The impact of targeted therapies on molecular alterations identified by an institutional molecular tumor board: an approach based on ESCAT classification

Abstract

Background

The European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets (ESCAT) classification system provides a standardized framework for categorizing genomic alterations (GAs) of patients with recurrent, metastatic, or rare cancer. This study aimed to present outcomes of patients discussed at the molecular tumor board (MTB) in general and according to ESCAT.

Patients and methods

We included 1226 patients with recurrent and/or metastatic cancer presented at the MTB from 2018 to 2022. Clinical and demographic data collected included age, gender, type of specimen, tumor type, number of prior treatments received, techniques used for molecular analyses, GAs identified, MTB recommendations, and inclusion or not into a clinical trial. The clinical endpoints collected were overall response rate (ORR), progression-free survival (PFS), and overall survival (OS), and were correlated with ESCAT.

Results

Successful molecular profiling was carried out in 895 of 1226 (73%) patients. Actionable GAs were found in 595 (49%) patients, and 206 (17%) patients were oriented to matched therapies. Eventually, 101 (8%) patients received a matched therapy. For these patients, PFS and OS were significantly longer for GAs classified as ESCAT tiers I/II, compared with tiers III/IV (P = 0.009 and P = 0.014, respectively).

Conclusions

Detection of actionable GAs through MTB molecular screening enabled to treat 8% of patients with matched therapy. Patients treated with matched therapy based on ESCAT tiers I/II had statistically longer PFS and OS, compared with ESCAT tiers III/IV.

Highlights

• •49% of patients discussed in the MTB had an actionable GA.
• •8% of patients discussed in the MTB received matched therapy.
• •Matched therapy-treated patients based on ESCAT tiers I/II had a longer survival.
• •ESCAT classification appears to be a suitable aid for prioritizing actionable GAs.

Introduction

Precision medicine in oncology aims to tailor treatments to the individual genomic profile of patients’ tumors, allowing development of customized cancer treatment plans. Advances in genomic sequencing technologies have made it possible to identify genomic alterations (GAs) driving tumor growth, facilitating the development of personalized treatment options. In addition, administering therapies tailored to driver GAs has shown favorable clinical outcomes across various cancer types.1, 2, 3, 4, 5, 6

Molecular tumor boards (MTBs) have emerged as valuable platforms for multidisciplinary discussions, integrating genomic data and pathological and clinical expertise to optimize treatment decisions for patients with cancer. MTBs play an increasingly important role in optimizing therapeutic management by interpreting molecular profiling.⁷ The main goal is to translate this molecular information into matched therapy recommendations⁷ aiming to improve patients’ outcome.⁸ Molecular profiling can also lead to the identification of acquired, primary, or germline molecular resistances, which may influence response to treatment. In addition, MTBs play a crucial role in revealing inherited cancer syndromes,⁹ enabling recommendations for additional germline genetic tests and appropriate counseling for early cancer detection in both patients and their at-risk family members.

In the literature, the assessed impact of MTBs in oncology is highly variable.¹⁰ According to a systematic review on 14 articles presenting results of MTBs, the frequency of receiving MTB-recommended matched therapy ranged from 11% to 43%, with overall response rates (ORRs) ranging from 0% to 67%.⁶ Prior analysis of our MTB showed that 10% of patients received therapies matching GAs based on MTB recommendations.¹¹ According to Heinrich et al.,¹² matched therapy recommendations were provided for 376/1000 patients (38%), but only 17% of them (6% in total) received matched therapy following MTB recommendations. These results were consistent with Giacomini et al.¹³ According to Liu et al.,¹⁴ only 3 out of 115 included patients (3%) received matched therapy, and only 1 of these patients experienced an improved clinical outcome. These heterogeneous results may be due to differing interpretations of actionability levels and evidence of potential therapeutic targets. With the increasing detection of GAs, their interpretation has consequently become more complex. The degree of interpretation variability potentially undermines the efficacy of matched therapies. To standardize patients’ orientation according to their molecular profile, the ESMO published in 2018 a scale based on actionability of GAs assessed in clinical or preclinical studies [European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets (ESCAT)].¹⁵ Its goal is to rank a GA and associated matched therapy according to six levels of evidence defined as tiers, based on their actionability according to the current literature and trials results.

In 2022, Martin-Romano et al.¹⁶ assessed the clinical impact of ESCAT in a prospective study including 552 patients. ESCAT tier I was significantly associated with improved ORR and progression-free survival (PFS). In the SAFIR-02 Breast randomized trial, patients with GAs classified as tiers I and II on ESCAT classification who received a matched therapy had longer PFS compared with patients receiving standard maintenance therapy.¹⁷ These studies suggest an added value of ESCAT classification for clinical routine in MTB and molecular screening programs.

Considering all these elements, the objective of this work was to conduct a retrospective study to present the outcomes of patients discussed at our MTB from 2018 to 2022, and to assess the clinical impact of prioritizing matched therapies according to the actionability level of GAs as defined by ESCAT in patients with recurrent and/or metastatic cancer.

