Abstract
Background
We report the results of the pre-planned secondary analysis of radiologic responses (RRs) of ISG-STS 1001, a randomized trial comparing anthracycline + ifosfamide (AI) versus histology-tailored (HT) neoadjuvant chemotherapy for primary localized high-risk soft-tissue sarcomas of the extremities and trunk wall.
Patients and methods
Patients with undifferentiated pleomorphic sarcoma (UPS), leiomyosarcoma (LMS), malignant peripheral nerve sheath tumor, synovial sarcoma or myxoid liposarcoma (MLPS) were randomized, whereas patients with myxofibrosarcoma, pleomorphic liposarcoma, pleomorphic rhabdomyosarcoma or unclassified sarcoma were included in the observational arm (O) and treated with AI. Patients with UPS, LMS or MLPS needing concurrent preoperative radiotherapy were included in O. We evaluated associations between: disease-free survival (DFS)/overall survival (OS) and centrally reviewed RR, assessed with RECIST 1.1 and as percent dimensional variation (D; both dichotomized and continuous); DFS/OS and histology; RR and histology.
Results
Four hundred and thirty-five patients were included (287 randomized, 148 observed). The analysis of RRs comprised 236 patients (154 randomized, 82 observed) with measurable disease and available for central review. RECIST best responses were: 28 (11.9%) partial response (PR), 195 (82.6%) stable disease (SD), 13 (5.5%) progressive disease (PD). RECIST significantly correlated with DFS [PD versus PR: hazard ratio (HR) 8.18, 95% confidence interval (CI) 2.96-22.58; SD versus PR: HR 2.96, 95% CI 1.30-6.75] and OS (PD versus PR: HR 12.61, 95% CI 3.40-46.84; SD versus PR: HR 4.24, 95% CI 1.34-13.47). The median value of D was −1.6%. Patients with D >−1.6% had worse clinical outcomes than those with D <−1.6% (DFS: HR 1.73, 95% CI 1.19-2.50; OS: HR 1.86, 95% CI 1.21-2.86). D in continuous scale inversely correlated with DFS (HR 1.53, 95% CI 1.25-1.87) and OS (HR 1.78, 95% CI 1.41-2.25).
Conclusions
These results confirm the prognostic value of RRs as per RECIST and D and demonstrate that any variation in size predicts the proportional efficacy of treatment.
Key words: sarcoma, randomized clinical trial, neoadjuvant chemotherapy, prognosis, tumor response, RECIST
Highlights
-
•
RECIST 1.1 significantly correlated with both DFS and OS.
-
•
Patients with D >−1.6% had worse DFS/OS than those with D <−1.6%.
-
•
D in continuous scale inversely correlated with both DFS and OS.
-
•
No significant association was observed between histology and RR as per RECIST 1.1.
-
•
A weak association was demonstrated between histology and D in continuous scale.
Introduction
Over the past few years, the role of (neo)adjuvant chemotherapy for the treatment of primary localized high-risk soft-tissue sarcomas (STSs) of the extremities or trunk wall has been investigated by two main distinct randomized clinical trials (RCTs).1, 2, 3, 4
We have already reported that the administration of three cycles of full-dose standard chemotherapy [anthracycline + ifosfamide (AI)] in the neoadjuvant setting was not inferior to the administration of three neoadjuvant plus two additional adjuvant cycles of the same regimen.1,2,5
ISG-STS 1001, an RCT comparing AI versus histology-tailored (HT) neoadjuvant chemotherapy in primary localized high-risk STSs (grade 3; largest diameter >5 cm; deeply seated) showed superiority of AI across all histological types, except for high-grade myxoid liposarcoma (HG-MLPS), for which an equivalence between AI and HT (namely, trabectedin) was observed.3,4,6, 7, 8
While the main study was closed to randomization, the HG-MLPS cohort was thus reopened to validate this observation. In the expanded cohort of HG-MLPS, trabectedin proved indeed non-inferior to AI while at the same time being endowed with a more favorable toxicity profile.9
Response Evaluation Criteria in Solid Tumors (RECIST) are based on tumor size and employ discrete thresholds to categorize tumor response to treatment.10,11 In their original formulation, they were not conceived as prognostic tools, but were primarily designed for standardizing tumor response assessment in clinical trials.12
In previous work, we demonstrated that radiologic responses (RRs) to preoperative chemotherapy [with or without radiotherapy (RT)] were associated with better clinical outcomes and that Choi criteria were better prognosticators than RECIST.13
In an attempt to provide a comprehensive assessment of the prognostic value of objective responses in primary localized high-risk STSs of the extremities or trunk wall treated with neoadjuvant chemotherapy (with or without RT), we hereby report the final results of the pre-planned secondary analysis of RRs of ISG-STS 1001.
Patients and methods
Study design
ISG-STS 1001 is a phase III, open-label, randomized, controlled study comparing AI versus HT neoadjuvant chemotherapy for the treatment of primary localized high-risk STSs of the extremities and trunk wall, registered at ClinicalTrials.gov (identifier NCT01710176).
The trial protocol and all amendments were approved by the independent ethics committee at each trial center. The study was conducted in accordance with the provisions of the Declaration of Helsinki. All patients provided written informed consent before enrollment.
The full protocol is available in the Supplementary Appendix, available at https://doi.org/10.1016/j.esmoop.2025.104299.
Participants
Patients were eligible for the randomized study if ≥18 years of age; had a histologically confirmed and centrally reviewed diagnosis (before random assignment) of localized, resectable with curable intent STS originating in an extremity or trunk wall, belonging to HG-MLPS (cellular component >5%), leiomyosarcoma (LMS), synovial sarcoma, malignant peripheral nerve sheath tumor (MPNST) or undifferentiated pleomorphic sarcoma (UPS); grade 3 according to the Federation Nationale des Centres de Lutte Contra le Cancer (FNCLCC) grading system; and >5 cm in the longest diameter at baseline radiologic assessment.
Patients were eligible for the observational arm of the study if affected by myxofibrosarcoma, pleomorphic liposarcoma, pleomorphic rhabdomyosarcoma, unclassified sarcoma; grade 3; and >5 cm in the longest diameter at baseline radiologic assessment. Also patients affected by LMS, HG-MLPS and MPNST candidate on clinical grounds to preoperative concurrent RT were included in the observational study since HT chemotherapy was not combinable with RT.
