Abstract
Material/water equilibrium binding constants (Eb) were determined for 11 organic solutes and 2 plastic materials commonly used in pharmaceutical product containers (plasticized polyvinyl chloride and polyolefin). In general, solute binding by the plasticized polyvinyl chloride material was greater, by nearly an order of magnitude, than the binding by the polyolefin (on an equal weight basis). The utilization of the binding constants to facilitate container compatibility assessments (e.g., drug loss by container binding) for drug-containing products is discussed.
1. Introduction
Plastic materials are widely used in medical items, such as solution containers, transfusion sets, transfer tubing, and devices. The physiochemical nature of these materials provides medical products with their necessary, desirable performance characteristics. While an important performance characteristic of plastics used in medical application is chemical inertness, interactions between a plastic material and a contacted pharmaceutical product are well documented (1–4). Such interactions may include sorption, the uptake of product components by the plastic material, or leaching, the release of plastic material components to the product. Since both sorption and leaching can materially affect product safety and efficacy, it is necessary to establish a particular system's propensity for interaction as part of the product development process.
Investigation of the material/solution interaction under conditions of actual product use represents the most direct and least “contentious” approach for assessing compatibility. However, numerous practical considerations can make such a direct approach difficult to implement in a scientifically and financially responsible manner. Prominent among such considerations are the following:
-
Utilization of the material in a number of configurations and/or with a number of product applications
-
Long contact durations
-
Cost, availability, and/or safety factors associated with the actual product
-
Analytical constraints associated with the actual product matrix
Considering these constraints, efforts have been made to develop strategies to estimate the potential magnitude of the material/solution interaction based on fundamental properties of the materials and the interacting entities. For example, drug binding can be modeled using fundamental properties of the drug and the contacting material. Such a thermodynamic model establishes the maximal equilibrium binding of the drug and thus delineates the potential that the product/material couple will fail from an efficacy perspective. Such a model would be useful in terms of
-
Comparing the performance of two different materials
-
Extrapolating existing data to estimate the behavior of a material versus a new drug entity
-
Evaluating the effect of product configuration changes on drug binding
Solute binding by a material is essentially a partition-mediated process, and the equilibrium distribution of a solute between a solution (i.e., the drug product) and material phase (i.e., the plastic container) can be expressed as the equilibrium interaction constant, Eb. In a model approach, an organic solvent system is defined that mimics the binding behavior of the material. For example, the octanol/water and hexane/water systems have been proposed as surrogates for plastic materials (4–13). If the solute-binding properties of a material are established by measuring its binding of model compounds with known octanol/water partition coefficients (Po/w), then the material's ability to bind drugs can be established if the drug's Po/w (adjusted to formulation conditions such as pH) is known. If the binding properties of two (or more) materials have been established via binding models, such information can be used to establish the relative binding properties of the two (or more) materials.
It is well established in the pharmaceutical literature that polyolefin-based containers exhibit, in general, a decreased propensity to bind drugs than do plasticized polyvinyl chloride (PVC) containers (for example, 14). In this study, we quantify the magnitude of this difference by establishing the thermodynamic nature of the equilibrium partitioning between these materials and various model compounds. Such information allows for the determination of a solute's (i.e., drug's) distribution in container/solution configurations.
2. Materials and Methods
2.1. Plastic Materials
The plastic materials used in this study (plasticized PVC and polyolefin) were obtained as extruded sheeting and are representative of the types of materials used in commercially available products. One potential application of such a container system is drug admixture. In this application, a drug-containing vial product is admixed with an LVP (large-volume parenteral) diluent (in the container system), stored for a short period time (consistent with the documented drug stability and/or microbial concerns), and then administered. In such a situation, drug binding to the container, should it occur, could be influenced by the fact that the container has been subjected to the LVP's steam sterilization process, which could modify the container's physiochemical properties somewhat. Thus, prior to the initiation of the binding study, the materials examined were subjected to steam sterilization conditions similar to those experienced by a LVP product during routine manufacturing.
2.2. Interaction Conditions
The binding assessment involved generating a binding model for each of the test materials. The test materials were contacted with aqueous donor solutions containing known amounts of model organic compounds. These model organic compounds were ones whose theoretical partitioning behavior, as expressed by classical octanol/water and hexane/water partition coefficients, are well known and which thus serve as analytically expedient surrogates for components of pharmaceutical formulations. The container materials and donor solutions were equilibrated, and the equilibrium concentration of each model compound in the donor solution was analytically measured. Differences between the initially prepared and final equilibrium concentrations of the models in the donor solution were utilized to calculate binding constants.
