Predicting in vivo behavior of injectable, in situ-forming drug-delivery systems

Most desirable drug administration regimens strive for a constant plasma drug concentration. However, conventional routes of administration often result in rapid increases of plasma drug concentrations that are only temporarily within the therapeutic window and require frequent administration to maintain effective local concentrations. Many therapeutics such as chemotherapeutic agents or antibiotics suffer from low water solubility, stability and dose-limiting toxicities which can make it more difficult to achieve local therapeutic concentrations. To overcome such challenges, countless innovative dosage forms from polymeric or lipid nano- and micro-particle formulations (e.g., the quintessential liposome), have been designed over the past several decades to package and deliver drugs intravenously. However, even with most sophisticated modifications, these systems are susceptible to the unrelenting transport issues and numerous barriers associated with intravenous administrations, often unpredictable side effects such as immune responses, severe off-target accumulation and toxicity.

Controlled release injectable formulations for sustained delivery of therapeutics are an attractive alternative to oral and intravenous dosing as well as the conventional preformed drug eluting depots. Compositions involving local injection of drug–polymer solutions or suspensions, drug-loaded microparticles, oil-based drug suspensions or aqueous concentrated proteins have been clinically utilized for treatment of a broad range of diseases from alcohol dependence to schizophrenia and HIV [1–3]. In most of these applications, the formulations are administered as an intramuscular or subcutaneous injection with the overall goal of consistent drug dosing at the target concentration achieved by long-term sustained drug release. The release is more gradual and extended for periods from days to months, which may mitigate some toxicities associated with intravenous dosage forms. The formulations also improve patient compliance by reducing the number of individual injections in nonsustained release formulations.

In situ-forming implants for long-term drug release

In situ-forming implants (ISFIs) [4] are a subclass of the injectable parenteral formulations. These liquid solutions of biodegradable polymers mixed with bioactive therapeutic agents can be injected directly into a target tissue and solidify into a solid drug-eluting depot. The solid implant formation process can be initiated as a response to a stimulus such as a solvent exchange, temperature or pH. The straightforward, easily personalized formulation of ISFI systems, conceptualized over 25 years ago by Dunn et al. [5], makes them an attractive alternative to other, more complex injectable formulations. An iteration of a solvent-exchange based system has been applied clinically for the treatment of advanced prostate cancer (Eligard^®, TOLMAR Pharmaceuticals) [6,7]. Phase inversion occurs with the precipitation of a hydrophobic polymer dissolved in a water-miscible solvent upon injection into an aqueous environment [8]. The addition of poly(lactic-co-glycolic acid) (PLGA) to a solvent, such as 1,2-N-methyl pyrrolidone (NMP, LD50 4000 mg/kg orally in rats), creates a low viscosity, low toxicity liquid polymer that can then be mixed with any agent (even fragile proteins since the fabrication requires no heat) and injected into the desired site with a hypodermic needle. Eligard delivers an active agent for up to 6 months and has been shown to reduce testosterone levels in up to 98% of patients [6,7]. A thermosensitive formulation, ReGel consists of triblock copolymers consisting of PLGA and PEG dissolved in water. Upon an increase in temperature over the polymer's lower critical solution temperature, the free flowing solution spontaneously gels. Oncogel (ReGel/Paclitaxel) (Macromed Inc. USA), which has been developed to deliver a chemotherapeutic agent locally to solid tumors, reached Phase II clinical trials in 2007.

Factors that lead to poor correlations of in vitro & in vivo outcomes of ISFIs

A significant limitation of all injectable parenteral systems is a lack of appropriate in vitro test beds that can streamline research and development and enable future applications of injectable technologies. The in vivo environment is highly complex and difficult to replicate. It is also difficult to decouple contributions from the various components that play an active role in the process. In many cases, the typical dissolution set-up does not take into account the processes governing the journey of the active agents immediately after their release from the parenteral formulation. These vary drastically from that of drugs administered orally or intravenously. In injectable formulations, drug is released into the interstitial space and has to diffuse to eventually reach the vasculature (either a capillary or lymphatic venuole). Although the fundamental composition of the interstitium is comparable between different sites, there are also drastic differences in extent of vascularization, morphology of the collagen network, protein type and presence of ions, to name a few. Foreign body response, enzymatic cleavage, pH and local byproducts are also considered to play a role. All of these can make predicting the performance of the formulation in vivo challenging and can eventually lead to interpatient variability and therapeutic inconsistencies [3,9].