Patients who were referred to participating institutions after an unplanned excision and who were candidate to re-excision were also included in the study according to the criteria above and could participate in the efficacy endpoint (DFS/OS) but not in the activity endpoint (response and outcome), since they did not have measurable disease at study entry.
Patients were ineligible if they had metastatic disease.
Other inclusion and exclusion criteria can be found in the Supplementary Appendix, available at https://doi.org/10.1016/j.esmoop.2025.104299.
Randomization and blinding
Patients were randomly assigned, in a 1 : 1 ratio, to receive three cycles of neoadjuvant AI or HT chemotherapy. Randomization procedures have already been reported elsewhere.3,4
Patients in the observational study were uniformly treated with three cycles of neoadjuvant AI followed by radical surgical resection and RT.
Procedures
Dosages and treatments have already been described elsewhere.3,4,9
Surgery was planned 3-4 weeks after the administration of the last preoperative cycle. RT was administered in the post-operative setting when indicated.
Follow-up was carried out every 4 months for the first 2 years after the end of treatment, then every 6 months from the third to the fifth year and annually after the sixth year.
Radiologic evaluation
Baseline imaging was acquired within 4 weeks before the first cycle of neoadjuvant chemotherapy.
RRs were assessed locally after the first and third cycles of neoadjuvant chemotherapy.
Central radiologic review was foreseen and the correlation between RRs and clinical outcomes was among the secondary endpoints of the study.
Scans were centrally reviewed at the Fondazione IRCCS Istituto Nazionale dei Tumori by radiologists with dedicated expertise in the diagnosis and management of STSs.
RRs were defined both as per RECIST 1.1 and as percent dimensional variation after the third cycle of neoadjuvant chemotherapy, calculated as follows: (longest diameter after the third cycle of neoadjuvant chemotherapy − longest diameter at baseline)/longest diameter at baseline × 100.
Statistical analysis
Associations between DFS/OS, RRs and tumor histology were evaluated by fitting both univariate and multivariate Cox regression models.14 The multivariate model included RECIST 1.1 [partial response (PR) as reference category], percent dimensional variation in continuous scale, tumor histology (MLPS as reference category), treatment (AI versus HT) and preoperative RT (not carried out versus carried out) as covariates. The multivariate model incorporated preoperative RT and treatment as adjusting covariates to account for the variables included in the subgroup analyses.
The relationship between percent dimensional variation and clinical outcomes was investigated by resorting to a regression model based on restricted cubic splines.
A Cox regression model including histology and RECIST 1.1 as well as the first-order interaction term was implemented. The hazard ratio (HR) and the corresponding 95% confidence interval (CI) for percent dimensional variation in continuous scale were reported as 1-standard deviation increase.
The Kaplan–Meier method was used to estimate the patterns of DFS and OS for RECIST, dichotomized percent dimensional variation and histology.15
Survival curves were compared using the log-rank test. Associations between categorical variables were assessed by means of chi-square or Fisher’s exact test whenever appropriate and associations between categorical variables and dimensional variation were assessed by means of the Wilcoxon test.
Associations between continuous variables were assessed with Spearman correlation coefficient.
Associations between RECIST 1.1 responses and patients’ baseline characteristics (preoperative RT, treatment, site, grade, age, size and histology) were evaluated by means of univariate logistic regression models.
Statistical analyses were conducted using SAS Studio (SAS Institute, Milan, Italy) and RStudio (The R Foundation for Statistical Computing, Vienna, Austria; https://www.R-project.org/), employing a nominal α level of 5%.
Results
Study design
This study recruited a total of 435 patients (Figure 1). For the overall series of patients who entered the trial with measurable disease (n = 366), 5-year probabilities of DFS and OS were 0.49 (95% CI 0.44-0.54) and 0.69 (95% CI 0.64-0.74), respectively.
Figure 1.
CONSORT diagram detailing study design and treatment arms.aRandomized patients: 65/93 in arm A and 63/64 in arm B (1 patient was randomized as UPS, but histological central review concluded for MFS, non-randomizable histology as per study protocol; the patient was excluded from our analyses). AI, anthracycline + ifosfamide; CONSORT, Consolidated Standards of Reporting Trials; HT, histology-tailored chemotherapy; MFS, myxofibrosarcoma; RT, radiotherapy; UPS, undifferentiated pleomorphic sarcoma.
Median follow-up was 88 months (interquartile range 73-101 months).
The analysis of RRs included all patients assessable with RECIST 1.1 and available for central radiologic review (n = 236); 5-year probabilities of DFS and OS were 0.53 (95% CI 0.47-0.59) and 0.69 (95% CI 0.63-0.74), respectively.
Patients’ demographics and clinical characteristics are reported in Table 1 and Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2025.104299.
Table 1.
