,Jinping Liu,＊,Junlin Wng,Yi Zho

aDepartment of Pharmaceutics,China Pharmaceutical University,Nanjing 210009,China

bJiangsu Aosaikang Pharmaceutical Co.,Ltd.,Nanjing 210009,China

A quality by design(QbD)case study on entericcoated pellets:Screening of critical variables and establishment of design space at laboratory scale

Shuling Kana,b,Jing Lua,Jianping Liua,＊,Junlin Wanga,Yi Zhaoa

aDepartment of Pharmaceutics,China Pharmaceutical University,Nanjing 210009,China

bJiangsu Aosaikang Pharmaceutical Co.,Ltd.,Nanjing 210009,China

A R T I C L E I N F O

Article history:

Received 27 January 2014

Received in revised form

7 July 2014

Accepted 11 July 2014

Available online 28 August 2014

Fluid-bed coating

Plackett-Burman design

Box-Behnken design

Naproxen

Quality by design

The study aims to prepare naproxen enteric-coated pellets(NAP-ECPs)by f l uid-bed coating using QbD principle.Risk assessment was f i rstly performed by using failure mode and effect analysis(FMEA)methodology.A Plackett-Burman design was then used for assessment of the most important variables affecting enteric-coated pellets characteristics.A Box-Behnken design was subsequently used for investigating the main,interactive, and quadratic effects of these variables on the response.By FMEA we discovered that eight factors should be considered to be high/important risk variables as compared with others. The responses of acid resistance and cumulative drug release were taken as critical quality attributes(CQAs).Pareto ranking analyses indicated that the coating weight gain(X7), triethyl citrate percentage(X1)and glycerol monostearate percentage(X2)were the most signif i cant factors affecting the selected responses out of the eight high-risk variables. Optimization with response surface method(RSM)further fully clarif i ed the relationship between X7,X1,X2and CQAs,and design space was established based on the constraints set on the responses.Due to the extreme coincidence of the predicted value generated by model with the observed value,the accuracy and robustness of the model were conf i rmed. It could be concluded that a promising NAP-ECPs was successfully designed using QbD approach in a laboratory scale.

? 2014 Shenyang Pharmaceutical University.Production and hosting by Elsevier B.V.All rights reserved.

1. Introduction

Pharmaceutical development involves traditional and systematic approaches.The former mainly depends on empirical evaluation of product and process performance,which pays little attention to the impact of multivariable on formulation and the process understanding.Product quality is identif i ed mainly by restricting f l exibility in the manufacturing process and end product testing(so called quality-by-testing,QbT)[1].There exist some lags to ensure product quality during the pharmaceutical development.The latter is pharmaceutical Quality by design(QbD)philosophy,which is a more scientif i c, risk-based,holistic and proactive approach with proper feedforward and feed-back control strategies to pharmaceutical development.Theapproach builds qualityinto productduring the pharmaceutical development but not merely testing for it [2],which helps to thoroughly understand the root-cause, such as critical material attributes and process parameters that impacting the predef i ned quality attributes[3].The following equation clearly illustrates what affects the product quality:Pharmaceutical Quality=F(drug substance,excipients,manufacturing,packaging,et al).The function F in the equation contributes to understand how the formulation variables and process parameters inf l uence the end product prof i le[3],hence developing accurate and robust product. Building from the QbD paradigm,methods can be used upstream at the beginning stages of the research,development and design phases[4],meanwhile the product quality should be proactively controlled in the manufacturing process.By developing a product with QbD principles,end product testing would be only used for the conf i rmation of product quality[1].

QbD is concerned with the certain predictable quality through linking the critical material attributes(CMAs)and critical process parameters(CPPs)into the critical quality attributes(CQAs)of drug product.The application of QbD concept in pharmaceutical development is presented graphically in Fig.1.Firstly,the potential risk factors are determined by risk assessment in the initial design during product development.Then,to improve process knowledge,multivariate experiments are carried out using design of experiments(DOE).DOE method can link the inputs to the outputs, as such,the relationships between CPPs and CQAs are well understood in mathematical form and the design space(DS)is further established[5].The most common used of DOE is Plackett-Burman,which can quickly screen the main factors among numerous inputs variables[6].However,the disadvantage of Plackett-Burman design is that interactions between variables are generally confounded and cannot be easily determined,as there are not enough degrees of freedom.Box-Behnken design were usually used to establish DS for it can determine the variables range which requires fewer runs than a central composite design[7].