Patients and methods

Patient selection

Our study was based on the analysis of MTB reports and collection of various clinical data of 1226 consecutive patients discussed at the MTB between January 2018 and December 2022. Inclusion criteria were adult patients with recurrent and/or metastatic cancer or a rare cancer, histologically proven with an available pathology report.

The MTB meets on a weekly basis and includes medical oncologists, radiologists, pathologists, geneticists, medical biologists, and bioinformaticians. Patients are discussed on medical oncologists’ requests. Institutional informed consent for clinical data collection, tumor samples, and molecular analysis was obtained from patients enrolled in this study. The MTB aims to guide patients toward early-phase clinical trials to access innovative treatments based on their molecular tumor profiling. Our MTB is part of the France Genomic Medicine Initiative 2025 (FGMI 2025; https://pfmg2025.aviesan.fr/). The overall MTB workflow is detailed in Supplementary Figure S1, available at https://doi.org/10.1016/j.esmorw.2024.100092.

Molecular profiling

The techniques used included next-generation sequencing (NGS) DNA panels covering 23-571 genes of theranostic interest, using a custom NGS panel known as DRAGON (Detection of Relevant Alterations in Genes involved in Oncogenetics by NGS), commercially available as SureSelect CD Curie CGP by Agilent. Furthermore, RNA analysis was conducted locally using the Archer VariantPlex Comprehensive Thyroid and Lung kit, which detects 20 gene fusions. Whole-genome sequencing, whole-exome sequencing, and RNA sequencing were carried out at the SeqOIA platform via FGMI2025. Tumor DNA/RNA extraction was carried out after evaluation of tumor cells content to allow macrodissection of the sample to increase tumor cellularity. Screening of mutations, microsatellite instability (MSI) status, tumor mutational burden (TMB) evaluation, and homologous recombination deficiency scoring were carried out by targeted NGS using several panels ranging from 23 to 571 genes of theranostic interest or using whole-exome sequencing and whole-genome sequencing. Fusion transcripts of theranostic interest were screened using Archer VariantPlex CLT kit or whole transcriptome RNA-seq.

Bioinformatics

After tumor sequencing, bioinformatics analyses were carried out to detect single-nucleotide variants and indels, copy number variations, MSI status, mutational signatures, TMB scores, homologous recombination deficiency status, and fusion transcripts (Supplementary Materials and Methods, available at https://doi.org/10.1016/j.esmorw.2024.100092). Preparation of samples, sequencing protocol, bioinformatics analyses, and validation of detected alterations are detailed in the Supplementary Materials and Methods,18, 19, 20, 21, 22 available at https://doi.org/10.1016/j.esmorw.2024.100092.

Levels of actionability of genomic alterations according to ESCAT

The ESCAT classification was used to retrospectively classify GAs from patients who received matched therapies according to MTB recommendations.¹⁵ Each matched therapy and associated GA required a comprehensive literature search. For GAs indicating matched therapy in the type of cancer approved by the health authorities, primarily the FDA (Food and Drugs Agency), ESCAT tier was I. ESCAT tier II was assigned to matched therapies assessed on a GA in trials with significant clinical benefit but without any available overall survival (OS) data. ESCAT tier III-A was assigned for GAs supporting matched therapy in a different cancer type. Tier III-B means that matched therapy was approved for a GA belonging to the same signaling pathway. In the absence of clinical evidence with preclinical results on actionability, the GAs were classified as tier IV. Finally, ESCAT tier X was assigned for GAs that had no available data on actionability in the literature. Sources were searched on PubMed and then discussed with two MTB experts (CLT and MK). The final classification was validated with one clinical biologist (IB), two MTB experts (CLT and MK) and a medical oncologist (RS).

Clinical variables collection

The following clinical and demographic data were collected: (i) pre-MTB data, including age, gender, cancer type, previous GAs detected, and number of prior treatments, defined as number of lines of treatment received from the diagnosis of metastatic disease; (ii) MTB data, including samples selected, molecular analyses carried out, GAs identified, and MTB recommendations; and (iii) MTB follow-up data, including treatment received after the MTB, response to matched therapy received according to RECIST version 1.1 reviewed by a specialist in oncology in the corresponding cancer type, and survival outcome with the date of progression defined as first progression associated with matched therapy received after inclusion in our MTB according to RECIST version 1.1, date of last follow-up, and date of death. MTB-related data were collected weekly after each MTB into an Excel (Microsoft Corp., Redmond, WA) file from patients’ medical file by the MTB coordination team from January 2018 to December 2022. From 2 January to 31 May 2023, these data were retrospectively verified, updated, and completed from patients’ medical files by a medical and Master of Sciences student (KRNA) supervised by an oncologist (RS).

Time-to-event endpoints, such as PFS and OS, were analyzed by Kaplan–Meier estimates, and were compared using log-rank tests for univariate analysis or Cox proportional hazard regression model for multivariate analysis (GraphPad Prism v10; GraphPad Software Inc./Insight Partners/Dotmatics, Boston, MA).

