Skip to main content
. 2020 Aug 8;15(2):273–286. doi: 10.1093/ecco-jcc/jjaa166

Table 3.

Multivariate analyses for development of B2 stricturing complications.

Model 1 Model 2 Model 3
Baseline predictor Estimate p-value Estimate p-value Estimate p-value
8.ECM transcriptome 1.5 [1.14–1.96] 0.0039 1.45 [1.1–1.91] 0.0067 1.4 [1.1–1.86] 0.0145
CBir [>25] - - 4.2 [1.44–14] 0.0116 3.4 [1.12–11.6] 0.0364
I gG.ASCA [>40] - - - - 4.32 [1.43–13] 0.0093
Model characteristics
AUC 0.73 [0.59–0.87] 0.79 [0.68–0.9] 0.82 [0.70–0.94]
AIC 112.4 107.5 102.8
R2 0.03 0.06 0.12
Sensitivity 75% [48–93] 81% [54–96] 88% [62–98]
Specificity 67% [61–74] 75% [69–80] 70% [63–76]
PPV 15% [11–42] 19% [15–56] 18% [14–66]
NPV 97% [92–98] 98% [94–99] 99% [95–99]
Model comparison
Model 2 vs model 1 Model 3 vs model 1
Likelihood ratio test - - LR = 6.92 0.0085 LR = 13.5 0.0012
Comparison of AUC - - 0.26 0.0103
Model 3 vs model 2
Likelihood ratio test LR = 6.62 0.01
Comparison of AUC 0.43

Baseline predictors are shown with odds ratios with 95% confidence intervals [CIs] in parentheses. Model characteristics were based on a predicted probability of 0.5. Models comparisons were completed with likelihood ratio test and DeLong’s test [to compare AUC between models].

AUC, area under the ROC curve; AIC, Akaike’s information criteria; PPV, positive predictive value; NPV, negative predicitive value; LR, likelihood ratio.