Lessons learned from regulatory submissions involving endogenous therapeutic analyte bioanalysis

Background & challenges

Endogenous substances are compounds that are naturally present in the body because either the body produces them or they are present in the normal diet [1]. Endogenous substances have been utilized as biomarkers to study physiological phenomena or they have been developed as drugs or biologics for therapeutic use (i.e., endogenous therapeutic substances). Though called endogenous therapeutic substances, these medications are exogenous, made by purification and concentration of animal extracts, synthetic or semisynthetic processes, or recombinant DNA technologies. They are administered to mimic or enhance the effects of endogenous substances already present in the body or diet, thereby supplementing or regulating the body’s natural processes. Examples of exogenously administered endogenous therapeutic substances include hormones, neurotransmitters, vitamins, fatty acids, inorganic elements and others (Table 1). By targeting specific biological pathways, these medications can help compensate for deficiencies or modulate cellular functions. The first discovered endogenous therapeutic substance is insulin, and 2023 marks the 100th anniversary of the discovery of insulin and its being commercially available as a treatment for diabetes [2]. Insulin, an endogenous hormone, is naturally synthesized and secreted by the pancreas. Diabetic patients, who may have insufficient insulin production or reduced sensitivity to insulin, can take insulin as one treatment option to regulate blood sugar levels. In the last century, there have been great innovations in insulin therapy, from crude extracts of animal pancreas to recombinant human insulin and insulin analogs available in rapid-acting, short-acting and long-acting forms to cater to the specific needs of people living with diabetes. Certain fatty acids, such as omega-3 fatty acids, are essential fats in the body for various physiological functions, but the body cannot produce them on its own and therefore they must be obtained from the diet. The two most important omega-3 fatty acids are eicosapentaenoic acid and docosahexaenoic acid [3]. They are found primarily in fatty fish, nuts, seeds and plant oils. The US FDA has approved several omega-3 fatty acids as an adjunctive therapy for reducing the risk of cardiovascular events in adults with elevated triglyceride levels. Endogenous substances include some biomarkers, and biomarker bioanalysis is one of the areas with substantial interest and growth recently. Biomarker bioanalysis is a broad topic with many varied modalities, and additional considerations for these new modalities are warranted. Strategic approaches for biomarker bioanalysis should be evaluated and selected based on the principles presented in this article. The regulatory perspective of biomarker bioanalysis has also been shared previously elsewhere [4,5] and will not be discussed in detail here.

For these endogenous therapeutic substances, there have been some unique challenges for bioavailability (BA) and bioequivalence (BE) assessments [6]. From the study design/data analysis perspective, both endogenous substances and administered exogenous therapeutic analytes are identical, and both contribute to the total circulating therapeutic analyte concentration. If unaccounted for, the presence of endogenous therapeutic analytes may bias BA/BE assessments. In addition, endogenous analyte concentrations may not be constant and can be impacted by circadian rhythm, dietary intake, or homeostasis. Further, there may be significant interindividual variability of these endogenous analyte concentrations influenced by factors such as genetics, age, gender and disease state. Study design with dietary restriction, enrolling participants deficient in the endogenous substance that is being investigated, data analysis with baseline correction, or other approaches can be used to address these challenges and support BA/BE assessment of endogenous therapeutic substances [7].

From the bioanalysis perspective, endogenous therapeutic analytes cannot be distinguished from administered exogenous therapeutic substances, adding complexity for bioanalytical method development and sample analysis. For example, the lack of endogenous analyte-free blank matrix to prepare calibration standards (CSs) is a common challenge. The recently published ICH “M10 Guidance for Industry: Bioanalytical Method Validation and Sample Analysis” [8] outlines some approaches (e.g., background subtraction, standard addition, using surrogate matrix and using surrogate analyte) that can be utilized to measure the concentration of endogenous therapeutic analytes in study samples when matrices without interference are not available.

In this article we present case examples from clinical studies involving endogenous therapeutic substances in regulatory submissions and highlight the common challenges encountered and lessons learned. Additionally, we provide practical tips and strategies to consider in endogenous therapeutic analyate bioanalysis from a regulatory perspective.

