Zhengguo Xu (1, 2), Matilde Merino-Sanjuan (2, 3), Victor Mangas-Sanjuan* (2, 3), Alfredo GarcÃa-Arieta (4)
(1) Department of Pharmacokinetics, Towa Pharmaceutical Europe, S.L., PolÃgono Industrial de Martorelles, 08107, Barcelona, Spain. (2) Department of Pharmacy and Pharmaceutical Technology and Parasitology, University of Valencia, Valencia, Spain. (3) Interuniversity Research Institute for Molecular Recognition and Technological Development, Polytechnic University of Valencia–University of Valencia, Valencia, Spain. (4) División de FarmacologÃa y Evaluación ClÃnica, Departamento de Medicamentos de Uso Humano, Agencia Española de Medicamentos y Productos Sanitarios, Calle Campezo 1, Edificio 8, 28022, Madrid, Spain
Introduction/Objectives: Dissolution tests are essential in the development of medicinal products. Among the multiple methods available to compare dissolution profiles, the most widely used is the similarity factor f2. Due to various drawbacks associated with the similarity factor f2 that lead to several restrictions of its use as written in regulatory guidelines, alternative methods such as the model-independent multivariate statistical distance (MSD) method is recommended in situations where the f2 method is inappropriate [1–2]. However, recent studies showed that the MSD method is less discriminative and sensitive than the f2 method. Therefore, the confidence interval of f2 using bootstrap methodology has been recommended instead [3–5]. As neither details of the estimator nor the types of confidence intervals are described in the guidelines and the literature on this topic is scarce, the accuracy and precision of several estimators and types of confidence intervals were investigated in this study by simulation. In addition, the uncertainty associated with the current practice of using f2 point estimate alone for the statistical inference was evaluated.
Methods: Five estimators were studied: the estimated f2 (denoted as est.f2 in the abstract), calculated using sample means, the bias-corrected f2 (bc.f2), the variance- and bias-corrected f2 (vcbc.f2), the expected f2 (exp.f2), calculated based on the mathematical expectation of f2, and the variance-corrected expected f2 (vcexp.f2). Fourteen types of confidence intervals were investigated: the Normal interval, the basic interval, two bias-corrected and accelerated (BCa) intervals, and ten types of percentile intervals. There were 68 simulated populations of dissolution profiles in total: one group of 34 populations with low variability and another group of 34 with high variability. One million individual dissolution profiles were simulated for the reference population and for each of the 33 test populations with the predefined target population f2 values of 25, 35, 45, 46, …, 75, respectively. Random samples of size 6, 12, 18 or 24 units were chosen from the reference and the test populations without replacement. From each pair of random samples, five f2 estimators were calculated, and fourteen types of confidence intervals were obtained using 5000 bootstrap samples. Comparisons were performed with variability scenarios of R.LV versus T.LV, R.LV versus T.HV, and R.HV versus T.HV, where R and T denote the reference and test products, and LV and HV denote the low and high variability of the population profiles, respectively. The whole process was repeated 10000 times and the percentage of the similarity conclusions was measured.
Results: The estimators bc.f2 and vcbc.f2 are not suitable when the variability of the dissolution profiles is high. The precision of the estimators exp.f2 and vcexp.f2 is better than that of est.f2 and the type I errors were acceptable when they were combined with any of the ten percentile intervals. However, they have the drawback of low power, which might be addressed by increasing the sample size. The estimator est.f2 showed lower bias compared to exp.f2 and vcexp.f2, but the associated type I errors were higher than 5% when combined with different types of confidence intervals with the exception of the combination with the basic interval when the sample size is 24 units. The Normal interval is not suitable since it leads to type I errors higher than 10% when combined with any of the five estimators. With the exception of the combination with est.f2 for sample size of 24 units, the basic interval is also not suitable due to high Type I error rates. Likewise, the two BCa intervals showed higher-than-acceptable type I errors when combined with the five estimators. The sample size of 6 units is insufficient. To properly control the type I error, samples with at least 12 units should be used.
Conclusions: The best combinations of estimator and type of confidence interval are exp.f2 and vcexp.f2 combined with any of the ten types of percentile intervals. The type I error rates of those combinations are all acceptable. When the sample f2 value is close to 50, the use of the confidence interval of f2 is recommended even when the variability of the dissolution profiles is low and the prerequisites defined in the regulatory guidelines for using the conventional f2 method are fulfilled in order to control the type I error rate.
References:
[1] European Medicines Agency, Guideline on the investigation of bioequivalence (2010).
[2] U.S. FDA, Guidance for Industry. Dissolution testing of immediate release solid oral dosage forms (1997).
[3] B. M. Davit, E. Stier, X. Jiang, O. Anand, Expectations of the US-FDA regarding dissolution data in generic drug regulatory submissions, Biopharma Asia 2 (2) (2013).
[4] V. Mangas-Sanjuan, S. Colon-Useche, I. Gonzalez-Alvarez, M. Bermejo, A. García-Arieta, Assessment of the regulatory methods for the comparison of highly variable dissolution profiles., AAPS J. 18 (6) (2016) 1550–1561. doi:10.1208/s12248-016-9971-5.
[5] P. Paixão, L. F. Gouveia, N. Silva, J. A. G. Morais, Evaluation of dissolution profile similarity—comparison between the f2, the multivariate statistical distance and the f2 bootstrapping methods., Eur. J. Pharm. Biopharm. 112 (2017) 67–74. doi:10.1016/j.ejpb.2016.10.026.
Reference: PAGE 29 (2021) Abstr 9721 [www.page-meeting.org/?abstract=9721]
Poster: Methodology - Other topics