1 Biorelevant dissolution testing of a weak base: Interlaboratory

2 reproducibility and investigation of parameters controlling in vitro

3 precipitation
4 Philippe Berben1, Lee Ashworth2, Stefania Beato3, Jan Bevernage4, Jean-Luc Bruel5, James Butler6,
5 Jennifer Dressman7, Kerstin Schäfer8, Paul Hutchins9, Lukas Klumpp7, James Mann2, Janine
6 Nicolai10, Krista Ojala11, Sanjaykumar Patel12, Sarah Powell2, Karin Rosenblatt13, Irena
7 Tomaszewska14, James Williams6, Patrick Augustijns1.

8
1
9 Drug Delivery and Disposition, University of Leuven (Leuven, Belgium)
2
10 Pharmaceutical Technology & Development, AstraZeneca (Macclesfield, United Kingdom)
3
11 Technical Research and Development, Novartis Pharma AG (Basel, Switzerland)
4
12 Pharmaceutical Sciences, Janssen Pharmaceutica, Johnson & Johnson (Beerse, Belgium)
5
13 Research & Development, Sanofi-Aventis (Vitry-sur-Seine, France)
6
14 GlaxoSmithKline R&D, Pharmaceutical Development & Supply (Ware, UK)
7
15 Institute of Pharmaceutical Technology, Goethe University (Frankfurt am Main, Germany)
8
16 Pharmaceutical Development, Boehringer Ingelheim Pharma GmbH (Biberach an der Riß, Germany)
9
17 Drug Product Science and Technology, Bristol-Myers Squibb (Moreton, UK)
10
18 Chemical and Pharmaceutical Development, Bayer AG (Wuppertal, Germany)
11
19 Pharmaceutical Sciences, Analytical Development, Orion Pharma (Turku, Finland)
12
20 Analytical Sciences, MRL, Merck & Co., Inc. (Kenilworth NJ, USA)
13
21 Drug Product Development, AbbVie Deutschland GmbH & Co. KG (Ludwigshafen, Germany)
14
22 Analytical Research and Development, Pfizer Ltd. (Sandwich, UK)

23
24
25 Corresponding author:

26 Patrick Augustijns

27 Drug Delivery and Disposition, KU Leuven, Gasthuisberg O&N II, Herestraat 49 – box 921, 3000 Leuven,

28 Belgium

29 Tel: +32 16 33 03 01 Fax: +32 16 33 03 05

30 Patrick.augustijns@kuleuven.be

1
31 KEY WORDS
32 Indinavir sulfate

33 Dissolution testing

34 USP 2 Apparatus

35 OrBiTo

36 Interlaboratory reproducibility

37 Supersaturation/precipitation

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

2
62 ABSTRACT
63 Following a previous study which aimed to determine the interlaboratory reproducibility of biorelevant

64 dissolution testing in the USP 2 apparatus for commercial formulations of two weak acids (ibuprofen

65 and zafirlukast), this study attempts to determine the interlaboratory reproducibility using a similar

66 protocol for a commercially available formulation of a weak base, indinavir. Fourteen partners

67 including twelve industrial and two academic partners participated in this study. To ensure uniformity,

68 all partners were provided with a standardized protocol to perform (i) a single medium dissolution test

69 in fasted state simulated gastric and intestinal fluids (FaSSGF and FaSSIF, respectively) and (ii) a two-

70 stage dissolution experiment simulating gastrointestinal transfer. Optionally, partners could run a

71 single-stage dissolution test in fed state simulated intestinal fluid (FeSSIF). For each dissolution test,

72 one Crixivan® capsule (containing 400 mg indinavir as its sulfate salt) was added as dose of interest.

73 For the single medium dissolution test in FaSSIF, all partners observed rapid release of indinavir

74 resulting in supersaturated concentrations, followed by precipitation to equilibrium solubility. The

75 degree and period of supersaturation varied among the participating laboratories. Average dissolution

76 profiles in FeSSIF appeared to be highly reproducible with dissolved concentrations remaining lower

77 than the thermodynamic solubility of indinavir in FeSSIF. For the two-stage dissolution test, most

78 partners observed supersaturated concentrations in the intestinal compartment; two partners

79 observed no supersaturation due to immediate precipitation. Given the fact that a high interlaboratory

80 but low intralaboratory variability was observed when supersaturation/precipitation occurred, an

81 undefined factor was hypothesized as a potential cause of the variability in precipitation. Hence, the

82 impact of several experimental factors on the supersaturation and precipitation behavior of indinavir

83 was investigated in a next step.

84 The investigation indicated that variability is likely attributable to a combination of factors, especially,

85 the time elapsed between sampling and dilution of the sample with the mobile phase. Therefore, when

3
86 designing a test in which supersaturation and precipitation is anticipated, stringent control of the test

87 methodology, especially regarding sampling and dilution, is needed.

88

89

90

91

92

93

94

95

96

97

4
 98 1. INTRODUCTION
99 Despite dissolution traditionally being used for quality control (QC) purposes, the application of

100 dissolution to predict the in vivo biopharmaceutical performance of oral dosage forms is gaining

101 considerable interest [1]. For example, regulatory authorities like the European Medical Agency (EMA)

102 advise the use of USP 1 and 2 dissolution test apparatus as a safe and reliable strategy for biowaiver

103 testing of immediate release (IR) formulations of BCS class I and III drugs [2] since these dissolution

104 tests are easy-to-handle and robust. Additionally, formulation scientists often use these dissolution

105 tests during the development of new drug candidates in order to select the formulation strategy with

106 the highest potential for reaching the market [3–5]. Although a multitude of intrinsic and extrinsic

107 parameters may affect the dissolution profile, it is crucial that test results can be integrated into the

108 drug development process with confidence, so the optimal formulation can be selected without

109 compromising in vivo performance or reducing development timelines [6].

