Analyte Recovery in LC-MS/MS Bioanalysis: An Old Issue Revisited

Abstract

There are several challenges associated with LC-MS/MS bioanalytical method development and validation. Low and variable recovery of some analytes, especially the more hydrophobic ones, is often challenging. Analytes can be lost to various extents throughout the process of sample collection, storage, before, during, and/or after sample preparation and analysis. The calculation of overall extraction recovery can detect problems of low recovery during sample preparation but does not identify the source(s) of analyte losses. Low overall analyte recovery is the net result of losses that can happen for multiple reasons at all steps of sample preparation and analysis. Therefore, identifying the source(s) of analyte loss during sample preparation can help guide the optimization the bioanalysis conditions to minimize these losses. In this article we propose a practical protocol to systematically identify and quantify the sources of low analyte recovery. This allows the proper choice of strategies to optimize the relevant bioanalytical conditions to minimize analyte losses and improve overall recovery.

1. Introduction

Bioanalytical support is essential for all stages throughout the drug discovery and development process. Bioanalysis is required for thousands-millions of samples generated from all preclinical and clinical studies, which exist in very diverse matrices. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is currently the gold standard for quantitative analysis, in support of drug discovery and development, and biomedical research in general, due to its sensitivity and selectivity [1]. All bioanalytical methods must be validated to various extents to ensure they are rugged, reproducible, and reliable to produce accurate and precise data. Extensive guidelines from the regulatory agencies including the FDA and EMA are continuously updated to establish the minimum requirements for the validation of bioanalytical methods used in drug discovery and development [2, 3].

There are several challenges associated with LC-MS/MS method development and validation [4–7]. In particular, low and variable recovery of some analytes, especially the more hydrophobic ones, is often challenging [5]. There has been an increasing need for the analysis of hydrophobic compounds in many therapeutic areas. About 40% of FDA approved drugs are categorized as practically insoluble in water, and it was recently predicted that nearly 90% of drugs in drug development pipeline are water insoluble [8, 9]. In addition, non-drug hydrophobic analytes such as phospholipids, cholesterol, eicosanoids, and bile acids are often encountered in other areas such as biomarker-, metabolomics-, and lipidomics-related research [10, 11].

FDA guidelines defines recovery as “extraction efficiency of an analytical process, reported as a percentage of the known amount of an analyte carried through the sample extraction and processing steps of the method” [2]. This is a broad definition and does not necessarily address all aspects of analyte losses during sample collection, storage, and analysis. The simplest and most fundamental form of bioanalytical recovery is the overall recovery, which can be calculated as the ratio of the measured vs. nominal concentrations of quality control standards. In reality, however, recovery determination is more complex and often require further investigation than simple overall recovery determination. For example, what standards are used, how they are prepared, in what matrix, at what concentration, how they are stored and for how long, etc., are all details that can make the difference between false vs. accurate bioanalytical recovery determination [6].

Furthermore, digging deeper beneath the surface with overall recovery determination is often required to help troubleshoot low overall recovery. For example, why overall recovery was low? where was the analyte lost? was it pre-extraction, during extraction, reconstitution, post extraction, or was it due to matrix effect? etc., are all questions that require answers before bioanalytical recovery can be optimized. Breaking down the individual components of overall recovery is essential to identify where analyte losses took place before bioanalytical recovery can be improved.

Analytes can be lost to various extents throughout the process of sample collection, storage before, during, and/or after sample preparation and analysis. In this article, we will only consider losses during sample preparation and analysis, which are inherent to the bioanalytical method conditions, and can be minimized by optimizing bioanalytical method parameters. In contrast, analyte losses at other stages such as sample generation and collection are inherent to the test systems themselves, which in many cases are not a part of the bioanalytical method itself. Table 1 summarizes the common sources of analyte losses during sample preparation and analysis. The sources of analyte losses can be divided into four categories, according to what stage of sample preparation they take place, including: (i) pre-extraction, (ii) during-extraction, (iii) post-extraction, and (iv) Matrix effect. Furthermore, losses at every stage can be divided into immediate/time-independent losses that take effect the moment the analyte is spiked into the matrix vs. time-dependent losses, which exacerbate over time, most likely due to chemical and/or biological degradation. As explained later, dividing the sources of analyte losses into these categories facilitate their identification and quantification. Identification of the particular sources/mechanisms of analyte losses is required for the efficient optimization of bioanalytical conditions to improve overall recovery.