Clinical and histopathological data of patients who entered the study with measurable disease and were available for central radiologic review (n = 236), overall and per treatment
| Patients’ characteristics | Treatment |
Total (n = 236) | |
|---|---|---|---|
| AI (n = 160) | HT (n = 76) | ||
| Age (years, median) | 52.5 | 50 | 52 |
| Min-max | 20-75 | 19-72 | 19-75 |
| IQR | 41-60 | 40.5-61 | 41-60.5 |
| Gender, n (%) | |||
| Male | 107 (67) | 46 (61) | 153 (65) |
| Female | 53 (33) | 30 (39) | 83 (35) |
| Size (mm, median) | 111.5 | 91 | 105.5 |
| Min-max | 10-353 | 38-218 | 10-353 |
| IQR | 84.5-149.5 | 74-136.6 | 81-147.5 |
| Site, n (%) | |||
| Upper limb | 15 (9) | 4 (5) | 19 (8) |
| Pelvic girdle | 16 (10) | 13 (17) | 29 (12) |
| Lower limb | 110 (69) | 54 (71) | 164 (70) |
| Other | 19 (12) | 5 (7) | 24 (10) |
| Microscopic surgical margins, n (%) | |||
| R0 | 132 (82) | 60 (79) | 192 (81) |
| R1 | 25 (16) | 12 (15) | 37 (16) |
| R2 | 1 (1) | 2 (3) | 3 (1) |
| Unknown | 2 (1) | 2 (3) | 4 (2) |
| Type of surgery, n (%) | |||
| Conservative | 149 (93) | 72 (95) | 221 (94) |
| Amputation | 9 (6) | 3 (4) | 12 (5) |
| No surgery | 2 (1) | 1 (1) | 3 (1) |
| Preoperative RT, n (%) | |||
| Not carried out | 93 (58) | 64 (84) | 157 (67) |
| Carried out | 67 (42) | 12 (16) | 79 (33) |
| Confirmed histology, n (%) | |||
| Myxoid-round cell liposarcoma | 37 (23) | 16 (21) | 53 (22) |
| Synovial sarcoma | 18 (11) | 18 (24) | 36 (15) |
| Malignant peripheral nerve sheath tumor | 10 (6) | 4 (5) | 14 (6) |
| Leiomyosarcoma | 13 (8) | 7 (9) | 20 (9) |
| Undifferentiated pleomorphic sarcoma | 43 (27) | 30 (40) | 73 (31) |
| Myxofibrosarcoma | 24 (15) | 1 (1) | 25 (11) |
| Unclassified spindle cell | 4 (3) | 0 (0) | 4 (2) |
| Pleomorphic liposarcoma | 8 (5) | 0 (0) | 8 (3) |
| Pleomorphic rhabdomyosarcoma | 3 (2) | 0 (0) | 3 (1) |
| Grade, n (%) | |||
| G2 | 10 (6) | 7 (9) | 17 (7) |
| G3 | 150 (97) | 69 (91) | 219 (93) |
| Dimensional variation after the third cycle of neoadjuvant chemotherapy (%, median) | −7 | −1 | −5 |
| Min-max | −64 to 46 | −56 to 44 | −64 to 46 |
| IQR | −22 to 0 | −10 to 5 | −19 to 1 |
| RECIST 1.1 centralized, n (%) | |||
| PD | 7 (4) | 6 (8) | 13 (5) |
| SD | 131 (82) | 64 (84) | 195 (83) |
| PR | 22 (14) | 6 (8) | 28 (12) |
AI, anthracycline + ifosfamide; HT, histology tailored; IQR, interquartile range; PD, progressive disease; PR, partial response; RT, radiotherapy; SD, stable disease.
Prognostic value of RECIST 1.1 responses after the third cycle of neoadjuvant chemotherapy
On the overall series of patients assessable with RECIST 1.1 and available for central radiologic review (n = 236), RECIST 1.1 best responses after the third cycle of neoadjuvant chemotherapy were: 28/236 (11.9%) PR, 195/236 (82.6%) stable disease (SD) and 13/236 (5.5%) progressive disease (PD) (Table 1 and Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2025.104299).
For patients treated with AI (n = 160), RECIST 1.1 best responses were: 22/160 (13.8%) PR, 131/160 (81.9%) SD and 7/160 (4.4%) PD.
For patients treated with HT (n = 76), RECIST 1.1 best responses were: 6/76 (7.9%) PR, 64/76 (84.2%) SD and 6/76 (7.9%) PD.
At univariate Cox regression analysis, RECIST 1.1 responses were significantly associated with both DFS (PD versus PR: HR 8.18, 95% CI 2.96-22.58, P < 0.0001; SD versus PR: HR 2.96, 95% CI 1.30-6.75, P = 0.01) and OS (PD versus PR: HR 12.61, 95% CI 3.40-46.84, P = 0.0002; SD versus PR: HR 4.24, 95% CI 1.34-13.47, P = 0.014).
The 5-year DFS and OS probabilities according to RECIST 1.1 are reported in Figure 2A and B.
Figure 2.
Disease-free and overall survival probabilities according to RECIST 1.1. (A) Disease-free survival and (B) overall survival probabilities of RECIST 1.1 (n = 236). (C) Disease-free survival and (D) overall survival probabilities of dichotomized percent dimensional variation after the third cycle of neoadjuvant chemotherapy (n = 236). (E) Disease-free survival and (F) overall survival probabilities of different histologies on the overall series of patients with measurable disease (n = 366). LMS, leiomyosarcoma; MFS, myxofibrosarcoma; MLPS, myxoid liposarcoma; MPNST, malignant peripheral nerve sheath tumor; SYNOV, synovial sarcoma; UPS, undifferentiated pleomorphic sarcoma.
A consistent yet not statistically significant association between RECIST 1.1 and DFS was observed also in the subsets of patients treated with neoadjuvant chemotherapy (either AI or HT) without RT (n = 157; PD + SD versus PR: HR 2.23, 95% CI 0.82-6.12, P = 0.118), in those who received AI without RT (n = 93; PD + SD versus PR: HR 2.44, 95% CI 0.76-7.88, P = 0.136) and in randomized patients who received AI without RT (n = 65; PD + SD versus PR: HR 1.49, 95% CI 0.45-4.92, P = 0.510).
RECIST 1.1 retained their prognostic value at multivariate analysis (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Prognostic value of percent dimensional variation
Firstly, we computed the median value of percent dimensional variation after the third cycle of neoadjuvant chemotherapy (D) in deceased patients (−1.6%) and demonstrated that patients with D >−1.6% had worse DFS/OS than those with D <−1.6% (n = 236; DFS: HR 1.73, 95% CI 1.19-2.50, P = 0.004; OS: HR 1.86, 95% CI 1.21-2.86, P = 0.005). The 5-year DFS and OS probabilities for dichotomized D are reported in Figure 2C and D.
Then, we demonstrated that D in continuous scale inversely correlated with both DFS (n = 236; HR 1.53, 95% CI 1.25-1.87, P < 0.0001) and OS (n = 236; HR 1.78, 95% CI 1.41-2.25, P < 0.0001) (Figure 3).
Figure 3.
Disease-free and overall survival probabilites according to dichotomized percent dimensional variation after the third cycle of neoadjuvant chemotherapy. (A) Disease-free survival and (B) overall survival probabilities of percent dimensional variation after the third cycle of neoadjuvant chemotherapy in continuous scale at median follow-up (n = 236).
These associations were also observed in patients treated with neoadjuvant chemotherapy (AI or HT) without RT (n = 157; DFS: HR 1.31, 95% CI 1.02-1.68, P = 0.033; OS: HR 1.47, 95% CI 1.10-1.96, P = 0.009) and in randomized patients who received HT (n = 63; DFS: HR 1.73, 95% CI 1.17-2.55, P = 0.006; OS: HR 1.60, 95% CI 1.04-2.46, P = 0.031).