Case examples & lessons learned

As mentioned above, the accuracy of concentration measurements following the administration of a therapeutic substance poses a challenge if the therapeutic analyte and its endogenous counterpart cannot be distinguished. The impact of this becomes more significant when the primary efficacy endpoint is defined by the therapeutic analyte concentration in the presence of its endogenous counterpart; for example, testosterone (T) that is used in testosterone replacement therapy (TRT). Below are some case examples highlighting the challenges and importance of conducting the bioanalysis of endogenous therapeutic analytes adequately.

Example 1

A relative BA study was conducted for bridging the efficacy and safety of a TRT product under development to the listed drug [9]. While blank plasma obtained from females was used to prepare CSs and quality controls (QCs), the T concentrations of CSs and QCs were not adjusted to account for the endogenous T concentration in the blank matrix used to prepare them. Therefore, the accuracy of the T concentrations in the incurred samples from study subjects could not be assured. The T concentrations for the CSs, QCs and incurred study samples had to be recalculated by adding the endogenous T concentration that was derived by employing the standard addition approach [10] that is discussed in detail later, in the Discussion section.

Example 2

The CSs in a phase III efficacy and safety clinical study were prepared in artificial matrix constituted with 4% bovine serum albumin (BSA) in 0.9% saline. This artificial matrix was used instead of serum with low endogenous concentrations of the analytes of interest (i.e., T, dihydrotestosterone [DHT] and estradiol) [11]. An investigation comparing the responses of CSs for T, DHT and estradiol prepared in artificial matrix versus those CSs prepared in human serum had to be carried out. The performance of CSs prepared in both matrices was parallel and showed linear correlation regression slopes of the CSs near to one between responses from both matrices for all three analytes of interest. The precision and accuracy using QCs prepared in both matrices were also evaluated. It was demonstrated that the performance of CSs prepared in artificial matrix was comparable to that of calibrators prepared in human serum and therefore assured reliable measurements.

Example 3

Development of prodrugs, with an ester-linked sidechain, is a strategy used to improve BA of a drug [12]. In most cases the labile ester bond is cleaved rapidly following prodrug absorption and only the active drug is present in the systemic circulation. Measurement of drug concentrations may be confounded if there is ex vivo conversion of the prodrug to the drug during blood sample collection and processing. This becomes a more significant concern if the drug administered is endogenous. For example, testosterone undecanoate (TU) is a prodrug of T formed by esterification of a hydroxyl group on the D-ring at position 17. Oral TU is metabolized partly in the intestinal wall into 5-α-dihydrotestosterone undecanoate (DHTU). The ester linkage of TU is readily cleaved and hydrolyzed to produce T and DHTU to DHT in blood by circulating nonspecific esterases [12]. As mentioned above, the primary efficacy endpoint for TRT is the responder rate based on the average serum or plasma T concentration at the end of active treatment. In addition, the key secondary endpoint is the percentage of TRT-treated subjects with serum or plasma T peak concentrations (C_max) within the predetermined ranges. Therefore an accurate and precise measurement of T is pivotal in efficacy and safety clinical studies for TRT.

There are several factors to consider for the bioanalysis of oral TU. First, even before considering the potential TU-to-T ex vivo conversion, endogenous T concentrations are usually higher in serum compared with those in plasma (i.e., when blood samples are collected from the same subject at the same time point and then split up for analysis) due to matrix effect. The target normal total T concentration range (i.e., reference concentrations) may vary across matrices and is dependent on the bioanalytical method employed in clinical studies. For example, during the development of a TRT product, a clinical study was conducted in 105 healthy, eugonadal males to determine their normal T concentration range. No TRT was administered in this study as the objective was to define the normal (i.e., eugonadal) T concentration range based on endogenous T. Subjects were fasted for at least 8 h before blood collection. Two types of samples (i.e., serum and sodium fluoride/EDTA plasma) were collected for the analysis of T and DHT. The reference range, based on the central 95% of the study population, for NaF/EDTA plasma was 222–800 ng/dl, while it was 286–991 ng/dl for serum [13]. The reference range for total T in serum derived from this study was in agreement with the generally well-accepted serum T concentration range of 300–1000 ng/dl in eugonadal males.