Table 1.

Common sources of analyte losses pre-, during-, and post-Extraction.

Pre-Extraction

Instability: Chemical + Biological degradation by matrix

Irreversible Binding to matrix components (proteins, RBC, salt complexation, etc.)

NSB to vial walls/ Insolubility/Precipitation

During-extraction

Instability: Chemical + Biological degradation by matrix in the presence of ACN

Extraction Inefficiency: Inability to liberate analyte bound to matrix components

NSB to vial walls/ Insolubility/Precipitation in the presence of ACN

Evaporation: Analyte degradation during evaporation/concentration

Post-extraction

Reconstitution Issues: Irreversible binding to the residual matrix components and/or NSB to vial walls

Instability: Chemical + Biological degradation by the unextracted residual matrix components in the presence of the reconstitution solvent

Matrix Effect

Ionization suppression/enhancement by interfering endogenous compounds in the MS source

Matrix effect and nonspecific binding (NSB) are the most widely known and investigated mechanisms for analyte losses in the literature [6, 7, 12-15]. NSB is the result of the sorption of analytes by the inner surfaces of labware including vials, tubes, pipette tips, etc. In addition, sorption of drugs to plastic polyvinylchloride (PVC) infusion bags and to plastic intravenous tubing is well documented, and might cause treatment failure [16]. Th extent of NSB can be severe, with >90% analyte losses previously reported [12]. Sorption includes both adsorption due to analyte binding to the outer surfaces and absorption involving penetration and diffusion of analytes across the surface of the labware material (leaching) [12, 14]. Depending on the analyte and the labware material, analytes sorption could be described as a hybrid process of relatively fast adsorption (outer surface binding) and slower absorption (leaching) processes to various extents [17-19]. Therefore, analyte disappearance from a labware due to NSB is typically a biphasic curve consisting of a fast (minutes) and steep segment followed by a much slower (hours-days) and shallower stage [12, 19, 20].

A variety of materials are used to manufacture labware including, Polyvinyl chloride (PVC), polystyrene plastic, polypropylene (PP), polytetrafluoroethylene, glass, etc. [21]. Analyte interaction with the labware surface is the net result of both ionic/electrostatic as well as hydrophobic/van der Waals interactions [22]. Electrostatic interaction takes place between the ionic functional groups of drugs with those on a labware surface such as silanol groups on glass surfaces, while hydrophobic interaction takes place between hydrophobic drugs and hydrophobic surfaces such as plastic. Because drug candidates are becoming more hydrophobic and because lab conventional hydrophobic plastics is now more common than glassware, hydrophobic adsorption is typically the main driving force responsible for NSB [12, 23, 24].

To overcome NSB to plastic surfaces, low-adsorption plates and vials with modified surfaces via the introduction of hydrophilic groups are available, [24, 25]. However, hydrophilic coating can also enhance NSB due to ionic interaction [23]. Similarly, glass surfaces can be treated with silane coupling agents, which masks the active free silanol moieties with a variety of organofunctional groups [26]. The extent of adsorption depends on the properties of the analyte, container, temperature, and matrix (e.g., ionic strength pH, lipids, and protein content) of the biological samples [27]. For example, lower sample pHs than the analytes pKa, will lower and enhance NSB of acidic and basic drugs, respectively, to negatively charged surfaces such as glass, polypropylene, or polystyrene [28]. Matrix and sample composition also play a major role. For example, NSB is magnified with hydrophobic analytes in matrices lacking binding partners (e.g. proteins, lipids, and other macromolecules) such as urine, CSF, plasma dialysate, filtrate, centrifugate, or infusion solutions [16, 20] as opposed to matrices rich with macromolecules, such as plasma, bile, and tissue homogenates [29, 30]. Similarly, samples with low organic solvent content such as buffer samples from in vitro assays (e.g. protein binding or permeability assays) show high NSB [23, 31]. Alternatively, anti-adsorptive agents are used to solubilize the analytes and/or block their absorption to the labware walls such as bovine serum albumin (BSA) [30], 3-[3-cholamidopropyl)-dimethylammonio]-1-propane sulfonate (CHAPS) [29], sodium dodecyl benzenesulfonate (SDBS) [32], Tween 20 or 80 [33], α/β/γ-cyclodextrin, Tween 80 [29], organic solvents (DMSO, methanol, acetonitrile, DMSO, methyl-t-butyl) [34], quaternary ammonium salts, and higher buffer concentration [12]. However, these agents may interfere with the chromatography and/or ionization in the MS source [7, 35].