Percent dimensional variation after the third cycle of neoadjuvant chemotherapy retained its prognostic value at multivariate analysis (Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Ancillary analyses exploring the impact of RT and the prognostic value of percent dimensional variation after the first cycle of neoadjuvant chemotherapy are reported in the Supplementary Results (including Supplementary Tables S14-S17 and Supplementary Figures S4 and S5), available at https://doi.org/10.1016/j.esmoop.2025.104299.
Prognostic value of tumor histology
The associations between tumor histology and DFS/OS were evaluated on the overall series of patients with measurable disease (n = 366).
Overall, MLPS was the histotype endowed with the most favorable outcome and MPNST the one with the worst OS (Figure 2E and F).
As reported in Supplementary Table S4, available at https://doi.org/10.1016/j.esmoop.2025.104299, histology other than MLPS represented an adverse prognostic factor, both in terms of DFS and OS.
Of note, we were able to replicate this association for DFS in patients who received preoperative chemotherapy alone (n = 254; Supplementary Table S5, available at https://doi.org/10.1016/j.esmoop.2025.104299) and in randomized patients treated with HT (n = 101; Supplementary Table S6, available at https://doi.org/10.1016/j.esmoop.2025.104299). A consistent yet nonsignificant association was observed in patients who received AI without RT (n = 152; Supplementary Table S7, available at https://doi.org/10.1016/j.esmoop.2025.104299) and in randomized patients treated with AI without RT (n = 103; Supplementary Table S8, available at https://doi.org/10.1016/j.esmoop.2025.104299).
The prognostic value of histology, both in terms of DFS and OS, was confirmed in the multivariate analysis (Supplementary Tables S2 and S3, available at https://doi.org/10.1016/j.esmoop.2025.104299).
No statistically significant interaction between tumor histology and RECIST 1.1 for both DFS (P = 0.85) and OS (P = 0.99) was observed.
Prognostic value of baseline patients’ characteristics other than tumor histology
At univariate Cox regression analysis, for the 236 patients who entered the trial with measurable disease and were available for central radiologic review, among baseline characteristics (preoperative RT, treatment, site, grade, age and size), only tumor size was significantly associated with clinical outcomes (Supplementary Table S9, available at https://doi.org/10.1016/j.esmoop.2025.104299).
However, as tumor size is partially redundant with RECIST 1.1 and percent dimensional variation, it was excluded from the multivariate models to avoid collinearity.
Association analyses
No statistically significant association was observed between tumor histology and RECIST 1.1 responses after the third cycle of neoadjuvant chemotherapy (Supplementary Table S10, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Percent dimensional variation in continuous scale after the third cycle of neoadjuvant chemotherapy was significantly associated with tumor histology (P = 0.037). Pairwise comparisons between different histologies highlighted a borderline statistically significant difference in the distribution of dimensional variation only between MLPS and UPS (P value adjusted for multiple comparisons = 0.046; Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Due to this association, we fitted a multivariate model also without histology to avoid collinearity, and demonstrated overlapping results when compared with the model which considered both dimensional variation and tumor histology (Supplementary Tables S11 and S12, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Associations between RECIST 1.1 responses after the third cycle of neoadjuvant chemotherapy and patients’ baseline characteristics (preoperative RT, treatment, site, grade, age, size and histology) were evaluated by means of univariate logistic regression models, demonstrating a significant association for tumor size alone (Supplementary Table S13, available at https://doi.org/10.1016/j.esmoop.2025.104299).
Percent dimensional variation and Sarculator
Sarculator nomograms confirmed their prognostic value in patients assessable with RECIST 1.1 and available for central review (n = 236; lower- versus higher-risk DFS: HR 0.50, 95% CI 0.34-0.73; OS: HR 0.46, 95% CI 0.29-0.72).
A statistically significant association was observed between RECIST 1.1 responses after the third cycle of neoadjuvant chemotherapy and Sarculator scores, both at 5 and 10 years (Wilcoxon test P = 0.006), with patients exhibiting PR as best responses having higher Sarculator scores than patients with SD or PD (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2025.104299).
This association was replicated when stratifying patients in low (predicted OS >60%) versus high (predicted OS <60%) risk as defined by Sarculator scores at 10 years (P = 0.035).
A correlation, although not as equally strong, was observed also between percent dimensional variation after the third cycle of neoadjuvant chemotherapy and Sarculator scores (Spearman correlation coefficient −0.19, 95% CI −0.31 to --0.16).
Sarculator scores were calculated at each participating center utilizing radiologic measurements before central review.
Discussion
We hereby present the definitive results of the pre-planned secondary analysis of RRs of ISG-STS 1001.
Consistent with previous work, we confirmed that RECIST 1.1 correlate with both DFS and OS.13
Of note, we report that clinical outcomes of patients with PR were significantly better than those of patients with SD, both in terms of DFS and OS.
These findings partially diverge from data we have already published elsewhere, when we described overlapping outcomes between patients with PR and SD as best responses.13
These discrepancies may be—to a certain extent—imputable to the different follow-up times of the two studies.
The first attempts to provide a formal, quantitative definition of objective responses date back to 1976, when 16 observers were challenged to measure spheres of different sizes randomly allocated underneath a mattress. Of note, ‘the purpose was not to predict long-term efficacy but to determine what change in bidimensional size could be confidently considered a change’ and ‘it should not necessarily be regarded as influencing the management of the patient’.16
RECIST were first conceptualized in the year 2000 by a joint task force composed by the European, American and Canadian cancer research organizations after analyzing data from over 4600 patients enrolled in 14 clinical trials.
A major update was implemented in 2009 (version 1.1), when new rules were introduced regarding the evaluation of lymph nodes and the number of target lesions.10,11
Objective responses were not intended, at least in their original formulation, to serve as prognostic tools, but were primarily designed for standardizing response detection and description in clinical trials, with no reference to patients’ outcomes.12
To challenge the long-standing and somewhat arbitrary thresholds introduced with RECIST, we employed percent dimensional variation after neoadjuvant chemotherapy, both dichotomized and in continuous scale, as covariate of our survival analyses.
