Alternatives to titanium dioxide in tablet coating

These red tablet cores were coated withfour different white coating suspensions.. To study the opacity of the different coatings,the cores were coated with a mass gain of 7%.. The placeb

Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=iphd20

Pharmaceutical Development and Technology

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/iphd20

Alternatives to titanium dioxide in tablet coating Juliana Radtke, Raphael Wiedey & Peter Kleinebudde

To cite this article: Juliana Radtke, Raphael Wiedey & Peter Kleinebudde (2021) Alternatives to

titanium dioxide in tablet coating, Pharmaceutical Development and Technology, 26:9, 989-999, DOI: 10.1080/10837450.2021.1968900

To link to this article: https://doi.org/10.1080/10837450.2021.1968900

Published online: 23 Aug 2021

Submit your article to this journal

Article views: 846

View related articles

View Crossmark data

Citing articles: 2 View citing articles

RESEARCH ARTICLE

Alternatives to titanium dioxide in tablet coating

Juliana Radtke, Raphael Wiedey and Peter Kleinebudde

Institute of Pharmaceutics and Biopharmaceutics, Heinrich Heine University, Universitaetsstrasse 1, Duesseldorf, Germany

ABSTRACT

Titanium dioxide (TiO2) is one of the most commonly used pharmaceutical excipients It is widely used as

a white pigment in tablet and pellet coatings However, it has recently been under massive criticism as a

number of studies suggest a cancerogenic potential It can therefore no longer be taken for granted that

TiO2will continue to be universally available for drug products Finding suitable alternatives is hence of

special relevance In this study, a number of different pigments were coated on tablets and their covering

potential analyzed None of the alternative pigments showed comparable effectiveness and efficiency to

TiO2, though the CaCO3/CaHPO4-based coating showed the second-best results Regarding the ability to

protect photosensitive active ingredients, ZnO showed a comparable potential as TiO2, while all other

pigments failed Using the alternative pigments as markers for in-line Raman spectroscopy as a process

analytical technology was challenging and led to increased prediction errors Again, the CaCO3/CaHPO4

-based coating was the only of the tested alternatives with satisfying results, while all other pigments led

to unacceptably high prediction errors

ARTICLE HISTORY

Received 29 April 2021 Revised 5 August 2021 Accepted 12 August 2021

KEYWORDS

Titanium dioxide; tablet coating; white pigments; Raman spectroscopy

Introduction

Titanium dioxide (TiO2) occurs in four different modifications in

nature: anatase, rutile, brookite, and riesite, of which only anatase

and rutile are frequently used in pharmaceutical products

(Balachandran and Eror1982; Tschauner et al.2020) As a widely

used white pigment in pharmaceutical coating formulations, it

ful-fills various functions On the one hand, it serves as a cosmetic

whitener and enhances the intensity of colored coatings On the

other hand, the presence of TiO2 in the coating layer provides

protection for photo-sensitive active pharmaceutical ingredients

(APIs) in the tablet core In the food industry, TiO2 is used under

the label E171 as a food additive, e.g as a visual embellishment

in icings, chewing gums and also coated tablets (Titanium

Dioxide Manufacturers Associationn.d.) The white pigment is also

contained in cosmetic products under the designation CI 7789

and as a UV filter/absorber in sunscreens (Titanium Dioxide

Manufacturers Associationn.d.)

TiO2 shows the highest covering potential of all white

pig-ments and has in addition a very high brightening capacity

(Titanium Dioxide Manufacturers Associationn.d.) This is reasoned

in its high refractive index and its birefringent character TiO2 in

the anatase modification has an average refractive index of 2.561

and in the rutile modification of 2.900 (at k ¼ 589 nm (Haynes

2014)) TiO2 also shows a high Raman activity Its frequent

pres-ence in pharmaceutical coating formulations simplifies the inline

process control of these coatings by Raman spectroscopy Here,

the applied coating mass during the process can be monitored

using the growing intensity of the characteristic TiO2 peaks

(M€uller et al.2012) The anatase modification shows characteristic

Raman peaks at wavenumbers of 640, 515, and 398 cm1, the

rutile modification at 612 and 448 cm1

TiO2showed a cancerogenic effect in several studies and

con-sequently its wide use is an increasingly subject of criticism with

high public awareness Intake of TiO2 can in principle take place orally, dermally or by inhalation The inhalation of very fine par-ticles, especially nanoparpar-ticles, is generally regarded as critical (Bakand et al 2012) Animal studies have shown that nanopar-ticles penetrate deep into the lungs and can lead to chronic inflammation (Ernst et al 2002; Muhle et al 1989; Baggs et al

1997) It was also observed that the inhalation of extremely high TiO2 concentrations over a very long period of time led to an increased formation of lung tumors in rats (Pott and Roller 2005; Heinrich et al.1995) For example, rats were exposed to an aero-sol containing 5 mg TiO2per m3for 24 months, 5 d a week, 6 h a day (Muhle et al.1989) These and other studies form the basis for

an ongoing European classification procedure for TiO2 according

to the‘Regulation on Classification, Labelling and Packaging (CLP)

of chemicals with particularly hazardous substance properties’ (EC

No 1272/2008) In the course of this, TiO2 was classified by the Risk Assessment Committee of the European Chemicals Agency (ECHA) as‘carcinogenic by inhalation’ in June 2017