LC-MS/MS with electrospray ionization (ESI) is currently the gold standard for the quantitative analysis in support of drug discovery and development and biomedical research, in general, due to its sensitivity and selectivity. However, ESI detection has a major advantage, that is the “Matrix Effect” Phenomenon. Matrix effect is caused by other matrix components that are co-extracted and co-eluted with the analyte(s) of interest, such as salts and phospholipids [11, 15, 36]. Interfering matrix components can suppress or enhance the ionization of the analyte of interest in the MS source. Exogenous causes of matrix effects are also possible, including buffer salts and ion-pairing reagents in mobile phase, and leached materials from the surface of labware and sample containers such as polymers [6, 7, 15, 37]. Matrix effect can be caused by both volatile and nonvolatile components. Nonvolatile solutes can prevent the formation of smaller droplets in the ion source by increasing surface tension and viscosity of charged droplets [38–40]. While volatile compounds can compete for available charges and access to droplet surfaces [41, 42]. Several studies have reported that matrix effect is, in general, stronger in positive compared to negative ionization mode [43]. This could be due to the fact that fewer compounds including matrix components such as proteins, amino acids and buffer salts are ionized in negative polarity mode, thereby, competition for charges and access to droplet surface is much lower than in the positive polarity mode.

In this article we propose a protocol to help address the challenges related to the poor recovery of some analytes during sample preparation for LC-MS/MS assays. The objective of the protocol is to define and quantify the various sources of analyte losses at all stages of sample collection, preparation, LC-MS/MS analysis, and storage before and during analysis. We then propose solutions and strategies to minimize the losses at each stage with the goal of improving analyte overall bioanalytical recovery. We have chosen plasma as the sample matrix, because it is the most common biological matrix encountered in drug discovery and development, but the same protocol could be applied to any other matrix. Similarly, protein precipitation with acetonitrile (ACN) followed by sample concentration and volume reduction using evaporation was used as the sample extraction method, but the same protocol could be applied to any other extraction methods.

2. Experimental

A description of all quality control (QC) standards prepared in this protocol is presented in Table 2. The same standards at every stage are also presentenced schematically in Fig. 1–4. We have followed a nomenclature system, where standard names start with the prefixes; Pre-, During-, and Post-, indicating when the matrix was spiked with the analyte stock solution in reference to extraction with ACN. Pre, i.e. Pre-extraction standards are spiked with the analyte before ACN extraction, while During- standards are spiked after ACN is added. Post-standards are spiked at the reconstitution step, after blank matrices are completely extracted and evaporated.

Table 2.

QC standards Prepared at various stages of the bioanalytical protocol.

Standard	Matrix	Pre-Extraction	During-Extraction	Post-Extraction
Tier-1 Standards:
Pre-i: Pre-Extraction-Overall	Plasma	30 min	2 h	48 h
During-ii: ACN-1st-Immediate	Plasma-ACN	-	Immediate	Immediate
Post-ii: Post-Extraction-Immediate	Extracted Plasma	-	-	Immediate
Neat	Sample Solvent	-	-	-
Tier-2 [A] Standards:
During-i: ACN-1st-2 h	Plasma-ACN	-	2 h	Immediate
Post-i: Post-Extraction-48 h	Extracted Plasma	-	-	48 h
Post-i-S: Post-Extraction-48 h-Sample Solvent	Sample Solvent	-	-	48 h
Tier-2 [B] Standards:
Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated	Plasma	Immediate	Up to 2 h	Up to 48 h
Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix	Inactive Plasma	Immediate	Up to 2 h	Up to 48 h
During-i-Acc: ACN-1st-Accelerated	Plasma-ACN	-	Up to 2 h	Up to 48 h
During-i-Acc-I: ACN-1st-Accelerated-Inactive Matrix	Inactive Plasma-ACN	-	Up to 2 h	Up to 48 h
Tier-2 [C] Standards:
Pre-i-Acc: Pre-Extraction-30 min-Accelerated	Plasma	30 min	Up to 2 h	Up to 48 h
Pre-i-Acc-I: Pre-Extraction-30 min-Accelerated-Inactive Matrix	Inactive Plasma	30 min	Up to 2 h	Up to 48 h