Solid dosage form owns a majority share in the market and pellets are achieving increasing attention as multiple units preparations for possessingremarkable advantages,including less effect by gastric emptying rhythm,homogeneous distribution in gastrointestinal(GI)tract thus maximizing drug absorption,reducing the risk of local GI tract irritation[8,9]. Film coating processes are widely used due to its outstanding functions in oral solid drug delivery system,such as masking unpleasant taste,improving stability,enhancing appearance, adding an active compound and controlling release rate[10]. Based on the above,the coating pellets dosage form is compelling and desirable,besides,and the f l uid-bed technique is a valuable approach to obtain coating pellets.Entericcoating is the most common method for manufacturing oral solid preparation especially when the drug acid stability/ dissolution or irritation to gastric mucosa is an issue.The coating pellets process consists of two phases:f i rstly,the pellets core containing drug should be obtained,and then the pellets are coated with enteric-coating materials.To date,the applicationofQbDforpharmaceuticaldevelopmentof enteric-coated pellets has yet been reported.Naproxen(NAP),a non-steroidal anti-inf l ammatory drug(NSAID),the solubility of which has a positive association of pH,meanwhile along with local gastric irritation,is used as a model compound to develop Naproxen enteric-coated pellets(NAP-ECPs).In previous work,we had successfully prepared NAP-loaded immediate-release pellets(NAP-IRPs)starter cores by extrusionspheronization by usingQbD concept(the relatedcontents are to be published).

Fig.1-QbD,risk management and quality management in formulation development.

The ultimate purpose of our study is to comprehensively link the CQAs,variables among DOE batches,to upstream manufacturing process and material attributes so as to identify factors affecting the CQAs.The elements in QbD are fourfold:(1)a risk assessment was performed to identify the main variables inf l uencing the selected quality attributes of NAP-ECPs;(2)a Plackett-Burman screening design was used to determine the most signif i cant factors affecting formulation composition and process parameters on acid resistance (Y1)and cumulative drug release(Y2);(3)a Box-Behnken optimization design was applied in the response surface method(RSM)study to obtain the exact relationship between the preparation CQAs and various factors.DS was established following the obtained response surface,control space(CS) was further achieved.(4)moreover,verif i cation experiments werecarriedout to identifythe robustnessand accuracyof the generated model.

2. Material and methods

2.1. Material

Naproxen(Volume average particle diameter of 32.822 μm) was purchased from Zhejiang Charioteer Pharmaceutical Co. Ltd.(Taizhou,China).Naproxen standard was purchased from National Institutes for Food and Drug Control(Beijing,China). Microcrystalline cellulose(MCC,Avicel?PH 101)and Croscarmellose Sodium(CCMC-Na,Ac-Di-Sol?SD-711NF)were kindlydonated fromFMC Biopolymer(Newark,U.S.A.). Lactose Monohydrate (GranuLac?200) and Polyvinylpyrrolidone(PVP K29-32)were gifts from Meggle GmbH (Wasserburg,Germany)and by China Division,ISP Chemicals Co.Ltd(Shanghai,China),respectively.Tween-80 was the product of Sinopharm Chemical Reagent Co.Ltd(Shanghai, China).Triethyl citrate were purchased from Aladdin Chemistry Co.Ltd(Shanghai,China).Glycerol monostearate were purchased from Nanjing Chemical Reagent Co.Ltd(Nanjing, China).Eudragit L30D-55 was kindly provided by Evonik Industries(Darmstadt,Germany).Hydrochloric acid was purchased from Nanjing Chemical Reagent Co.Ltd(Nanjing, China).Sodium phosphate dibasic were purchased from Shanghai Ling Feng Chemical Reagent Co.Ltd(Shanghai, China).Sodium dodecyl sulfate were purchased from Nanjing Chemical Reagent Co.Ltd(Nanjing,China).

2.2. Experimental methods

2.2.1. Preparation of NAP-IRPs starter cores

NAP-IRPs starter cores were prepared by extrusionspheronization method.Brief l y,NAP,MCC,PVP,GranuLac?200,CCMC-Na and Tween-80 were uniformly mixed at the weight ratio of 17.00:9.00:1.80:5.78:1.19:0.17.Distilled water was then added in a slow manner and mixed until a homogeneous and cohesive wet mass was obtained.The wet mass was then extruded and spheronized in an extrusionspheronization apparatus(JBZ-300 multifunctional pelleting and coating machine,YILIAN new-drug research institute, China).The resulting pellets were dried in hot air oven at 40°C for 24 h,afterwards screened through 24-30 meshes.

