ISSUES IMPORTANT TO IN VITRO-IN VIVO CORRELATION

IVIVR Workshop Open Discussion

26

Although this book addresses numerous topics in the area of in vitro-in vivo cor­relation (IVIVC), this chapter summarizes some of the scientific and regulatory issues important to IVIVC and the Draft FDA Guidelines on IVIVC. This chapter was based on the open discussion sessions at the IVIVR Workshop in Baltimore, Maryland on September 4---{), 1996. The chapter represents a compilation of comments and the con­sensus of the attendees at the meeting. This chapter is organized into five major topics with each issue listed under the topic and the consensus response from the attendees. For those few times that a consensus could not be reached, the different points of view are provided.

GASTROINTESTINAL PHYSIOLOGY

1. Our knowledge about GI physiology is greater than ever before but more re­search is needed.

The consensus was that this statement is true and a better understanding of GI physiology would provide a better understanding as to the limits of develop­ing and applying IVIVCs.

2. The use of animal models for IVIVC investigation needs to be considered.

The consensus was that animal models should not be used in the regulatory decision process for oral or extended release products. However, they may be a useful tool for product development.

3. In vitro and in situ animal and human systems are important to IVIVC.

The consensus was that these systems, in principal, have a role in IVIVC re­search but the role is not well defined except for permeability.

4. New in vitro and in situ animal and human models need to be developed and correlated to man.

The consensus was that additional research on permeability and regional per­meability will be valuable for the development of ER products.

In Vitro-in Vivo Correlations, edited by Young et at. Plenum Press, New York, 1997 289

290 Issues Important to in Vitro-in Vivo Correlation

5. Investigation of factors affecting BA.

This is an important area to investigate but the consensus was that it is out­side the realm of this conference on IVIVC.

DISSOLUTION

1. In vitro dissolution can be used for QC and as a surrogate for in vivo (IVIVC).

It was agreed that in vitro dissolution can be used for both QC and as a surro­gate for in vivo. More importantly, however, was the agreement that the sys­tems are not required to be the same. If they are not the same, a major concern was that the QC dissolution specifications would be based on batch history and not the anticipated in vivo consequences to the formulations.

A concern by a number of individuals was the appropriateness of using a different dissolution test for QC release vs. IVIVC. The problem is that the QC specifications from a dissolution system different from the IVIVC system does not ensure comparable in vivo response for different formulations.

In a similar vein, a number of attendees stated, "how can the dissolu­tion specifications be developed from a different system than the IVIVC dis­solution system which is a surrogate for the in vivo, if we are concerned about in vivo response to the product?" It was agreed that there may be finan­cial and/or logistical reasons why a company would not be able to use the dissolution system from the IVIVC, in this case the ideal situation would be to relate the two dissolution systems in some manner (e.g., mapping the for­mulations from one dissolution system to another).

2. We must define and further investigate the methodology needed to determine differences in dissolution.

It was agreed that further research in this area is appropriate.

3. We must define what is an acceptable vs. unacceptable difference.

For biowavers which require IVIVC, the acceptance criteria for the compari­son of curves should be based on in vivo consequences not just the in vitro curves.

4. The guidance states "If in vitro dissolution is shown to be independent of disso-lution conditions such as pH and agitation .... ". We must define independent.

The consensus was that the variables and range of variables to be investi­gated need to consider the characteristics of the product. The variables one might consider are: pH, agitation, ionic strength, lipids/surfactants, apparatus.

5. Optimization of dissolution systems to obtain IVIVC is important.

The attendees agreed that the optimization process can go to the extreme. A better approach would be to describe the IVIVC by more complex mathe­matical/statistical models without making heroic efforts to change dissolu­tion. This assumes that the existing dissolution system is discriminating.

Issues Important to in Vitro--in Vivo Correlation 291

6. We should be developing mechanistic (mimicking physicallbiological systems) as well as empirical dissolution systems, but this is not a regulatory issue.

The consensus was that theses dissolution systems may need to be developed and the FDA must be open to accepting these systems for IVIVC and specifi­cations.

7. The dissolution system used for IVIVC should be a discriminatory system based on the critical manufacturing variables that will affect in vivo response.

The attendees agreed with this statement and emphasized the fact that the critical manufacturing variables should be "critical" based on in vivo conse­quences not in vitro consequences.