110 Recently, a study among 16 partners participating in the European project OrBiTo (Oral

111 Biopharmaceutics Tools) was performed to evaluate the interlaboratory reproducibility of dissolution

112 testing in the paddle apparatus (USP 2) with biorelevant media; commercially available ibuprofen

113 tablets and zafirlukast tablets were selected as formulations of interest [7]. For both drugs, a high

114 degree of reproducibility was observed in terms of the dissolution profile for both a single medium and

115 a two-stage dissolution test. Since both drugs are weak acids, release is slow in the acidic environment

116 used in the first stage to represent the stomach but fast in the neutral medium used in the second

117 stage to represent the small intestine; supersaturation and precipitation were neither expected nor

118 observed. By contrast, the pH-shift due to transfer from the stomach to the small intestine may trigger

119 supersaturation of weak base drugs, which may result in precipitation. Under these circumstances,

120 attaining reproducible results in a representative dissolution test is more challenging.

121 Over the past 15 years, several in vitro setups have been developed to evaluate supersaturation and

122 precipitation, ranging from relatively simple to rather complex tools. For example, Kostewicz et al.

5
123 developed a simple transfer model by connecting two USP 2 vessels with a peristaltic pump for the

124 prediction of intestinal supersaturation and precipitation [4]. On the other hand, the TNO intestinal

125 model 1 (TIM-1) is a more advanced tool allowing high-quality prediction of the in vivo behavior of

126 orally administered drugs [8]. Given the complexity of supersaturation and precipitation, multiple

127 factors may affect this interplay in vivo including gastric emptying rate, gastrointestinal motility,

128 permeability, the physicochemical properties of the drug, and the volume and composition of

129 gastrointestinal fluids [9]. Since simple dissolution tests do not succeed to capture all these dynamic

130 processes, it is highly advised to combine in vitro data with physiologically-based pharmacokinetic

131 (PBPK) models for a more accurate prediction of the PK profile of oral dosage forms [10,11]. In the case

132 of weak bases, these software programs typically rely on the input from in vitro supersaturation and

133 precipitation experiments in combination with human physiology data and physicochemical properties

134 of the drug. As such, reliability of the in vitro methods must be guaranteed to accurately predict the in

135 vivo drug behavior.

136 Using the same approach as the previous study in which the interlaboratory reproducibility of two

137 weak acids was assessed, the current study aimed to evaluate the robustness of a biorelevant

138 dissolution test in the USP 2 apparatus for a weak base. Participating partners were required to

139 perform (i) a single medium dissolution test in FaSSGF and FaSSIF (and optionally in FeSSIF) and (ii) a

140 two-stage dissolution test simulating the gastrointestinal transfer. Crixivan®, containing indinavir

141 sulfate, was selected as the formulation of interest since the gastrointestinal behavior of indinavir has

142 been recently elucidated by administering Crixivan® capsules to human healthy volunteers in the

143 fasted state [12]. Fourteen partners participated in this cross-laboratory study including AbbVie

144 Deutschland GmbH & Co. KG, AstraZeneca, Bayer AG, Boehringer Ingelheim Pharma GmbH, Bristol-

145 Myers Squibb Pty. Ltd., GlaxoSmithKline, Goethe University, Janssen Pharmaceutica, KU Leuven, Merck

146 & Co. Inc., Novartis Pharma AG, Orion Pharma, Pfizer Ltd. and Sanofi-Aventis.

6
147 2. MATERIALS AND METHODS

148 2.1. Selection of the formulation of interest

149 Indinavir is a weak base with pKa values of 3.7 and 5.9. While the solubility of indinavir is very high in

150 an acidic environment (> 99.4 g/L at pH < 3.5), its solubility in a neutral environment is relatively low

151 (0.048 g/L in FaSSIF, pH 6.5) [13,14]. According to the criteria of the Biopharmaceutics Classification

152 System (BCS), indinavir has been reported to be either a class II (low solubility, high permeability) or

153 class IV (low solubility, low permeability) drug, as the intestinal permeability of indinavir sits on the

154 border of the permeability criterion of the BCS. Since indinavir shows poor aqueous solubility, the drug

155 is formulated as a sulfate salt (Crixivan®, 400 mg indinavir) for which superior bioavailability has been

156 demonstrated compared to the free base [15]. Furthermore, a clinical trial has been recently

157 performed to elucidate the intraluminal behavior of Crixivan® [12].

158 2.2. Chemicals and media

159 All participating partners were initially provided with a small amount of pure indinavir sulfate and with

160 commercially available capsules of indinavir sulfate (Crixivan®) originating from the same batch

161 (L006343). Subsequent investigations of the source of variability were conducted using a different

162 capsule batch (N006110). Each partner was expected to purchase the remaining material and

163 chemicals independently implying that, most likely, different filters and different batches of

164 FaSSIF/FeSSIF/FaSSGF powder were used.

165 Biorelevant media were prepared according to instructions provided by Biorelevant.com. Fasted state

166 simulated gastric fluid (FaSSGF) was obtained by adding FaSSIF/FeSSIF/FaSSGF powder (0.06 mg/mL)

167 to blank FaSSGF (1.99 mg/mL NaCl, adjusted to pH 1.6 with 1 M HCl). Fasted state simulated intestinal

168 fluid (FaSSIF) was prepared by dissolving FaSSIF/FeSSIF/FaSSGF powder (2.24 mg/mL) in blank FaSSIF

169 (3.95 mg/mL NaH2PO4.H2O, 6.19 mg/mL NaCl, 0.42 mg/mL NaOH, adjusted to pH 6.5) while double-

170 concentrated FaSSIF (2x FaSSIF) was obtained by dissolving double amounts of NaH2PO4.H2O, NaCl,

171 NaOH and FaSSIF/FeSSIF/FaSSGF powder compared to normal FaSSIF, and adjusted to pH 7.5 with 2M

7
172 of NaOH, so that, after addition of 250 mL of 2x FaSSIF to 250 mL of FaSSGF, the final pH amounts to

173 6.5. Fed state simulated intestinal fluid (FeSSIF) was prepared by dissolving FaSSIF/FeSSIF/FaSSGF

174 powder (11.2 mg/mL) in blank FeSSIF (4.04 mg/mL NaOH, 8.65 mg/mL glacial acetic acid, 11.87 mg/mL

175 NaCl, adjusted to pH 5.0). Media were equilibrated prior to use to allow for micelle particle size

176 stabilization and were considered to be stable for 48h at room temperature [16].