The probable reason for this cancerogenic effect is that par-ticles can induce a chronic inflammation in the lungs (Bakand et

al 2012) This immune reaction most likely leads to an increased inflammation-based cancer risk In the Annex published by ECHA, this new classification of TiO2 is justified by the occurrence of increased inflammation in rats that have inhaled large amounts of TiO2 (Pott and Roller2005) A study also showed an increase in squamous cell carcinomas and bronchiolalveolar adenomas (Lee

et al.1985)

With regard to oral intake however, the Risk Assessment Committee concluded that there was no evidence of a carcino-genic effect of TiO2 after oral intake (Annex 1 2017) Also, the European Food Safety Authority (EFSA) concluded in 2016 that there were no indications of health risks for consumers based on data available to date (European Food Safety Authority [EFSA] CONTACT Peter Kleinebudde kleinebudde@hhu.de Institute of Pharmaceutics and Biopharmaceutics, Heinrich Heine University, Universitaetsstrasse 1, 40225 Duesseldorf, Germany

ß 2021 Informa UK Limited, trading as Taylor & Francis Group

2021, VOL 26, NO 9, 989 –999

https://doi.org/10.1080/10837450.2021.1968900

2016) The low absorption and bioavailability of TiO2 (< 0.1% of

orally ingested amount) are emphasized This factum is however

not unanimous, since data suggesting a partial oral absorption of

TiO2are known in the literature for a long time (B€ockmann et al

2000) After the publication of four new studies stating a potential

toxicity after oral intake (Bettini et al 2017; Proquin et al 2017;

Guo et al.2017; Heringa et al.2016) the EU Commission called for

a reassessment of this conclusion in 2018

Above all, the study by Bettini et al from 2017 should be

men-tioned here (Bettini et al.2017) TiO2was administered to ratsvia

a gavage probe over seven days or over 100 days via drinking

water The used TiO2 was a product marketed as E 171 with

44.7% of the particles being smaller than 100 nm (mass fraction)

Among the observations made in the rats were effects on the

immune system, changes in the intestinal mucosa, and increased

inflammatory parameters A possible tumor-promoting effect was

derived from this (Bettini et al.2017) EFSA concluded, even after

a re-evaluation, that the 2016 assessment should not be revised

However, data gaps regarding possible effects on the

reproduct-ive system are stated and further studies are recommended In

addition, EFSA has set up a working group to further define the

specifications of food additives, e.g with regard to particle size

distribution

Despite the above assessments, in April 2019 the French

gov-ernment issued a regulation banning the placing on the French

market of food containing the food additive E 171 for a period of

one year starting 1 January 2020 (Le ministre d’Etat, ministre de

la transition ecologique et solidaire, et le ministre de l’economie

et des finances2019) The decision is based on an expert opinion

of the French Agency for Food Safety, Environment and Health at

Work (ANSES) On 21 December 2020, the ban was expanded for

another year (Le ministre d’Etat, ministre de la transition

ecologique et solidaire, et le ministre de l’economie et des

finan-ces2020)

This expert opinion underlines the lack of scientific data which

is not compatible with the classification of TiO2as safe for health

(Additif alimentaire E 1712019) It calls for the collection of

fur-ther data to characterize the different physico-chemical forms of

TiO2 and additional toxicological data on the possible effects of

their uptake In June 2019, EFSA responded with a statement

con-cluding that the ANSES opinion does not contain significant new

evidence that would justify a reassessment of TiO2 (EFSA 2019)

As early as 2017, the European Commission published a‘Call for

Data’ calling for the ESFA recommended studies on reproductive

toxicity to be carried out and for more accurate characterization

by August 2019 A final EFSA assessment was expected by the

end of 2020, but has not been published to date

Due to the extent of the debate and first precaution measures

by authorities, it can no longer be taken for granted that TiO2will

continue to be universally available for drug products Finding

suitable alternatives is hence of special relevance

Despite this relevance, only little work on this question has

been published so far A study by the paint manufacturer Akzo

Nobel studied Zinc sulfide, zirconium dioxide, calcium carbonate

(CaCO3), and barium sulfate as alternatives to TiO2 in paint (de

Jong and Flapper 2017) The study concluded that none of the

studied pigments could achieve the opacity of a TiO2-containing

formulation For pharmaceutical applications, no study has been

published so far to the best of the author’s knowledge

The aim of this study was therefore to test alternative white

pigments, with a special focus on recently introduced

ready-to-use mixtures that are specifically advertised as TiO2-free white

pigments The suitability of the pigments should be investigated

regarding three important functions of TiO2in drug products: cre-ating a clear white surface with a high covering potential, protect-ing photo-sensitive APIs from light and servprotect-ing as a marker for in-line Raman-spectroscopy as a PAT-tool during coating