2.2.2. Preparation of NAP-ECPs

Adding Tween-80 and TEC to the 40%of the total amount deionized water which is preheated to 80°C and the mixture was dispersed 2 min at 6000 rpm by homogenizer (XHF-D high-speed homogenizer,Ningbo Scientz Biotechnology Co.Ltd,China).Then GMS was added and being dispersed 8 min at 6000 rpm under 80°C water bath, following by adding the residual water and continuously stirred to the ambient temperature.Thereafter,the above suspended liquid was added to Eudragit L30D-55 aqueous dispersion slowly,stirring at middle rate for 1 h.Prior to the f i lm coating procedure,the blend aqueous dispersion was sieved with a 60 meshes.Deposition of enteric materials on the NAP-IRPs starter cores was performed in a f l uid-bed granulator and coater(JHQ-100,Shenyang,China).And next coating aqueous dispersion was bottom-sprayed onto NAP-IRPs starter cores from a 1.0 mm diameter nozzle attached to a peristaltic pump(HL-2,Shanghai,China) under the condition of 31-32°C coating temperature, 100-150 ml min-1air f l ow rate,besides,the various spray rate and atomizing pressure by reference to Plackett-Burman design.The coating weight gain was calculated using the following equation F=(Wb-Wa)/Wa × 100%.In which, Wa and Wb are the pellets weight before and after coating respectively.The content of coating pellets was 47.92%. NAP-ECPs containing 250 mg of NAP were sealed in hard gelatin capsules with a manual capsule f i lling machine (CapsulCN,Zhejiang,China).

2.2.3. Determination of NAP

Concentrations of NAP in dissolution medium were quantif i ed by UV-visible spectroscopy(spectrophotometer WFZ UV-2000,Nanjing,China)at wavelengths of 331 nm.The linearity of the method was studied in the range of the drug concentration 2.0-120.0 μg/ml(r=0.9997).The RSD of the intraday and interday precision forNAP werelessthan 2%.The recovery rates for NAP were in the range of 98-102%,and the RSD were below 2%.

2.2.4. In vitro dissolution study of the NAP-ECPs

Dissolution studies were carried out according to USP 34 XXIII, apparatus II paddle method at a rotation speed of 50 rpm and the temperature was maintained at 37 ± 0.5°C.In brief,the total duration of dissolution was 2 h and 45 min.During the f i rst 2 h,the preparations were subjected to simulated gastric media[0.1 M HCl solution containing 0.5%sodium lauryl sulfate(SLS)]and during the later 45 min,the preparations underwentsimulatedintestinal media(BufferpH 6.8).Acidstage: the dissolution medium was 300 mL of 0.1 M HCl solution containing 0.5%SLS.Each NAP-ECPs capsule containing 250 mg NAP was put into each vessel.After 2 h,5 mL of thedissolution sample was withdrawn and substituted with the same amount of fresh dissolution medium.The samples were f i ltered through 0.45 μm f i lter and analyzed by UV spectrophotometer at 331 nm.Buffer stage:After 2 h operation in the acid stage,600 ml of 0.1 M K2HPO4preheated to 37°C was immediately added into the previous f l uid.If necessary,adjust by 1 M HCl or 1 M NaOH to a pH of 6.8 ± 0.05(PBS 6.8).The operation was continued for45 min.Atthe endof 45min,5 mL of the dissolution sample was withdrawn and passed through 0.45 μm f i lter and analyzed by UV-visible method for NAP as described above.

2.2.5. Risk assessment

Fish-bone diagram was constructed to identify the potential risks and corresponding causes.Specif i cally,acid resistance and cumulative drug release were identif i ed as the two CQAs. Based on previous knowledge and initial experimental data, failure mode and effect analysis(FMEA)method were further applied in the risk analysis of the parameters of the pellets coating.Each variable(potential failure mode)was scored in terms of severity(S),detectability(D)and probability(P).More broadly,Severity is a measure of the possible consequences of a failure mode affecting on the safety and eff i cacy of the f i nal product.Detectability def i ned that a failure mode can be detected.The f i nal parameter probability is considered as the occurrence probability or the likelihood of a failure.For each risk,S,D,P scores were multiplied together to produce a “Risk Priority Number”(RPN),RPN=S × D × P,which represents the overall magnitude of the risk.We ranked S,D and P of 5 as worst-case,1 as best-case value and 3 as moderate-case value, and then a maximum RPN of 125 and a minimum RPN of 1 are possible.