Firstly, we computed the median value of dimensional variation after the third cycle of neoadjuvant chemotherapy (D) in deceased patients (−1.6%), and demonstrated that patients with D >−1.6% had an increased risk of disease progression or recurrence and lower OS than those with D <−1.6%.
These results imply that hardly any variation from baseline (−1.6% may be considered a rough approximation of 0%) affected clinical outcomes, with almost any dimensional reduction carrying a positive impact and almost any increase in size a corresponding detrimental effect.
One may argue that inter-timepoint and/or interobserver variability may hamper the introduction of this new cut-off in clinical practice.
To address this hypothetical objection, we proved that D in continuous scale inversely correlated with both DFS and OS. This implies that any variation from baseline affected clinical outcomes in a way that was proportional to its magnitude. In more simplistic words, the greater the dimensional reduction, the stronger the positive impact on clinical outcomes, and the greater the increase in size, the stronger the corresponding detrimental effect.
Strikingly, survival analyses employing percent dimensional variation in continuous scale highlighted that patients with SD as RECIST 1.1 best response displayed a significant degree of variability in terms of clinical outcomes, with DFS and OS probabilities spanning from 30% to 70%, approximately.
These unexpected and intriguing findings clearly demonstrate that RECIST coarsely fail to capture clinical heterogeneity and retain unsatisfactory prognostic value in patients with subthreshold (in either direction) dimensional responses.
Ancillary survival analyses employing histology as covariate confirmed its prognostic impact on DFS and OS, consistent with previous publications.13
MLPS was the histotype endowed with the most favorable outcome—with tumor histology other than MLPS representing an independent adverse prognostic factor, both in terms of DFS and OS—and MPNST the one with the worst OS.
To rule out confounding factors, we demonstrated that no statistically significant association existed between histology and RECIST 1.1 responses.
Due to the existence of a weak association between histology and percent dimensional variation, we fitted the multivariate model also without histology to avoid collinearity, and demonstrated overlapping results when compared with the model which considered both dimensional variation and tumor histology.
This way, we proved that patients who achieved more profound dimensional responses had better clinical outcomes irrespective of tumor histology, and could therefore conclude that dimensional variation was indeed an independent prognostic factor.
This study has some limitations. To begin with, images had to be centrally reviewed at the coordinating center as per study protocol. While this maximized robustness and enhanced the quality and consistency of radiologic evaluations, it inevitably led to the exclusion of those patients whose images were not available for centralized archiving.
Patients who received concurrent preoperative RT were also included in this work.
We reckon that this may have injected unwanted technical variability and tackled this potential issue replicating our results in the subsets of patients treated with preoperative chemotherapy alone, demonstrating that the association between RRs and clinical outcomes was not significantly influenced by the administration of RT, and that RT per se did not have a significant impact on DFS/OS.
Another peculiar issue of sarcoma research is the extensive heterogeneity of these diseases and their often unpredictable clinical behavior.17, 18, 19, 20, 21, 22, 23, 24
Ideally, clinical trials focusing on sarcomas should thus be conducted on individual histotypes.
At the time of writing, this is not feasible due to the rarity of these tumors, which would make the accrual process painstakingly long if not frankly unmanageable in most scenarios.
The observed number of objective responses as per RECIST 1.1 was relatively low, with only 28/236 (11.9%) patients achieving PR after the third cycle of neoadjuvant chemotherapy.
This suggests that three cycles of neoadjuvant chemotherapy, which we have already proved non-inferior to five cycles in terms of clinical benefit, may not be enough to achieve maximal RECIST responses.1,2
This discrepancy further emphasizes the prognostic relevance of percent dimensional variation in continuous scale, as it might allow us to identify those patients who have not been classified as responses as per RECIST 1.1, as they approach yet not reach the corresponding threshold, but will nonetheless achieve systemic disease control and clinical benefit.
One may object that non-dimensional responses may suffice to tip the balance in patients who sit at the interface between distinct RECIST categories. We have already reported elsewhere that Choi criteria applied to magnetic resonance imaging (MRI) were better prognosticators of OS and freedom from progression than RECIST in patients with high-risk STSs of the extremities or trunk wall treated with neoadjuvant chemotherapy without concurrent RT.13
However, we reckon that the translation of Choi criteria to MRI, among the secondary endpoints of this study, needs some fine-tuning, as it poses several technical challenges yet to be fully understood.13,25,26 These and other considerations will be distilled into forthcoming work from our group.
Sarculator nomograms represent a reliable prognostic tool for the stratification of patients affected by STSs of the extremities and retroperitoneum.
However, computing Sarculator scores requires bits of information that become available only after a surgical specimen has been obtained, whereas RRs after preoperative chemotherapy may inform clinical management already in the neoadjuvant setting.
The amount of clinically relevant information that can be inferred from variables as simple as dimensional responses is mesmerizing. However, in the era of so-called ‘personalized’ and data-driven medicine, more advanced tools are needed. Among others, radiomics—leveraging the gargantuan body of data embedded in bioimages—represents the ideal candidate to revolutionize cancer research, especially via the integration with other ‘-omics’ sciences.27, 28, 29, 30, 31
In the near future, we might be able to combine ‘naked-eye’ evaluations routinely carried out by radiologists with quantitative data extracted from bioimages in research and clinical practice.
Our results confirm the prognostic relevance of objective responses and reaffirm their fundamental role in clinical trials and cancer care.
However, well aware of the criticalities of the rigid and—to some extent—arbitrary model introduced with RECIST, we went further and proceeded to demonstrate that percent dimensional variation in continuous scale represents a reliable surrogate endpoint for DFS and OS in this clinical scenario.
We are persuaded that such intriguing results should spark a debate in the scientific community and initiate a critical revision of response criteria in solid tumors and their future role in cancer research.