The donor experiment was performed at a temperature of approximately 40 °C for a period of 21 days. To establish whether equilibrium was obtained under such conditions, the donor solutions were assayed after 7, 14, and 21 days of contact. A specified amount of material was contacted with 75 mL of the donor solutions containing approximately 10 mg/L of the model compounds. Donor solutions were prepared to contain either groups of the model compounds or individual compounds (in order to assess the potential for competitive solute binding). Compounds with similar binding characteristics were grouped together so that the magnitude of compound loss due to binding was optimized. In the case of multiple compound donor solutions, triplicate test articles were prepared for each material/donor solution couple. In the case of single-compound donor solutions, a single test article was prepared for each material/compound couple. The donor solutions were acidified to ensure that the ionizable solutes were in their protonated (neutral) forms. The suite of model compounds used is consistent with the suites used in previous, similar investigations (8–13) and is summarized in Table I. The individual model compounds, as well as all analytical reagents, were obtained commercially (e.g., Aldrich Chemical, Milwaukee, WI) as substances of known, high purity.
Partition Coefficients and Binding Constants (Log Eb) Calculated from the Analytical Results
2.3. Analytical Method
The analytical method employed for donor solution analysis was high-performance liquid chromatography (HPLC) with UV detection. Operating conditions for the HPLC method are summarized in Table II. The HPLC analyses were performed with Agilent (Wilmington, DE) model 1100 chromatography systems.
Operating Parameters, HPLC/UV Analysis
Standards for the target analytes were prepared at two levels, approximately equal to the initial analyte levels in the donor solution and a factor of ten lower than this (≈10 and ≈1 mg/L, respectively). Analyte concentrations were determined by putting the sample's peak area into a linear calibration curve (standard area versus standard concentration).
2.4. Mathematical Analysis, Equilibrium Interaction Constant, Eb
The equilibrium interaction constant, Eb, is essentially a partition coefficient describing the equilibrium distribution of a solute between a material phase (p) and a solution phase (s):
In a binding study as performed herein, Cs is measured as the equilibrium concentration of the analyte remaining in the donor solution. While Cp cannot be measured directly, it can be calculated from the experimental parameters and Ci, the initial analyte concentration in the donor solution as follows. The amount of the solute bound by the material (Am) is calculated from Ci, Cs, and the donor solution volume (Vs) as
The concentration of solute in the material phase (Cp) is Am divided by the weight of material use (Wm):
While having both Cp and Cs should allow for a direct calculation of Eb, there is the issue of compatible concentration units. Specifically, Cp is a mass concentration—for example, mg/kg (ppm)—while Cs is a solution concentration. If the density (D) of the solution can be measured (or assumed to be 1 g/mL), this difference can be reconciled.
The overall equation for calculating Eb is as follows, illustrating units and conversions.
As an example calculation, consider actual data obtained for solute DMP and the polyolefin material:
Ci = 10.300 mg/L (preparation concentration of the donor solution)
Cs = 6.815 mg/L (mean of replicate binding samples)
Wm = 20.011 g (mean of replicate binding samples) = 0.020011 kg
Vs = 75 mL = 0.075 L
D = 1 g/mL
3. Results and Discussion
3.1. Discussion of the Experimental Conditions
As the model compounds span nearly four orders of magnitude in terms of their partition coefficients, it is reasonable to anticipate that no single contact condition is most appropriate for all the model compounds. In general, the most accurate estimation of Eb is obtained if the compound's concentration in the equilibrated donor is between 20% and 80% of the compound's concentration in the initial donor solution. In order to accomplish this objective in this study, the compounds were grouped in terms of similar partition coefficients and the material weights used in the binding samples were set accordingly. Thus, for example, the material weight in the binding samples for the polyolefin material was approximately 25 grams for the poorly binding compounds in group 1 (Table I) and 6.0 grams for the highly binding compounds in group 3.
Two significant issues associated with a binding assessment are the achievement of equilibrium and the potential for competition between members of a compound group. Because partitioning is a thermodynamic process, one must be sure that equilibrium is achieved in the binding assessment. Analysis of the donor solution composition over time allows one to assess whether equilibrium has been achieved over the course of the experiment. As shown in Tables III and IV, little if any change occurs in the composition of the donor solutions between the second and third week of contact and one therefore concludes that equilibrium has been achieved.
Equilibration Concentrations of the Model Compounds, Plasticized PVC Material
Equilibration Concentrations of the Model Compounds, Polyolefin Material
The second issue addresses whether the binding characteristics of a single model compound are influenced by the presence of two (or more) other compounds in the donor solution. This question is relevant to this study because the model compounds were grouped into mixtures (due to the practical reason that so doing results in fewer samples for analysis), and to “real world” situations wherein a pharmaceutical product may contain more than one ingredient that would be prone to binding by a container. This issue was addressed in this study by comparing the binding characteristics of all of the model compounds determined individually and in their respective analytical groups. As shown in Tables III and IV, the concentration of the model compounds in the donor solutions after 3 weeks of contact with the test materials was generally the same (within analytical imprecision) whether the compounds were individually or collectively present in the donor solutions. Thus little or no competition between model compounds was observed in this study. However, as the difference in compound concentration, single versus groups, for two of the analytes (EAB and BAB) was somewhat larger than the difference observed for the other model compounds, the single compound binding constants for these compounds were used in the subsequent determination of the binding models.
