Phase Ib/II Dose Expansion Study of Lenvatinib Combined with Letrozole in Postmenopausal Women with Hormone Receptor–Positive Breast Cancer

Hormone receptor–positive (HR⁺) breast cancer is the most common breast cancer subtype. In metastatic breast cancer (MBC), inhibitors of the CDK4/6 and PI3K/AKT/mTOR pathway (1), in combination with endocrine therapy (ET), have delayed use of chemotherapy in the absence of visceral crisis. CDK4/6 inhibitors combined with ET have emerged as standard-of-care first-line therapy, showing significant progression-free survival (PFS) and overall survival (OS) gains with consistent hazard ratios (HR) across drugs within the same class (2–6). Post CDK4/6 inhibitor exposure, few randomized studies have offered head-to-head comparison to identify the optimal second-line treatment strategy (7–10). In real-world practice, treatment choice is largely based upon physician preference and shared decision-making with the patient.

RET is an estrogen response gene with preclinical studies demonstrating cross-talk between the RET and estrogen receptor (ER) pathway (11). A recent study testing RET expression in 990 biopsy tissue from patients with primary breast cancer showed RET overexpression (defined by moderate to strong intensity staining on IHC) in 45.4% of hormone receptor–positive, HER2-negative tumors (12). Increased RET expression has been demonstrated in hormone-resistant breast cancer cell lines, with greater cell kill observed in combination of ET and RET inhibitor compared with single-agent ET (13). Lenvatinib is a multikinase inhibitor with potent activity against RET, vascular endothelial growth factor receptor (VEGFR), KIT, and fibroblastic growth factor receptor pathways, with FDA approval for use as monotherapy and combination therapy in thyroid, liver, endometrial, and renal cell carcinomas (14–18). Preclinical studies from our laboratory showed at least additive effects of lenvatinib with ET in HR⁺ breast cancer cell lines (manuscript in preparation). An investigator-initiated, single-center phase Ib/II study investigating safety and tolerability of the combination of lenvatinib plus letrozole, followed by dose expansion at recommended phase II dose (RP2D) was conducted. Primary objectives included safety and RP2D determination in phase Ib, and objective response rates (ORR) in phase II dose expansion.

Patients ages 18 years or older, with histologically confirmed HR⁺/HER2-negative breast cancer, Eastern Cooperative Oncology Group (ECOG) performance status of 0–1, measurable disease according to Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1 (19), and adequate organ function were eligible. No restrictions to prior line(s) of chemotherapy or ET were placed, and patients with prior exposure to letrozole were eligible if they progressed on letrozole more than 1 year after adjuvant treatment, or more than 6 months in the metastatic setting. Patients who were premenopausal were eligible if treated with medical ovarian suppression with postmenopausal levels of estradiol; patients with symptomatic brain metastases were excluded. Detailed inclusion and exclusion criteria are provided in Supplementary Data.

The study was conducted in accordance with the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice Guideline and approved by the relevant regulatory and independent institutional review board. Written informed consent was obtained from enrolled patients.

This was an investigator-initiated phase Ib/II dose expansion, open-label study (ClinicalTrials.gov: NCT02562118) investigating the safety and tolerability of lenvatinib plus letrozole in patients with HR⁺/HER2⁻ breast cancer at the National University Cancer Institute, Singapore (NCIS). The trial was initially intended to recruit early-stage breast cancer patients receiving neoadjuvant ET, but due to poor recruitment, was subsequently amended to enroll MBC patients after the first four patients were enrolled.

A 3 + 3 dose deescalation design was applied in phase Ib, with a standard dose of letrozole at 2.5 mg daily combined with varying doses of lenvatinib. Lenvatinib dosing commenced at 20 mg daily, with deescalation to lower doses if required. A 2-week safety run-in of lenvatinib monotherapy was administered prior to addition of letrozole (Fig. 1A). After RP2D was determined, dose expansion occurred to explore potential efficacy signals. Treatment was continued until disease progression, withdrawal of consent, or unacceptable toxicity.

Serial tumor biopsies were carried out at up to 4 time points for translational studies, including baseline, 2 weeks after lenvatinib safety run-in, 4 weeks post combination therapy, and an optional biopsy at disease progression.