Acknowledgements
We would like to thank the following investigators who participated in the study: ISG Investigators: Carla Dani (Department of Epidemiology, IRCCS Azienda Ospedaliera Universitaria San Martino, IST Istituto Nazionale per la Ricerca sul Cancro, Genova, Italy); Chiara Villa, Antonella Messina and Lorella Rusi (Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy); Andrea Marrari (Istituto Clinico Humanitas, Rozzano, Italy); Anna Maria Nuzzo (Fondazione del Piemonte per l’Oncologia IRCCS Candiolo, Turin, Italy); Carmen Nuzzo (Istituto Regina Elena, Roma, Italy); Antonino De Paoli e Angela Buonadonna (Centro di Riferimento Oncologico, Aviano, Italy); Alessandro Comandone e Antonella Boglione (Ospedale Gradenigo, Torino, Italy); Lorenzo Livi e Daniela Greto (Ospedale Careggi ed Università di Firenze, Firenze, Italy); Nada Riva e Manuela Monti (Istituto Scientifico Romagnolo per lo Studio e la Cura dei Tumori, Meldola, Italy); Elisabetta Pennacchioli e Tommaso De Pas (Istituto Europeo di Oncologia, Milan, Italy); Vincenzo Ippolito (Ospedali Civili, Brescia, Italy). GEIS Investigators: Patricio Ledesma (Grupo Español de Investigación en Sarcomas); Andres Redondo (Hospital Universitario La Paz, Madrid, Spain); Claudia Valverde (Hospital Universitari Vall d’Hebron, Barcelona, Spain); Raquel Bratos (MD Anderson Cancer Center, Madrid, Spain); Josefina Cruz (Hospital Universitario de Canarias, Tenerife, Spain); Javier Martinez Trufero (Hospital Universitario Miguel Servet, Zaragoza, Spain); Ricardo Cubedo (Hospital Universitario Puerta de Hierro, Madrid, Spain); Isabel Sevilla (Hospital Universitario Virgen de la Victoria, Málaga, Spain); Pablo Luna (Hospital Universitari Son Espases, Palma de Mallorca, Spain); Rafael Lopez (Hospital Clínico Universitario de Santiago, Santiago de Compostela, Spain); Pilar Sancho (Hospital Universitario Virgen del Rocío, Sevilla, Spain). FSG Investigators: Olivia Bally, Mehdi Brahmi, Isabelle Ray-Coquard, Philippe Cassier, Nathalie Marques, Luis Tassy (Centre Léon Bérard, Lyon, France); Pascaline Boudou-Rouquette, Camille Tlemsani, Jerome Alexandre, Francois Goldwasser (Hopital Cochin, Paris, France); Emmanuelle Bompas, Frederic Rolland (ICO Rene Gauducheau, Saint Herblain, France); Christophe Perrin, Marie Talarmin (Centre Eugene Marquis, Rennes, France); Antoine Italiano, Maud Toulmonde (Institut Bergonie, Bordeaux, France); Mathieu Laramas (Hopital Universitarie, Grenoble, France); Jacques-Olivier Bay, Pascale Dubray-Longeras (Centre Jean Perrin, Clermont Ferrand, France). Polish Investigators: Piotr Rutkowski (Centrum Onkologii Instytut im Marii Sklodowskiej-Curie, Warszawa, Poland).
Funding
Pharmamar® provided trabectedin for the HG-MLPS cohort. The study was partially funded through a European Union grant [grant number EUROSARC FP7 278472; AG]. The French sites were supported by awards to J-YB from NETSARC, LYRICAN [grant number INCA-DGOS-INSERM 12563] and DEPGYN [grant number RHU4]. This work was partially funded by the Italian Ministry of Health (“Ricerca corrente” funds).
Disclosure
AG: honoraria: Novartis, Pfizer, Lilly, PharmaMar, Deciphera; consulting or advisory role: Novartis, Bayer, Lilly, Pfizer, Nanobiotix, PharmaMar, SpringWorks Therapeutics, Boehringer Ingelheim; research funding: PharmaMar [my institution (Inst)], Nanobiotix (Inst); travel, accommodations, expenses: PharmaMar, Nanobiotix. EP: consulting or advisory role: Daiichi Sankyo, Deciphera, SynOX. JMB: consulting or advisory role: PharmaMar, GlaxoSmithKline, Novartis, Amgen, Bayer, Roche, Lilly, Tecnofarma, Asofarma, Boehringer Ingelheim; speakers’ bureau: PharmaMar; research funding: GlaxoSmithKline (Inst), PharmaMar, Novartis (Inst), Eisai (Inst), Lilly (Inst), Bayer (Inst), Eisai, Novartis, Lixte Biotechnology (Inst), Karyopharm Therapeutics (Inst), Deciphera (Inst), Blueprint Medicines (Inst), Nektar (Inst), FORMA Therapeutics (Inst), Amgen (Inst), Daiichi Sankyo (Inst), Immix BioPharma (Inst), Bristol Myers Squibb (Inst), Pfizer (Inst), Celgene (Inst), PharmaMar (Inst), Arog (Inst), Adaptimmune, Rain Therapeutics (Inst), InhibRx (Inst), Ayala Pharmaceuticals (Inst), Philogen (Inst), CEBIOTEX (Inst), PTC Therapeutics (Inst), SpringWorks Therapeutics (Inst); expert testimony: PharmaMar, Novartis; travel, accommodations, expenses: PharmaMar, Pfizer. GG: consulting or advisory role: Deciphera, GlaxoSmithKline, Eisai; travel, accommodations, expenses: PharmaMar. AB: consulting or advisory role: GlaxoSmithKline, PharmaMar; travel, accommodations, expenses: PharmaMar. J-YB: leadership: Innate Pharma; honoraria: Roche, AstraZeneca, PharmaMar, MSD, BMS, Bayer, Ignyta, Deciphera; consulting or advisory role: Roche, PharmaMar, Deciphera, Blueprint Medicines, Bayer, Karyopharm Therapeutics; research funding: GlaxoSmithKline (Inst), PharmaMar (Inst), Novartis (Inst), Bayer (Inst), Roche (Inst), BMS (Inst), MSD (Inst), Deciphera (Inst), AstraZeneca (Inst), OSE Pharma (Inst); travel, accommodations, expenses: Roche. VF: consulting or advisory role: Bristol Myers Squibb, Novartis, GlaxoSmithKline; speakers’ bureau: Bristol Myers Squibb, Novartis, Pierre Fabre, MSD Oncology, PharmaMar; travel, accommodations, expenses: Pierre Fabre. IL: honoraria: Amgen, Roche, BMS, Janssen, AstraZeneca, Agenus, Macrogenics, Celon Pharma, MSD, Menarini, Pfizer, Sanofi, BeiGene, Jacobio, Loxo, Rhizen