Precipitation of a drug in the gastrointestinal (GI) tract, especially to a less soluble polymorph, can negatively impact bioavailability [1]. A precursor to precipitation is generation of a supersaturated solution in the intestinal lumen. For a weakly basic compound, this could occur simply as the drug transits from the acidic stomach into the relatively alkaline duodenum [2]. Supersaturation and precipitation are also impacted by gastric emptying, rate of drug absorption from the intestine, and composition of the intestinal milieu [[3], [4], [5]]. Simulating intestinal precipitation and dissolved drug concentration in vitro is an essential prerequisite for accurate prediction of the plasma concentration–time profile [6]. This entails development of biorelevant in vitro dissolution testing methods that take into account all key parameters influencing drug precipitation.

In the past decade, media simulating gastric and intestinal luminal conditions of pH, bile concentration, buffer capacity and osmolarity have been developed [[7], [8], [9]]. These biorelevant conditions implemented in USP dissolution testing apparati are often sufficient to predict in vivo dissolution of Biopharmaceutics Classification System (BCS) – I drug molecules [10]. However, an absorptive compartment is critical when evaluating supersaturating formulations of low solubility, BCS-II molecules, which by definition have high permeability. Non-sink dissolution testing of such formulations can overestimate drug precipitation and underpredict drug absorption, especially of weakly basic drugs that can completely solubilize in acidic gastric media but can generate metastable supersaturated solutions as they transit through the more alkaline intestinal lumen [11,12]. Rapid absorption from the intestinal lumen reduces the dissolved drug concentration and in turn reduces the driving force for precipitation, which is not possible in a closed dissolution testing apparatus. For instance, Carlert et al. observed significant precipitation of a weakly basic drug during in vitro dissolution testing, but a dose proportional increase in plasma exposure of the drug did not suggest significant precipitation in vivo [6]. In that study, the authors recommended developing in vitro methods where there is continuous removal of drug from the dissolution media for future studies with formulations of BCS-II drug molecules. Psachoulias et al. developed a three compartment in vitro setup consisting of a duodenal, gastric and reservoir compartment [13]. Drug solution was emptied from the gastric compartment at a first order rate, as drug free media from the reservoir simultaneously diluted the drug in the duodenal compartment to simulate gastric emptying and absorption. The sum of incoming flow rates of gastric and reservoir media into the duodenal compartment was made equal to the outgoing flow rate of the duodenal media. This in vitro system accurately predicted the intestinal concentration of dissolved and precipitated weakly basic model compounds due to the simulation of drug removal by dilution. However, the operating parameters of this system were selected arbitrarily and could not be rationally determined based on the absorption rate constant of a drug.

The USP Apparatus 4 and biphasic dissolution testing system can mimic in vivo drug removal rates, but have limitations [14,15]. USP Apparatus 4 uses unphysiologically large volumes of the dissolution media, while the biphasic system lacks a membrane, allowing uncontrolled partitioning of drug species [12]. Additionally, common sampling and analytical methods may not accurately distinguish free drugs from bound species, potentially overestimating the bioavailability of amorphous solid dispersions (ASDs) [16,17]. Membrane based assays such as Parallel Artificial Membrane Permeability Assays (PAMPAs), ultrathin large area membrane (UTLAM), and diffusion cells (e.g., side-by-side, MicroFLUX) provide insights into drug permeation but struggle with scaling up for physiologically relevant absorption rates due to their small surface area-to-volume (SA/V) ratios resulting in extended, non-physiological experimental times (e.g., 16 h versus 2–4 h of gut transit) [[18], [19], [20], [21]]. Human SA/V is 1.9–2.3 cm−1, much higher than the 0.23 cm−1 of flat sheet membranes [14,22]. Vertical membrane flux cells (SA/V of 1 cm−1) and hollow fiber systems (SA/V of 2.3 cm−1) offer more relevant in vitro SA/V ratios [[22], [23], [24], [25], [26]].

A recently introduced Artificial Gut Simulator (AGS), equipped with an absorption module, can be used to conduct biorelevant dissolution testing with small samples [11,12,24]. The absorption module consists of a manifold of hollow fibers that provides a large surface area for absorption. During dissolution testing, the absorption module is suspended in the drug donor compartment wherein the concentration gradient across the hollow fiber membrane drives drug molecules to diffuse into the intraluminal fluid. Drug free receiver media is continuously pumped through the hollow fibers, thereby maintaining an absorptive sink. Using a previously developed theory of the device, the AGS operating parameters can be tuned to absorb drug at a biorelevant rate [24].

It has been demonstrated that for a BCS-II compound like ketoconazole and the amorphous solid dispersion formulations thereof, precipitation may be largely overestimated in the absence of an absorptive sink [6,11,12]. With an AGS tuned to absorb drug at a physiological rate, a reduction in both amorphous and crystalline precipitation may be achieved [11,12]. However, these results have not been verified against in vivo observations. The objectives of the present study are twofold. First is to demonstrate that intestinal dissolved drug concentration and precipitate fraction upon dosing a weakly basic BCS-II compound such as dipyridamole (DPD) are more closely simulated in the presence of an absorptive sink with AGS than with a “closed no-absorption” dissolution testing system. The second objective is to demonstrate that a scheme introduced earlier with caffeine to successfully predict intestinal absorption with AGS and subsequently the plasma concentration–time profile [24], can be applied to DPD as well.

DPD is a weakly basic drug with poor aqueous solubility but high intestinal permeability (logP 2.74, pKa 6.24) [2]. Gastric emptying and oral absorption of 30, 50 and 90 mg doses of DPD were simulated using the AGS, which was tuned to absorb drug at a physiological rate [2,27]. A one stage compartment based approach was used to predict the plasma concentration–time profile following oral administration of DPD solution [28]. A compartmental model describing DPD disposition pharmacokinetics (PK) in vivo was developed by fitting the PK model to a published plasma concentration–time profile obtained upon Intravenous (IV) bolus dose administration of 20 mg DPD [27]. Next, the amount of drug absorbed by the AGS was directly input into the central compartment of the PK model to predict the plasma concentration–time profile. The performance of the proposed approach was validated by comparing the predicted plasma concentration–time profile with the in vivo profile [27]. Additionally, the dissolved concentration and fraction of DPD precipitate in the AGS donor were compared with measurements made in human duodenum [2].