Materials and methods

Materials Tablet cores

The biconvex placebo cores consisted of 50% lactose (TablettoseVR

80, Molkerei MEGGLE Wasserburg GmbH & Co KG, Wasserburg

am Inn, Germany), 49.5% microcrystalline cellulose (MCC, AvicelVR

PH-102, FMC Corporation, Philadelphia, PA) and 0.5% magnesium stearate (Peter Greven GmbH & Co KG, Bad M€unstereifel, Germany) The nifedipine cores consisted of 5% nifedipine (Bayer

AG, Wuppertal, Germany), 35% MCC (SanaqVR

102, Pharmatrans-Sanaq AG, Allschwil, Switzerland), 59% lactose (FlowLacVR

100, Molkerei MEGGLE Wasserburg GmbH & Co KG, Germany) and 1% magnesium stearate (ParteckVR

LUB MST, Merck KGaA, Germany) The properties of the tablet cores are given inTable 1

Coating suspensions

For the investigation of the opacity of different coating suspen-sions, colored tablet cores were needed Therefore, four batches

of placebo cores were coated with a HPMC-based red immediate release coating suspension (AquapolishVR

P red, Biogrund GmbH,

H€unstetten, Germany) These red tablet cores were coated with four different white coating suspensions The first suspension con-tained 15% polyvinyl alcohol/polyethylene glycol graft copolymer (KollicoatVR

IR BASF, Ludwigshafen, Germany), 3% titanium dioxide (TiO2; KRONOS Worldwide, Inc., Dallas, TX) in the anatase modifi-cation, 0.5% sodium lauryl sulfate (SDS) and 81.5% demineralized water The second suspension contained 15% polyvinyl alcohol/ polyethylene glycol graft copolymer (KollicoatVR

IR, BASF, Germany), 3% zinc oxide (ZnO; Grillo Zinkoxid GmbH), 0.5% SDS, and 81.5% demineralized water For the preparation of these sus-pensions, SDS and the pigment were added to water and dis-persed with an Ultra-TurraxVR

(IKAVR

-Werke GmbH & CO KG, Staufen, Germany) to achieve a homogenous suspension KollicoatVR

IR was dissolved separately in water and then added to the pigment suspension In addition, two HPMC-based ready to use coating mixtures were used Both were applied as a mixture

of 15% coating suspension and 85% demineralized water AquaPolishVR

P white 014.117 (APP117, Biogrund GmbH,

H€unstetten, Germany) is composed of hydroxymethylcellulose, hydroxypropylcellulose, polyethylenglycol, calcium carbonate Table 1 Tablet core properties, n ¼ 20, x ± relative standard deviation Cores

Diameter/mm (%)

Height/mm (%)

Band height/mm (%)

Mass/g (%)

Table 2 Investigated pigments with refractive indices.

(CaCO3) and dicalcium phosphate AquaPolishVR

P white 014.123 contains hydroxymethylcellulose, hydroxypropylcellulose, glycerin,

magnesium carbonate, MCC, and dicalcium phosphate Table 2

indicates the pigments which were included in the investigation

and their refractive index

For the opacity study, the four coating suspensions were

applied to the red cores To investigate the suitability of Raman

spectroscopy, the four coating suspensions were applied to

pla-cebo cores

Methods

Particle size distribution

The particle size distribution of the pigments was determined by

laser diffraction (Mastersizer 3000, Malvern Instruments, Malvern,

UK) For this purpose, all samples were dispersed in water and

measured three times using the wet-dispersion unit The

concen-tration of sample in water was selected in such a way that an

optimal laser obscuration of 2 6% was guaranteed Any

agglom-erates of particles were deagglomerated by ultrasound prior to

each measurement Using the corresponding software, the particle

size distribution was determined from the data based on the Mie

theory and given as volume distributions The refractive index

was adjusted depending on the material For ZnO a refractive

index of 2.0034 and for TiO2of 2.493 was applied (Bodurov et al

2016) Since the ready-to-use mixtures (APP117 and APP123)

con-tained i.a dibasic calcium phosphate, the refractive index of

dical-cium phosphate (1.55) was applied for the respective

measurements The x10 quantile and the x50 quantile from the

obtained distribution curves were used to describe the particle

size Determination was challenging for the ready-to-use mixture,

since they contained further excipients like polymers and

stabil-izers All other excipients except the pigments were however

sol-uble and therefore expected not to interfere with the particle size

determination This was especially the case, since the sample was

strongly diluted with water before measurement

Coating of tablets

All coating processes were performed in a laboratory drum coater

with a drum size of 5 l (BFC 5, L.B Bohle Maschinenþ Verfahren

GmbH, Ennigerloh, Germany) Two 1.0 mm nozzles (D€usen-Schlick

GmbH, Untersiemau, Germany) were installed for the application

of the coating suspensions The distance between the nozzles and

the tablet bed was 10 cm Process parameters for the application

of the different coating suspensions for the opacity study are

given in Table 3 To study the opacity of the different coatings,

the cores were coated with a mass gain of 7% Samples of 100 tablets were taken during the coating process at mass gains of 1,