177 2.3. Protocols to perform dissolution tests

178 All partners were asked to perform two types of dissolution tests with Crixivan® including (i) a single

179 medium dissolution test in FaSSGF and FaSSIF (and optionally in FeSSIF) and (ii) a two-stage dissolution

180 test simulating the gastrointestinal transfer. Below, the protocols for the single medium (cfr. 2.3.1.)

181 and two-stage (cfr. 2.3.2.) dissolution test are described as they were shared with the participating

182 partners.

183 2.3.1. Single-stage biorelevant dissolution

184 Single-stage biorelevant dissolution tests were performed according to the following standardized

185 protocol:

186  Apparatus: the USP 2 Apparatus (rotating paddle) was used with a paddle speed of 75 rpm,

187 at 37 °C, using a volume of 250 mL for FaSSGF, and 500 mL for FaSSIF and FeSSIF.

188  Sampling time points: samples were taken after 5, 10, 20 and 30 min, and (if the product

189 proved to dissolve slowly) every subsequent 30 min until a plateau was reached. Partners

190 were asked to add extra early time point(s) if plateau was reached within 10 minutes or

191 less.

192  Dose: protocol instructions indicated to mimic the dose of interest in humans, even if this

193 means adding multiple dose units to the vessel. In this case, dissolution experiments were

194 performed using one capsule of Crixivan®. Instructions were to run at least 3 replicates for

195 each type of experiment; most partners ran six replicates.

8
196  Sampling and analysis: although sampling methodology was allowed to vary among the

197 partners, it was recommended to use manual sampling via syringe (5 mL, from which an

198 approximate 0.5 mL analysis sample was taken), followed by analysis using a method

199 appropriate for the drug. To this end, the first 2-3 mL was passed through the filter and

200 any excess of the sample in the syringe was to be returned to the vessel. Moreover,

201 samples should be appropriately diluted with mobile phase and analyzed on the same day,

202 unless stability could be confirmed over a longer period. Other sampling and analysis

203 methods (e.g. semi-automated sampling, UV detection) were allowed at the discretion of

204 the partner provided they were compatible with the sample properties. Issues to be

205 considered included sample precipitation prior to analysis during semi-automated

206 sampling and the potential impact of variable background UV absorbance if using UV

207 detection without HPLC.

208  Filtration: to separate undissolved particles from dissolved particles, a suitable filter

209 should be selected, with a check for adsorption of the drug onto the filter. It was not

210 allowed that the adsorbed fraction exceeded 2%. If a larger fraction was adsorbed, then

211 this should be clearly stated in the report. Typically, a PTFE, GHP or glass fiber filter was

212 used (pore size 0.45 µm, diameter 25 mm). In addition, if there was an excess of

213 undissolved excipient/drug present, it was advised to use a pre-filter to avoid filter

214 clogging.

215  Product specific modifications to the method: if supersaturation is likely in the samples

216 taken, then samples should be diluted immediately after sampling to avoid (further)

217 precipitation.

218 The partners were also advised that, when using a volume of 250 mL in a standard 1 L dissolution

219 vessel, the use of bathless dissolution vessels should be avoided due to heating zone issues.

220 Additionally, partners were informed that the paddles are very close to the medium surface at this

221 (lower than typical) volume implying that sampling challenges might result in some variability.

9
222 2.3.2. Two-stage biorelevant dissolution (representing the gastrointestinal transfer)

223 Two-stage biorelevant dissolution tests were performed according to the following standardized

224 protocol:

225  Apparatus: the USP 2 Apparatus (rotating paddle) was used with a paddle speed of 75 rpm,

226 at 37 °C, using an initial volume of 250 mL for FaSSGF (pH 1.6) for 30 min. Subsequently,

227 250 mL of 2x FaSSIF (pH 7.5) was added to end up with a final volume of 500 mL of normal

228 FaSSIF (pH 6.5).

229  Sampling time points: during the gastric stage, samples were taken after 5, 10, 20 and 30

230 min. Following the addition of 2x FaSSIF, additional samples were taken after 35, 40, 50

231 and 60 min, and (if slowly dissolving) every subsequent 30 min until a plateau was reached.

232 It was recommended to add extra early time point(s) if the plateau was reached within 10

233 min or less after addition of 2x FaSSIF.

234  Dose selection, sampling, filtration and analysis were performed as described for the

235 single-stage dissolution test. Any product-specific modifications to the methodology were

236 expected to be clearly reported.

237 2.4. Variations from the protocol

238 As the protocols provided some flexibility in the experimental design, there were several instances in

239 which variations on the protocol occurred or a partner chose to explore some of the parameters that

240 were not explicitly specified in the protocol. For instance, although the protocol intended that the

241 media should be preheated before use, one partner ran the studies without preheating the media.

242 Despite the sampling time points being clearly defined in the protocol, some partners took samples at

243 additional time points or stopped the experiment earlier than instructed. In addition, one partner used

244 a semi-automated UV system, meaning that indinavir concentrations were measured online, without

245 a dilution step. Furthermore, the protocol clearly stated that if supersaturation is likely to occur (which

246 is the case for indinavir), samples should be diluted immediately after sampling to stop ongoing

10
247 precipitation. Nevertheless, the time interval between taking and diluting the samples appeared highly

248 variable among the different partners (as discussed below).

249 2.5. Analytical methods

250 To determine the concentrations of indinavir, most partners developed an in-house HPLC or UPLC

251 method whilst one partner used a semi-automated UV system. Exact details of the analytical methods

252 from the individual partners can be obtained upon request. A typical method for the HPLC analysis of

253 indinavir is shown in Table 1.

