narrow therapeutic index drug, or only two release rates were used to develop the IVIVC, or if the internal predictability criteria are not met (for criteria, see p. 438). However, since the IVIVC will potentially be used to predict the in vivo performance for future changes, it is of value to evaluate external predictability when additional data are available. An important concept is that the less data available for initial IVIVC development, the more additional data may be needed to define completely the IVIVC's predictability. Some combination of three or more formulations with different release rates is considered optimal.
Internal and External Predictability. Estimation of prediction error internally: Internal predictability should be evaluated for all IVIVCs (irrespective of the therapeutic index of the drug).
Estimation of prediction error externally. This is appropriate in some situations, particularly when only two formulations with different release rates are used to develop the IVIVC model, when calculation of prediction error internally is inconclusive, or when a narrow therapeutic index drug is studied.
The additional test data sets used for external prediction error calculation may have several differing characteristics compared to the data sets used in IVIVC development. Although formulations with different release rates provide the optimal test of an IVIVC's predictability, data from other types of formulations may be considered. In each case, bioavailability data should be available for the data set under consideration.
The following represent, in decreasing order of preference, formulations that may be used to estimate prediction error externally:
• A formulation with a different release rate than those used in IVIVC development.
• A formulation with the same or similar release rate, but involving some change in the manufacture of this batch (e.g., composition, process, equipment, manufacturing site).
• A formulation with the same or similar release rate obtained from another batch/lot with no changes in manufacturing.
Methods and Criteria for Evaluation of Predictability. The objective of IVIVC evaluation is to estimate the magnitude of the error in predicting the in vivo bioavailability results from in vitro dissolution data. Any appropriate approach related to this objective may be used for evaluation of predictability. One approach is to predict the in vivo plasma concentration-time profile from the in vitro dissolution data. This procedure is shown in Fig. 7 below, where the in vitro dissolution rate is converted to absorption rate using the IVIVC model and then convolved to predict the plasma
profile. The Cmax and AUC from the predicted profiles should be compared to those from the observed profile to calculate % prediction errors on Cmax and AUC (Fig. 8).
Absolute % prediction error on Cmax and AUC:
Internal predictability: The recommended approach involves the use of the IVIVC model to predict each formulation's (formulations used in developing
the IVIVC) plasma concentration profile (or Cmax and/or AUC for a multiple level C IVIVC) from each respective formulation's dissolution data. Calculate the % prediction error on Cmax and AUC. Criteria
• Average absolute percent prediction error (% PE) of 10% or less for Cmax and AUC establishes the predictability of the IVIVC. In addition, the % PE for each formulation should not exceed 15%.
• If these criteria are not met, that is, if the internal predictability of the IVIVC is inconclusive, evaluation of external predictability of the IVIVC should be performed as a final determination of the ability of the IVIVC to allow the use of in vitro dissolution as a surrogate for bioequivalence.
External predictability: This involves using the IVIVC to predict the in vivo performance of a formulation with known bioavailability that was not used in developing the IVIVC model. Criteria
• The percent prediction error of 10% or less for Cmax and AUC establishes the external predictability of an IVIVC.
• The percent prediction error between 10 and 20% indicates inconclusive predictability and the need for further study using additional data sets. Results of estimation of PE from all such data sets should be evaluated for consistency of predictability.
• The percent prediction error greater than 20% generally indicates inadequate predictability
In the evaluation of internal predictability, it is recommended that the PK parameter estimates used (e.g., for unit impulse response) in predicting the in vivo performance should be the average values or population estimates. Individual PK parameters should not be used to predict individual PK profiles which then are averaged to obtain the predicted average concentration—time profiles. This is due to the following three problems:
1. One does not have dissolution data on the dosage unit that the individual subject was administered. Therefore the input function is based on average parameters. Use of average in vitro parameters and individual in vivo parameters is not appropriate.
2. The percent prediction error calculated in this manner for internal predictability will always look better since the IVIVC was developed using the same individual values, and one is trying to predict the same data using the same individual estimates.
3. Further, since IVIVC will be used to obtain bio waivers when changes are made in future, based on in vitro dissolution data (and no in vivo data), one does not know what the individual parameters will be in each patient that is likely to use the drug. Therefore use of population estimates or mean PK parameters is recommended.
