Machine Learning-Assisted Recurrence Prediction for Patients With Early-Stage Non-Small-Cell Lung Cancer

PURPOSE

Stratifying patients with cancer according to risk of relapse can personalize their care. In this work, we provide an answer to the following research question: How to use machine learning to estimate probability of relapse in patients with early-stage non-small-cell lung cancer (NSCLC)?

MATERIALS AND METHODS

For predicting relapse in 1,387 patients with early-stage (I-II) NSCLC from the Spanish Lung Cancer Group data (average age 65.7 years, female 24.8%, male 75.2%), we train tabular and graph machine learning models. We generate automatic explanations for the predictions of such models. For models trained on tabular data, we adopt SHapley Additive exPlanations local explanations to gauge how each patient feature contributes to the predicted outcome. We explain graph machine learning predictions with an example-based method that highlights influential past patients.

RESULTS

Machine learning models trained on tabular data exhibit a 76% accuracy for the random forest model at predicting relapse evaluated with a 10-fold cross-validation (the model was trained 10 times with different independent sets of patients in test, train, and validation sets, and the reported metrics are averaged over these 10 test sets). Graph machine learning reaches 68% accuracy over a held-out test set of 200 patients, calibrated on a held-out set of 100 patients.

CONCLUSION

Our results show that machine learning models trained on tabular and graph data can enable objective, personalized, and reproducible prediction of relapse and, therefore, disease outcome in patients with early-stage NSCLC. With further prospective and multisite validation, and additional radiological and molecular data, this prognostic model could potentially serve as a predictive decision support tool for deciding the use of adjuvant treatments in early-stage lung cancer.

Determination of essential biomarkers in lung cancer: a real-world data study in Spain with demographic, clinical, epidemiological and pathological characteristics

Background

The survival of patients with lung cancer has substantially increased in the last decade by about 15%. This increase is, basically, due to targeted therapies available for advanced stages and the emergence of immunotherapy itself. This work aims to study the situation of biomarker testing in Spain.

Patients and methods

The Thoracic Tumours Registry (TTR) is an observational, prospective, registry-based study that included patients diagnosed with lung cancer and other thoracic tumours, from September 2016 to 2020. This TTR study was sponsored by the Spanish Lung Cancer Group (GECP) Foundation, an independent, scientific, multidisciplinary oncology society that coordinates more than 550 experts and 182 hospitals across the Spanish territory.

Results

Nine thousand two hundred thirty-nine patients diagnosed with stage IV non-small cell lung cancer (NSCLC) between 2106 and 2020 were analysed. 7,467 (80.8%) were non-squamous and 1,772 (19.2%) were squamous. Tumour marker testing was performed in 85.0% of patients with non-squamous tumours vs 56.3% in those with squamous tumours (p-value < 0.001). The global testing of EGFR, ALK, and ROS1 was 78.9, 64.7, 35.6% respectively, in non-squamous histology. PDL1 was determined globally in the same period (46.9%), although if we focus on the last 3 years it exceeds 85%. There has been a significant increase in the last few years of all determinations and there are even close to 10% of molecular determinations that do not yet have targeted drug approval but will have it in the near future. 4,115 cases had a positive result (44.5%) for either EGFR, ALK, KRAS, BRAF, ROS1, or high PDL1.

Conclusions

Despite the lack of a national project and standard protocol in Spain that regulates the determination of biomarkers, the situation is similar to other European countries. Given the growing number of different determinations and their high positivity, national strategies are urgently needed to implement next-generation sequencing (NGS) in an integrated and cost-effective way in lung cancer.

Multimodal prediction of response to neoadjuvant nivolumab and chemotherapy for surgically resectable stage IIIA non–small cell lung cancer

