Development of in vitro - in vivo correlations for newly optimized Nimesulide formulations

Use of the human volunteers in bioequivalence studies is being discouraged by the Food and drug administration after the introduction of biowaiver approaches. In-vitro in-vivo correlation (IVIVC) with the level A is accepted for the registration of new molecules. In the present study deconvolution technique with numeric approaches was applied after compressing and in vitro validating the 100mg Nimesulide immediate, intermediate and slow release tablets. Single centered, crossover, randomized study was conducted in four phases with a two-week washout period to obtain the plasma drug concentration data after administrating test and reference products in male healthy volunteers. KineticaTM 4.4.1 (Thermoelectron corp, USA) was used for the calculation of two ways ANOVA with 90% CI from both log transformed and non- transformed data and Phoenix WinNonlin 7 and it's IVIVC toolkit version 7.0 was used for the application of numeric approaches of IVIVC. Results revealed that the individual internal percentage prediction error for AUCinf and Cmax were found to be < 15% while their average values were < 10% in all medium. Numeric values of % PE at pH 6.8 and pH 7.4 (50 rpm in USP II and 100 rpm in USP I and II apparatus) were found to be (2.5842, 2.9789 and, 7.1732; 7.0944, 2.4721 and 4.350) for AUCinf and (2.5842, 0.5736 and 4.6928; 5.6214, 3.0551 and -2.4711) values for Cmax respectively. The low values of prediction errors demonstrate that the correlation model is projecting the in vivo response of each formulation. Percentage External error (% PE) was not required because individual values of percentage internal error (%PE) of Cmax and AUClast were not >15. In order to predict point to point correlation between fraction drug dissolved and drug absorbed, their mean r2 value was found to be > 0.9112 which showed a linear correlation in slightly alkaline pH.


Introduction
Bioequivalence studies are considered very much important for the establishment of new generic dosage forms. It is a significant method to estimate the in vivo performance of the compound which can be used as a surrogate to determine the therapeutic efficacy [1]. Due to the extensive availability of generic compounds locally in the market, the need for bioequivalence studies are much more required as compared to the past decade [2]. Use of the human being for the bioequivalence studies, cost of the study, subject to subject variation, unavailability of the expertise for the bioequivalence studies made them more complicated especially in case of generic product development. Pharmaceutical scientists have an urge to develop such techniques which are not only cost-effective but also useful for the establishment of generic products.
In vitro-in vivo correlations (IVIVC) plays a significant role in the product development and optimization process which is a very time consuming and costly procedure. Optimization of formulation requires modification in the composition of formulation, batch sizes and equipment and manufacturing procedures. If one or more such alterations are carried out in the formulation, then in vivo studies are required to be conducted to prove the comparison of the new formulation with a conventional product which will increase the total optimization cost and also increase the expenditure of carrying out multiple bioequivalence studies. In order to avoid these problems, IVIVC studies are carried out for the development of the pharmaceutical product [3,4]. Several studies demonstrating the utilization of in vitro dissolution assessment as success criteria for the prediction of the bioavailability studies [5]. IVIVC not only reduces the time and also minimizes the in vivo experiments, but also recommended for regulatory purposes. Scientists reported that successful correlation can be used as a surrogate for bioequivalence studies and to support the biowaiver studies. Such studies can be useful to develop suitable dissolution specifications [6]. IVIVC is also adequate for the rationalization of therapeutically significant drug release specifications of the formulations [7].
Following FDA guidelines, four levels of IVIVC i.e. Level A, B, C and multiple C were present. Correlation level depends upon the ability of the correlation to demonstrate the plasma level profile completely, which may due to the administration of the given dosage form. Pharmacokinetic studies are considered very significant in the development of innovator dosage form but found to be the most expensive task for the pharmaceutical company. Three different types of guidelines have been established by the Food and Drug Administration (FDA) for biowaiver studies of all types of drugs. First one is about the release pattern if the dosage forms show more than 85% release of drug within 30min it will be accepted without in vivo studies. The second one is the comparison of the prepared dosage form with the already established standard dosage form by similarity and dissimilarity factors. In vivo-in vitro (IVIVC) studies is the third type in which in vivo behavior can be predicted by using the in vitro data. BCS class II drugs categorized as a very potential candidate for IVIVC studies due to their in vitro profile as a rate-limiting step. Four levels i.e., A, B, C and multiple C of IVIVC can be developed after obtaining the in vitro and in vivo profile. Level A is very significant due to its point to point relationships of in vitro with in vivo data profile [8].
Level A showed a linear correlation and presented a point to point correlation between in vitro drug release studies and in vivo input rate. According to FDA guidelines, for the establishment of Level A correlation, formulations should be developed with altered drug release rates i.e. immediate (IR), intermediate (IntR) and slow release (SR) products or if drug release rate is condition independent then the single drug release rate is essential [9]. Level A correlation is of two-step method i.e. deconvolution step is followed by the assessment of the fraction of compound absorbed and dissolved. The detected fraction of drug absorbed is assessed by the numerical deconvolution procedure. The predicted fraction of drug absorbed is assessed using a detected fraction of drug dissolved. Now, these predicted drug absorbed values are used to estimate the predicted plasma concentrations by convolution method. The validity of the applied model is then assessed by computing the percentage prediction error (% PE) by comparing the difference between the predicted and observed values of several pharmacokinetic parameters i.e. C max & AUC 0-1 . Sirisuth and Edington in 2002 estimated IVIVC model for naproxen and metoprolol [10]. Macha et al in 2009 reported a level A correlation for nevirapine formulations using WinNonlin IVIVC Toolkit and found < 10% average %PE for different pharmacokinetic parameters [11].
Nimesulide is a selective COX-2 inhibitor and recommended for inflammation, pain, and fever [12]. Nimesulide is available in 100mg tablets, microcapsules and in controlled release formulations [13]. It belongs to BCS class II and showed less solubility and high permeability.
The aim of the study was to establish a Level A IVIVC to illustrate the relationship between in vitro release and in vivo behavior of different Nimesulide formulations immediate release (IR) intermediate release (InR) and slow release (SR) formulations. Phoenix WinNonlin 7 and its IVIVC toolkit version 7.0 was used for the establishment of correlation. Furthermore, the validity of the applied model was tested by computing the prediction errors.