Fig. 1.

QC strands and recovery ratios used in Tier-1 to quantify analyte losses related to overall recovery, matrix effect, evaporation and reconstitution, and all other remaining sources lumped together.

Fig. 4.

Strategies proposed to optimize bioanalytical method conditions and improve recovery at various stages.

The prefix in the standard names is followed by the roman numbers “i” and “ii”. Standards carrying the number “i” in their names are incubated for a certain period of times before, during, and/or post-extraction as described below. While standards labeled “ii”, are progressed to the next step immediately.

In addition, analyte losses were quantified under full incubation periods in Tier-1 (anticipated to process an average sample batch) of 30 min, 2 h, and 48 h pre-, during-, and pos-extraction, respectively. Therefore, in the following stages of the protocol, we use accelerated incubation periods up to the tested ones in Tier-1. The term “acc”, i.e. accelerated, was added to the names of these standards. Finally, standards prepared in pre-inactivated matrix were marked with the letter “I”, while standards prepared in the sample solvent were marked with the letter “S”.

2.1. Tier-1 Standards

Pre-i: Pre-Extraction-Overall: A 100 μL blank plasma is spiked with 5 μL of 20-X spiking solution in 100% organic solvent to prepare a standard containing 1-X analyte concentration and 5% organic solvent. Samples are then incubated for 30 minutes* before adding one mL of ACN. Samples are then vortexed and incubated for two hours* before evaporation to dryness using a speed vacuum. Dried samples are reconstituted with 100 ul of 50% organic mobile phase in water and incubated for a total of 48 hours (8 h on the bench and 40 h in the −4 °C fridge and/or autosampler)* before injected in the LC-MS/MS system.

During-ii: ACN-1^st-Immediate: One mL ACN is added to a 100 μL blank plasma first and vortexed before spiking with 5 μl of the same 20-X spiking solution in 100% organic solvent. Standards are then immediately transferred to a speed vacuum for evaporation. Dried samples are reconstituted with 100 μL of 50% organic mobile phase in water and immediately injected in the LC-MS/MS system.

Post-ii: Post-Extraction-Immediate: 100 ul blank plasma is extracted by adding 1 mL ACN, vortexted, and evaporated to dryness using a speed vacuum. Evaporated samples are reconstituted with 100 μL of 50% organic mobile phase in water, and 95 μL of the resulting solution is spiked with 5 μL of the same 20-X spiking solution in 100% organic solvent to prepare a standard containing 1-X analyte concentration, which is immediately injected in the LC-MS/MS system.

Neat: Ninety-five μL of 50% organic mobile phase in water is spiked with 5 μL of 20-X spiking solution in 100% organic solvent to prepare a standard containing 1-X analyte concentration, vortexed, and immediately injected in the LC-MS/MS system.

* A 30-min pre-extraction incubation period with the matrix was selected because, during the preparation of standard curve samples, it will not take the analyst more than 30 min, in average, from the time all blank plasma samples are spiked with the spiking solution until ACN is added.

*A 2-h incubation period of the spiked plasma standards with ACN was selected because during the preparation of standard curve samples it will not take the analyst more than two hours to add ACN to all samples of an average batch size before proceeding to the evaporation step.

*A 48-h post-extraction incubation period was selected because it will not take more than 48 h to inject/run all samples of an average batch size after they are completely extracted and prepared. This 48 h include up to 8 hours, in average, on the bench, while samples are being handled before they are placed in the −4 °C fridge and/or autosampler.

