Skip to main content
ESMO Open logoLink to ESMO Open
. 2026 Mar 3;11(3):106083. doi: 10.1016/j.esmoop.2026.106083

Real-world external control arm for the single-arm LIBRETTO-001 trial of selpercatinib in RET-mutation-positive medullary thyroid cancer: RECALIB-RET

J Hadoux 1,2, J Hernando 3, KH Wong 4, C Do Cao 2,5, H Lasolle 2,6, C Buffet 2,7, D Benisvy 2,8, F Illouz 2,9, O Schneegans 2,10, R Varnier 2,11,12, D Drui 2,13, S Zanetta 2,14, S Hescot 2,15, M Muller 2,16, C Nascimento 2,17, N Roudaut 2,18, M Khanal 19, S Zheng 20, X Li 5,19, R Singh 19, W Tian 19, AS Afonso 19, B Houot 1,2, I Borget 2,21,22, G Segall 19,∗
PMCID: PMC12969821  PMID: 41780123

Abstract

Background

The single-arm phase I/II LIBRETTO-001 trial demonstrated durable efficacy with selpercatinib in patients with rearranged during transfection (RET)-mutation-positive medullary thyroid cancer (MTC). RECALIB-RET compared effectiveness outcomes using a real-world external control (EC) arm of patients with RET-mutation-positive MTC treated with standard of care (SoC) versus selpercatinib (LIBRETTO-001).

Patients and methods

In this retrospective study, the selpercatinib arm comprised first-line (1L) or second-and-later-line (≥2L) patients from LIBRETTO-001. The EC arm comprised SoC-treated patients, pooled across (i) the French ENDOCAN-TUTHYREF (Refractory Thyroid Tumours) database (1L and ≥2L); (ii) a chart review of European patient medical records (1L and ≥2L) and (iii) a published United States chart review (≥2L). Index treatment was cabozantinib or vandetanib (1L) and any SoC (≥2L; including multikinase inhibitor, chemotherapy or immunotherapy). The primary endpoint was progression-free survival (PFS). The safety profile of SoC in the EC arm was a secondary endpoint. Propensity score matching (PSM) balanced baseline characteristics between treatment arms.

Results

Baseline characteristics of the 1L selpercatinib (n = 116) and EC (n = 107) arms were balanced; PSM reduced the sample size by <30% across arms (n = 84 per arm after PSM). 1L selpercatinib conferred statistically significant PFS benefit over SoC both pre-PSM [median: not reached (NR) versus 24.0 months; P < 0.001] and post-PSM (median: NR versus 26.1 months; P < 0.001). In ≥2L (selpercatinib, n = 179; EC, n = 51), PSM increased attrition, reducing sample sizes by 78.8% (selpercatinib) and 25.5% (EC), to n = 38 per arm. The unadjusted median PFS was significantly longer with ≥2L selpercatinib versus SoC (35.6 months versus 11.6 months; P = 0.005); the difference did not reach significance after PSM. Safety findings were consistent with previously reported findings for SoC.

Conclusion

Selpercatinib conferred significant PFS benefit over SoC in treatment-naïve (1L) patients with RET-mutation-positive MTC, with inconclusive results in the ≥2L setting. Matching retrospective real-world data to prospective trial data is feasible. EC arms to single-arm trials may provide evidence supporting the evaluation of comparative effectiveness.

Key words: selpercatinib, real-world external control arm, multikinase inhibitors, LIBRETTO-001, medullary thyroid cancer, RET mutation

Highlights

  • •

    HTA and regulatory agencies are increasingly accepting EC arms as comparators to single-arm clinical trials.

  • •

    The effectiveness of selpercatinib in patients with MTC was compared post hoc with a matched RWD EC arm.

  • •

    Results support a PFS benefit with 1L selpercatinib versus SoC, both before and after PSM.

  • •

    Matching of retrospective RWD with prospective trial data was achieved using PSM, allowing a comparative analysis.

  • •

    Well-matched or weighted real-world EC arms can be effective comparators to single-arm clinical trial data.

Introduction

Medullary thyroid cancer (MTC) is rare, accounting for <5% of thyroid cancers, but it contributes a disproportionate rate (up to 13%) of thyroid cancer-related deaths.1 Rearranged during transfection (RET) proto-oncogene mutations, detected in ∼70% of MTCs, can be hereditary or somatic and may confer a more aggressive disease course.2, 3, 4 Until the approval of selpercatinib, approved options for the first-line (1L) systemic treatment of advanced MTC were two multikinase inhibitors (MKIs): cabozantinib and vandetanib.5, 6, 7

Benefit with cabozantinib or vandetanib (C/V) over placebo was demonstrated in phase III randomised controlled trials (RCTs) in patients with advanced or metastatic (a/m)MTC irrespective of RET mutational status: ZETA and EXAM, respectively. Both MKIs significantly prolonged median progression-free survival (mPFS) versus placebo, regardless of prior systemic treatment.8,9 No overall survival (OS) benefit versus placebo was reported for either MKI.10,11 In a post hoc analysis of the ZETA trial, the mPFS was 21.4 months with vandetanib in patients with locally a/mMTC,10 and in EXAM, the mPFS was 11.2 months with cabozantinib.9 While both MKIs may be used long term, clinical disease progression is observed. Pre-existing or acquired RET V804M mutations constitute a primary resistance mechanism,12, 13, 14, 15 highlighting the need for a more selective RET inhibitor (sRETi) for patients with RET-driven MTC.

Selpercatinib, a first-in-class, highly selective RET kinase inhibitor, is approved in the United States (US) and Europe for the treatment of patients with RET-mutation-positive a/mMTC who require systemic therapy.16, 17, 18, 19 Accelerated approvals were based largely on the finding of durable efficacy with selpercatinib in the phase I/II, single-arm LIBRETTO-001 trial of selpercatinib in adult patients with RET-mutation-positive a/mMTC (NCT03157128), which reported mPFS durations of 41.4 months in MKI-pretreated patients and not reached (NR; median follow-up 42.4 months) in treatment-naïve (1L) patients.17,18,20,21

The causal interpretation of selpercatinib efficacy estimation in LIBRETTO-001 is hampered by the lack of a randomised control arm, the feasibility of which is highly challenging in this ultra-rare cancer.20 In this clinical setting, an external control (EC) arm (or ‘synthetic comparator’) based on real-world data (RWD) can ‘mimic’ an RCT control arm. This is accepted by regulatory authorities and health technology assessment (HTA) agencies as a valid comparator, per their guidelines and recommendations,22, 23, 24, 25, 26, 27, 28 provided the EC arm is supported by a prespecified protocol and statistical analysis plan that includes strategies to (i) maximise the comparability of the population’s demographic and clinical characteristics, and the study endpoints and methods of data collection; and (ii) minimise bias, confounding and data misclassification.24,29,30

RECALIB-RET explored the feasibility and results of comparison between selpercatinib-treated patients with RET-mutation-positive a/mMTC from LIBRETTO-001 and a matched, standard-of-care (SoC)-treated population in a real-world EC arm, initiated before the interim analysis of the phase III LIBRETTO-531 RCT of selpercatinib versus the physician’s choice of C/V.12 We describe here the findings of this comparative outcomes analysis, separately for 1L and second-line-or-later (≥2L) settings. Real-world safety outcomes during SoC treatment are also reported.

Patients and methods

Data sources

The RWD sources used in this retrospective study were (i) the French Refractory Thyroid Tumours (Tumeurs de la Thyroîde Refractaire; ENDOCAN-TUTHYREF) database, a nationwide network funded by the French National Cancer Institute containing data provided by physicians dedicated to the care of all patients with refractory thyroid cancer (www.tuthyref.com); (ii) a chart review of patient medical records at seven sites (high-volume treatment centres that test for RET alterations) across four countries (excluding France), from which data were extracted into a study-specific case report form by physicians who regularly treat patients with RET-mutation-positive MTC; and (iii) the dataset of a published US chart review including patients with previously treated MTC (≥2L).31 Supplementary Appendix Figure S1, available at https://doi.org/10.1016/j.esmoop.2026.106083 depicts the study design.

All study-mandated variables from the pre-existing databases of the TUTHYREF network and the US study were mapped to those of the study-specific case report form, enabling data pooling (Supplementary Appendix Tables S1 and S2, available at https://doi.org/10.1016/j.esmoop.2026.106083 describe the mapping approaches). Clinical judgement was applied where categorical values did not fully overlap. Data for treatment-effectiveness analyses were available from all RWD sources; safety data were not available from the US chart review.

