Comparison of the morphologic criteria (RECIST) and metabolic criteria (EORTC and PERCIST) in tumor response assessments: a pooled analysis
이 논문의 관계도
List view
Radiomics and Radiogenomics연결 근거: radiomics, radiomic
- Su 2023Radiogenomic-based multiomic analysis reveals imaging intratumor heterogeneity phenotypes and t…연결어: radiomics, radiomic, radiogenomic
- Henry 2022Investigation of radiomics based intra-patient inter-tumor heterogeneity and the impact of tumo…연결어: radiomics, radiomic, subsampling
- R. T. Mull 1984Mass Estimates By Computed Tomography Physical Density연결어: radiomics, radiomic
- Ploeg 2014Van Der Ploeg 2014 Modern Modelling Techniques Are Data Hungry A연결어: radiomics, radiomic
- Reuben 2017Genomic And Immune Heterogeneity Are Associated With연결어: radiomics, radiomic
- +27 more
Thoracic CT / PET Imaging연결 근거: ct, pet
- Kim 2007Pulmonary involvement in Churg-Strauss syndrome: an analysis of CT, clinical, and pathologic fi…연결어: ct, pulmonary, egpa
- Cha 2016Pulmonary Intravascular Lymphomatosis: Clinical, CT, and PET Findings, Correlation of CT and Pa…연결어: ct, pet, pulmonary
- DeFreitas 2021Complications of Lung Transplantation: Update on Imaging Manifestations and Management연결어: ct, pulmonary, lung transplantation
- Chen 2022Dynamic 18 F Fdg Petct Can Predict연결어: ct, pet, pulmonary
- Cottin 2016Respiratory manifestations of eosinophilic granulomatosis with polyangiitis (Churg-Strauss)연결어: pulmonary, egpa, churg-strauss
- +45 more
Shared Tags이 논문의 tags: CT, PET/CT, chemotherapy, deep learning, radiomics
- Shur 2021Radiomics in Oncology: A Practical Guide공유 tag: CT, PET/CT, deep learning, radiomics
- R. T. Mull 1984Mass Estimates By Computed Tomography Physical Density공유 tag: CT, deep learning, radiomics
- Cha 2016Pulmonary Intravascular Lymphomatosis: Clinical, CT, and PET Findings, Correlation of CT and Pa…공유 tag: CT, PET/CT, chemotherapy
- Reuben 2017Genomic And Immune Heterogeneity Are Associated With공유 tag: CT, PET/CT, radiomics
- Song 2017Imaging Phenotyping Using Radiomics To Predict Micropapillary공유 tag: CT, PET/CT, radiomics
- +54 more
요약
본 연구는 고형암 환자에서 morphologic criteria(RECIST)와 metabolic criteria(EORTC, PERCIST) 간 tumor response assessment의 일치도를 비교한 pooled analysis이다. RECIST 1.0/1.1과 EORTC/PERCIST를 적용한 기존 문헌을 체계적으로 수집하여 κ statistics와 McNemar test를 통해 반응률 및 일치율을 분석하였다. morphologic criteria 대비 metabolic criteria를 적용할 때 ORR가 통계적으로 유의하게 증가하며, 특히 RECIST상 SD인 환자에서 metabolic response로의 전환이 빈번하게 관찰되었다. 두 기준 간 일치도는 moderate에서 fair 수준에 머물러, 단일 영상 기반 평가의 한계를 정량적으로 보여준다.
방법
- Study design: PRISMA 지침을 따르는 pooled analysis 및 systematic review.
- Cohort/Patient population: RECIST vs EORTC 비교군 216명(8편 문헌), RECIST vs PERCIST 비교군 407명(9편 문헌). 폐암, 대장암, 유방암, 두경부암 등 다양한 고형암 환자 포함. 치료제는 targeted agents(41.2%)와 cytotoxic chemotherapy(58.8%)로 구성됨.
- Imaging/Profiling method: computed tomography(CT) 기반 RECIST 1.0/1.1과 [18F]-FDG positron emission tomography(PET) 기반 EORTC/PERCIST 기준을 비교.
- Endpoint/Statistical method: CR, PR, PD, SD 및 metabolic 반응 지표(CMR, PMR, PMD, SMD)를 endpoint로 사용. ORR는 CR/PR 또는 CMR/PMR 비율로 정의되었으며, McNemar test로 비교하였다. 기준 간 일치도는 unweighted κ statistics로 평가하였다(κ < 0: poor, 0-0.20: slight, 0.21-0.40: fair, 0.41-0.60: moderate, 0.61-0.80: substantial, >0.80: almost perfect).
주요 결과
- RECIST vs EORTC: 216명 중 일치도는 moderate(κ = 0.447; 95% CI 0.356-0.537)였으며, 86명(39.8%)에서 불일치 발생. EORTC 적용 시 70명은 tumor response가 upgrade, 16명은 downgrade됨. RECIST 기준 SD인 68명 중 56명(82.46%)이 CMR 또는 PMR로 전환됨. ORR는 EORTC 기준에서 53%로 RECIST(28%) 대비 유의하게 증가(p < 0.0001).
- RECIST vs PERCIST: 407명 중 일치도는 fair(κ = 0.389)였으며, 181명(44.5%)에서 불일치 발생. PERCIST 적용 시 151명은 upgrade, 30명은 downgrade됨. ORR는 PERCIST 기준에서 55%로 RECIST(30%) 대비 유의하게 증가(p < 0.0001).
- 불일치 양상: morphologic criteria와 metabolic criteria 간 tumor response 평가가 상이하며, 특히 targeted agents 치료 시 종양의 necrosis 또는 cystic change로 인해 CT 기반 크기 측정의 한계가 PET 기반 대사 영상에서 더 잘 포착되는 것으로 나타났다.
강점
단일 소규모 연구의 편향을 줄이고 다양한 고형암 및 치료군을 아우르는 pooled analysis를 통해 RECIST와 metabolic criteria 간 불일치 규모를 정량화하였다. 특히 targeted agents와 cytotoxic chemotherapy 모두에서 morphologic SD가 metabolic response(CMR/PMR)로 전환되는 빈도를 명확히 제시하여, 치료 초기 반응 평가 시 functional imaging(PET)의 임상적 가치를 뒷받침한다. wiki 내에서 radiomics 및 AI 모델의 ground truth 설정 시 RECIST만 의존할 경우 발생할 수 있는 endpoint mismatch 문제를 경고하고, multimodal imaging 기준을 마련하는 데 직접적인 참고 자료로 활용 가능하다.
해석
본 연구는 oncology 및 biomedical-imaging 문헌에서 RECIST가 고형암 치료 반응 평가의 gold standard로 자리 잡은 배경과 한계([18F]-FDG PET 기반 metabolic 변화 조기 포착 불가, obscuring margins 또는 cystic lesion 존재 시 측정 오류)를 정량적으로 입증한다. 특히 targeted therapy가 종양 크기 감소 대신 necrosis를 유도하는 메커니즘을 고려할 때, morphologic 영상만으로는 치료 효능을 과소평가할 수 있음을 보여준다. LLM Wiki의 imaging/AI 문헌들과 연결하면, radiomics나 deep learning 모델이 CT와 PET를 fusion하여 반응 예측 성능을 높일 때, RECIST 기반 라벨링의 한계를 인지하고 PERCIST/EORTC와 같은 metabolic endpoint를 보조 지표로 포함해야 함을 시사한다. 이는 향후 multimodal response assessment 기준과 AI validation framework 수립에 중요한 근거가 된다.