The RPN threshold was set at 60,and any formulation variable or process parameter with an RPN 60 or above was regarded as a potential critical factor,that is,potential risks are evaluated by subsequentprocess characterization studies since it possibly has a potential impact on CQAs and in consequence on product safety and eff i cacy,while factors with a lower RPN can be eliminated from further study[11].

2.2.6. Plackett-Burman design screening study

Based on the risk assessment results,Plackett-Burman study was used to screen signif i cant factors inf l uencing selected CQAs.The coating aqueous dispersion always contained Tween-80 as emulsif i er which was 40%of the GMS amount. The Plackett-Burman design screening study with each factor evaluated at low(-1)and high(+1)levels were summarized in Table 1.The determination of the low and high values was derived from the preliminary study results.The responses evaluated were Y1and Y2.

2.2.7. Box-Behnken design optimization study

Relied on the results of the Plackett-Burman screening study, RSM were applied in order to rapidly achieve the optimal NAP-ECPswith Design Expertsoftware(Version 8.0.6). Box-Behnken design was specif i cally selected here for aforementioned reasons of requiring fewer runs than a central composite design[7].The DOE details were listed in Table 2. The relationship between the material attributes/process parameters and CQAs was delineated in the DS.DS was determined from the common region of successful operating ranges for multiple CQAs(Table 2).The successful operating ranges for the Y1,Y2,were determined Y1≤ 10%and Y2≥ 80%, respectively.Based on the prior knowledge space,the CS was also determined.It is expected that operation within the CS will result in a product possessing the desired CQAs.

2.2.8. Conf i rmation tests of model

To verify the accuracy and robustness of the model,three different combinations were got at low,medium or high levels of the selected factors within CS.Formulations at those compositions were analyzed and further compared the observed responses with the predicted.

2.2.9. Statistical analyses

The results of Plackett-Burman study were analyzed via the statistical software of Minitab(version 16.1.0),the inf l uence of each parameter on the responses was demonstrated in the constructed Pareto charts,in which the length of each bar stood for the magnitude of the impact on the response.

Table 1-Plackett-Burman screening Design of Experiments and their results.X1:TEC percentage;X2:GMS percentage;X3: spray rate;X4:atomizing pressure;X5:batch size;X6:coating aqueous dispersion solid content;X7:coating weight gain;X8: curing time.The response acid resistance(Y1)and drug cumulative release(Y2)were reported as mean ± SD.

Table 2-Box-Behnken optimization Design of Experiments and their results.Acid resistance(Y1)and Drug cumulative release(Y2)were reported as mean ± SD.

For Box-Behnken analyses,the regression equation describestheeffectsof thevariablesontheresponsesintermsof linear,interactive and quadratic.The equation followed as:

where,b0is intercepted,and Y is the measured response associated with the factors(X1,X2and X3),their interactions (X1X2,X1X3and X2X3)and quadratic(X21,X22and X23).The pvalues related to the regression coeff i cients indicated the signif i cance of the factors on the response.ANOVA and the coeff i cient of determination(R2)were also applied to determine the suitability of the model[12].

3. Results and discussion

3.1. Risk assessment

Risk assessment aims to obtain all the potential high impact factors which will be subjected to a DOE study to establish a product or process DS.Just like outlined in the ICH Q9 document,risk identif i cation and risk analysis are two basic components of risk assessment[13].The f i rst step in the risk assessment was to systematically gather up all the possible factors that could inf l uence product quality.Basing on the literature data,previous study experiences[14,15]and preformulation data,f i sh-bone diagrams[16]was applied to organize hierarchically these factors(see Fig.2).The RPN scores using FMEA methodology was depicted Fig.3.To initiate the FMEA,we broke the failure modes down into those coming from the formulation,process,people,environment, and equipment inputs.Except the formulation and process, other factors leading to the variability in product quality were considered to be lower risk since all the development work was conducted under the conditions which were usually f i xed by preliminary experiments and/or prior knowledge.In addition,variables that could affect in vivo performance have generally been scored high.

Eight high-risk factors identif i ed in a risk analysis study have potential impact on Y1and Y2which were taken as indexes evaluating CQAs.These independent factors included: TEC percentage(X1,compared to polymer),GMS percentage (X2,compared to polymer),spray rate(X3),atomizing pressure (X4),batch size(X5),coating aqueous dispersion solid content (X6,amount of all the solid compared to the amount of the coating aqueous dispersion),coating weight gain(X7)and curing time(X8),these eight factors would be used for further screening study to obtain the signif i cant factors inf l uencing selected CQAs by Plackett-Burman design.

