Fig. 2

Schematic overview of the causal path between ACPA and DMARD-free sustained remission that was studied using a mediation model as described by Baron and Kenny [25]. According to the Baron and Kenny model there are three requirements for a true mediation relationship. The first step is to regress the dependent variable on the independent variable to confirm that the independent variable is predictive for the dependent variable. In our data, the absence of anticitrullinated protein antibodies (ACPA) was indeed predictive for disease-modifying antirheumatic drug (DMARD)-free sustained remission (HR 0.16, p < 0.001). The second step is to regress the mediator variable on the independent variable to confirm that the independent variable is a predictor of the mediator. The association between ACPA and bone marrow edema (BME) was not statistically significant although there was a strong tendency towards significance (β = 1.58, p = 0.055). The final step would be to regress the dependent variable on both the mediator and independent variable to confirm that the mediator is a significant predictor of the dependent variable and the effect of the independent variable from the first step is greatly reduced. This was not the case (HR for ACPA corrected for BME 0.12, p < 0.001). As shown in Table 2, BME scores were not associated with DMARD-free sustained remission (HR 0.99, p = 0.56). Thus, this mediation analysis revealed that magnetic resonance imaging (MRI)-detected BME does not mediate the association of ACPA with not achieving DMARD-free sustained remission