miRNA | Cohort | Crude OR | Crude p value | Adjusted OR | Adjusted p value |
|---|
miR-99a | Discovery | 6.78 | 0.03 | 6.82 | 0.06 |
Validation | 0.32 | 0.28 | 0.85 | 0.91 |
miR-143 | Discovery | 0.45 | 0.04 | 0.39 | 0.04 |
Validation | 1.75 | 0.31 | 3.43 | 0.24 |
miR-23a | Discovery | 4.08 | 0.03 | 3.82 | 0.05 |
Validation | 2.81 | 0.31 | 1.86 | 0.65 |
miR-197 | Discovery | 4.32 | 0.02 | 5.00 | 0.03 |
Validation | 1.41 | 0.68 | 0.99 | 0.99 |
- In order to test if clinical parameters influenced the association between miRNA levels and response, each univariately selected miRNA was first inserted in a logistic regression model on response (crude) in the discovery and validation cohort (–ΔΔCrt and –ΔΔCt values respectively). Then the baseline clinical parameters that were different between discovery and validation (Additional file 2, p < 0.10) were added to a separate model (adjusted). The parameters that were used to adjust the association of miR-99a and miR-143 with response to ADA were (log-tranformed) CRP, DAS28, (the square root of) SJC and VAS-GH. The association of miR-23a and miR-197 with response to ETN was adjusted for age and (log-transformed) ESR. If clinical parameters would be the main cause for the inability to validate the findings, the adjusted ORs of the miRNA in discovery and validation should be comparable. The analyses showed that relationship between each miRNA and response was significantly influenced by clinical parameters in most cases; however, because the adjusted ORs between discovery and validation after correction are not comparable, the clinical parameters do not explain the found differences between the different cohorts
-
ADA adalimumab, CRP C-reactive protein, DAS28 disease activity score based on a 28-joint count, ESR erythrocyte sedimentation rate, ETN etanercept, OR odds ratio, SJC swollen joint count, VAS-GH visual analogue scale of general health
