In vitro-in vivo correlation (IVIVC) technology allows formulation and manufacturing (CMC) professionals to understand how dissolution parameters affect in vivo drug performance. IVIVC can be performed at one of three levels—A, B, and C. Getting the most out of this tool requires knowing when and how to use which level of IVIVC. Here are some questions (and answers) that I’ve gotten regarding best practices for this approach.
Q: Do you prefer using an arithmetic mean or a geometric mean when running Level C correlation on mean data?
A: You can use arithmetic mean or geometric mean. The difference is not important if you don’t have any outliers. Arithmetic means are very sensitive to extreme values. If you want to be more stable, use the geometric mean and median.
You can also use all the raw data to establish the IVIVC. Then the linear regression will result in parameters (slope and intercept) close to those of arithmetic means but variability will be associated with the estimates.
If you use arithmetic means, you have to take into account that if you use all the raw data to perform IVIVC Level C, your IVIVC Level C is going to go through the arithmetic mean. Geometric means are closer to what you calculate in bioequivalence (BE) studies because you make a log transformation in these studies.
Q: When is a Level C IVIVC useful compared to a Level A IVIVC?
A: A Level C IVIVC can be very useful when you don’t have enough data to make a Level A, or you don’t know if it’s possible, and you don’t want to invest a lot of time. You can easily try with mean values for Cmax, AUC, and Tmax to see if you have a match with your dissolution data. If yes, and it matches for the three parameters, then you can consider investing the time to perform Level A IVIVC.
The second case is when, for example, you have a pro drug which has a limited plasma concentration compared to the active metabolite. In this case, the metabolite concentration levels will reflect the initial pre-systemic metabolism, but it will also reflect the current elimination for the plasma. Thus, performing Level A IVIVC would be misleading.
Lastly, sometimes you don’t have enough data. For example, you may lack intravenous (IV) or fast release data as the unit impulse response (UIR) to perform deconvolution. In this case, you’re forced to make a Level C IVIVC. These are the three main cases in which a Level C IVIVC is indicated.
Q: Can I perform a Level C IVIVC with only one or two formulations?
A: No. But, you can use two formulations when performing Level A IVIVC. That’s because Level A uses all the plasma concentration data. By contract, Level C restrict the plasma concentration curve to one point which can be Cmax or the AUC. If you have only two profiles, or two formulations A and B, and you’ve restricted it to Cmax and AUC and only to the mean values, of course, you will have a straight line between the two mean values. That is one of the drawbacks of Level C IVIVC. Thus, this approach is less powerful than Level A.
Q: Could Level C IVIVC be used as a surrogate of in vivo bioequivalence?
A: According the guidelines, no. But, some authorities accept it when you can correlate all the bioavailability parameter values (Cmax and AUCs) with dissolution. That means you can Level C to predict the risk that the formulations are not bioequivalent. In this case, it can be used as an in vivo surrogate, and some authorities will accept it.
One advantage of Level C is that you don’t have to perform any deconvolution. That could be the case if you don’t have the UIR in your dataset. That means as you could use the Cmax and AUC values directly. This can be a simple, straightforward approach. But, if you have the data to perform Level A IVIVC, the authorities are not going to accept a Level C IVIVC.
Q: Whenever we do dissolution testing, the time points in the dissolution test don’t match the in vivo time points. Can we do perform an IVIVC for such cases?
A: For Level A, no unless you can interpolate the dissolution data or perform time scaling. For Level C, yes. Because for Level C, you restrict the plasma concentration data to Cmax and AUC. You don’t have any notion of time. That means you don’t need to be at the time of Cmax (Tmax) to make the correlation between the Cmax and the dissolution. You’re only trying to see which dissolution data will fit with the in vivo data.
Level A is at the opposite extreme. This approach requires data between your dissolution and in vivo studies which are going to fit each other temporally. Or you will need to model your dissolution data to extract the time points which are going to fit in vivo data after time scaling.
Q: When do we need to perform deconvolution in IVIVC?
A: For Level A IVIVC, you need to perform either a classical numerical, Wagner-Nelson, or Loo-Riegelman deconvolution to extract the absorption curve. That is for Level A, not Level C. In Level C, you don’t need the absorption curve; you are going to work mainly with the PK parameters. So deconvolution is for Level A, not Level C.
Interested in learning how to perform Level C IVIVC in Phoenix? Please watch the webinar I presented on this topic.