Univariable | Multivariable | ||
---|---|---|---|

OR (95% CI) | OR (95% CI)^{a} | ||

Model for TILs^{b} (as an outcome variable; n = 1,039) | |||

Non–MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 1.83 (1.08–3.09) | 1.91 (1.12–3.25) | |

MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 0.47 (0.23–0.94) | 0.45 (0.22–0.92) | |

P_{interaction}^{c} | 0.002 | 0.002 | |

Model for intratumoral periglandular reaction^{b} (as an outcome variable; n = 1,040) | |||

Non–MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 1.98 (1.02–3.85) | 1.97 (1.00–3.86) | |

MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 0.48 (0.24–0.96) | 0.43 (0.21–0.87) | |

P_{interaction}^{c} | 0.004 | 0.002 | |

Model for peritumoral lymphocytic reaction^{b} (as an outcome variable; n = 1,036) | |||

Non–MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 1.53 (0.81–2.89) | 1.44 (0.76–2.76) | |

MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 0.66 (0.34–1.26) | 0.54 (0.27–1.05) | |

P_{interaction}^{c} | 0.068 | 0.038 | |

Model for Crohn-like lymphoid reaction^{b} (as an outcome variable; n = 874) | |||

Non–MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 2.48 (1.42–4.31) | 2.86 (1.62–5.06) | |

MSI-high | F. nucleatum-negative | 1 (referent) | 1 (referent) |

F. nucleatum-positive | 0.89 (0.45–1.76) | 0.99 (0.49–1.98) | |

P_{interaction}^{c} | 0.022 | 0.020 |

↵

^{a}The multivariable logistic regression model initially included sex, age at diagnosis, year of diagnosis, family history of colorectal cancer, tumor location, CpG island methylator phenotype-specific promoter status, long interspersed nucleotide element-1 methylation level, and*KRAS, BRAF*, and*PIK3CA*mutations. A backward elimination with a threshold*P*of 0.05 was used to select variables for the final models. The variables that remained in the final models for TILs and intratumoral periglandular reaction were shown in Supplementary Table S1.↵

^{b}To avoid violation of the proportional odds assumption, the most common score or less (i.e., 0 for TILs and Crohn-like lymphoid reaction; and 0–1 for intratumoral periglandular reaction and peritumoral lymphocytic reaction) was categorized as low, and higher score was categorized as high.↵

^{c}*P*_{interaction}(two-sided) was calculated using the Wald test for the cross-product of*F. nucleatum*status (positive vs. negative) and MSI status (high vs. non-high) in the logistic regression model.