Table 3.
No. of cases | Colorectal cancer-specific mortality
|
Overall mortality
|
|||||
---|---|---|---|---|---|---|---|
No. of events | Univariable HR (95% CI) |
Multivariable HR (95% CI)a |
No. of events | Univariable HR (95% CI) |
Multivariable HR (95% CI)a |
||
PTGS2-absent/low cancer | |||||||
MIR21 expression level | |||||||
Quartile 1 (lowest) | 73 | 25 | 1 (reference) | 1 (reference) | 44 | 1 (reference) | 1 (reference) |
Quartile 2 | 73 | 14 | 0.49 (0.26–0.95) | 0.57 (0.29–1.11) | 33 | 0.63 (0.40–1.00) | 0.71 (0.45–1.12) |
Quartile 3 | 76 | 26 | 0.99 (0.57–1.71) | 0.90 (0.52–1.57) | 43 | 1.07 (0.70–1.63) | 1.01 (0.66–1.56) |
Quartile 4 (highest) | 73 | 22 | 0.91 (0.51–1.62) | 0.61 (0.34–1.10) | 39 | 1.04 (0.68–1.61) | 0.89 (0.57–1.38) |
Ptrendb | 0.74 | 0.22 | 0.39 | 0.91 | |||
PTGS2-high cancer | |||||||
MIR21 expression level | |||||||
Quartile 1 (lowest) | 119 | 26 | 1 (reference) | 1 (reference) | 60 | 1 (reference) | 1 (reference) |
Quartile 2 | 117 | 30 | 1.21 (0.72–2.05) | 1.17 (0.69–1.98) | 65 | 1.29 (0.90–1.83) | 1.20 (0.84–1.72) |
Quartile 3 | 116 | 35 | 1.46 (0.88–2.43) | 1.29 (0.77–2.15) | 63 | 1.22 (0.86–1.74) | 1.06 (0.74–1.51) |
Quartile 4 (highest) | 118 | 53 | 2.45 (1.53–3.93) | 2.28 (1.42–3.67) | 82 | 1.86 (1.33–2.60) | 1.75 (1.25–2.45) |
Ptrendb | < 0.0001 | 0.0004 | 0.0008 | 0.004 | |||
Pinteractionc | 0.007 | 0.0004 | 0.06 | 0.036 |
Abbreviations: CI, confidence interval; HR, hazard ratio.
The multivariable Cox regression model included sex, age, year of diagnosis, family history of colorectal cancer in parent or sibling, tumor location, disease stage, tumor differentiation, microsatellite instability, CpG island methylator phenotype, KRAS, BRAF, and PIK3CA mutations, and long interspersed nucleotide element-1 (LINE-1) methylation level. A backward elimination with a threshold of P = 0.05 was used to select variables in the final models.
Test for a linear trend was conducted across ordinal quartile categories (1 to 4) of tumor MIR21 expression level as a continuous variable in the Cox regression model.
Pinteraction values (two-sided) were calculated by the Wald test on the cross-product term of tumor MIR21 expression (ordinal quartile categories [1 to 4]) and PTGS2 expression (ordinal categories; absent [1], low [2], and high [3]) variables in the Cox regression model.