2, 3, 4, 5, 5.5, 6, and 6.5% The placebo cores for the Raman inves-tigation were coated with a mass gain of 3% for the TiO2- and ZnO-containing coatings and 5% for APP117 and APP123 For each coating suspension, a calibration and a test batch was coated To also study the effect of coatings on drug stability, 15 nifedipine tablet cores were added to the coating process They were of larger size (12 mm) than the other tablets and could therefore be manually selected from the batches after coating The coating parameters for this study are given inTable 4 All cores were coated with an inlet air volume of 100 m3/h The tab-let batch size was 3800 g for all coating runs

Scanning of coated tablets

To determine the opacity of various white coatings, the tablets were scanned at different sample times and at the end of the process using a standard computer scanner (Epson Perfection V800 Photo, Suwa, Japan) The images were taken at a resolution

of 300 dpi with a color depth of 48 bits No color correction was applied Per sample time 40–50 tablets were measured, in add-ition six white tablets were measured as reference The tablets were arranged in rows of 5 The edges of the scan support were covered with white paper Since the scanner could not be closed during the measurement, all measurements were carried out under exclusion of light Each scan was performed three times, an average image of the three images created and used for evaluation

Image analysis for opacity determination

The image analysis was performed in Python version 3.7 (Python Software Foundation, Wilmington, DE) The OpenCV library was used The RGB color values obtained were converted into the cor-responding HSV and Lab color values In the HSV color space, H (hue) describes the hue, S (saturation) the color saturation, and V (value) the light value The V-value can take values between 0 for black and 1 for white The H and S values from the HSV color space were included in the evaluations of this work The Lab color space describes all perceptible colors in a three-dimensional color space The brightness value L (luminance) is perpendicular to the color planes a and b The coordinate indicates the color type and intensity between green and red, the b coordinate between blue and yellow The luminance can take values between 0 and 100, where L¼ 0 stands for black and L ¼ 100 for white From the Lab color space, only the L values were used directly for evaluation to track changes in brightness during coating In addition, the color

Table 3 Coating process parameters, opacity trial.

Table 4 Coating process parameters, Raman trial.

distance, Delta E, was calculated in this color space Here, the

color distance to white was calculated and evaluated According

to EN ISO 11664-4 (International Organization for Standardization

2008) Delta E between two colors is calculated as Euclidean

dis-tance usingEquation (1):

DE1, 2¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðL1L2Þ2

þ ða1a2Þ2

þ ðb1b2Þ2

q

(1)

The image analysis was performed for a circular area in the

center of each tablet The area of the analyzed region was half of

the tablet cap The OpenCV-based evaluation leads to the HSV

and Lab values for each tablet recorded These were then

statistic-ally evaluated by calculating the mean value, standard deviation

and confidence interval (95%) The accuracy of the measurements

was checked using the values of the white reference tablets

Coating thickness determination

The coating thickness of the three coatings at a mass gain of 7%

was calculated using the volume and the density of the applied

coating For density measuring, coated films were produced for

all coating formulations Films were casted using a Coatmaster

510 (Erichsen, Hemer, Germany) The plate was heated to 40C to

simulate the conditions during the coating process The

suspen-sions were casted using a coating knife with a gap width of

800mm The film forming was completed after a drying time of

30 min The density of the casted film was measured by gas

pycn-ometry (AccuPyc 1330, Micrometics Instrument Corp., Norcross,

GA) For all measurements, the temperature was kept constant at

25C The volume of the coating was calculated from the total

tablet surface of the tablets and the applied coating mass The

change of the tablet surface during the coating process was

assumed to be negligible The coating efficiency was determined

by weight analysis after each coating process and included in the

calculations

Crushing strength and tablet geometry

Crushing strength, diameter, height, and mass of the tablets were determined using the Smart Test50 tablet tester (Dr Schleuniger Pharmatron, Thun, Switzerland) For each core type, 20 tablets were measured The breaking force was standardized according

to Fell and Newton by calculating the tensile strength (Fell and Newton1970) The cap height was measured for 20 tablets with a calliper (Digital ABS Caliper, Mitutoyo Corporation, Kawasaki, Japan)

Photostability study Storing conditions

The embedded nifedipine tablet cores were protected from light exposure by a TiO2-containing coating or one of the three alterna-tive TiO2-free white coatings (ZnO-containing, APP117, or APP123) Nifedipine is an aromatic compound of the dihydropyri-dine-type and, like other members of this group, shows a pro-nounced sensitivity to light (Ebel et al.1978) As shown in Figure

1, nifedipine degrades to a nitrophenylpyridine analog (impurity A) and a nitrosophenylpyridine analog (impurity B) under the influence of light

According to Lehto et al., nifedipine shows maximum instabil-ity in the solid-state at a wavelength of 455 nm (Lehto et al

1999) The protective capacity of the four coatings was investi-gated on the basis of a stability test in a light chamber at a wave-length between 315 and 400 nm, which is not the worst-case wavelength, but still was considered to be high relevance The light chamber was equipped with four fluorescent tubes with a UVA radiation power of 3.5 W (SUPRATEC 18 W/73, OSRAM GmbH,