Patients had vital sign measurements, physical examinations, ECOG performance status determination, full blood count, chemistry panel, electrocardiography, echocardiogram and computed tomography scan performed at baseline. Serial thyroid function and urine dipstick monitoring were performed for lenvatinib-specific toxicities including hypothyroidism and proteinuria. Toxicities were assessed using CTCAE version 4.0 (20). Frequency of assessments are described in Data Supplement.

A 3 + 3 dose deescalation design was applied in phase I, followed by dose expansion at RP2D with target enrolment of at least 30 patients, based on Ahern's design assuming 80% power, one-sided α of 0.1, alternative hypothesis of an ORR of 50%, and null hypothesis of ORR <30%. Primary objectives of the study included the safety and determination of RP2D in phase Ib, and determination of ORR at RP2D in phase II dose expansion. All statistical analyses were generated using STATA Version 16.0 and performed on an intention-to-treat (ITT) basis, except for ORR and clinical benefit rate (CBR) where modified ITT was additionally performed to include only those who were evaluable. Details are included in Supplementary Data.

Pharmacokinetics (PK) modeling was conducted using a noncompartmental extravascular model for plasma with Phoenix WinNonLin Software version 8.0 (Centera). Pharmacodynamics (PD) biomarker analysis was carried out with IHC assessment of RET, pERK, Ki67, CD31, and SMA using formalin-fixed, paraffin-embedded tissue sections.

Hematoxylin and eosin–stained sections were studied for tumor adequacy, and biopsies with <5% tumor in them were excluded from analysis. Automated staining was done using Leica biosystems BOND-MAX system for the primary antibodies RET (ABCAM, EPR2871) and ERK1/2 MAPK 44/45 pERK (Cell Signaling; #9101), Ki67 (DAKO, #M7240), CD31 (Dako, #M0823), and SMA (Dako, #M0851). Semiquantitative scoring of marker expression was done as H-score (intensity times % of +cells) for RET and pERK, proliferation index % (Ki67), microvascular density (MVD), and vascular normalization index (CD31 and SMA).

MVD was calculated as the number of CD31⁺ vessels in the highest vascular hotspot microscopic field at 200× magnification. In case of >1 hotspots, MVD of the hotspots was averaged. In very tiny biopsy samples, the entire biopsy was used for MVD calculation. Vascular normalization index (VNI) was calculated as the percentage of dual CD31⁺ vessels coexpressing SMA on the pericyte layer.

The H-score for RET expression levels was categorized as negative/0(0–49), 1+(50–99), 2+(100–149), and 3+(150–300). H-score ≥50 (cat 1–3+) was considered RET-positive tumors, and H-score <50 was considered RET-negative tumors.

The mixed-effect model assuming a random intercept was implemented to examine the effect of response and time on the IHC assessments. This takes into account possible intrasubject correlation in repeated assessments at baseline, 2 weeks, and 4 weeks. Statistical evaluations were made assuming a two-sided test at the 5% level of significance.

The data generated in this study are available within the article and its Supplementary Data files.

Forty-seven patients were enrolled and included in the safety analysis (16 subjects in dose escalation, 31 in dose expansion). The initial four patients enrolled received study treatment as neoadjuvant therapy for locally advanced disease, after which only patients with MBC were enrolled. All patients were evaluable for toxicity. All 31 patients in phase II were included for survival analysis, and median duration of follow-up for this cohort was 12.3 months. One patient was unevaluable for response due to death by suicide prior to first response assessment (Supplementary Fig. S1).

Median age at enrolment was 61 (range, 31–72) years, and median lines of prior therapy in the metastatic setting was 5 (range, 0–11). Among patients with MBC (n = 43), 70% had prior chemotherapy, 77% had prior ET plus CDK4/6 inhibitor therapy, and 12% had prior ET as single-agent therapy without combination with targeted therapy (e.g., CDK4/6 inhibitor or mTOR inhibitor) in the metastatic setting (Table 1).