Patient selection

The selpercatinib arm comprised prospective patient-level data of individuals with RET-mutation-positive a/mMTC enrolled in LIBRETTO-001 who received 1L (1L cohort) or ≥2L (≥2L cohort) selpercatinib. Enrolment commenced in May 2017; data until the January 2023 data cut-off were analysed. The EC arm comprised retrospective patient-level data of individuals with RET-mutation-positive a/mMTC who received SoC therapy (1L or ≥2L setting) from June 2013 to September 2023. Patients were observed from the date of diagnosis to the earliest of death or the date of data extraction.

The comparability of the LIBRETTO-001 trial selpercatinib and the EC arms, including the eligibility criteria, was assessed according to guidance from Thorlund et al.32 (Supplementary Appendix Table S3, available at https://doi.org/10.1016/j.esmoop.2026.106083).

Study objectives

The main objective was to use an EC arm to compare effectiveness, via an adjusted comparison of mPFS, between selpercatinib-treated patients in LIBRETTO-001 and SoC-treated patients in a real-world context. The primary endpoint was propensity score matching (PSM)-adjusted PFS; secondary endpoints were OS and overall response rate (ORR). Another secondary objective was to describe the safety profile of SoC treatment during the index line of therapy (LoT) in the EC arm.

All adverse events (AEs) or serious AEs (SAEs) documented in clinical charts and notes of patients in the EC arm were collected from therapy initiation to 30 days after the last dose or patient death, whichever occurred first. Number and frequency (by patient), distribution by grade (per Common Terminology Criteria for Adverse Events version 4.033) and relatedness to index treatment (determined and documented in each patient’s medical records by their treating physician) were described.

Statistical analysis

Consistent with guidance for the application of ‘synthetic comparators’,25,34,35 the protocol-prespecified balancing and outcomes analysis involved a three-step process to maximise the validity of planned comparisons: (i) definition of the estimand, (ii) adequate selection of ECs from RWD and (iii) choice of an adequate statistical approach targeting the treatment effect [e.g. population-adjusted indirect comparisons using methods such as PSM, inverse probability treatment weighting (IPTW) and entropy balancing (EB)].25,36 The estimand of interest for both IPTW and EB was the average treatment effect in treated individuals (ATT). Estimation of ATT could be biased, as patients in the selpercatinib arm were excluded due to a lack of patients in the real-world EC arm within the specified calliper. We ensured, as far as possible, the similarity of data-collection processes, populations and outcome definitions between arms.32

For primary analyses, PSM was utilised to balance covariates between study arms, with a calliper width of 0.2 standard deviations (SDs) on the logit of propensity scores and a 1 : 1 matching ratio. PSM was chosen as one of the most established and widely used approaches,32,37,38 and is included in guidelines by HTA agencies, including the National Institute for Health and Care Excellence (NICE) and Haute Autorité de Santé.39,40 Sensitivity analyses used IPTW and EB weighting methods. Selection of covariates included in the adjustment model was informed by factors identified through a systematic literature review to be predictive or prognostic of outcomes of patients with MTC, and availability (including completeness) in both arms.

Definitions of time-to-event outcomes (including censoring rules) are provided in Supplementary Appendix Tables S4 and S5, available at https://doi.org/10.1016/j.esmoop.2026.106083. The primary endpoint of PFS in the selpercatinib arm was assessed by investigators. Sensitivity analysis compared PFS assessed in LIBRETTO-001 by an independent review committee (IRC) with physician-assessed PFS in the EC arm. Time-to-event outcomes were analysed using the Kaplan–Meier method and log-rank test. Comparison of time-to-event outcomes used Cox proportional-hazards models, adjusting for the corresponding variables. Outcomes were compared after PSM, IPTW and EB, using logistic regression models, adjusting for covariates.

In the selpercatinib arm, the index LoT was that in which selpercatinib was first administered. For the EC treatment-naïve (1L) cohort, the index LoT was the first LoT (C/V). For the EC pretreated (≥2L) cohort, a proportional randomisation algorithm, corresponding to those treated in 2L, third line (3L) or beyond in LIBRETTO-001, selected the index LoT.41 Any sRETi or LoTs received after sRETi exposure were ineligible as the index LoT.

Safety data were analysed using descriptive statistics.

Results

Patient characteristics and index treatment

Study population attrition is shown in Figure 1. The selpercatinib arm included 116 1L and 179 ≥2L patients, and the EC arm included 107 1L and 51 ≥2L patients. Patient numbers in the EC arm after matching were 84 (1L) and 38 (≥2L). Demographic and clinical characteristics by cohort are provided in Table 1, and by RWD source in Supplementary Appendix Tables S6 (1L) and S7 (≥2L), available at https://doi.org/10.1016/j.esmoop.2026.106083.

Figure 1.

Figure 1

Figure 1

Patient selection in RECALIB-RET study arms. (A) Patient selection for the selpercatinib arms from the phase I/II single-arm LIBRETTO-001 trial. (B) Patient selection for the SoC EC arms from RWD. Green boxes indicate patient exclusions; bold boxes indicate the final RECALIB-RET study arms. 1L, first line; (≥)2L, second (-and-later) line; b.i.d., biduum (2 days); EC, external control; q.d., once daily; MTC, medullary thyroid cancer; RET, gene encoding rearranged during transfection protein; RWD, real-world data; SoC, standard of care. aPatients who received multiple lines of treatment were eligible for both the 1L and ≥2L cohorts; hence, the number of unique patients was <178.

Table 1.

Demographic, clinical and molecular characteristics of patients with a/mMTC in the RECALIB-RET selpercatinib (LIBRETTO-001) and EC (RWD) arms, stratified by prior MTC treatment status (1L versus ≥2L)

Characteristics Selpercatinib arm
EC arm
Treatment-naïve (1L) cohorta (N = 116) Pretreated (≥2L) cohort (N = 179) Treatment-naïve (1L) cohorta (N = 107) Pretreated (≥2L) cohort (N = 51)
Sex
 Male 71 (61.2) 109 (60.9) 80 (74.8) 32 (62.7)
 Female 45 (38.8) 70 (39.1) 27 (25.2) 19 (37.3)
Age at initiation of index LoT (years)
 Number of patients, n 116 179 107 51
 Mean ± SD 55.2 ± 15.4 56.4 ± 15.0 57.4 ± 14.0 53.4 ± 13.9
 Median (IQR) 57.0 (44.0-66.5) 58.0 (47.0-67.0) 60.0 (49.0-68.0) 55.0 (39.0-65.0)
 Range 15.0-87.0 17.0-90.0 24.0-84.0 29.0-77.0
ECOG PS at initiation of index LoT
 0 53 (45.7) 58 (32.4) 44 (41.1) 10 (19.6)
 1 59 (50.9) 108 (60.3) 21 (19.6) 20 (39.2)
 2 4 (3.4) 13 (7.3) 5 (4.7) 5 (9.8)
 3 0 (0.0) 0 (0.0) 0 (0.0) 1 (2.0)
 Missing 0 (0.0) 0 (0.0) 37 (34.6) 15 (29.4)
Country of treatment
 France 10 (8.6) 25 (14.0) 79 (73.8) 30 (58.8)
 Spain 1 (0.9) 4 (2.2) 13 (12.1) 9 (17.6)
 United Kingdom 1 (0.9) 2 (1.1) 13 (12.1) 0 (0.0)
 Germany 0 (0.0) 5 (2.8) 1 (0.9) 0 (0.0)
 The Netherlands 0 (0.0) 0 (0.0) 1 (0.9) 0 (0.0)
 United States 83 (71.6) 115 (64.2) 0 (0.0) 12 (23.5)
Stage at initial diagnosis
 I NR NR 4 (3.7) 4 (7.8)
 II NR NR 7 (6.5) 1 (2.0)
 III NR NR 4 (3.7) 5 (9.8)
 IVA NR NR 39 (36.4) 12 (23.5)
 IVB NR NR 4 (3.7) 2 (3.9)
 IVC NR NR 39 (36.4) 22 (43.1)
 Unknown 116 (100) 179 (100) 10 (9.3) 33 (64.7)
Metastasis at initiation of index LoT 114 (98.3) 175 (97.8) 106 (99.1) 51 (100.0)
 Number of metastatic sites at initiation of index LoT
 Number of patients, n NR NR 82 24
 Missing NR NR 25 27
 Mean ± SD NR NR 2.6 ± 1.1 3.3 ± 0.8
 Median (IQR) NR NR 3.0 (2.0-3.0) 3.0 (3.0-4.0)
 Range NR NR 1.0-6.0 2.0-5.0
Germline RET mutation identified
 Yes 0 (0.0) 0 (0.0) 17 (15.9) 4 (7.8)
 No 0 (0.0) 0 (0.0) 86 (80.4) 32 (62.7)
 Unknown 116 (100.0) 179 (100.0) 4 (3.7) 15 (29.4)
RET M918T mutation
 Yes 66 (56.9) 122 (68.2) 58 (54.2) 28 (54.9)
 No 50 (43.1) 57 (31.8) 30 (28.0) 7 (13.7)
 Unknown 0 (0.0) 0 (0.0) 19 (17.8) 16 (31.4)
Coalterations identified 2 (1.7) 1 (0.6) 0 (0.0) 4 (7.8)
 Unknown 12 (10.3) 3 (1.7) 0 (0.0) 0 (0.0)
Time from the diagnosis of advanced-stage disease to initiation of index LoT (months)
 Number of patients, n 114 179 107 51
 Missing 2 0 0 0
 Mean ± SD 69.1 ± 100.2 76.9 ± 67.6 28.5 ± 48.3 46.2 ± 53.3
 Median (IQR) 32.4 (5.7-97.3) 55.8 (29.9-106.3) 9.4 (3.0-25.2) 31.5 (11.5-59.0)
 Range 0.5-593.1 0.5-341.5 –1.9-251.3 0.7-318.6
Time from the initiation of index LoT to the end of follow-up (months)
 Number of patients, n 116 179 107 47
 Missing 0 0 0 4
 Mean ± SD 41.3 ± 11.7 36.6 ± 17.3 59.5 ± 40.1 35.1 ± 27.1
 Median (IQR) 43.2 (37.3-48.3) 40.5 (23.8-49.7) 52.2 (28.0-83.0) 31.0 (10.8-57.6)
 Range 1.9-63.3 0.4-66.8 7.6-184.8 1.8-110.5