8542

Background: The NADIM trial (NCT03081689), led by the Spanish Lung Cancer Group, assessed the antitumor activity and safety of neoadjuvant chemoimmunotherapy for resectable stage IIIA NSCLC. Patients received neoadjuvant nivolumab and paclitaxel-carboplatin for three cycles before surgical resection, followed by one year of adjuvant nivolumab. At 24 months, progression-free survival (PFS) was 77%, suggesting that neoadjuvant chemoimmunotherapy represents a promising option in this setting. Pathological complete response (pCR) could potentially be used as an important surrogate endpoint for survival. We present here a re-analysis of the NADIM cohort aiming to develop a machine learning algorithm to predict the pCR status based on multimodal baseline data. Methods: We combined baseline clinical data (e.g., age, smoking status), biological data (e.g., tumor histology, mutations), radiology reports and radiomics analysis of the baseline CT scan in a multimodal analysis. While 46 patients were enrolled in the NADIM trial, only 28 had a complete set of data available for this retrospective study. For each patient, tumors were segmented on the baseline CT-scan in 3D by a Deep Learning algorithm. Radiomics features were extracted following the IBSI standards and combined with the other data modalities. A filter-based variable selection method was applied before training several machine learning algorithms. The optimization criterion was the Area Under the ROC Curve (AUC). Due to the small size of the cohort, a leave-one-out cross-validation approach was used to properly estimate the model performance. For a sub-cohort of 20 patients for which data have been collected longitudinally during the neoadjuvant treatment, an additional Delta-radiomics model was used to predict the pCR status. Results: An XGBoost algorithm with a linear base learner displayed an AUC of 0.69, a precision of 75%, a sensitivity of 83% and a specificity of 50%. Features with highest weight in the algorithm were a mix of radiological, radiomics, biological and clinical features (including the neutrophils to lymphocytes ratio, mutations and histology) highlighting the importance of a truly multimodal analysis. Indeed, withdrawing a specific data modality (e.g., radiomics or biological features), led to a decrease of ̃15% of the AUC. Inclusion of the Delta-radiomics analysis on the data collected longitudinally prior to surgery led to an improved AUC of 0.76 in that patient sub-cohort. Conclusions: This study is, to our knowledge, the first to offer a multimodal analysis of the response to neoadjuvant treatment for surgically resectable stage IIIA NSCLC and is a proof of concept that a machine learning algorithm can be used to predict the pCR in this context. These preliminary results are being confirmed in the ongoing NADIM II trial. Clinical trial information: NCT03838159.

Prognostic model of long-term advanced stage (IIIB-IV) EGFR mutated non-small cell lung cancer (NSCLC) survivors using real-life data

BACKGROUND

There is a lack of useful diagnostic tools to identify EGFR mutated NSCLC patients with long-term survival. This study develops a prognostic model using real world data to assist clinicians to predict survival beyond 24 months.

METHODS

EGFR mutated stage IIIB and IV NSCLC patients diagnosed between January 2009 and December 2017 included in the Spanish Lung Cancer Group (SLCG) thoracic tumor registry. Long-term survival was defined as being alive 24 months after diagnosis. A multivariable prognostic model was carried out using binary logistic regression and internal validation through bootstrapping. A nomogram was developed to facilitate the interpretation and applicability of the model.

RESULTS

505 of the 961 EGFR mutated patients identified in the registry were included, with a median survival of 27.73 months. Factors associated with overall survival longer than 24 months were: being a woman (OR 1.78); absence of the exon 20 insertion mutation (OR 2.77); functional status (ECOG 0-1) (OR 4.92); absence of central nervous system metastases (OR 2.22), absence of liver metastases (OR 1.90) or adrenal involvement (OR 2.35) and low number of metastatic sites (OR 1.22). The model had a good internal validation with a calibration slope equal to 0.781 and discrimination (optimism corrected C-index 0.680).

CONCLUSIONS

Survival greater than 24 months can be predicted from six pre-treatment clinicopathological variables. The model has a good discrimination ability. We hypothesized that this model could help the selection of the best treatment sequence in EGFR mutation NSCLC patients.

Abstract 560: High levels of baseline ctDNA constitute a poor prognostic factor in progression-free survival in patients receiving neo-adjuvant chemo-immunotherapy: Results from NADIM clinical trial

There is growing evidence supporting that long-term survival of neoadjuvant chemo-immunotherapy for locally advanced NSCLC patients can be achieved. However, some patients invariably progress within the short-term. Identification of patients at high risk of progression is needed to achieve a better control of the disease. Concentrations of baseline ctDNA have been shown to be of prognostic significance.

Patients and methods 42 pre-treatment plasma samples from the NADIM clinical trial (NCT03081689), in which resectable stage IIIA NSCLC patients were treated with neoadjuvant chemo-immunotherapy with Nivolumab, were analyzed by NGS, using the Oncomine Pan-Cancer Cell-Free Assay™ (Thermo Fisher Scientific®). Variant calling, annotation and filtering were performed on the Ion Reporter (v5.14) platform using the OncomineTagSeq Pan-Cancer Liquid Biopsy workflow (v2.3). The final variant matrix was obtained from vcf files as generated from Ion Reporter (v5.14) platform and applying an internal pipeline (R-code is available upon request). Progression disease was evaluated by RECIST criteria V1.1.