A predictive IVIVC can empower in vitro dissolution to act as a surrogate for in vivo bioavailability/bioequivalence. This can be used to grant biowaivers and to set meaningful dissolution specifications that take into account the clinical consequences.
Biowaivers. The Guidance outlines five categories of biowaivers. These are described in detail below.
1. Biowaivers without an IVIVC.
2. Biowaivers using an IVIVC: Nonnarrow therapeutic index drugs.
3. Biowaivers using an IVIVC: Narrow therapeutic index drugs.
4. Biowaivers when in vitro dissolution is independent of dissolution test conditions.
5. Situations for which an IVIVC is not recommended for biowaivers.
Ideally, one would like to be able to predict the in vivo performance of the drug product from its in vitro dissolution. Therefore, with a predictive IVIVC, waivers for in vivo bioavailability studies may be granted for manufacturing site changes, equipment changes, manufacturing process changes, and formulation composition changes. The biowaivers section deals with changes ranging from situations such as minor changes, which are insignificant for product performance, to major changes for which an IVIVC is not sufficient to justify the change, for a regulatory decision. The IVIVC guidance in this area complements the SUPAC-MR guidance (Scale Up and Post Approval Changes—Modified Release Dosage Forms) . An IVIVC can be used to support those drug product changes in SUPAC-MR that might have required a biostudy. However, there are situations such as those outlined under category 5, where an IVIVC cannot be used.
The mechanism of drug release from the drug product should remain the same when changes are made to a formulation for an IVIVC to be applicable. If the release mechanism changes (e.g., from a diffusion-controlled release to an osmotic release; beads to a matrix tablet), a previously developed IVIVC is not applicable.
The two criteria for granting a biowaiver for a new formulation, where an IVIVC has been established, are that the differences in predicted means of Cmax and AUC are no more than 20% from that of the reference product and, where applicable, the new formulation meets the application or compendial dissolution specifications (see Fig. 9).
Biowaivers with and without an IVIVC
Category 1: Biowaivers Without an IVIVC
This section relates to waivers for lower strengths (beaded capsules as well as tablets), changes made to lower strengths and certain preapproval changes—see biowaivers chapter and IVIVC Guidance for details.
Category 2: Biowaivers Using an IVIVC: Nonnarrow Therapeutic Index Drugs 
a. Two Formulations/Release Rates
A biowaiver is possible for an ER drug product using an IVIVC developed with two formulations/release rates for (1) Level 3 manufacturing site changes as defined in SUPAC-MR and (2) Level 3 nonrelease controlling excipient changes as defined in SUPAC-MR, with the exception of complete removal or replacement of excipients (see below).
b. Three Formulations/Release Rates
A biowaiver is possible for an ER drug product using an IVIVC developed with three formulations/release rates (or developed with two formulations/release rates with establishment of external predictability) for (1) Level 3 process changes as defined in SUPAC-MR; (2) complete removal of or replacement of nonrelease controlling excipients as defined in SUPAC-MR; and (3) Level 3 changes in the release controlling excipients as defined in SUPAC-MR.
c. Biowaivers for Lower Strengths
If an IVIVC is developed with the highest strength, waivers for changes made on the highest strength and any lower strengths may be granted if these strengths are compositionally proportional or qualitatively the same, the in vitro dissolution profiles of all the strengths are similar, and all strengths have the same release mechanism.
d. Biowaiver for New Strengths
This biowaiver is applicable generally to strengths lower than the highest strength (in some instances under an NDA (such as for compositionally proportional formulations), waiver for higher strengths may be possible if scientifically justified especially using an established IVIVC). For details on biowaiver and criteria for new strengths (in an NDA or an ANDA as a generic), see biowaivers chapter.
e. Changes in Release-Controlling Excipients
Changes in release-controlling excipients in the formulation should be within the quantitative range of release-controlling excipients
(used in the different release rate formulations) of the established correlation.
f. Obtaining Category 2a, 2b, and 2c Biowaivers: The difference in predicted means of Cmax and AUC should be no more than 20% from that of the reference product and, where appropriate, the new formulation should meet the application/compendial dissolution specifications.
