additional treatments are needed. Moreover, there are currently no targeted therapies approved for ATC. Our study represents the largest cohort to date showing the clinical benefit of a nonkinase inhibitor in advanced thyroid cancer. While radiographic response rates were low, 74% of patients demonstrated SD with a median PFS of just over a year among rapidly progressing DTC patients, and exceeding 1 year in the MTC subgroup. Our findings are consistent with similar studies that have reported SD rates of 65%–76% in those treated with everolimus (15, 16). These findings are particularly notable given that we selected patients with clinically aggressive disease. It is also important to acknowledge that almost half of patients in our cohort received prior TKI therapy yielding a 70.6% SD rate; therefore, mTOR inhibition may prove a viable second-line option post-TKI progression. This was evaluated in a single-arm phase II study of DTC treated with everolimus plus sorafenib, following progression on sorafenib alone, where median PFS was 13.7 months (17), comparable with our cohort treated with mTOR inhibition alone.

In comparing our cohort with 399 thyroid cancer samples generated by The Cancer Genome Atlas (TCGA) Research Network (18), we observed several important differences: (i) our cohort represents a genetically enriched population with a higher mutational burden in that we selected for aggressive tumors and (ii) a significant proportion of our patients had mutations in the PI3K/mTOR/Akt (16/38, 42.1%) compared with TCGA (<5%). Our favorable outcomes may result from targeting the PI3K/mTOR/Akt pathway. Murine models with constitutive mTOR activation have demonstrated decreased in vitro tumor growth with the use of everolimus (19). We expected patients with alterations in the PI3K/mTOR/Akt to derive the greatest benefit from everolimus. That was, however, not the case with the exception of the PDTC/ATC subgroup where all patients with a significant response to treatment demonstrated mutations directly linked to mTOR activation.

Patients with DTC histology who were only BRAF-mutated experienced a PFS benefit greater than that of all other subgroups (43.4 months), fitting with their histology and low mutational burden. While hypothesis-generating, this observation could also have been due to chance alone or other confounders. Of interest, the presence of a coexisting TP53 mutation decreased the median PFS to just 6.8 months among PTC patients, in line with prior studies that indicate that TP53 mutations are an independent marker of poor outcomes in cancer (20). The three PI3K/mTOR/Akt–mutated patients among the FTC and HCTC subgroups had a median PFS of 18.6 months. The sequenced HCTC subgroup demonstrated a median PFS better than expected given that only two patients were TSC2-mutated. Although speculative, this may be explained by alternative mechanisms, such as a reduction in angiogenesis or immunomodulatory effects.

RET oncogene activation was observed in nearly all sequenced MTC patients (87.5%). Their median PFS approached 1 year with 2-year PFS comparable with DTC patients. RET-mutant cell lines demonstrate sustained growth suppression following mTOR inhibition (21), which could account for these findings despite the low number of mTOR alterations we identified. In our PDTC/ATC cohort, best response was SD for two patients and another achieved a near-CR and remained progression-free until approximately 18 months (23). This patient had an exceptional response to mTOR inhibition from a TSC2 nonsense mutation and subsequently developed acquired resistance with a new mTOR^F2108L mutation resulting in allosteric inhibition. The median PFS for the four ATC/PDTC patients who had PI3K/mTOR/Akt alterations (TSC2, FLCN, or NF1) was impressive (15.2 months). The coexistence of a TP53 mutation observed in two of these patients did not decrease survival. Although speculative, PI3K pathway alterations may represent a dominant oncogene in ATC.

Our genomic analysis has some important limitations, mainly related to sample size of each histologic subgroup. Our sequencing platform did not assess for translocations or epigenetic events. In addition, histologic subgroups analyzed were not always mutually exclusive of one another, given the overlap in genomic mutations and the possibility of mixed histologies.

The progression-free benefit of everolimus in this cohort is in keeping with previously studied TKIs and supports mTOR inhibition as a clinically meaningful target in thyroid cancer. Everolimus demonstrated a manageable toxicity profile and may represent a suitable alternative for those patients who cannot tolerate VEGFR inhibitor–specific side effects. This clinically aggressive cohort showed genomic heterogeneity, but enriched for PI3K/mTOR/Akt alterations. It may be that prior VEGFR TKI exposure promotes alternative pathway coactivation to enable tumor progression. PDTC/ATC appeared to benefit most from everolimus when a PI3K/mTOR/Akt alteration was present. While genomic profiling of tumors does not currently guide therapeutic selection in thyroid cancer, these data have important implications when selecting an mTOR inhibitor versus VEGFR TKI. This remains an active area of investigation as we continue to explore additional targeted therapies and the appropriate sequence of treatments in an era of personalized medicine.

N.L. Busaidy reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Eisai, reports receiving commercial research grants from Novartis, and reports receiving commercial research support from Bayer. L.J. Wirth is a consultant/advisory board member for Eisai, Loxo and Novartis. R.I. Haddad is a consultant/advisory board member for Astra-Zeneca, Bristol-Myers Squibb, Eisai, Merck and Pfizer. G. Rabinowits is a consultant/advisory board member for EMD Serono and Pfizer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: G.J. Hanna, N.L. Busaidy, L.J. Wirth, R.I. Haddad, K. Misiukiewicz, T. Thomas, J.H. Lorch

Development of methodology: G.J. Hanna, N.L. Busaidy, L.J. Wirth, K. Misiukiewicz, T. Thomas, J.H. Lorch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.J. Hanna, N.L. Busaidy, N.G. Chau, L.J. Wirth, J.A. Barletta, A. Calles, R.I. Haddad, S. Kraft, G. Rabinowits, E.K. Alexander, F.D. Moore, K. Misiukiewicz, T. Thomas, E. Marqusee, S.L. Lee, J.H. Lorch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.J. Hanna, N.L. Busaidy, N.G. Chau, L.J. Wirth, R.I. Haddad, A. O'Neill, E.K. Alexander, K. Misiukiewicz, T. Thomas, P.A. Janne, J.H. Lorch

Writing, review, and/or revision of the manuscript: G.J. Hanna, N.L. Busaidy, N.G. Chau, L.J. Wirth, J.A. Barletta, A. Calles, R.I. Haddad, M.E. Cabanillas, G. Rabinowits, A. O'Neill, S.A. Limaye, E.K. Alexander, F.D. Moore, T. Thomas, M.A. Nehs, E. Marqusee, S.L. Lee, P.A. Janne, J.H. Lorch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.J. Hanna, A. Calles, R.I. Haddad, A. O'Neill, J.H. Lorch

Study supervision: G.J. Hanna, R.I. Haddad, M.E. Cabanillas, J.H. Lorch

We would like to thank Tyler Haddad, Vicky Vergara, Julian Huang, Sarah Reilly, and Marge Suda for their assistance with data gathering. We thank the Center for Cancer Genome Discovery (CCGD) at the Dana-Farber Cancer Institute for their assistance in analyzing the sequencing data. We are grateful to our patients for their participation in this research. This investigator-initiated study was financially supported by Novartis and the National Institutes of Health/National Cancer Institute Cancer Center Support Grant (grant number P30-CA016672; to M.E. Cabanillas and N.L. Busaidy).

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.