Sang-In Park, Dong-Wha Kang(2,3), Seok Kyu Yoon(1), Hyungsub Kim(1), Kwan Cheol Pak(2), Hyeong-Seok Lim(1,2)
(1) Department of Clinical Pharmacology and Therapeutics, Asan Medical Center, Seoul, Korea, (2) University of Ulsan College of Medicine, Seoul, Korea (3) Department of Neurology, Asan Medical Center
Objectives: Stroke is a main cause of long-term disability producing considerable socioeconomic burdens worldwide. The prognosis of stroke varies significantly between individual patients. The identification of predictors for the prognosis of patients with acute stroke and their quantitative predictability is vital to understanding disease progression and would enable the development of rational treatment strategies for patients. Prediction of the time course of the degree of disability in acute stroke patients may also be helpful for making proper management plan. The current modeling and analysis study analyzed the longitudinal degree of disability data expressed in modified Rankin scale (mRS) which were collected from patients with acute ischemic stroke in Korea. Through this analysis, we tried to identify the predictors for the disabilities of the patients including the changes of lesion volumes in magnetic resonance imaging (MRI) early after onset of the disease.
Methods: A longitudinal data obtained from 193 patients for 24 months after the onset of acute ischemic stroke were used for the analysis. The patients were enrolled from those whose diseases were confirmed via initial magnetic resonance imaging (MRI) within 24 hours after onset of the symptoms. Stroke lesion volumes were measured at the initial stage (initial lesion volume, ILV) and at follow-up (5 ± 1 days) MRI (follow-up lesion volume, FLV). Degree of disability was classified into 6 categories based on the mRS as follows:
- Grade 0: no symptoms at all
- Grade 1: no significant disability despite symptoms
- Grade 2: slight disability
- Grade 3: moderate disability
- Grade 4: moderately severe disability
- Grade 5: severe disability
- Grade 6: dead
The mRS of each patient was assessed at scheduled times: 0, 1, 3, 6, 9, 12, 15, 18, 21, and 24 months after the onset of acute stroke. The grade 5 and 6 data are combined in this analysis.
A proportional odds cumulative logit model with disease progression model was implemented in NONMEM software (version 7.4, Icon Development Solution, Hanover, MD, USA).
Various structural models for disease progression, including exponential, weibull, log-logistic models were investigated. Difference in lesion volume (DLV in cm3), calculated as FLV – ILV, ILV, FLV, age, sex, baseline NIHSS scores, Trial of Org 10172 in Acute Stroke Treatment stroke subtype (TOAST), diabetes mellitus, smoking, hyperlipidemia, previous stroke, hypertension were tested with regard to whether they improve the model significantly or not. Graphical assessment of optimum fit and statistical significance criteria. Stepwise forward selection and backward elimination (p ≤ 0.01, p ≤ 0.005).
Results: Inclusion of disease progression model in the baseline proportional odds cumulative logit model in the form significantly increase the model fit. Disease progression model in the form of an exponential decay model well described the time course of the probability of mRS, the half-life of which was estimated as 13.6 months. Inclusion of TDLV in the baseline proportional odds cumulative logit model in the form of a simple maximum effect model best described the relationship between TDLV and the logit probabilities over time of mRS scores, and improved the model fit significantly, and Emax and ED50 were estimated as 4.6 and 46.7 cm3. The mean baseline logit probability for each mRS score (0, 1, 2, 3, 4, 5/6) were 3.72, -2.76, -4.08, -2.99, and -2.87, respectively.
In visual predictive check plots, the final model predicted the observed probabilities of each score of mRS over time reasonably well.
Conclusions: The model developed in this study well described the probability of the degree of disability according to time. This modeling result can be applied to predict the functional outcome over time after a stroke in order to prepare a proper management plan in acute stroke patients.
References:
[1] Lim, H. S., Kim, S. M., & Kang, D. W. (2018). Quantitative Predictive Models for the Degree of Disability After Acute Ischemic Stroke. The Journal of Clinical Pharmacology, 58(4), 549-557.
[2] Cook, S. F., & Bies, R. R. (2016). Disease progression modeling: key concepts and recent developments. Current pharmacology reports, 2(5), 221-230.
Reference: PAGE () Abstr 9473 [www.page-meeting.org/?abstract=9473]
Poster: Drug/Disease Modelling - CNS