Evaluation of the Rainbow Dynamic Dissolution Monitor™ Semi-automatic Fiber Optic Dissolution Tester

Caspar Schatz, Michel Ulmschneider, Rolf Altermatt, Stephan Marrer
Pharmaceutical Quality Assurance and Quality Control, F. Hoffmann-La Roche Ltd, Basel, Switzerland

email:caspar.schatz@roche.com

Summary
The Rainbow Dynamic Dissolution Monitor™ (Delphian Technology LP, Ardsley, USA) is a simple and convenient UV absorbance technique for acquiring precise, accurate, reproducible and robust dissolution profiles of drug formulations containing a single active ingredient. The instrument and its software are GMP compliant. Benefit analysis shows that it has significant advantages for dissolution over systems using filtering and flow -through cells. The Rainbow Dynamic Dissolution Monitor™ is thus suitable for routine dissolution analysis in pharmaceutical quality control.

Introduction
The Rainbow Dynamic Dissolution Monitor™ uses 12 fiber optic immersion probes residing in vessels of two dissolution baths throughout the dissolution test. Two deuterium lamps are used as a light source. After interacting with the sample, the light is guided to a series of 12 photo diode array ultraviolet monolithic miniature spectrometers (UV MMS, Carl Zeiss, Jena, Germany) [1] which measure from 200 to 400 nm with an absolute wavelength accuracy of 0.2 nm and a temperature drift less than 0.005 nm/K. The spectral pixel spacing is 0.8 nm, giving a Rayleigh resolution of about 3 nm. Stray light measured at 240 nm using a deuterium lamp and potassium iodide solution (10 g/l) is 0.3 % [2]. Each spectrometer unit and its probe are referred to as a channel.

Before performing a dissolution run the system collects 0% and 100 % transmission blanks and standard absorbance scans per channel. Hence the amount of dissolved active compound is determined with a single point calibration. To eliminate standard preparation errors, a second quality control standard is measured for control purposes only. During a run, the system can acquire absorbance scans every 10 seconds. The software incorporates two methods to correct for turbidity and scattering effects. The first method uses two wavelengths: one to determine the active compound, the other to act as a compensation wavelength. The second method is based on a second derivative algorithm using a wavelength range [1]. It uses a very simple algorithm to estimate the second derivative and a form of co-addition of several wavelengths to improve the signal to noise ratio.

Experimental
All the experiments were performed using the Rainbow Dynamic Dissolution Monitor™ with 10 mm pathlength Hellma ultra mini-immersion probes, (type 661.673-UV, Hellma GmbH & Co., Müllheim/Baden, Germany) to acquire UV measurements. Only six channels/probes were evaluated, always using one scan per measurement in each case.

System suitability
Suitability was assessed in terms of fiber optic unit transmission and the linear range of the spectroscopic assembly.

Transmission
The relative energy of 100% transmission spectra of artificial intestinal fluid pH 7.5 (reference Anticoagulant Tablets Section, page 10, for fluid composition) was plotted against wavelength as a measure of channel and probe transmission

Linear range
The linear range of all six evaluated channels was tested using a dilution series of potassium dichromate (spectroscopy grade, Fluka Chemie AG, Buchs, Switzerland) in 0.01 N sulphuric acid [3]. A stock solution was diluted to concentrations giving absorbance values ranging from 0.2 to 2.0. Based on the absorbance spectrum of potassium dichromate in 0.01 N sulphuric acid, absorbance was measured at 258 nm.
A correlation coefficient was calculated using the absorbance values from the two weakest standard solutions; the same process was repeated for increasing strengths of standard solutions to determine the upper end of absorbance linearity (taken as 99.9% correlation with prediction).

Anticoagulant tablets
The Rainbow Dynamic Dissolution Monitor™ was evaluated using anticoagulant tablets containing 3 mg of active compound, corn starch white, lactose powder, magnesium stearate, and talc.

