Erwin Dreesen (1), Bram Verstockt (2), Marc Ferrante (2), Séverine Vermeire (2), Thomas Bouillon (1), Ann Gils (1)
(1) Department of Pharmaceutical and Pharmacological Sciences, University of Leuven, Belgium (2) Department of Gastroenterology and Hepatology, University Hospital Leuven, Belgium
Objectives: Vedolizumab (VDZ) is a monoclonal antibody used to treat patients with ulcerative colitis (UC) and Crohn’s disease (CD). Patients receive fixed 300 mg IV doses of VDZ at week (w)0, w2 and w6 (induction therapy) and q8w thereafter (maintenance therapy). Despite a significantly higher response rate to VDZ than to placebo, only a minority of patients achieved disease remission upon VDZ induction therapy [1-3]. VDZ exposure differs widely between patients and exposure-response (E-R) relationships have been established [4]. We previously demonstrated that VDZ dose optimisation to a w2 trough concentration (TC)>28.9 mg/L predicted 75% of patients with UC to achieve endoscopic remission (ER; decrease in Mayo endoscopic subscore [MES] from 3 [severe disease] or 2 [moderate disease] at baseline to 1 [mild disease] or 0 [inactive disease] at w14 of therapy). We aimed to develop a population pharmacokinetic (popPK) model and an E-R model to support individualised VDZ dose optimisation to a predefined TC target and the associated therapeutic outcome.
Methods: A total of 939 consecutive trough samples (from w2 to w30) of 178 patients (66 UC, 112 CD; excl. 1/179 patients with antibodies to VDZ) was used to develop a popPK model [4]. We analysed these data under a known 2cmt model with parallel linear and nonlinear clearance by using prior distributions from the GEMINI popPK model [5] to support estimation of PK parameters that were poorly supported by the current data (NONMEM 7.4 with $PRIOR). Classical stepwise covariate selection procedure was employed (forward α=0.01, backward α=0.001). An E-R Markov model was implemented to explore the relationship between individual model-predicted VDZ TC at w2 and the probability of achieving ER at w14 in patients with UC (incl. 54/66 patients who had MESbaseline>1). The predicted VDZ TC were modelled to affect the transition probabilities between MES states from baseline to w14 [6]. Simulations were performed using Berkeley-Madonna 8.3.18.
Results: Our model with fully data-driven estimation of the linear clearance (CLL; 0.207 L/day [3%], typical value [relative standard error]) and volume of distribution in the central compartment (Vc; 4.62 L [9%]) showed good predictive ability. Linear terminal elimination half-life of VDZ was 23.2 days. Lower albumin, mean platelet volume and haemoglobin, and higher C-reactive protein and fat-free mass [7] were associated with higher CLL, thus predicting lower VDZ exposure. Only 11% and 4% of the interindividual variability (IIV) of CLLand Vc was explained by these covariates, leaving 28% and 42% of the IIV unexplained.
Given the large unpredictable IIV and the absence of a safety exposure limit, it is reasonable to provide an optimised, fixed starting dose to all patients [8]. Optimising this starting dose from 300 mg to 600 mg predicted the probability of attaining the 28.9 mg/L w2 target to increase from 44% to 96% (N=2000 simulated patients). Model-based dose individualisation may be implemented from w2 onwards to target all patients to the w6 exposure target and reduce drug expenditures due to unnecessary overexposure.
ER was achieved in 32/54 patients with UC. The objective function value (OFV) dropped with 18.2 points from the null model, where transition probabilities between MES were driven by chance alone, to the E-R model informed by the individual predicted VDZ TC at w2. A VDZ TC at w2 of 7.0 mg/L [31%] was estimated to yield a 50% probability of going from MESbaseline3 to MESw142. A VDZ TC at w2 of 14.1 mg/L [28%] was estimated to yield a 50% probability of going from MESbaseline 2 to MESw14 1 or 0. Targeting all patients in our cohort (63% and 37% at MESbaseline 3 and 2, resp.) to the previously established 28.9 mg/L VDZ TC at w2 predicted a 59% probability of achieving ER:
0.63x[(28.9/7.0)/(1+(28.9 /7.0))x(28.9 /14.1)/(1+(28.9/14.1))]+0.37x[(28.9/14.1)/(1+(28.9/14.1))]
=0.63×0.54+0.37×0.67
=0.59
Conclusions: Patients with UC may benefit from a double VDZ starting dose. Our models may be implemented in a therapeutic drug monitoring (TDM) software tool to support dose optimisation for precise attainment of exposure targets after w2 [4] and the associated outcome probabilities. TDM can aid the go/no-go decision for continuing to VDZ maintenance therapy. When a patient does not achieve ER despite ‘sufficient’ VDZ exposure, pharmacodynamic failure is implicit and the patient may be switched to therapy with another mechanism of action.
References:
[1] Feagan BG et al. N Engl J Med (2013) 369, 699–710.
[2] Sandborn WJ et al. N Engl J Med (2013) 369, 711–21.
[3] Sands BE et al. Inflamm Bowel Dis (2017) 23, 97–106.
[4] Dreesen E et al. Clin Gastroenterol Hepatol (2018) [Epub ahead of print]
[5] Rosario M et al. Aliment Pharmacol Ther (2015) 42, 188-202.
[6] Dreesen E et al. Br J Clin Pharmacol (2019) [Epub ahead of print]
[7] Janmahasatian S et al. Clin Pharmacokinet (2005) 44(10), 1051-65.
[8] Holford N et al. Ther Drug Monit (2012) 34(5), 565-8.
Reference: PAGE 28 (2019) Abstr 8986 [www.page-meeting.org/?abstract=8986]
Poster: Clinical Applications