Category 3: Biowaivers Using an IVIVC: Narrow Therapeutic Index Drugs 
If external predictability of an IVIVC is established, the following waivers (all waivers described under category 2 above including major site changes and nonrelease-controlling excipient changes) are possible if at least two formulations/release rates have been studied for the development of the IVIVC.
a. Manufacturing changes
A biowaiver is possible for an ER drug product using an IVIVC for (1) Level 3 process changes as defined in SUP AC-MR; (2) complete removal of or replacement of nonrelease-controlling excipients as defined in SUP AC-MR; and (3) Level 3 changes in the release-controlling excipients as defined in SUPAC-MR.
b. Biowaivers for Lower Strengths—see category 2c above for details c. Approval of New Strengths—see category 2d above for details Obtaining category 3c biowaivers: see requirements for obtaining 2d biowaivers d. Changes in Release-Controlling Excipients—see category 2e above e. Obtaining Category 3a and 3b Biowaivers: see requirements under category 2f above.
Category 4: Biowaivers When In Vitro Dissolution Is Independent of Dissolution Test Conditions 
Situations in which biowaivers are likely to be granted for both narrow and nonnarrow therapeutic index drugs:
a. Categories 2 and 3 biowaivers are likely to be granted with an IVIVC established with one formulation/release rate.
b. Obtaining Category 4 Biowaivers
• Biowaivers may be granted if dissolution data are submitted in application/compendial medium and in three other media (e.g., water, 0.1 NHCl, USP buffer at pH 6.8) and the in vitro dissolution is shown to be independent of dissolution test conditions after the change is made in drug product manufacturing.
• The difference in predicted means of Cmax and AUC should be no more than 20% from that of the reference product and, where appropriate, the new formulation should meet the application/compendial dissolution specifications. For new strengths, see 2d above.
Category 5: Situations for which an IVIVC Is Not Recommended 
a. Approval of a new formulation of an approved ER drug product when the new formulation has a different release mechanism.
b. Approval of a dosage strength higher or lower than the doses that have been shown to be safe and effective in clinical trials.
c. Approval of another sponsor's ER product even with the same release-controlling mechanism.
d. Approval of a formulation change involving a nonrelease-controlling excipient in the drug product that may significantly affect drug absorption.
Setting Dissolution Specifications . Once an IVIVC is developed, this should be used to set dissolution specifications for the product. An IVIVC provides in vivo relevance to in vitro dissolution specifications, beyond batch-to-batch quality control. In this approach, the in vitro dissolution test becomes a meaningful predictor of in vivo performance of the formulation, and dissolution specifications may be used to minimize the possibility of releasing lots that would be different in in vivo performance.
1. Setting Dissolution Specifications Without an IVIVC
• The recommended range for dissolution specifications at any time point is ±10% of the label claim deviation from the mean dissolution profile obtained from the clinical/bioavailability batches. In certain cases, reasonable deviations from the ±10% range can be accepted provided that the range at any time point does not exceed 25%. Specifications greater than 25% may be acceptable based on evidence that lots (side batches) with mean dissolution profiles that are allowed by the upper and lower limits of the specifications are bioequivalent.
• A minimum of three time points are recommended to set the specifications. These time points should cover the early, middle, and late stages of the dissolution profile. The last time point should be the time point where at least 80% of drug has dissolved, or the time when the plateau of the dissolution profile has been reached.
• Specifications should be established based on average dissolution data (n= 12) for each lot under study, equivalent to USP Stage 2 testing. Specifications that allow all lots to pass at Stage 1 of testing may result in lots with less than optimal in vivo performance passing these specifications at USP Stage 2 or Stage 3. USP acceptance criteria for dissolution testing are recommended unless alternate acceptance criteria are specified in the ANDA/ NDA.
2. Setting Dissolution Specifications Where an IVIVC Has Been Established
If an IVIVC has been established, it should be used to set dissolution specifications. Optimally, specifications should be established such that all lots that have dissolution profiles within the upper and lower limits of the specifications are bioequivalent. Less optimally but still possible, lots exhibiting dissolution profiles at the upper and lower dissolution limits should be bioequivalent to the clinical/bioavailability lots or to an appropriate reference standard.
a. Level A Correlation Established
• Specifications should be established based on average data (n=12).
• A minimum of three time points that cover the early, middle, and late stages of the dissolution profile is recommended to establish the specifications. The last time point should be the time point where at least 80% of drug has dissolved or the time where the plateau of the dissolution profile has been reached.
