Discovery-stage PK analysis assists in the design and selection of optimal candidates that can be advanced to become therapeutics. At this stage in development, prior to regulatory TK studies, there are numerous operational limitations, including a lack of resources, reagents and time. Despite these limitations, it is critical to use scientific justification and data value-driven approaches to define the scope of assay performance and analytical qualification Citation[1]. These approaches must be transformed into efficient processes as well as operationally feasible and reliable assays. While not in the scope of this article, it should be noted that the data generated in the discovery setting are increasingly used to support applications for investigational new drugs. Therefore a risk-based approach should also ensure the quality of the assay and data.
Know what you are measuring
A fundamental element in discovery PK is the need to understand and be able to communicate to project teams what the PK assay was designed to measure. Teams are often comparing multiple candidates and asking questions regarding the PK concentration time relationship (Cmax, AUC and terminal half-life). Unexpected profiles for the modality in question, such as a long half-life for a fusion protein or short half-life for an antibody, will always raise questions regarding the analytical method; will the assay detect metabolites of a fusion protein? Will high concentrations of target interfere in the assay or does it measure total drug? How sensitive is a specific antibody–drug conjugate assay to drug-to-antibody ratio? Answers to these questions start with the reagents used and the format selected Citation[2,3]. Distinguishing between anti-drug antibodies (ADA), soluble target interference, matrix interference and target-mediated disposition can be particularly challenging.
ADA is common in multidose efficacy and exploratory toxicology studies and can interfere in the PK assay measurement and/or change the half-life of the drug. A conventional bridging ADA assay can be developed if an abnormal PK profile is observed, to help confirm that ADA is responsible. There are also alternative approaches to the use of these assays that may better suit the needs of discovery Citation[4]. The additional effort to assess drug tolerance and/or to use pretreatments such as acid dissociation is generally not used at this stage. Thus, samples earlier in a PK profile with higher levels of drug may read as ADA negative but ADA may be still be present.
In programs where shed or soluble targets are present, and when total target levels may increase as a result of drug binding, a minimal understanding of the risk of target interference in the PK assay is important. If the PK assay was designed with the intent to measure a certain form of the analyte (i.e., free or total drug), then experimental data would need to be generated to demonstrate this ability Citation[2]. The effort to measure target directly will increase the understanding of the assay properties, reduce the overall risk of erroneous reporting and inform the PK/PD understanding for the project.
Matrix interference is generally addressed through the use of minimum required dilutions in an appropriate assay buffer using pooled control matrix. However, control matrices may not perfectly represent incurred samples. Individually spiked matrix samples can serve as an alternative and more rigorous approach to avoiding matrix interference. This approach should be considered when testing models that produce serum samples with known interference, such as lipemia or autoantibodies. Incurred sample dilution linearity, when scientifically justified, confirms what you are measuring and is the most definitive approach to address abnormal PK, which leads to questions regarding the immunological similarity of the analyte to the reference standard. Lack of parallelism between the diluted sample and reference standard curve suggests immunological detectable changes in the analyte post-dosing, which may be due to metabolism. Lack of dilution linearity for initial dilutions followed by proportional dilution in subsequent dilutions indicates a matrix interference issue and not a lack of immunological similarity.
Understand analytical error related to goal
To help decide what constitutes a ‘fit-for-purpose’ assay, one parameter that should be assessed is analytical error. Regulatory TK studies require guidance-defined accuracy and precision to assign appropriate safety margins related to exposure. In the discovery phase, the goals are very different and what constitutes accepted analytical error may also differ. One scientific approach to establishing minimum requirements for acceptable precision and accuracy is to conduct a statistical power calculation Citation[5]. The power calculation will determine if the study design and the associated analytical and biological error can meet the intended purpose. In the case of PK analysis, the goal may be to differentiate two therapeutics on the basis of clearance.
Total error can be computed from the inter-assay precision and the relative bias Citation[6,7] and used as the error parameter in the power calculation. For example, using a simulation to determine the minimum difference that could be detected using a study design of three mice per treatment group with a p-value of 0.05 and a power of 0.8, it can be shown that a method with a 10% error could detect a 20% difference with a 0.8 power in a sample size of three mice. A method with 40% error could detect a 60% difference between groups with 0.8 power. If the goal was to look for large differences in clearance (>60%), then a method with a total error of 40% would be sufficient; however, if smaller differences were expected, then the study design would need to be altered or the assay revamped. The power calculator provides a rational approach to determining method acceptance in the discovery setting.
Utility of generic assays
As has been described above, one parameter of therapeutic selection may be to compare the PK of several very similar candidates. In approaching this request, it is important to understand the molecules that are being quantitated and the desired outcome of the study. If the molecules differ only by a few amino acids, it is possible to develop one assay using only one of the candidates as a standard, and interpolate the other candidate samples against that reference candidate. This is only possible if the changes in the candidates do not affect the binding of the reagents used in the assay, hence the need to understand the molecules being measured. With this in mind, we favor the use of polyclonal antibodies to minimize the effect of amino acid differences and increase the probability of assessing all the candidates with one assay. Assay development and qualification for the one candidate (surrogate standard) would be executed as usual. For the other candidates, at a minimum, assessing the accuracy and precision of the measurement of the candidates when interpolating off the surrogate standard curve is important. However, more informative is an assessment of immunological similarity (parallelism) of the candidates in comparison with the surrogate standard. A lack of linearity, as previously described, may indicate a lack of immunological similarity, and a separate assay for the candidate would need to be developed. The threshold for developing a new assay as well as the bias and CV expectations should be defined by the desired outcome as described above. Studies designed to aid in candidate selection are not usually powered to detect minor changes in PK parameters, since PK is only one consideration in candidate selection. In cases where the number of constructs is large, a screen approach may be taken, where a small subset of samples will be analyzed as opposed to the entire PK profile, thereby limiting the study’s ability to detect only gross differences in PK.
