Statistical optimization of dithranol-loaded solid lipid nanoparticles using factorial design

This st�dy describes a 32 f�ll factorial experimental design to optimize the form�lation of dithranol (DTH) loaded solid lipid nanoparticles (SLN) by the pre�em�lsion �ltrasonication method. The variables dr�g� lipid ratio and sonication time were st�died at three levels and arranged in a 32 factorial design to st�dy the infl�ence on the response variables particle size and % entrapment efficiency (%��). From the statistical analysis of data polynomial eq�ations were generated. The particle size and %�� for the 9 batches (R1 to R9) showed a wide variation of 219�348 nm and 51.33� 71.8� %, respectively. The physical characteristics of DTH�loaded SLN were eval�ated �sing a particle size analyzer, differential scanning calorimetry and X�ray diffraction. The res�lts of the optimized form�lation showed an average particle size of 219 nm and entrapment efficiency of 69.88 %. Ex-vivo dr�g penetration �sing rat skin showed abo�t a 2�fold increase in localization of DTH in skin as compared to the marketed preparation of DTH.


INTRODUCTION
Solid lipid nanoparticles (SLN) were developed at the beginning of the 199�s as an alternative carrier system to the existing traditional carriers, s�ch as em�lsions, lipo� somes and polymeric nanoparticles (Müller et al., 1995).Solid lipid nanoparticles (SLN) are the new generation of nanopartic�late active s�bstance vehicles and are attracting major attention as novel colloidal dr�g carriers for topical �se.Small lipid vesicles in the range of nanometers have advantages, b�t avoid the disadvantages, of other colloidal carriers (Utreja, Jain, 2��1).Compared with polymeric nanoparticles, SLN have lower toxicity beca�se of the absence of solvents in the prod�ction process and also rela� tively low cost excipients.SLN offer combined advantages s�ch as controlled release, biodegradable, negligible skin irritation and protection of active compo�nds.(Müller, Mäder, Gohla, 2���; Sylvia, Müller, Wissing, 2��3).The 5�4 major advantage of SLN is the possibility of prod�ction on an ind�strial scale (Müller,Lippacher,Gohla,2���).The lipids which are �sed in making these carriers have an approved stat�s, offer low systemic toxicity and also low cytotoxicity (Müller, et al., 1997).Moreover, the small particle size of SLN ens�re close contact with the strat�m corne�m (SC), th�s increasing penetration of encaps�lated active agent into the skin (Mei et al.,2��3).
Psoriasis is one of the most common h�man skin diseases and is considered to have key genetic �nderpin� nings.�t is characterized by excessive growth and aberrant differentiation of keratinocytes (Lowes, Bowcock, Kr�e� ger, 2��7).Dithranol or Anthralin is a hydroxyanthrone, anthracene�derivative medicine applied to the skin of in� divid�als with psoriasis, exhibiting both anti�proliferative and anti�inflammatory properties.Dithranol acc�m�lates in mitochondria, res�lting in a red�ction of adenosine triphosphatase (ATP) synthesis, which leads to inhibition of DNA replication and repair and hence slows the exces� sive cell division that occ�rs in psoriatic plaq�es (Gerrit� sen, 2��7).�n addition, modification of DNA bases and inhibition of vario�s enzymes have also been described.�pidermal calmod�lin has been reported to be elevated in psoriasis.Dithranol was demonstrated to be a potent competitive antagonist of calmod�lin.(T�cker et al., 1986).SLN appears to be an interesting vehicle for topical administration of DTH (Carlotti et al.,2��9).SLN co�ld protect DTH from degradation ca�sed by UVA irradiation.
The objective of this st�dy was to develop a mathe� matical model �sing 3 2 experimental design in order to ded�ce the adeq�ate conditions for preparation of DTH� loaded SLN with desired characteristics able to improve the localization of DTH in skin.Ex-vivo penetration and localization of the optimized form�lations with maxim�m dr�g entrapment efficiency and minim�m particle size were also examined.

