Lead Investigator: Siddharth Singh, University of California San Diego
Title of Proposal Research: Predicting the short- and long-term risk of serious infections in patients with Crohn’s disease: Analysis of the PYRAMID Registry
Vivli Data Request: 6551
Funding Source: NIDDK K23DK117058; I will also be applying for NIH funding, through NIDDK PA-18-741
Potential Conflicts of Interest: Dr. Singh reports: Research grant from Janssen and AbbVie
Personal fees from Pfizer for ad-hoc fellowship grant review.
COI Management:
Research grant from AbbVie: The lead PI is not receiving any salary or consulting support from AbbVie, for this independent investigator-initiated project. The project being supported by the grant is independent of the current proposal.
Research grant from Janssen: The lead PI is not receiving any salary or consulting support from Janssen. The project being supported by this grant has been completed in July 2020.
Personal fees from Pfizer: This fees is not for consulting or advising Pfizer. Instead, this is for ad hoc grant review, for research and fellowship grants administered by Pfizer. Pfizer does not have any input on these grants, which are managed by grant review panels.
Summary of the Proposed Research:
The overall objectives of this proposal are to accurately predict the short- and long-term risk of serious infections with medications used in the treatment of patients with Crohn’s disease (CD). We hypothesize that pre-treatment clinical factors can identify patients at risk of infections when starting adalimumab (a leading biologic agent used to treat Crohn’s disease) in the short-term (first 6 months). Once patients are on treatment, and disease is better controlled, the risk of serious infection may decrease, by controlling disease effectively and decreasing the need for corticosteroids and/or opiate pain medications, all of which make patients more susceptible to infections. We will also look to predict the long-term risk of infection once patients are on medication and responding to it. This information will provide patients and providers knowledge to tailor therapy based on patients’ predicted risk of treatment-related complications like serious infections.
Statistical Analysis Plan:
Statistical Analysis Approach:
Aim #1 – We will calculate unadjusted incidence rate (per 100 PY) of outcomes in adalimumab-treated patients, in year 1, year 2, and year 3. We will graphically evaluate the time-dependent change in corticosteroid use and time-dependent change in rate of remission (assessed based on PGA, and SIBDQ), by examining responses gathered at predetermined intervals.
Aim #2 – Model derivation and validation will be reported in accordance with the TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) statement. We will first calculate the short-term risk of serious infections overall. (A) Model derivation: Consistent with prior approach, we will develop a multivariable logistic regression prediction model with short-term risk of serious infections as the dependent variable. Baseline variables with p-value <0.15 on univariable analyses will be included with backward variable selection, after assessment for collinearity, clinical importance, and interpretability. Interaction terms will be assessed individually and included in the final model if they have a p-value <0.10 on multivariable analyses. (B) Model validation: The model will be internally validated using the optimism-bootstrap resampling technique with 300 replications to determine reproducibility. We will report an optimism-corrected AUC to account for apparent over-fitting during model fitting. Bootstrapping allows the use of entire dataset for model derivation, and provides reliable estimates compared with other validation techniques such as split-sample validation analyses (pessimistic estimates of performance, large variability). (C) Model performance: The predictive accuracy of the model will assessed using: 1) discrimination, measured by area under the receiver operator curve (AUC); 2) prediction error, measured by Nagelkerke’s R-squared and the Brier score; and 3) calibration, assessed by the Hosmer-Lemeshow goodness-of-fit test. (D) Model transformation into a clinical decision support tool (CDST): For ease of risk estimation, we will transform the model into a CDST by converting coefficients from the multivariable model to a points system as described by Sullivan et al. Based on quartiles of predicted probability of serious infections within 6m of starting therapy, we will determine cut points for low- (quartile 1), intermediate- (Q2-3), and high-risk groups (Q4). We will calculate the sensitivity, specificity, positive and negative predictive value, for the high and low risk groups.
Aim #3 – (I) We will estimate on-treatment evolution of corticosteroid and opiate use, both in the short-term (within 6m) and long-term (>6m) after starting adalimumab, measured both as proportion and cumulative burden of use (in mg equivalents if reported and extractable). (II) To estimate the relative contribution of treatment effect on long-term risk of serious infections, we will apply two different statistical models to calculate adjusted incidence rate ratio: the first model will only adjust for pre-treatment baseline risk factors (model A), and the second model will additionally adjust for on-treatment time-fixed and time-varying risk factors (model B). The difference in the estimates of the trend parameter between models A and B will provide the relative contribution of treatment effect to long-term risk of serious infections. (III) Finally, we will derive a multivariable Cox proportional hazard prediction model with long-term risk of serious infections on adalimumab, as the dependent variable, based on pre-treatment baseline, on-treatment time-fixed and time-varying risk factors. Because multiple infections in individual patients are not independent of each other, we will account for intra-patient correlations using generalized estimating equations. The model derivation, validation, performance evaluation and transformation in CDST will follow similar steps as in Aim #2.
Requested Studies:
An Open-Label, Multicenter, Efficacy and Safety Study to Evaluate Two Treatment Algorithms in Subjects With Moderate to Severe Crohn’s Disease
Data Contributor: AbbVie
Study ID: NCT01235689
Sponsor ID: M11-271
A Long-Term Non-Interventional Registry to Assess Safety and Effectiveness of Humira® (Adalimumab) in Subjects with Moderately to Severely Active Crohn’s Disease (CD)
Data Contributor: AbbVie
Study ID: NCT00524537
Sponsor ID: NCT00524537
Public Disclosures:
- Ahuja, D., Luo, J., Qi, Y., Syal, G., Boland, B.S., Chang, J., Ma, C., Jairath, V., Xu, R. and Singh, S., 2024. Impact of Treatment Response on Risk of Serious Infections in Patients with Crohn’s Disease: Secondary Analysis of the PYRAMID Registry. Clinical Gastroenterology and Hepatology. Doi: 10.1016/j.cgh.2024.01.003
- Ahuja, D., Luo, J., Qi, Y., Syal, G., Boland, B., Chang, J., Ma, C., Jairath, V., Xu, R. and Singh, S., 2024. IMPACT OF TREATMENT RESPONSE WITH ADALIMUMAB ON RISK OF SERIOUS INFECTIONS IN PATIENTS WITH CROHN’S DISEASE: SECONDARY ANALYSIS OF THE PYRAMID REGISTRY. Inflammatory Bowel Diseases, S1(30), p.S1. Doi: 10.1093/ibd/izae020.001