Sample integrity & risk-based stability assessment
Sample integrity is the foundation for reliable measurements; however, there are limited data collected prior to sample analysis and minimal attention is paid to long-term storage, because samples are often analyzed soon after collection (<2 weeks). For monoclonal antibodies, serum is often the preferred matrix. For other therapeutics, consideration should be given to what matrix should be collected based on knowledge of the analyte. For other more complicated or potentially labile modalities, sample collection procedures should be evaluated prior to collection to avoid artifacts Citation[8]. Analysis of therapeutics in the tissue is important to understand exposure at the site of action. Tissue analysis requires additional attention to avoid proteolytic degradation of the therapeutic during sample processing as well as storage Citation[9].
Utilize in-study performance
Prestudy assay qualification is an essential part of defining that a method meets its intended purpose; however, these experiments are the best approximation of the actual performance. It is very difficult to simulate all of the possible issues that arise during sample analysis. As a result, in-study validation or analytical monitoring provides a more realistic estimate of the analytical error and true performance. In addition to analytical performance estimates, in-study monitoring may uncover temporal effects related to sample loading and incubation, plate edge effects and dilutional preparation bias.
In order to monitor these potential effects, QC samples must be included with every run and appropriately distributed across the plate. A large number of QCs can be used when there is uncertainty at the top and bottom of the curve, but a minimum of two QC levels are required, one at close to the high end of the range and the other close to the low end of the range, to better monitor assay drift over time. Ideally, QC samples should also be stored in the same matrix as samples, for the same length of time, and processed in an identical manner as the study samples. Dilutional controls cannot guarantee that all samples are diluted properly; however, they can provide insight into universal differences in dilution processing that may occur across operators or with large multistep dilutions Citation[10]. Monitoring of QC sample performance should be done daily for run acceptance rejection, and periodically throughout the lifecycle of the analysis, to address the concerns described above.
For generic assays that will be used routinely, it is important to conduct more rigorous prospective evaluation of the performance characteristics of the assay. Prospective monitoring will alert the scientist to changes in the assay and ultimately ensure that the assay meets the intended purpose. In order to assess the performance, the QC samples are plotted over time and can be evaluated statistically Citation[11,12]. Once those characteristics are defined, the scientist must decide if the method meets the intended purpose or if additional development is needed. In-study performance can be monitored the same way.
Discovery PK analysis requires a flexible approach to developing assays that includes appropriate use of prestudy and in-study assessments, of performance and reliability, to ensure that the PK data will meet the intended use. Given the resource and time constraints, bioanalytical scientists need to use sound scientific judgment to balance these constraints, and may need to deploy a wide range of tools for some assays and utilize a minimalist approach for others. In all cases, it is important to clearly communicate what the assay measures and its strengths and limitations.
Financial & competing interests disclosure
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
No writing assistance was utilized in the production of this manuscript.
References
- Hartmann C , Smeyers-VerbekeJ, MassartDL, McDowallRD. Validation of bioanalytical chromatographic methods. J. Pharm. Biomed. Anal. 17(2), 193–218 (1998).
- Lee JW , KelleyM, KingLEet al. Bioanalytical approaches to quantify ‘total’ and ‘free’ therapeutic antibodies and their targets: technical challenges and PK/PD applications over the course of drug development. AAPS J. 13(1), 99–110 (2011).
- Kaur S , XuK, SaadOM, DereRC, Carrasco-TrigueroM. Bioanalytical assay strategies for the development of antibody-drug conjugate biotherapeutics. Bioanalysis5(2), 201–226 (2013).
- Clark T , KingLE. Discovery biotherapeutics bioanalysis; challenges and possible solutions. Bioanalysis4(9), 1003–1007 (2012).
- Boulanger B , DeweW, GilbertA, GovaertsB, Maumy-BertrandM. Risk management for analytical methods based on the total error concept: conciliating the objectives of the pre-study and in-study validation phases. Chemometr. Intell. Lab. Syst. 86, 198–207 (2007).
- Findlay JWA , SmithWC, LeeJWet al. Validation of immunoassays for bioanalysis: a pharmaceutical industry perspective. J. Pharm. Biomed. Anal. 21(6), 1249–1273 (2000).
- DeSilva B , SmithW, WeinerRet al. Recommendations for the bioanalytical method validation of ligand-binding assays to support pharmacokinetic assessments of macromolecules. Pharm. Res. 20(11), 1885–1900 (2003).
- Ray CA , DumaualC, WilleyMet al. Optimization of analytical and pre-analytical variables associated with an ex vivo cytokine secretion assay. J. Pharm. Biomed. Anal. 41(1), 189–195 (2006).
- Neubert H , FountainS, KingLet al. Tissue bioanalysis of biotherapeutics and drug targets to support PK/PD – a perspective. Bioanalysis4(21), 2589–2604 (2012).
- Pandya K , RayCA, BrunnerL, WangJ, LeeJL, DeSilvaB. Strategies to minimize variability and bias associated with manual pipetting in ligand binding assays to assure data quality of protein therapeutic quantification. J. Pharm. Biomed. Anal. 53, 623–630 (2010).
- Westgard JO , GrothT, AronssonT, FalkH, VerdierCH. Performance characteristics of rules for internal quality control: probabilities for false rejection and error detection. Clin. Chem. 23, 1857–1887 (1977).
- Viswanathan CT , BansalS, BoothBet al. Quantitative bioanalytical methods validation and implementation: best practices for chromatographic and ligand binding assays. Pharm. Res. 24(10), 1962–1973 (2007).