• Predict the plasma concentration time profile using convolution techniques or other appropriate modeling techniques and determine whether the lots with the fastest and slowest release rates that are allowed by the dissolution specifications result in a maximal difference of 20% in the predicted Cmax and AUC (see Fig. 10). An established IVIVC may allow setting wider dissolution specifications. This would be dependent on the predictions of the IVIVC (i.e., 20% differences in the predicted Cmax and AUC). However, if based on the IVIVC, the dissolution specifications justified are less than the 20% range allowed with an IVIVC, a minimum range of 20% will be generally allowed unless there are clinical concerns.
• USP acceptance criteria for dissolution testing are recommended unless alternate acceptance criteria are specified in the ANDA/ NDA.
b. Multiple Level C Correlation Established
If a multiple-point Level C IVIVC has been established, establish the specifications at each time point such that there is a maximal difference of 20% in the predicted Cmax and AUC. Additionally, the last time point should be the time point where at least 80% of drug has dissolved.
c. Level C Correlation Based on Single Time Point Established
This one time point may be used to establish the specification such that there is not more than a 20% difference in the predicted AUC and Cmax. At other time points, the maximum recommended range at any dissolution time point specification should be ±10% of label claim deviation from the mean dissolution profile obtained from the clinical/bioavailability lots. Reasonable deviations from ± 10% may be acceptable if the range at any time point does not exceed 25%.
If the release characteristics of the formulation can be described by a zero-order process for some period of time (e.g., 5%/hr from 4 to 12 hours), and the dissolution profile appears to fit a linear function for that period of time, a release-rate specification may be established to describe the dissolution characteristics of that formulation. Such a specification may provide for a better control of the in vivo performance of the product. A release rate specification may be (i) an addition to the specifications established on the cumulative amount dissolved at the selected time points, or (ii) may be the only specification along with a cumulative dissolution specification for time when at least 80% of drug has dissolved.
IVIVC can impart in vivo meaning to the in vitro dissolution test and can be useful as surrogate for bioequivalence. IVIVCs can thus decrease regulatory burden by decreasing the number of biostudies required in support of a drug product. As an additional benefit to the sponsors, IVIVC can support wider in vitro dissolution specifications, where justified. FDA strongly encourages the development and evaluation of IVIVCs during ER product development. Generally IVIVC development adds value to the overall drug development process by providing an understanding of the relevance of the in vitro dissolution data leading to better utilization of the in vitro dissolution test. Usually this IVIVC development can be done without conducting new studies. One can use the early development studies where multiple releaserate formulations are generally incorporated in the bioavailability studies. IVIVCs can thus be useful in decreasing the regulatory burden with no undue penalty to the companies that develop these correlations.
EMEA GUIDANCE THAT DEALS WITH IVIVC 
The EMEA Guidance on Quality of MR products and transdermal products covers some of the considerations in development and evaluation of IVIVC and some applications of IVIVC. Similar to U.S. FDA, sponsors are asked to consider development of an IVIVC. If an IVIVC is established, the dissolution test, after proper validation, can be used as a "qualifying control method with in vivo relevance" rather than just a quality control test.
• Levels of correlations are defined in a similar manner to the FDA Guidance.
• Development of IVIVC: Development considerations of levels A, B, and C IVIVC are briefly discussed in this guidance. For a level A IVIVC, generally one formulation tested at different dissolution conditions should be compared to aqueous solution. This seems to be different (although not explicit) from the FDA Guidance where there is a need to study multiple release-rate formulations.
• Evaluation of predictability: Methods and criteria for predictability are the same as in the FDA Guidance; however, there is no explicit discussion of situations with condition-independent dissolution or narrow therapeutic index drugs.
• Applications—Biowaivers: While the FDA Guidance provides detailed situations for biowaivers, the EMEA Guidance provides a summary to state that when a Level A IVIVC has been established and the release specification is not changed, type II
variations (e.g., major changes in nonrelease-controlling excipients, insignificant changes in release-controlling excipients or major changes in method of manufacturing) may be accepted on the basis of in vitro data, the therapeutic index of the drug substance and predictability of the IVIVC. In general, BA/BE data are needed for products with an established level B or C correlation or no IVIVC, unless justified.
Applications—Dissolution specifications: If IVIVC is established, it is used to set specifications. However, there are some differences from the FDA Guidance.
(A) Level A: The specification is based on a 1:1 correlation between the dissolution profile in vivo and in vitro (FDA Guidance is not restricted to a 1:1 correlation).
