Altered Oral Bacteria Abundance Associated with Disease Specific Survival in Oral Squamous Cell Carcinoma

Purpose/Objective(s) The oral microbiome is altered in the presence of oral squamous cell carcinoma (OSCC). It is unknown if clinical outcomes, such as disease specific survival (DSS), are associated with differences in the abundance of select bacterial genera. The aim of this study is to determine if there are microbiome differences based on DSS in OSCC using The Cancer Microbiome Atlas (TCMA) and an in-tandem analysis of a prospectively collected database. Materials/Methods TCMA is a publicly available database containing curated, decontaminated microbial profiles for tumors from 1,772 patients. Data utilized from this database was limited to microbiome profiles and clinicopathologic features for OSCC patients. Separately, our institution collects oral swab samples from OSCC tumors prior to surgical treatment for 16S rRNA sequencing and follows outcomes prospectively. Statistical analysis was performed using R Studio. Predictor variables for DSS were analyzed using logistic regression models reported with hazard ratio (HR) and 95% confidence interval (CI). R Studio was used to plot Beta Diversity and PCoA using Vegan package, and the Wilcoxon signed-rank and Kruskal-Wallis test were performed to evaluate differential abundance between the bacterial genera. Statistical significance was defined as p<0.05. Results One hundred five patients with OSCC were included from TCMA with a mean age of 60 (std 13, min 19, max 90), 65% male (N=68) and 92% white (N=95) with diverse oral cavity primary sites with oral tongue most common (N=55, 52%). Twenty-eight (26%) were Stage I-II and 77 (74%) were Stage III-IVA. There were no patients with distant metastases. Rates of lymphovascular invasion (LVI) were 23% (N=18), perineural invasion (PNI) were 56% (N=47) and microscopic or gross extranodal extension (ENE) were 26% (N=20). Negative surgical margins occurred in 83 patients (84%), and the majority had no prior cancer diagnosis (N=100, 95%). Forty-one patients (59%) were disease free with a mean follow up of 35 months (std 31, min 2.2 months, max 173 months). Clinicopathologic features that were predictive of DSS included ENE (HR 2.73, 95% CI 1.18, 6.34, p=0.019) and being a current smoker (HR 3.12, 95% CI 1.08, 9.04, p=0.036). There were 40 bacterial genera identified in TCMA for OSCC tumors. No difference was observed in beta diversity between patients with recurrence versus no recurrence. Examining relative abundance of bacterial genera revealed that Leptotrichia (p=0.023) and Haemophilus (p=0.045) were differentially enriched based on DSS. Preliminary analysis of a prospectively collected database of OSCC microbiome oral swab samples also showed that changes in the relative abundance of Leptotrichia (p=0.00002) and Haemophilus (p=0.00023) were associated with recurrence. Conclusion Changes in the relative abundance of select oral bacteria genera are associated with recurrent OSCC. Shifts in the microbiome are seen prior to surgical treatment and may be predictive of clinical outcomes. The oral microbiome is altered in the presence of oral squamous cell carcinoma (OSCC). It is unknown if clinical outcomes, such as disease specific survival (DSS), are associated with differences in the abundance of select bacterial genera. The aim of this study is to determine if there are microbiome differences based on DSS in OSCC using The Cancer Microbiome Atlas (TCMA) and an in-tandem analysis of a prospectively collected database. TCMA is a publicly available database containing curated, decontaminated microbial profiles for tumors from 1,772 patients. Data utilized from this database was limited to microbiome profiles and clinicopathologic features for OSCC patients. Separately, our institution collects oral swab samples from OSCC tumors prior to surgical treatment for 16S rRNA sequencing and follows outcomes prospectively. Statistical analysis was performed using R Studio. Predictor variables for DSS were analyzed using logistic regression models reported with hazard ratio (HR) and 95% confidence interval (CI). R Studio was used to plot Beta Diversity and PCoA using Vegan package, and the Wilcoxon signed-rank and Kruskal-Wallis test were performed to evaluate differential abundance between the bacterial genera. Statistical significance was defined as p<0.05. One hundred five patients with OSCC were included from TCMA with a mean age of 60 (std 13, min 19, max 90), 65% male (N=68) and 92% white (N=95) with diverse oral cavity primary sites with oral tongue most common (N=55, 52%). Twenty-eight (26%) were Stage I-II and 77 (74%) were Stage III-IVA. There were no patients with distant metastases. Rates of lymphovascular invasion (LVI) were 23% (N=18), perineural invasion (PNI) were 56% (N=47) and microscopic or gross extranodal extension (ENE) were 26% (N=20). Negative surgical margins occurred in 83 patients (84%), and the majority had no prior cancer diagnosis (N=100, 95%). Forty-one patients (59%) were disease free with a mean follow up of 35 months (std 31, min 2.2 months, max 173 months). Clinicopathologic features that were predictive of DSS included ENE (HR 2.73, 95% CI 1.18, 6.34, p=0.019) and being a current smoker (HR 3.12, 95% CI 1.08, 9.04, p=0.036). There were 40 bacterial genera identified in TCMA for OSCC tumors. No difference was observed in beta diversity between patients with recurrence versus no recurrence. Examining relative abundance of bacterial genera revealed that Leptotrichia (p=0.023) and Haemophilus (p=0.045) were differentially enriched based on DSS. Preliminary analysis of a prospectively collected database of OSCC microbiome oral swab samples also showed that changes in the relative abundance of Leptotrichia (p=0.00002) and Haemophilus (p=0.00023) were associated with recurrence. Changes in the relative abundance of select oral bacteria genera are associated with recurrent OSCC. Shifts in the microbiome are seen prior to surgical treatment and may be predictive of clinical outcomes.