In routine dissolution analysis, the anticoagulant tablets are dissolved in 900 ml of artificial intestinal fluid pH 7.5 (comprised of 80.5g anhydrous dipotassium hydrogen phosphate and 15.6 g of potassium dihydrogen phosphate dihydrate in 10 liters ofdistilled water), stirred at 50 rpm in apparatus 2 [4]. The medium is degassed and heated to 37.0 ± 0.5 °C. The 20-minute Q value used for release analysis is 75% [5].

Linearity of absorbance readings
To identify a suitable detection wavelength, triplicate absorbance spectra were acquired of solutions equivalent to 25, 50, 75, 100, and 125% of active compound dissolved in artificial intestinal fluid pH 7.5, with approximate concentrations of 0.00083, 0.00167, 0.00250, 0.00333, and 0.00417 mg/ml, respectively. The resulting correlation coefficient was plotted against wavelength.

Linearity of compensation methods
Based on the absorbance and second derivative spectrum of active compound in artificial intestinal fluid pH 7.5 determined in an earlier experiment, the linearity of three different turbidity compensation methods was investigated (Table 1).

Table 1: Three different turbidity compensation methods.

Method	Type of spectrum used	Wavelengths [nm]
1	Absorbance	310 (detection) 350 (compensation)
2	Absorbance	310 (detection) 376 (compensation)
3	Second derivative	300 to 320

All methods were evaluated in triplicate using 25, 50, 75, 100, and 125% solutions of active compound dissolved in artificial intestinal fluid pH 7.5 with approximate concentrations of 0.00083, 0.00167, 0.00250, 0.00333, and 0.00417 mg/ml, respectively, and using clear medium as well as medium containing a concentration of placebo powder equivalent to one 130.0 mg tablet dissolved in 900 ml artificial intestinal fluid pH 7.5.

The validation of analytical methods (VoAM) program, version 3.0 [6], was used, with the following acceptance criteria [7, 8]: correlation coefficient > 0.99; y intercept within the 95% confidence interval of 2% of the reference x value (100% solution of active compound); precision, expressed as the standard deviation of relative repeatability (treating each set of triplicate data as one group), < 2.00% assuming data and mean recovery between 98.00 and 102.00%.

Comparison of turbidity compensation methods
Six absorbance readings of active compound solution in artificial intestinal fluid pH 7.5 before and after addition of placebo powder were acquired in triplicate to compare the accuracy and efficacy of the three turbidity compensation methods, using 100% solutions of active compound (approximately 0.00333 mg/ml).

The VoAM 3.0 program [6] was used to determine statistical equivalence, with the following acceptance criterion: the 95% confidence interval of the mean of the test method had to lie entirely within 2.00% either side of the mean of the reference method.

Robustness of the turbidity compensation methods
Two absorbance measurements were acquired at 12 different positions (hence different bending radii) of the fiber optic immersion probes and cables to test the robustness of each compensation method with respect to obligatory movement by the fiber optic immersion probes during the performance of a dissolution run. This experiment was performed using artificial intestinal fluid pH 7.5 containing active compound at approximately 0.00417 mg/ml, equivalent to the extent of dissolution of 125%.

Method comparison
The two turbidity compensation wavelength methods were compared in dissolution runs using a dissolution bath (Distek Premiere 5100, Distek Inc., North Brunswick, USA) and three lots of anticoagulant tablets (six tablets per lot). Active compound release was quantified at 20 minutes using the two turbidity compensation methods and the corresponding reference methods. With the reference methods, a 20 ml aliquot was manually removed from each vessel and membrane-filtered (0.45 mm pore size, Gelman Acrodisc, product no. 4496, Pall Gelman Sciences, Ann Arbor, USA) [5]. Single point calibration was used for quantification on a diode array spectrometer (HP 8452 A, Agilent Technologies, Rockaway, USA) in combination with the same compensation wavelength used with the Rainbow Dynamic Dissolution Monitor™.
Equivalence was defined as a maximum deviation of ± 2.0% per tablet, with post-calculation rounding.

