Skip to main content
. 2020 Nov 4;18:298. doi: 10.1186/s12916-020-01769-6

Table 7.

Treatment-specific prediction of the presence of nADA in the replication data

Preparation Model Cohort OR 95% CI p AUC R2 Sensitivity Specificity
IFNβ-1a s.c. Without genetics KI 0.65 0.08 0.66 0.52
IFNβ-1a s.c. Without genetics TUM 0.60 0.06 0.71 0.42
IFNβ-1a s.c. PRS nADA presence 5 × 10−08 KI 3.89 2.35–6.45 1.44 × 10−07 0.85 0.42 0.78 0.78
IFNβ-1a s.c. PRS nADA presence 5 × 10−08 TUM 2.56 1.56–4.21 2.11 × 10−04 0.76 0.24 0.68 0.65
IFNβ-1a s.c. PRS top 30% vs. bottom 30% KI 73.86 11.77–463.61 4.42 × 10−06 0.91 0.59 0.78 0.90
IFNβ-1a s.c. PRS top 30% vs. bottom 30% TUM 13.78 3.00–63.28 7.45 × 10−04 0.83 0.38 0.80 0.76
IFNβ-1a s.c. SNP rs77278603 additive coding KI 4.49 2.41–8.36 2.14 × 10−06 0.82 0.36 0.74 0.74
IFNβ-1a s.c. SNP rs77278603 additive coding TUM 3.88 1.78–8.47 6.67 × 10−04 0.73 0.21 0.63 0.74
IFNβ-1a s.c. SNP rs77278603-A dominant coding KI 9.16 2.48–33.79 8.79 × 10−04 0.78 0.31 0.57 0.76
IFNβ-1a s.c. SNP rs77278603-A dominant coding TUM 3.85 1.10–13.49 3.51 × 10−02 0.72 0.20 0.56 0.68
IFNβ-1b s.c. Without genetics KI 0.70 0.15 0.48 0.82
IFNβ-1b s.c. Without genetics TUM 0.58 0.02 0.82 0.20
IFNβ-1b s.c. PRS nADA presence 1 × 10−06 KI 2.40 1.45–3.97 6.46 × 10−04 0.78 0.33 0.57 0.85
IFNβ-1b s.c. PRS nADA presence 1 × 10−06 TUM 2.15 1.43–3.23 2.28 × 10−04 0.73 0.22 0.73 0.58
IFNβ-1b s.c. PRS top 30% vs. bottom 30% KI 10.16 2.30–44.95 2.25 × 10−03 0.83 0.46 0.58 0.87
IFNβ-1b s.c. PRS top 30% vs. bottom 30% TUM 5.97 2.03–17.52 1.14 × 10−03 0.78 0.33 0.69 0.75
IFNβ-1b s.c. SNP rs28366299 additive coding KI 4.51 1.72–11.80 2.14 × 10−03 0.77 0.31 0.57 0.82
IFNβ-1b s.c. SNP rs28366299 additive coding TUM 6.91 3.18–15.03 1.07 × 10−06 0.79 0.32 0.74 0.61
IFNβ-1b s.c. SNP rs28366299-A dominant coding KI 9.78 2.68–35.74 5.62 × 10−04 0.83 0.40 0.62 0.83
IFNβ-1b s.c. SNP rs28366299-A dominant coding TUM 7.56 3.01–19.02 1.71 × 10−05 0.80 0.33 0.77 0.57

Predictors in the model without genetics: sex, age, treatment duration, and titration site. The genetic models contained the same base model plus the indicated genetic factors and ancestry components. The top models are indicated in bold font. OR odds ratio; CI 95% confidence interval; p p value of the genetic component; AUC area under the receiver operating characteristic curve; R2 Nagelkerke's pseudo-R2; KI Karolinska Institutet, Sweden; TUM Technical University of Munich, Germany