Additional research has shown that implants made with PLGA and NMP and injected subcutaneously in vivo had increased burst release and rate of phase inversion when compared with those formed in vitro in a phosphate-buffered saline (PBS) bath [10]. Implant reproducibility, which includes ultimate implant shape, size and microstructure have all been shown to be partially responsible for these differences [10]. Implant shape and size can deviate depending on rate of injection, type of needle used or formulation [11]. Implant microstructure on the other hand has been found to depend on the rate of phase inversion [12,13]. More importantly, current in vitro dissolution studies do not accurately mimic the physical or chemical properties of tissues. The chemical composition of intramuscular or subcutaneous injection sites can be modeled as a mixture of water, organic solvent, proteins and salts, which is significantly different than the solvents used in in vitro dissolution studies. While only a few studies have been done to investigate the effect of bath-side composition, none were found to account for differences of implant behavior between in vitro and in vivo [12,14].

Another rarely examined factor is the mechanical property of the milieu surrounding the implant, resulting in poor correlations between in vitro and in vivo implant behavior [15]. While primary applications of ISFIs have been focused on subcutaneous and intramuscular injections, many new approaches focus more on site-specific delivery into specific tissues such as tumors, brain, eye, bone and cartilage [16,17], and periodontal bone defects [18]. The formation and subsequent morphology (shape, porosity, density) of ISFIs is intimately related to the mechanical properties of the injection site. This, in turn, has broad implications on the ultimate rate of burst and overall drug release as well as degradation rates [10,19], and therefore there is a strong need for an accurate in vitro predictor of in vivo implant microstructure and drug release kinetics. Our group has investigated the effect of injection site stiffness on implant formation and release [19]. When injected in vivo, implants were flatter with a less uniform shape as compared with the spherical implants that were formed when dropped in a PBS bath in vitro. More interestingly, they found that the deviation of in vitro burst release from in vivo correlated with implant polymer molecular weight. The deviations in behavior are a function of implant swelling within a constrained environment, resulting in a mechanical reaction force on the depot. This force causes a reduction in implant volume due to an imbalance in the solvent/bath-side exchange that typically leads to swelling, but because the implant is constrained here, it can no longer expand. Subsequently, a reaction force is generated and leads to the rapid release of drug and solvent due to mechanically induced convection within the implant [20]. Release is modulated by the osmotic activity of the implant (swelling) and the stiffness of the injection site [20].

Need for better models

Appropriate in vitro models that can accurately recapitulate the in vivo environment are an essential step in facilitating future clinical translation of injectable delivery systems. This is particularly important in ISFIs, where the duration of release can be on the order of months. Such long-term animal studies could be difficult to complete and of high cost. A better predictor of in vivo implant morphology can also lead to faster translation of ISFIs with applications in tissue engineering where porosity and interconnectivity are vital to cell infiltration or delivery. The typical dissolution set-up does not take into account most of the local factors [15]. Seeing the need for standardization of in vitro release methods, several agencies including the US FDA and United States Pharmacopeia (USP) have collaborated and published guidelines for improving in vitro–in vivo correlations for injectable dosage forms. However, these recommendations focus primarily on evaporation prevention, pH, osmolality and microbial contamination [15], and failed to standardize a release protocol. Nonetheless, despite concerted efforts, there are still, to date, no regulatory standards for parenteral injectable delivery systems (including microsphere, suspension and in situ-forming implants) [1,2]. The in vivo dissolution setup sanctioned by the USP was designed for oral not parenteral formulations. There are now several USP systems available (1: basket, 2: paddle, 3: reciprocating cylinder, 4: flow-through cell, 5: paddle over disc, 6: cylinder and 7: reciprocating holder), but none are currently able to reproduce accurately the in vivo parameters affecting drug release, distribution and eventual absorption of agents released from injectable or implantable dosage forms [1,15].