Results
System suitability
Evidence for the suitability of fiber optic transmission and system linear range is given below.

Transmission
The plot of relative probe energy in artificial intestinal fluid pH 7.5 (Figure 1) shows values exceeding 30% from about 230 to 390 nm. Hence this wavelength range is suitable for measuring UV absorbance.

Figure 1 Plot of relative immersion probe energy in artificial intestinal fluid pH 7.5 against wavelength (nm)

Linear range
The spectrum of potassium dichromate in 0.01 N sulphuric acid shows an absorbance maximum at 258 nm being used to evaluate linear range. Table 2 gives the upper limits of the linear ranges examined with potassium dichromate in 0.01 N sulphuric acid at 258 nm with all six probes (channels).

Table 2: Upper end of linear range for all probes (channels).

Probe	Upper limit of linear range [absorbance units (AU)]
1	1.6
2	1.8
3	1.6
4	1.5
5	1.5
6	1.5

Since probes 4 to 6 showed values of 1.5 AU, the linear range of the whole system also had to be set to 1.5 AU.

Anticoagulant tablets
Linearity of the absorbance readings
Figure 2 plots the correlation coefficient against wavelength with active compound dissolved in artificial intestinal fluid pH 7.5 (25­125% solutions, with approximate concentrations of 0.00083­ 0.00417 mg/ml). The active compound absorbance spectrum in this figure indicates decreases in absorbance at 260 and 340 nm, arising from a decrease in system linearity owing to a decrease in system signal to noise ratio. The further slight decrease near 220 nm arises from the reduced energy available in the shortwave UV region.

Figure 2 Linearity experiment in triplicate (a­c) with active compound at a concentration equivalent to 100% dissolution in artificial intestinal fluid pH 7.5 (approximately 0.00333 mg/ml): plot of correlation coefficient against wavelength, incorporating the absorbance spectrum of active compound

It is clear that the whole wavelength range from 220 to 340 nm is suitable for method development

Linearity of turbidity compensation methods
Figure 3 shows the UV absorbance spectrum of active compound and its estimate of the second derivative used in the Rainbow Dynamic Dissolution Monitor™ software.
Based on the absorbance spectrum with a maximum at 310 nm, two turbidity compensation methods using the peak wavelength and compensation wavelengths of 350 and 376 nm respectively were chosen. The 300­320 nm range was used in the case of the second derivative.

Figure 3 Absorbance spectrum of active compound at a concentration equivalent to 100% dissolution in artificial intestinal fluid pH 7.5, co-plotted with its estimate for the second derivative used in the Rainbow Dynamic Dissolution Monitor'

Table 3 gives the relevant statistical parameters for validating the three methods using clear as well as placebo-spiked solutions.

