Table 2.

Mediation analysis in 698 patients with rheumatoid arthritis from the Leiden Early Arthritis Clinic, with radiographic severity of joint damage over 7 years as the outcome variable

Step 1: effect of age on possible mediators
	Effect (β)a	95 % CI	p value
SJC	1.00	1.00–1.01	0.11
TJC	1.00	1.00–1.01	0.55
Symptom duration	0.992	0.988–0.996	<0.001
CRP	1.016	1.011–1021	<0.001
RF	0.99	0.98–1.00	0.09
ACPA	0.98	0.97–0.99	<0.001
Step 2: effect of age on radiographic joint damage
	Effect (β)	95 % CI	p value
Ageing	1.034	1.029–1.040	<0.001
Step 3: effect of age and possible mediator on radiographic joint damage
	Effect (β)	95 % CI	p value
SJC	1.00	0.99–1.00	0.23
Ageing	1.035	1.029–1.040	<0.001
TJC	1.00	0.98–1.01	0.76
Ageing	1.037	1.030–1.044	<0.001
Symptom duration	1.003	1.002–1.005	<0.001
Ageing	1.035	1.029–1.040	<0.001
CRP	1.003	1.001–1.005	0.003
Ageing	1.033	1.027–1.038	<0.001
ACPA	1.37	1.16–1.60	<0.001
Ageing	1.035	1.030–1.040	<0.001
RF	1.30	1.10–1.52	0.002
Ageing	1.034	1.029–1.039	<0.001

^aThe effect size (β) of swollen joint count (SJC), tender joint count (TJC), symptom duration and C-reactive protein (CRP) reflect the increase per year increase of age. For example, the β for CRP is 1.016 this means that for every year increase in age there is 1.016-fold increase in CRP. A β of 0.992 indicates an increase 0.992- fold, hence actually a decrease. The effect size of anti-citrullinated protein antibody (ACPA) and rheumatoid factor (RF) reflect the odds ratio. Step 1, 2 and 3 of the mediation analyses are explained in Fig. 1. In step 1 a linear or logistic regression was used, in step 2 and 3 a multivariate normal regression analysis was used [27]. Also here the effects are per unit. For example, the β for age on joint damage is 1.034/year this means that for every year increase in age there is an increase of 3.4 % this is equal to an increase of 95.2 % every 20 years (1.034^20). All features (SJC, TJC, ACPA, RF, symptom duration, and age) were assessed at baseline