3.2. Inf l uence of various factors on Y1and Y2by Plackett-Burman screening DOE

The goal of this study was to identify the most signif i cant factors affecting the CQAs using Plackett-Burman design. Plackett-Burman design can estimate the signif i cance of the main factor from large numbers of factors with very high eff i ciency and accuracy[17],thus assuring to quickly reduce the number of high-risk factors needed to be studied in the next step.An eight factors-two levels-12 runs Plackett-Burman screening study was performed using Minitab statistical experiment design software and the responses were Y1and Y2.

Fig.2-An Fish-bone diagram illustrating factors that may have impact on acid resistance and cumulative drug release.

As learned from Table 1,Y1and Y2varied from 3.37%(PB-8) to 12.47%(PB-6),and from 72.47%(PB-8)to 90.11%(PB-6), respectively,for the various factor combinations.Fig.4 indicated that among all of the factors,coating weight gain,TEC percentage and GMS percentage(P ＜ 0.05)strikingly inf l uenced acid resistance,while cumulative drug release was signif i cantly impacted by coating weight gain and GMS percentage(P ＜ 0.05).As recorded in Table 3,the relative strength of each factor inf l uencing CQAs was further in detail depicted by the “Effect”value.A positive value indicates an effect that helps to enhance the response value;conversely a negative sign value represents an inverse relationship between the response and the factor.The higher the absolute value the greater the effect of that factor on the responses.Lower Y1and higherY2weredesired.In thisstudy,theresultsdemonstrated that decreased TEC/GMS percentage and increased coating weight gain would contribute to lower Y1.While both increasing the GMS and decreasing the coating weight gain would resulted in the increase of Y2.Eudragit L30D-55 was enteric-coated material,which dissolves merely in pH 5.5 or abovemedium.Thelargerthecoatingweight gainsthe thicker of coating f i lm.An attempt to decrease the coating f i lm excessively would result in discontinuous coating f i lm so that leading to too much drug release in the gastric region.In contrary,exorbitant thick f i lm might handicap drug releasing from formulation and Y2turned out to be too low.TEC was plasticizer which can decrease the f i lm forming temperature to favor coating.Besides,TEC was also hydrophilic material, which was easily dissolved in water thus playing a poreforming agent role in the enteric-coated f i lm.For this reason,too high levels of TEC will result in too much drug release in gastric environment,while the low value TEC can't decrease the f i lm forming temperature during coating process in eff i ciency.GMS was an antisticking agent,which candecrease the viscosity of the coating aqueous dispersion to avoid the adhesion among pellets and it also benef i ts for coating process.As emulsif i er Tween-80 was always accompany with GMS,which was 40%of the amount of GMS.It is well known that Tween-80 is well hydrophilic surfactant, when GMS was in the high content meaning Tween-80 was also in the high levels,for the same reason like TEC which would result in poor product quality.Too low GMS content could not decrease the viscosity of coating aqueous dispersion,but would lead to the failure of the adhesion among pellets during development and coating.A good correlation was obtained between the observed and predicted values as indicated by the R2value of 0.9510 and 0.9701for Y1and Y2.

Fig.3-Pareto chart showing RPN scores for the operating parameters for ECPs coating process.Parameters that had RPN scores higher than the threshold(RPN=60)were considered for further experimentation.

Fig.4-Standard Pareto charts showing the effects of independent variables on acid resistance and on cumulative drug release.

Basing on the result of the screening study,it is concluded that the responses were impacted signif i cantly by coating weight gain,TEC percentage and GMS percentage.These three parameters were further examined for their interactions and their effects on product quality attributes via the Box-Behnken DOE.From the Coef.of Table 3,it inferred that spray rate has an inverse impact on the Y1and Y2,and which was at a low value favoring to improve product quality;all of the residual factors of atomizing pressure,batch size,coating aqueous dispersion solid content and curing time were found to be less signif i cant and hence kept at 0.02 MPa,5 g,20%,and 15 min,respectively,considering cost and time saving in the next phase.

