Interpretable artificial intelligence to optimise use of imatinib after resection in patients with localised gastrointestinal stromal tumours: an observational cohort study.

BACKGROUND: Current guidelines recommend use of adjuvant imatinib therapy for many patients with gastrointestinal stromal tumours (GISTs); however, its optimal treatment duration is unknown and some patient groups do not benefit from the therapy. We aimed to apply state-of-the-art, interpretable artificial intelligence (ie, predictions or prescription logic that can be easily understood) methods on real-world data to establish which groups of patients with GISTs should receive adjuvant imatinib, its optimal treatment duration, and the benefits conferred by this therapy. METHODS: In this observational cohort study, we considered for inclusion all patients who underwent resection of primary, non-metastatic GISTs at the Memorial Sloan Kettering Cancer Center (MSKCC; New York, NY, USA) between Oct 1, 1982, and Dec 31, 2017, and who were classified as intermediate or high risk according to the Armed Forces Institute of Pathology Miettinen criteria and had complete follow-up data with no missing entries. A counterfactual random forest model, which used predictors of recurrence (mitotic count, tumour size, and tumour site) and imatinib duration to infer the probability of recurrence at 7 years for a given patient under each duration of imatinib treatment, was trained in the MSKCC cohort. Optimal policy trees (OPTs), a state-of-the-art interpretable AI-based method, were used to read the counterfactual random forest model by training a decision tree with the counterfactual predictions. The OPT recommendations were externally validated in two cohorts of patients from Poland (the Polish Clinical GIST Registry), who underwent GIST resection between Dec 1, 1981, and Dec 31, 2011, and from Spain (the Spanish Group for Research in Sarcomas), who underwent resection between Oct 1, 1987, and Jan 30, 2011. FINDINGS: Among 1007 patients who underwent GIST surgery in MSKCC, 117 were included in the internal cohort; for the external cohorts, the Polish cohort comprised 363 patients and the Spanish cohort comprised 239 patients. The OPT did not recommend imatinib for patients with GISTs of gastric origin measuring less than 15·9 cm with a mitotic count of less than 11·5 mitoses per 5 mm2 or for those with small GISTs (<5·4 cm) of any site with a count of less than 11·5 mitoses per 5 mm2. In this cohort, the OPT cutoffs had a sensitivity of 92·7% (95% CI 82·4-98·0) and a specificity of 33·9% (22·3-47·0). The application of these cutoffs in the two external cohorts would have spared 38 (29%) of 131 patients in the Spanish cohort and 44 (35%) of 126 patients in the Polish cohort from unnecessary treatment with imatinib. Meanwhile, the risk of undertreating patients in these cohorts was minimal (sensitivity 95·4% [95% CI 89·5-98·5] in the Spanish cohort and 92·4% [88·3-95·4] in the Polish cohort). The OPT tested 33 different durations of imatinib treatment (<5 years) and found that 5 years of treatment conferred the most benefit. INTERPRETATION: If the identified patient subgroups were applied in clinical practice, as many as a third of the current cohort of candidates who do not benefit from adjuvant imatinib would be encouraged to not receive imatinib, subsequently avoiding unnecessary toxicity on patients and financial strain on health-care systems. Our finding that 5 years is the optimal duration of imatinib treatment could be the best source of evidence to inform clinical practice until 2028, when a randomised controlled trial with the same aims is expected to report its findings. FUNDING: National Cancer Institute.

Developing Novel Genomic Risk Stratification Models in Soft Tissue and Uterine Leiomyosarcoma.