Preparation of SLN
SLN were prepared by pre�em�lsion followed by the �ltrasonication method (Fang,et. al.,2��8).Briefly, lipid phase consisted of dithranol, Compritol 888 ATO and Span 6� (3% w/v) maintained at 7� °C.An aq�eo�s phase was prepared by dissolving poloxamer 4�7 (3% w/v) in distilled water (s�fficient to prod�ce 5� ml of preparation) and heated to the same temperat�re as oil phase.Hot aq�eo�s phase was added to oil phase and homogenization was carried o�t at a temperat�re of 7�°C �sing an Omni TH homogenizer (Make Omni USA) at 46564.7 ×g for 3 min.Coarse hot oil in water em�lsion th�s obtained was s�bjected to f�rther size red�ction �sing an �ltrasonicator (make Sonic VCX 75�, USA) for 1��3� min.

Experimental design and statistical analysis
Most form�lation st�dies involve variation of one factor at a time, keeping other factors constant.Factorial design enables all factors to be varied sim�ltaneo�sly, allo� wing q�antification of the effects ca�sed by independent variables and interactions between them.�n this st�dy, a 3 2 f�ll factorial experimental design was introd�ced to opti� mize the form�lation of nanoparticles.�nitial st�dies were �ndertaken to decide on the excipients and their levels in the experimental design.
The choice of lipid was made on the basis of sol�� bility and partitioning of dithranol in the lipid.Aq�eo�s phase s�rfactant and lipid phase s�rfactant were selected on the basis of stability of dispersion prepared by �sing different s�rfactants.
�n order to optimize the preparation of form�lations, the dr�g� lipid ratio (X1) and sonication time (X2) were chosen as independent variables.These two factors that might affect the nanoparticle form�lation and three levels of each factor were selected (Table �) and arranged accor� ding to a 3 2 f�ll factorial experimental table (

Particle size analysis
The particle size analysis of the form�lations was performed �sing a Malvern Mastersizer 2��� MS device (Malvern �nstr�ments, Worcestershire, UK) and laser diffraction with a beam length of 2.4� mm, range lens of 3�� RF mm, at 14.4% obsc�ration.The mean diameter of each batch is recorded in Table ���.

Entrapment efficiency
For determination of entrapment in SLNs, the dr�g loaded lipid nanoparticles were separated from free dr�g by �ltra�centrif�gation (Beckmann �nstr�ments, �taly) at abo�t 158476.5 ×g.Free dr�g, determined spectrophoto� metrically from the added dr�g, remained �nentrapped in s�pernatant liq�id which was obtained after �ltra�centri� f�gation.The collected samples were added in chloroform and warmed to dissolve completely, and then extracted with dimethyl formamide (DMF) which dissolved only dithranol.The sol�tion was filtered, dil�ted with methanol and dithranol content determined spectrophotometrically.
�� was calc�lated according to the following eq�ation� Amo�nt of entrapped dr�g in SLN ��% = ---------------× 1�� Amo�nt of entrapped dr�g in SLN and free dr�g in dispersion Differential Scanning Calorimetry (DSC) study Differential Scanning Calorimetry (DSC) was per� formed on a Mettler�Toledo DSC 821 e (Col�mb�s, OH) instr�ment, and an empty standard al�min�m pan was �sed as reference.DSC scans were recorded at a heating rate of 1� °C/min in a temperat�re range of 3��3�� °C.DSC mea� s�rements were carried o�t on p�re compritol 888 ATO and dithranol as b�lk material and SLN loaded with dithranol.

X-ray diffraction
X�ray scattering meas�rements were carried o�t with a Philips PAN analytical expert PRO X�ray diffractometer 178�.The samples were irradiated with mono�chromati� zed C�Ka radiation and analyzed between 2�8� ºC 2Ø.The patterns were collected with voltage of 3�kV and c�rrent of 3� mA, respectively.The scanning rate (2Ø/min) was set at 1� ºC/min.

Fourier Transmission Infrared Spectroscopy (FTIR) studies
A Jasco FT�R spectrophotometer (Jasco FT�R� 4�1, Japan) was �sed for infrared analysis of samples.Abo�t 1�2 mg of sample was mixed with dry potassi�m bromide and the samples were examined at transmission mode over a wave n�mber range of 4��� to 4��cm �1 .FT�R st�dies were carried o�t on p�re compritol 888 ATO and dithranol as b�lk material and SLN loaded with dithranol.