Data are n (%) unless stated otherwise.

1L, first line; ≥2L, second and later line; a/m, advanced/metastatic; EC, external control; ECOG PS, Eastern Cooperative Oncology Group performance status; IQR, interquartile range; LoT, line of therapy; MTC, medullary thyroid cancer; NR, not recorded; RET, gene encoding rearranged during transfection protein; RWD, real-world data; SD, standard deviation.

a

Patients who received cabozantinib or vandetanib as 1L therapy.

1L cohorts

In the 1L cohorts, the median age at index therapy initiation was 57.0 years in the selpercatinib arm and 60.0 years in the EC arm; 38.8% and 25.2%, respectively, were female, and 56.9% and 54.2%, respectively, harboured a RET M918T mutation. Metastases were present at index LoT initiation in 98.3% of the selpercatinib arm and 99.1% of the EC arm. In total, 10 (8.6%) and 79 (73.8%) patients, respectively, were treated in France. The median [interquartile range (IQR)] time from diagnosis of advanced-stage disease to initiation of index LoT, and from initiation of index LoT to end of follow-up was 32.4 months (5.7-97.3 months) and 43.2 months (37.3-48.3 months), respectively, in the selpercatinib arm and 9.4 months (3.0-25.2 months) and 52.2 months (28.0-83.0 months), respectively, in the EC arm (Table 1). Vandetanib was the predominant index treatment in the EC arm [78.5%, versus 21.5% (cabozantinib)]. Treatment was ongoing at the end of the study in 14.0% of the EC arm; the most frequent reason for index LoT discontinuation was disease progression (47.7%). An sRETi was received after index LoT in 57.0% of patients (Supplementary Appendix Table S8, available at https://doi.org/10.1016/j.esmoop.2026.106083).

≥2L cohorts

The median age at initiation of index therapy was 58.0 years (selpercatinib) and 55.0 years (EC), and 39.1% and 37.3%, respectively, were female. A RET M918T mutation was detected in 68.2% (selpercatinib) and 54.9% (EC) of patients, and metastases at initiation of the index LoT were detected in 97.8% and 100%, respectively. In total, 25 (14.0%) patients in the selpercatinib arm and 30 (58.8%) in the EC arm were treated in France (Table 1). The median (IQR) time from diagnosis of advanced-stage disease to initiation of index LoT, and from index LoT initiation to the end of follow-up was 55.8 months (29.9-106.3 months) and 40.5 months (23.8-49.7 months), respectively, for the selpercatinib arm, and 31.5 months (11.5-59.0 months) and 31.0 months (10.8-57.6 months), respectively, for the EC arm (Table 1). The predominant index treatment in the EC ≥2L cohort was an MKI (66.7%), being C/V in 51.0%, and the index LoT for >95% of patients was 2L or 3L (Supplementary Appendix Table S8, available at https://doi.org/10.1016/j.esmoop.2026.106083). A chemotherapy-based regimen was the index treatment for a greater proportion of patients treated in France versus the overall EC arm (Supplementary Appendix Table S9, available at https://doi.org/10.1016/j.esmoop.2026.106083). Treatment was ongoing at the study end in 19.6% of patients; the most frequent reason for discontinuation of the index LoT was disease progression (41.2%). Almost half of the EC ≥2L cohort received an sRETi after the index LoT (Supplementary Appendix Table S8, available at https://doi.org/10.1016/j.esmoop.2026.106083).

Balancing of covariates

Covariates that were captured and highly complete in both arms (included in the adjustment model) are listed in Supplementary Appendix Table S2; the impact of PSM on balancing of covariates is demonstrated in Supplementary Appendix Figure S2, available at https://doi.org/10.1016/j.esmoop.2026.106083.

Although identified as predictive or prognostic factors, data on Eastern Cooperative Oncology Group performance status, tumour grade, RET M918T status, tumour progression status before treatment initiation, the median time from initial diagnosis and the median time from diagnosis of metastatic disease were incomplete, not captured or unavailable in at least one data source in the 1L and ≥2L cohorts. These factors were excluded from the model for one or both treatment settings (Supplementary Appendix Table S1, available at https://doi.org/10.1016/j.esmoop.2026.106083).

1L cohorts

Before adjustment, imbalances in patient demographics and disease characteristics between treatment arms included a higher proportion of females and longer median times from both initial and advanced diagnosis to index LoT initiation in the 1L selpercatinib arm versus EC. These imbalances were reduced after PSM; however, sample sizes for the 1L cohorts were reduced by 26.3% (selpercatinib) and 21.5% (EC; Table 2). Supplementary Appendix Tables S10 and S11, available at https://doi.org/10.1016/j.esmoop.2026.106083, present patient characteristics after IPTW and EB, respectively.

Table 2.

Characteristics of the RECALIB-RET selpercatinib (LIBRETTO-001) and EC (RWD) 1L and ≥2L cohorts of patients with RET-mutation-positive a/mMTC, before and after PSM