(B) Level B correlation can also be used to set specifications, although methodology details are not provided (Level B correlations are not useful for waivers or setting dissolution specifications according to the FDA Guidance).
(C) For any level of correlation, i.e., levels A, B, C, or multiple level C, specifications should be set such that the maximal difference in predicted AUC is 20% and, predicted Cmax only if relevant (FDA Guidance requires both AUC and Cmax).
1. Marroum, P.J. Presentation on Bioavailability/Bioequivalence for Oral Controlled Release Products, Controlled Release Drug Delivery Systems: Scientific and Regulatory Issues, Fifth International Symposium on Drug Development, East Brunswick, NJ, May 15-17, 1997.
2. FDA, Guidance for Industry: SUPAC-MR: 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, September 1997.
3. FDA, Guidance for Industry: Bioavailability and Bioequivalence Studies for Orally Administered Drug Products—General Considerations, March 2003.
4. FDA, Guidance for Industry: Extended Release Oral Dosage Forms: Development, Evaluation, and Application of in vitro/in vivo Correlations, September 1997.
5. Code of Federal Regulations 21 section 320.
6. FDA, Guidance for Industry: Food-effect Bioavailability and Fed Bioequivalence Studies, December 2002.
7. FDA, Guidance for Industry: Providing Clinical Evidence of Effectiveness for Human Drug and Biological Products, May 1998.
8. El-Tahtawy, A.A.; Jackson, A.J.; Ludden, T.M. Comparison of Single and Multiple Dose Pharmacokinetics Using Clinical Bioequivalence Data and Monte Carlo Simulations. Pharmaceutical Research 1994, 11 (9), 1330-1336.
9. Uppoor, V.R.S. Regulatory Perspectives on in vitro (Dissolution)/in vivo (Bioavailability) Correlations. Journal of Controlled Release 2001, 72, 127132.
10. Yamaoka, K.; Nakagawa, T.; Uno, T. Statistical Moments in Pharmacokinetics. Journal of Pharmacokinetics and Biopharmaceutics 1978, 6(6), 547-548.
11. Marroum, P.J. Cardizem CD, Biopharmaceutics Review, Center for Drug Evaluation and Research, Food and Drug Administration, June 1991.
12. Eddington, N.D.; Marroum, P.; Uppoor, R.; Hussain, A.; Augsburger, L. Development and Internal Validation of an in vitro-in vivo Correlation for Hydrophilic Metoprolol Tartrate Extended Release Tablet Formulations. Pharmaceutical Research 1998, 15, 464-471.
13. Mahayni, H.; Rekhi, G.S.; Uppoor, R.S.; Marroum, P.; Hussain, A.S.; Augsburger, L.L.; Eddington, N.D. Evaluation of External Predictability of an in vitro-in vivo Correlation for an Extended-Release Formulation Containing Metoprolol Tartrate. Journal of Pharmaceutical Sciences, 2000, 89(10), 13541361.
14. Langenbucher, F. Numerical Convolution/Deconvolution as a Tool for Correlating in vitro and in vivo Drug Availability. Pharm. Ind. 1982, 44 (11), 1166-1171.
15. Langenbucher, F. Improved Understanding of Convolution Algorithms Correlating Body Response with Drug Input. Pharm. Ind. 1982, 44 (12), 12751278.
16. Gillespie, W.R. Convolution-Based Approaches for in vivo-in vitro Correlation Modeling, in in vitro-in vivo Correlations. Advances in Experimental Medicine and Biology 1997, 423, 53-65.
17. Langenbucher, F.; Mysicka, J. In vitro and in vivo Deconvolution Assessment of Drug Release Kinetics from Oxprenolol Oros Preparations. British Journal of Clinical Pharmacology 1985, 19 (Suppl. 2), 151S-162S.
18. Brockmeier, D. In vitro-in vivo Correlation, A Time Scaling Problem? Evaluation of Mean Times. Arzneim-Forsch (Arzneimittel-Forschung) 1984, 34 (11) 1604-1607.
19. EMEA Guideline CPMP/QWP/604/96: CPMP Note for Guidance on Quality of Modified Release Products: A: Oral Dosage Forms B: Transdermal Dosage Forms Section 1 (Quality), 29 July 1999.
Was this article helpful?