Results

Patient characteristics

Among the 1226 consecutive patients included from 8 January 2018 to 19 December 2022, 288 patients (23%) had samples with a low tumor cell content (i.e. low tumor purity), resulting in NGS results that did not accurately reflect tumor analysis. In addition, 43 patients (3%) had insufficient DNA/RNA quality or quantity after extraction, leading to technical failure. Consequently, molecular profiling was contributive for 895 patients (73%; Figure 1 and Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2024.100092). Actionable GAs were found in 595 out of these 895 patients (66%). Eventually, 206 patients (17%) were suggested to receive matched therapies (when available), and 101 patients (8%) eventually received therapy matching a GA as part of clinical trials (n = 55, 54%) or off-label use (n = 46, 46%). Reasons for not pursuing treatment were poor clinical condition (n = 22, 2%), no disease progression (n = 19, 1%), treatment already received (n = 17, 1%), therapeutic alternative (n = 14, 1%), death (n = 12, 1%), lost to follow-up (n = 11, 1%), ineligible patient for the suggested clinical trial (n = 4, <1%), no drug available (n = 2, <1%), patient refusal (n = 2, <1%), screening failure for the suggested clinical trial (n = 1, <1%), refusal of Authorization for Temporary Use by pharma (n = 1, <1%).

Figure 1.

Flowchart. MTB, molecular tumor board. ^aPoor clinical condition, no available material, no indication for molecular analyses or death before analysis. ^bReasons for not receiving matched therapy: death (n = 12), screen failure (n = 1), immediate progression (n = 1), ineligible patient (n = 4), refusal of authorization for temporary use by the laboratory (n = 1), lost from sight (n = 11), no trial available (n = 2), patient refusal (n = 2), poor clinical condition (n = 22), stable disease (n = 19), therapeutic alternative (n = 13), treatment already received (n = 17). ^cDeath (n = 3), stable disease (n = 5) and complete response (n = 1). ^dDeath (n = 1), progressive disease (n = 2), and stable disease (n = 3).

In the total population, the median age was 59 years (range 18-95 years), with the majority being females (68%; Table 1). The most frequent cancer types were gastrointestinal (n = 306, 25%), followed by breast (n = 275, 22%), and gynecological cancers (n = 180, 15%). The median number of prior lines of treatment was 2 (range 0-17). Patients had previously received chemotherapy (n = 969, 79%), molecularly targeted therapy (n = 479, 39%), and/or immunotherapy (n = 240, 20%). The characteristics of the overall population were reflected in the population of 101 patients with actionable GAs and treated with matched therapy (Table 1).

Table 1.

Patient characteristics

Characteristics	Cancer type	All patients (n = 1226)	Patients with actionable genomic alterations and matched therapy (n = 101)	Progression-free survival P valuea (n = 101)	Overall survival P valuea (n = 101)
Age, years				0.4810	0.6027
Median		59	60
Range		18-95	18-95
Sex, n (%)				0.6361	0.4934
Male		389 (32)	30 (30)
Female		837 (68)	71 (71)
Tumor type, n (%)				0.1843	0.2981
Gastrointestinal		306 (25)	34 (34)
	Pancreatic cancer	111 (9)	13 (13)
	Colorectal cancer	95 (8)	10 (11)
	Cholangiocarcinoma	40 (3)	6 (6)
	Gastroesophageal cancer	31 (2)	2 (2)
	Biliary tract cancer	4 (<1)	0 (0)
	Peritoneal cancer	5 (<1)	1 (1)
	Bowel cancer	8 (<1)	0 (0)
	Anal cancer	10 (<1)	2 (2)
	Liver cancer	1 (<1)	0 (0)
Breast		275 (22)	19 (18)
	Hormone receptor-positive breast cancer	170 (14)	10 (10)
	Triple-negative breast cancer	105 (8)	8 (8)
Gynecologic		180 (15)	18 (18)
	Ovarian cancer	86 (7)	5 (5)
	Endometrial cancer	38 (3)	5 (5)
	Cervical cancer	36 (3)	5 (5)
	Vulvar cancer	11 (1)	2 (2)
	Vaginal cancer	7 (<1)	0 (0)
	Pelvic cancer	2 (<1)	0 (0)
	Fallopian tube cancer	1 (<1)	1 (1)
Head and neck		78 (6)	4 (4)
	Squamous cell carcinoma	45 (4)	3 (3)
	Parotid adenocarcinoma	7 (<1)	0 (0)
	Adenoid cystic carcinoma	15 (1)	1 (1)
	Other (adenocarcinoma, undifferentiated carcinoma, glomus tumors, paraganglioma)	11 (<1)	0 (0)
Genitourinary		36 (3)	3 (3)
	Bladder cancer	16 (1)	2 (2)
	Prostate cancer	11 (1)	1 (1)
	Kidney cancer	5 (<1)	0 (0)
	Urinary tract cancer	4 (<1)	0 (0)
Sarcoma		84 (7)	8 (8)
	Leiomyosarcoma	14 (1)	2 (2)
	Undifferentiated sarcoma	10 (<1)	0 (0)
	Osteosarcoma	9 (1)	1 (1)
	Liposarcoma	7 (<1)	0 (0)
	Sarcomatoid carcinoma	5 (<1)	1 (1)
	Carcinosarcoma	4 (<1)	0 (0)
	Malignant peripheral nerve sheath tumor	4 (<1)	1 (1)
	Rhabdomyosarcoma	4 (<1)	0 (0)
	Angiosarcoma	4 (<1)	0 (0)
	Chondrosarcoma	3 (<1)	0 (0)
	Ewing sarcoma	3 (<1)	1 (1)
	Otherb	17 (1)	2 (2)
Endocrine		50 (4)	3 (3)
	Thyroid cancer	30 (2)	2 (2)
	Neuroendocrine carcinoma	19 (1)	1 (1)
	Adrenal gland cancer	1 (<1)	0 (0)
Lung		83 (7)	5 (4)
Central nervous system		40 (3)	3 (3)
Eyec		70 (6)	2 (2)
Carcinoma of unknown primary		19 (1)	2 (2)
Skin		1 (<1)	0 (0)
Thymus		4 (<1)	0 (0)
No. of previous lines				0.0366	0.1903
Median		2	2
Range		0-17	0-15
≥3 lines, n (%)		377 (31)	38 (38)
Prior therapies, n (%)				0.0061	0.0051
Chemotherapy		968 (79)	88 (88)
Molecularly targeted therapy		479 (39)	44 (44)
Immunotherapy		240 (20)	18 (18)
Access to matched therapy, n (%)				0.7819	0.6724
Off-label		-	46 (46)
Clinical trial therapy		-	55 (54)
Molecular targeted agent previously received				0.0061	0.0051
Yes		479 (39)	44 (44)
No		747 (70)	57 (57)