254 Table 1. Typical conditions for the HPLC analysis of indinavir

HPLC WatersTM 600 Controller

Column Novapak C18 (4 µm, 8 mm x 100 mm, Waters)

Mobile phase Acetate buffer (25 mM)/methanol (25/75)

pH mobile phase 3.5

Flow rate 1 mL/min

Detection UV: 260 nm or Fluo: Ex: 256 nm/ Em: 290 nm

Injection volume 50 µL

Retention time 5.1 min

Analysis time 8 min

255

256

257

258

259

11
260 3. RESULTS AND DISCUSSION

261 3.1. Single-stage dissolution profiles

262 The average dissolution profiles of indinavir in FaSSGF from all partners are depicted in Figure 3 as the

263 first 30 min of the two-stage dissolution test reflect the dissolved fraction in FaSSGF. In this acidic

264 environment, a fast dissolution of Crixivan® was observed whereby almost 50% of the drug was

265 dissolved after 5 min; in most cases, over 90% of indinavir was dissolved within 10 min. Furthermore,

266 dissolved concentrations of indinavir were highly reproducible among the different participating

267 partners (e.g. RSD = 5.7% after 30 min). Most likely, the fast and complete dissolution of indinavir are

268 the result of its excellent solubility in the acidic environment (>99.4 g/L at pH<3.5) and the immediate

269 release properties of the formulation. Similar behavior has been demonstrated in the fasted stomach

270 of human healthy volunteers following the oral administration of a single capsule of Crixivan® [12].

271 In contrast to the gastric medium, indinavir is predominantly unionized at the pH of the intestinal

272 environment. Together with the partition coefficient (LogDpH 6.5 2.4), this makes it likely that indinavir

273 shows a high affinity for colloidal structures such as bile salts and phospholipids, as previously

274 demonstrated by Holmstock et al. [17]. Biorelevant media, containing these colloidal structures, are

275 therefore recommended for dissolution testing of indinavir rather than simple pharmacopeial buffers.

276 As a result, partners were asked to perform a dissolution test in FaSSIF and optionally in FeSSIF. Figure

277 1 depicts the average dissolution profiles from all partners in FaSSIF. Every partner observed

278 concentrations of indinavir higher than its thermodynamic solubility in FaSSIF (dotted line), indicating

279 that the sulfate salt of indinavir enables generation of supersaturation during dissolution in FaSSIF. In

280 all cases, the supersaturated indinavir levels induced precipitation. As a wide range of maximum

281 concentrations (Cmax) were observed (maximum dissolved fraction ranging from 29.1 % - 72.9 %), the

282 degree of supersaturation was highly variable among the different participating laboratories. In

283 addition, most partners observed a Cmax after 10 min; however, four partners measured a maximum

284 concentration of indinavir at the first sampling time point (5 min).

12
285
286 Figure 1: Dissolution of indinavir from Crixivan® 400 mg capsules in FaSSIF. Each colored curve represents the
287 results from one partner (mean ± SD, n=3-6). The dotted line represents the dissolved fraction corresponding to
288 the thermodynamic solubility of indinavir in FaSSIF.

289

290 As the dissolution test in FeSSIF was optional, only four partners performed a single-stage dissolution

291 test in simulated intestinal medium representing the fed state. Figure 2 illustrates that all partners

292 measured a nearly complete dissolution of indinavir from its formulation 30 min (RSD = 7.6%) after the

293 initiation of the dissolution test. Dissolution profiles were highly reproducible among the four

294 participating partners. The high reproducibility in this experiment may be linked to the lack of

295 supersaturation, and thus precipitation: the theoretical maximum concentration of one capsule

296 Crixivan® (0.8 g/L) in 500 mL FeSSIF is lower than the thermodynamic solubility of indinavir in FeSSIF

297 (1.36 g/L).

13
298
299 Figure 2: Dissolution of indinavir from Crixivan® 400 mg capsules in FeSSIF. Each colored curve represents the
300 results from one partner (mean ± SD, n=3-6). The dotted line represents the dissolved fraction corresponding to
301 the thermodynamic solubility of indinavir in FeSSIF.

302

303 3.2. Two-stage dissolution test

304 In addition to the single medium dissolution experiments, a two-stage dissolution test was performed

305 simulating the gastrointestinal transfer, i.e. from the acidic environment of the stomach to the neutral

306 environment of the small intestine. The average dissolution profiles for fourteen different partners are

307 illustrated in Figure 3. Since one partner repeated the experiments, fifteen different dissolution

308 profiles were obtained. As previously described (cfr. 3.1.), all partners observed almost complete

309 dissolution of the weak base indinavir in the acidic environment. Upon addition of 250 mL of 2x FaSSIF

310 (pH 7.5), dissolution profiles diverged. Most partners (13 out of 15) clearly observed supersaturated

311 concentrations of indinavir following the drastic shift in pH; however, the degree and time span of the

312 intestinal supersaturation varied substantially among the participating partners. For instance, two

313 partners observed extensive supersaturation for more than 20 min while two other partners measured

314 extensive intestinal precipitation immediately upon the gastrointestinal transfer. One hour after the

315 pH-shift, all partners measured indinavir concentrations (RSD = 37.5%) approaching its thermodynamic

316 equilibrium solubility in FaSSIF (0.048 g/L) which corresponds to 5.2% of the administered dose (see

317 Figure 3, dotted line) [13].

14
318 Even though all partners were provided with standardized protocols to perform the dissolution

319 experiments, interlaboratory reproducibility was relatively poor (e.g. RSD = 64.0% and 92.5% at 40min

320 and 50 min, respectively), particularly for time points where supersaturation/precipitation (range 30-

321 90 min) took place. Given the complexity of this interplay, there are several factors which could explain

322 the interlaboratory variability. Moreover, vessel-to-vessel data for each partner were rather consistent

323 suggesting that the protocol was interpreted consistently by each partner but not among partners.

324 This indicates that the in vitro precipitation may be affected by one or more uncontrolled factors.

325 Therefore, further experiments were conducted to unravel the impact of various individual factors on

326 the in vitro precipitation behavior of indinavir (cfr. section 3.3).