Ex-vivo skin penetration studies
Ointment containing DTH�loaded SLN was prepa� red with white soft paraffin.Ointment containing plain DTH (marketed form�lation) was acq�ired from the ma� rket.Ex-vivo skin penetration st�dies of DTH ointment were performed with rat skin (Li�,et al.,2��7;Shah,et al.,2��7;Lv,et al.,2��9) �sing Franz diff�sion cell.Rat skin was taken from the abdominal region, after removing hair and s�bc�taneo�s fat tiss�e, by p�nching o�t a disc of ap� proximately 2.5�cm 2 in area.This slice was mo�nted on the Franz diff�sion cell.Phosphate b�ffer (pH 7.4) served as receptor fl�id.A small q�antity (�.1 g) of the ointment was applied to the skin s�rface.Serial sampling was performed at specified time intervals (1,2,3,4,5,6,7,8,9,1�,12 ho�rs) by removing the contents of the receptor compartment and replacing it with fresh medi�m.The samples were analyzed �sing UV�V�S spectrophotometer (Shimadz� UV 18��) and mean c�m�lative amo�nt diff�sed Q (mg/cm 2 ) at each sampling time point was calc�lated.At the end of 12 ho�rs, the amo�nt of dr�g in the receptor compartment, the dr�g remaining on the skin, and the dr�g concentration in the skin was determined by extraction into a s�itable solvent followed by spectrophotometric analysis �sing an UV�V�S spectrophotometer.

Experimental design and statistical analysis
The objective of this st�dy was to prepare solid li� pid nanoparticles of dithranol by pre�em�lsion �ltrasoni� cation method and to optimize the effects of form�lation variables on response parameters.Based on preliminary st�dies, compritol 888 ATO, Span 6� and Poloxamer 4�7 were chosen as lipid, lipid phase s�rfactant and aq�eo�s phase s�rfactant respectively.Dr�g�lipid ratio and soni� cation time were selected as variables and entrapment efficiency and particle size as response parameters.A 3 2 f�ll factorial design was selected as it helps st�dy the effect on response parameters by changing both variables sim�ltaneo�sly with a minim�m n�mber of experimental r�ns.
The particle size and �� for the 9 batches (R1 to R9) showed a wide variation 219�348 nm and 51.33� 71.�8%, respectively (Table ���).The data clearly indicated strong dependence of response variables on the selected inde� pendent variables.�n order to q�antify the effect of form�lation varia� bles on the response parameters, it was necessary to cons� tr�ct a mathematical model which wo�ld help in predicting val�es of response parameters at any selected val�es of form�lation variables within the bo�ndaries of the design.�t may be the case that the levels of form�lation variables which are intermediate between the selected levels yield optim�m form�lation.Design �xpert 7.1 software was �sed to generate a mathematical model for each response parameter and the s�bseq�ent statistical analysis.
The coefficients of the polynomial eq�ations genera� ted �sing MLRA (Design expert 7.1) for particle size and %�� of DTH�loaded SLN dispersion st�died are listed in (Table �V) along with the val�es of r 2 .Five coefficients (a to e) were calc�lated with k as the intercept.
The eq�ation was �sed to obtain estimates of the res� ponses at vario�s factor combinations, where the optim�m combination was fo�nd to be similar to that corresponding to R 7 and hence R 7 was treated as the optimized batch.
For particle size response, the Model F�val�e of 15.55 implies the model is significant.There is only a 2.35% chance that a "Model F�Val�e" this large co�ld occ�r d�e to noise.P val�e were fo�nd to be �.�235, with a val�e less than �.�5�� indicating model terms are significant.
The "Predicted R�Sq�ared" of �.6�33 is in rea� sonable agreement with the "Adj�sted R�Sq�ared" of �.9��9."AdeqPrecision" meas�res the signal to noise ra� tio.A ratio greater than 4 is desirable.The ratio of 11.2�7 indicates an adeq�ate signal th�s the proposed model can be �sed to navigate the design space.
For %�� response, the Model F�val�e of 26.66 implies the model is significant.There is only a 1.�9% chance that a "Model F�Val�e" this large co�ld occ�r d�e to noise.P val�e were fo�nd to be �.�1�9, with a val�e less than �.�5�� indicating model terms are significant.
The "Predicted R�Sq�ared" of �.7614 is in rea� sonable agreement with the "Adj�sted R�Sq�ared" of �.9413."AdeqPrecision" meas�res the signal to noise ra� tio.A ratio greater than 4 is desirable.The ratio of 12.862 indicates an adeq�ate signal.Th�s the proposed model can be �sed to navigate the design space.
Since the val�es of r 2 are relatively high for both the responses, i.e., �.9629 for particle size and �.978� for %��, the polynomial eq�ations form an excellent fit to the experimental data and are highly statistically valid.
Three�dimensional response s�rface plots for each response parameter were constr�cted to st�dy the effects of both form�lation variables sim�ltaneo�sly along with the behavior of the system.
Fig�re 1 shows response s�rface plot for particle size.�t can be observed from the fig�re that dr�g� lipid ratio had a positive effect on particle size i.e. particle size  increased with increase in dr�g� lipid ratio.Least particle size was observed at the lowest level of dr�g�lipid ratio.
A greater amo�nt of lipid may have res�lted in increasing the size of SLN.Sonication time had the opposite effect on particle size.�t can be observed from Fig�re 1 that particle size de� creased with increased sonication time.Fig�re 1 indicates that the sonication time had a greater impact on particle size compared to lipid content, moreover the effect of lipid content was more prono�nced at a high level of sonication which might be d�e to less effort req�ired to disperse the small lipid agglomerates.At low sonication time, lipid concentration had little infl�ence.
Fig�re 2 shows sim�ltaneo�s effect of sonication time and dr�g� lipid ratio on entrapment efficiency.�t can be observed that �� increased with increased sonication time.�ncreased sonication time res�lted in decreased particle size thereby increasing the total s�rface area and favoring entrapment.Dr�g� lipid ratio also had a similar effect.The val�e of �� was maximal when both form�la� tion variables were employed at their highest levels.The reasons can be attrib�ted to the maxim�m amo�nt of lipid present for entrapment of the dr�g.The effect of sonication on entrapment was not evident at high lipid levels.At low lipid levels however, the decrease in sonication red�ced entrapment which may be acco�nted for by less availabi� lity of lipid and decrease in lipid phase sol�bility of dr�g d�e to red�ced sonication.