Treatment-naïve cohorts (1L)
Characteristics Before PSM (unadjusted)
After PSM (adjusted)
Selpercatinib arm (N = 114)a EC arm (N = 107) Std diffb VRc Selpercatinib arm (N = 84) EC arm (N = 84) Std diffb VRc
Sex (female) 44 (38.6) 27 (25.2) 0.29 NA 24 (28.6) 27 (32.1) –0.08 NA
Age at initiation of index LoT (years)
 Number of patients, n 114 107 — — 84 84 — —
 Missing, n 0 0 — — 0 0 — —
 Mean ± SD 54.9 ± 15.4 57.4 ± 14.0 –0.17 1.2 55.6 ± 14.6 55.4±14.5 0.01 1.02
 Median (IQR) NR 60.0 (49.0-68.0) — — NR NR — —
 Range NR 24.0-84.0 — — NR NR — —
Presence of RET M918T mutation 64 (56.1) 58 (54.2) –0.04 NA 49 (58.3) 49 (58.3) 0.00 NA
Time from initial diagnosis to initiation of index LoT (months)
 Number of patients, n 114 107 — — 84 84 — —
 Missing, n 0 0 — — 0 0 — —
 Mean ± SD 101.0 ± 127.5 68.3 ± 82.7 0.30 2.38 73.5 ± 97.8 67.7 ± 74.8 0.05 1.71
 Median (IQR) NR 28.0 (7.0-110.6) — — NR NR — —
 Range NR 0.0-384.2 — — NR NR — —
Time from advanced-stage diagnosis to initiation of index LoT (months)
 Number of patients, n 114 107 — — 84 84 — —
 Missing, n 2 0 — — 0 0 — —
 Mean ± SD 69.1 ± 100.2 28.6 ± 48.3 0.52 4.3 45.4 ± 56.1 34.5 ± 52.9 0.14 1.12
 Median (IQR) 32.4 (5.7-97.3) 9.4 (3.0-25.2) — — NR NR — —
 Range 0.5-593.1 0.0-251.3 — — NR NR — —
Pretreated cohort (≥2L)
Before PSM (unadjusted)
After PSM (adjusted)
Selpercatinib arm (N = 179) EC arm (N = 51) Std diffb VRc Selpercatinib arm (N = 38) EC arm (N = 38) Std diffb VRc
Sex (female) 70 (39.1) 19 (37.3) 0.04 NA 16 (42.1) 16 (42.1) 0.00 NA
French (yes) 25 (14.0) 30 (58.8) –1.05 NA 14 (36.8) 17 (44.7) 0.19 NA
Age at initiation of index LoT (years)
 Number of patients, n 179 51 — — 38 38 — —
 Missing, n 0 0 — — 0 0 — —
 Mean ± SD 56.4 ± 15.0 53.4 ± 14.0 0.21 1.16 53.3 ± 16.1 54.4 ± 14.5 –0.08 1.24
 Median (IQR) 58.0 (47.0-67.0) 55.0 (39.0-65.0) — — NR NR — —
 Range 17.0-90.0 29.0-77.0 — — NR NR — —
Number of prior lines of therapy received before the index LoT
 Number of patients, n 179 51 — — 38 38 — —
 Missing, n 0 0 — — 0 0 — —
 Mean ± SD 1.9 ± 1.3 1.3 ± 0.5 0.61 6.18 1.3 ± 0.5 1.3 ± 0.6 0.00 0.68
 Median (IQR) 1.0 (1.0-2.0) 1.0 (1.0-2.0) — — NR NR — —
 Range 1.0-8.0 1.0-3.0 — — NR NR — —
Duration of prior line relative to the median (8.9 months)
 ≥Median 86 (48.0) 22 (43.1) 0.10 NA 19 (50.0) 17 (44.7) –0.11 NA
 <Median 81 (45.3) 27 (52.9) –0.15 NA 19 (50.0) 19 (50.0) 0.00 NA
Time from advanced diagnosis to initiation of index LoT (months)
 Number of patients, n 179 51 — — 38 38 — —
 Missing, n 0 0 — — 0 0 — —
 Mean ± SD 76.9 ± 67.6 46.2 ± 53.3 0.51 1.61 50.9 ± 43.3 48.3 ± 57.7 0.04 0.56
 Median (IQR) 55.8 (29.9-106.3) 31.5 (11.5-59.0) — — NR NR — —
 Range 0.5-341.5 0.7-318.6 — — NR NR — —
Cabozantinib exposure before index LoT 83 (46.4) 7 (13.7) 0.76 NA 8 (21.1) 7 (18.4) –0.06 NA
Vandetanib exposure before index LoT 120 (67.0) 38 (74.5) –0.16 NA 30 (78.9) 30 (78.9) 0.00 NA
Other MKI exposure before index LoT 41 (22.9) 7 (13.7) 0.24 NA 4 (10.5) 5 (13.2) 0.07 NA
Chemotherapy exposure before index LoT 21 (11.7) 8 (15.7) –0.12 NA 5 (13.2) 4 (10.5) –0.08 NA
PD as best response to treatment immediately before index LoT 42 (23.5) 10 (19.6) 0.09 NA 9 (23.7) 7 (18.4) –0.13 NA

Data are n (%) unless stated otherwise.

1L, first line; ≥2L, second and later line; a/m, advanced/metastatic; EC, external control arm; IQR, interquartile range; LoT, line of therapy; MKI, multikinase inhibitor; MTC, medullary thyroid cancer; NA, not available; NR, not reported; PD, progressive disease; PSM, propensity score matching; RET, gene encoding rearranged during transfection protein; RWD, real-world data; SD, standard deviation; std diff, standard difference; VR, variance ratio.

a

Two patients were excluded from this analysis due to missing time from advanced diagnosis to initiation of index LoT.

b

Standard difference is the difference in means between groups divided by a measure of SD; values close to 0 indicate better balance.

c

Variance ratio closer to 1 indicates better balance.

≥2L cohorts

Before adjustment, imbalances in patient demographic and disease characteristics included greater proportions of patients treated in France and those with a duration of prior LoT less than the median (across both ≥2L arms), lower proportions with cabozantinib or other MKI exposure before index LoT, and a shorter median time from advanced-stage diagnosis to the initiation of index LoT in the ≥2L EC versus selpercatinib arms. These imbalances were mitigated after PSM, but at the expense of sample size reductions of 78.8% (selpercatinib) and 25.5% (EC; Table 2). Supplementary Appendix Tables S10 and S11, available at https://doi.org/10.1016/j.esmoop.2026.106083, present patient characteristics after IPTW and EB, respectively.

Treatment outcomes

Outcomes were evaluated before and after PSM for patients who received selpercatinib versus SoC (Table 3; Figure 2 and Supplementary Appendix Figure S3, available at https://doi.org/10.1016/j.esmoop.2026.106083). Supplementary Appendix Tables S12 (1L) and S13 (≥2L) present the results of sensitivity analyses of comparisons conducted using IPTW and EB, and Supplementary Appendix Figure S4 presents sensitivity analysis comparing IRC-assessed PFS in the LIBRETTO-001 selpercatinib arm with physician-assessed PFS in the SoC-treated EC arm, all available at https://doi.org/10.1016/j.esmoop.2026.106083.

Table 3.

Efficacy outcomes in the selpercatinib (LIBRETTO-001) versus SoC (EC, RWD) 1L and ≥2L cohorts of patients with RET-mutation-positive a/mMTC, before and after PSM

Treatment naïve cohort (1L)
Pretreated cohort (≥2L)
Before PSM (unadjusted)
After PSM
Before PSM (unadjusted)
After PSM
Selpercatinib arm (N = 116) SoC (N = 107) Selpercatinib arm (N = 84) SoC (N = 84) Selpercatinib arm (N = 179) SoC (N = 51) Selpercatinib arm (N = 38) SoC(N = 38)
INV-assessed PFS (months)
 Median (95% CI) NR (50.7-NR) 24.0 (15.8-35.6) NR (50.7-NR) 26.1 (18.7-38.7) 35.6 (27.6-44.3) 11.6 (4.1-16.3) 27.4 (18.3-44.3) 11.6 (5.4-80.8)
 HR (95% CI) 0.34 (0.22-0.54) 0.34 (0.20-0.58) 0.43 (0.24-0.77) 0.59 (0.31-1.14)
 P value <0.0001 <0.0001 0.005 0.116
 Probability (%) of PFS (95% CI)
 3 years 77.9 (68.7-84.8) 38.9 (28.7-49.0) 77.9 (66.9-85.7) 41.7 (29.8-53.1) 49.8 (41.9-57.3) 25.4 (11.8-41.5) 41.1 (25.3-56.2) 39.7 (22.4-56.6)
 5 years 61.3 (45.3-73.9) 26.7 (17.6-36.7) 59.0 (41.0-73.2) 31.0 (20.0-42.6) 30.5 (21.9-39.5) 25.4 (11.8-41.5) 26.0 (12.1-42.3) 39.7 (22.4-56.6)
 Follow-up durationa (months)
 Median (95% CI) 41.7 (39.1-44.8) 82.1 (55.4-114.1) 45.5 (41.6-47.2) 82.1 (52.5-114.1) 47.0 (44.0-49.7) 16.1 (12.0-34.6) 49.7 (39.5-55.2) 14.8 (10.6-26.7)
OS
 Median (95% CI) NR (NR-NR) 105.8 (82.1-172.1) NR (NR-NR) 105.8 (82.1-172.1) 64.3 (64.3-NR) 63.7 (42.1-NR) 64.3 (33.3-NR) 63.7 (36.3-NR)
 HR (95% CI) 0.46 (0.23-0.91) 0.41 (0.18-0.93) 1.33 (0.69-2.57) 1.25 (0.58-2.68)
 P value 0.026 0.033 0.400 0.564
 Estimate of OS (%), (95% CI)
 3 years 89.9 (82.5-94.3) 83.9 (74.7-90.0) 91.4 (82.8-95.8) 83.7 (73.0-90.4) 71.1 (63.6-77.3) 71.2 (55.3-82.2) 63.0 (45.7-76.2) 71.8 (52.4-84.4)
 5 years 88.7 (80.9-93.4) 68.5 (57.4-77.3) 89.9 (80.8-94.8) 70.8 (58.3-80.2) 61.9 (52.9-69.8) 55.5 (36.6-70.8) 57.6 (40.3-71.5) 56.0 (34.0-73.2)
 Follow-up durationa (months)
 Median (95% CI) 44.5 (42.6-46.3) 71.1 (58.6-83.3) 46.0 (44.5-47.9) 70.9 (55.4-83.3) 46.6 (44.8-49.1) 45.5 (34.6-64.9) 46.6 (41.9-50.8) 45.5 (31.0-64.9)
ORR 84.5 26.2 85.7 28.6 73.7 21.6 76.3 26.3
 OR (95% CI) 17.4 (8.5-35.7) 17.6 (7.8-39.7) 15.8 (6.3-39.7) 24.5 (NR-NR)b
 P value <0.0001 <0.0001 <0.0001 NR
DCR 97.4 77.6 96.4 77.4 93.9 62.7 94.7 60.5
 OR (95% CI) 9.9 (2.6-37.4) 8.1 (2.2-29.2) 42.2 (9.6-185.7) 151.8 (NR-NR)b
 P value 0.001 0.002 <0.0001 NR

