13

SETTING DISSOLUTION SPECIFICATIONS FOR MODIFIED-RELEASE DOSAGE FORMS

Deborah A. Piscitelli and David Young

IVIVC Cooperative Working Group Pharmacokinetics-Biopharmaceutics Laboratory School of Pharmacy University of Maryland at Baltimore Baltimore, Maryland 21201

1. BACKGROUND

Dissolution specifications are limits for the percent of drug released at specific times. Setting these boundaries assures that all formulations which meet these limits per­form similarly. For years dissolution specifications have served as an in vitro quality as­surance (e.g., in stability testing). As a quality control measure, dissolution specifications are defined by the Sponsor and the Food and Drug Administration (FDA). Acceptance cri­teria is often based on three time points: 1) an early time point to identify if dose dumping occurs, 2) a time point to characterize the release profile and demonstrate the extension of release, and 3) a time point to prove that most of the intended entire dose is delivered. The USP dissolution acceptance criteria for diltiazem hydrochloride extended release capsules is based on time points at 3, 9, and 12 hours. The following chart describes the USP disso­lution specifications for diltiazem hydrochloride l ;

Time (hours)

3 9 12

Amount dissolved

between 10% and 25% between 45% and 85% not less than 70%

In the past, dissolution specifications were based primarily on the in vitro perform­ance of the product without any consideration to in vivo significance. More recently, the FDA has expanded the role of dissolution specifications in the guidelines for scale up and post approval changes.2,3 In these guidances, a scaled-up formulation or a minor post ap­proval change to the formulation may not require a bioavailability study, if it can be shown to fall within the dissolution specifications. Thus, the ideal situation would be to set the dissolution specifications such that scaled up and post approval changed formula-

In Vitro-in Vivo Co"elations, edited by Young et al. Plenum Press, New York, 1997 159

160 D. A. Piscitelli and D. Young

tions as well as the stability lots fall within the dissolution boundaries.4 Dissolution speci­fications can also serve a similar role in pre-approval formulation development where for­mulations can be optimized using the dissolution specifications as a guide rather than depending on multiple bioavailability studies.

At the present time, this broaden goal requires that the dissolution system no longer serves as a quality control dissolution test but instead is a surrogate for in vivo bioavail­ability. By doing this, the in vitro dissolution test becomes a predictor of in vivo perform­ance of the formulation, hence, the specifications can be set to minimize the possibility of releasing formulations that would be different in their in vivo performance. In order to ac­complish this, it is crucial to incorporate in vivo data when setting dissolution limits. Dis­solution specifications can be set so that all formulations that have dissolution profiles within the specifications are bioequivalent, or minimally, all formulations should be bio­equivalent to the pivotal (bio) batch. In this chapter, various methods for setting dissolu­tion specifications which serve as predictors of in vivo performance are reviewed. First, however, it is important to review the significant characteristics when choosing a method for setting dissolution specifications.

2. CHARACTERISTICS FOR A DISSOLUTION SPECIFICATION METHOD

2.1. Discriminating Dissolution System

The dissolution system is not as important as the resulting dissolution profiles. If the system is nondiscriminating, this may create a problem when developing the specifications. Formulations may appear similar in vitro, yet result in bioinequivalence in vivo. Shah and Williams demonstrate this point with quinidine gluconate. 5 Two quinidine gluconate formula­tions produced similar in vitro profiles when the dissolution test was performed in an acidic medium. However, the formulations produced different plasma concentration-time curves and were in fact bioinequivalent. When the dissolution medium was changed to pH 5.4 acetate buffer, the dissolution profiles became different; thus more accurately representing the in vivo process. It is best to have a discriminating system so that if a formulation is bioinequivalent in vivo, this may be identified in vitro. Dissolution testing may need to be performed under a va­riety of conditions: changing medium (e.g., pH, osmolarity, surfactant), mixing speed, or ap­paratus (e.g., USP Apparatus I, II, and III). If the system is over-discriminating, the resultant dissolution specifications will be wider when encompassing all bioequivalent formulations.

2.2. Incorporation of in Vivo Data

In order for the dissolution system to serve as a surrogate, the dissolution data must be correlated to the in vivo data. There are numerous chapters in this book that describe in vitro-in vivo correlations (IVIVC) methods and this chapter will not expand on this topic further. However, it is important to emphasize that only with a defined IVIVC can one be confident that formulations within the dissolution specifications will most likely result in bioequivalent products.