Lenvatinib treatment commenced at dose level (DL) 1 of 20 mg daily (n = 4). Of 4 patients, 2 experienced dose-limiting toxicities (DLT) including grade 3 (G3) palmar–plantar erythrodysthesia (PPE; n = 1) and G3 proteinuria (n = 1), resulting in dose deescalation to DL-1 of 16 mg lenvatinib daily. At DL-1 (n = 6), no DLTs were observed, but all 6 patients required dose reductions, a majority (4/6) occurring within 6 weeks of initiation of combination therapy (G3 hypertension (3/6) and G3 wound pain (1/6)), deeming DL-1 intolerable. Six patients were enrolled at DL-2 of lenvatinib 14 mg daily and none required dose reduction, thus DL-2 was declared as RP2D. Phase II dose expansion occurred at RP2D, with 31 patients enrolled (Fig. 1B).

In dose-finding phase, the most common all-grade treatment-emergent adverse events (TEAE) included hypothyroidism (75%), hypertension (62.5%), proteinuria (62.5%), and PPE (62.5%; Table 2), which were deemed treatment related. The most common grade 3/4 toxicities were hypertension (43.8%), proteinuria (18.8%), and PPE (18.8%), which were managed with medication and resolved without sequelae.

In dose expansion, the most common all-grade TEAEs were hypertension [64.5% (grade 3/4, 45.2%)], hypothyroidism [38.7% (grade 3/4, 0%)], and fatigue [38.7% (grade 3/4, 6.5%); Table 2]. Toxicities were deemed treatment related and consistent with lenvatinib toxicity profile, and were managed by standard guidelines. One patient experienced grade 4 hyponatremia that was deemed unrelated to study procedures, and no grade 5 TEAEs were observed. Median time for normalization of thyroid function after commencement of thyroxine for hypothyroidism was 12.6 weeks (range, 1.6–26.4 weeks). There was variable practice between investigators in the management of thyroid function post discontinuation of lenvatinib, with some continuing thyroxine replacement without attempting to taper, whereas others actively reduced the dose to facilitate weaning. At the time of data analysis, 2 patients were weaned off thyroxine at 3.6 and 32.9 weeks, respectively, after discontinuation of lenvatinib. Hypothyroidism is a known class effect of tyrosine kinase inhibitors that affect the VEGFR pathway, and studies looking at long term-effects have shown varying levels of recovery, ranging from complete resolution to permanent hypothyroidism if there is destruction of thyroid cells with thyroid gland atrophy (21, 22).

Antitumor activity of letrozole plus lenvatinib is detailed in Table 3 and Fig. 2. All patients (n = 16) in the dose-finding and 30 of 31 patients in the dose-expansion cohorts were evaluable for treatment response. The key primary endpoint for efficacy was measured based on ORR in the dose-expansion cohort.

ORR in the dose-finding cohort was 56.3% (95% CI, 29.9%–80.2%). Among patients with MBC (n = 12), 5 of 6 (83%) achieved RECIST partial response (PR) at DL-1 (16 mg lenvatinib; Table 3), despite all 6 patients requiring dose reductions of lenvatinib to 10 or 14 mg daily. Patient LL-006 is a patient with de novo MBC who commenced on study after progressing on first-line paclitaxel. She commenced on lenvatinib with letrozole as first-line ET and has ongoing treatment response of 32.7 months despite 2 dose reductions to 10 mg of lenvatinib. Another patient LL-010 with de novo MBC who progressed on 3 prior lines of ET, including CDK4/6 inhibitor, achieved ongoing PR of 25.6 months on treatment despite dose reduction to 14 mg lenvatinib since week 11 of combination therapy.

At DL-2 (14 mg lenvatinib), 2 of 6 patients achieved PR and 4 patients achieved SD as best response, with 2 patients having SD of more than 6 months (SD ≥ 6 months). A patient with 5 median prior lines of therapy, including ET, CDK4/6 inhibitors, and chemotherapy, achieved sustained SD for 15.5 months.

In the dose-expansion cohort, response was observed in 7 of 30 evaluable patients, with ORR of 23.3% (95% CI, 9.9%–42.3%; Table 3). Median duration of response (DoR) was 6.9 months (IQR 5.9–13.1). Additionally, 8 patients had SD ≥ 6 months, giving a CBR of 50.0% (95% CI, 31.3%–68.7%). Twenty-two of 31 patients had progressed, with median time-to-progression of 6.2 months (IQR, 2.1–10.8; Fig. 2). At the time of data cutoff, 14 patients have passed on, and 1-year OS was 59.7% (95% CI, 38.1%–75.9%). Similar estimates were observed in the ITT population, with ORR and CBR of 22.6% (95% CI, 9.6%–41.1%) and 48.5% (95% CI, 30.2%–66.9%), respectively.

