The dissolution profiles of ibuprofen are summarized in Fig. 1 for the two-stage methodology and Fig. 2 for the transfer methodology. In two-stage dissolution methods, the dissolution curves obtained using experimental conditions within the above-mentioned constraints exhibited similar profiles regardless of buffer volume, buffer species, buffer pH, rotational speed, and duration of the first (gastric) stage (Fig. 1). In the transfer method dissolution profiles, only the dissolution profiles in the small intestinal vessel(s) are plotted (Fig. 2). The dissolution profiles from Institutes D and G both achieved 100% dissolution, in line with the solubility of ibuprofen at intestinal pH. By contrast, the profile from Institute F reached only 50% dissolution of ibuprofen, which was explained post hoc by the fact that 50% of the drug was retained in the gastric stage rather than being transferred. The profile from Institute H exhibited even lower dissolution of ibuprofen because the second (“duodenal”) chamber was held at a constant volume of 30 mL, by discarding fluid transferred out of that chamber. Since Institute H only measured the drug concentration in the second chamber, ibuprofen dissolution was essentially measured in just 30 mL of the 0.01N HCl/SIF mixture.

Ibuprofen dissolution profiles generated using various two-stage method setups. The first stage dissolution was conducted for either the first 20 or 30 min before initiating the second, small intestinal, stage. *Institute B1 -HPLC method and Institute B2 – UV probe

Ibuprofen dissolution profiles in the small intestinal chamber generated using various transfer model setups

Since ibuprofen is a weakly acidic drug and the small intestinal environment is favorable for its dissolution, for modeling purposes, it is necessary to account for all transferred material from the gastric to the small intestinal stage. For this reason, the observed concentration and the volume out of duodenal chamber at each time point were used to generate the full dissolution profiles in the small intestinal region for the Institute H results (Fig. 2).

The dissolution profiles of dipyridamole are summarized in Fig. 3 for the two-stage methodologies and in Fi, 4 for the transfer methodologies. Using two-stage dissolution methodologies, three institutes (Institute A, B, and D) displayed complete dissolution of dipyridamole and no or close to no precipitation regardless of buffer species and capacity, while Institute C observed precipitation following pH adjustment to 6.8 with NaOH (Fig. 3).

Dipyridamole dissolution profiles generated using the two-stage method. The first stage dissolution was conducted for either the first 20 or 30 min before the second, small intestinal condition was initiated. *The buffer at the second stage: SIF – simulated intestinal fluid, 1/10 SIF – 1/10th concentration of SIF, Maleate – 7 mM Maleate buffer

In transfer methodologies, only the dissolution profiles in the small intestinal vessel(s) are plotted (Fig. 4). The results from Institutes D, E, and I exhibited more than 80% dissolution of dipyridamole, while Institutes G and F reported only 50–60% of dipyridamole dissolution at the end of the experiment. As for ibuprofen, Institute H reported very low dissolution of dipyridamole because, once again, the dissolution profile was calculated based on the volume in the second (“duodenal”) chamber multiplied by the drug concentration at the time points of sample withdrawal. Dipyridamole is a weak base drug and the gastric environment, pH, stirring speed, volume, and the residence time might all be important to describe its dissolution from the tablet. Since the pKa of dipyridamole (pKa 6.4) is close to the pH of the dissolution buffers representing the small intestine (pH 6.5–7.5, depending on the institute), the buffer capacity, species and volume are all expected to affect its dissolution. This was reflected by the large range of results recorded.

Dipyridamole dissolution profiles in the small intestinal region generated using the transfer model

The plasma profiles of ibuprofen were simulated using the dissolution profiles produced by the WG as the in vivo dissolution input. The purpose of this simulation was not to provide a fully accurate prediction of the observed clinical plasma profile (e.g., no attempt was made to further optimize, e.g., disposition parameters in the model) but rather directed at assessing the criticality of different in vitro conditions by different members of the WG (including pH, buffer species, buffer capacity, volumes of media, media change conditions and stirring rate). The prediction of ibuprofen absorption after oral dosing of 400 mg with the biopredictive dissolution profiles from the two-stage test was all BE with the clinical data (Fig. 5). The simulation results presented in Table IV suggest that all of the two-stage dissolution conditions studied were indeed biopredictive for ibuprofen, even though there were some differences in buffers, transfer rate/timing and other experimental conditions. Part of the reason for this is that for weakly acidic drugs like ibuprofen, the environment in the small intestine is far more favorable than the one in the stomach for dissolution, so as long as the buffer pH is higher than the drug’s pKa and the pH is adequately maintained over the experiment, modest differences in dissolution conditions are unlikely to affect the outcome of the simulation with respect to BE.