PURPOSE: Leiomyosarcomas (LMS) are clinically and molecularly heterogeneous tumors. Despite recent large-scale genomic studies, current LMS risk stratification is not informed by molecular alterations. We propose a clinically applicable genomic risk stratification model. EXPERIMENTAL DESIGN: We performed comprehensive genomic profiling in a cohort of 195 soft tissue LMS (STLMS), 151 primary at presentation, and a control group of 238 uterine LMS (ULMS), 177 primary at presentation, with at least 1-year follow-up. RESULTS: In STLMS, French Federation of Cancer Centers (FNCLCC) grade but not tumor size predicted progression-free survival (PFS) or disease-specific survival (DSS). In contrast, in ULMS, tumor size, mitotic rate, and necrosis were associated with inferior PFS and DSS. In STLMS, a 3-tier genomic risk stratification performed well for DSS: high risk: co-occurrence of RB1 mutation and chr12q deletion (del12q)/ATRX mutation; intermediate risk: presence of RB1 mutation, ATRX mutation, or del12q; low risk: lack of any of these three alterations. The ability of RB1 and ATRX alterations to stratify STLMS was validated in an external AACR GENIE cohort. In ULMS, a 3-tier genomic risk stratification was significant for both PFS and DSS: high risk: concurrent TP53 mutation and chr20q amplification/ATRX mutations; intermediate risk: presence of TP53 mutation, ATRX mutation, or amp20q; low risk: lack of any of these three alterations. Longitudinal sequencing showed that most molecular alterations were early clonal events that persisted during disease progression. CONCLUSIONS: Compared with traditional clinicopathologic models, genomic risk stratification demonstrates superior prediction of clinical outcome in STLMS and is comparable in ULMS.

An interpretable AI model for recurrence prediction after surgery in gastrointestinal stromal tumour: an observational cohort study.

Background: There are several models that predict the risk of recurrence following resection of localised, primary gastrointestinal stromal tumour (GIST). However, assessment of calibration is not always feasible and when performed, calibration of current GIST models appears to be suboptimal. We aimed to develop a prognostic model to predict the recurrence of GIST after surgery with both good discrimination and calibration by uncovering and harnessing the non-linear relationships among variables that predict recurrence. Methods: In this observational cohort study, the data of 395 adult patients who underwent complete resection (R0 or R1) of a localised, primary GIST in the pre-imatinib era at Memorial Sloan Kettering Cancer Center (NY, USA) (recruited 1982-2001) and a European consortium (Spanish Group for Research in Sarcomas, 80 sites) (recruited 1987-2011) were used to train an interpretable Artificial Intelligence (AI)-based model called Optimal Classification Trees (OCT). The OCT predicted the probability of recurrence after surgery by capturing non-linear relationships among predictors of recurrence. The data of an additional 596 patients from another European consortium (Polish Clinical GIST Registry, 7 sites) (recruited 1981-2013) who were also treated in the pre-imatinib era were used to externally validate the OCT predictions with regard to discrimination (Harrell's C-index and Brier score) and calibration (calibration curve, Brier score, and Hosmer-Lemeshow test). The calibration of the Memorial Sloan Kettering (MSK) GIST nomogram was used as a comparative gold standard. We also evaluated the clinical utility of the OCT and the MSK nomogram by performing a Decision Curve Analysis (DCA). Findings: The internal cohort included 395 patients (median [IQR] age, 63 [54-71] years; 214 men [54.2%]) and the external cohort included 556 patients (median [IQR] age, 60 [52-68] years; 308 men [55.4%]). The Harrell's C-index of the OCT in the external validation cohort was greater than that of the MSK nomogram (0.805 (95% CI: 0.803-0.808) vs 0.788 (95% CI: 0.786-0.791), respectively). In the external validation cohort, the slope and intercept of the calibration curve of the main OCT were 1.041 and 0.038, respectively. In comparison, the slope and intercept of the calibration curve for the MSK nomogram was 0.681 and 0.032, respectively. The MSK nomogram overestimated the recurrence risk throughout the entire calibration curve. Of note, the Brier score was lower for the OCT compared to the MSK nomogram (0.147 vs 0.564, respectively), and the Hosmer-Lemeshow test was insignificant (P = 0.087) for the OCT model but significant (P < 0.001) for the MSK nomogram. Both results confirmed the superior discrimination and calibration of the OCT over the MSK nomogram. A decision curve analysis showed that the AI-based OCT model allowed for superior decision making compared to the MSK nomogram for both patients with 25-50% recurrence risk as well as those with >50% risk of recurrence. Interpretation: We present the first prognostic models of recurrence risk in GIST that demonstrate excellent discrimination, calibration, and clinical utility on external validation. Additional studies for further validation are warranted. With further validation, these tools could potentially improve patient counseling and selection for adjuvant therapy. Funding: The NCI SPORE in Soft Tissue Sarcoma and NCI Cancer Center Support Grants.

Novel Genomic Risk Stratification Model for Primary Gastrointestinal Stromal Tumors (GIST) in the Adjuvant Therapy Era.