Biorelevant phantoms for improved in vitro prediction of in vivo behavior

The dissolution method most commonly used for polymeric implants and microparticles is the ‘sample and separate’ method, where the formulation is introduced into a vial containing a release media (most commonly PBS, pH 7.4). The vial is kept at physiological temperature while under constant agitation. At predetermined time intervals, aliquots are taken from the release media (replaced with fresh media) and analyzed for release of drug. The constant agitation introduces convection and ensures immediate homogenization of drug and eroded polymer in the release media, maintaining high concentration gradients. Although this standard method ensures ‘sink conditions’ when properly designed, it misrepresents the mass transport phenomena in living tissues, where natural barriers such as connective tissue can locally violate sink conditions near the implant. An alternative approach is to simulate the tissue using a hydrogel phantom. Hydrogels, which are cross-linked hydrophilic polymers with high water content, can be tuned to have biologically relevant mechanical properties, and have been investigated as in vitro tissue mimicking phantoms [21,22]. PLGA ISFIs injected directly in to polyacrylamide phantoms have been shown to better mimic in vivo implant drug release, shape and microstructure than the standard sample and separate dissolution method [22]. These studies have shown the importance of injection site stiffness as a major factor in implant performance, which in the past has been otherwise ignored. Most importantly, while it is impossible to mimic all of the conditions found in vivo, with proper design these biorelevant phantoms can be used to study specific in vivo aspects in a high-throughput, controlled and reproducible manner.

In vivo characterization with medical imaging

The use of homogenous phantoms allows for high contrast and limited artifacts under fluorescence, ultrasound and MRI imaging. In addition to better approximating the local tissue environment, tissue-mimicking hydrogel phantoms can also provide a high-throughput method to characterize multiple parameters, such as implant solidification, drug penetration distance and erosion through the use of different imaging tools. Our group has also developed nondestructive techniques for monitoring both ISFI morphology, phase inversion and implant erosion using diagnostic ultrasound [23] and ultrasound elastography [24], respectively. Erosion of implants formed inside of similar acrylamide phantoms was found to follow similar trends to those implanted in vivo [24]. Additionally, the use of a hydrogel system along with fluorescence imaging, enables the evaluation of spatial drug distributions. As these implants are increasingly being used for treatment of local diseases, distribution studies would be beneficial in optimizing formulations to increase in vivo treatment area. Many other quantitative imaging techniques for robust nondestructive analysis of biomaterial performance in a number of implantable and injectable applications are currently under development [25–28].

Future perspective

Numerous features present at the site of injection of in situ-forming, extended release dosage forms are not easily recapitulated via available in vivo testing approaches. Predictive in vitro models are vital to reducing the cost and time associated with the progression to the clinical use of drug-delivery systems. Yet, even with modification of the simple standard USP dissolution setups, physical morphology of the tissue, presence of various proteins, salts, variations in pH and their potential heterogeneous distribution, particularly when examining approaches for direct intratumoral delivery, can significantly influence the outcomes [29]. Another parameter which is rarely examined is the mechanical property of the injection environment, specifically density and modulus. Additional complications are introduced when the activity of the released drug is meant to be local not systemic. This is seen in formulations delivering active agents directly to tumors, into the eye or into bone or cartilage. Because their activity is not meant to be systemic, plasma drug concentrations provide limited information as to local implant performance. In turn, local tissue sampling is difficult if not impossible to carry out in humans. As such, animal models become crucial, but development can be additionally facilitated by use of appropriate in vitro models capable of predicting local drug release and distribution. To this end, hydrogel-based phantoms may be a suitable setup that can be used to conduct both drug release and distribution experiments with in situ-forming implants and other injectable formulations. If properly designed, phantoms capable of recapitulating the injection site environment could replace the use of small animals for distribution studies and more accurately predict treatment volumes in humans. Likewise, analysis techniques applied to both in vitro and in vivo models will be essential. Because much more meaningful information can be gained from nondestructive direct analysis of the implant at the site of interest, imaging methods can play a powerful role in future development of parenteral formulations. Finally, the issues of valid, straightforward predictive in vitro models are not unique to drug delivery applications. For example, the issues are similar for biomaterials-based solutions in regenerative medicine. A recent meta-analysis found a poor correlation between the ability of several in vitro tests to predict the in vivo outcomes of these materials. A covariance score of 58% was reported in this study, and the authors concluded that “there was no significant overall correlation between in vitro and in vivo outcome” [30]. This clearly highlights the need for improved in vitro testing platforms and nondestructive analysis techniques that are able to provide accurate in vivo information in a broad range of biomaterials applications.