Fast, medium and slow release tablets
Tablets of three different release fashions were planned by using the central composite rotatable design (CCRD). Nine formulations F-IR1, F-IR2 and F-IR3 from immediate release F-InR1, F-InR2 and F-InR3 from intermediate release and F-SR1, F-SR2, F-SR3 were from the slow release were optimized (already published) [14][15][16] due to their excellent preformulation results. All the formulations reported in Table 1 were in weights ranges from 400 ± 20 mg were compressed by direct compression with single punch machine under the controlled conditions of humidity and temperature.

Dissolution studies
Dissolution studies of marketed reference and compressed formulations F-IR1, F-IR2, F-IR3 from immediate release and F-InR1, F-InR2, F-InR3 from intermediate release F-SR1, F-SR2,  [17][18][19]. Six tablets of each formulation were placed in a 900 ml dissolution medium by setting the temperature limits 37 ± 0.5˚C at 100 rpm. Effect of surfactant was analyzed by using 1-3% sodium lauryl sulphate (SLS) after dissolving it in phosphate buffer of pH 7.4. An aliquot of 5 ml medium was taken out from vessels at different time intervals and an equal volume was replaced by fresh medium. Syringe filter of 45 μm was used for filtration process and drug concentration was calculated by UV spectrophotometer at 297 nm each experiment was repeated three times.

In vivo studies
In-vivo absorption studies were a single centered, crossover, randomized, in four phases with two week washout period in male healthy volunteers (age: 18-27 years) under the complete guidelines of FDA (www.fda.gov). Weight range of the volunteers was between ±10 percent of the ideal body weight. All physical examinations and medical examinations were within normal limits, Allergy history was also analyzed which was found negative. Those volunteers whose, weights and heights were not in normal range, their diagnostic tests failed in case of medical examination in a clinical situation, smokers like having more than 10 cigarettes daily and any other addiction like alcohol or volunteers on special diet user i.e., spicy, vegetarian, rich diet were excluded from the study. The study was conducted under the supervision of principal investigator and physician in a private hospital in Karachi after getting the ethical approval from Pharmacy, Ethics Committee, Bahauddin Zakariya University Multan. Volunteers were initially informed about the pros and cons of the study and a written consent was taken in this regard. For the bioequivalence study, optimized formulations from immediate release F-IR1 were compared with standard marketed brand "Nimaran 1 " 100mg (Bosch Pharmaceutical). Optimized intermediate F-InR2 and slow release F-SR3 formulations were used for the IVIVC considerations with deconvolution approaches. These formulations were given to subjects in a fasting condition with 250 ml of water at 8:00 am in the morning, who already having an overnight fast condition of around 10 hrs.

In vitro data analysis
Where n is the Number of samples, Rj and Tj are the Percentage release of reference and tests brands at different times respectively. Two formulations should be considered as similar if the f 2 value is more than 50.

In vivo data analysis
Pharmacokinetic parameters of reference (marketed research brand) and compressed F-IR1, F-InR2 and F-SR3 were calculated using Kinetica TM 4.4.1 (Thermoelectron, USA). Data was successfully fitted into oral two-compartment model and various other compartmental and noncompartmental parameters were calculated. Similarly, bioequivalence of F-IR1 and innovator products were carried out. Different bioequivalence attributes i.e., AUC last , AUC 0-1 , AUC tot , C maxcalc and T maxcalc of both reference and test products were assessed using two-way ANOVA methods. Schirmann's two-one sided t-test was applied for the verification of the bioequivalence assessment. Products bioequivalence were established by using 90% confidence interval (CI) values for reference and test products and the ratio of the values was targeted in the range of 0.8-1.25 for log-transformed and 0.8-1.20 for non-log transformed data respectively. For nonparametric assessment and carry-over effect, statistical software SPSS 20.0 (SPSS Inc.) was used.

