Guillaume Bonnefois (1), Raphaël Vlavonou (1), Pierre-Olivier Tremblay (1), Mario Tanguay (1, 2)
(1) Syneos Health Clinique, Montreal, Canada, (2) Faculty of Pharmacy, Université de Montréal, Canada.
Introduction: Most of drugs that are likely to be administered to patients with renal impairment should be investigated in this population to assess the effects of this condition on the pharmacokinetics (PK) and to provide appropriate dosing recommendations, if necessary.
These PK studies would generally include the enrollment of control subjects who should match the renal patients with respect to age, gender, race, weight or body mass index (BMI). However, based on existing regulatory guidance documents and literature, there is no well-established “matching” methodology that would ensure appropriate comparability in terms of demographics, while taking into considerations the recruitment challenges for the control subjects [1-4].
Various strategies are applied for the matching procedure [5]: one approach, referred to as “mean matching”, would consist in recruiting a single cohort of control subjects about equal size of each renal impairment cohort that would match the mean characteristics of the patients. A flexibility is usually allowed for age and BMI parameters, but these limits are somewhat arbitrary (e.g., mean age ±10 years, mean BMI ±20%). Another approach consists in a “one-to-one pairing” of the prospective control subject to the renal patient. Again, certain flexibility may be allowed for age and BMI parameters. In that case, each category of renal impairment (e.g., mild, moderate and severe) would have its own control [5].
One of the current challenges with the above methods is to ensure a similar distribution of demographics between patient cohorts and control subjects.
Objectives: The objectives of the work were to address this challenge by:
- Applying statistical concepts to propose a more robust and quantitative matching approach that would limit bias in the PK comparison;
- Developing a new computational platform to guide clinicians and to facilitate the selection of the control subjects.
Methods: This work relies on the demographic characteristics’ distribution of the patient cohorts. For the mean matching approach, a three-step process was developed. In the first step, the normality of demographics of each patient group was assessed. In the second step, the mean and standard deviation (SD) were calculated and the statistical differences between the means of the three patient groups were investigated using the Levene and Anova tests. A non-significant statistically difference implied pooling the three groups together. Otherwise, the control group should be matched according to the three separate patient groups. Using an empirical rule, i.e. proportion of patients within 1-, 2-, and 3 SD of the mean or 68.27%, 95%, 99%, respectively, the third step enabled to well distribute the variables of the control group to attain similar patient distributions.
In the one-to-one pairing strategy, a two-step iterative process was established. For each demographic characteristics and each renal impaired group, the empirical distribution was firstly estimated by a non-parametric method: the kernel density estimation [6,7]. Thereafter, this empirical probability density function was divided in two or more parts according to the total number of renal impaired patients. The corresponding intervals and densities were calculated to obtain the number of control subjects within each interval. These methods were then implemented into a web-site application, which was developed using R-Shiny [8, 9].
Results: A computational approach was developed to implement both updated matching strategies through the Shiny application. Demographics variables can be tested such as age, BMI, or gender to develop an integrated tool that would be convenient for clinicians. The creatinine clearance or the glomerular filtration rate of each patient can be calculated (e.g. using Cockcroft-Gault, Modification of Diet in Renal Disease, or other equations) to automatically classify this patient within the three patient cohorts, if necessary.
Conclusions: An R-Shiny application was developed through a user-friendly interface. This tool provides a quantitative assessment of control subjects to facilitate and guide the selection of these subjects. The tool may require further validation to confirm its practical application to a renal impairment studies using real patient datasets.
References:
[1] Guidance for Industry “Pharmacokinetics in Patients with Impaired Renal Function”, FDA, March 2010, Available at: https://www.fda.gov/downloads/drugs/guidances/ucm204959.pdf Accessed on February 27th, 2019.
[2] Guidance for Industry “Pharmacokinetics in Patients with Impaired Hepatic Function”, FDA, May 2003. Available at: https://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm072123.pdf. Accessed on February 27th, 2019.
[3] Guideline on the evaluation of the pharmacokinetics of medicinal products in patients with impaired hepatic function”, EMA, August 2005. Available at: https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-evaluation-pharmacokinetics-medicinal-products-patients-impaired-hepatic-function_en.pdf . Accessed on February 27th, 2019.
[4] Guideline on the evaluation of the pharmacokinetics of medicinal products in patients with decreased renal function”, EMA, August 2014. Available at: https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-evaluation-pharmacokinetics-medicinal-products-patients-decreased-renal-function_en.pdf. Accessed on February 27th, 2019.
[5] Paglialunga S et al. Update and trends on pharmacokinetic studies in patients with impaired renal function: practical insight into application of the FDA and EMA guidelines. Expert Review of Clinical Pharmacology. 2017 10(3): 273–283
[6] Rosenblatt, M (1956). “Remarks on Some Nonparametric Estimates of a Density Function”. The Annals of Mathematical Statistics. 27 (3): 832–837.
[7] Parzen, E (1962). “On Estimation of a Probability Density Function and Mode”. The Annals of Mathematical Statistics. 33 (3): 1065–1076.
[8] Chang W et al (2018). Shiny: Web Application Framework for R. R package, version 1.2.0. https://CRAN.R-project.org/package=shiny
[9] R Core Team (2018). R: A language and environment for statistical computing, version 3.5.0. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/
Reference: PAGE 28 (2019) Abstr 8999 [www.page-meeting.org/?abstract=8999]
Poster: Methodology - Other topics