Skip to main content
. 2017 Jan 31;6(3):e1284720. doi: 10.1080/2162402X.2017.1284720

Table 3.

Ordinal logistic regression analysis to assess the association of tumor SQSTM1 expression level (predictor) with the density of T cells (outcome).

  Univariable OR Multivariable OR
  (95% CI) (95% CI)*
Model for CD3+ cell density (n = 579, as an ordinal quartile outcome variable)
Tumor SQSTM1 expression level Low 1 (referent) 1 (referent)
  Intermediate) 0.91 (0.62–1.33) 0.85 (0.58–1.25)
  High 0.91 (0.61–1.36) 0.89 (0.59–1.32)
  ptrend 0.68 0.63
Model for CD8+ cell density (n = 573, as an ordinal quartile outcome variable)
Tumor SQSTM1 expression level Low 1 (referent) 1 (referent)
  Intermediate 1.07 (0.73–1.57) 1.11 (0.75–1.64)
  High 1.44 (0.96–2.17) 1.46 (0.97–2.20)
  ptrend 0.06 0.06
Model for CD45RO+ cell density (n = 586, as an ordinal quartile outcome variable)
Tumor SQSTM1 expression level Low 1 (referent) 1 (referent)
  Intermediate 1.24 (0.84–1.81) 1.35 (0.92–1.99)
  High 1.22 (0.81–1.82) 1.28 (0.85–1.92)
  ptrend 0.40 0.31
Model for FOXP3+ cell density (n = 557, as an ordinal quartile outcome variable)
Tumor SQSTM1 expression level Low 1 (referent) 1 (referent)
  Intermediate 0.69 (0.47–1.02) 0.66 (0.45–0.98)
  High 0.55 (0.36–0.83) 0.55 (0.36–0.83)
  ptrend 0.005 0.006

Abbreviations: CI, confidence interval; OR, odds ratio.

*

The multivariable ordinal logistic regression model initially included age, sex, year of diagnosis, family history of colorectal cancer in any parent or sibling, tumor location, microsatellite instability, CpG island methylator phenotype, KRAS, BRAF, and PIK3CA mutations, and LINE-1 methylation level. A backward stepwise elimination with a threshold of p = 0.05 was used to select variables in the final models.

ptrend value was calculated by the linear trend across the ordinal categories of tumor SQSTM1 expression level (low, intermediate, and high) in the ordinal logistic regression model for the density of CD3+, CD8+, CD45RO (PTPRC)+, or FOXP3+ cells (an ordinal quartile outcome variable). Because we assessed four primary outcome variables, we adjusted two-sided α level to 0.01 (= 0.05/4) by simple Bonferroni correction.