Pharmaceuticals, Takeda, Cullinan Oncology; consulting or advisory role: Boehringer Ingelheim; research funding: Roche (Inst), Agenus (Inst); travel, accommodations, expenses: BMS. DMD: honoraria: Waldemar Link GmbH & Co KG; travel, accommodations, expenses: Waldemar Link GmbH & Co KG. AB: honoraria: Gentili; consulting or advisory role: Gentili; travel, accommodations, expenses: Gentili. SP: research funding: Ikena Oncology (Inst), Astex Pharmaceuticals (Inst). SS: honoraria: PharmaMar, Aadi, Gentili, Boehringer Ingelheim; consulting or advisory role: Bayer, PharmaMar, Daiichi Sankyo, Bavarian Nordic, Novartis, Astex Pharmaceuticals, Ikena Oncology, Rain Therapeutics, Regeneron, Boehringer Ingelheim, Agenus, Servier, PharmaEssentia, NEC OncoImmunity AS; research funding: Lilly (Inst), Bayer (Inst), Pfizer (Inst), Novartis (Inst), Daiichi Sankyo (Inst), Epizyme (Inst), PharmaMar (Inst), Amgen (Inst), Advenchen Laboratories (Inst), Karyopharm Therapeutics (Inst), Blueprint Medicines (Inst), SpringWorks Therapeutics (Inst), GlaxoSmithKline (Inst), Hutchison MediPharma (Inst), Deciphera (Inst), Boehringer Ingelheim (Inst), Rain Therapeutics (Inst), InhibRx (Inst); travel, accommodations, expenses: PharmaMar. APDT: consulting or advisory role: Bayer, Roche, Boehringer Ingelheim. PGC: research funding: Amgen/Dompé (Inst), Bayer (Inst), GlaxoSmithKline (Inst), Novartis (Inst), Pfizer (Inst), PharmaMar (Inst), Eisai (Inst), Lilly (Inst), Advenchen Laboratories (Inst), Arog (Inst), Blueprint Medicines (Inst), Deciphera (Inst), Epizyme (Inst), Karyopharm Therapeutics (Inst), Daiichi Sankyo (Inst), Boehringer Ingelheim (Inst), Rain Therapeutics (Inst), Fogorn Ther Inc (Inst), Hutchinson MediPharam Lt (Inst), InhibRx (Inst), PTC Ther (Inst), SpringWorks Ther (Inst). ALP: honoraria/grants: Pharmamar. All other authors have declared no conflicts of interest.
Contributor Information
A. Vanzulli, Email: andrea.vanzulli@istitutotumori.mi.it, andrea.vanzulli@unimi.it.
A. Gronchi, Email: alessandro.gronchi@istitutotumori.mi.it.
Supplementary data
References
- 1.Gronchi A., Frustaci S., Mercuri M., et al. Short, full-dose adjuvant chemotherapy in high-risk adult soft tissue sarcomas: a randomized clinical trial from the Italian Sarcoma Group and the Spanish Sarcoma Group. J Clin Oncol. 2012;30(8):850–856. doi: 10.1200/JCO.2011.37.7218. [DOI] [PubMed] [Google Scholar]
- 2.Gronchi A., Stacchiotti S., Verderio P., et al. Short, full-dose adjuvant chemotherapy (CT) in high-risk adult soft tissue sarcomas (STS): long-term follow-up of a randomized clinical trial from the Italian Sarcoma Group and the Spanish Sarcoma Group. Ann Oncol. 2016;27(12):2283–2288. doi: 10.1093/annonc/mdw430. [DOI] [PubMed] [Google Scholar]
- 3.Gronchi A., Ferrari S., Quagliuolo V., et al. Histotype-tailored neoadjuvant chemotherapy versus standard chemotherapy in patients with high-risk soft-tissue sarcomas (ISG-STS 1001): an international, open-label, randomised, controlled, phase 3, multicentre trial. Lancet Oncol. 2017;18(6):812–822. doi: 10.1016/S1470-2045(17)30334-0. [DOI] [PubMed] [Google Scholar]
- 4.Gronchi A., Palmerini E., Quagliuolo V., et al. Neoadjuvant chemotherapy in high-risk soft tissue sarcomas: final results of a randomized trial from Italian (ISG), Spanish (GEIS), French (FSG), and Polish (PSG) sarcoma groups. J Clin Oncol. 2020;38(19):2178–2186. doi: 10.1200/JCO.19.03289. [DOI] [PubMed] [Google Scholar]
- 5.Palassini E., Ferrari S., Verderio P., et al. Feasibility of preoperative chemotherapy with or without radiation therapy in localized soft tissue sarcomas of limbs and superficial trunk in the Italian sarcoma group/grupo español de investigación en sarcomas randomized clinical trial: three versus five cycles of full-dose epirubicin plus ifosfamide. J Clin Oncol. 2015;33(31):3628–3634. doi: 10.1200/JCO.2015.62.9394. [DOI] [PubMed] [Google Scholar]
- 6.Gronchi A., Bui B.N., Bonvalot S., et al. Phase II clinical trial of neoadjuvant trabectedin in patients with advanced localized myxoid liposarcoma. Ann Oncol. 2012;23(3):771–776. doi: 10.1093/annonc/mdr265. [DOI] [PubMed] [Google Scholar]
- 7.Gronchi A., Hindi N., Cruz J., et al. Trabectedin and RAdiotherapy in Soft Tissue Sarcoma (TRASTS): results of a phase I study in myxoid liposarcoma from Spanish (GEIS), Italian (ISG), French (FSG) sarcoma groups. EClinicalMedicine. 2019;9:35–43. doi: 10.1016/j.eclinm.2019.03.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sanfilippo R., Hindi N., Cruz Jurado J., et al. Effectiveness and safety of trabectedin and radiotherapy for patients with myxoid liposarcoma: a nonrandomized clinical trial. JAMA Oncol. 2023;9(5):656–663. doi: 10.1001/jamaoncol.2023.0056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gronchi A., Palmerini E., Quagliuolo V., et al. Neoadjuvant chemotherapy in high-grade myxoid liposarcoma: results of the expanded cohort of a randomized trial from Italian (ISG), Spanish (GEIS), French (FSG), and Polish sarcoma groups (PSG) J Clin Oncol. 2024;42:898–906. doi: 10.1200/JCO.23.00908. [DOI] [PubMed] [Google Scholar]
- 10.Eisenhauer E.A., Therasse P., Bogaerts J., et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1) Eur J Cancer. 2009;45(2):228–247. doi: 10.1016/j.ejca.2008.10.026. [DOI] [PubMed] [Google Scholar]