Establishing in vitro-in vivo correlation (IVIVC)
Phoenix WinNonlin 7 and it's IVIVC toolkit version 7.0 (Certara USA, Inc., 100 Overlook Center, Suite 101, Princeton, NJ, USA) was used to estimate the absorption of Nimesulide. Invitro release evaluation of Nimesulide was conducted using the Weibull model and curve shape was evaluated by shape factor β. Implicit, numeric and analytical are three different deconvolution method approaches used in previous literature among them numeric method with point area was used for the IVIVC studies. For the establishment IVIVC of level A the percentage prediction error of C max and AUC was calculated.

Predictability error for IVIVC studies (Level A)
Internal (%PE) and external (%PE) prediction errors were used in the determination of IVIVC. Internal prediction for individual formulation was estimated by its C max and AUC values. Prediction error was used to assess the comparison between observed and predicted bioavailability. In the present study, external prediction error was not used because the values of internal % PE of AUC and C max were within the adequate limits. For the IVIVC predictability, the accepted limits for %PE for C max and AUC were 10 [9]. The %PE AUC and % PE Cmax can be calculated using the following equations:

In vitro dissolution studies
More than 85% of Nimesulide was released from F-IR1, F-IR2 and F-IR3 within 60 minutes. pH-dependent percentage of Nimesulide released was observed in the different dissolution medium. Comparatively less amount of drug release was observed at pH 1.2 and 4.5 medium, which was may be due to weak acidic nature and pKa values of HPMC K4M. It was increased pH 6.8 and 7.4, may be due to its dissociation and micelle formation after pH 6.5 of Nimesulide. A higher percentage of drug was released with 1% SLS already reported [8]. In F-InR1, F-InR2 and F-InR3 the average drug release were nearly 10% reduced in comparison to immediate release formulations. Reason may be due to concentration of HPMC which showed the controlled release behaviour when used in the range of 5-15% [21][22][23]. Hydrophilic polymer HPMC K4M showed inverse relationship with release rate of Nimesulide due to the presence of carboxylic group and swelling nature [21,24].   Table 2 showed the correlation (r 2 ) values of different release formulations in different dissolutions apparatuses and dissolution mediums.

In vivo pharmacokinetic studies
For the log-transformed values, the geometric mean ratio of C max calculated for reference and test was 0.995 μg/ml while geometric mean C max values were 6.087 ± 1.072 and 6.060 ± 1.073 μg/ml respectively. ANOVA results were found to be in significant range. P values of formulations and periods were insignificant whereas sequence and subject were found to be significant. reference product were statistically lies in the acceptable limits in both (log and non-log transformed) data. Bernareggi et al., in 1998 found the similar results after conducting bioequivalence studies on granules, tablets, and suspension [13].

Development of IVIVC correlations
Multiple formulations of F-IR1, F-InR2 and F-SR3 were designed and developed using HPMC as a rate controlling polymer. Data indicated that the as the concentration of HPMC reduces the drug release rate. In this study, in vitro release assessment were conducted at several dissolution media i.e. pH 1.2, 4.5, 6.8, 7.4; FaSSGF, FaSSIF and FeSSIF using USP dissolution Apparatus I (at 100 rpm) and Apparatus II (at 50 and 100 rpm). Fig 2 showed [25]. For AUC inf and C max , the %PE value of each formulation was found to be < 15% while the average values were < 10%. The low values of prediction errors demonstrate that the correlation model is projecting the in vivo response of each formulation. Thus indicated a valid Level A IVIVC according to the FDA guidance document [26]. External (%PE) was not required because individual values of internal (%PE) error of C max and AUC last were not >15. As compared with other nonlinear models, the Weibull model was considered to be the best fit for in vitro-in vivo data. Estimation of regression values (r 2 ) is significantly important for point to point correlation. Correlation (r 2 ) values after comparison of in-vivo absorbed vs. in-vitro drug dissolved values at different dissolution media were successfully calculated. Linear correlation found at pH 6.8 and 7.4 i.e. r 2 = 0.999. Jantratid et al in 2006 reported excellent value of correlation i.e. r 2 = 0.968 for cimetidine tablets to develop IVIVC model [27]. The reported results are quite similar to the Tandt et al., in 1995 reported the IVIVC correlation at phosphate buffer pH 6 [28]. Results of the present study indicated that IVIVC correlation was obtained at pH 6.8 and pH 7.4 showed good correlation Also, internal prediction error (%PE) of AUC inf and C max were calculated as presented in Table 3. In the previous literature, the fed state of gastric and intestinal medium also considered as the best for correlation [18,19]. Leu et al in 2008 established the IVIVC model for hemibenzathine and demonstrated that best medium was found to be fed state to predict AUC and C max showed < 10% prediction error [29].

Conclusion
In this study, a Level A IVIVC was established which demonstrate the relationship between in vivo absorption data and in vitro release data for all Nimesulide products. Results of internal validation were found to be within the adequate limits indicating the prediction of correlation models.