Method 1 (310 and 350 nm): clear solutions
Parameter	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
r	0.99999	0.99999	0.99999	0.99994	0.99977	0.99997
SDrel [%]	0.15	0.13	0.15	0.14	0.16	0.12
intercept	+	+	+	+	+	+
recovery [%]	100.16	100.01	100.13	100.26	101.03	100.04
Method 2 (310 and 376 nm): clear solutions
Parameter	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
r	0.99999	0.99999	0.99999	0.99990	0.99968	0.99995
SDrel [%]	0.14	0.13	0.14	0.15	0.16	0.13
intercept	+	+	+	+	+	+
recovery [%]	100.12	100.02	99.88	100.19	101.29	100.04
Method 3 (derivative from 300 to 310 nm): clear solutions
Parameter	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
r	0.99998	0.99992	0.99997	0.99996	0.99930	0.99997
SDrel [%]	0.22	0.23	0.25	0.26	0.26	0.22
intercept	+	+	+	+	+	+
recovery [%]	101.45	101.26	101.39	101.62	101.18	101.16
Method 1 (310 and 350 nm): placebo-spiked, turbid solutions
Parameter	Probe 1*	Probe 2	Probe 3	Probe 4*	Probe 5	Probe 6
r	0.99846	0.99879	0.99989	0.8973	0.99965	0.99983
SDrel [%]	0.58	0.35	0.21	0.28	0.16	0.23
intercept	+	+	+	-	+	+
recovery [%]	98.34	100.82	99.88	99.63	100.34	99.90
Method 2 (310 and 376 nm): placebo-spiked, turbid solutions
Parameter	Probe 1*	Probe 2	Probe 3	Probe 4*	Probe 5	Probe 6
r	0.99673	0.99858	0.99985	0.98045	0.99938	0.99978
SDrel [%]	0.71	0.41	0.28	0.34	0.16	0.24
intercept	-	+	+	-	+	+
recovery [%]	96.93	100.67	99.51	99.17	99.99	99.47
Method 3 (derivative from 300 to 310 nm): placebo-spiked, turbid solutions
Parameter	Probe 1*	Probe 2	Probe 3	Probe 4*	Probe 5	Probe 6
r	0.99972	0.99966	0.99970	0.99847	0.99927	0.99988
SDrel [%]	0.49	0.49	0.42	0.37	0.38	0.26
intercept	+	+	+	+	+	+
recovery [%]	100.79	100.04	100.84	101.03	100.89	100.47

In the case of the clear solutions all parameters were inside the acceptance limits. There were no significant differences (p = 95%) in method validation. However, correlation coefficients were clearly lower, and standard deviations of relative repeatability and recovery rates clearly higher, with the second derivative method (Method 3) than with either compensation wavelength method (Methods 1 and 2). Therefore it can be concluded that the second derivative algorithm is less accurate and less precise than either wavelength method when examining clear solutions.

When performing the same method validation experiments with placebo-spiked solutions, there were air bubbles in the measurement compartments of probes 1 and 4 in at least one of the triplicate measurements. As can be seen in Figure 4, in contrast to the wavelength-independent baseline offset caused by tablet excipients, air bubbles have a wavelength-dependent impact on the baseline owing to the wavelength dependency of refraction and diffraction. This explains why in Method 1, where the compensation wavelength approximates to the analytical wavelength, only probe 4 failed the validation acceptance limits; in Method 2, on the other hand, where the compensation wavelength is further from the analytical wavelength, probes 1 and 4 failed the acceptance criteria. Since the second derivative algorithm corrected for sloping baseline offsets, all probes met the acceptance criteria, making this the most robust method. Although there were no significant differences (p = 95%) in method validation, the second derivative algorithm tended to have a slightly higher standard deviation of relative repeatability.

Figure 4 Probe 4: Absorbance spectra of measure 2 per triplicate (Methods 1­3) at a concentration equivalent to 75% dissolution (approximately 0.0025 mg/ml), comparing the wavelength-dependent baseline offset caused by air bubbles in the measuring compartment (Expts. 2 & 3) vs the air bubble-free spectrum (Expt. 1) which shows only the wavelength-independent offset caused by excipient turbidity.

Based on the validation acceptance criteria for method equivalence there were no significant differences (p = 95%) between the methods used for turbidity compensation. But as can be seen from Figure 5, which shows the mean concentrations of six measurements with their standard deviations as error bars for the three methods before and after addition of placebo powder, the two-wavelength compensation methods (methods 1 and 2) have a smaller standard deviation than the second derivative method (method 3) and show less probe to probe variation. Hence methods 1 and 2 are more rugged from this standpoint.

Figure 5 Concentrations measured using the three methods before and after addition of placebo powder (means of six measures). Error bars: 2 SD.