M€unchen, Germany) In each case, six tablets per coating were stored in the light cabinet and exposed to UV light for 2 or

4 weeks The tablets were flipped over every week to ensure a uniform light irradiation from both sides of the tablets Three tab-lets each were removed after 2 weeks and 3 after 4 weeks and the content determined by HPLC In addition, the content of

Figure 1 Molecular structure of nifedipine and its degradation products: a) impurity A; b) impurity B.

uncoated nifedipine tablets was determined after 2 and 4 weeks

of storage in the light cabinet and after 4 weeks of storage in

the dark

HPLC analysis

The coated and non-coated nifedipine tablets were weighed after

removal from the light cabinet and each was dissolved in 20 ml

methanol and diluted with the mobile phase to 50.0 ml The

mobile phase consisted of nine parts acetonitrile, 36 parts

metha-nol, and 55 parts distilled water To ensure a complete dissolution

of the API, the samples were treated in an ultrasonic bath for

20 min and shaken regularly The samples were handled under

exclusion of light and the dissolved sample was transferred into

light-protected amber glass vials Three tablets per coating

prep-aration were examined, whereby the content was determined

three times by HPLC The analysis was carried out according to

Ph Eur 2.2.29 (European Pharmacopoeia 9.0 2017), which

describes the test for related substances of nifedipine The HPLC

(VWR Hitachi HPLC, VWR International GmbH, Darmstadt,

Germany) was equipped with a LiChrospherVR

RP-18 5mm column (Merck KGaA, Gernsheim, Germany) Nifedipine was identified by

the retention time given by the Pharm Eur and by comparison

with chromatograms of pure nifedipine The impurities were

iden-tified by relative retention times from literature: 0.72 for the

nitro-phenylpyridine analog and 0.86 for the nitrosonitro-phenylpyridine

analog (Florey1990) In absence of pure degradation products for

reference, impurities A and B were quantified by comparing the

peak areas of the nifedipine peaks and the peak areas of the

deg-radation products The absolute values for concentration might

therefore by to a certain extent flawed, relative comparisons

are valid

Raman spectroscopy

Raman spectra were measured in-line using a Raman RXN2

ana-lyzer (Kaiser Optical Systems, Inc., Ann Arbor, MI) with a PhAT

probe As this non-contact optic device forms a spot with a

diam-eter of 6 mm, it allows the measurement of a larger sample area

The diode laser operates at 785 nm with a laser power of 400 mW

The PhAT probe was installed through the front door of the

coater During measurements, the optic of the PhAT probe was

dedusted permanently with compressed air The probe was

installed with a distance of 21 cm to the tablet bed The iC

RamanTMsoftware package (Kaiser Optical Systems, Ann Arbor, MI)

was used for data acquisition During the coating processes one

spectrum was collected every 10 s The exposure time was set

between 2 and 3 s

Data analysis methods

Data pretreatment

Before model building, the measured Raman spectra were

pre-treated using MatlabVR

R 2018b (The MathWorks, Inc., Portola Valley, CA) A moving average with a window size of 12 was

applied to the raw spectra Then, spectra were preprocessed using

standard normal variate (SNV) The preprocessed range of

wave-numbers depended on the applied coating layer and was chosen

accordingly with regard to the model performance parameters

For each coating process, models were built and tested with

dif-ferent spectral ranges to find the range, which led to the smallest

root mean square error of prediction (RMSEP)

Partial least squares regression (PLSR)

Partial least squares regression (PLSR) was performed using MatlabVR

R 2018b The models were built by regression of the inline-measured Raman spectra (X-data) against the applied mass

of coating suspension (Y-data) As the signal of the Raman active ingredients increases linearly with the mass of applied coating suspension, a linear increase of the Raman spectra during the coating process was assumed Data of the entire coating run were included in the PLSR model The spectral range and optimal num-ber of factors were defined individually for each coating formula-tion in dependence of the model predicformula-tion performance ability Models were built with the data of the calibration data set and tested with the new data of a test data set

Results and discussion

Covering capacity Visual inspection

In dependence of the coating formulation, a mass gain of 7% cor-responded to a different layer thickness The TiO2-containing coat-ing showed a maximum layer thickness of about 90mm, the ZnO-containing coating of 86mm, APP117 of 77 mm, and APP 123 of

86mm The TiO2 and ZnO-containing coating layers both had a pigment content of 20% Since APP117 and APP123 are ready-to-use mixtures, the pigment content is unknown here Figure 2

shows tablets, which were coated with the different coatings with

a mass gain between 1 and 7% By the visual inspection, it becomes clear that the coatings showed different cover-ing capacities

Even at smaller mass gains, the TiO2-containing coatings showed a high opacity Thus, the tablets appear almost white even with a mass gain of 3%, with the application of higher mass gains only slight changes could be achieved This also explains the fact that non-functional TiO2-containing coatings are usually applied with a target mass gain between 3 and 4% A comparable opacity is only achieved with APP117 at mass gains > 5% APP117 showed the second-highest opacity, despite the slightly lower coating thickness The optical differences between the ZnO-containing coating and APP123 were hardly visible, whereby the opacity of the ZnO-containing coating appeared slightly higher Both coatings showed an insufficient opacity at mass gains< 7% and were not able to completely cover the red color of the tablet cores Even at mass gains of 7% the red color of the cores is still visible

Image analysis

The results, which are shown in Figure 2, are only based on optical considerations, these observations should be supple-mented with quantifiable results using a computer scanner

Figure 2 Red tablets with increasing mass of coated white pigment.

