Validating robustness of HPLC related substances methods using statistical design of experiments

Abstract

In this study, robustness testing of five analytical related substances methods was carried out. These methods are routine analytical methods used to determine the amount of impurities in the active pharmaceutical ingredient or in the finished dosage form using high performance liquid chromatography (HPLC). Robustness testing was done by varying slightly, selected factors, such as mobile phase composition, flow rate, column temperature, injection volume and wavelength in narrow ranges. Factors were varied using the design of experiments (DOE) approach. Two-level screening designs namely fractional factorial and Plackett-Burman designs were utilized. The results obtained using both design types were also compared for Benazepril HCl methods. The experimental designs for each chosen method were created 11sing Minitab 15 statistical software and all runs were executed using HPLC Waters Alliance 2695 module equipped with a UVNIS 2487 detector. Robustness evaluation was done both graphically and statistically using Minitab 15 statistical software. For each method, factor effects were calculated and using regression analysis p-values, significant factors were identified. Significant factors were identified, statistically, using two different error estimation methods and also graphically. Robustness evaluation was based upon three chromatographic separatory responses namely principal peak capacity and tailing factors and upon the most critical resolution between two peaks. System suitability limits for each method were also derived theoretically using worst-case factor level combinations. On comparing both design types, it was found that less factors resulted as being significant using the PB design rather than the FF design. This may be due to the fact that the fractional factorial design is more resolved and no confounding effects are present to mask any significant effects. On comparing both error estimation methods, it was concluded that estimation of error is best done using negligible dummy factors in PB or using negligible two-factor interactions in fractional factorial designs rather than using Dong's Algorithm. Benazepril HCl isocratic method using PB design was found to be robust to capacity factor and tailing factor responses but not robust with respect to resolution. Trandolapril, Amlodipine Besylate and gradient Benazepril HCl methods were found to be robust to all three responses while, Mirtazapine method was not found robust with respect to capacity and tailing factor responses. Drift effects in this method might have contributed to some significant effects.