In addition, TU-to-T ex vivo conversion may occur during sample handling. The following factors were found to contribute to the TU-to-T ex vivo conversion that affects the concentration measurements in both serum and plasma [10,14]:

• Post-collection incubation temperature: lowering the temperature reduces conversion.

• Post-collection incubation time: TU-to-T ex vivo conversion occurs most rapidly during the first 30 min post collection. Reducing the incubation time will help reduce the TU-to-T ex vivo conversion.

• TU concentration: the TU-to-T ex vivo conversion is TU concentration dependent.

• Presence of esterase inhibitor in test tubes: the presence of an esterase inhibitor (e.g., NaF in NaF/EDTA tubes) further reduces the TU-to-T ex vivo conversion.

TU-to-T ex vivo conversion can be prevented more efficiently in NaF/EDTA plasma compared with serum as an esterase inhibitor, NaF, is present in the tubes. In addition, plasma samples can be placed in an ice bath at a lower temperature compared with serum, which needs to sit at room temperature for the first 30 min. Lower temperature helps prevent the TU-to-T ex vivo conversion. Deviation from standard procedures of sample handling and processing may lead to unexpectedly higher T concentrations from plasma compared with serum prepared from blood when collected at the same time point from the same subject. The consequence of this will be significant as it can cause false positives (i.e., higher plasma T concentration than the actual T concentration in vivo) that will falsely increase the efficacy responder rate for the TRT product. The analyte’s stability in the sample should be evaluated and demonstrated beginning from the blood drawn into a collection tube through the separation of plasma or serum from the red blood cells and other blood components, as a part of the method development and validation process. In the case of TU, it is important to demonstrate that the ester does not undergo hydrolysis into T during sample processing. The circulating TU concentration is almost 20-times higher than the T concentration [15], and therefore even a minimal degradation of TU will impact the measured T concentration.

Lessons learned from examples 1–3

CSs should be prepared using the same biological matrix as the study samples and should be free of the endogenous analyte – in these cases, T. The QC samples should be prepared by spiking known concentrations of analyte(s) in the same biological matrix as the study samples. The endogenous concentrations of analyte(s) in the biological matrix should be evaluated prior to QC preparation (e.g., by replicate analysis). The QC concentrations should account for the endogenous concentrations in the biological matrix (i.e., additive) and be representative of the expected study sample concentration range. Traditional approaches of preparing CS and QC samples for endogenous therapeutic analytes include using matrix with low endogenous T (e.g., from females), stripping endogenous analytes from the matrix, or using the standard addition method. When artificial matrices are used, parallelism evaluation needs to be carried out to ensure that the performances of the CSs prepared in different matrices are parallel and demonstrate a linear correlation regression with slopes being near to one. The bottom line is that the bioanalysis of endogenous therapeutic analytes should be able to address the questions that need to be answered during drug development. As for the development of TRT products, the T concentration values play a pivotal role in the assessment of efficacy and safety. Therefore it is not an overstatement to say the bioanalysis of the primary analyte of interest, T, which is an endogenous compound, is pivotal for successful TRT development.

Example 4

The reliability of the bioanalytical method used for quantifying free hormone analog may be compromised by several issues, impacting the accuracy of pharmacokinetic data in clinical pharmacology studies. Free hormone analog in human serum was quantified using a specific sandwich ELISA, and polyethylene glycol solution was used in a precipitation step. First, the assay’s recovery of free hormone analog from human serum exhibited high variability due to significant matrix effects. While the regular human hormone and the analogs are minimally protein bound, only about 50–60% of hormone analogs were accounted for as free with this methodology. Second, incurred study samples went through a different sample preparation process compared with the CSs and QCs, resulting in uncertainty of the accuracy and precision of the incurred study samples. The CS and QC samples were prepared in precipitated human serum instead of human serum, which was used for the incurred study samples, to avoid the matrix effects on recovery observed in the incurred study samples. Finally, the uncertainty surrounding the reliability of quantified free hormone analog concentrations in incurred study samples from patients arose from the use of a single QC sample in study sample analysis, providing no assurance for the concentrations measured within the broad concentration range of the incurred study samples. Additionally, the calculation of precision and accuracy for the QC samples using experimental concentrations rather than standard nominal concentrations introduced bias and was not in compliance with the bioanalytical method validation guidance [4].