Results A total of 116 variants were detected in 88.10% (N=37) of the plasma samples collected before neoadjuvant treatment. The average number of variants detected per sample was 3.13. The most frequently mutated genes were TP53, which accounts for 59.52% of the detected variants, followed by PIK3CA (30.95%), MAP2K1 (30.95%), APC (23.81%), MTOR (9.52%) and KIT (9.52%). Patients in whom a GNA11 mutation was detected in the plasma sample by NGS showed worsen progression free survival (PFS) (HR: 14. 95%; CI: 2.6-71, P-value with Fold Discovery Rate correction: 0.019). Finally, ctDNA levels at Mutant Allele Frequency (MAF) below 1% at baseline were associated with improved PFS (P=0.025). At 30 months, PFS was 80.30% for these patients compared with 58.33% in patients with ctDNA levels ≥ 1%.

Conclusions Molecular profiling of liquid biopsies collected before neoadjuvant chemo-immunotherapy using NGS can identify patients at high risk of progression who might require more aggressive adjuvant treatment in order to achieve a better control of the disease.

Citation Format: Mariano Provencio, Roberto Serna-Blasco, Ernest Nadal, Amelia Insa, M. Rosario Garcia-Campelo, Diego Pereiro Corbacho, Manuel Domine, Margarita Majem, Delvys Rodriguez-Abreu, Alex Martinez-Marti, Javier de Castro, Manuel Cobo, Guillermo Lopez-Vivanco, Edel del Barco, Reyes Bernabe, Nuria Viñolas, Isidoro Barneto, Santiago Viteri, Eva Pereira, Ana Royuela, Marta Casarrubios, Clara Salas, Edwin R Parra, Ignacio Wistuba, Virginia Calvo, Raquel Laza-Briviesca, Bartomeu Massuti, Alberto Cruz-Vermudez, Atocha Romero. High levels of baseline ctDNA constitute a poor prognostic factor in progression-free survival in patients receiving neo-adjuvant chemo-immunotherapy: Results from NADIM clinical trial [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 560.

COVID-19 disease in patients with LUNG cancer in Spain: GRAVID LunG canceR pAtients coVid19 Disease (GRAVID STUDY)

e18709

Background: Patients with cancer may be more susceptible to infection and at increased risk of more severe COVID-19 disease; however, prognostic factors are not yet clearly identified. The LunG canceR pAtients coVid19 Disease (GRAVID) study aimed to describe clinical characteristics, outcomes, and predictors of poor outcome in patients with lung cancer and COVID-19. Methods: In this large nationwide study, we reviewed medical records of patients with lung cancer and confirmed COVID-19 diagnosis from 65 Spanish hospitals. Clinical features, treatments and disease outcomes were collected. The primary endpoint was to determine all-cause mortality; secondary endpoints were hospitalization and admission to intensive care units (ICU). Risk factors for poor prognosis were identified by univariate and multivariate logistic regression models. Results: Overall, 447 patients were included for analysis. Mean age was 67·1 ± 9·8 years; 332 (74·3%) were men, and 383 (85·7%) current/former smokers. NSCLC was the most frequent type of cancer (377, 84·5%), consisting mainly of adenocarcinoma (228, 51·0%), and stage III metastatic or unresectable disease (354, 79·2%). Two-hundred and sixty-six (59·5%) patients were receiving anticancer treatment, mostly first-line chemotherapy. In total, 350 (78·3%) patients were hospitalized for a mean of 13·4 ± 11·4 days, nine (2·0%) patients were admitted to the ICU, and 146 (32·7%) died. Advanced disease and the use of corticosteroids to treat COVID-19 during hospitalization were predictors of mortality. Hospitalized, non-end-of-life stage patients with lymphocytopenia and high LDH had an increased risk of death. Severity of COVID-19 correlated to higher mortality, ICU admission, and mechanical ventilation rates. Conclusions: Due to their underlying comorbidities and immunocompromised status, patients with lung cancer and COVID-19 show high hospitalization and mortality rates. These outcomes, alongside the identification of prognostic factors, may inform physicians on the risks and benefits in this population, in order to provide individualized oncological care.