PURPOSE: Traditional risk stratification schemes in gastrointestinal stromal tumors (GIST) were defined in the pre-imatinib era and rely solely on clinicopathologic metrics. We hypothesize that genomic-based risk stratification is prognostically relevant in the current era of tyrosine kinase inhibitor (TKI) therapeutics. EXPERIMENTAL DESIGN: Comprehensive mutational and copy-number profiling using MSK-IMPACT was performed. We integrated clinicopathologic and genomic parameters and utilized an elastic-net penalized Cox proportional hazards machine learning model for outcome risk stratification. RESULTS: A 3-tier genomic risk stratification model for recurrence-free survival (RFS) in 152 primary localized gastric and 80 small bowel GISTs was proposed. Gastric GISTs were classified as high risk if chr1p deletion or SDHB loss was present, and intermediate risk if chr14q deletion was present or KIT exon 11 mutation was absent. Small bowel GISTs were classified as high risk if MAX/MGA/MYC, CDKN2A, or RB1 alterations were present, and intermediate risk if chr1p deletion or chr5q amplification was present. Compared with conventional risk stratification, genomic risk stratification both upgrades and downgrades, suggesting that conventional risk stratification may underestimate or overtreat some high-risk and low-risk patients, respectively. Longitudinal sequencing detected most KIT-independent genomic alterations at baseline. Subanalysis in 26 SDH-deficient GISTs revealed that presence of TP53 mutations or chr1q amplifications portends worse RFS and disease-free survival. CONCLUSIONS: We developed a novel, next-generation genomic risk stratification model for primary gastric and small bowel GISTs, complementing traditional clinicopathologic models. Future independent validation of our model in external cohorts is essential.

Histology-Specific Prognostication for Radiation-Associated Soft Tissue Sarcoma.

PURPOSE: Radiation-associated sarcomas (RAS) are rare but aggressive malignancies. We sought to characterize the histology-specific presentation and behavior of soft tissue RAS to improve individualized prognostication. METHODS: A single-institutional prospectively maintained database was queried for all patients with primary, nonmetastatic RAS treated with surgical resection from 1982 to 2019. Patients presenting with the five most common RAS histologies were propensity-matched to those with sporadic tumors of the same histology. Incidence of disease-specific death (DSD) was modeled using cumulative incidence analyses. RESULTS: Among 259 patients with RAS, the five most common histologies were malignant peripheral nerve sheath tumor (MPNST; n = 19), myxofibrosarcoma (n = 20), leiomyosarcoma (n = 24), undifferentiated pleomorphic sarcoma (UPS; n = 55), and angiosarcoma (AS; n = 62). DSD varied significantly by histology (P = .002), with RAS MPNST and UPS having the highest DSD. In unadjusted analysis, RAS MPNST was associated with increased DSD compared with sporadic MPNST (75% v 38% 5-year DSD, P = .002), as was RAS UPS compared with sporadic UPS (49% v 28% 5-year DSD, P = .004). Unadjusted DSD was similar among patients with RAS AS, leiomyosarcoma, or myxofibrosarcoma and sporadic sarcoma of the same histology. After matching RAS to sporadic patients within each histology, DSD only differed between RAS and sporadic MPNST (83% v 46% 5-year DSD, P = .013). Patients with RAS AS presented in such a distinct manner to those with sporadic AS that a successful match was not possible. CONCLUSION: The aggressive presentation of RAS is histology-specific, and DSD is driven by RAS MPNST and UPS histologies. Despite the aggressive presentation, standard prognostic factors can be used to estimate risk of DSD among most RAS. In MPNST, radiation association should be considered to independently associate with markedly higher risk of DSD.

Clinical sequencing of soft tissue and bone sarcomas delineates diverse genomic landscapes and potential therapeutic targets.

The genetic, biologic, and clinical heterogeneity of sarcomas poses a challenge for the identification of therapeutic targets, clinical research, and advancing patient care. Because there are > 100 sarcoma subtypes, in-depth genetic studies have focused on one or a few subtypes. Herein, we report a comparative genetic analysis of 2,138 sarcomas representing 45 pathological entities. This cohort is prospectively analyzed using targeted sequencing to characterize subtype-specific somatic alterations in targetable pathways, rates of whole genome doubling, mutational signatures, and subtype-agnostic genomic clusters. The most common alterations are in cell cycle control and TP53, receptor tyrosine kinases/PI3K/RAS, and epigenetic regulators. Subtype-specific associations include TERT amplification in intimal sarcoma and SWI/SNF alterations in uterine adenosarcoma. Tumor mutational burden, while low compared to other cancers, varies between and within subtypes. This resource will improve sarcoma models, motivate studies of subtype-specific alterations, and inform investigations of genetic factors and their correlations with treatment response.

