Artificial Intelligence to Help Non-Small Cell Lung Cancer Patients: Measure Lung Cancer Biology and Treatment Response Via Imaging
试验速览
- 阶段
- 不适用
- 状态
- 尚未招募
- 入组人数
- 1,000
- 主要终点
- Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at time of study enrolment
研究概览
简要总结
SALMON is a prospective, multi-center, multi-country, biomarker validation study that synergizes an extensive non-interventional biomarker discovery study on diagnostic images and tissue biopsies of non-small cell lung cancer NSCLC (rATLAS) with a smaller biomarker minimally interventional study on patients with metastases who undergo liquid biopsy and imaging follow-up for 2 years (aRECIST). A total of 1120 patients will be screened to get 1000 participants enrolled in rATLAS, and a subset of 250 participants will be screened to then recruit 150 participants also for aRECIST. The study will end after one visit for participants in rATLAS while there is a 2-years follow-up period for participants in aRECIST. Participants will not receive any treatment specific for this study, but might receive standard of care therapy or investigational products in the framework of another clinical study following the baseline visit.
The objectives of optimizing AI based tools for the assessment of EGFR status (rATLAS) and automated Response Evaluation Criteria in Solid Tumours 1.1 (RECIST 1.1) (aRECIST) will be achieved using a trial design that combines a biomarker discovery study design (cross-sectional for rATLAS) with a reader study design (follow-up study in aRECIST). Medical treatments in the aRECIST cohort are not dictated by study protocol, rather determined by the clinicians in line with standard clinical practice.
详细描述
Lung cancer is the leading cause of cancer-related death worldwide, accounting for an estimated 2.1 million deaths in 2018. About 80%-85% of lung cancers are NSCLC. The main subtypes of NSCLC are adenocarcinoma, squamous cell carcinoma, and large cell carcinoma. These subtypes, which start from different types of lung cells, are grouped together as NSCLC. For these reasons, the choice of NSCLC for this trial is the most relevant, as there is a clear societal burden / clinical need that will be addressed in SALMON.
A pressing need exists to uncover the genetic and molecular composition of tumors, as this information would accelerate the development of more effective cancer therapies. Since tumors differ in their biological makeup, treatments can now often be tailored towards individual patients, in a strategy termed personalized medicine. NSCLC perfectly illustrates this paradigm, with treatments either targeted at well-defined oncogenic pathways (epidermal growth factor receptor (EGFR) mutations, Anaplastic lymphoma kinase (ALK) & receptor tyrosine kinase 1 (ROS1) gene rearrangements, B-Rapidly Accelerated Fibrosarcoma gene (BRAF) mutations), and for so-called "wild-type" NSCLC, immunotherapy (with or without chemotherapy). The recent development of different types of immunotherapy has led to promising advances in the treatment of patients with NSCLC in advanced or metastatic disease.
Medical imaging plays a pivotal role in the assessment of tumors, including lung cancer, as it provides non-invasively the features to identify, characterize and stage the local tumor and its overall metastatic burden. The information provided by imaging can be still improved with a thorough post-processing analysis of the images, with "radiomics" being the most sophisticated tool for quantitative imaging. Through radiomics the imaging data obtained can be linked to specific biological properties of the tumor, acquired either from tissue or liquid biopsies. However, the knowledge on this is fragmented to date.
The first objective of this trial is to generate a radiomics atlas (rATLAS) that provides a broader link between biological descriptors and medical imaging features in NSCLC, to eventually identify oncologic pathways through medical images and quantify the overall burden of specific markers for targeted therapies. In particular, imaging biomarkers that can predict immunosensitivity and more accurately predict the prognosis than the existing ones will be assessed on the medical images and compared to tissue and liquid biopsies.
Apart from NSCLC diagnosis and staging, imaging plays a key role also in the evaluation of patient's follow-up, representing the gold standard for the assessment of response to therapy. However, the evaluation of medical images is amenable to radiologist's interpretation and many attempts have been made to face this issue in the last decades. The RECIST is a one-dimensional measure (shortest/longest diameter in the plane of measurement) created in 2000, updated in 2009 (RECIST 1.1), with a branch for immunotherapy finalized in 2017 (iRECIST). RECIST, in all its different declination, is used to assess whether a tumor in cancer patients is progressing, regressing or did not change before and after some event, such as therapy. RECIST criteria are recognized both in the EU and in US. Although representing an international attempt to overcome the subjective evaluation of tumor response to cancer treatment, biases of measurements remain as it relies on human assessments. In 1976 Moertel and Hanley acknowledged that "the culmination of most experimental therapeutic trials for solid tumors occurs when a [physician] places a ruler or caliper over a lump and attempts to estimate its size," and with this measurement comes the inevitable component of human error. In 2021 this issue is still present in clinical practice, but in the last years Artificial Intelligence (AI) has demonstrated remarkable progress in image-recognition tasks in research.