In the dose-expansion cohort, 25 patients had prior ET plus CDK4/6 inhibitor treatment. Among these patients, 5 achieved PR (ORR 20.0%; 95% CI, 6.8%–40.7%), and median DoR was 6.9 months (IQR 5.9–13.1 months). Ten patients achieved SD as best response, of whom 8 achieved SD ≥ 6 months, resulting in a CBR of 52.0%. In this subgroup of 25 patients deemed to have CDK4/6 inhibitor resistance, median PFS was 6.2 months (95% CI, 2.1–10.8 months), and 1-year OS rate was 62.4% (95% CI, 37.1%–79.9%; Fig. 2). Among these 25 patients, 8 received CDK4/6 inhibitor plus ET as immediate prior therapy before trial enrolment. Interestingly, among these 8 patients, 5 demonstrated clinical benefit with letrozole plus lenvatinib [PR (n = 2), SD ≥6 months (n = 3)], yielding CBR of 62.5%. Presently, 4 patients remain on trial with continued benefit. Only 3 patients in the dose-expansion cohort received prior single-agent ET therapy without targeted therapy. These 3 patients had a median of 5 prior lines of therapy in the metastatic setting (range, 2–6), with at least 2 prior lines of ET before study enrolment. Among the three patients, no responses were observed, and median PFS was 3.91 months, with a lower 95% confidence limit of 1.61 months. However, the upper 95% confidence limited cannot be estimated.

Lenvatinib PK studies using blood samples collected during monotherapy run-in and combination therapy showed no significant differences in C_max or peak-to-trough ratios, indicating that the overall PK profile of lenvatinib was similar in single-agent or combination treatment, with no significant interaction between lenvatinib and letrozole (Supplementary Fig. S2).

Of the 47 patients, 38 baseline biopsies were available for IHC biomarker evaluation, of whom 30 patients had paired samples at baseline and post 2 weeks of single-agent lenvatinib (Fig. 3A). Twenty-seven patients had a tumor biopsy after 4 weeks of combination therapy. After 2 weeks of lenvatinib monotherapy, significant downregulation of RET (mean H-score change −15.6; 95% CI, −29.2 to −2.0, P = 0.026) and numerical downregulation of pERK1/2 (mean H-score change −14.6; 95% CI, −31.0 to 1.8, P = 0.080) was observed, suggesting on-target effect of the RET pathway. Surrogate markers for VEGFR pathway including MVD and VNI were studied, and we observed significant decrease in mean MVD (mean H-score change −8.4; 95% CI, −15.4 to −1.5, P = 0.019) and increase in VNI (mean H-score change 9.0; 95% CI, 3.9–14.1, P = 0.001) after 2 weeks of single-agent lenvatinib (Supplementary Table S1). Taken together, these suggest target engagement of lenvatinib and effect on downstream pathways.

We first studied the association between baseline tumor RET expression and subsequent clinical benefit from letrozole plus lenvatinib. Of the 38 patients recruited in phase Ib/II with available baseline biopsies, 37 of 38 were evaluable for clinical response. Twenty-four of 37 (64.9%) and 13/37 (35.1%) patients, respectively, had RET-positive (IHC H-score categories, 1–3+) and RET-negative (IHC H-score category, 0) tumors at baseline. Twenty-two of 37 patients (59.5%) were good responders, defined as patients who achieved PR or SD ≥ 6 months. Although more patients with RET-positive tumors achieved a good response compared with those with RET-negative tumors at baseline (67% vs. 46%, P = 0.19), this difference was not statistically significant.

We next studied serial changes in RET pathway signaling using biopsies across 3 time points. A total of 37, 29, and 26 tumor samples were collected at baseline, 2 weeks after lenvatinib monotherapy, and 4 weeks after combination therapy, respectively, and included in this analysis (Fig. 3B).