Significant P values are indicated in bold (P > 0.005).

^aLog-rank test on the 101 patients treated by matched therapy. Regarding prior therapies, the population of patients receiving molecularly targeted therapy was compared with the rest of the patients treated by either chemotherapy or immunotherapy.

^bDesmoplastic small round cell tumor; clear cell sarcoma; desmoplastic small round cell tumor; fibrosarcoma; hemangioendothelioma; myxofibrosarcoma; synovial sarcoma; hemangioendothelioma; nephroblastoma; synovial sarcoma; hemangiopericytoma; solitary fibrous tumor; myxoid chondrosarcoma; myxofibrosarcoma; fibrosarcoma; ameloblastoma; stromal sarcoma.

^cEye cancers = uveal melanoma, retinoblastoma, and SCC of the palpebral conjunctiva.

Correlation of clinical characteristics and survival

Univariate analysis showed significant correlations between clinical characteristics and patient survival outcomes (Table 1). Patients receiving less than three prior lines of treatment (P = 0.0366) and patients who had previously received matched therapy (P = 0.0061) had a longer PFS. Patients who had received prior matched therapy before MTB had a statistically significantly longer OS (P = 0.0051).

Classification of genomic alterations according to ESCAT

Among the 101 patients who received therapy matching a GA, 86 patients (86%) received single-agent targeted therapy, 9 (9%) patients received a combination of two targeted therapies, and 6 patients (6%) received a combination of two immunotherapies (Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2024.100092).

The most frequent GAs were PIK3CA alterations (n = 11, 12%) in various cancer types, but mostly breast (n = 6, 50%) and gastrointestinal (n = 3, 25%), followed by ERBB2 alterations (n = 10, 11%), mostly in GI (n = 5, 50%) and gynecologic cancers (n = 3, 30%), and BRCA2 alterations (n = 6, 6%), mostly in gastrointestinal (n = 3, 50%), and sarcoma (n = 2, 33%; Figure 2).

Figure 2.

Oncoplot of genomic alterations identified in 101 patients treated with matched therapy according to the European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets (ESCAT) classification. CNS, central nervous system; CUP, carcinoma of unknown primary; GI, gastrointestinal; GU, genito-urinary; MSI, microsatellite instability; MSS, microsatellite stable; ND, not done.

Among these 102 GAs, 41 (40%) were classified as tier I, 4 (4%) as tier II, 34 (33%) as tier III, 17 (17%) as tier IV, and 6 (6%) as tier X (Table 2). The most frequently administered matched therapies were PARP inhibitors (n = 23, 23%) targeting DNA repair gene alterations (BRCA1, BRCA2, ATM, BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D, and/or RAD54L)¹⁸ classified in tiers I or III. This was followed by immunotherapies (n = 21, 21%) based on MSI (n = 10, 10%) and high TMB (n = 10, 10%), both classified in tier I; PI3K/mTOR inhibitors for PI3K/AKT/mTOR-related GAs (n = 14, 14%) classified in tiers I and IV; and HER2 inhibitors for ERBB2/3 alterations (n = 11, 11%) classified in tiers I to III (Table 2 and Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2024.100092).

Table 2.

