(W1130-12-78) Unlocking Residuals: Shelf-Life Predication for Oral Tablets Using Moisture-Modified Arrhenius Modeling

Purpose: This study was conducted to determine the shelf-life of two tablet formulations containing 25:75 COMPOUND X:HPMCAS-H amorphous solid dispersion (ASD). COMPOUND X is a poorly soluble candidate for an oncology indication. Poor pharmacokinetics resulting from insufficient absorption and exposure was observed with the crystalline form from Phase 1 studies in patients. Reformulation efforts led to the development of amorphous forms to mitigate the poor solubility observed with the active pharmaceutical ingredient (API). To support an initial clinical use period for the reformulated product, and expedite the time to submission, the Accelerated Stability Assessment Program (ASAP) was utilized. Tablet samples were placed on stability, and the concentration of each degradant in the stressed samples were modeled using the moisture-modified Arrhenius equation, ln⁡k=ln⁡A-E_a/RT+B(RH). The quality of the fit of the degradation data to the moisture-modified Arrhenius model is explored in this study by evaluating the residual error of each model.

Methods: Thirty stress conditions were selected, and the stressed samples were analyzed for impurities via HPLC, and these concentrations were modeled.

Results: Five degradant species were modeled for each tablet formulation for a total of 10 models. The residual error or residuals for each model are shown in Figures 1 and 2. As seen in the figures, some of the residuals do not span the zero-line: 50°C/ 81%,RH, 60°C/50%RH, 70°C/41%RH, 80°C/0%RH, and a slight trend is observed for the degradant species at RRT 0.78 in both the 5 mg and 20 mg tablets. The trend observed showed at low and high RH, the residual error was below the zero-line while middle values for RH were above the zero-line. Typically, the residuals should be randomly distributed across the zero line. Since the trend of higher residuals at low temperatures and smaller residuals at high temperatures was observed, there may be other factors affecting the degradation rate of the species at RRT 0.78 other than the model parameters. This trend could indicate a phase change, although this is not thought to be the case since it was only observed for RRT 0.78. Error bars should overlap the zero line and if they do not, the true uncertainty in the data may not be well represented by the model. Alternative fitting methods were explored with the same or worse results; therefore, the zero-order fitting method was used for all model fits.

Conclusion: This study examined the residual error for each model and possible reasons for not meeting the expected criteria of random distribution and spanning zero. Alternative fitting methods were explored with no improvement to the residual errors, and therefore zero-order kinetics, or the default, fitting method was used for all degradant species. The conditions that did not span zero were kept in the modeling as this criterion alone did not justify removing those conditions. The COMPOUND X tablets are predicted to pass the specification limit, i.e., having probabilities of passing greater than 95%, after 3-years at 25°C/60%RH in 30 cc HDPE/HIS bottles, 28 count with 0.5 g silica desiccant. In the packaging configuration simulations, it was observed that the addition of at least 0.5 g silica desiccant and a heat induction seal increased the probabilities of passing the specification limit and is recommended in the packaging configuration.

Model Residuals for the five degradant species modeled - RRT 0.74, RRT 0.78, RRT 0.98, RRT 1.12, and RRT 1.41 - for Tablet Formulation A.

Model residuals for the five degradant species modeled - RRT 0.74, RRT 0.78, RRT 0.98, RRT 1.12, and RRT 1.41 - for Tablet Formulation B.