No significant differences in baseline mean H-scores of tumor RET and pERK were observed in good compared with poor responders (RET 82.6 vs. 52.7, P = 0.129; pERK 59.3 vs. 37.9, P = 0.289). Changes in RET and pERK IHC over time with treatment were similar in both groups, likely indicating on-target effect of treatment regardless of RET or pERK expression. Good responders had decreased tumor Ki67 in response to treatment, compared with poor responders whose tumors demonstrated progressive increase in Ki67 over time. Interestingly, a significant increase in VNI was seen in good responders (P < 0.001) but not poor responders (P = 0.078). In concordance, a downward trend in MVD at the threshold of statistical significance (P = 0.082) was noted among good responders but not poor responders. Both VNI and MVD are markers of vascular normalization and taken together suggest more prominent target engagement and PD effect of lenvatinib in clinical responders than in nonresponders (Fig. 3B).

In this study, we demonstrated that the novel combination of lenvatinib and letrozole at biologically effective doses resulted in durable clinical benefit in patients with MBC with a manageable toxicity profile, including those who previously demonstrated resistance to CDK4/6 inhibitors.

Overall, treatment-related toxicities were reversible and were known toxicities of lenvatinib such as hypertension and hypothyroidism, with minimal evidence of additive toxicity when combined with letrozole. Toxicities were effectively managed with early, proactive interventions using medical therapy where clinically indicated, such as antihypertensives and thyroxine replacement. Dose finding commenced at the dose level of lenvatinib 20 mg plus letrozole 2.5 mg daily, but dose deescalation was required due to DLT of PPE and proteinuria. Although there appears to be a dose–response relationship with a higher ORR observed among patients treated with levatinib 16 mg daily compared with 14 mg daily, lenvatinib 16 mg daily was deemed intolerable due to increased toxicity requiring frequent and early dose reductions. After careful review of the risk–benefit ratio, 14 mg lenvatinib was declared as the RP2D. Lenvatinib is currently approved by the FDA at varying doses depending on tumor subtype (14–18), ranging from 12 mg daily for hepatocellular carcinoma to 24 mg daily for thyroid cancer. The declared RP2D of lenvatinib in our study is within this range of approved dosing, and no evidence of significant interaction on PK studies was observed when letrozole was coadministered with lenvatinib. This could be explained by the fact that lenvatinib and letrozole are metabolized mainly by different cytochrome enzymes, CYP3A and CYP2A, respectively (23, 24). In terms of PD effects, tumor IHC confirmed on-target effects with downregulation of RET and pERK1/2 expression as well as a decrease in MVD and increase in VNI, indicating RET and VEGFR pathway modulation, respectively.

Tyrosine kinase inhibitors with activity against RET, including cabozantinib and vandetinib, have previously been studied in patients with MBCs in phase II trials. Single-agent cabozantinib resulted in an ORR of 13.6% (95% CI, 6%–25.7%) and a 12-week disease control rate of 46.7% (95% CI, 31.7%–61.6%) in a study involving 45 patients with MBC, of whom 96% had ER-positive breast cancer with a median of 3 prior lines of chemotherapy in the metastatic setting (25), whereas another study treated 46 patients with metastatic breast cancer with prior exposure to both anthracyclines and taxane therapy with single-agent vandetinib did not observe any objective responses (26). Combination of vandetinib with fulvestrant was further studied in patients with bone-only metastatic disease who have progressed on 1 to 2 prior lines of ET, with a CBR of 41% and median PFS of 5.8 months (27). The combination of lenvatinib and letrozole has shown evidence of clinically meaningful antitumor activity in the metastatic setting. In the phase II dose-expansion cohort, despite a median of 5 prior lines of therapy, ORR was 23.3% with a CBR of 50%. Median time-to-progression was 6.2 months, with a median DoR of 6.9 months. An encouraging 1-year OS of 59.7% was observed in this heavily pretreated population.