Prognostic stratification of clinical and molecular epithelioid hemangioendothelioma subsets.

Epithelioid hemangioendothelioma is a low-grade malignant vascular tumor with an intermediate clinical behavior between benign hemangiomas and high-grade angiosarcomas. Pathologic or molecular factors to predict this clinical heterogeneity are not well defined. A WWTR1-CAMTA1 fusion is present in most classic epithelioid hemangioendothelioma, regardless of their clinical behavior, suggesting that additional genetic abnormalities might be responsible in driving a more aggressive biology. A small subset of cases show distinct morphology and are characterized genetically by a YAP1-TFE3 fusion. Two histologic grades have been described in classic epithelioid hemangioendothelioma of the soft tissue. However, proposed criteria do not apply to other clinical presentations and have not been assessed in the YAP1-TFE3 positive tumors. Furthermore, no previous studies have compared the survival of these two molecular subsets. In this study we investigate the clinicopathologic and molecular findings of a large cohort of 93 translocation-positive epithelioid hemangioendothelioma managed at our institution. Patient characteristics, histologic features, treatment outcomes, and genetic abnormalities were investigated and these factors were correlated with overall survival. In 18 patients (15 with WWTR1-CAMTA1 and 3 with YAP1-TFE3) Memorial Sloan Kettering-IMPACT targeted DNA sequencing was performed to identify secondary genetic alterations showing more than half of tumors had a genetic alteration beyond the disease-defining gene fusion. Patients with conventional epithelioid hemangioendothelioma with WWTR1-CAMTA1 fusion had a less favorable outcome compared with the YAP1-TFE3 subset, the 5-year overall survival being 59% versus 86%, respectively. Soft tissue epithelioid hemangioendothelioma were frequently solitary, followed an uneventful clinical course being often managed with curative surgery. Multifocality, pleural involvement, lymph node or distant metastases had a significantly worse outcome. Patients with pleural disease or lymph node metastases had an aggressive clinical course akin to high-grade sarcomas, with 22% and 30%, respectively, alive at 5 years, compared with >70% survival rate in patients lacking these two adverse factors.

Yield of Colonoscopy in Identification of Newly Diagnosed Desmoid-Type Fibromatosis with Underlying Familial Adenomatous Polyposis.

BACKGROUND: Desmoid-type fibromatosis can arise in patients with familial adenomatous polyposis (FAP), therefore patients with desmoids often undergo colonoscopy to rule out FAP. Because finding FAP is uncommon, we sought to define subsets of desmoid patients in whom colonoscopy frequently identified FAP. METHODS: Patients with desmoid-type fibromatosis were identified from surgery and pathology databases at a single institution, and information on colonoscopy and FAP diagnosis was collected retrospectively. CTNNB1 mutation status was defined by Sanger sequencing and digital polymerase chain reaction of archived specimens. RESULTS: Among 626 patients with desmoids, 26 were diagnosed with FAP. In 20 patients, FAP diagnosis predated the desmoid diagnosis. Among patients without prior FAP diagnosis, 161 underwent colonoscopy, which identified only six cases of FAP (diagnostic yield 3.7%). Yields were substantially higher among patients with four characteristics: age < 40 years (11% yield), intra-abdominal or retroperitoneal tumors (5.4%), multifocal disease (29%), and family history (8%) (all p < 0.001). All cases of FAP were detected in patients younger than 40 years of age and with at least one of the other three characteristics. CTNNB1 mutation status was available in 82 patients with known FAP status. None of the 61 patients with CTNNB1 mutations were diagnosed with FAP, while 7 of the 21 patients with no CTNNB1 mutation detected (24%) were FAP patients. CONCLUSIONS: Patients with desmoid-type fibromatosis and undiagnosed FAP generally have multiple risk factors, which may be used to selectively recommend colonoscopic screening. Routine CTNNB1 sequencing may also rule out FAP and allow for deferral of colonoscopy.