Lessons learned from example 4

It is essential to use validated bioanalytical methods that utilize CS and QC samples in appropriate matrix (e.g., human serum), following the recommended sample preparation procedures. It is highly recommended to be in compliance with the validation acceptance criteria outlined in the recently published ICH M10 guidance [8]. When conducting analytical runs, it is important to include a blank sample, CSs covering the expected sample concentrations, multiple levels of QCs (e.g., at LLOQ and low, medium and high QC concentrations) in at least two sets or as a minimum of 5% of the study sample count, in addition to the incurred study samples. Reliable results from hormone bioanalysis can be obtained by following these best practices and guidance.

Example 5

Calcitriol is a synthetic vitamin D analog which is indicated for the management of hypocalcemia and the resultant metabolic bone disease in patients undergoing chronic renal dialysis. The product-specific guidance (PSG) for calcitriol capsule recommends BE to be demonstrated based on the 90% CIs of baseline-corrected calcitriol [16]. In one instance, a bioanalytical method using methanol as the surrogate matrix was developed; however, calcitriol in human serum was measured in the pivotal BE study. In the method validation report, the recovery data were not reported for QCs in methanol, but those of QCs in human serum were. In addition, the accuracy and precision were validated using low QCs in methanol but medium and high QCs in human serum and using a standard curve established with CSs prepared in methanol. Given that the concentrations of QC samples were not comparable between methanol and human serum and did not represent the concentration range of study samples, additional cross-validation was needed to demonstrate that the differences in these two matrices did not impact the accuracy and precision of calcitriol measurements in human serum using standard curves constructed in a methanol matrix. During study sample analysis, the data from the first several subjects demonstrated that the majority of sample concentrations were below the medium QC concentration and the predose concentrations were around the low QC concentration. Therefore the concentrations of QCs in human serum in the BE study did not adequately represent the concentrations obtained from the study samples. As a result, additional QCs of relevant concentrations in human serum were warranted in the study sample analysis.

Lessons learned from example 5

As per the ICH M10 guidance [8], surrogate matrices may be acceptable for bioanalytical methods in cases where it is difficult to obtain an identical matrix to that of study samples. The choice of surrogate matrix should be scientifically justified. The use of stripped serum has been reported in the literature for bioanalysis of calcitriol [17]. When matrices different from the authentic matrix (i.e., surrogate matrices) are used, justification with supporting data should be provided. When the matrix of CSs is different from that of study samples, there may be potential differences in recovery and matrix effect. Therefore cross-validation should be conducted to support the accuracy and precision of calcitriol measurements in both methanol and serum matrices. A parallelism test is warranted to detect potential matrix effects.

Example 6

Similar to the example above, a surrogate matrix approach was adopted for another endogenous hormone when developing the bioanalytical method to analyze study samples from the BE study. In the bioanalytical method, CSs in phosphate-buffered saline (PBS) with 2% BSA were used to construct the standard curve. In the method validation report, the recovery data were submitted for QCs in PBS with 2% BSA only but not for those with the authentic matrix (i.e., serum). As explained above, recovery data of both surrogate and authentic matrices should be provided to ensure that there is no significant matrix effect. Parallelism demonstrates that the serially diluted incurred sample response curve is parallel to the calibration curve; parallelism is a performance characteristic that can detect potential matrix effects [8]. In this case, a parallelism study was included as part of the method validation. However, only one run each from the surrogate matrix (i.e., PBS with 2% BSA) and human serum was performed. Also, linear regression was used in the parallelism study, while quadratic regression was used in the incurred study sample analysis. In addition, the concentrations of QCs in human serum in both method validation and study sample analysis were not representative of study sample concentrations.