Images of samples, which were taken during the coating

proc-esses and at the end of the process, were obtained using this

scanner For each sample time point, approximately 50 tablets

were scanned and examined by image analysis The values of the

Lab- and HSV-color space were determined using the obtained

images and calculations were based on a circular area in the

mid-dle of the tablet cap The v-, S-, and L-values were considered for

evaluation In addition, the color distance to white (Delta E) was

calculated Figure 3 shows the results, which were obtained by

image analysis

As the white coatings should cover the red tablet core, a

decrease of saturation was expected during the coating process

and with an increasing weight gain The saturation decreased by

the application of all four coating formulations (Figure 3) At a

mass gain of 1% the TiO2-containing coating could reduce the

saturation under 0.1 In comparison, a value of 0.23 was achieved

using APP117, 0.27 using APP12, and 0.31 with the

ZnO-contain-ing coatZnO-contain-ing The initial saturation of the uncoated tablet cores

was 0.48 At a mass gain of 3%, tablets, which were coated with

the TiO2-containing coating, showed a saturation of 0.035 and

even the application of higher mass gains resulted only a slight

decrease in saturation With a final mass gain of 7% a saturation

of 0.031 was achieved These results confirmed the visual

observa-tions, in which the TiO2-containing coating showed a high

cover-ing ability even at lower mass gains Also, the application of

APP117 reduced the saturation to 0.03 at a mass gain of 7%

However, a higher mass gain of 5% was necessary to achieve

comparable results as the TiO2-containing coating The

ZnO-con-taining coating and APP123 led to a lower decrease in saturation,

at a mass gain of 7% both showed a saturation about 0.06 With

the TiO2-containing coating this value was already achieved with

a mass gain of 2% and with APP117 with a mass gain of 4% As

‘v’ describes the brightness in the HSV color space, it is expected

to increase while coating colored cores with white coating formu-lations Such an increase was observed during the application of all four coating formulations (Figure 3) However, the TiO2 con-taining coating again showed a clear superiority with a brightness value of 0.87 at the end of the process Among the TiO2-free coat-ings, APP117 showed the highest brightness value (0.79) With the ZnO-containing coating and APP123 a maximum brightness of 0.72 was achieved at the end of the process These results are consistent with the visual observations, which were described in 3.1.1 The TiO2 containing coating appeared whitest, followed by the APP117 The two other coatings did not show a sufficient cov-ering ability even for higher mass gains.‘L’ indicates the lumines-cence and describes the whiteness in the Lab color space As expected, the progression of L was very similar to the progression

of v So, the observations, which were based on v, can be trans-ferred to L Also, considering Delta E, the TiO2-containing coating was the most efficient and effective coating in terms of covering capacity Delta E was calculated as the color distance to pure white with an L-value of 100 At a mass gain of 7%, the tablets, which were coated using the TiO2-containing coating, showed a Delta E value of 13.4 This was the lowest achieved Delta E value, however, a visual difference to white could still be recognized in front of a white background

The results are in good alignment with results from Rowe, who studied the opacity of various pigments using a colorimeter (Rowe

1984) The white pigments CaCO3, calcium sulfate, talc, and TiO2, as well as a number of non-white pigments were comprised in the coating layers and compared regarding opacity The opacity was determined as the contrast ratio of measurements in front of a black and a white background Without the addition of a pigment, the contrast ratio was 33.3%, with TiO as a pigment 91.6%, with Figure 3 Change of color measures with increased coating mass.

CaCO346.7%, and with talc 46.4% Also in the study of Rowe, TiO2

shows a significantly higher opacity compared to the other

pig-ments which is in agreement with the results of this study

(Rowe1984)

Particle size of pigments

The opacity of white pigments is largely due to their ability to

scatter incident light In addition to the properties of the incident

light, the scattering depends on the optical properties of the

par-ticle and its parpar-ticle size, shape, surface texture, spatial

orienta-tion, arrangement of the particles, etc (Nelson and Deng 2008)

Up to a certain level the light scattering and thus the opacity of a

particle can be increased by reducing the particle size Below a

certain particle size however, the efficiency of light scattering

decreases again, for example TiO2 particles with sizes < 0.1 mm

show a decrease in light scattering and thus in the resulting

opa-city (Diebold2014) Since the human eye shows the highest

sensi-tivity to yellow-green light (wavelength around 0.55mm), the

optimum diameter of commercial white pigments is on average

0.2–0.3 mm (Winkler 2013) The particle sizes of the pigments of

the four white coatings used were determined by laser diffraction

using wet dispersion in water TiO2 showed the smallest particle

size with a x50 value of 0.403mm and a x10 value of about

0.007mm (Figure 4)