8. Dissolution specifications should be developed from IVIVC models.

The attendees agreed with this statement but again emphasized that, if the QC dissolution system is different, QC specifications would not be based on IVIVC. Some type of relationship (possibly mapping) may be required to re­late QC dissolution to IVIVC dissolution. The attendees believed that the specifications should be based on IVIVC models, if possible, but that the use of an IVIVC model should not be a requirement.

9. Should the formulations be bioequivalent at the extremes of the dissolution specifications or should the extremes of the dissolution specifications be bio­equivalent to the pivotal batch?

The consensus was that the upper/lower dissolution specification limits should be equivalent as a product but studies should not be required to verify that the top and bottom are equivalent. There was not, however, an agreement about the definition or criteria of equivalent. It was agreed that equivalent is a broader term than bioequivalent.

Differing opinions existed as to the criteria for equivalent. Some of the comments are provided below: I) The development of the upper and lower dissolution specifications should be based on a desire to have the formula­tions bioequivalent following our present criteria (if we do not desire bio­equivalence at our present criteria, the science is moving backwards), 2) A 20% mean range for Cmax and AVC is a reasonable position at this time for ER products, 3) A 20% mean range for Cmax and AVC is arbitrary and is a step backward in our science, 4) A 10% difference in the mean Cmax and AVC between the pivotal batch and upper or lower specification is reason­able, 5) a 10% difference in the mean Cmax and AVC is appropriate for the upper and lower specifications, 6) The innovator must be bioequivalent across all batches otherwise new generic products may target the top or bot­tom of the innovator and therefore produce a wide range of generic products which may not be bioequivalent to all batches of the innovator, 7) Bioequiva­lence standards that are drug specific may be the most appropriate and the criteria should depend on the drug product itself rather than 1 criteria for all products.

There are really two situations: I) specifications developed from IVIC and 2) specifications developed from side batches when an IVIVC does not

292 Issues Important to in Vitro-in Vivo Correlation

exist. In the second situation, a clinical study may be needed to confirm equivalence.

STUDY DESIGN

1. Clinical studies should be conducted as early as possible after choosing the dos­age form.

This was strongly recommended by attendees.

2. Food studies may play an important role in IVIVC and should be performed when necessary.

If an IVIVC developed in the fasted state is capable of predicting the in vivo characteristics in the fed state, no additional IVIVC work under fed condi­tions is required. If not, more investigation may be required and one might even consider performing an IVIVC study under fed conditions.

3. Crossover studies are preferable to parallel designed studies. If parallel or cross-over studies are used in IVIVC, a common reference is needed.

The attendees recommended that the statement "a common reference is needed" in the Draft IVIVC Guidelines be changed to "a common reference may be useful". There was no recommendation on the nature of the common reference but further investigations may be needed to determine the optimal choice (solution, IR, IV) for specific circumstances.

4. How many formulations are needed to develop an IVIVC model? What factors does it depend on?

Multiple formulations should be used with the exception of one formulation as stated in the Draft IVIVC Guidelines. An exact number could not be de­fined.

5. How many subjects are needed for an IVIVC study?

The number of subjects required for IVIVC study needs to be determined af­ter the modelling method and a statistical criteria are established.

6. One should consider the potential site of absorption and release from the ER for­mulation when choosing the unit impulse (solution vs. IR solid dosage form vs. IV). The unit impulse should be chosen to match the sites of absorption as much as possible.

The attendees agreed with this statement.

MODELLING

1. The present deconvolution and convolution methods have provided a starting point for IVIVC but further research in these two area is required (e.g., Verotta, Gillespie ).

Issues Important to in Vitrtr-in Vivo Correlation 293

The consensus was that more research is needed in these areas and that a user friendly deconvolution and convolution computer program is needed.

2. Both empirical and mechanistic models (e.g., accounting for in vivo release, mechanism of release, dosage form-GI interaction) need to be investigated and evaluated further.

The consensus was that more research is required for both types of modelling approaches.

3. If the in vivo or in vitro profiles is faster, the dissolution system should be changed or alternative models (e.g., time scaling) developed.

The consensus was both approaches were appropriate.

4. We should be developing Level A correlations with the entire curves. What about early vs. late points?

The consensus was that the in vitro dissolution should predict the entire in vivo curve up to the post-absorption phase, which needs to be defined. If, for example, there is site limited absorption then it may not be necessary to use the entire in vitro curve.