Setting Dissolution Specifications for Modified-Release Dosage Forms 161

2.3. Including Variability

Ideally, variability should be included in setting dissolution specifications. This in­cludes the variability in the dissolution data; however, this is usually small. More impor­tantly, this includes variability in the in vivo data, since this is usually larger than the in vitro variability. Since each drug has a certain amount of variability when administered to patients, it would be beneficial to quantify this variability and include it when setting dis­solution specifications. For example, a Level A correlation may be determined for 10 sub­jects. The mean and variance can then be calculated for the parameters of the correlation (i.e., slope and intercept). This intersubject variability can then be incorporated into the method used for setting the dissolution specifications. In this way the specifications would be wider for those drugs with a larger amount of variability and narrower for those drugs with a smaller amount of variability.

2.4. Plasma Concentration-Time Curves

A method for setting dissolution specifications should demonstrate in some fashion what these dissolution profiles would look like as plasma concentration-time curves. If the specifications are being used to predict in vivo performance, it is necessary to see the in vivo profiles. This may be done by administering the upper and lower formulations in a bioavailability study or perhaps more efficiently by convoluting the upper and lower lim­its to achieve the associated plasma concentration-time curves. Additionally, if some form of intersubject/intertablet variability can be incorporated into the plasma concentration­time profiles, bioequivalence using the typical bioequivalence matrices can be evaluated.

50,

50

& &

_40 &

oJ c;. &

&

U

.5. & C 0 30 i. :oJ g C II U C 0 ()

20

10

10 15 20 25

Time (Hours)

Figure 1. Plasma concentration-time profiles convoluted from dissolution specifications. The solid and broken lines represent the upper and lower dissolution specifications, respectively. The triangles represent the pivotal batch.

162 D. A. Piscitelli and D. Young

1. Manufacture a range of formulations which produce a range of dissolution profiles.

2. Administer these formulations in a bioavailability study and test for bioequivalence.

3. The dissolution range that results in bioequivalent formulations would be the specifications.

Figure 2. Steps for setting dissolution specifications using the set range method.

An example of typical plasma concentration-time profiles convoluted from dissolution specifications can be seen in Figure 1. The upper and lower limits are represented by the solid and broken line, respectively. The triangles represent the pivotal batch.

3. METHODS

It is suggested in the FDA guidance that a minimum of three time points be required to set dissolution specifications.4 Ideally, these time points should cover the early, middle and late segments of the dissolution profile. The final time point should occur when at least 80% of the label claim has been released. If the maximum amount released is less than 80% of the label claim, then the last time point should be the time where the plateau in the profile has been achieved. This section of this chapter outlines several methods which may be used to develop dissolution specifications.

3.1. Set Range Method

The FDA guidance outlines this method when an IVIVC does not exist.4 This method does not incorporate in vivo data or intersubject variability. The steps are outlined in Figure 2. In this method, the upper and lower limits are not shown as plasma concentra­tion profiles, thus there is no assurance that these limits result in bioequivalent plasma profiles unless a separate bioavailability study is performed.

3.2. Empirical Method

The empirical method is described in Figure 3. The empirical method includes in vivo data, intersubject variability, and results in plasma concentration-time profiles. This method uses several formulations; the more formulations included, the more confident one can be that the specifications are appropriate. This method is costly both in time and money. Quite often this method turns out to be a retrospective analysis of several inde-

1. Perform dissolution testing on the clinical lots and mean the profile. 2. The maximum suggested range at any dissolution time point is

.± 10% of the label claim deviation from the mean dissolution profile. (Deviations from the + 10% criteria may be accepted by the FDA provided that the range at any time point does not exceed 25%.)

Figure 3. Steps for setting dissolution specifications using the empirical method.

Setting Dissolution Specifications for Modified-Release Dosage Forms

1. Establish an IVIVC. 2. Calculate the 90% confidence intervals around the IVIVC parameters. 3. Using the IVIVC and the midpoint from the dissolution specifications,

calculate the mean amount absorbed/released. 4. Calculate the upper and lower dissolution boundary using the mean in vivo

data and the upper and lower 90% confidence intervals for the parameters.

Figure 4. Steps for setting dissolution specifications using the 90% confidence interval method.

163

pendent bioequivalence studies. Thus, this analysis may lead to an additional bioequiva­lence study to ensure that the two outside dissolution profiles (upper and lower) are bio­equivalent.