1L, first line; ≥2L, second and later line; a/m, advanced/metastatic; CI, confidence interval; DCR, disease control rate; EC, external control; HR, hazard ratio; INV, investigator; MTC, medullary thyroid cancer; NR, not reached; OR, odds ratio; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; PSM, propensity score matching; RET, gene encoding rearranged during transfection protein; RWD, real-world data; SoC, standard of care.

a

Calculated using the reverse Kaplan–Meier method.

b

The generalised Hessian matrix was not a positive definite matrix; the iteration was terminated, and the standard error of the parameter estimates was not output. Although the parameter estimates were available, they should not be used for inference.

Figure 2.

Figure 2

Figure 2

Kaplan–Meier plots for INV-assessed PFS in the selpercatinib (LIBRETTO-001) and SoC(EC,RWD) 1L and ≥2L cohorts. (A) Unadjusted 1L. (B) PSM-adjusted 1L. (C) Unadjusted ≥2L. (D) PSM-adjusted ≥2L. Red line, selpercatinib; blue line, EC. 1L, first line; 2L, second line; CI, confidence interval; EC, external control; HR, hazard ratio; INV, investigator; mPFS, median PFS; NR, not reached; PFS, progression-free survival; PSM, propensity score matching; RWD, real-world data; SoC, standard of care.

1L setting

Unadjusted investigator-assessed mPFS was statistically significantly longer with 1L selpercatinib versus SoC [NR, 95% confidence interval (CI) 50.7 months to NR versus 24.0 months, 95% CI 15.8-35.6 months; hazard ratio (HR) 0.34, 95% CI 0.22-0.54; P < 0.0001]. Benefit persisted after PSM (NR, 95% CI 50.7 months to NR versus 26.1 months, 95% CI 18.7-38.7 months; HR 0.34, 95% CI 0.20-0.58; P < 0.0001; Table 3, Figure 2A and B). The median follow-up for both PFS and OS was markedly longer for those who received SoC. Both unadjusted and PSM-adjusted median OS were 105.8 months (95% CI 82.1-172.1 months) with SoC, and NR (95% CI NR-NR) with selpercatinib due to insufficient follow-up (unadjusted: HR 0.46, 95% CI 0.23-0.91, P = 0.026; adjusted: HR 0.41, 95% CI 0.18-0.93; P = 0.033; Table 3, Supplementary Appendix Figure S3A and B, available at https://doi.org/10.1016/j.esmoop.2026.106083). At the median follow-up for the selpercatinib arm of 46.0 months, the probability of (PSM-adjusted) OS for 1L selpercatinib and SoC was 89.9% (95% CI 80.8% to 94.8%) and 77.5% (95% CI 65.7% to 85.6%), respectively (Supplementary Appendix Figure S3B, available at https://doi.org/10.1016/j.esmoop.2026.106083).

After a median follow-up of 43.2 months (selpercatinib arm) and 52.2 months (EC arm; Table 1), ORR and disease control rate were significantly higher with selpercatinib than with SoC, both before and after PSM (Table 3).

≥2L setting

Unadjusted investigator-assessed mPFS was 35.6 months (95% CI 27.6-44.3 months) and 11.6 months (95% CI 4.1-16.3 months) for ≥2L selpercatinib and SoC, respectively (HR 0.43, 95% CI 0.24-0.77; P = 0.005). After PSM, the mPFS was 27.4 months (selpercatinib) and 11.6 months (SoC); the difference was not statistically significant and had wide confidence intervals (HR 0.59, 95% CI 0.31-1.14; P = 0.116; Table 3, Figure 2C and D). The follow-up duration was approximately three times longer for the selpercatinib arm (both before and after PSM; Table 3). The unadjusted median OS was similar between selpercatinib- and SoC-treated patients (HR 1.33, 95% CI 0.69-2.57; P = 0.400), and remained so after PSM (HR 1.25, 95% CI 0.58-2.68; P = 0.564) after comparable follow-up durations (Table 3, Supplementary Appendix Figure S3C and D, available at https://doi.org/10.1016/j.esmoop.2026.106083). At a median follow-up of 46.6 months for the selpercatinib arm, the probability of (PSM-adjusted) OS for ≥2L selpercatinib and SoC was 57.6% (95% CI 40.3% to 71.5%) and 56.0% (95% CI 34.0% to 73.2%), respectively (Supplementary Appendix Figure S3D, available at https://doi.org/10.1016/j.esmoop.2026.106083).

As the PSM-adjusted models failed to converge for ORR and disease control rate, no inferences could be made about these parameters.

Safety outcomes

Patient-level safety data for selpercatinib-treated patients in the LIBRETTO-001 safety analysis set, aggregated across lines of therapy (N = 324), are published.20 Patient- and AE-level safety summaries for the EC 1L and ≥2L cohorts during index LoT are provided in Supplementary Appendix Table S14, available at https://doi.org/10.1016/j.esmoop.2026.106083.

AEs in the SoC 1L cohort (n = 127) were grade ≥3 in 29.9% of patients. The most frequently occurring grade ≥3 AEs were electrocardiography QT prolongation (6.3%), skin toxicity (5.5%) and diarrhoea (4.7%). SAEs occurred in 14.2% of patients, with the most common ones being vomiting, renal failure and skin toxicity (1.6% of patients each).

In the SoC ≥2L safety analysis population (n = 39), AEs were grade ≥3 in 25.6% of patients. The most common grade ≥3 AEs were weight decrease (10.3%), diarrhoea (5.1%) and asthenia (5.1%). SAEs were reported in 10.3% of patients. The most frequently reported SAE was weight decrease (5.1%); all others occurred in just one (2.6%) patient each.

Discussion

In RECALIB-RET, we successfully generated a real-world EC arm for selpercatinib-treated patients with RET-mutation-positive a/mMTC from the single-arm LIBRETTO-001 trial, allowing exploration of whether RWD can effectively mimic a clinical trial control arm. Furthermore, the data confirmed the findings of the comparative phase III LIBRETTO-531 trial.

The primary endpoint of PSM-adjusted mPFS was longer with selpercatinib versus SoC in both the 1L and ≥2L settings (NR versus 26.1 months, and 27.4 months versus 11.6 months, respectively). In 1L, the magnitude of benefit was statistically significant by all adjustment methods, with HRs comparable to those observed in LIBRETTO-531 (HR 0.31-0.34 versus 0.28).12 In ≥2L, the PFS benefit with selpercatinib over SoC was non-significant when using the primary covariate-balancing method (PSM), but statistically significant when using the secondary method (IPTW).

Treatment effect estimates after PSM for the ≥2L setting may have been affected by the small sample size (N = 38), with wider confidence intervals (0.31-1.14 versus 0.20-0.58) and lower statistical significance (P = 0.116 versus P < 0.001) than for the 1L setting. Residual imbalance in baseline covariates (e.g. number of prior lines before index LoT), and potential for greater uncontrolled confounding and greater natural variability of disease history and OS in the pretreated setting may also have contributed. For example, there was greater heterogeneity of index treatment, whereby almost one-third (29.4%) of patients received chemotherapy regimens. This may be attributable in part to the relatively high proportion of patients who received chemotherapy in France (43.3%), where this is more common practice than elsewhere.42

RECALIB-RET used RWD from three sources, and fitness-for-use was assessed according to guidance provided by HTA authorities based on comparability, similarity, consistency of outcome definitions, reliability and comprehensiveness compared with the prospective data.25,32 Despite efforts to reduce bias by implementing mapping and matching procedures, we cannot rule out residual bias in the EC arm due to the use of multiple RWD sources that may be heterogeneous regarding factors that are generally controlled for in clinical trials, including potential variation in data definitions and patient management, such as assessment methods and frequencies. We conclude that using EC methodology has challenges, especially regarding the balancing of prognostic factors. Although a variety of prognostic factors have been described for MTC [e.g. age, disease stage, calcitonin doubling time (Ct-DT), RET mutational status, number of metastatic sites, time between 1L treatment initiation and diagnosis], only a few studies have focused on patients with metastatic disease, representing a limitation of EC arms for this disease.

