Lenvatinib in Radioiodine-Refractory Differentiated Thyroid Cancer: Comparing SELECT Pivotal-Trial Efficacy with Real-World Effectiveness

Introduction

Lenvatinib (Lenvima), an oral multikinase inhibitor targeting vascular endothelial growth factor receptors (VEGFR) 1–3, fibroblast growth factor receptors (FGFR) 1–4, platelet-derived growth factor receptor-α (PDGFRα), RET, and KIT, is widely recommended as a first-line systemic therapy option for progressive radioiodine-refractory differentiated thyroid cancer (RR-DTC). Approval was underpinned by the pivotal phase 3 SELECT trial, which demonstrated an unprecedented progression-free survival (PFS) advantage over placebo. However, the SELECT population—predominantly fit, younger, and closely monitored—does not fully represent the heterogeneous patients encountered in routine oncology practice. Post-approval real-world evidence (RWE) from North America, Europe, Japan, Taiwan, and other regions has since provided critical insights into how lenvatinib performs across broader populations, with important implications for dosing strategy, adverse event (AE) management, and overall survival (OS) interpretation.

SELECT Trial: Efficacy and Safety Benchmarks

The SELECT trial randomized 392 patients (261 lenvatinib, 131 placebo) with progressive RR-DTC, requiring measurable disease and documented radiologic progression within 13 months prior to randomization. Approximately 95% of patients had Eastern Cooperative Oncology Group (ECOG) performance status 0–1. Lenvatinib was initiated at a fixed 24 mg once daily, with protocol-specified stepwise dose reductions permitted for toxicity 57.

The primary endpoint was met with statistical and clinical significance: median PFS was 18.3 months with lenvatinib versus 3.6 months with placebo (hazard ratio [HR] 0.21; 99% CI 0.14–0.31; p<0.001). Updated analyses confirmed a median PFS of 19.4 months 5. The objective response rate (ORR) was 64.8% versus 1.5%, and disease control was achieved in approximately 87% of lenvatinib-treated patients 57. Among responders, median duration of response reached 33.1 months.

Despite these impressive efficacy results, SELECT revealed a substantial toxicity burden. Grade ≥3 treatment-related adverse events (TRAEs) occurred in 75.9% of lenvatinib-treated patients. The most frequent grade ≥3 AEs were hypertension (41.8%), diarrhea (18.8%), and fatigue/asthenia (13.8%). Proteinuria occurred in 31% of patients (grade ≥3 in 10%). Dose interruptions were required in 82.4% of patients, dose reductions in 67.8%, and AE-driven permanent discontinuation in 14.2%. The mean achieved dose was 17.2 mg/day despite a 24 mg planned starting dose—a finding that underscores the inherent need for dose modification even under tightly controlled trial conditions 15.

OS interpretation from SELECT is problematic: median OS was not reached in either arm and the between-arm HR was 0.73 (p=0.10), not statistically significant, attributable primarily to a 71.4–89% crossover rate from placebo to open-label lenvatinib 57. Post hoc analyses do suggest an OS benefit in patients aged ≥65 years (HR 0.53; p=0.02) and in those with lung metastases ≥1 cm 79.

Real-World Populations: Key Differences from SELECT

Post-approval cohorts consistently document older, more comorbid, and more heavily pretreated patients than those enrolled in SELECT. The Austrian RELEVANT cohort (n=43) had a median age of 70 years, compared with approximately 63–64 years in SELECT, and included patients with brain metastases (12%) and high rates of comorbidities 6. Canadian series reported 62% of patients with ECOG ≥2 at lenvatinib initiation—a dramatic contrast to the 5% ECOG ≥2 prevalence in SELECT 1. A Japanese real-world cohort (median age 70.5 years, median body weight 54.7 kg) reflected a distinctly smaller-build and frailer population 3. A large US multicenter chart review (n=308) documented a broader case-mix where 62% started at 24 mg, though adverse event data were not systematically collected 2. The Italian cohort explicitly noted that ECOG performance status 0–1 and radioactive iodine (RAI)-avid disease were independently associated with superior PFS and OS 6.

Dose Modification Patterns in Routine Practice

A defining feature of real-world lenvatinib use is the frequency and magnitude of dose modifications—in many cohorts exceeding even the already high rates documented in SELECT. The Austrian RELEVANT study reported dose reductions in 91% of patients, with a median maintenance dose of 14 mg/day 6. In the Taiwanese cohort (n=65), the median maintenance dose within the first three months was 10 mg/day 6. The US multicenter chart review reported that 90.3% of patients remained on their starting dose, though AE data were absent, limiting dose-change interpretation 2.

A particularly instructive analysis from Japan compared full-dose initiation (24 mg/day; n=18) with reduced-dose initiation (≤14 mg/day; n=20). Among full-dose starters, 60.8% discontinued within approximately 30 days due to intolerable AEs, with five patients never resuming therapy. In the reduced-dose group, only 25.0% discontinued within the same window (p=0.018) 38. Crucially, when daily dose was normalized to body weight, the exposure was comparable between groups (0.10 vs. 0.11 mg/kg/day; p=0.26), and multivariate analysis found that starting dose was not an independent predictor of PFS or OS 3. These findings collectively suggest that body-weight-adjusted dosing may be more physiologically meaningful than a fixed nominal dose—especially in smaller-build Asian populations—and that preserving continuous drug exposure matters more than achieving peak dose intensity.