327
328 Figure 3: Dissolution of indinavir from Crixivan® 400 mg capsules in the two-stage dissolution test. Each colored
329 curve represents the results from one partner (mean ± SD, n=3-6). The dotted line represents the dissolved fraction
330 corresponding to the thermodynamic solubility of indinavir in FaSSIF.

331

332 In human healthy volunteers, intestinal precipitation of indinavir has been observed upon

333 gastrointestinal transfer [12]. The dissolution profiles from the two-stage dissolution test also

334 consistently showed precipitation, but it was generally more pronounced in vitro than in vivo. Since in

15
335 vivo, the supersaturated drug has the choice to precipitate or to permeate across the intestinal

336 epithelium, it is likely that in vivo absorption reduces intestinal supersaturation and subsequently the

337 driving force for precipitation. As a result, when assessing supersaturation in vitro, the precipitated

338 fraction might be overestimated in the absence of an absorptive sink compartment, as is the case for

339 the single medium run in FaSSIF and the two-stage dissolution test using the USP 2 dissolution

340 apparatus.

341 3.3. Experimental factors affecting in vitro precipitation

342 The relatively high variability in the intestinal dissolution profiles obtained in single- and two-stage

343 dissolution tests (see Figure 1 and 3) is most likely related to variations in precipitation from the

344 metastable state of supersaturation. Indeed, dissolution profiles in the media where no

345 supersaturation occurs (FaSSGF in Figure 3, FeSSIF in Figure 2), demonstrate good reproducibility

346 between partners. To better understand supersaturation/precipitation behavior in the USP 2

347 dissolution apparatus, it is necessary to identify the sources of the interlaboratory variability. Hence,

348 in further experiments, it was investigated whether slight changes in media pH, time of sample dilution

349 and the possibility of heterogeneous nucleation might contribute to the observed variability.

350 3.3.1. pH

351 When adding 2x FaSSIF of pH 7.5 to FaSSGF with a pH of 1.6, the pH of the final medium is expected to

352 decrease to 6.5, which is close to the second pKa of indinavir (5.9). Accordingly, a slight deviation from

353 pH 6.5 might significantly change the solubility and dissolution behavior of indinavir, and thus be a

354 potential source of variation. To test this hypothesis, the two-stage dissolution experiment was

355 repeated with 2x FaSSIF buffers containing a pH of 7.3 or 7.7. Using these slightly different buffers, the

356 pH after the pH-shift amounted to 6.33 and 6.70, respectively. However, the average dissolution

357 profiles appeared in close agreement to the reference experiment which used a 2x FaSSIF buffer with

358 a pH of 7.5, as illustrated in Figure 4. When 2x FaSSIF buffer was adjusted to pH of 6.5 corresponding

359 to FaSSIF itself, the dissolution/precipitation profile differed substantially, as depicted in Figure 4. In

16
360 this case, the pH of the final medium (4.4) was much lower than the pKa of indinavir and, due to a

361 marked increase in indinavir solubility at this pH (0.94 g/L at pH 4.5 compared to 0.048 g/L at pH 6.5),

362 no indinavir precipitation was observed upon the pH-shift (Figure 4, circles). Ultimately, these results

363 suggest that slight differences in starting buffer pH are not the main source of the variable

364 precipitation. Even so, it is recommended that the pH of 2x FaSSIF should be as accurate as possible.

365
366 Figure 4: Dissolution of indinavir in the two-stage dissolution test using 2x FaSSIF solutions with a different pH.
367 Data are presented as mean ± SD, n=3.

368

369 3.3.2. Sample dilution

370 The protocol clearly stated that, if supersaturation is likely to occur (which is the case for indinavir in

371 media simulating the intestinal environment), samples should be diluted immediately after collection

372 to circumvent further precipitation. Nevertheless, post-hoc questioning indicated that the time

373 elapsed between sample collection and dilution appeared to be highly variable among the different

374 partners, as listed in Table 2. Only five partners diluted their samples immediately after collection, as

375 required by the protocol. Five other partners diluted their samples within 10 to 30 min after sampling,

376 and two partners diluted their samples when the dissolution experiment was complete. Furthermore,

377 one partner did not perform a dilution step and one partner measured indinavir concentrations online,

378 using a semi-automated system. When filtered samples are not diluted immediately after collection,

379 further in-vial precipitation could take place, resulting in an underestimation of the indinavir

17
380 concentration in solution. Therefore, variations in the time span between sample collection and

381 dilution might contribute to the observed interlaboratory variability.

382 In fact, one partner who observed immediate precipitation upon the pH-shift (Figure 3, orange colored

383 curve) did not perform a dilution step, indicating that precipitation was indeed ongoing in the vial. The

384 other partner who recorded a similar dissolution profile did dilute the samples (Figure 3, red colored

385 curve), but only 15 minutes after sampling rather than immediately.

386 On the other hand, two partners observed extensive supersaturation for more than 20 min post pH-

387 shift (Figure 3, salmon pink and lilac colored curve). One of these partners measured concentrations

388 of indinavir using a semi-automated UV system, so it is possible that variable background UV

389 absorbance due to the presence of precipitated particles interfered with the UV signal. Recently, Jede

390 et al. proposed an in-line derivative spectroscopic method to measure drug concentrations online in

391 highly turbid samples [18]. This methodology could be considered as an interesting alternative as the

392 increased sampling frequency and lack of sample preparation allow better capturing of

393 supersaturation/precipitation.