Here, the x50 value was closest to a particle size of 0.2–0.3 mm

described as optimal The high opacity of the TiO2-containing

coating can therefore be explained not only by the high refractive

index and the birefringence character of TiO2, but also by the

small particle size All other pigments used had significantly

higher particle sizes ZnO and APP117 showed comparable x10

and x50 values, whereas pure ZnO with a x50 value of 3.5mm and

a x10 value of 1.2mm showed slightly smaller particles than

APP117 (x50: 4.1mm, x10: 1.9 mm) Despite the small particle size

and a rather high refractive index of 2.0 (Haynes2014), the

ZnO-containing coating showed insufficient opacity, as seen inFigure

3 This can be explained by the wide energy band gap of ZnO at

Eg3.3 eV (Srikant and Clarke 1998) It causes an increased light

transmission at wavelengths in the visible range (above 400 nm)

(Struk et al.2010) Since opacity is the reciprocal of transmission,

this results in a reduction of opacity APP117 showed a

signifi-cantly higher opacity compared to APP123 This corresponds to

the result of the particle size measurement With a x50 value of

16.3mm and a x10 value of 6.2 mm, APP123 had by far the largest

particles APP117 contains CaCO3 with a refractive index of 1.66

(Haynes 2014) and dibasic calcium phosphate with a refractive

index of 1.55 (Haynes2014) as insoluble components The

refract-ive indices of the insoluble components of APP123 are 1.54 for

magnesium carbonate (Haynes2014) and 1.46 for microcrystalline

cellulose (Sultanova et al 2013), and by that only slightly below the values for APP117 The higher particle size can therefore be regarded as one of the reasons for the inferior opacity of APP123 The quantitative composition of APP123 and APP117 is not known, so more precise conclusions are not possible

Due to the generally known dependence of opacity on particle size, the data shown could only be partially explained In addition

to particle size, other factors, such as refractive index, surface properties, particle spacing, and special features, such as the band gap of the ZnO described above must also be taken into account

It should also be noted that the pigment content in the coatings investigated could not always be kept constant Above all, the high opacity of TiO2 and the low opacity of APP123 can be explained by the large differences in particle size

Effect on photostability

Many APIs show a light sensitivity due to their photo reactivity (Albini and Fasani 1998) The European Pharmacopeia requires light protection for more than 250 APIs This sensitivity to light poses a challenge both for formulation development and for the manufacturing process of pharmaceuticals If it is a tablet, light protection can be ensured by an opaque coating If such a coat-ing is not applied, light protection is only guaranteed by the packaging Here, no permanent light protection can be guaran-teed, especially during the handling of the dosage form by the patient The light protection provided by the four different coat-ings used in this article was investigated using the model drug nifedipine

The nifedipine tablets were coated with a mass gain of 7% The coated tablets were then stored under UV light (315–400 nm) The content of degradation products was determined after 2 and

4 weeks (Figure 5)

While the European Pharmacopoeia requires that concentra-tions of degradation products are limited to 0.1% of the nifedi-pine content, reference tablets without any coating showed concentrations of 0.16% (impurity A) and 3.33% (impurity B) after

2 weeks (Figure 5) After 4 weeks the levels increased to 0.24 and 4.02% Tablets stored in the dark showed concentrations of 0.01 and 0.00% after 4 weeks

Regarding impurity A, ZnO-containing showed the lowest con-centrations after 2 as well as after 4 weeks For both points in time, the concentrations were below 0.1% and the requirements

of the European Pharmacopeia therefore fulfilled TiO2 showed similar results as ZnO for impurity B and slightly higher concentra-tions with higher variability for impurity A

The tablets coated with APP117 and APP123 exceeded the specified limits both after 2 and after 4 weeks Compared to the Figure 4 50% and 10% quantiles of the particle size distributions determined by laser diffraction.

other two coatings, more nifedipine was converted to impurity A,

whereby the content of the tablets coated with APP117 after

2 weeks was at comparable levels as the un-coated

refer-ence tablets

The conversion of nifedipine into impurity B in both the TiO2

-and the ZnO-containing coating was completely prevented during

the first 2 weeks, as the concentrations were below the limit of

detection After 4 weeks under light irradiation, however, values

close to or just above 0.1% were also achieved here This does

exceed Pharmacopeia limitation, but it should be considered that

extreme conditions were chosen here Furthermore, comparison

to the other coatings clearly shows significant differences APP117

and APP123 could not sufficiently protect the active substance

against photo-induced conversion Here, more than 2% of the

nifedipine was converted to impurity B, so that the conversion to

it could only be slightly reduced compared to the tablets that

were not coated There was no difference in the determined

con-tents of impurity B after 2 and after 4 weeks

The best light protection was achieved by the ZnO-containing

coating In addition, this coating was the only one that could

suf-ficiently prevent the conversion of nifedipine into impurity A This

can be explained by the pronounced photoprotective property of

ZnO, which is particularly pronounced in the wavelength range of

UVA radiation (320–400 nm) (Smijs and Pavel 2011) This

wave-length range corresponds to the wavelength range of

315–400 nm used for the experiments shown TiO2also has a

pho-toprotective character, which is mainly observed in the UVB

radi-ation range (290–320 nm) (Smijs and Pavel 2011) TiO2 and ZnO

are often found in sunscreens In that context a higher protection

by ZnO compared to TiO2 in the UV range has already been

described in the literature (Pinnell et al.2000)