3.3. The responses surface for both Y1and Y2were obtained by Box-Behnken DOE

This study aimed at understanding the effects and interactions between coating weight gain,TEC percentage and GMS percentage on Y1and Y2.The levels used for selected parameters and the experimental results are listed in Table 2. As shown in Table 2,Y1and Y2varied from 4.51%to 13.12% and from 56.17%to 100.13%,respectively.Table 4 illustrated the statistical analysis results using Design Expert software; the values of the regression coeff i cients(coded)of the variables are associated with the inf l uence on the CQAs.The largest part of the absolute values for the coeff i cients(coded) meant the variables had the most potential effect on the response.Analysis of variance(ANOVA)was performed to evaluate the model signif i cance.A model will be considered statistically signif i cant if the P-value represented by “Prob ＞ F”is 0.05 or less.F-ratio is the “mean square between”divided by the ‘‘mean square within’’.The low value of the F-ratio's means more errors in the model.The adequacy of the developed models were estimated by “l(fā)ack of f i t”,R2,adjust R2[R2(adj)]and predicted R2[R2(pred)].The “l(fā)ack of f i t”estimates the error variance independently of the model.A signif i cant‘‘Lack of Fit’’(P ＞ 0.05)indicates that the variability measured by the replicates does not explain the gap between predicted and experimental data points.The R2value is the maximumsquaredregressioncoeff i cientthat canbeachievedby a model using only the variables in it,which is an indication of how well the model f i ts the experimental data,and the closer the value is to 1,the better the model is.The R2(adj)is a modif i ed form of R2considering the number of terms used within the modelandtheR2(pred)isan estimationof howwellthemodel predictsaresponsevalue.Table4demonstrated thesuggested most suitable model were quadratic for both responses of Y1and Y2.

Table 3-Estimated effects and coeff i cients for acid resistance(Y1)and for cumulative drug release(Y2)(coded units).

Table 4-Estimated regression coeff i cients for acid resistance and for cumulative drug release(Quadratic).

InTable4,itwasobservedthatY1wassignif i cantlyaffected by coating weight gain,TEC percentage and GMS percentage (P ＜ 0.05).Analysis of variance(ANOVA)in Table 5 manifested that the model was statistically signif i cant in its prediction of Y1,as depicted by Prob ＞ F of 0.0032,F-ratio of 20.02 and the Prob ＞ Ffor “l(fā)ackoff i t”of0.3882.ThemodelwasgoodwithR2of 0.9973,and can well depict independent factors on the response(Y1)with good the R2(adj)and the R2(pred).Table 4 implied no coating weight gain-TEC percentage-GMS percentage interaction on Y1for all of the P-value ＞ 0.05.

The three studied factors were also investigated inY2.As learned from regression analysis,coating weight gain was inversely correlated with increasing Y2(P ＜ 0.05)(Table 4).The increase in TEC and GMS percentage was also observed to signif i cantly increase Y2(P ＜ 0.05).There was a striking interaction between the coating weight gain and the TEC percentage with respect to an increase in the Y2(P ＜ 0.05). ANOVA statistics elucidated that a “Prob ＞ F”of 0.0205,F-ratio of8.57andtheProb ＞ F for “l(fā)ack off i t”of0.1639,indicating that the factors in the model were signif i cant.In addition,the model was speculated to well predict the response with R2,R2(adj)and the R2(pred)were 0.9964,0.9899 and 0.9479, respectively.

Counter and response surface plots were also analyzed to visualize the effects of the parameters and their interactions on the responses.Fig.5 showed the effects of coating weight gain,TEC percentage and GMS percentage on Y1and Y2.

3.4. Establishment and evaluation of the DS

DS was def i ned by the ICH Q8 as “the multidimensional combination and interaction of input variables(material attributes)and process parameters that have been demonstrated to provide assurance of quality.Working within the design space is not considered as a change;however the movement out of the design space is considered a change and would normally initiate a regulatory post approval change process.Design space is proposed by the applicant and is subject to the regulatory assessmentand approval”[5].The DS makes QbD a reality and the wider the DS,the more robust and f l exible the process is to accommodate variations.In this study,RSM in conjunction with optimization was applied to establish DS.