394 Table 2. Overview of sample dilution conditions for each partner

Partner Dilution Length of time to dilute Preheat

1 No* N/A* Yes
2 Yes < 30 min Yes
3 Yes Immediate Yes
4 Yes** < 15 min Yes
5 Yes Immediate Yes
6 Yes Post run completion (<2 hours) No
7 Yes Post run completion (<2 hours) Yes
8 Yes < 30 min Yes
9 Yes Immediate Yes
10 Yes Immediate Yes
11 Yes Immediate Yes
12 Yes < 10 min Yes
13 Yes < 15 min Yes
14 No N/A Yes
395
396 *Online analysis using a semi-automated UV-system with no fixing dilution
397 **Fixing dilution at 35 and 40 min time points, only in the two-stage dissolution test

398

18
399 To gain further insight into the impact of the time span between sample collection and dilution,

400 additional experiments were performed in which the filtered samples were diluted at predetermined

401 time points including (i) immediately after sampling, (ii) after each time point, (iii) after the experiment,

402 (iv) 2 hours after the experiment, and (v) 24 hours after the experiment. This experiment revealed that

403 the time elapsed between sampling and dilution greatly affects the obtained dissolution/precipitation

404 profiles of indinavir. As depicted in Figure 5, this effect is more pronounced for samples in which

405 extensive supersaturation (35-50 min) was observed. For instance, in samples collected 10 min after

406 the pH-shift, the dissolved fraction of indinavir decreased 34% if samples were diluted after the

407 experiment instead of immediately; when filtered samples were only diluted 24 hours after the

408 experiment, a 63% decrease was observed. Although none of the partners waited 24 hours to dilute

409 their samples, these data suggest that immediate dilution is highly recommended to circumvent

410 further in-vial precipitation in samples with supersaturated concentrations. In-vial precipitation may

411 further cause chromatographic issues including blockage of the column or baseline and retention time

412 variability. This may introduce additional variability due to difficulties in selecting consistent

413 integration parameters. Furthermore, the protocol stated that, upon filtration, any excess in the

414 syringe should be returned to the vessel, thereby potentially inducing precipitation.

415 Overall, these data clearly indicate that it’s highly advisable to increase the level of the instructions of

416 the sampling methodology (sample dilution and preparation) in the standard operating protocols.

19
417
418 Figure 5: Impact of time span to dilute the collected samples on the apparent precipitation behavior of indinavir.
419 Data are presented as mean ± SD, n=3. Dotted line represents the dissolved fraction corresponding to the
420 thermodynamic solubility of indinavir in FaSSIF.

421

422 3.3.3. Heterogeneous nucleation

423 The crystal growth theory indicates that undissolved particles or impurities on the surface could induce

424 precipitation by acting as nuclei for crystal growth [19]. We therefore investigated the potential role

425 of heterogeneous nucleation (i.e., nucleation on impurity surfaces) in causing the observed

426 interlaboratory variability.

427 3.3.3.1. Apparatus material

428 In the study protocol, the apparatus material for paddles, shafts and sinkers was not explicitly

429 specified, making it a potential source of variability if other materials were used by the different

430 participating partners. For example, it has been reported that PTFE (polytetrafluoroethylene, Teflon)

431 might induce heterogeneous nucleation [20]. To investigate the effect of the apparatus material on

432 the dissolution/precipitation behavior of indinavir, an additional set of experiments was performed by

433 using instruments consisting of PTFE or stainless-steel. The dissolution/precipitation profiles were very

434 similar, as depicted in Figure 6, implying that the type of apparatus material could be excluded as a

435 potential source of variation in the indinavir results. Additionally, it would be useful for future

436 experiments to explore whether different filter materials and/or pressure forces could affect the

20
437 precipitation rate of indinavir. Therefore, when revising the standard operating protocols, specification

438 of the apparatus and filter material may help to achieve improved reproducibility in future

439 experiments.

440 Besides, it is known that the mechanism of hydrodynamics (shaking vs. stirring) can drastically affect

441 the precipitation rate of supersaturated drugs. Typically, stirring conditions overestimate the intestinal

442 motility while shaking conditions may better resemble the intestinal motility of the fasted state [21].

443 Indeed, the applied stirring speed helps to overcome the activation energy for nucleation, resulting in

444 a faster and a higher degree of drug precipitation compared to shaking conditions. However, it is

445 difficult to state if shaking conditions will result in an increased/decreased interlaboratory

446 reproducibility. For instance, the homogeneity within the vessel might be decreased, resulting in local

447 concentration gradients and thus a different supersaturation/precipitation event. A comparative study

448 between stirring and shaking conditions could thus be of interest for a future study.

449
450 Figure 6: Dissolution of indinavir from Crixivan® 400 mg capsules in the two-stage dissolution test in which
451 paddles, shafts and sinkers were made either completely of stainless steel or of PTFE. Data are presented as mean
452  SD, n=3.
453
454 3.3.3.2. Vessel cleanliness
455 In a next step, the impact of undissolved particles, possibly remaining from a previous experiment, on

456 the precipitation behavior of indinavir was investigated. To compare the obtained results

21
457 quantitatively, the area under the curve of supersaturation (SAUC) was calculated. To this end, the

458 average dissolution profile was plotted along with the thermodynamic equilibrium solubility of

459 indinavir at pH 6.5, as illustrated in Figure 7. Assuming that the pH changes immediately from 1.6 to

460 6.5 upon transfer, the solubility was set at 0.048 g/L, which is equivalent to 5.2% of the administered

461 dose [13]. SAUC values were determined using an AUC macro in Microsoft Excel (Figure 7; blue area)

462 for concentrations up to 60 minutes upon pH shift, as this was the crucial phase of indinavir

463 supersaturation and precipitation.

464

465 Figure 7: Example graph showing the calculation of SAUC (blue area) for a pH shift dissolution experiment.

466
467 Using the same dissolution equipment (Sotax AT7), it was noticed that repetition of the experiment

468 resulted in a decrease of the SAUC with each subsequent experiment, as illustrated in Figure 8 (black

469 bars). The mean SAUC values were determined to be statistically significant (p<0.05) for experiments

470 2, 3 and 4 compared to experiment 1 and between experiments 2 and 4. A reason with respect to the

471 higher SAUC value for experiment 1 might have been the naivety of the Sotax AT7 bath to indinavir.

472 Each subsequent experiment had an increased probability of remaining indinavir particles present in

473 the dissolution vessel, which could cause a seeding effect resulting in heterogeneous nucleation. The

474 Sotax AT7 bath is prone to this effect as it utilizes hollow shaft sampling with a coarse mesh on which

475 residues could remain undetected. Furthermore, an additional experiment was executed, in parallel

22
476 with the performance of experiment 4, where 2x FaSSIF was added one hour later than defined in the

477 protocol. It was observed that this prolonged residence time in FaSSGF resulted in a larger SAUC (14.2

478 instead of 9.8), implying that an increased time in FaSSGF, in which indinavir is highly soluble, might

479 decrease the extent of intestinal precipitation upon pH-shift.