- 11.Fournier L., de Geus-Oei L.F., Regge D., et al. Twenty years on: RECIST as a biomarker of response in solid tumours an EORTC imaging group – ESOI joint paper. Front Oncol. 2022;11 doi: 10.3389/fonc.2021.800547. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Booth C.M., Eisenhauer E.A. Progression-free survival: meaningful or simply measurable? J Clin Oncol. 2012;30(10):1030–1033. doi: 10.1200/JCO.2011.38.7571. [DOI] [PubMed] [Google Scholar]
- 13.Stacchiotti S., Verderio P., Messina A., et al. Tumor response assessment by modified Choi criteria in localized high-risk soft tissue sarcoma treated with chemotherapy. Cancer. 2012;118(23):5857–5866. doi: 10.1002/cncr.27624. [DOI] [PubMed] [Google Scholar]
- 14.Cox D.R. Regression models and life-tables. J R Stat Soc. 1972;34(2):187–220. [Google Scholar]
- 15.Kaplan E.L., Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53(282):457–481. [Google Scholar]
- 16.Moertel C.G., Hanley J.A. The effect of measuring error on the results of therapeutic trials in advanced cancer. Cancer. 1976;38(1):388–394. doi: 10.1002/1097-0142(197607)38:1<388::aid-cncr2820380156>3.0.co;2-a. [DOI] [PubMed] [Google Scholar]
- 17.IARC . International Agency for Research on Cancer (IARC); Lyon, France: 2020. WHO Classification of Soft Tissue and Bone Tumours. [Google Scholar]
- 18.Gamboa A.C., Gronchi A., Cardona K. Soft-tissue sarcoma in adults: an update on the current state of histiotype-specific management in an era of personalized medicine. CA Cancer J Clin. 2020;70(3):200–229. doi: 10.3322/caac.21605. [DOI] [PubMed] [Google Scholar]
- 19.Callegaro D., Miceli R., Bonvalot S., et al. Impact of perioperative chemotherapy and radiotherapy in patients with primary extremity soft tissue sarcoma: retrospective analysis across major histological subtypes and major reference centres. Eur J Cancer. 2018;105:19–27. doi: 10.1016/j.ejca.2018.09.028. [DOI] [PubMed] [Google Scholar]
- 20.Callegaro D., Miceli R., Bonvalot S., et al. Development and external validation of two nomograms to predict overall survival and occurrence of distant metastases in adults after surgical resection of localised soft-tissue sarcomas of the extremities: a retrospective analysis. Lancet Oncol. 2016;17(5):671–680. doi: 10.1016/S1470-2045(16)00010-3. [DOI] [PubMed] [Google Scholar]
- 21.Pasquali S., Colombo C., Pizzamiglio S., et al. High-risk soft tissue sarcomas treated with perioperative chemotherapy: improving prognostic classification in a randomised clinical trial. Eur J Cancer. 2018;93:28–36. doi: 10.1016/j.ejca.2018.01.071. [DOI] [PubMed] [Google Scholar]
- 22.Pasquali S., Pizzamiglio S., Touati N., et al. The impact of chemotherapy on survival of patients with extremity and trunk wall soft tissue sarcoma: revisiting the results of the EORTC-STBSG 62931 randomised trial. Eur J Cancer. 2019;109:51–60. doi: 10.1016/j.ejca.2018.12.009. [DOI] [PubMed] [Google Scholar]
- 23.Pasquali S., Palmerini E., Quagliuolo V., et al. Neoadjuvant chemotherapy in high-risk soft tissue sarcomas: a Sarculator-based risk stratification analysis of the ISG-STS 1001 randomized trial. Cancer. 2022;128(1):85–93. doi: 10.1002/cncr.33895. [DOI] [PubMed] [Google Scholar]
- 24.Gronchi A., Miah A.B., Dei Tos A.P., et al. Soft tissue and visceral sarcomas: ESMO–EURACAN–GENTURIS Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2021;32(11):1348–1365. doi: 10.1016/j.annonc.2021.07.006. [DOI] [PubMed] [Google Scholar]
- 25.Choi H., Charnsangavej C., Faria S.C., et al. Correlation of computed tomography and positron emission tomography in patients with metastatic gastrointestinal stromal tumor treated at a single institution with imatinib mesylate: proposal of new computed tomography response criteria. J Clin Oncol. 2007;25(13):1753–1759. doi: 10.1200/JCO.2006.07.3049. [DOI] [PubMed] [Google Scholar]
- 26.Stacchiotti S., Collini P., Messina A., et al. High-grade soft-tissue sarcomas: tumor response assessment - pilot study to assess the correlation between radiologic and pathologic response by using RECIST and Choi criteria. Radiology. 2009;251(2):447–456. doi: 10.1148/radiol.2512081403. [DOI] [PubMed] [Google Scholar]
- 27.Shur J.D., Doran S.J., Kumar S., et al. Radiomics in oncology: a practical guide. Radiographics. 2021;41(6):1717–1732. doi: 10.1148/rg.2021210037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.van Timmeren J.E., Cester D., Tanadini-Lang S., Alkadhi H., Baessler B. Radiomics in medical imaging—“how-to” guide and critical reflection. Insights Imaging. 2020;11(1):91. doi: 10.1186/s13244-020-00887-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Subramanian A., Nemat-Gorgani N., Ellis-Caleo T.J., et al. Sarcoma microenvironment cell states and ecosystems are associated with prognosis and predict response to immunotherapy. Nat Cancer. 2024;5:642–658. doi: 10.1038/s43018-024-00743-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Arthur A., Orton M.R., Emsley R., et al. A CT-based radiomics classification model for the prediction of histological type and tumour grade in retroperitoneal sarcoma (RADSARC-R): a retrospective multicohort analysis. Lancet Oncol. 2023;24(11):1277–1286. doi: 10.1016/S1470-2045(23)00462-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Pasquali S., Iadecola S., Vanzulli A., et al. Radiomic features of primary retroperitoneal sarcomas: a prognostic study. Eur J Cancer. 2024;213 doi: 10.1016/j.ejca.2024.115120. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.