RET M918T status, which is reported to indicate poor prognosis in MTC regardless of grade, was used in the present analysis as a balancing factor. Although the prognostic influence of RET M918T in vandetanib-treated patients was inconclusive in the ZETA trial, the response rate was greater in those with a RET M918T mutation,8 and its presence appeared to have a positive prognostic effect on PFS and OS in patients treated with cabozantinib.11 However, no such prognostic impact compared with other RET mutations has been observed in patients treated with selpercatinib, as reported in a pooled post hoc analysis of LIBRETTO-001 and LIBRETTO-531.43 The most promising and robust prognostic factor reported to date is the International MTC Grading System (IMTCGS) grade (recently validated in a metastatic MTC cohort), whereby Ki-67 proliferative index ≥5% and/or mitotic index ≥5 per 2 mm2 and/or necrosis define a higher grade.44,45 IMTCGS grade is associated with another key prognostic parameter (Ct-DT).44,46 As neither IMTCGS grade nor Ct-DT were available herein, the surrogate parameters time from initial diagnosis to initiation of index LoT, and time from advanced diagnosis to initiation of index LoT were used as best alternative indicators of tumour growth rate and tumour burden. Although pre-PSM time from advanced diagnosis to initiation of 1L therapy was much longer in the selpercatinib arm versus the EC arm (potentially driven by varying disease aggressiveness and/or varying clinical management in the respective treating countries), this covariate was well-balanced after PSM. In MTC clinical trials, progressive disease ≤14 months before enrolment serves as an indirect indicator of pretreatment tumour growth rate.11,20 However, this information was not available in either LIBRETTO-001 or any of the RWD comprising the EC arm. Taken together, these missing data and the quest for the best prognostic parameters highlight the bias in evaluating the accuracy, and weaken the strength of our comparisons. Moving forward, the use of data from a previous clinical trial conducted in a similar clinical setting to generate an EC arm will be key towards reducing these comparability issues.

Another challenge with using RWD to generate an EC arm is the higher likelihood of introducing bias than in the clinical-trial setting, in which strict protocols direct prospective data collection including, for example, AE data.47,48 Inherent constraints associated with the retrospective nature of RWD collection in this study (e.g. bias, incomplete or inconsistent data) and typical under-reporting of AEs in routine practice49 limit the validity of comparison with the trial arm. Caution is thus warranted when interpreting our safety findings. However, this study was not designed to enable a valid comparison of safety events between an SoC-treated EC arm and the selpercatinib-treated populations of LIBRETTO-001 or LIBRETTO-531, due to differences in data collection, study setting and index LoT. This emphasises that the use of data from previous clinical trials or high-quality RWD to generate an EC arm is key to a more accurate evaluation and comparison of safety endpoints.

Despite these limitations, our results are strengthened by the inclusion of sensitivity analyses. Although there is no single best method, selection thereof should be based on the best balance after matching or weighting. Therefore, it is important to consider balance diagnostics, such as standardised mean difference and variance ratio, which were achieved with the primary method (PSM). Results with IPTW in the 1L setting were statistically significant and, although balance after PSM was better, the direction of the HR point estimate was the same. In the ≥2L setting, PSM yielded a non-significant PFS treatment effect, while IPTW yielded statistical significance. The EB method did not converge in that setting, indicating the infeasibility of that method in this analysis. In addition, for PSM, the lack of complete matching resulted in substantial sample attrition. For IPTW, the applied weighting using one arm as the reference (weighting of 1) indicates a lesser (<1) or greater (>1) similarity to the patients in the reference arm, yielding an effective sample size (ESS); smaller ESSs reflect greater variability in the results. In RECALIB-RET in the ≥2L setting, PSM resulted in matching of 38 patients from each arm; the remaining 141 (78.8%; selpercatinib) and 13 (25.5%; EC) were excluded. Meanwhile, with IPTW, weighting of the EC arm using the selpercatinib arm (n = 179) as the reference yielded an ESS in the EC arm of 28.

As even the most robust and reliable endpoints may depend on a variety of factors, we included assessments of secondary endpoints. While the longer follow-up period for the EC arm versus the selpercatinib arm facilitated analysis of OS in the former, comparative median OS analyses should be interpreted with caution, considering that (i) in a real-world setting, OS can be confounded by subsequent treatment exposures and (ii) insufficient follow-up in the selpercatinib arm may limit the reliability of comparing median OS between study arms. Consequently, real-world EC arms may detect trends in clinical benefit but may not provide unconfounded, statistically powered and reliable conclusions regarding OS. Furthermore, OS was more variable than PFS in both the 1L and ≥2L settings across all adjustment methods. PFS may be less sensitive to variation in criteria for response than ORR (e.g. RECIST versus non-RECIST), but could still be sensitive to variations in assessment of disease progression.

HTA agency guidelines suggest various methods to address confounding, including PSM.39 For example, NICE guidelines indicate that an absolute standardised difference of <0.1 (as seen herein) generally reflects good balance.39 In addition, the US Food and Drug Administration (FDA) highlights in its guidelines on the use of ECs to support drug approvals the importance of implementing analyses to balance trial arm populations and to address missingness, although no specific strategies are recommended.34 The importance ascribed to the appropriate use of RWD to support new drug submissions is reflected by numerous examples of feedback from regulatory agencies, including the US FDA and European Medicines Agency (EMA), declaring the evidence to be inadequate to support approval due to data missingness, bias, small patient numbers, inappropriate patient populations and confounding, among others.50, 51, 52 As an example, real-world evidence derived from the literature and a compassionate-use programme was submitted to the EMA in support of the efficacy/effectiveness and safety of emapalumab (indicated to treat primary haemophagocytic lymphohistiocytosis). This evidence was rejected because the historical comparator derived from the literature deviated too far from the clinical trial population; risk of bias associated with the population selection method; lack of predefined statistical analyses; extremely wide confidence intervals obtained with non-randomised comparison between the historical control (presumptive external comparator) and the pivotal study; and insufficiently robust data on compassionate use, which precluded conclusions on efficacy.51 In another case, the FDA rejected real-world evidence proffered to support the original marketing application of erdafitinib for locally advanced or metastatic urothelial carcinoma with susceptible FGFR3 or FGFR2 genetic alterations, due to the small sample size, selection bias, misclassifications and missing data.53

Our findings show that the use of RWD to build an EC arm relies on several important elements of methodology, data quality, statistical analysis and design. Although this work might have been insufficient to obtain full EMA approval for selpercatinib and reimbursement from HTA agencies without data from an RCT such as LIBRETTO-531, RECALIB-RET highlights the challenges that must be overcome to fully realise the promise of synthetic control arms for rare diseases. These include data quality and completeness, clear delineation of prognostic parameters in the respective clinical settings, an a priori statistical analysis plan and a plan to address safety data.

In conclusion, this analysis of PSM-adjusted RWD versus single-arm trial data supports the benefit of selpercatinib over SoC in patients with RET-mutation-positive a/mMTC. The quality of the EC RWD was such that with PSM adjustment we were able to match retrospective RWD with prospective trial data and achieve results comparable to an RCT. RECALIB-RET highlights the challenges associated with using a real-world EC arm to replicate head-to-head RCT results while simultaneously achieving longer follow-up, and with providing comparative data in circumstances in which phase III trials are not available.

Acknowledgements

The authors acknowledge Anna Xue (TechData Service Company LLC) for her work on programming the data analysis; Hayley Fenton (IQVIA) and Lizzi Esterberg (RTI Health Solutions) for real-world data source mapping and transformations; Deelan Mareena D’Souza (Eli Lilly and Company) for LIBRETTO-001 trial data management support; Montse Pedros (IQVIA) and Daniel Cuadras (IQVIA) for study protocol and statistical analysis plan support and Eleonora Zonta (IQVIA), Paula Peressini-Lopez (IQVIA), Annabelle Monnot (IQVIA), Ayomikun Ilevbare (IQVIA), Steve Branco (IQVIA), Kirill Dushkin (IQVIA) and Rohan Parikh (RTI Health Solutions) for their project management support throughout data acquisition and data management. Medical writing support was provided by Jacqueline Kolston, PhD, and Greg Plosker (Rx Communications Ltd, Mold, UK), and funded by Eli Lilly.