The protection of light-sensitive APIs by a TiO2-containing

coating was already investigated in a study by Bechard et al.,

which examined the light protection by an HPMC-based coating

(TiO2 content: 29.5% w/w) with different mass gains (Bechard et

al.1992) The coated cores contained 12% (m/m) nifedipine They

were coated with mass applications of 2, 4, 6, 10, and 15% and

then irradiated with white fluorescent light (12 15 W) in a light

cabinet A clear coating without the addition of a pigment

showed no light protection, there was no difference in the

con-version of nifedipine compared to the uncoated cores In their

study, nifedipine showed an initially high decomposition rate in

the first 2–3 d This is in agreement with the results of this study,

in which most of the degradation took place before the first

sampling point Even with TiO2concentrations up to 29.5%, a film thickness between 24 and 68mm was not sufficient to protect nifedipine from light degradation The film thickness was deter-mined as a key variable for the light protection of nifedipine Good light protection was provided with a film thickness of

145mm In this study, a good light protection of nifedipine was achieved with lower TiO2concentrations (16.2% w/w) and a lower film thickness (90mm)

Inline monitoring using Raman spectroscopy

To test the applicability of inline monitoring using Raman spec-troscopy for the TiO2-free alternative coatings, PLSR models were built using the data of the four calibration data sets For each coating formulation a prediction model was built which was used

to predict the applied coating mass of the corresponding test data set Model building and prediction parameters are given in

Table 5 A number of two or three components were used for model building

The PLSR model for the prediction of the TiO2-containing coat-ing showed the smallest calibration error (0.37%) and the highest

R2 (0.9998) This model resulted in the smallest prediction error under 1% As shown in the observed vs predicted plot, the applied coating mass was predicted very precisely (Figure 6) With values between 1.09 and 2.34% the PLSR models of the alterna-tive coating formulations showed higher calibration errors In add-ition, the R2

values were smaller Here, the APP117 prediction model showed the most promising calibration parameters with a RMSEC of 1.09% and aR2

of 0.9986 This model was able to pre-dict the application of APP117 with an acceptable RMSEP of 2.06% However, the predicted coating mass showed higher devi-ations from the observed coating mass compared to the TiO2 -con-taining coating (Figure 6)

With an RMSEP of 6.97%, it was not possible to build a reliable PLSR model for the ZnO-containing coating As shown in Figure

6, the applied coating mass was overestimated during the entire

Figure 5 Concentration of degradation products after 2 and 4 weeks storage under UV light.

Table 5 PLSR-model building and prediction parameters.

coating process Also, the application of APP123 could not be

pre-dicted with an acceptable prediction error In the beginning and

at the end of the process the applied coating mass was

underesti-mated, which led to an RMSEP of 3.82% The results can be

explained with regard to the spectral changes during the coating

process As TiO2 is a strong Raman marker, the related peaks

showed a distinctly and linear intensity increase during the

coat-ing process Also, the increascoat-ing intensity of the CaCO3peaks

dur-ing the application of APP117 enabled a model builddur-ing with an

acceptable predictive ability Compared to TiO2, CaCO3 showed a

lower Raman intensity which resulted in smaller changes in the

Raman spectra during the process and explains the comparatively

higher prediction error During the application of the

ZnO-con-taining coating and APP123 no sufficient spectral changes were

obtained to build a reliable PLSR model

Conclusion

The TiO2-free white coatings used in this study are inferior to the

TiO2-containing white coating in some of the investigated

proper-ties Especially with regard to opacity and appearance, no

alterna-tive white coating could achieve a similar result as the TiO

-containing coating TiO2was superior regarding efficiency as well

as effectiveness, i.e the achieved opacity was the highest and was reached using the minimal mass gain The CaCO3 and CaHPO4-based coating were the second-best regarding both crite-ria Regarding the protection of a photosensitive API, ZnO seemed

to be a suitable alternative to TiO2 while CaCO3-, MgO-, and CaHPO4-based coatings could not protect a photosensitive API in the coated tablet at increased light exposure, which led to increased degradation

In-line process control of the coating processes of all four coat-ings used by means of Raman spectroscopy showed that replac-ing TiO2would be a major challenge Acceptable prediction errors could only be achieved for one of the TiO2-free alternatives (APP117) But also here the prediction errors were in a signifi-cantly higher range compared to the TiO2-containing coating due

to the reduced Raman activity of the pigment Difference spectra and the implementation of a moving average could increase the prediction capability of PLSR models However, this was only the case, if the applied coating showed sufficient changes in the Raman spectra over process time

In conclusion, the presented data demonstrate that some alter-native pigments – especially the combination of CaCO and Figure 6 Coating mass predicted by PLSR plotted against observed coating mass.