研究设计
- 研究类型
- Observational
- 观察模型
- Cohort
- 时间视角
- Prospective
入排标准
- 年龄范围
- 18 Years 至 —（Adult, Older Adult）
- 性别
- All
- 接受健康志愿者
- 否
入选标准
- •Participant must be aged at least 18 years
- •Willing and able to comply with clinic visits and study-related procedures.
- •Willing and able to provide signed informed consent.
- •Participant must be at first diagnosis of NSCLC and have the largest diameter of the primary tumor equal or greater than 2 cm.
- •Participant must be treatment naïve (includes radiotherapy).
- •Participant must have received a CT scan for the diagnosis of NSCLC according to "Imaging Protocol" document (Appendix 1).
- •Participant with confirmed availability of representative tumor specimens in formalin-fixed, paraffin-embedded (FFPE) blocks or ≥25 unstained slides (at least 10 unstained slides). Participant without adequate archival tumor specimens cannot be included
- •Additional inclusion criteria specific to aRECIST cohort:
- •Participant must be diagnosed with NSCLC Stage IV.
- •Participant must have a life expectancy ≥ 3 months.
- •Participant must have at least one lesion that is suitable for accurate repeated assessment (according to RECIST criteria).
- •Participant must be able to comply with standard of care visits for imaging purposes to follow-up on treatment response.
- •Participant must need to agree to undergo a liquid biopsy at baseline and at follow-up visits.
- •Participant must undergo either chemotherapy or immunotherapy after baseline visit, according to SoC.
排除标准
- •Pregnant or breast-feeding participants (to avoid radiation exposure)
- •Participant is either an employee of Radiomics or the investigational center or an immediate relative of an employee of Radiomics or the investigational center.
- •Participant with total body CT scan already performed at a different site with acquisition parameters different from those reported in the Imaging Protocol
- •Additional inclusion criteria specific to aRECIST cohort:
- •Participant who previously underwent or are planned for curable cancer surgery (lobectomy, wedge resection, pneumonectomy) or ablative radiotherapy on metastases.
结局指标
主要结局
Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at time of study enrolment
时间窗: At time of study enrolment
Equivalence will be assessed by comparing manual RECIST target response (central readings) with the target response as generated by the automated Radiomics aRECIST workflow. Categorical similarity measures between central panel RECIST and aRECIST will be computed through Cohen's kappa coefficient. aRECIST is considered successful if kappa is at least 0.7 (lower bound), in the full dataset of 150 patients.
Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at month 24
时间窗: Month 24
Equivalence will be assessed by comparing manual RECIST target response (central readings) with the target response as generated by the automated Radiomics aRECIST workflow. Categorical similarity measures between central panel RECIST and aRECIST will be computed through Cohen's kappa coefficient. aRECIST is considered successful if kappa is at least 0.7 (lower bound), in the full dataset of 150 patients.
Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at month 3
时间窗: Month 3
Equivalence will be assessed by comparing manual RECIST target response (central readings) with the target response as generated by the automated Radiomics aRECIST workflow. Categorical similarity measures between central panel RECIST and aRECIST will be computed through Cohen's kappa coefficient. aRECIST is considered successful if kappa is at least 0.7 (lower bound), in the full dataset of 150 patients.
Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at month 6
时间窗: Month 6
Equivalence will be assessed by comparing manual RECIST target response (central readings) with the target response as generated by the automated Radiomics aRECIST workflow. Categorical similarity measures between central panel RECIST and aRECIST will be computed through Cohen's kappa coefficient. aRECIST is considered successful if kappa is at least 0.7 (lower bound), in the full dataset of 150 patients.
Equivalence between aRECIST and manual RECIST in the evaluation of target lesions at month 12
时间窗: Month 12
Equivalence will be assessed by comparing manual RECIST target response (central readings) with the target response as generated by the automated Radiomics aRECIST workflow. Categorical similarity measures between central panel RECIST and aRECIST will be computed through Cohen's kappa coefficient. aRECIST is considered successful if kappa is at least 0.7 (lower bound), in the full dataset of 150 patients.
Identification of imaging biomarkers that discriminate EGFR status in patients with NSCLC with a minimum area under the curve of 0.65
时间窗: At time of study enrolment
A radiomics-based EGFR mutation prediction model will be trained and tested. The EGFR mutation prediction model is considered successful if its AUC of ROC is ≥ 0.65 in the independent test set of 200 patients.
次要结局
- Identification of imaging biomarkers that correlate with major oncogenic biomarkers to help guide drug development and therapy choice in NSCLC, with a minimum AUC of 0.65(At time of study enrolment)
- Reduction in diagnostic time and inter-reader variability compared to manual RECIST (local reading) in determining therapeutic response on target lesions.(Month 24)