Salvatore D'Agate (1), Timothy Wilson (2), Chandrashekhar Chavan (3), Burkay Adalig (4), Michael Manyak (5), Juan Manuel Palacios-Moreno (6), Oscar Della Pasqua (7)
(1) Clinical Pharmacology & Therapeutics Group, University College London, UK; (2) Clinical Statistics, Parexel, USA; (3) Global Medical Urology, GSK, India; (4) Global Medical Urology, GSK, Turkey; (5) Global Medical Urology GSK, USA; (6) Global Medical Urology GSK, Spain; (7) Clinical Pharmacology Modelling & Simulation, GSK, UK
Objectives: Lower Urinary Tract Symptoms (LUTS) caused by Benign Prostatic Hyperplasia (BPH) are typically measured by the International Prostate Symptom Score validated questionnaire. The IPSS provides information on LUTS improvement as well as deterioration, the last being the key outcome of BPH progression. While most of the available evidence refers to mean or median changes, little attention has been given to individual IPSS trajectories [1,2]. The current investigation aimed to develop and validate a drug-disease model describing individual IPSS trajectories in moderate and severe LUTS/BPH patients receiving placebo, dutasteride, tamsulosin or combination treatment. Clinical trial simulations were then performed to demonstrate the impact of the underlying disease progression and covariate factors on the deterioration of symptoms, disentangling it from the drug effects.
Methods: A meta-analytical approach was used including pooled data from moderate and severe LUTS/BPH patients (N=10238) enrolled into six clinical studies [3-6]. For consistency and standardization purposes, baseline measurements were defined as those collected on the last day of the placebo run-in phase. To ensure that patient and disease-specific fluctuations in IPSS trajectory were disentangled from treatment-specific changes, individuals treated with placebo only were used for model development. Watchful waiting was considered as a non-pharmacological intervention and handled as an active treatment arm. A nonlinear mixed effects model was developed using NONMEM v7.2, in which disease specific characteristics were parameterised independently from drug effects; consequently, treatment response was then evaluated as a covariate effect on the underlying disease model parameters. Standard graphical and statistical methods were used for model building, covariate selection and evaluation. A sensitivity analysis was subsequently performed using the final model to investigate how individual parameter estimates, treatment and covariate factors affect individual IPSS trajectories. Predictive performance was assessed using internal and external validation procedures.
Individual IPSS vs time profiles were simulated for a virtual BPH population including baseline characteristics resampled from the original clinical studies. Mean profiles along with 95%-CI were used to illustrate the effect of different treatments in patients with varying rates of disease progression. Results were summarised based on the predicted absolute change in IPSS at predefined visits. For the purposes of these simulations, clinical response was defined as improvement in IPSS ≥25% relative to baseline.
Results: Despite considerable noise and interindividual variability in IPSS measurements, improvement and deterioration of IPSS was characterized by a Gompertz function. Covariate factors (namely, baseline IPSS, BMI, duration of symptoms and alcohol user status) were found to affect both the parameters describing the disease progression and placebo effect. Interindividual variability was identified for the parameters describing the zero-order disease progression parameter as well as the magnitude and half-life of placebo effect. Residual variability in IPSS was described by a proportional and additive error model.
Exploratory simulations showed that treatment response depends on the underlying disease progression rate and baseline IPSS, making it difficult to distinguish the contribution of each factor to the response. Our results also indicate that individual IPSS trajectories can be predicted over the course of treatment and may be clustered according to the underlying rate of disease progression (i.e., fast, moderate, slow progressing phenotypes). No independent prognostic baseline factor could be identified that can be used as a predictor of response or the time course of symptoms.
Conclusion: Individual IPSS trajectories can be characterised by a longitudinal model. In addition to the identification of baseline covariates, the use of a longitudinal model enables characterization of the initial improvement followed by slowly progressive changes in IPSS. Initial simulations shed further light into the factors that explain interindividual differences in the rate of disease progression and consequently in the deterioration of symptoms.
This investigation was sponsored by GlaxoSmithKline.
References:
[1] Barry MJ, et al. J Urol 1992; 148(5):1549-64.
[2] Verhamme KMC, et al. Eur Urol 2002; 42: 323–328.
[3] Roehrborn CG, et al. Eur Urol. 2010; 57(1): 123-131.
[4] Roehrborn CG, et al. BJU Int. 2015; 116(3): 450-459.
[5] Barkin J, et al. Eur Urol 2003; 44: 461–6.
[6] Roehrborn CG, et al. BJU Int. 2005; 96(4): 572-577.
Reference: PAGE 27 (2018) Abstr 8794 [www.page-meeting.org/?abstract=8794]
Poster: Drug/Disease Modelling - Other Topics