Since concentrations are higher before than after the addition of placebo powder, all three turbidity compensation methods overcompensate. Although the differences are not significant (p = 95%) for method validation, the differences between clear and turbid solutions are smallest when using the two-wavelength method with a compensation wavelength approximating to the analytical wavelength. Hence method 1 is most suitable in terms of the accuracy of turbidity correction.
Robustness of turbidity compensation methods

Table 4 shows that the method to method difference, expressed as the relative SD, was not significant (p = 95%) in method validation. All three methods are therefore equivalent in terms of the robustness of moving fiber optic probes.

Table 4 Relative standard deviations (SDrel) of two measurements in 12 positions per method

Method 1 (310 and 350 nm)
	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
SDrel [%]	0.16	0.14	0.11	0.08	0.09	0.08
Method 2 (310 and 376 nm)
	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
SDrel [%]	0.20	0.19	0.14	0.10	0.11	0.10
Method 3 (2nd derivative from 300 to 310 nm)
	Probe 1	Probe 2	Probe 3	Probe 4	Probe 5	Probe 6
SDrel [%]	0.14	0.16	0.16	0.12	0.11	0.16

Method comparison
Every dosage form met the acceptance criteria using method 1 and 2 (Table 5). Both methods give accurate results on the Rainbow Dynamic Dissolution Monitor™.

Table 5 Amounts (%) of dissolved active compound using methods 1 and 2. With the Hewlett-Packard spectrometer, sample absorbances were measured after filtration. Some calculated differences do not quite match the percentages in the results column due to the calculation being performed before rounding.

Lot 1	Method 1 (310 and 350 nm)			Method 2 (310 and 376 nm)
Tablet	HP	Rainbow	Difference	HP	Rainbow	Difference
1	98.6	97.7	-0.9	98.4	98.4	-0.1
2	100.3	98.5	-1.8	100.4	99.0	-1.4
3	98.9	97.3	-1.6	98.9	97.9	-1.0
4	101.0	99.0	-1.9	100.6	99.0	-1.6
5	99.3	97.6	-1.8	99.0	97.6	-1.4
6	94.9	95.8	-0.9	95.5	94.9	-0.7
Lot 2	Method 1 (310 and 350 nm)			Method 2 (310 and 376 nm)
Tablet	HP	Rainbow	Difference	HP	Rainbow	Difference
1	95.6	96.0	0.4	95.7	96.0	0.2
2	95.6	94.3	-1.3	95.6	94.3	-1.3
3	95.8	94.4	-1.4	95.9	94.4	-1.5
4	99.0	97.7	-1.3	99.0	97.7	-1.4
5	100.2	98.7	-1.5	99.8	98.7	-1.1
6	101.7	99.8	-1.9	101.4	99.8	-1.5
Lot 3	Method 1 (310 and 350 nm)			Method 2 (310 and 376 nm)
Tablet	HP	Rainbow	Difference	HP	Rainbow	Difference
1	96.2	95.3	-0.8	96.4	96.1	-0.3
2	101.5	99.6	-1.9	101.9	100.0	-1.9
3	99.7	98.0	-1.7	99.7	98.3	-1.4
4	98.4	97.1	-1.3	98.6	97.5	-1.1
5	100.6	98.7	-1.9	100.9	98.9	-2.0
6	98.9	98.4	-0.5	100.1	98.9	-1.2

Benefit analysis
Table 6 presents the results of a benefit analysis comparing the Rainbow Dynamic Dissolution Monitor™ with a conventional system using filtration and flow-through cuvettes to determine the amount of dissolved active compound, in terms of the following parameters: laboratory work, validation burden, maintenance, analytical information and GMP compliance.The total scores show that the Rainbow Dynamic Dissolution Monitor™ outperforms a semi-automatic filtering and flow-through cuvette measurement on-line system without loss of GMP compliance.

Table 6 Benefit analysis: Rainbow Dynamic Dissolution MonitorÔ vs a conventional online system, in terms of criteria ranked using a weighting factor (0-100%). Mark (1-5): system approximation to criteria. Score: weighting factor x  mark.