25
S A U C ( m in m g /m L )

20

15

10

0
1 2 3 4 t 1 2
x x x x n k k
e
ta ta ta ta il te te
o o o o g is is
S S S S A D D
480

481 Figure 8: Mean SAUC values for the pH-shift dissolution of indinavir using different brands of USP 2 apparatus
482 including a Sotax AT7, Distek 2100C and Agilent 708-DS bath. All Sotax and Distek replicates were carried out on
483 the same instrument. Data are presented as mean  SD.

484
485 To further investigate the potential nucleation-inducing effect of retained indinavir particles, the two-

486 stage dissolution test was repeated using two other types of instruments known to be naïve to

487 indinavir: a Distek 2100C and an Agilent 708-DS. The SAUC values for both instruments were similar to

488 experiment 3 and 4 for the Sotax AT7 apparatus. Although the Agilent 708-DS was new and therefore

489 naïve for indinavir, it had been calibrated using the prednisone performance verification test (PVT),

490 directly prior to the experiment with indinavir. As the prednisone PVT results in an incomplete

491 dissolution, the presence of residual solid particles, which are not removed during cleaning

492 procedures, could have contributed to heterogeneous nucleation of indinavir. The Distek 2100C

493 apparatus also generated SAUC values similar to the Agilent 708-DS experiment. Although naïve to

494 indinavir, this apparatus had been used for various other projects. As it is uncommon practice to

495 conduct a wash of the system prior to starting a dissolution in the company which conducted this

23
496 follow-up experiments, heterogeneous nucleation by the presence of undissolved residues and/or dust

497 particles might have induced precipitation.

498 In general, it seems that vessel cleanliness might be a more important factor than naivety to the drug

499 in question with respect to in vitro precipitation. Vessels should be clean and free of retained

500 undissolved particles from previously performed experiments to properly conduct any

501 supersaturation/precipitation experiment. As the protocol did not provide specific cleaning guidelines

502 and differences across institutes exist with respect to the cleaning procedure (solvent, number and

503 frequency of wash cycles), it would be relevant to implement specific cleaning guidelines for

504 dissolution apparatus in the revised protocols to improve the interlaboratory reproducibility of a

505 straightforward precipitation test. Indeed, some partners washed their vessels using tap water while

506 other partners followed an in-house standardized cleaning procedure consisting of different wash

507 cycles using several solvents.

508 4. CONCLUSION
509 This study describes the results of cross-laboratory biorelevant USP 2 dissolution testing of a

510 commercially available formulation of indinavir, a weak base. Fourteen partners involved in the

511 OrBiTo-consortium participated in these studies. The average dissolution profiles in FaSSGF and FeSSIF

512 were shown to be highly consistent among partners. However, in testing where supersaturation and

513 precipitation occurred, i.e. in FaSSIF (due to the conversion of the salt back to the free base) and in the

514 two-stage test (due to the pH-shift), results were less reproducible. Post-hoc evaluation of various

515 parameters suggested that the reduced reproducibility in the experiments with

516 supersaturation/precipitation was likely attributable to deviations in the protocol with respect to the

517 time elapsed between sample collection and dilution, although other factors may also play a role. To

518 improve the reproducibility of dissolution tests in which supersaturation/precipitation occurs, it’s

519 necessary to improve the precision of the instructions in the protocols (e.g. controlled sampling,

520 dilution methodology, cleaning procedure) and strictly adhere to these when running the experiments

24
521 as these dissolution tests are widely used within pharmaceutical industry for different research and

522 quality control purposes.

523 5. ACKNOWLEDGEMENTS
524 This work has received support from the Innovative Medicines Initiative Joint Undertaking

525 (http://www.imi.europa.eu) under Grant Agreement No. 115369, resources of which are composed of

526 financial contribution from the European Union’s Seventh Framework Program and EFPIA companies’

527 in kind contribution.