Phase II multicenter randomized trial to assess the efficacy and safety of first line nivolumab in combination with paclitaxel in subjects with R/M HNSCC unable for cisplatin-based chemotherapy (NIVOTAX): A TTCC study

TPS6086

Background: Nivolumab is the standard of care for patients (pts) with R/M HNSCC in the platinum-refractory setting. Up to 20% of R/M HNSCC pts are ineligible for cisplatin-based CT due to poor performance status and/or comorbidities. ERBITAX (weekly cetuximab + paclitaxel) is a recommended regimen for this patient population according to the Spanish Society of Medical Oncology guidelines. Preclinical data suggests a role for paclitaxel as immuno-modulator, mainly by increasing tumor infiltrating CD8+ (Galluzzi L., et al 2015). NIVOTAX trial aims to evaluate efficacy and safety of nivolumab + paclitaxel vs ERBITAX as first-line treatment for R/M HNSCC pts with platinum-refractory disease or ineligible for platinum-based chemotherapy. Methods: NIVOTAX (NCT04282109) is a randomized, open-label, multicenter, phase II trial sponsored by the Spanish Group of Head and Neck Cancer Treatment (TTCC) including R/M HNSCC pts not amenable for curative-intent therapy, previously untreated for R/M disease and not candidates for cisplatin-based chemotherapy. Population is distributed in 3 Groups: 1= Platinum-refractory; 2=Platinum-sensitive but unable to receive cisplatin due to: Karnofsky performance status (KPS) 70% and/or major comorbidities (renal/heart failure, grade ≥2 hearing loss) and/or previous allergic reactions to platinum compounds; 3= Platinum-sensitive but cumulative cisplatin dose received ≥225 mg/m² for locally-advanced disease. Pt are stratified according to: KPS (70% vs 80-100%); PD-L1 by Combined Positive Score (CPS ≥1 vs < 1); and HPV positivity (HPV+ oropharynx vs HPV-/non-oropharyngeal). 141 Pt are being randomized 2:1 to NIVOTAX (nivolumab 240 mg q2 weeks + weekly paclitaxel at 80 mg/m2 up to 12 weeks followed by maintenance nivolumab 480 mg/ q4 weeks) or ERBITAX ( weekly 250 mg/m2 plus paclitaxel 80 mg/m2 up to 12 weeks followed by maintenance cetuximab 250 mg/m2 weekly). Both arms will be continued up to a maximum of 24 months. Primary end-point is to evaluate treatment efficacy in terms of 2-year overall survival (2-y OS). It is assumed that 2-y OS in the NIVOTAX arm will be at least 26% (10% gain when compared to the expected 16% 2-y OS rate in this pt population). Secondary objectives include progression free survival (PFS), overall response rate, disease control rate, duration of response, 6m PFS, 5y-OS and safety profile. Response endpoints will be assessed using RECIST 1.1 criteria. As of February 12, 2021, 64 pts have been randomized. Planned safety data review for the first 10 pts treated with NIVOTAX regimen did not show any unexpected AE. Clinical trial information: NCT04282109.

Seroprevalence and immunological memory against SARS-CoV-2 in lung cancer patients (p): SOLID study