Although there is no standard-of-care ET in the second-line-and-beyond setting, approved ET combination therapies include everolimus with exemestane (8), and alpelisib with fulvestrant (7). In the BOLERO2 and SOLAR1 studies, combination of ET with everolimus or alpelisib yielded ORR of 9.5% and 26.6%, and median PFS of 10.6 months and 11.0 months, respectively, albeit in populations that were largely CDK4/6 inhibitor naïve. Although reported outcomes of alpelisib plus fulvestrant appear to be more favorable than everolimus plus exemestane, a majority of patients enrolled in SOLAR-1 were less heavily pretreated and received fulvestrant as ET backbone, which has demonstrated superiority to aromatase inhibitors as monotherapy. The SOLAR1 trial was also benefited by biomarker selection of patients most likely to respond to alpelisib. Despite a more heavily pretreated and biomarker-unselected patient population, the combination of letrozole plus lenvatinib investigated in our study has shown promising clinical activity comparable to that reported by BOLERO-2 and SOLAR-1. Interestingly, we noticed that 23 patients (74%) treated in the dose-expansion cohort had previously progressed on letrozole monotherapy in the metastatic setting, yet 11 of 23 (47.8%) patients demonstrated clinical benefit from letrozole plus lenvatinib. Further studies are under way to understand the mechanism of action of lenvatinib in ER⁺ breast cancer that have progressed on ET.

We found the activity of this combination to be highly encouraging in the subgroup of patients who previously progressed on CDK4/6 inhibitors, with 20% ORR and 52% CBR at RP2D. In a subset of 8 patients in the dose expansion cohort who progressed on CDK4/6 inhibitors immediately prior to enrolment, ORR and CBR were higher at 25.0% and 62.5%, respectively, with a median PFS of 6.2 months. In a recently presented study on the use of alpelisib plus fulvestrant after progression on ET plus CDK4/6 inhibitor, ORR of 21%, CBR of 42% with median PFS of 7.3 months, and 6-month PFS of 50.4% were observed in PI3K-mutant tumors (10). This appears comparable with the observed efficacy of letrozole plus lenvatinib despite the lack of biomarker enrichment in our patient population. Overall, this supports further exploration of lenvatinib plus letrozole in CDK4/6 inhibitor–resistant MBC.

We carried out correlative tumor IHC studies to determine the effects of lenvatinib on known target pathways, including the RET, pERK, and VEGFR pathways. Inhibition of RET signaling can target downstream ERK signaling pathway (13), whereas inhibition of VEGFR can reduce neomicrovasculature and improve normalization of tumor vessels, which improves intratumoral drug delivery (28). Our study showed downregulation of RET expression and also its downstream pERK pathway, with IHC studies showing corresponding decrease in pERK IHC expression with treatment. This is similar to a pilot study (n = 10) done in patients with hormone receptor–positive, HER2-negative breast cancer treated with 10 days of vandetinib, another RET inhibitor, preoperatively, which led to significant downregulation of the pERK pathway (29). Comparison between good and poor responders to lenvatinib in our study did not reveal statistically significant differences in baseline levels or degree of change in RET expression levels on IHC, suggesting that future studies need not limit patient selection to only those with RET-positive tumors. IHC also showed evidence of inhibition of the VEGFR pathway, another known target of lenvatinib, through decreased MVD and increased VNI, indicating decreased neovascularization and normalization of vessels through recruitment of pericytes around leaky tumor vasculature. Further correlative studies may uncover potential selective biomarkers for this promising regimen.

A potential limitation of our study was that the original study design targeted to recruit patients in a neoadjuvant setting. However, due to challenges in patient recruitment, the protocol was amended to treat patients with metastatic disease instead, and thus the initial statistical assumption of 50% ORR would have been overly optimistic. We have retrospectively performed sample size estimation assuming an ORR of 10% under the null hypothesis, and an ORR of 25% under the alternative, which would have been more realistic for a population of patients with metastatic disease. In order to achieve a power of 80% with a one-sided alpha of 10%, a total of 31 patients would be required, with at least 6 responses to be considered for further investigation. In our study, we recruited 31 patients in the dose expansion cohort, of whom 30 had evaluable disease. Among these patients, 7 achieved partial response, thus meeting the minimum target for response, which is encouraging. Nonetheless, we acknowledge that this estimation was conducted post hoc and thus should be interpreted with caution.

