Quantitative assessment of nocturia patient’s urination pattern and drug treatment effects using a joint ordered categorical model

Abstract

Nocturia occurs in about 70% of the elderly, lowering the quality of life seriously. However, due to the lack of understanding of the disease and objective tool, drug effects have been evaluated based on empirical bases. To develop a model for quantitative analysis of drug effect assessments given routine clinical data and for characterizing the diurnal pattern of nocturia patients’ main symptoms. This was a routine clinical study conducted in 20 male Korean outpatients with severe nocturia (≥ 3 urinations /night) over 3 periods, where each patient was evaluated before treatment (Period 1), after 1 month of mono-therapy of tamsulosin (Period 2), and after 3 months of combination therapy of tamsulosin and solifenacin (Period 3). Outcome measurements were the urination frequency (FREQ) and the average urine volume per void (FBC, functional bladder capacity) every 2-hour within a 24-hour interval, taken from Frequency Volume Chart (FVC) recorded by study patients each period. FREQ and FBC were analyzed using the joint ordered categorical model after being categorized. The final model was selected by maximizing the joint likelihood of the two categorical variables using NONMEM 7.2.Each variable was divided into 3 categories; for FREQ, category 0, 1, 2 for 0, 1, 2 urinations, respectively, and for FBC, category 1, 2, 3 for <100, 100-200, ≥200 mL, respectively. The baseline at period 1 for FREQ was best described in the logit scale by a constant (category ≥ 1), decreased by a circadian function with 24-hr period (category = 2), whereas that for FBC by a circadian one with 12-hr period (category ≥ 2), decreased by a constant (category = 3). As for drug effect, for FREQ it was best described in the logit scale by an exponential decay with half-life of 3.2 hour (period 2) plus an additional constant decrease of 0.29 (period 3) added to the baseline, whereas for

FBC, it was negligible at period 2 but increased at period 3 by 36% from the baseline mesor. When drug effect was assessed by symptom improvements in nocturia, the model prediction for the probability of no-urination category at nighttime increased from 17% up to 45% and 52% in period 2 and 3, respectively, while that of large FBC category increased from 26% up to 41% in period 3, where the latter result of the improvement on FBC was unable to be found by the p-value based approach conventionally used. The new finding identified from this model was solifenacin has significant effect on decreasing nocturia frequency, which is liable to be hidden with the conventionally used analysis method and causes controversy on the use of this drug on nocturia patients. The other results of the two drugs’ treatment were identical with the previously studies ones. One limitation with the current work is that no demographics information was included in the model, resulting in the model with no prediction ability. This limitation, however, will be overcome by a prospective study that we are conducted in a larger patient population receiving a different amount of dose each other. Nevertheless, we expect that the developed model can be used as a supportive tool in the treatment of nocturia. These results demonstrate the feasibility of applying the proposed method to quantitative understanding of nocturia characteristics and objective assessment of drug effects.