Funding

This work was supported by Eli Lilly and Company (no grant number).

Disclosure

JHa has received honoraria from Lilly, AAA, Ipsen, Eisai, Bayer, HRA Pharma and ITM Radiopharma; advisory/consultancy fees from Roche and PharmaMar; research/grant funding from Lilly, Novartis and Sanofi; and travel/accommodation expenses from Lilly and Roche. JHe has received speakers’ bureau fees from Eli Lilly, Eisai, Ipsen, Novartis, Angelini and Esteve. KHW has received grants from Bayer; consulting fees from Eli Lilly and Bayer; payment/honoraria for lectures, presentations or speakers’ bureaus from Eli Lilly, Eisai and Bayer; and has participated on advisory boards for Eli Lilly and Bayer. CDC has received honoraria for a lecture from Eli Lilly. HL has received payment or honoraria for lectures, presentations or speakers’ bureaus from Ipsen, and support for attending meetings and/or travel from Merck, Pfizer and Recordati Rare Disease. Her institution has received grants from Ligue Contre le Cancer, Société Française d’Endocrinologie and Fondation ARC pour la recherche contre le cancer. CB has received payment or honoraria for lectures, presentations or speakers’ bureaus from Lilly France. SH has received payment or honoraria for lectures, presentations or speakers’ bureaus from Eisai, and support for attending meetings and/or travel from Eisai. CN’s institution has received funding from Lilly for the LIBRETTO-001 study, and payment or honoraria for lectures, presentations or speakers’ bureaus from Eisai. She has received support for attending meetings and/or travel from Eisai and Ipsen. IB declares that a contract was made between her institution (Gustave Roussy) and Eli Lilly for the extraction of data used for RECALIB-RET. MK, XL, RS, WT, ASA, ASA and GS are employees and minor shareholders of Eli Lilly and Company. SZh was an employee of Eli Lilly and Company at the time the work was conducted. DB, FI, OS, RV, DD, SZa, MM, BH and NR declare no conflicts of interest.

Data Sharing

Lilly provides access to all individual participant data collected during the trial, after anonymisation, with the exception of pharmacokinetic or genetic data. Data are available to request 6 months after the indication studied has been approved in the United States and European Union and after primary publication acceptance, whichever is later. No expiration date of data requests is currently set once data are made available. Access is provided after a proposal has been approved by an independent review committee identified for this purpose and after receipt of a signed data sharing agreement. Data and documents, including the study protocol, statistical analysis plan, clinical study report, blank or annotated case report forms, will be provided in a secure data-sharing environment. For details on submitting a request, see the instructions provided at www.vivli.org.

Supplementary data

Supplementary Material
mmc1.docx (740.1KB, docx)