8531

Background: Coronavirus disease 2019 (COVID-19) is diagnosed by detecting the virus by reverse transcription polymerase chain reaction (RT-PCR). The majority of p go on to develop antibodies (Ab) against viral proteins. However, it is not known how long these antibodies last nor whether cancer treatments could affect the duration of immune response. The prognosis and greater or lesser vulnerability of the oncological population are also unknown. Methods: This prospective, longitudinal, multicenter serological study in the setting of SARS-CoV-2 was carried out in 50 Spanish hospitals. Eligibility criteria was a diagnosis of any thoracic cancer. The first determinations were performed between April 21, 2020 and June 3, 2020, either for p in follow up or in active treatment. Between September 10, 2020, and November 20, 2020, the second antibody (Ab) determination was performed in all previously seropositive p. Clinical and treatment data were collected, as was their clinical situation at study end. Study objectives were to prospectively determine seroprevalence in unselected lung cancer p during the first wave of the pandemic; the natural history of these p; the persistence of immunity more than 4 months after first determination; protection or lack thereof against reinfection after this period, and the nature of such protection; and the influence of treatments on maintenance or loss of immunity. Results: Of 1,500 p studied, 128 were seropositive, representing an overall prevalence of 8.5% seropositivity [95% confidence interval [CI], 7.2%, 10.1%]. Seventy-five percent were in active cancer treatment. COVID-19 infection was suspected in 47.7% [95% CI, 38.8%, 56.6%]. A second determination was performed on average 4.5 months later [IQR: 4; 5] and obtained for 104 of the initially seropositive p (81%). A second determination could not be obtained in 24 p, the majority due to death caused by disease progression (73%). In the second determination, IgG was not detected in 30.8% (32/104) of p. The severity of the infection, the need for hospitalization (p: 0.032) and the presence of symptoms at diagnosis (p: 0.02), including fever (p: 0.005) and nasal congestion (p: 0.005), were associated with persistence of immunity in the second determination. No variables or treatments received were associated with Ab loss. At time of last follow-up among those p for whom a second determination was performed, 89% (93 p) had completely recovered from the virus, with no lasting after effects. Only 1 of the 128 (0.78%) seropositive p had died from COVID-19. Conclusions: The prevalence of infection in lung cancer p is similar to that of the general population. Immunity against SARS-CoV-2 does not appear to be compromised by treatment, persisting beyond 4 months. Neither do mortality rates appear to be particularly high in this unselected population.

Abstract S12-04: COVID-19 disease in patients with lung cancer in Spain: GRAVID Lung Cancer Patients Disease (GRAVID study)

Background: Previous reports indicate that lung cancer patients are at an increased risk of severe COVID-19 disease and higher mortality rate compared to general population. However, prognostic factors are not yet clearly identified. The LunG canceR pAtients coVid19 Disease (GRAVID) study aimed to describe clinical characteristics, outcomes and predictors of poor prognosis in patients with lung cancer and COVID-19. Methods: In this large nationwide prospective study, medical records of lung cancer patients with COVID-19 diagnosis from 65 spanish hospitals were included. Clinical features, treatments and disease outcomes were collected. The primary endpoint was to determine any-cause mortality; secondary endpoints were hospitalization and admission at intensive care units (ICU). Risk factors of poor prognosis were identified by univariable and multivariable logistic regression models. Results: Overal, 447 patients were analysed. Mean age was 67.1 ± 9·8 years, and the majority were men (332, 74·3%) and current/former smokers (383 (85.7%). NSCLC was the most frequent cancer type (377, 84.5%), being adenocarcinoma (228, 51·0%) the predominant histology. 354 patients (79.2%) had unresectable stage III or metastatic disease, and 266 (59.5%) where receiving anticancer treatment, mostly first-line chemotherapy. 350 (78.3%) patients were hospitalized for a mean of 13·4 ± 11·4 days, 9 (2.0%) patients were admitted to ICU, and 146 (32.7%) patients died. Advanced disease and corticosteroid treatment at hospitalization were predictors of mortality. Non-terminal stage hospitalized patients with lymphocytopenia and high LDH showed an increased risk of death. Severity of COVID-19 correlated to mortality, admission at ICU and mechanical ventilation. Conclusion: With underlying comorbidities and immunocompromised status, patients with lung cancer and COVID-19 present high hospitalization and mortality rates. These outcomes, alongside the identification of prognostic factors, may inform physicians on risks and benefits for this population to provide individualized oncological care.

Citation Format: Jose Maria Mazarico Gallego, Antonio Calles, Monica Antoñanzas, Cristina Pangua, Xabier Mielgo Rubio, Ernest Nadal, Rafael Lopez Castro, Ana Lopez-Martin, Edel del Barco, Manuel Domine, Virginia Calvo, Pilar Diz, Carmen Sandoval, Elia Sais Girona, Ivana Gabriela Sullivan, M. Angeles Sala, Gema Garcia Ledo, Jose Ramon Jarabo, Rosa Alvarez Alvarez, Javier Baena, Maria Gonzalez-Cao, Mariano Provencio. COVID-19 disease in patients with lung cancer in Spain: GRAVID Lung Cancer Patients Disease (GRAVID study) [abstract]. In: Proceedings of the AACR Virtual Meeting: COVID-19 and Cancer; 2021 Feb 3-5. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(6_Suppl):Abstract nr S12-04.