PK of i.d. and s.c. adalimumab administration
The adalimumab concentration time profile is displayed in Figure 3D . First, a non-compartmental analysis of pharmacokinetics (PK) was performed. After exclusion of subjects where any leakage occurred during injection, in the remaining subjects Cmax was significantly higher after i.d. injection compared to s.c. injection (90% CI 0.57-0.90, p=0.02). No difference was detected in AUC0-inf (90% CI 0.55-1.09, p=0.22) or AUC0-last (90% CI 0.60-1.07, p=0.20) (per protocol subjects in Table 2, all enrolled subjects inSupplementary table 1 ). These data show that i.d. administration of adalimumab yields a higher maximum concentration than s.c. administered adalimumab.
To further examine PK and to be able to correct for inter-individual variation in the kinetics of adalimumab and the formation of anti-adalimumab antibodies, a population PK model was developed. After exclusion of subjects in which any spill of adalimumab occurred during administration, data from 10 s.c. and 9 i.d. injections was available for model development using 275 adalimumab measurements that were above the lower limit of detection (LOD). A total of 4% of the measurements was below the LOD and therefore excluded from analysis. A significant effect between the time-varying titre levels and the CL was identified (p<0.001), indicating that the CL of adalimumab increases in the presence of high titre levels. However, a bias in the absorption kinetics for s.c. and i.d. was identified with linear absorption kinetics. Subsequent exploration of different structural absorption models resulted in a model event time (MTIME) function for the absorption rate constant (ka) after i.d. administration and two separate absorption compartments with equal kas and one with an absorption lag time for s.c. administration to be best fit for purpose (Figure 3E ). In this revised structural model, significant (p<0.01) inter-individual variability on the titre-CL relationship and the central volume of distribution was identified. Additionally, a significant (p<0.01) improvement in model fit was quantified after estimating a 29% higher bioavailability (F) after i.d. administration of adalimumab compared to s.c. administered adalimumab. A negative age-CL relationship and a positive weight-CL relationship were identified. Both covariates gave p<0.001 improvement in the model fit. The developed model showed an overall accurate description of the absorption and elimination phase of adalimumab (Supplementary figure 2A-B ). Model parameters (Table 3 ) were estimated with high precision and were comparable to literature values (23) . Simulations of the typical adalimumab absorption rates over time showed a clear difference between both administration routes, in which the i.d. dose had a fast initial phase which decreased after MTIME, whereas the s.c. administration had a slower initial phase and a small increase in the absorption rate, approximately 2 hours after dosing (Figure 3F ).
Cytokine production was assessed by stimulating ex vivo whole blood with LPS and aluminium hydroxide, driving NFκB and NLRP3 inflammasome activation. Results are shown in Figure 4 . Free TNFα levels after both s.c. and i.d. administration sharply decreased from pre-dose to post-dose (mean levels pre-dose i.d. 897 pg/mL, i.d. 48h post-dose 50 pg/mL, s.c. pre-dose 928 pg/mL, s.c. 48h post dose 74 pg/mL), as has been reported earlier (16), and returned to baseline at the end of study (i.d. 70 days post-dose 1149 pg/mL, s.c. 70 days post dose 850 pg/mL). No significant differences in inhibition of cytokine release were detected when i.d. adalimumab administration was compared to s.c. adalimumab administration (IFNγ p=0.61; IL-6 p=0.31; IL-8 p=0.81; IL-1β p=0.61; TNFα p=0.80). For LPS/aluminium hydroxide induced IFNγ production after adalimumab administration, a gender effect has been reported (14). A gender effect was not detected in this study (IFNγ p=0.99, IL-6 p=0.80; IL-8 p=0.96; IL-1β p=0.75; TNFα p=0.08).