In conclusion, lenvatinib plus letrozole achieved clinically meaningful antitumor activity that is comparable to contemporary combination endocrine therapies in the second-line-and-beyond setting with a manageable toxicity profile. Encouraging antitumor activity was observed in a subgroup of patients despite prior resistance to aromatase inhibitor monotherapy and/or prior CDK4/6 inhibitor therapy. Our results support the further development of the combination of lenvatinib and letrozole in HR⁺/HER2⁻ advanced breast cancers. A randomized phase II clinical trial of lenvatinib in combination with letrozole in comparison with fulvestrant in a second-line setting is currently in development.

J.S. Lim reports personal fees from AstraZeneca, Roche, DKSH, MSD, Eisai, Novartis, and Pfizer, and grants from CTI biopharma during the conduct of the study. A.L. Wong reports other support from Otsuka Pharmaceuticals, Pfizer, Novartis, and Eisai outside the submitted work. S.G. Ow reports personal fees from Pfizer, AstraZeneca, Roche, Novartis, and Lilly during the conduct of the study. D.S. Tan reports grants and nonfinancial support from Eisai during the conduct of the study; grants, personal fees, and nonfinancial support from AstraZeneca, Eisai, and MSD, grants from National Medical Research Council Clinician Scientist Award, personal fees from GSK, grants and nonfinancial support from Roche, and grants from Pangestu Family Foundation Gynaecological Cancer Research Fund outside the submitted work. R. Sundar reports other support from Bristol-Myers Squibb, Merck, Eisai, Bayer, Taiho, Novartis, MSD, Eli Lilly, BMS, Roche, Taiho, AstraZeneca, DKSH, Roche, AstraZeneca, Taiho, and Eisai and grants from Paxman Coolers outside the submitted work. B. Tai reports other support from Boehringer Ingelheim outside the submitted work, and received royalty from Wiley-Blackwell for the publication of books. B. Goh reports grants and nonfinancial support from MSD and Adagene, and nonfinancial support from Taiho pharmaceuticals and BMS outside the submitted work. S. Lee reports grants, personal fees, and other support from Eisai during the conduct of the study; grants, personal fees, and other support from Pfizer, personal fees from Novartis, AstraZeneca, Eli Lilly, Daiichi Sankyo, and Roche, grants and personal fees from MSD, grants from Karyopharm, and other support from Taiho and ASLAN Pharmaceuticals outside the submitted work. No disclosures were reported by the other authors.

J.S.J. Lim: Data curation, formal analysis, validation, investigation, writing–original draft, writing–review and editing. A.L.A. Wong: Investigation, writing–review and editing. S.G.W. Ow: Investigation, writing–review and editing. N.Y.L. Ngoi: Investigation, writing–review and editing. G.H.J. Chan: Investigation, writing–review and editing. Y.L.E. Ang: Investigation, writing–review and editing. W.Q. Chong: Investigation, writing–review and editing. S.E. Lim: Investigation, writing–review and editing. Y.W. Lim: Investigation, writing–review and editing. M. Lee: Investigation, writing–review and editing. J.R.E. Choo: Investigation, writing–review and editing. H.L. Tan: Investigation, writing–review and editing. W.P. Yong: Investigation, writing–review and editing. R.A. Soo: Investigation, writing–review and editing. D.S.P. Tan: Investigation, writing–review and editing. C.E. Chee: Investigation, writing–review and editing. R. Sundar: Investigation, writing–review and editing. K. Yadav: Investigation, writing–review and editing. S. Jain: Data curation, formal analysis, validation, investigation, writing–original draft, writing–review and editing. L. Wang: Data curation, formal analysis, writing–review and editing. B. Tai: Data curation, software, formal analysis, validation, investigation, writing–original draft, writing–review and editing. B.C. Goh: Investigation, writing–review and editing. S.-C. Lee: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, writing–review and editing.

This study was supported by the National Medical Research Council, Singapore (NMRC/CSA/015/2009, NMRC/CSA-SI/0004/2015, NMRC/CG/012/2013, NMRC/CG/M005/2017_NCIS, MOH-000414) and Eisai Pharmaceuticals. We are grateful for all patients who participated, without whom the study would not have been possible. We would also like to thank Eisai for provision of lenvatinib, and the National Medical Research Council for grant provision for manpower and study support (NMRC/CSA/015/2009, NMRC/CSA-SI/0004/2015, MOH-000414, NMRC/CG/M005/2017_NCIS, and NMRC/CG/012/2013).

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.