Particle size analysis
The d (9�) for nanopartic�late dispersions showed a size ranging from 219 -348 nm (Table ���).The effect of lipid concentration on the particle size is evident from the particle size of samples R 3 , R 6 and R 9 (342 nm, 319 nm and 285 nm, respectively) where it is at high level and so� nication time is at low, middle and low levels.Samples R 4 FIGURE 1 -Three�dimensional response s�rface plots for particle size.The second pop�lation of particles, as shown in the above fig�re, may be d�e to inadeq�ate sonic energy at the periphery of the dispersion or to particle growth d�ring the time period between sonication and size analysis.

Entrapment efficiency
A high amo�nt of dr�g was incorporated in nano� particle dispersion.The %�� of different form�lations prepared (Table ���) indicates the positive infl�ence of lipid content on dr�g entrapment.The form�lations R 3 , R 6 and R 9 which have lipids at high level b�t s�rfactant at low middle and high levels, respectively showed % entrapment of 7�.3, 69.5 and 71.8�, respectively.This indicates overwhelming infl�ence of lipid on entrapment, irrespective of s�rfactant content.Form�lations R 1 , R 4 , which had lipid content at low level and s�rfactant at low and medi�m level, respectively show less % �� (< 6�%).For form�lation R 7 , the effect of high s�rfactant level was evident in the form of higher entrapment (69.88%).The partitioning of dr�g between lipid and water phases d�ring pre�em�lsion formation affects dr�g entrapment in nanoparticles.This in t�rn depends on the amo�nt of lipid, sol�bility of dr�g in lipid, process temperat�re and s�rfactant concentration.Therefore, the positive infl�en� ce of lipid content on entrapment is explained.