Criterion	Weighting factor [%]	Rainbow		Conventional
		Mark	Score	Mark	Score
Laboratory work	15	3	45	3	45
Qualification burden	15	4	60	2	30
Maintenance	15	3	45	2	30
Analytical information	15	5	75	2	30
GMP compliance	40	5	200	5	200
Total	100	20	425	14	335

Laboratory work
Laboratory workload, in terms of preparing the bath and standards, is similar with both systems. During operation, no more hardware problems are to be expected with the Rainbow Dynamic Dissolution Monitor™ than with the conventional system since the fiber optic immersion probes and related mechanics are quite robust [9].

Qualification burden
Since the Rainbow Dynamic Dissolution Monitor™ incorporates no filtration facility or liquid pump, there is less equipment to qualify. The UV detectors are also simpler to qualify than spectrometers used for UV/VIS precision measurements [10].

Maintenance
The Rainbow Dynamic Dissolution Monitor™ is easier to maintain and less labor- intensive due to the elimination of sample removal and filtration.

Analytical information
The Rainbow Dynamic Dissolution Monitor™ can supply a data point every 10 seconds, giving dissolution profiles containing a lot of information. It also eliminates problems due to dead volume, time differences between sampling and measuring, and filter clogging, leading to greater accuracy.

GMP compliance
Both systems are GMP compliant.

Conclusions
The analytical results confirm that the Rainbow Dynamic Dissolution Monitor™ can be used to measure dissolution by methods which meet the acceptance criteria for linearity, accuracy, precision, and reproducibility stipulated in current validation of analytical methods guidelines [7]. Both instrument and software are GMP compliant. Benefit analysis shows that it outperforms dissolution measurement systems employing filtering and flow-through cells. The advantages have an impact on the high acquisition costs, though. The Rainbow Dynamic Dissolution Monitor™ is thus suitable for routine dissolution analysis in pharmaceutical quality control.

Both the two-wavelength compensation method and the second derivative algorithm are suitable for monitoring dissolution. The former is generally more precise and accurate, especially for non-disintegrating formulations where the medium stays clear. For formulations giving a background resulting in a sloping offset, the second derivative algorithm is preferable, as also when there are air bubble problems.

References
[1] Bynum K, Kraft E, Pocreva J, Ciurczak E, Palermo P, In Situ Dissolution Testing Using a Fiber Optic Probe Dissolution System, Dissolution Technologies, 6 (4) 8­10 (1999)
[2] Product information: MMS UV Monolithic Miniature Spectrometer, Carl Zeiss, OEM Spekralsensorik, Jena
[3] European Pharmacopoeia, 3rd ed., Council of Europe, Strasbourg, Absorbance Spectrometry, Ultraviolet and Visible, 28-29 (1997)
[4] USP 24 NF 19, United States Pharmacopeial Convention, Inc., Rockville MD, Dissolution, 1941-1943 (1999)
[5] Pharma Switzerland, Quality Assurance and Quality Control, Analysis Instruction for Anticoagulant Tablets, F. Hoffmann-La Roche Ltd, Basel (2000)
[6] VoAM 3.0, Program for Validation of Analytical Methods, F. Hoffmann-La Roche Ltd, Basel (1999)
[7] Pharma Switzerland, Quality Assurance and Quality Control, Guideline for Validation of Analytical Methods, F. Hoffmann-La Roche Ltd, Basel (1998)
[8] International Conference on Harmonization, Validation of Analytical Procedures: Methodology, ICH Harmonised Tripartite Guideline Q2B (1995)
[9] Schatz C, Ulmschneider M, Altermatt R, Marrer S, Altorfer H, Manual In Situ Fiber Optic Dissolution Analysis in Quality Control, Dissolution Technologies, 7 (2) 6-13 (2000)
[10] Pharma Switzerland, Quality Assurance and Quality Control, SOP for Testing and Maintenance of UV/VIS Precision Spectrophotometers, F. Hoffmann-La Roche Ltd, Basel (1999)