Lessons learned from example 6

For a parallelism study, it is expected that at least three sets of parallelism data comparing both matrices should be obtained, and the same regression model and weighting factor should be used in both the parallelism study and pivotal BE study. In addition, QCs in human serum should be representative of the measured concentration range of study samples in the parallelism study. Ideally, the matrix used for method development and validation of an endogenous substance should be the same as the matrix of the incurred study samples. If the availability of endogenous analyte-free authentic matrix is very limited, the use of surrogate matrix may be acceptable with adequate scientific justification for its choice. Due to the potential differences between surrogate and authentic matrices, the impact of different matrix effect and recovery in both matrices should be evaluated during method development and validation prior to study sample analysis. A well-designed parallelism study is also critical to assure that the observed changes in response per given changes in analyte concentrations are comparable between the surrogate and authentic matrices, which indicates there is no concerning matrix effect in the adopted surrogate matrix.

Inspections & compliance

The aim of inspections is to ensure data quality and integrity, in addition to human subject protection. Here we present some observations found through inspections from a compliance perspective. As stated above, drug products with endogenous substances present challenges when it comes to method development, validation and baseline correction.

The following case illustrates the difficulties of comparing drug products where the endogenous concentrations are relatively high compared with those of the exogenously administered substance. The FDA inspected a study for a drug approved for the treatment of cholic acid disorders. The main objective of the study was to compare the BA of the clinically tested formulation to the to-be-marketed formulation. The study involved administration of multiple doses of the to-be-marketed formulation and clinically tested formulation to healthy male subjects using a randomized, replicate design. Inspection of the bioanalytical component of the study observed that a surrogate matrix (i.e., buffer) was utilized instead of plasma to generate calibration curves. A deficiency was noted because matrix effect was not compared, thus possibly impacting accuracy and precision calculations in addition to dilution integrity. In addition, the absence of long-term stability data was identified as another issue.

A survey was conducted covering 12 years of the FDA Office of Study Integrity and Surveillance’s inspections and remote record assessments (a total of 164) of drug products which are also endogenous substances. These included levothyroxine, estrogen and estradiol products, progesterone, T, potassium chloride, potassium citrate, iron, vitamin K, vitamin D, ursodiol and omega-3 fatty acid products. Most of the inspections were for sex hormone products (64% of the 164 inspections), reflecting the large number of applications that the agency receives for these products. In terms of outcomes for on-site inspections, approximately 33% were classified as ‘Voluntary Action Indicated’, 66% as ‘No Action Indicated’ and a single case as ‘Official Action Indicated’, implying significant concerns about data quality and possible rejection of the study. For remote record assessments conducted during the COVID-19 epidemic, there were two cases out of eight where not all data were found reliable. From this, it was observed that bioanalytical laboratories, in general, have been able to conduct studies with endogenous substances without significant compliance issues and concerns; however, challenges in bioanalysis of endogenous therapeutic analytes still remain, and precautions are needed to ensure reliable results.

Discussion: things to consider

Bioanalysis of endogenous analytes is always challenging on either analytical platform – LC–MS/MS or ligand binding assay (LBA). Vigilant method development and validation strategies are warranted when working on these endogenous analytes. If available, biological matrix to prepare CSs and QCs should be the same as the study samples (i.e., authentic biological matrix) and it should be free of matrix effect and interference. The ICH M10 guidance illustrates some approaches that can be used in those cases where matrices without interference are not available: the background subtraction approach; the standard addition approach; the surrogate matrix approach; and the surrogate analyte approach [8]. Figure 1 & Table 2 illustrate these approaches.