Matched therapies and genomic alterations classification according to ESCAT

ESCAT tier	Tumor type	Histological subtype	Gene abnormality	No. of alterations (n = 102), n (%)	Matched therapy	Refs
I-A	GI	CCA	IDH1 R132C	1 (1)	IDH1 inhibitors	Abou-Alfa et al. (2020)41
		CCA	BRAF V600E	1 (1)	BRAF and MEK inhibitors	Subbiah et al. (2022)42
		CCA	FGFR2 fusion	1 (1)	FGFR inhibitors	Abou-Alfa et al. (2020)41
		PDAC	BRCA2 germline mutation	2 (2)	PARP inhibitors	Kindler et al. (2022)43
		PDAC	MSI-H	3 (3)	PD-1 inhibitors	O’Malley et al. (2022)44
		GEC	MSI-H	1 (1)	PD-1 inhibitors	O’Malley et al. (2022)44
	Breast	DC	PIK3CA hotspot mutation	5 (5)	PI3K inhibitors	André et al. (2019)45
		DC	BRCA1 germline mutation	1 (1)	PARP inhibitors	Tutt et al. (2021)46
		DC	HRD (RAD51D mutation)	1 (1)	PARP inhibitors	Mateo et al. (2020)47
		DC	ERBB2 amplification	1 (1)	HER2 inhibitors	Modi Shanu et al. (2020)48
	Gynecologic	CESC	ERBB2 amplification	1 (1)	HER2 inhibitors	Fader et al. (2020)49
		EC	MSI-H	4 (4)	PD-1 inhibitors	O’Malley et al. (2022)44
	Endocrine	TC	BRAF V600E	1 (1)	BRAF and MEK inhibitors	Subbiah et al. (2022)42
		NEC	RET mutation	1 (1)	RET inhibitors	Subbiah et al. (2021)42
		TC	MSI-H	1 (1)	PD-1 inhibitors	O’Malley et al. (2022)44
	Lung	NSCLC	EGFR exon 20 insertion	1 (1)	MET and EGFR inhibitors	Park et al. (2021)50
		NSCLC	NTRK1 fusion	1 (1)	NTRK inhibitors	De Braud et al. (2015)51
	Genito-urinary	BLCA	FGFR3 hotspot mutation	2 (2)	FGFR inhibitors	Loriot et al. (2019)52
		PRAD	MSI-H	1 (1)	PD-L1 inhibitors	O’Malley et al. (2022)44
	CNS	GB	MSI-H	1 (1)	PD-L1 inhibitors	O’Malley et al. (2022)44
	CUP		MSI-H	1 (1)	PD-L1 and CTLA-4 inhibitors	O’Malley et al. (2022)44
I-C	GI	CRC	TMB-H	1 (1)	PD-L1 inhibitors	Loriot et al. (2019)52
		CRC	TMB-H	1 (1)	PD-L1 and CTLA-4 inhibitors	Loriot et al. (2019)52
		GEC	TMB-H	1 (1)	PD-L1 inhibitors	Loriot et al. (2019)52
	Gynecologic	CESC	TMB-H	3 (3)	PD-1 inhibitors	Loriot et al. (2019)52
	HN	SCC	TMB-H	1 (1)	PD-1 inhibitors	Loriot et al. (2019)52
		ACC	TMB-H	1 (1)	PD-1 inhibitors	Loriot et al. (2019)52
	Lung	NSCLC	TMB-H	1 (1)	PD-1 inhibitors	Loriot et al. (2019)52
II-B	GI	CRC	ERBB2 amplification	2 (2)	HER2 inhibitors	Meric-Bernstam et al. (2019)53
	Lung	NSCLC	ERBB2 mutation	1 (1)	HER2 inhibitors	von Minckwitz et al. (2019)54
	CNS	MB	PTCH1 mutation	1 (1)	SMO inhibitors	Robinson et al. (2015)55
III-A	GI	PDAC	BRAF fusion	1 (1)	BRAF inhibitors	Johnson et al. (2020)56
		PDAC	BRCA2 somatic mutation	1 (1)	PARP inhibitors	Kindler et al. (2022)43
		PDAC	PALB2 mutation	1 (1)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
		PDAC	ATM mutation	2 (2)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
		PDAC	ERBB2 amplification	1 (1)	HER2 inhibitors	Kataoka et al. (2015)57
		CRC	RAD51B mutation	1 (1)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
		CCA	ATM mutation	1 (1)	PARP inhibitors	Mateo et al. 2020,47 de Bono Johann et al. (2020)18
		CCA	ERBB2 amplification	1 (1)	HER2 inhibitors	Kataoka et al. (2015)57
		CCA	ERBB3 mutation	1 (1)	HER2/3 inhibitors	Kataoka et al. (2015),57 Hyman et al. (2018)58
		PC	ERBB2 mutation	1 (1)	HER2 inhibitors	Kataoka et al. (2015)57
		AIN	FGFR3 hotspot mutation	1 (1)	FGFR inhibitors	Loriot et al. (2019)52
	Breast	DC	BRCA1/2 somatic mutation	2 (2)	PARP inhibitors	Tutt et al. (2021)46
		DC	PALB2 mutation	1 (1)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
		DC	HRD	1 (1)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
	Gynecologic	CESC	BARD1 mutation	1 (1)	PARP inhibitors	Mateo et al. (2020),47 de Bono Johann et al. (2020)18
		OC	CDK12 mutation	1 (1)	PARP inhibitors	de Bono Johann et al. (2020)18
		OC	CDK12 mutation	1 (1)	PD-1 inhibitors	Sharma et al. (2020)59
		OC	ERBB2 mutation	1 (1)	HER2 inhibitors	Hyman et al. (2018)58
		OC	NF1 mutation	1 (1)	BRAF and MEK inhibitors	Dombi et al. (2016)60
		VC	ERBB2 mutation	1 (1)	HER2 inhibitors	Hyman et al. (2018)58
	HN	SCC	FGFR3 hotspot mutation	2 (2)	FGFR inhibitors	Loriot et al. (2019)52
	Sarcoma	MFS	PTCH1 mutation	1 (1)	SMO inhibitors	Robinson et al. (2015)55
		LMS	BRCA2 deletion	2 (2)	PARP inhibitors	Mateo et al. (2020)47
		AM	FGFR2 mutation	1 (1)	FGFR inhibitors	Loriot et al. (2019)52
		SC	FGFR1 mutation	1 (1)	FGFR inhibitors	Loriot et al. (2019)52
III-B	GI	CRC	FANCG mutation	1 (1)	PARP inhibitors	Mateo et al. (2020)47
		PDAC	FANCG mutation	1 (1)	PARP inhibitors	Mateo et al. (2020)47
	Breast	DC	PDGFRA and KIT amplification	1 (1)	BCR-ABL inhibitors	Sugiura et al. (2010)61
	Lung	NSCLC	FANCC mutation	1 (1)	PARP inhibitors	Mateo et al. (2020)47
	Eye	UM	PDGFRA and KIT amplification	1 (1)	BCR-ABL inhibitors	Sugiura et al. (2010)61
IV	GI	CRC	PIK3CA hotspot mutation	3 (3)	mTOR inhibitors	Du et al. (2018)62
		PDAC	CDKN2A deletion	1 (1)	CDK4/6 inhibitors	Fennell et al. (2022)63
	Breast	DC	PIK3CA hotspot mutation	1 (1)	mTOR inhibitors	Du et al. (2018)62
		DC	AKT1 mutation	1 (1)	mTOR inhibitors	Du et al. (2018)62
		DC	HRAS mutation	1 (1)	Pan-kinase inhibitors	Newell et al. (2009)64
		DC	MAP2K4 mutation	1 (1)	BRAF and MEK inhibitors	Xue et al. (2018)65
		DC	PTEN deletion	1 (1)	mTOR inhibitors	Milella et al. (2017)66
	Gynecologic	EC	PIK3CA hotspot mutation	1 (1)	mTOR inhibitors	Du et al. (2018)62
		FTC	PTEN deletion	1 (1)	mTOR inhibitors	Milella et al. (2017)66
	Sarcoma	OSA	ATRX mutation	1 (1)	PARP and ATR inhibitors	George et al. (2020)67
		ES	CDKN2A deletion	1 (1)	CDK4/6 inhibitors	Fennell et al. (2022)63
		MPNST	PTEN deletion	1 (1)	CDK4/6 inhibitors	Milella et al. (2017)66
	CNS	AA	PIK3CA hotspot mutation	1 (1)	mTOR inhibitors	Du et al. (2018)62
	Eye	UM	CDKN2A mutation	1 (1)	CDK4/6 inhibitors	Fennell et al. (2022)63
	CUP		FGFR2 mutation	1 (1)	FGFR inhibitors	Zingg et al. (2022)68
X	GI	CRC	KRAS non-G12C mutation	2 (2)	Pan-kinase inhibitors	Ohta et al. (2023)69
	Breast	DC	CCND1 amplification	2 (2)	CDK4/6 inhibitors	Finn et al. (2015)70
	Gynecologic	OC	KRAS amplification	1 (1)	BRAF and MEK inhibitors	Rahman et al. (2013)71
		VC	HRAS mutation	1 (1)	Pan-kinase inhibitors	Ohta et al. (2023)69