5. All formulations need to fit the same model.

Since all formulations within the range of the formulation and manufacturing specifications should fit the IVIVC, a single IVIVC should be developed for multiple formulations.

6. We should account for correlation of data within a subject.

New approaches which account for the correlation of data is required to ap­propriately model the IVIVC but a user friendly program is needed in order to make the approach practical.

7. We should account for variability in the in vitro dissolution.

New approaches are needed to account for the variability in the in vitro disso­lution but the mathematical and statistical methods need to be further devel­oped and implemented in a user friendly manner in order to make the approaches practical.

8. Do we develop a Level A for I formulation or multiple formulations (similar to cross formulation approach in Level C)?

Since all formulations within the range of the formulation and manufacturing specifications should fit the IVIVC, mUltiple formulations should be used when developing an IVIVC.

The exception is that stated in the guidelines when the dissolution re­mains the same for dissolution systems that have differences in, for example, pH, agitation, system. In this case, the study is considered a confirmatory study, not a modelling study since the model is based on the in vitro dissolu­tion results.

9. Clarify internal vs. external predictions and when is one better than another?

294 Issues Important to in Vitro--in Vivo Correlation

One major difference between internal vs. external predictions is that the ex­ternal procedure predicts the in vivo response for formulation(s) that are in no way used in the development of the IVIVC model. Internal prediction, on the other hand, uses formulation(s) that are used to develop the model. An exam­ple of the internal procedure is cross-validation where one formulation is left out of the IVIVC model and predicted from the IVIVC model developed from the other formulations. This is repeated for each formulation.

The external prediction procedure will generally prove to be the best but the conditions where it performs better than the internal prediction proce­dure has not been defined. Further research is needed to define the various conditions where the two predictions will result in false positive and false negative findings.

EVALUATION / VALIDATION

1. An investigation of the appropriate prediction error metrics needs to be under­taken.

The attendees agreed with this.

2. More research into the area of model validation or the criteria for IVIVC accept­ability needs to be undertaken.

The attendees strongly agreed with this and stated this should be a major area of future research.

3. A definition of what is acceptable and what is not acceptable needs to be devel-oped.

The attendees strongly agreed with this and stated that it is extremely diffi­cult for the industry to implement validation procedures without a criteria for acceptable and not acceptable. The acceptance criteria for a valid IVIVC may be different for drugs with a narrow therapeutic index.

OTHER ISSUES

1. The biopharrnaceutical classification for ER products needs to be continued.

The biopharmaceutical classification for IR formulations has been beneficial and an expansion into ER formulations would be helpful. Attendees were concerned about the ability to develop such a classification.

2. Additional research on the IVIVCs for highly variable drugs is required.

There was a consensus that further research in this area is required.

3. Methods need to be developed to deal with the following: Truncated absorption (change in F) and saturable presystemic metabolism.

There was a consensus by the attendees that the modelling procedures for these areas is required.

Issues Important to in Vitro-in Vivo Correlation 295

4. IVIVC models need to be developed early to assist in product development.

There was a strong consensus that early development of IVIVC models would be beneficial to product development.

5. Define the role ofIVIVC in generic drugs

IVIVC may be important for generic drugs (e.g., in the development of disso­lution specifications), but further research in the general area of IVIVC is re­quired. The same principles for developing IVIVC models for NDA and ANDA applications should apply.

6. What is the cost / benefit ratio for IVIVC development?

IVIVC can be extremely useful in product development, product optimiza­tion, and post-approval changes. The costibenefit ratio, however, could not be defined but the attendees felt that it was the sponsors decision to determine the cost/benefit ratio for developing IVIVC models.

7. Can this be extended to products administered other than po.

It was agreed that the principals of IVIVC can be extended to products ad­ministered by other routes such as 1M.

8. Explain further what is a formulation.

For IVIVC, a formulation is not defined based on the type or quantity of polymer and/or excipients within the formulation, but a more important issue is the choice of the range of differences between the formulations. Formula­tions may have different release rates but they should posses the same mecha­nism of release.

9. Semantics within the guidelines

It is important for the FDA to realize that although many guidelines use the terms "desires" as something less than "required", the pharmaceutical indus­try interprets the term "desires" as equal to required.