Background subtraction approach

With the background subtraction approach, similar recovery and matrix effect between samples and CSs can be obtained with the use of the same matrix for the calibration curve and study sample analysis. However, the LLOQs of these methods are limited by the background endogenous analyte concentrations in particular batches used as blanks in constructing the calibration curves. In addition, there may be different background endogenous analyte concentrations in different batches of pooled matrices, resulting in an inconsistent result from background subtraction over time or between different laboratories. For example, the exogenous estrogens are metabolized in the same manner as endogenous estrogens. Per the current draft PSG for conjugated estrogen tablets, baseline-adjusted unconjugated estrone, baseline-adjusted total estrone, unconjugated equilin and total equilin in plasma are recommended to be measured [19]. Estrone is an endogenous substance, while equilin (i.e., an estrogen from horse) is not endogenous in humans. The PSG for conjugated estrogen tablets recommends BE studies to be conducted in physiologically or surgically postmenopausal women. Blank plasma from postmenopausal women usually contains relatively high baseline concentrations of estrone. When plasma without endogenous estrone interference is not available, the background subtraction approach can be used in the bioanalysis of study samples from a BE study. Typically, the pooled blank plasma that is used for CS and QC preparation is analyzed with the internal standard (IS) in multiple replicates (e.g., in triplicate) and the mean area response ratio of the analyte to IS in blank plasma is subtracted from each of the CS and QC samples in the same analytical run to correct for endogenous baseline concentrations. When multiple analytical runs are performed for study samples from a BE study, the area response ratio of the analyte to IS in blank plasma from each run is expected to be constant, because the same pooled blank plasma is used to prepare CSs and QCs for all the analytical runs.

Standard addition approach

The standard addition approach is typically used to determine the concentration of endogenous analytes in authentic matrix used for CSs and QCs; it can be used for measurements in study samples as well. For study sample analysis, every sample is divided into aliquots of equal volume, and all aliquots but one are separately spiked with known and varying amounts of the analyte standards to construct a calibration curve for either the authentic blank matrix or every study sample (e.g., with three to five points). The endogenous blank concentration or study sample concentration is determined as the negative x-intercept of the calibration curve prepared in that particular study sample. This method uses the exact same matrix of every study sample for the construction of its own calibration curve, avoiding batch-to-batch matrix differences. However, it may involve labor-intensive sample preparation and can be time-consuming. In addition, the standard addition approach should only be used for analytical platforms with linear response. In Example 1 above, blank matrix was obtained from female subjects and the endogenous T concentration was low. However, the blank matrix free from T (i.e., the analyte) could not be obtained. Thus accurate measurement of T concentration could only be obtained with a correction using the standard addition method.

Surrogate matrices approach

The surrogate matrices approach, especially using artificial matrices, is often used. This is a cost-efficient approach if an appropriate method validation strategy is incorporated to avoid matrix effect. While using surrogate matrices may appear to be more convenient and simple, it requires more early up-front effort in method development [20]. Scientific justification for the choice of surrogate matrix (e.g., neat solution, artificial matrices, or stripped matrices) should be provided. Similarity of matrix composition may not be required as long as the accuracy and precision, parallelism and dilution linearity (e.g., for LBA) are adequately demonstrated [21]. The use of alternate matrices for the preparation of CSs is generally not recommended unless the analyte-free biological matrix is not readily available nor can be prepared. In Example 2 mentioned above, surrogate (i.e., artificial) matrix was used to prepare CSs and QCs. While the performance of CSs prepared in both surrogate matrix and authentic matrix was comparable and showed linear regression correlation slopes, the endogenous concentration of the therapeutic analyte was not accounted for in quantification of the incurred study samples and resulted in questioning the accuracy of the measured endogenous therapeutic analyte concentrations. In the calcitriol example (Example 5) above, neat solution matrix (i.e., methanol) was used. As stripped serum matrix has been reported in the literature to be used in calcitriol bioanalysis, justification for the selection of a neat solution matrix over stripped serum would be warranted [22]. Additional cross-validation data should be submitted to support the accuracy and precision of analyte measurements in both the surrogate matrix (i.e., neat solution) and authentic matrix (i.e., serum). A well-designed parallelism study is critical to assure that the observed changes in response to changes in analyte concentrations are equivalent for the surrogate and authentic matrices. This also indicates there is no concerning matrix effect in the employed bioanalytical method. It is expected that at least three sets of parallelism data comparing both matrices, using the same regression model and weighting factor as the pivotal BE study, to be reported. QCs in human plasma should be representative of the measured concentrations of study samples in the parallelism study. Given the importance of demonstrating parallelism, Jones et al. took the approach of employing standard addition to evaluate parallelism [23]. This approach was used for both surrogate matrices and surrogate analytes. The endogenous concentration that was derived by obtaining the negative x-intercept through the extrapolation from the authentic analyte calibration curve in authentic matrix was compared with that derived from direct intrapolation from the surrogate matrix or surrogate analyte calibration curves; if they agreed, it was found that parallelism was demonstrated.