3.3. 90% Confidence Interval

The 90% confidence method includes in vivo data and intersubject variability. It does not predict plasma concentration-time profiles, thus it is unable to test if the specifi­cations result in bioequivalent formulations. The steps are outlined in Figure 4.

3.4. 95% Confidence Method

Dissolution specifications are justified using 95% confidence intervals around the the mean plasma concentration-time profile.? This profile was obtained following the ad­ministration of the pivotal batch in a bioavailability study. The 95% confidence method includes in vivo data, intersubject variability, and results in plasma concentration-time profiles. The steps are outlined in Figure 5.

Although this method predicts plasma concentration-time profiles for the upper and lower dissolution limits, it does not require the limits to be tested for bioequivalence.

3.5. Artificial Neural Network Method

Artificial neural network (ANN) had been described in two other chapters within this book. These ANN methods include in vitro and in vivo data. They can provide plasma concentration-time profiles for given in vitro profiles. The steps are outlined in Figure 6.

1. Establish an IVIVC. 2. Calculate the in vivo release using the working dissolution limits and the

IVIVC. 3. Convolute the in vivo release to generate the upper and lower plasma

concentration profiles. 4. Administer the pivotal batch in vivo. 5. Construct 95% confidence intervals around the observed data. 6. If the working dissolution specifications are within the 95% confidence

intervals, then any formulation with a dissolution profile that falls within the dissolution specifications would be bioequivalent to the pivotal batch.

Figure 5. Steps for setting dissolution specifications using the 95% confidence interval method.

164 D. A. Piscitelli and D. Young

1. Determine appropriate data set to train the ANN. 2. Determine an optimal ANN configuration (i.e., type of ANN, training

paradigm, number of nodes, etc.) 3. Train the ANN (Take steps to ensure memorization of the ANN is not

occurring). 4. Validate the ANN with a data set tat was not used for training. 5. Predict in vivo plasma profile with the ANN using the proposed dissolution

specifications.

Figure 6. Steps for setting dissolution specifications using the ANN method

3.6. Convolution-Based Method

In the convolution-based method the in vivo release/absorption is modeled as a func­tion of the in vitro release.s Thus, this method requires an oral solution, intravenous, or immediate release reference dose. The convolution-based method uses the in vitro dissolu­tion specifications, an established IVIVC, intersubject variability, and results in plasma concentration-time profiles. The steps for this method can be found in Figure 7.

The limits may be tested for bioequivalence. Intersubject variability may be added to the reference dose parameters and/or the IVIVC parameters.

4. ADDITIONAL CONSIDERATIONS

As seen, it is valuable if the method for setting the dissolution specifications can predict the plasma curves for the upper and lower dissolution limits. These plasma curves, which represent the mean plasma curves for formulations with the upper and lower disso­lution limits, can then be compared to each other or the pivotal batch. Expanding the pre­diction to include intersubject variability would provide a more realistic representation of the in vivo performance: Bioequivalence could then be tested for the formulations repre­senting the upper, lower, and/or pivotal batch dissolution profiles. In Figure 8, a popula­tion simulation for 12 subjects has been performed for the upper dissolution limit. The solid and broken line represent the mean plasma profiles for the upper and lower dissolu­tion specifications, respectively. As shown in this figure, intersubject variability can be an important factor when evaluating the dissolution specifications.

1. Establish an IVIVC. 2. Convolute the dissolution limits and the pivotal batch. 3. Test for bioequivalence between the upper and lower limits. 4. If the limits are bioinequivalent, test for bioequivalence between each limit

and the pivotal batch. 5. If the limits are bioequivalent, accept the dissolution specifications and

consider expanding the limits.

Figure 7. Steps for setting dissolution specifications using the convolution based method.

Setting Dissolution Specifications for Modified-Release Dosage Forms 165

o

- '" ~ U

.§. &: 0 30 ~

~ &:

B &: 0

0 20

••

• K

x x + •

+ x ~

,. ... j o ·

" •

••

.'.

Time {Ho ..... '

Figure 8. Population simulation for the formulation with the upper specification dissolution profile. The solid and broken lines represent the mean plasma concentration-time profiles for the mean upper and lower dissolution pro­files, respectively. The individual symbols represent the simulated plasma concentrations for 12 subjects.