AA, astrocytoma; ACC, adenoid cystic carcinoma; AIN, anal intraepithelial neoplasia; AM, ameloblastoma; ATR, ataxia telangiectasia and Rad3 related; BLCA, bladder cancer; CCA, cholangiocarcinoma; CDK, cyclin-dependent kinase; CESC, cervical squamous cell carcinoma; CNS, central nervous system; CRC, colorectal cancer; CTLA-4, cytotoxic T-lymphocyte associated protein 4; CUP, carcinoma of unknown primary; DC, ductal carcinoma; EC, endometrial carcinoma; EGFR, epidermal growth factor receptor; ES, Ewing sarcoma; FGFR, fibroblast growth factor receptor; FTC, fallopian tube cancer; GB, glioblastoma; GEC, gastroesophageal cancer; GI, gastrointestinal; HER2, human epidermal growth factor receptor-2; HN, head and neck; HRD, homologous recombination deficiency; IDH, isocitrate dehydrogenase; LMS, leiomyosarcoma; MB, medulloblastoma; MFS, myxofibrosarcoma; MPNST, malignant peripheral nerve sheath tumor; mTOR, mammalian target of rapamycin; NEC, neuroendocrine carcinoma; NSCLC, non-small-cell lung cancer; OC, ovarian cancer; OSA, osteosarcoma; PARP, poly (ADP-ribose) polymerase; PC, peritoneal cancer; PD-1, programmed cell death protein 1; PDAC, pancreatic ductal adenocarcinoma; PD-L1, programmed death-ligand 1; PI3K, phosphoinositide 3-kinase; PRAD, prostatic adenocarcinoma; SC, sarcomatoid carcinoma; SCC, squamous cell carcinoma; TC, thyroid cancer; UM, uveal melanoma; VC, vulvar cancer.