Surrogate analyte approach

While it is costly, the surrogate analyte approach allows direct and sensitive quantification of analytes without interference of endogenous substance. While stable-isotope-labeled analytes are used as surrogate standards to construct the calibration curve for the quantification of endogenous analytes, this approach is only applicable to mass spectrometry-based platforms. In addition, as this approach is based on an assumption that the physicochemical properties of the authentic and surrogate analytes are the same with the exception of molecular weight, the ratio of the mass spectrometric responses (i.e., the response factor) of the labeled to unlabeled analytes should be close to unity (1) and remain constant over the entire calibration range. Though not commonly observed, the surrogate analyte approach is reported in literature. Agrawal et al. used ¹³C₃-cortisol as a surrogate analyte and cortisol-d₆ as the IS to develop a validated LC–MS/MS method using a surrogate analyte for endogenous cortisol quantification in human whole blood in a clinical trial with an inhaled nonsteroidal gluococorticoid receptor modulator [24]. The geometric mean ratio of cortisol signal response to ¹³C₃-cortisol was 0.972 (range: 0.910–1.02), indicating similar ionization efficiency between these two molecules. In addition, this method also demonstrated acceptable accuracy, precision and reproducibility, making it a suitable bioanalytical method to support at-home sampling for cortisol in clinical trials. An LC–MS/MS method using ¹³C₅-NAD⁺ as a surrogate analyte was also developed and validated for the quantitative determination of NAD⁺ in human whole blood [25]. The ultimate goal of a bioanalytical method is the accurate measurement of the analyte of interest with minimal variability; therefore an appropriate predefined development and validation strategy is warranted. The strategy includes inclusion of the appropriate quality measurements by incorporating suitable QC samples. These quality measurements range from simple precision and accuracy evaluation to various stability assessments. The requirement for quality measurements depends on the sample handling conditions and bioanalytical method [18].

In a recently reported clinical study, a randomized clinical trial with IFN-β1 products to better understand the potential role for pharmacodynamic biomarkers in biosimilar development, neopterin and myxovirus resistance protein 1 (MxA) were listed as one of the endpoints [26]. The proposed endogenous analytes were very diverse in nature and needed to be measured in different matrices (i.e., serum and hemolyzed blood). In addition, they needed different analytical platforms to perform the analysis (i.e., LC–MS/MS for neopterin and LBA for MxA). In this regard, a surrogate matrix approach was chosen by considering its feasibility and cost-efficiency. Appropriate surrogate matrices (i.e., fetal bovine serum for neopterin and lysis buffer for MxA) were optimized during method development. Though the analytical platforms are diverse, the strategy for method validation was very similar. The respective study matrices were screened (i.e., ten lots of blank serum/hemolyzed blood) during the initial stage and the endogenous concentrations ascertained (i.e., ng/ml, pg/ml) in a representative matrix after repeated measurements (n >3). During method validation, the CSs were prepared in surrogate matrices. QC samples at the LLOQ, low QC, medium QC, high QC and ULOQ levels for MxA were utilized in surrogate and study matrix to evaluate accuracy and precision. QCs at LLOQ in study matrix were not prepared for neopterin due to higher endogenous concentrations. In addition, the benchtop stability in both matrices was evaluated for both compounds, while freeze–thaw stability and long-term stability were evaluated only in study matrix because the surrogate matrix samples were freshly spiked as needed. The matrix effect was evaluated by performing parallelism (i.e., comparing standard curve slopes for neopterin; back-calculating endogenous concentration by extrapolation; dilution of study samples for MxA). During study sample analysis, respective QCs (i.e., low, medium and high QC) in both matrices (i.e., surrogate and study matrix) were incorporated into each analytical run. Table 3 summarizes the suggested method development and validation strategy for all four proposed approaches.