528

529

530

531

532

533

534

535

536

537

538

539

540

541

25
542 6. REFERENCES
543 [1] E.S. Kostewicz, B. Abrahamsson, M. Brewster, J. Brouwers, J. Butler, S. Carlert, P.A. Dickinson, J. Dressman,
544 R. Holm, S. Klein, J. Mann, M. McAllister, M. Minekus, U. Muenster, A. Müllertz, M. Verwei, M. Vertzoni,
545 W. Weitschies, P. Augustijns, In vitro models for the prediction of in vivo performance of oral dosage
546 forms, Eur. J. Pharm. Sci. 57 (2014) 342–366. doi:10.1016/j.ejps.2013.08.024.
547 [2] European Medical Agency, 2010, (n.d.).
548 http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2010/01/WC500070039.
549 pdf (accessed April 10, 2018).
550 [3] A. Ruff, T. Fiolka, E.S. Kostewicz, Prediction of Ketoconazole absorption using an updated in vitro transfer
551 model coupled to physiologically based pharmacokinetic modelling, Eur. J. Pharm. Sci. Off. J. Eur. Fed.
552 Pharm. Sci. 100 (2017) 42–55. doi:10.1016/j.ejps.2016.12.017.
553 [4] E.S. Kostewicz, M. Wunderlich, U. Brauns, R. Becker, T. Bock, J.B. Dressman, Predicting the precipitation of
554 poorly soluble weak bases upon entry in the small intestine, J. Pharm. Pharmacol. 56 (2004) 43–51.
555 doi:10.1211/0022357022511.
556 [5] L. Di, B. Feng, T.C. Goosen, Y. Lai, S.J. Steyn, M.V. Varma, R.S. Obach, A perspective on the prediction of
557 drug pharmacokinetics and disposition in drug research and development, Drug Metab. Dispos. Biol. Fate
558 Chem. 41 (2013) 1975–1993. doi:10.1124/dmd.113.054031.
559 [6] I. Kola, J. Landis, Can the pharmaceutical industry reduce attrition rates?, Nat. Rev. Drug Discov. 3 (2004)
560 711–715. doi:10.1038/nrd1470.
561 [7] J. Mann, J. Dressman, K. Rosenblatt, L. Ashworth, U. Muenster, K. Frank, P. Hutchins, J. Williams, L.
562 Klumpp, K. Wielockx, P. Berben, P. Augustijns, R. Holm, M. Hofmann, S. Patel, S. Beato, K. Ojala, I.
563 Tomaszewska, J.-L. Bruel, J. Butler, Validation of Dissolution Testing with Biorelevant Media: An OrBiTo
564 Study, Mol. Pharm. 14 (2017) 4192–4201. doi:10.1021/acs.molpharmaceut.7b00198.
565 [8] M. Verwei, M. Minekus, E. Zeijdner, R. Schilderink, R. Havenaar, Evaluation of two dynamic in vitro models
566 simulating fasted and fed state conditions in the upper gastrointestinal tract (TIM-1 and tiny-TIM) for
567 investigating the bioaccessibility of pharmaceutical compounds from oral dosage forms, Int. J. Pharm. 498
568 (2016) 178–186. doi:10.1016/j.ijpharm.2015.11.048.
569 [9] B. Hens, S.M. Pathak, A. Mitra, N. Patel, B. Liu, S. Patel, M. Jamei, J. Brouwers, P. Augustijns, D.B. Turner, In
570 Silico Modeling Approach for the Evaluation of Gastrointestinal Dissolution, Supersaturation, and
571 Precipitation of Posaconazole, Mol. Pharm. 14 (2017) 4321–4333.
572 doi:10.1021/acs.molpharmaceut.7b00396.
573 [10] E.S. Kostewicz, L. Aarons, M. Bergstrand, M.B. Bolger, A. Galetin, O. Hatley, M. Jamei, R. Lloyd, X. Pepin, A.
574 Rostami-Hodjegan, E. Sjögren, C. Tannergren, D.B. Turner, C. Wagner, W. Weitschies, J. Dressman, PBPK
575 models for the prediction of in vivo performance of oral dosage forms, Eur. J. Pharm. Sci. Off. J. Eur. Fed.
576 Pharm. Sci. 57 (2014) 300–321. doi:10.1016/j.ejps.2013.09.008.
577 [11] N. Kaur, A. Narang, A.K. Bansal, Use of biorelevant dissolution and PBPK modeling to predict oral drug
578 absorption, Eur. J. Pharm. Biopharm. Off. J. Arbeitsgemeinschaft Pharm. Verfahrenstechnik EV. 129 (2018)
579 222–246. doi:10.1016/j.ejpb.2018.05.024.
580 [12] J. Rubbens, J. Brouwers, J. Tack, P. Augustijns, Gastrointestinal dissolution, supersaturation and
581 precipitation of the weak base indinavir in healthy volunteers, Eur. J. Pharm. Biopharm. Off. J.
582 Arbeitsgemeinschaft Pharm. Verfahrenstechnik EV. 109 (2016) 122–129. doi:10.1016/j.ejpb.2016.09.014.
583 [13] P. Augustijns, B. Wuyts, B. Hens, P. Annaert, J. Butler, J. Brouwers, A review of drug solubility in human
584 intestinal fluids: implications for the prediction of oral absorption, Eur. J. Pharm. Sci. Off. J. Eur. Fed.
585 Pharm. Sci. 57 (2014) 322–332. doi:10.1016/j.ejps.2013.08.027.
586 [14] J.H. Lin, I.W. Chen, K.J. Vastag, D. Ostovic, pH-dependent oral absorption of L-735,524, a potent HIV
587 protease inhibitor, in rats and dogs, Drug Metab. Dispos. Biol. Fate Chem. 23 (1995) 730–735.
588 [15] K.C. Yeh, P.J. Deutsch, H. Haddix, M. Hesney, V. Hoagland, W.D. Ju, S.J. Justice, B. Osborne, A.T. Sterrett,
589 J.A. Stone, E. Woolf, S. Waldman, Single-dose pharmacokinetics of indinavir and the effect of food,
590 Antimicrob. Agents Chemother. 42 (1998) 332–338.
591 [16] B. Kloefer, P. van Hoogevest, R. Moloney, M. Kuentz, M.L.S. Leigh, J. Dressman, Study of a Standardized
592 Taurocholate– Lecithin Powder for Preparing the Biorelevant Media FeSSIF and FaSSIF, Dissolution
593 Technol. 17 (2010) 6–13.
594 [17] N. Holmstock, T. De Bruyn, J. Bevernage, P. Annaert, R. Mols, J. Tack, P. Augustijns, Exploring food effects
595 on indinavir absorption with human intestinal fluids in the mouse intestine, Eur. J. Pharm. Sci. Off. J. Eur.
596 Fed. Pharm. Sci. 49 (2013) 27–32. doi:10.1016/j.ejps.2013.01.012.

26
597 [18] C. Jede, C. Wagner, H. Kubas, C. Weber, W. Weitschies, In-line derivative spectroscopy as a promising
598 application to a small-scale in vitro transfer model in biorelevant supersaturation and precipitation testing,
599 J. Pharm. Pharmacol. 70 (2018) 1315–1323. doi:10.1111/jphp.12991.
600 [19] L. Lindfors, S. Forssén, J. Westergren, U. Olsson, Nucleation and crystal growth in supersaturated solutions
601 of a model drug, J. Colloid Interface Sci. 325 (2008) 404–413. doi:10.1016/j.jcis.2008.05.034.
602 [20] Bernland, Karin Maria, Nucleating and clarifying polymers, 2010.
603 [21] S. Carlert, A. Pålsson, G. Hanisch, C. von Corswant, C. Nilsson, L. Lindfors, H. Lennernäs, B. Abrahamsson,
604 Predicting intestinal precipitation - a case example for a basic BCS class II drug, Pharm. Res. 27 (2010) 2119-
605 2130. doi:10.1007/s11095-010-0213-8.

606

27