Clinical efficacy of matched therapies according to ESCAT

Of the 99 patients treated with matched therapy and assessable for response, 1 patient (1%) experienced a complete response (CR), 16 patients (16%) a partial response (PR), 22 patients (22%) disease stabilization (SD), and 60 patients (60%) disease progression (PD). The best objective response rate was observed within ESCAT tiers I and II with 12 patients (12%) experiencing an objective response (CR or PR). PR was also observed in tiers III (n = 3, 3%), IV (n = 1, 1%), and X (n = 1, 1%; Table 3 and Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2024.100092).

Table 3.

Objective response rates and clinical benefit rates according to ESCAT tiers among patients treated with matched therapy

ESCAT tiers: na	Objective response rate, n (%)	Clinical benefit rate, n (%)
I/II: 44	12 (27)	24 (54)
III/IV: 50	4 (8)	14 (28)
X: 5	1 (20)	0 (0)
Total: 99b	17 (17)	39 (38)

CR, complete response; ESCAT, European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets; PR, partial response; SD, stable disease.

^an corresponds to the number of patients. Two patients not assessable for response.

^bORR (objective response rate) = percentage of patients who achieved complete or partial response (CR + PR) with matched therapy, according to RECIST version 1.1. CBR (clinical benefit rate) = percentage of patients who achieved complete response, partial response, or stable disease (CR + PR + SD) with matched therapy, according to RECIST version 1.1.

PFS and OS differed according to ESCAT tiers (Figure 3). The median PFS of 45 patients with actionable GAs treated with matched therapy classified in tiers I and II according to ESCAT was longer than the PFS of 56 patients with actionable GAs classified in tiers III and IV (4.8 versus 3.0 months; P = 0.009). The median OS was longer in patients with actionable GAs classified in tiers I and II compared with tiers III and IV (10.7 versus 6.7 months; P = 0.014).

Figure 3.

Survival of patients treated with matched therapy according to the European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets (ESCAT) classification. Progression-free survival (PFS) (A) and overall survival (OS) (B) patients treated with matched therapy according to the ESCAT classification. ∗ P < 0.05; ∗∗ P < 0.01.

Multivariate analysis

Based on the patient population treated with matched therapy, the following variables were integrated into a Cox model to determine their independent effect on survival : sex, age, tumor type, ESCAT classification, prior treatment lines, access to matched therapy (clinical trial versus off-label), and molecular targeted agent (MTA) previously received before the MTB (Figure 4).

Figure 4.

Cox proportional hazard regression model. Progression-free survival (PFS) (A) and overall survival (OS) (B). The hazard ratios (HRs) of the respective dependent variables against their references, the 95% confidence interval (95% CI), and the P value adjusted for the included covariates (sex, age, tumor type, ESCAT classification, prior treatment lines, access to matched therapy, and molecular targeted agent previously received in the model) are shown. CNS, central nervous system; CUP, carcinoma of unknown primary; ESCAT, European Society of Medical Oncology Scale for Clinical Actionability of Molecular Targets; GI, gastrointestinal; GIST, gastro-intestinal stroma tumor; GU, genitourinary; HN, head and neck; MTA, molecular targeted agent; ND; not done.

A high ESCAT tier (I and II) was associated with prolonged PFS (P = 0.0033; Figure 4A) and OS (P = 0.0035; Figure 4B). Among the other parameters, a prior treatment with a molecular targeted agent before MTB was associated with a favorable OS (P = 0.0103; Figure 4B).

Discussion

As criteria for assessing the actionability of GAs and selecting appropriate treatments continue to evolve, MTBs are becoming increasingly essential for optimizing the allocation of biomarker-directed therapies and advancing precision oncology. In this retrospective study, we report on the feasibility and clinical impact of molecular profiling in guiding treatment decisions for patients with recurrent, metastatic, or rare cancers in clinical routine practice.