A post hoc SELECT analysis further corroborated this principle: patients with dose interruptions lasting <10% of total treatment duration had significantly longer PFS and higher ORR than those with longer interruptions (median PFS not reached vs. 12.8 months), with the shorter-interruption group also achieving a higher mean dose intensity (20.1 vs. 14.6 mg/day) 56. In the prospective COLLECT study (n=262), patients who underwent planned drug holidays—proactive, scheduled treatment breaks using schedules such as 14-days-on/7-days-off—achieved significantly better 1-year OS (95.8% vs. 81.5%; HR 0.311; p=0.0047) and 1-year PFS (94.5% vs. 83.5%; HR 0.307; p=0.019) compared with continuous daily dosing, without compromising ORR (66.7% vs. 61.6%) 8.

Adverse Event Management and AE-Driven Discontinuation

Across all real-world cohorts, AE incidence exceeds 90% of treated patients. Hypertension remains the most common and clinically impactful toxicity, reported in 67.8% of SELECT patients (grade ≥3: 41.8%), 71% in the Austrian cohort (grade ≥3: 26%), and up to 87% in Japanese patients—of whom 80% experienced grade ≥3 hypertension, likely reflecting pharmacokinetic differences related to lower body weight 16. Despite higher hypertension rates, Japanese real-world cohorts achieved lower AE-driven discontinuation (3.3%) versus SELECT (14.2%), attributable to more aggressive early dose reduction and antihypertensive management 56.

Proteinuria occurred in 31% of SELECT patients (grade ≥3: 10%), rising to 71% in Austria (grade ≥3: 7%) and 63% in Japanese SELECT patients 156. The lower grade ≥3 rate despite higher overall incidence in real-world cohorts reflects systematic urine monitoring with early dose interruption before severity escalation. Additional clinically relevant AEs include diarrhea, weight loss (up to 70% in Austrian cohort), fatigue, hand-foot syndrome (palmar-plantar erythrodysesthesia syndrome [PPES]), and anorexia. Gastrointestinal fistula, though rare in SELECT (1.5%), reached 14.7% in a single-center analysis of 95 RR-DTC patients, with tumor infiltration of the trachea, bronchus, or esophagus and poorly differentiated histology identified as significant risk factors 56. In the COLLECT cohort, planned drug holidays were associated with 0% fistula incidence versus 2.8% in the continuous-dosing group 8.

Real-world AE-driven permanent discontinuation ranges from 3.1% (Taiwan) to 48% (Canadian single-center, a high-comorbidity cohort where nearly all patients started at ≤20 mg) 126. The Canadian series documented grade ≥3 AEs in 92.6% of patients, including rare but serious events such as posterior reversible encephalopathy syndrome (PRES) and Takotsubo cardiomyopathy, reinforcing the need for vigilant monitoring beyond standard toxicities 1.

Overall Survival: Interpretation and Limitations

OS comparison between SELECT and real-world cohorts requires particular caution. The SELECT trial did not demonstrate a statistically significant OS benefit as a primary analysis, primarily because 71.4–89% of placebo-arm patients crossed over to open-label lenvatinib upon progression 157. Post hoc analyses accounting for crossover and age subgroups suggest an OS advantage, particularly in patients aged ≥65 years, though conclusions depend on modeling assumptions 7.

In real-world settings, OS is further confounded by prognostic heterogeneity (ECOG status, RAI avidity, molecular alterations including TP53 mutations identified as predictors of early discontinuation in the Canadian cohort), post-progression therapies (cabozantinib, BRAF/RET-directed agents, pembrolizumab combinations, radiotherapy), and competing causes of death in elderly populations 12. The Austrian cohort reported 74% two-year OS; the Canadian cohort reported median OS of 72.9 months; the Italian cohort reported 70 months; the US multicenter cohort (n=308) estimated two-year OS at 78.4% and six-year OS at 57.0%, with median OS not reached 26. The Japanese cohort reported a median OS of 45.4 months 38. These figures are broadly encouraging but are not directly comparable with SELECT due to all the confounders enumerated above. PFS and ORR remain the most reliable pragmatic endpoints for evaluating lenvatinib benefit in routine practice.

Clinical Optimization: Patient Selection and Practical Implications

Evidence synthesis supports several actionable recommendations. First, the approved 24 mg/day starting dose remains appropriate for fit patients (ECOG 0–1) with robust monitoring infrastructure, but individualized lower starting doses (10–20 mg/day, or ≤14 mg/day in patients with body weight <55–60 kg or age >70 years) reduce early discontinuation without clearly compromising body-weight-adjusted dose exposure or PFS/OS 38. Second, proactive, multidisciplinary AE prevention—including home blood pressure monitoring with antihypertensive initiation or escalation for grade ≥2 hypertension, baseline and monthly urinalysis for proteinuria, early dietitian involvement for weight loss, and dermatologic support for PPES—is the single most impactful strategy for maintaining dose intensity and avoiding unnecessary treatment cessation 156. Third, minimizing interruption duration through rapid-cycle dose modifications—rather than waiting for AE resolution before restarting—appears more important than preserving the nominal dose level, as demonstrated by the SELECT post hoc analysis and COLLECT planned drug holiday data 58. Fourth, patient selection favoring earlier initiation before ECOG deterioration, NLR elevation, or excessive tumor burden accumulation yields the most favorable outcomes 56. Finally, shared decision-making should realistically frame the expectation that dose reduction and interruption are not treatment failures but are integral to the therapeutic strategy, and that >90% of patients will require modification to maintain lenvatinib long-term 13.

Comparison Table: SELECT Pivotal Trial versus Real-World Evidence