3.2. Comparison of the Binding Properties of the Materials Studied
The measured log Eb values are summarized in Table I. Figure 1 is a comparison plot for the materials examined in this study. The interpretation of such a diagram is straightforward. Two materials that behave exactly alike will have a line with an intercept of 0 and a slope of 1. A non-zero intercept suggests that one material has a larger “intrinsic binding capacity” than does the other material. A non-unit slope suggests that one material is more “sensitive”, in terms of its compound binding, to the chemical nature of the compound than is the second material.
Comparative binding characteristics, plasticized PVC versus polyolefin. Two materials that have equivalent binding properties will have a relationship that is described by the line of equivalence (intercept = 0, slope = 1). The plasticized PVC material has, in general, the same “sensitivity” to solute properties as does the polyolefin material (near unit slope) but has a higher “intrinsic binding capacity” than does the polyolefin (positive intercept). This means that on an equal weight basis, drug binding will be less for the polyolefin material than for the plasticized PVC material.
As shown in Figure 1, the two materials (plasticized PVC and polyolefin) have, in general, a similar “sensitivity” in terms of how they respond, from a binding perspective, to changes in the chemical nature of the bound compound (as indicated by the nearly unit slope of the comparison line in Figure 1). However, the intercept of the comparison line (+0.92) suggests that the intrinsic binding capacity of the plasticized PVC is approximately 8 times greater (100.92 = 8.3) than that of the polyolefin. More conservatively, the minimum difference between the intrinsic binding capacities of the two materials can be assessed as the smallest difference in measured Log Eb for the model compounds. From Table I, the model compound that exhibits the minimal difference is EBA, which has a ratio of binding constants (plasticized PVC versus polyolefin) of 3.6.
3.3. Utilization of the Binding Models
As an example of the utility of the binding models, consider the following situation. A drug-containing formulation (250 mL) is stored in an appropriately sized plastic container. Due to several considerations, a container made from plasticized PVC will weigh more than an equivalently sized container made from a polyolefin (for the purpose of this example, the 250-mL container would weigh approximately 20 g for the PVC and 12 g for the polyolefin). The drug has a log Po/w of 1.6 and a log Ph/w of 0.8 (similar to the model solute DMP). The container can be made from either a plasticized PVC or polyolefin. The relevant questions are
-
What is the magnitude of the difference in drug binding, plasticized PVC versus polyolefin material?
-
Does either material exhibit an acceptable level of drug binding (defined here as less than 10% drug loss)?
The relationship between the fractional binding (Fb) and a solute's Eb takes the form:
where Wc = the container weight (in kg), Vs is the solution volume (in L), and D is the solution's density (in kg/L). The Eb values are obtained from Table I: Eb = 45.19 and 1.92 for the plasticized PVC and polyolefin materials, respectively. Substituting these values into the Fb equation (and assuming a solution density of 1 kg/L) produces Fb,polyolefin = 0.084, Fb,PVC = 0.783. Thus the polyolefin is a suitable container material for this application, while drug loss to a plasticized PVC container would be excessive.
4. Conclusions
The relative compound binding properties of two materials used as drug product storage containers, plasticized PVC and polyolefin, has been quantitatively established in this study. In general, drug binding is greater, on a proportional weight basis, by 4 to 8 times for the plasticized PVC versus the polyolefin material. This difference is exacerbated in the utilization of these materials by the practical observation that for a given container volume, polyolefin containers generally weigh significantly less than plasticized PVC containers. The combination of these two factors leads to the following generalizations related to the relative drug binding properties of these two materials:
-
Drugs that are compatible (from a binding perspective) with a plasticized PVC container will be compatible (from a binding perspective) with a polyolefin container.
-
Drugs that are not compatible (from a binding perspective) with a plasticized PVC container may be compatible (from a binding perspective) with a polyolefin container.
-
If the fractional drug binding in a PVC container is known, the fractional drug binding in a polyolefin container can be estimated (and visa versa).
Although it is the conclusion of this study that drug binding may be reduced, polyolefin versus plasticized PVC containers, it is noted that many drug formulations and admixtures are compatible with plasticized PVC containers. This is the case as drug solubility requirements for aqueous drug products dictate that a drug's effective partition coefficients (log Po/w and log Ph/w) are typically lower than the values used in the numerical example provided herein.
- © PDA, Inc. 2010