Differential scanning calorimetry
The DSC techniq�e was employed to characterize the DTH�loaded SLN.DSC thermograms (Fig�re 4) indicate the melting points and corresponding enthalpies of DTH, Compritol and SLN.The enthalpy indicates ab� sol�te heat energy �ptake and is given by the area �nder the transition peak.The sharp melting endotherm of DTH was observed at 182.1�°C with corresponding enthalpy of �35.43 J/g while compritol b�lk showed a melting endo� therm at 75 °C and enthalpy of �98.64 J/g.The DTH SLN showed an endothermic peak at 72.11°C and enthalpy of �65.61 J/g.A sho�lder peak was also observed alongside at 59.1°C.�n general, melting point depression is observed when the b�lk lipid is transformed to nanopartic�late form.The decrease in melting point and formation of a sho�lder peak is attrib�ted to smaller particle size, lattice defects and formation of amorpho�s regions arising o�t of incor� poration of dr�g molec�les (Chen,et al.,2��6;Sch�bert,M�¨ller�Goymann,2��5).This leads to the corresponding decrease in enthalpy from �98.64 to �65.61 J/g.

X-ray diffraction
X�ray diffraction data listed in the following Fig�re 5 was in good agreement with res�lts established by DSC meas�rements.The diffraction pattern of the b�lk matrix showed a marked difference from those of the SLN, as they showed a relatively sharper peak than the SLN.�t was clear that from DTH�loaded SLN, the less ordered crystals were the majority and the amorpho�s state contrib�ted to the higher dr�g loading capacity as seen previo�sly.There was a significant difference between the diffraction patterns of dithranol and DTH�loaded SLN.�t was confirmed that DTH existed in amorpho�s state in the DTH�SLN beca�se of the disappeared sharp peak of DTH in the diffraction pattern.

FTIR studies
From FT�R st�dy, the characteristic peak of dr�g s�ch as of the aromatic C=O (1597 cm �1 ), aliphatic C�OH (2923 cm �1 ), aromatic C�H (731, 773 and 1459 cm -1 ) disa� ppeared and were replaced by the peak of compritol 888 ATO where remaining peaks also either shifted or were replaced in the �R spectr�m of the form�lation shown in Fig�re 6.This established dr�g entrapment in lipid matrix.

Ex-vivo skin penetration studies
The ex-vivo permeation of DTH thro�gh rat skin from DTH�loaded SLN ointment was eval�ated �sing a Franz diffusion cell.The mean c�m�lative amo�nt diff�sed Q (mg/cm 2 ) at each sampling time point was calc�lated.The res�lts of diff�sion st�dies are represented graphically as Q vs Time in Fig�re 7.
�n the present investigation, DTH�loaded SLN ointment of optimized form�lation prod�ced significantly higher deposition of DTH in skin (55 %) than marketed Ointment of DTH (27 %), as shown in Table V.The stra� t�m corne�m which represents the principle barrier of skin str�ct�re has pore diameters of abo�t 2� nm b�t in f�lly hydrated state the pore diameters increase to 4�� nm.Th�s, a dr�g�localizing effect in the skin seems possible beca�se of hydration of skin by an occl�sion effect of  M.S. Gambhire, M.R. Bhalekar, V.M. Gambhire 51� SLN, s�bmicron particle size (219 nm) and high adhesion d�e to very high s�rface may explain the increase in skin permeation of dithranol SLN.D�e to the lipoidal nat�re of SLN, the penetrated dr�g concentrates in the skin and remains localized for a longer period of time, th�s enabling dr�g targeting for the skin (Pople, Singh, 2��6).

CONCLUSION
The pre�em�lsion followed by �ltrasonication techniq�e was �sed to prepare solid lipid nanoparticles of reprod�cible sizes in the range of 219 to 348 nm by addressing the effects of processing parameters.The ap�   plication of 3 2 factorial design proved to be a �sef�l tool for optimization of DTH�loaded SLN.Using the factorial design one can select a s�itable composition of form�la� tion to obtain DTH�loaded SLN in the size range of 219 to 348 nm depending on the application of the system.The res�lts of the ex-vivo penetration st�dies demonstrated that abo�t a two�fold increase in localization of DTH in skin was obtained with DTH�loaded SLN entrapped ointment compared to plain DTH.

TABLE II -
A 3 2 F�ll factorial experimental design layo�t

TABLE III -
Val�es of particle size and entrapment efficiency of DTH�loaded SLN (R 1 � R 9 ) as per f�ll factorial design

TABLE IV -
Val�es of coefficients for polynomial eq�ations and r 2 for vario�s response variables of DTH�loaded SLN

TABLE V �
Mean amo�nt of dithranol deposited in Rat skin and remaining on the skin