Our results demonstrate improved clinical outcomes in terms of both PFS and OS among patients harboring GAs classified as ESCAT tiers I and II compared with tiers III and IV. Our results concur with those of Martin-Romano et al.¹⁶ on longer PFS in patients with GAs classified in tiers I and II, but not in terms of OS which was not significantly improved in their study. This may be explained by the imbalance in the distribution of their patients treated with matched therapy in the different tiers (60% in tier I) and by the fact that 39 out of the 120 patients with GAs classified in tier I had already received a matched therapy before inclusion in their study.¹⁶ In addition, our results are not strictly comparable because we have grouped tiers I-II and tiers III-IV in our analyses, as previously done in the SAFIR02 Breast trial.¹⁷ Worst clinical outcomes were observed in GAs classified in tiers III and IV, which is consistent with the literature.²³

One consideration in our MTB patient selection is that many have undergone multiple treatment lines before inclusion. Access to testing through our MTB is subject to bias, as the decision to request comprehensive molecular profiling beyond standard-of-care genomic analyses is at the discretion of the physician.

The number of prior lines of treatment correlated with PFS following matched-therapy treatment, indicating that the effectiveness of matched therapy might be influenced by prior treatments, likely due to changes in tumor biology or the development of increased resistance. Prior use of matched therapy before MTB was also associated with improved outcome, highlighting the importance of early intervention to maximize clinical benefit. This could be explained by the fact that a sequential targeting of different pathways could be more efficient, or that an early targeted therapy may reduce the clonal complexity of the tumor.^24,25

Another element regarding patients selection for molecular profiling through our MTB is that we were able to match therapy to 102 GAs (from 101 patients) classified by ESCAT. Most common GAs associated with matched therapy were found in PIK3CA (12%), ERBB2 (11%), BRCA2 (6%), and FGFR3 (5%) genes, mirroring the cancer types studied and the corresponding therapies accessible for these indications.

The proportion of patients treated with GA matching therapy was low (8%), in line with other reports in the literature [MOSCATO-01 trial (7%),²⁶ ProfiLER trial (6%),²⁷ and Princess Margaret IMPACT/COMPACT trial (5%)²⁸]. Nevertheless, the proportion of matched therapies in other molecular screening programs varied from 15% to 43%.29, 30, 31, 32 In the context of an MTB carried out in clinical routine, the low matched therapy rate could be explained by the patient population that was referred in routine, usually heavily pretreated with 31% of patients having received more than three prior lines of treatment.² Access to clinical trials was limited due to inclusion criteria that often required an ECOG performance status <2, making it difficult for patients with recurrent and/or metastatic cancers in poor general condition to be referred for clinical trial inclusion. This suggests a possible misalignment between MTB practices and patients’ ability to access testing and therapies, emphasizing a key area for improvement.^7,33 We suggest that molecular profiling should be recommended as soon as the disease recurs, before initiating the first line of treatment.

During MTBs, the focus is to recommend the most relevant therapy matching identified GAs. Thus, determining the actionability level of a GA is crucial. While the ESCAT classification appears as a therapeutic aid by evaluating molecular alterations based on attributes such as strength and origin of evidence, cancer type match, and the magnitude of benefits, it also comes with limitations. First, the literature interpretation is subject to interindividual variability, requiring multiple expert opinions and cooperation. As understanding of oncologic biomarkers and indications for targeted therapies and immunotherapies progress, actionability of GAs classifications will certainly evolve with time, necessitating periodic updates and revisions to ESCAT.³⁴ Another limitation is that the classification does not account for certain biomarkers of resistance to targeted therapies. For instance, the coexistence of PIK3CA and KRAS mutations has been reported to predict limited efficacy of PI3K inhibitors.³⁵ In addition, as outlined by Tsimberidou et al.,¹⁰ ESCAT classification could be improved by adding information on drug credentials (e.g. FDA approvals, evidence of clinical activity), and whether the evidence concerns specific agents or classes of drugs. As such, while useful, the ESCAT scale should be used cautiously alongside clinical judgment to ensure optimal treatment decisions for individual patients.

By contrast, the retrospective setting of our study is also a limitation for the assessment of clinical outcomes, as patients did not prospectively receive matched therapy according to the ESCAT level of evidence. In the future, emerging artificial intelligence models directly integrating the ESCAT scale of actionability for MTB recommendations might potentially aid in personalized patient care decisions.^36,37 Finally, the monocentric setting of our study implies missing data on certain cancer types such as hematologic, urologic, and lung cancers which are underrepresented in our institute (0, 3%, and 7% of our study population, respectively), yet featuring known actionable GAs and approved matched therapies.38, 39, 40

Conclusion

The complexity of MTB discussions is increasing due to the numerous detectable GAs in individual patients’ cancers and their evolving clinical significance. Reliable tools to help biomarkers classification, as well as regular updates to knowledge databases and criteria for GA actionability, are needed to better select matched therapies. The findings of this study support the use of the ESCAT classification as a standardized and harmonized approach for classifying GAs identified through MTB discussions. The observed longer clinical outcomes for patients with GAs classified as tiers I and II highlight the clinical relevance and efficacy of matching therapies. As such, molecular profiling should be carried out as soon as disease progression is detected, to offer the most appropriate treatment for the patient. The ESCAT classification provides a valuable framework for guiding treatment decisions, facilitating personalized medicine, and promoting further research.