The quadratic response surface of CQAs as a function of selected variables was given in Fig.5.The objective of optimization is to optimize input variables range for meeting a goal.A vital step of optimization is to achieve appropriate response functions for both dependences and independences. In Design Expert,the desirability response values were set Y1≤ 10%and Y2≥ 80%.When GMS was at low and high limits set in experiment,Fig.6A and B showed the proposed DS, comprised of the yellow overlap region of ranges for the two CQAs.As depicted in Fig.6C.the overlay part of the yellow regionin Fig.6A withB satisf i edbothY1≤ 10%andY2≥ 80%,in which GMS was from 3%to 10%.However,coating weight gain and TEC percentage were variables and it was diff i cult to determine the exact value in real operation during development.Hereby,In order to determine the range of TEC percentage and coating weight gain and to achieve the mostrobust and f l exible product,mathematics knowledge was applied here to develop the relationship between coating weight gain and TEC percentage.Specif i cally,a biggest inscribed parallelogram was plotted in the orange overlay region shown in Fig.6D,whose abscissa of X7and ordinate of X1,the four coordinates were(23%,10%),(33%,10%),(36.5%, 16%),and(46.5%,16%),respectively.It was easy to obtain the exact value of X1and X7inside the parallelogram with math knowledge.The DS was established which was delineated in the orange region in Fig.6D,the range of the independent was, GMS percentage of 3%-10%,coating weight gain and TEC percentage of the point inside the orange parallelogram region.Known from the variables range of DS,the range of TEC percentage and GMS percentage were thoroughly same with the Box-Behnken DOE.Based on the preformulation experience,levels of TEC and GMS must be in the certain range to ensure that the production was performed favorably.Such results don't mean that the f i xed range of variables was too small,but illustrated that the available operation range was wide in manufacture ensuring product quality.

Table 5-Summary of ANOVA and lack of f i t for testing model.

Fig.5-Response surface(3D)plot of the effects of variables on the acid resistance and on the cumulative drug release of prepared ECPs.(A)coating weight gain/GMS and(B)coating weight gain/TEC were on the acid resistance;(C)coating weight gain/GMS and(D)coating weight gain/TEC were on cumulative drug release.

3.5. Determination of control strategy of the prepared NAP-ECPs

For ensuring a product of required quality of robustness and consistency during producing,ICH Q10 def i nes the control strategy as “a planned set of controls,derived from the understanding of current product and process that assures process performance and product quality”[18].The normal operating ranges is CS which is def i ned as the upper and/or lower limits for the critical material attributes and CPP.In the CS,the parameters were routinely controlled during production in order to assure the reproducibility[1].The acceptable rangeofbothmaterialattributesand processparameters were determined basing on the knowledge space from screening design and DS,the detail information was following as Fig.7.

3.6. Conf i rmation tests

To evaluate the accuracy and robustness of the obtained model,a conf i rmation test was carried out with low,medium and high value of all the eight factors within CS.The model conf i rmation experiments design and results were shown in Table 6.The results showed that the predicted and observed responses values of the formulations with the different variables combinations were nearly similar.A good agreement was obtained between the model prediction and experimental observation.Thus,the validity of the model was established and the formulation variables and process parameters were robust within the control space.

Fig.6-Design space of prepared ECPs comprised of the overlap region of ranges for the three CQAs using GMS percentage of (A)3%and(B)10%;(C)the theory region and(D)the operating region.

Fig.7-The control space of the prepared NAP-ECPs.

Table 6-Model conf i rmation experiments design table.X1:TEC percentage(%w/w);X2:GMS percentage(%w/w);X3:spray rate(ml/min);X4:atomizing pressure(MPa);X5:batch size(g);X6:coating aqueous dispersion solid content(%w/w);X7: coating weight gain(%w/w);X8:curing time(min).The response acid resistance(Y1,%)and drug cumulative release(Y2,%) were reported as mean ± SD for experimental value.

In QbD,robustness estimation is moved into method optimization for the def i nition of DS to ensure the CQAs values which were deduced from any working inside the DS are acceptable[19].Once the DS is established,the validation becomes an exercise to demonstrate that the process will deliver a product of acceptable quality when operating within the DS[20].No current regulatory document provides guidelines on how to estimate the DS level[19]and no new concept exists to implement a control strategy in the pharmaceutical industry[21].Within our approach,basing on the initial risk assessment and DS,a CS created with all the eight factors was involved.The conf i rmation tests illustrated that both the unimportant factors had little effect on the selected CQAs and the robustness of the model.

4. Conclusion

This current case study demonstrated how QbD approach can be applied toward the development of the ECPs preparation. Fish-bone paragraph and FMEA analysis favors to identify critical formulation and process parameters that affect ECPs productquality.And next,the Plackett-Burman and Box-Behnken design were used for screening the signif i cant factors and optimizing the variables range,respectively.The f i nal aim of this approach is to achieve a process model of the ECPs preparation,thus a DS can be established based on it, and a CS could be further obtained.Conf i rmation tests were carried out at three levels of low,medium and high of the variables and the results manifested that the prediction and experimental observation were in a good agreement,which conf i rmed the accuracy and robustness of the model.

