CT-FAST Stochastic is based on object-oriented modeling of a discrete event system and Monte Carlo simulations. The simulation process is divided into two major parts:
The treatment process, which aims at forecasting package usage at the site level
The supply process, which evaluates the ability of the supply strategy to meet the usage forecast at the desired service level (i.e. with a very low and controlled risk of running out of drug).
 |
 |
 |
 |
 |
The treatment process model simulates patient enrollment at the investigational sites and the progress of each patient through their assigned treatment. The goal is to generate the medication usage forecast on a day-to-day basis along with an estimate of the uncertainty. The model includes all stochastic parameters using random variables with adequate statistical distributions:
Site opening, Enrollment rate: enrollment rate can vary from site to site and over time.
Stratification Randomization: patient randomization is modeled according to the randomization scheme (randomization level, ratio and block size). Visit window
Variable factor: for some trials, the kits dispensed to a patient at a given visit and dose level may depend on a variable factor (e.g. weight of the patient for pediatric studies in neuroscience). Variable factor ratios may vary from country to country.
Dose titration: titration is modeled as a random process with a probability attached to each possible dose level transition. Such transitions can be numerous in a trial and the a-priori quantification of the probability of each transition is generally not possible. In order to address this, the tool provides a convenient way to enter information about titration probabilities in a qualitative manner (i.e. low / medium / high probability), which is more manageable by the user. The tool then automatically performs the conversion of this information, combined with overall probabilities also entered by the user, into detailed figures used in the simulations.
Discontinuation/dropout: the probability of discontinuing the trial may be defined over each trial phase (visits comprised in trial phase are defined by the user). Discontinuation may either lead to an abrupt dropout or to a follow-up phase with a different probability defined in each case.
The supply process model simulates the day-to-day progress of medication kits through the distribution network. The network is represented with up to three layers: sites may be delivered either through a local depot or directly from a distribution center. An inventory management rule is attached to each node (warehouse) of the network. Two types of IVRS (re)supply rules are modeled: trigger-based and predictive.
The objective of this part is to optimize the inventory management rules, in order to determine the best balance between the overage and the risk of shortage. Several scenarios are investigated with different trigger levels, forecasting horizons and overages.
In practice, the scenarios are evaluated using a number of indicators such as the probability of shortage, the number and frequency of shipments and the portion of packages wasted. These indicators are presented to the user for each scenario in order to select the optimum one. Several production schedules may also be analyzed.