References

  • 1.Gild M.L., Clifton-Bligh R.J., Wirth L.J., Robinson B.G. Medullary thyroid cancer: updates and challenges. Endocr Rev. 2023;44(5):934–946. doi: 10.1210/endrev/bnad013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Ciampi R., Romei C., Ramone T., et al. Genetic landscape of somatic mutations in a large cohort of sporadic medullary thyroid carcinomas studied by next-generation targeted sequencing. iScience. 2019;20:324–336. doi: 10.1016/j.isci.2019.09.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Kato S., Subbiah V., Marchlik E., Elkin S.K., Carter J.L., Kurzrock R. RET aberrations in diverse cancers: next-generation sequencing of 4,871 patients. Clin Cancer Res. 2017;23(8):1988–1997. doi: 10.1158/1078-0432.CCR-16-1679. [DOI] [PubMed] [Google Scholar]
  • 4.Wirth L.J., Sherman E., Robinson B., et al. Efficacy of selpercatinib in RET-altered thyroid cancers. N Engl J Med. 2020;383(9):825–835. doi: 10.1056/NEJMoa2005651. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Filetti S., Durante C., Hartl D., et al. Thyroid cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2019;30(12):1856–1883. doi: 10.1093/annonc/mdz400. [DOI] [PubMed] [Google Scholar]
  • 6.CAPRELSA® (vandetanib) tablets, for oral use. U.S. Food and Drug Administration; 2022. https://www.accessdata.fda.gov/drugsatfda_docs/label/2022/022405s019lbl.pdf Available at. [Google Scholar]
  • 7.COMETRIQ® (cabozantinib) capsules, for oral use. U.S. Food and Drug Administration; 2023. https://www.cometriq.com/downloads/Cometriq_Full_Prescribing_Information.pdf Available at. [Google Scholar]
  • 8.Wells S.A., Jr., Robinson B.G., Gagel R.F., et al. Vandetanib in patients with locally advanced or metastatic medullary thyroid cancer: a randomized, double-blind phase III trial. J Clin Oncol. 2012;30(2):134–141. doi: 10.1200/JCO.2011.35.5040. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Elisei R., Schlumberger M.J., Muller S.P., et al. Cabozantinib in progressive medullary thyroid cancer. J Clin Oncol. 2013;31(29):3639–3646. doi: 10.1200/JCO.2012.48.4659. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Kreissl M.C., Bastholt L., Elisei R., et al. Efficacy and safety of vandetanib in progressive and symptomatic medullary thyroid cancer: post hoc analysis from the ZETA trial. J Clin Oncol. 2020;38(24):2773–2781. doi: 10.1200/JCO.19.02790. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Schlumberger M., Elisei R., Muller S., et al. Overall survival analysis of EXAM, a phase III trial of cabozantinib in patients with radiographically progressive medullary thyroid carcinoma. Ann Oncol. 2017;28(11):2813–2819. doi: 10.1093/annonc/mdx479. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Hadoux J., Elisei R., Brose M.S., et al. Phase 3 trial of selpercatinib in advanced RET-mutant medullary thyroid cancer. N Engl J Med. 2023;389(20):1851–1861. doi: 10.1056/NEJMoa2309719. [DOI] [PubMed] [Google Scholar]
  • 13.Heilmann A.M., Subbiah V., Wang K., et al. Comprehensive genomic profiling of clinically advanced medullary thyroid carcinoma. Oncology. 2016;90(6):339–346. doi: 10.1159/000445978. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Liu X., Shen T., Mooers B.H.M., Hilberg F., Wu J. Drug resistance profiles of mutations in the RET kinase domain. Br J Pharmacol. 2018;175(17):3504–3515. doi: 10.1111/bph.14395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Vodopivec D.M., Hu M.I. RET kinase inhibitors for RET-altered thyroid cancers. Ther Adv Med Oncol. 2022;14 doi: 10.1177/17588359221101691. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Kroiss M., Sherman E.J., Wirth L.J., et al. 1656P Durable efficacy of selpercatinib in patients (pts) with medullary thyroid cancer (MTC): update of the LIBRETTO-001 trial. Ann Oncol. 2022;33(suppl 7):S1299–S1300. [Google Scholar]
  • 17.RETEVMO® (selpercatinib) capsules, for oral use. Prescribing information. U.S. Food and Drug Administration; 2024. https://www.accessdata.fda.gov/drugsatfda_docs/label/2024/213246s011s013lbl.pdf Available at. [Google Scholar]
  • 18.RETSEVMO® (selpercatinib) capsules, for oral use. Summary of product characteristics. European Medicines Agency; 2024. https://www.ema.europa.eu/en/documents/product-information/retsevmo-epar-product-information_en.pdf Available at. [Google Scholar]
  • 19.FDA approves selpercatinib for medullary thyroid cancer with a RET mutation. 2024. https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-selpercatinib-medullary-thyroid-cancer-ret-mutation Available at.
  • 20.Wirth L.J., Brose M.S., Subbiah V., et al. Durability of response with selpercatinib in patients with RET-activated thyroid cancer: long-term safety and efficacy from LIBRETTO-001. J Clin Oncol. 2024;42(27):3187–3195. doi: 10.1200/JCO.23.02503. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.FDA approves selpercatinib for lung and thyroid cancers with RET gene mutations or fusions. 2020. https://www.fda.gov/drugs/resources-information-approved-drugs/fda-approves-selpercatinib-lung-and-thyroid-cancers-ret-gene-mutations-or-fusions Available at. [DOI] [PubMed]
  • 22.Jahanshahi M., Gregg K., Davis G., et al. The use of external controls in FDA regulatory decision making. Ther Innov Regul Sci. 2021;55(5):1019–1035. doi: 10.1007/s43441-021-00302-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Patel D., Grimson F., Mihaylova E., et al. Use of external comparators for health technology assessment submissions based on single-arm trials. Value Health. 2021;24(8):1118–1125. doi: 10.1016/j.jval.2021.01.015. [DOI] [PubMed] [Google Scholar]
  • 24.Mishra-Kalyani P.S., Amiri Kordestani L., Rivera D.R., et al. External control arms in oncology: current use and future directions. Ann Oncol. 2022;33(4):376–383. doi: 10.1016/j.annonc.2021.12.015. [DOI] [PubMed] [Google Scholar]
  • 25.Lambert J., Lengline E., Porcher R., Thiebaut R., Zohar S., Chevret S. Enriching single-arm clinical trials with external controls: possibilities and pitfalls. Blood Adv. 2023;7(19):5680–5690. doi: 10.1182/bloodadvances.2022009167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Mortier L., Blom A., van Hille B., et al. Avelumab as second-line or later treatment in patients with metastatic Merkel cell carcinoma: analysis of real-world outcomes in France using the CARADERM database linked to the French national healthcare database. Eur J Cancer. 2024;209 doi: 10.1016/j.ejca.2024.114261. [DOI] [PubMed] [Google Scholar]
  • 27.Vanier A., Fernandez J., Kelley S., et al. Rapid access to innovative medicinal products while ensuring relevant health technology assessment. Position of the French National Authority for Health. BMJ Evid Based Med. 2024;29(1):1–5. doi: 10.1136/bmjebm-2022-112091. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Zisis K., Pavi E., Geitona M., Athanasakis K. Real-world data: a comprehensive literature review on the barriers, challenges, and opportunities associated with their inclusion in the health technology assessment process. J Pharm Sci. 2024;27 doi: 10.3389/jpps.2024.12302. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Seeger J.D., Davis K.J., Iannacone M.R., et al. Methods for external control groups for single arm trials or long-term uncontrolled extensions to randomized clinical trials. Pharmacoepidemiol Drug Saf. 2020;29(11):1382–1392. doi: 10.1002/pds.5141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Gray C.M., Grimson F., Layton D., Pocock S., Kim J. A framework for methodological choice and evidence assessment for studies using external comparators from real-world data. Drug Saf. 2020;43(7):623–633. doi: 10.1007/s40264-020-00944-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Parikh R., Hess L.M., Esterberg E., Bhandari N.R., Kaye J.A. Diagnostic characteristics, treatment patterns, and clinical outcomes for patients with advanced/metastatic medullary thyroid cancer. Thyroid Res. 2022;15(1):2. doi: 10.1186/s13044-021-00119-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Thorlund K., Dron L., Park J.J.H., Mills E.J. Synthetic and external controls in clinical trials – a primer for researchers. Clin Epidemiol. 2020;12:457–467. doi: 10.2147/CLEP.S242097. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), Version 4.0. 2017. https://evs.nci.nih.gov/ftp1/CTCAE/CTCAE_4.03/Archive/CTCAE_4.0_2009-05-29_QuickReference_8.5x11.pdf Available at.
  • 34.United States Food and Drug Administration Considerations for the design and conduct of externally controlled trials for drug and biological products. Guidance for industry. 2023 https://www.fda.gov/media/164960/download Available at. [Google Scholar]
  • 35.European Medicines Agency Reflection paper on use of real-world data in noninterventional studies to generate real-world evidence for regulatory purposes. 2025. https://www.ema.europa.eu/en/documents/other/reflection-paper-use-real-world-data-non-interventional-studies-generate-real-world-evidence-regulatory-purposes_en.pdf Available at.
  • 36.Hainmueller J. Entropy balancing for causal effects: a multivariate reweighting method to produce balanced samples in observational studies. Polit Anal. 2012;20(1):25–46. [Google Scholar]
  • 37.Wan F. Propensity score matching: should we use it in designing observational studies? BMC Med Res Methodol. 2025;25(1):25. doi: 10.1186/s12874-025-02481-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Hermans S.J.F., van der Maas N.G., van Norden Y., et al. Externally controlled studies using real-world data in patients with hematological cancers: a systematic review. JAMA Oncol. 2024;10(10):1426–1436. doi: 10.1001/jamaoncol.2024.3466. [DOI] [PubMed] [Google Scholar]
  • 39.National Institute for Health and Care Excellence NICE Real-World Evidence Framework. 2022. https://www.nice.org.uk/corporate/ecd9 Available at. [DOI] [PMC free article] [PubMed]
  • 40.Haute Autorité de Santé (HAS) Real-world studies for the assessment of medicinal products and medical devices. 2021. https://www.has-sante.fr/jcms/p_3284524/en/real-world-studies-for-the-assessment-of-medicinal-products-and-medical-devices Available at.
  • 41.Cui Z.L., Khanal M., Chen Y., Li X., Winfree K.B. MSR105 Proportional randomization method to identify index line of therapy in externally controlled trials. Value Health. 2024;27(6) [Google Scholar]
  • 42.Segall G., Singh R., Jen M.H., et al. Real-world clinical profile, RET mutation testing, treatments, and PROs for MTC in Europe. Eur Thyroid J. 2024;13(1) doi: 10.1530/ETJ-23-0172. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Hadoux J., Elisei R., Hoff A., et al. Efficacy of selpercatinib by RET mutation in RET-mutant medullary thyroid cancer: a combined analysis of LIBRETTO-001 and LIBRETTO-531 trials. Thyroid. 2024;34(suppl 1):A-3–A-4. [Google Scholar]
  • 44.Xu B., Fuchs T.L., Ahmadi S., et al. International Medullary Thyroid Carcinoma Grading System: a validated grading system for medullary thyroid carcinoma. J Clin Oncol. 2022;40(1):96–104. doi: 10.1200/JCO.21.01329. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Porcelli T., Moog S., Bani M.A., et al. Validation of the International Medullary Thyroid Carcinoma Grading System in patients with distant metastases. Thyroid. 2025;35(11):1268–1276. doi: 10.1177/10507256251388028. [DOI] [PubMed] [Google Scholar]
  • 46.Barbet J., Campion L., Kraeber-Bodere F., Chatal J.F, GTE Study Group Prognostic impact of serum calcitonin and carcinoembryonic antigen doubling-times in patients with medullary thyroid carcinoma. J Clin Endocrinol Metab. 2005;90(11):6077–6084. doi: 10.1210/jc.2005-0044. [DOI] [PubMed] [Google Scholar]
  • 47.Blonde L., Khunti K., Harris S.B., Meizinger C., Skolnik N.S. Interpretation and impact of real-world clinical data for the practicing clinician. Adv Ther. 2018;35(11):1763–1774. doi: 10.1007/s12325-018-0805-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Dang A. Real-world evidence: a primer. Pharmaceut Med. 2023;37(1):25–36. doi: 10.1007/s40290-022-00456-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Shafei L., Mekki L., Maklad E., et al. Factors that influence patient and public adverse drug reaction reporting: a systematic review using the theoretical domains framework. Int J Clin Pharm. 2023;45(4):801–813. doi: 10.1007/s11096-023-01591-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Zong J., Rojubally A., Pan X., et al. A review and comparative case study analysis of real-world evidence in European regulatory and health technology assessment decision making for oncology medicines. Value Health. 2025;28(1):31–41. doi: 10.1016/j.jval.2024.09.007. [DOI] [PubMed] [Google Scholar]
  • 51.Bakker E., Plueschke K., Jonker C.J., Kurz X., Starokozhko V., Mol P.G.M. Contribution of real-world evidence in European Medicines Agency’s regulatory decision making. Clin Pharmacol Ther. 2023;113(1):135–151. doi: 10.1002/cpt.2766. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Purpura C.A., Garry E.M., Honig N., Case A., Rassen J.A. The role of real-world evidence in FDA-approved new drug and biologics license applications. Clin Pharmacol Ther. 2022;111(1):135–144. doi: 10.1002/cpt.2474. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Feinberg B.A., Gajra A., Zettler M.E., Phillips T.D., Phillips E.G., Kish J.K. Use of real-world evidence to support FDA approval of oncology drugs. Value Health. 2020;23(10):1358–1365. doi: 10.1016/j.jval.2020.06.006. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material
mmc1.docx (740.1KB, docx)

RESOURCES