A Clinical Decision Support Tool May Help to Optimise Vedolizumab Therapy in Crohn's Disease

Parambir S. Dulai; Aurelien Amiot; Laurent Peyrin-Biroulet; Vipul Jairath; Melanie Serrero; Jerome Filippi; Siddharth Singh; Benjamin Pariente; Edward V. Loftus Jr; Xavier Roblin; Sunanda Kane; Anthony Buisson; Corey A. Siegel; Yoram Bouhnik; William J. Sandborn; Karen Lasch; Maria Rosario; Brian G. Feagan; Daniela Bojic; Caroline Trang-Poisson; Bo Shen; Romain Altwegg; Bruce E. Sands; Jean-Frederic Colombel; Franck Carbonnel


Aliment Pharmacol Ther. 2020;51(5):553-564. 

In This Article


Data Sources and Participants

Methodology for the development and validation of our CDST has been published previously.[13] In the current study, individual participant data from the phase 3 VDZ in CD trial (GEMINI 2) were used in combination with observational cohort data from the VICTORY consortium and GETAID collaboration.[14,15] A treat-straight-through cohort was created from the GEMINI 2 clinical trial programs to mimic an observational cohort design. Patients from VICTORY and GETAID were included if they had started VDZ therapy for clinically or endoscopically active CD, had a follow-up clinical or endoscopic assessment of disease activity after VDZ initiation, and had baseline data available to calculate the CDST.

Clinical Decision Support Tool

The CDST is calculated using the following five variables:

  1. No prior bowel surgery (+2 points).

  2. No prior TNF-antagonist therapy (+3 points).

  3. No prior fistulising disease (+2 points).

  4. Baseline albumin (+0.4 points per g/L).

  5. Baseline C-reactive protein (−0.5 points if 3.0–10.0 mg/L; −3.0 points if >10 mg/L).

Patients with a score of 13 points or less are classified as low probability, >13 to 19 points as intermediate probability and >19 points as high probability.[13]


Our main objectives were to determine whether the previously created and validated CDST predicted measured VDZ concentrations (trough and peak) in the 52-week GEMINI 2 clinical trial and whether differences in measured drug exposure corresponded to differences in drug efficacy and rapidity of onset of action as assessed by reductions in Harvey-Bradshaw Index (HBI) over time (exposure-efficacy). The HBI was chosen given its widespread use internationally in routine practice, its availability in the GETAID cohort dataset, and its good correlation with the Crohn's Disease Activity Index.[16] Secondary objectives were to (a) externally validate the CDST in an independent multicentre cohort (GETAID collaboration) and (b) determine whether the CDST identified patients most likely to benefit from VDZ dose intensification for apparent lack of response. Finally, we assessed whether differences in predicted exposure-efficacy correlated with achievement of endoscopic remission and the likelihood of undergoing surgery for CD.

VDZ Pharmacokinetics

VDZ concentrations were assessed in the GEMINI 2 trials using serum samples with a direct VDZ capture PK assay. A sandwich ELISA assay was used for quantifying VDZ in human serum. Serum concentrations of VDZ were determined in accordance with good laboratory practice. The lower limit of detection was 0.125 μg/mL. Time points for trough concentration assessments taken 30 minutes before VDZ infusions were weeks 0, 2, 6, 22 and 46. Additional concentration assessments were taken at weeks 4, 14, 38 and 52. Time points for peak concentration assessments taken 2 hours post-infusion were weeks 0, 2, 6, 22 and 46.

Statistical Analysis

First, we evaluated the relationship between CDST-defined probability groups, changes in HBI scores and measured VDZ concentrations using the entire 52-week GEMINI 2 study dataset (exposure-efficacy relationship). Differences in median concentrations at each time point among the three probability groups were first assessed using nonparametric testing (Kruskal-Wallis); pairwise comparisons were subsequently performed for each group at each time point. A closed testing procedure was used to control the overall type I error such that each of the pairwise comparisons was conducted at the 0.05 level with no P value adjustments if the initial omnibus hypotheses that all of the probability groups showed equal (a) mean HBI scores and (b) median measured VDZ concentrations were first rejected at the 0.05 level. If the omnibus comparison was not significant at the 5% level, subsequent pairwise comparisons were not performed.

Second, we re-validated the CDST in the GETAID cohort for predicting differences in week 14 remission rates between patients classified as low probability and intermediate-high probability. Intermediate- and high-probability patients were pooled in the GETAID cohort for comparison because of the low number of patients being classified as high probability (<10%) in this cohort. Week 14 was chosen for analysis because it is specified in US Food and Drug Administration (FDA) labelling as the most appropriate time for evaluation of the success of induction therapy. Furthermore, over 90% of the GETAID cohort had prior TNF-antagonist exposure, and prior subgroup analyses of GEMINI have observed that these patients require at least 10 weeks of exposure to observe meaningful differences in remission rates compared to placebo.[17] Secondary analyses were performed comparing changes in HBI over time and rates of clinical remission and corticosteroid-free remission at weeks 6, 14, 22 and 30. Sensitivity analyses were done limiting the analyses to patients receiving Q8 week VDZ maintenance, as European labelling allows for an additional dose to be given at week 10 in patients with a suboptimal induction response. Categorical data were compared using chi-square or Fisher's exact test.

We then assessed response to VDZ dose intensification in the GETAID cohort and VICTORY consortium according to the CDST-defined baseline probability of response (low vs intermediate-high) to confirm whether the exposure-efficacy relationship observed could be modified by higher predicted drug exposure. The decision to dose escalate was made clinically by treating providers without consideration for CDST-defined probability of response as the providers were unaware of how the different variables were used to generate a score and how that CDST score might classify a patient's probability of response. Our a priori hypothesis was that the low-probability and possibly the intermediate-probability groups would most likely benefit from an extra infusion at week 10 or interval shortening to Q4 or Q6 weeks given that these patients would have lower drug exposure than the high-probability group. In the GETAID cohort, response to interval shortening was assessed using pre- and post-interval shortening HBI scores. In the VICTORY consortium, response was assessed using the physician global assessment, with a clinically meaningful response defined as a >50% reduction in symptom activity post-interval shortening. Within-patient and within-group changes in HBI were assessed using repeated-measure analysis of variance with the group-time interaction function.

Finally, in our prior publication, we observed differences in week 26 endoscopic remission rates according to CDST strata. Using data from the most recent VICTORY consortium cohort database, we assessed differences in 52-week cumulative rates of endoscopic remission (absence of ulcers) across probability groups among patients undergoing endoscopic follow-up, and whether these differences in endoscopic remission corresponded to differences in rates of surgery between the high-probability group and the intermediate- or low-probability groups (exposure-efficacy-complication relationship). This relationship was initially assessed by groupwise and pairwise log-rank analyses and univariable Cox proportional hazard analyses. Adjustment for hazard ratio (HR) estimates was then performed for the covariates known to influence risk of surgery that were not already included in the baseline prediction model, including disease duration >2 years, ileal disease location, age >60 years, prior CD-related hospitalisation and smoking status.

Ethics Compliance Statement

VICTORY consortium and GETAID collaboration datasets were collected after ethics/IRB approval at all participating sites. GEMINI data were collected as part of the phase 3 clinical trial (NCT00783692) with corresponding ethics/IRB approval. All authors had access to the study data results and have reviewed and approved the final manuscript.