PO.CL09.01 · 临床研究
精准肿瘤学中基因组改变与临床试验代表性之间的差异
Disparities between genomic alterations and clinical trial representation in precision oncology
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
背景：精准肿瘤学已从基于组织学的治疗转向以分子驱动的框架，但临床试验的优先方向是否与真实世界的基因组图谱相符仍不明确。本研究评估了2010至2025年间以生物标志物为导向的实体瘤试验，并将试验代表性与TCGA中的基因组流行情况进行比较。
方法：从AACT ClinicalTrials.gov数据库中提取干预性实体瘤试验（2010-2025年），并使用Python进行处理。通过结构化筛选识别成人实体瘤研究，并通过对标题、入组标准和干预措施进行文本挖掘检测30个预定义的生物标志物。人工审核去除假阳性，并在AACT和TCGA之间统一生物标志物名称。TMB-high定义为≥10个突变/兆碱基，MSI-high定义为MSIsensor评分≥10。试验按分期、申办方类型和组织特异性进行分类。计算代表性比率（试验频率÷TCGA基因组流行率），比率>1.5归为过度代表，0.5-1.5归为均衡，<0.5归为代表性不足。针对每个生物标志物进行了TCGA总生存期（OS）分析。
结果：在7,259项以生物标志物为导向的实体瘤试验中，19.4%为组织无关型，54.5%为企业申办，79.6%为早期（I/II期）。在所分析的30个生物标志物中，相对于其TCGA基因组流行率，6个（20%）过度代表，16个（53%）均衡，8个（27%）代表性不足。过度代表的生物标志物包括EGFR（28.4% vs 8%）、VEGFA（12.5% vs 2%）、FGFR3（6.0% vs 3%）和HER2（12.4% vs 6%），反映了对高度可成药致癌驱动因素的优先关注。相比之下，代表性不足的生物标志物包括TP53（2.0% vs 37%）、STK11（0.5% vs 8%）、KEAP1（0.2% vs 4%）、TMB-high（2.9% vs 12%）和CDKN2A（7.3% vs 17%），其中大多数是尚无获批靶向治疗的抑癌基因，且具有公认的免疫相关性。TP53、CDKN2A和KEAP1的改变在TCGA中与显著较短的OS相关（均p < 0.001），凸显了它们在临床试验代表性有限的情况下仍具有的预后意义。
结论：当前的肿瘤学试验不成比例地聚焦于已确立的、可靶向的致癌基因，而高流行率的抑癌基因改变和免疫相关生物标志物仍代表性不足。这种失配凸显了基因组流行率与临床试验设计之间的转化鸿沟，并强调了拓宽生物标志物纳入范围、扩展组织无关型框架以及整合真实世界基因组数据以推进公平精准肿瘤学的机遇。
查看英文原文 English abstract
Background: Precision oncology has shifted from histology-based treatment toward molecularly driven frameworks, but it remains unclear whether clinical trial priorities align with the real-world genomic landscape. This study evaluated biomarker-driven solid tumor trials from 2010 to 2025 and compared trial representation with genomic prevalence in TCGA.
Methods: Interventional solid tumor trials (2010-2025) were extracted from the AACT ClinicalTrials.gov database and processed in Python. Adult solid tumor studies were identified via structured filters, and 30 predefined biomarkers were detected through text-mining of titles, eligibility criteria, and interventions. Manual review removed false positives, and biomarker names were harmonized across AACT and TCGA. TMB-high was defined as ≥10 mutations per megabase, and MSI-high was defined as MSIsensor score ≥10. Trials were categorized by phase, sponsor type, and tissue specificity. Representation ratios (trial frequency ÷ TCGA genomic prevalence) were calculated, with ratios >1.5 classified as over-represented, 0.5-1.5 as balanced, and <0.5 as under-represented. TCGA overall survival (OS) analyses were performed for each biomarker.
Results: Among 7,259 biomarker-driven solid tumor trials, 19.4% were tissue-agnostic, 54.5% were industry-sponsored, and 79.6% were early-phase (I/II). Of the 30 biomarkers analyzed, 6 (20%) were over-represented, 16 (53%) were balanced, and 8 (27%) were under-represented relative to their TCGA genomic prevalence. Over-represented biomarkers included EGFR (28.4% vs 8%), VEGFA (12.5% vs 2%), FGFR3 (6.0% vs 3%), and HER2 (12.4% vs 6%), reflecting prioritization of highly druggable oncogenic drivers. In contrast, under-represented biomarkers included TP53 (2.0% vs 37%), STK11 (0.5% vs 8%), KEAP1 (0.2% vs 4%), TMB-high (2.9% vs 12%), and CDKN2A (7.3% vs 17%), most of which were tumor suppressors without approved targeted therapies and have recognized immune relevance. Alterations in TP53, CDKN2A, and KEAP1 were associated with significantly shorter OS in TCGA (all p < 0.001), underscoring their prognostic significance despite limited clinical trial representation.
Conclusions: Current oncology trials disproportionately focus on established, targetable oncogenes, while high-prevalence tumor suppressor alterations and immune-related biomarkers remain underrepresented. This mismatch highlights a translational gap between genomic prevalence and clinical trial design and underscores opportunities to broaden biomarker inclusion, expand tissue-agnostic frameworks, and integrate real-world genomic data to advance equitable precision oncology.
利益披露 Disclosure
J. Zhang, None..
R. Santiago, None..
K. Bakirhan, None..
T. Tamdin, None.