Table 1. Advantages and disadvantages for the dissolution specifications setting methods

Methods Advantages Disadvantages Empirical 1) Includes in vivo data 1) Costly in time and money

2) Inciudes variabiiity 2) Potential to be a retrospective analysis 3) Includes plasma profiles 3) May require an add~ional bioequivalence study

Set Range 1) Inexpensive 1) Does not include in vivo data 2) Does not include variability 3) Does not predict the plasma profiles for the specifocalions

90% Confidence Interval 1) Includes in vivo data 1) Does not predict the plasma profiles for the specifocalions 2) Includes variabilny 2) Does not prove bioequivalence between the upper and iowar

lim~s

95% Confidence Interval 1) Inciudes in vivo data 1) Does not predict the plasma profiles for the specifocalions 2) Inciudes variability 2) Does not prove bioequivalence between the upper and lower

lim~s 3) Requires both deconvolution and convolution programs

ArtifICial Neural Network 1) Inciudes in vivo data 1) Unable to have a priori structure for inter- and intra-subject 2) Predicts plasma profiles variability 3) May test for bioequivalence 2) Specifocalion of the types of inputs. ANN configuration between the upper and lower lim~s variables. and format of the input-output relationship can lead to

an exorMant number of possible ANN 3) The range of input values used to train a network becomes an important consideration in the successful prediction from an ANN. Desirable to predict from inputs weil wnhin range used to train; dissoiulion specifocalions are often at e~her end of the ranges. 4) Requires training in ANN

Convolution-Based 1) Includes in vivo data 1) Requires both deconvolution and convoiution programs 2) Can include variabil~y 3) Predicts plasma profiles 4) May test for bioequivalence between the upper and iowar lim~s

166 D. A. Piscitelli and D. Young

5. SUMMARY

Dissolution specifications are used for quality assurance and may also serve as a sur­rogate for in vivo bioavailability. These limits can guide formulation development and eliminate the need for bioavailability studies for scale up and post approval changes. Sev­eral methods for setting dissolution specifications have been reviewed in this chapter. A summary of the advantages and disadvantages for each method can be found in Table 1. When choosing a method for setting dissolution specifications, it is important to 1) have a discriminating dissolution system, 2) incorporate in vivo data, 3) include intersubject vari­ability, and 4) predict plasma concentration-time profiles. Predicting plasma concentration curves allows one to see how the change in formulation or dissolution limits perform in vivo. Dissolution specifications should be set so that all formulations that have dissolution profiles within the limits of the specifications are bioequivalent. This can be assured if the boundaries are tested for bioequivalence. Minimally, the formulations that have dissolu­tion profiles within the limits of the specifications should be bioequivalent to the pivotal batch. A population prediction of the plasma concentration-time profiles for the upper and lower limit would incorporate the true intersubject variability for the formulation.

REFERENCES

I. The United States Pharmacopeia 23 and The National Formulary 18, The United States Pharmacopeia I Convention, Inc. Supp. 1, 1995:2449-2450.

2. Guidance for Industry. Immediate Release Solid Oral Dosage Forms. Scale-Up and Post-Approval Changes: Chemistry, Manufacturing and Controls, In Vitro Dissolution Testing, and In Vivo Bioequiva­lence Documentation. Center for Drug Evaluation and Research, November 1995.

3. Guidance for Industry. Modified Release Solid Oral Dosage Forms. Scale-Up and Post-Approval Changes: Chemistry, Manufacturing and Controls, In Vitro Dissolution Testing, and In Vivo Bioequivalence Docu­mentation. Center for Drug Evaluation and Research, May 1996.

4. Guidance for Industry. Extended Release Solid Oral Dosage Forms Development, Evaluation and Applica­tion of In Vitro/ln Vivo Correlations. Center for Drug Evaluation and Research, July 1,1996.

5. Shah VP and Williams RL. In vivo and In vitro correlations: scientific and regulatory perspectives, in Ge­nerics and Bioequivalence. CRC Press, Inc., Boca Raton, Florida, 1994. 108-109.

6. In Vitro/In Vivo Correlations for Extended Release Oral Dosage Forms. Pharmacopeial Forum. July-Au­gust 1988:4160.

7. Leeson, L.J., In vitro-In vivo correlations. Drug Information Journal. 1995;29:903-915. 8. Gillespie WR. Convolution-based approaches for In vivo-In vitro correlation modeling. In Vitro-In Vivo

Relationship Workshop. Baltimore, MD. September 1996.