Acknowledgment

This study is f i nancially supported by the major project of National College Students Innovation Project for the R&D of Novel Drugs(No.J1030830).Thanks to FMC,Meggle GmbH,ISP and Evonik Industries for providing the excipients and enteric-coated material.

R E F E R E N C E S

[1]Lawrence XY.Pharmaceutical quality by design:product and process development,understanding,and control.Pharm Res 2008;25:781-791.

[2]Xie L,Wu H,Shen M,et al.Quality-by-design(QbD):effects of testing parameters and formulation variables on the segregation tendency of pharmaceutical powder measured by the ASTM D 6940-04 segregation tester.J Pharm Sci 2008;97:4485-4497.

[3]Lionberger RA,Lee SL,Lee.Quality by design:concepts for ANDAs.AAPS Pharm Sci Tech 2008;10:268-276.

[4]Wu H,Khan MA,Hussain AS.Process control perspective for process analytical technology:integration of chemical engineering practice into semiconductor and pharmaceutical industries.Chem Eng Commun 2007;194:760-779.

[5]Guideline ICH.Pharmaceutical development Q8(R2).Annex to pharmaceutical development,Step;2009.p.4.

[6]Plackett RL,Burman JP.The design of optimum multifactorial experiments.Biometrika 1946;33:305-325.

[7]Motwani SK,Chopra S,Talegaonkar S,et al. Chitosan-sodium alginate nanoparticles as submicroscopic reservoirs for ocular delivery:formulation,optimisation and in vitro characterisation.Eur J Pharm Biopharm 2008;68:513-525.

[8]Abdalla A,M¨ader K.Preparation and characterization of a self-emulsifying pellet formulation.Eur J Pharm Biopharm 2007;66:220-226.

[9]Gandhi R,Lal Kaul C,Panchagnula R.Extrusion and spheronization in the development of oral controlled-release dosage forms.Pharm Sci Technolo Today 1999;2:160-170.

[10]Dubey A,Boukouvala F,Keyvan G,et al.Improvement of tablet coating uniformity using a quality by design approach. AAPS Pharm Sci Tech 2012;13:231-246.

[11]Vogt FG,Kord AS.Development of quality-by-design analytical methods.J Pharm Sci 2011;100:797-812.

[12]Ragonese R,Macka M,Hughes J,et al.The use of the Box-Behnken experimental design in the optimisation and robustness testing of a capillary electrophoresis method for theanalysisofethambutolhydrochlorideinapharmaceutical formulation.J Pharm Biomed Anal 2002;27:995-1007.

[13]Guideline ICH.Quality risk management.Q9 Current Step; 2005.p.4.

[14]Li J,Liu P,Liu J-P,et al.Bioavailability and foam cells permeability enhancement of salvianolic acid B pellets based on drug-phospholipids complex technique.Eur J Pharm Biopharm 2012;83:76-86.

[15]Li J,Liu P,Liu J-P,et al.Novel Tanshinone II A ternary solid dispersion pellets prepared by a single-step technique: in vitro and in vivo evaluation.Eur J Pharm Biopharm 2012;80:426-432.

[16]Ahmed A,Kayis B,Amornsawadwatana S.A review of techniques for risk management in projects.Benchmarking: An Int J 2007;14:22-36.

[17]Deng L-Y,Tang B.Generalized resolution and minimum aberration criteria for Plackett-Burman and other nonregular factorial designs.Stat Sin 1999;9:1071-1082.

[18]Guideline ICH.Pharmaceutical quality system q10.Current Step;2008.p.4.

[19]Orlandini S,Pinzauti S,Furlanetto S.Application of quality by design to the development of analytical separation methods.Anal Bioanal Chem 2013;405:443-450.

[20]Rathore AS,Winkle H.Quality by design for biopharmaceuticals.Nat Biotechnol 2009;27:26-34.

[21]Adam S,Suzzi D,Radeke C,et al.An integrated Quality by Design(QbD)approach towards design space def i nition of a blending unit operation by Discrete Element Method(DEM) simulation.Eur J Pharm Sci 2011;42:106-115.

*Corresponding author.Biological Pharmacy and Microparticle Drug Delivery Systems,No.24,Tong Jia Xiang,Nanjing 210009,China. Tel./fax:+86 25 83271293.

E-mail address:liujianpingljp@hotmail.com(J.Liu).

Peer review under responsibility of Shenyang Pharmaceutical University.

http://dx.doi.org/10.1016/j.ajps.2014.07.005

1818-0876/? 2014 Shenyang Pharmaceutical University.Production and hosting by Elsevier B.V.All rights reserved.