IVIVC: Making In-Vitro Dissolution Predict In-Vivo Performance

Why we want a beaker to predict a patient

Every clinical bioequivalence study costs money, time and volunteer exposure. A dissolution test in a lab, by contrast, is cheap and repeatable — you drop the dosage form into a vessel of fluid and measure how fast the drug dissolves. The dream of an in-vitro-in-vivo correlation (IVIVC) is to make that beaker test predict the plasma curve in a real patient. If a reliable correlation exists, a change to the product can be cleared by re-running the dissolution test instead of dosing humans again.

An IVIVC is only physically plausible when dissolution is the slow, controlling step — when absorption is dissolution-rate-limited. For a fast-dissolving, highly permeable drug, dissolution finishes long before absorption is over, so the beaker has nothing meaningful to predict. That is why IVIVC is most valuable for modified-release products, where the formulation deliberately controls the release rate and that rate governs everything downstream.

The levels of correlation

Not all correlations are equally strong, so the framework defines levels. A Level A correlation is the gold standard: a point-by-point relationship between the entire in-vitro dissolution curve and the entire in-vivo absorption curve. A Level B correlation compares summary statistics — for example, the mean in-vitro dissolution time against the mean in-vivo residence time — using all the data but only as averages. A Level C correlation links a single dissolution point (say, percent dissolved at 60 minutes) to a single pharmacokinetic parameter such as AUC or Cmax. Level A is the most useful because it can predict the whole plasma curve, not just one number.

Building a Level A correlation

1. Make two or three versions of an extended-release product that release at deliberately different rates — slow, medium, fast.
2. Measure each one's in-vitro dissolution profile, and run a clinical study to obtain each one's plasma curve.
3. Deconvolute the plasma curves to recover the in-vivo absorption profile — the fraction of drug absorbed over time.
4. Plot fraction absorbed in vivo against fraction dissolved in vitro at matching times; a straight line near 45° confirms a Level A correlation.
5. Validate by predicting a withheld batch's plasma curve from its dissolution data and checking the prediction error.

Level A correlation - one matched time point pair

  Time (h)   % dissolved in vitro   % absorbed in vivo
     1              22                    20
     2              41                    39
     4              68                    66
     8              95                    93

Fit:  Fa(in vivo) = slope * Fd(in vitro) + intercept
      slope  ~ 0.98   intercept ~ 0  ->  near 1:1 line  =>  Level A

Validation on a new batch:
  predicted AUC = 41.0 ;  observed AUC = 42.3
  prediction error = |41.0 - 42.3| / 42.3 = 3.1%   (well under the 10-15% guide)