Predicted plasma profiles of ibuprofen based on biopredictive dissolution profiles. Dark blue dotted lines represent 80% and 125% of the average clinical plasma profile (23). a represents simulated results based on two-stage dissolution methodologies and b represents simulated results based on transfer methodologies

With respect to the transfer test, only the results from Institutes D, G, and H fell within the BE limits (Fig. 5). The failure of results from Institute F to meet the BE limits can be traced back to the experimental design: in the Institute F experiments, only 50% of ibuprofen was transferred to the small intestinal chamber from the gastric chamber. Even though the dissolution profile was re-generated to assume the same concentration was achieved in the rest of the small intestine as in the duodenal chamber, the concentration of ibuprofen in the duodenal chamber used to generate the simulated in vivo dissolution profile might have been too low to evaluate the biorelevant dissolution. Since a large volume of acid comes from the gastric chamber into the duodenal chamber (which holds only 30 mL of the small intestinal buffer), the pH in the duodenal chamber is lowered, and thus limits the extent of ibuprofen dissolution, leading to an underestimation of the predicted plasma profile of ibuprofen. Since ibuprofen dissolves well at intestinal pH (as shown in Institute D, and G setups) and is completely absorbed in vivo (23, 26), it seems that the short duration of dissolution, lowered pH and small volume used to represent the small intestinal environment in the Institute H setup, combine to produce a less extensive dissolution than would occur in the small intestine in vivo.

The plasma profiles of dipyridamole were simulated using the dissolution profiles produced by the WG as the in vivo dissolution input, analogous to the approach used for ibuprofen. Again, the purpose of this simulation was not to provide a fully accurate prediction of the observed clinical plasma profile but rather directed toward assessing the criticality of differences in the in vitro conditions among the different institutes. The results are shown in Fig. 6a and b.

Predicted plasma profiles of dipyridamole based on biopredictive dissolution profiles. Dark blue dot lines represent 80% and 125% of the calculated mean clinical plasma profile (43). a represents simulated results based on two-stage dissolution methodologies and b represents simulated results based on transfer methodologies

While nine of the eleven simulated plasma profiles satisfied equivalency with the clinical data with respect to Cmax but overestimated AUC, the dissolution profile by Institute F led to a simulation which satisfied the BE requirements for AUC0-24 but missed on Cmax (Table V). An issue with all of the simulations is that the clinical data chosen as the reference in vivo data exhibited a Tmax of 0.5 h, while the in silico simulations based on the biorelevant dissolution profiles consistently predicted a longer Tmax of  ~1.5 h (27). BCS class IIb drugs like dipyridamole dissolve well in the gastric environment and the earlier Tmax suggests a faster gastric emptying time. Indeed, the stomach has a rapid emptying mechanism described as “Magenstrasse,” which might potentially explain the relatively faster Tmax observed in the reference clinical data (28, 29). It also should be mentioned that other clinical data reported by Gregov et al. exhibited a Tmax of 1.06–1.58 h over the 25–200 mg dose range, in agreement with the simulation results, and, as seen, high inter-study and subject variabilities in the PK study of dipyridamole have been reported (10, 24, 30).

The dose of 50 mg of dipyridamole in this set of experiments is relatively low and its pKa is high (pKa 6.4). Since the total aqueous volume available for dissolution at the end of the experiment was  ~300 to 500 mL in all cases (except for Institute H), most of the dipyridamole dose dissolved, with little precipitation, and hence more absorption was predicted by most simulations than was observed clinically.

Although it is beyond the scope of the current evaluation, more investigation of the precipitation kinetics and of the intestinal pH, along with buffer capacity, would be required to fully simulate the entire plasma concentration profile for this poorly soluble but highly permeable weak base. Since dipyridamole is a weakly basic drug, the gastric environment is the more favorable for its dissolution, and it is consequently necessary to model the pH and dissolution time in the stomach (via the gastric emptying kinetics) in a way that reflects the physiology closely in order to attain a more predictive dissolution and simulation. Additionally, media volumes and hydrodynamics of the test method may need to be adjusted to adequately reflect the supersaturation and precipitation kinetics.