None of these approaches fit all cases, and a thorough understanding of the study (e.g., the type and endogenous concentrations of analyte; matrix type and availability; and required assay sensitivity) and the pros and cons of each approach is crucial in selecting the appropriate approach to meet the need. We hope that this article gives a better grasp of the important considerations in selecting a strategy in method development and validation, and eventually, conducting bioanalysis of endogenous therapeutic analytes in clinical studies to generate reliable data.

Matrix effect & sample preparation

Matrix interference presents a notorious challenge to accurate and precise quantitation of endogenous analytes, especially when using a surrogate matrix or analyte approach. Sample preparation plays a crucial role in accurate and precise quantitation by removing interferences and improving analytical performance (e.g., sensitivity) and stability. It is an essential step that needs to be performed before the endogenous analytes of interest are introduced into an analytical instrument. While reducing matrix interference at the sample preparation (i.e., clean-up) stage can be achieved through sample dilution or selective extraction [27], conventional sample extraction methods such as solid-phase extraction, liquid–liquid extraction and protein precipitation fall short in selective extraction and still suffer from matrix effect due to co-extraction of components with similar physicochemical properties. Traditional sample extraction methods require higher sample volumes which eventually pose more matrix interference by indulging more undesired co-existing elements. Therefore, microextraction techniques could be one of the viable options to combat matrix interference. The recent trends of these miniaturized, high-throughput and automated sampling techniques may help overcome these challenges by selective extraction with significantly lower sample volumes. For example, advancements in solid-phase microextraction [28] with selective coating, enable the selective extraction of the analyte of interest even in the presence of undesired endogenous components that have similar physicochemical properties. Similarly, other emerging sample preparation techniques, such as micelle-dominated distribution strategy for non-matrix-matched calibration without an IS [29], could be another good option. The advancement of these newer microextraction techniques minimizes the matrix effect by reducing the interference of sample matrix and provides more enriched samples [30].

Conclusion

Bioanalysis for endogenous therapeutic analytes poses more challenges than typical bioanalysis. No singular approach is perfect, and therefore the best strategy needs to be selected case by case. For example, use of surrogate matrix is one of the most often used approaches in endogenous analyte bioanalysis. Unless it is ensured that the surrogate matrix has consistent and comparable extraction recovery, accuracy and precision performance to the authentic matrix, reliable bioanalytical data may not be obtained using a surrogate matrix. However, if the compatibility between the surrogate matrix and authentic matrix can be demonstrated, using a surrogate matrix will have the advantage of providing a cleaner and simpler matrix with less endogenous interference.

The bottom line is that the question that needs to be addressed by conducting the study has to be clearly understood and the best approach needs to be employed to address the question. The importance of reliable, reproducible and robust bioanalytical methods in drug development cannot be overemphasized. It is always encouraged to engage in early communication with regulatory agencies to understand the regulatory perspective [31]. When we communicate and collaborate together, we can overcome challenges and advance public health.

Future perspective

Quantification of endogenous therapeutic analytes in biological matrices will be advanced by the evolution of bioanalysis. For example, future refinement of analyte extraction techniques (e.g., solid-assisted liquid–liquid extraction, microsampling extraction) [30] will help improve the accuracy and precision when measuring endogenous therapeutic analytes using surrogate matrices and as a result, reducing matrix interference and minimizing recovery differences. As discussed above in the ‘Matrix effect and sample preparation’ section, efficient sample preparation techniques are critical in removing the undesired interfering elements. As extraction methods alone would not completely remove matrix interference, advancement in analytical instrumentation (e.g., resolution of the mass spectrometer with improved sensitivity [32], in addition to significant enhancement of chromatographic separation) in conjunction with the use of surrogate analytes and/or surrogate matrices can significantly empower the bioanalysis of endogenous analytes. For example, advancement in instrumentation capabilities will minimize the differences in ionization between surrogate and authentic analytes in the presence of other matrix ions. While challenges will continue to be encountered in endogenous therapeutic analyte bioanalysis, the expected advances in extraction methodologies, analytical instruments and platforms will better equip us to overcome these challenges.