PO.PS01.05 · 人群科学
口腔细菌组及其与癌症特异性死亡率的关联
Oral bacteriome and its association with cancer-specific mortality
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
目的：癌症是美国的首要死因，每年造成超过600,000人死亡。新出现的证据表明，口腔微生物群（人体最多样化的微生物群落之一）的改变可能影响癌症结局。然而，微生物多样性和组成与癌症死亡率之间的关联仍知之甚少。我们旨在识别口腔微生物组多样性和细菌丰度与癌症特异性死亡率之间的关联。
方法：我们分析了来自2009-2012年全国健康与营养检查研究（NHANES）并链接至国家死亡索引（NDI）的441名参与者。对漱口水样本进行16S rRNA V4扩增子测序，并使用SILVA v123解析到属水平。在稀释至2,000条读长后计算α多样性。癌症死亡使用ICD-10恶性肿瘤编码定义。作为一项探索性分析，将监督式随机森林模型（1,000棵树；100个随机种子；"randomForest" R包）应用于中心对数比（CLR）转换的丰度值，以识别与癌症死亡相关的属，重要性通过平均准确度下降（Mean Decrease Accuracy）衡量。采用调查加权多变量Cox比例模型，按α多样性和属丰度估计癌症特异性死亡率的校正风险比，并校正年龄、性别、种族/族裔、贫困收入比、终生吸烟史和自评口腔健康。采用Bonferroni校正以应对多重假设检验。
结果：较高的α多样性与较低的癌症特异性死亡率相关，Shannon指数（aHR 0.72，95% CI 0.53-0.98）和逆Simpson指数（aHR 0.13，95% CI 0.02-0.73）呈显著负相关，而Faith系统发育多样性和观察到的ASV丰富度不显著。在使用随机森林的探索性分析中，Prevotella_7、Gemella、Bergeyella、Fusobacterium、Ruminococcaceae_UCG-014、Alloprevotella和Oribacterium成为区分癌症死亡与非癌症死亡的潜在判别属（MDA>5.75）。在校正的Cox模型中，Bergeyella丰度高于中位数的参与者（aHR 0.59；95% CI 0.35-0.99）和有Oribacterium的参与者（aHR 0.52；95% CI 0.32-0.81）的癌症特异性死亡率较低。校正后，探索性分析中识别出的其他属对癌症死亡不显著。
结论：口腔微生物组中存在可能对癌症特异性死亡率具有预测价值的细菌。尽管既往文献对Bergeyella和Oribacterium在癌症死亡中的作用支持有限，我们的结果强调需要进一步研究来验证我们的发现，并需考察用于微生物组指导型预防策略的生物标志物和靶点。
查看英文原文 English abstract
Objectives: Cancer is the leading cause of death in the United States, with more than 600,000 deaths annually. Emerging evidence suggests that alterations in the oral microbiota, one of the body's most diverse microbial communities, may influence cancer outcomes. However, associations between microbial diversity and composition with cancer mortality remain poorly understood. We aimed to identify associations between oral microbiome diversity and bacterial abundance with cancer-specific mortality.
Methods: We analyzed 441 participants from the 2009-2012 National Health and Nutrition Examination Study (NHANES) linked to National Death Index (NDI). Oral rinse samples were 16S rRNA V4 amplicon-sequenced and resolved to the genus level using SILVA v123. Alpha diversity was calculated after rarefaction to 2,000 reads. Cancer deaths were defined using ICD-10 malignant neoplasm codes. As an exploratory analysis, supervised random forest models (1,000 trees; 100 random seeds; “randomForest” R package) were applied to centered log-ratio (CLR) transformed abundance values to identify genera associated with cancer-mortality, with importance measured by Mean Decrease Accuracy. Survey-weighted multivariable Cox proportional models estimated adjusted hazard ratios for cancer-specific mortality by alpha diversity and genus abundance, adjusting for age, sex, race/ethnicity, poverty income ratio, lifetime smoking history, and self-rated oral health. Bonferroni correction accounted for multiple hypothesis testing.
Results: Higher alpha diversity was associated with lower cancer-specific mortality, with significant inverse associations for Shannon (aHR 0.72, 95% CI 0.53-0.98) and Inverse Simpson's Index (aHR 0.13, 95% CI 0.02-0.73), while Faith's Phylogenetic Diversity and observed ASV richness were not significant. In the exploratory random analysis using random forest, Prevotella_7, Gemella, Bergeyella, Fusobacterium, Ruminococcaceae_UCG-014, Alloprevotella, and Oribacterium emerged as potential discriminatory genera for cancer deaths versus non-cancer deaths (MDA > 5.75). In adjusted Cox models, participants with Bergeyella abundance above the median (aHR 0.59; 95% CI 0.35-0.99) and participants with Oribacterium (aHR 0.52; 95% CI 0.32-0.81) had lower cancer-specific mortality. After adjustment, the other genera identified in the exploratory analysis were not significant for cancer-mortality.
Conclusion: The oral microbiome harbors bacteria that may be potentially predictive of cancer-specific mortality. Although prior literature provides limited support for Bergeyella's and Oribacterium's roles in cancer mortality, our results underscore the need for further studies to validate our results and for examining biomarkers and targets for microbiome-informed prevention strategies.
利益披露 Disclosure
S. Han, None..
K. Ciarrocca, None.