*Longer or shorter pre-, during, and/or post-extraction periods can be selected based on the particular conditions of the analysis.

*Organic mobile phase is chosen based on the chromatography conditions.

2.2. Tier-2 [A] Standards

During-ii: ACN-1^st-Immediate: Repeated from Tier 1.

During-i: ACN-1^st-2 h: Prepared the same way as the “During-ii” standard except that incubation with ACN before evaporation is for two hours as compared to immediately transferring standards for evaporation after spiking.

Post-ii: Post-Extraction-Immediate: Repeated from Tier 1.

Post-i: Post-Extraction-48 h: Prepared the same way as the “Post-ii” standard except that standards are incubated for a total of 48 hours after extraction (8 h on the bench and 40 h in the −4 °C fridge and/or autosampler) as compared to immediately injecting in the LC-MS/MS system after spiking.

Post-i-S: Post-Extraction-48 h-Sample Solvent: Prepared the same way as the “Post-i” standard but with spiking a sample solvent (50% organic mobile phase in water) as compared to spiking the extracted plasma matrix.

Neat: Repeated from Tier 1.

2.3. Tier-2 [B] Standards

Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated: A 100 μL blank plasma is spiked with 5 μl of 20-X spiking solution in 100% organic solvent, vortexed, and one mL of ACN is immediately added. Samples are then vortexed and evaporated to dryness, within two hours, using a speed vacuum. Dried samples are reconstituted with 100 μL of 50% organic mobile phase in water and injected in the LC-MS/MS system within 48 hours.

Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix: Prepared the same way as the “Pre-ii-Acc” but using inactivated plasma instead of the plasma (active) matrix. Matrix can be inactivated by heat to stop any enzymatic reactions in biological matrices. Incubation at 65°C for 20 minutes inactivates the majority of enzymes, which usually have an optimal incubation temperature of 37°C. Enzymes that cannot be inactivated at 65°C can often be inactivated by incubation at 80°C for 20 minutes [44]. Alternatively, sodium fluoride (NaF), potassium fluoride (KF), bis(4-nitrophenyl) phosphate (BNPP) and ethylenediamine tetraacetate (EDTA) can be used to inactivate various hydrolases, while phenylmethylsulphonyl fluoride (PMSF) can be used for the inactivation of proteases [45]. A third approach to inactivate the matrix involve organic solvents such as ACN and methanol to denature enzymes [46, 47].

During-i-Acc: ACN-1^st-Accelerated: One mL ACN is added to a 100 μL blank plasma first and vortexed before spiking with 5 μL of 20-X spiking solution in 100% organic solvent. Then, standards are transferred to a speed vacuum for evaporation within two hours. Dried samples are reconstituted with 100 μL of 50% organic mobile phase in water and injected in the LC-MS/MS system within 48 hours.

During-i-Acc-I: ACN-1^st-Accelerated-Inactive Matrix: Prepared the same way as the “During-i-Acc” but using inactivated matrix.

*Accelerated conditions: At this point, since it has been established in Tier1 that there are no issues with during- and/or post-extraction stabilities, accelerated conditions with shorter incubation times up to the maximum tested incubation time frames, i.e. 2-h during- and 48-h post-extraction can be used.

2.4. Tier-2 [C] Standards

Pre-i-Acc: Pre-Extraction-30 min-Accelerated: Prepared the same way as the “Pre-i” standard except that incubations with ACN and post-extraction are within the maximum tested periods, i.e., up to two and 48 hours during- and post-extraction, respectively.

Pre-i-Acc-I: Pre-Extraction-30 min-Accelerated-Inactive Matrix: Prepared the same way as the “Pre-i-Acc” but using inactivated matrix.

Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated: Repeated from Tier-2 [B].

Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix: Repeated from Tier-2 [B].

During-i-Acc: ACN-1^st-Accelerated: Repeated from Tier-2 [B].

During-i-Acc-I: ACN-1^st-Accelerated-Inactive Matrix: Repeated from Tier-2 [B].

3. Results

3.1. Quantification of analyte losses during bioanalysis

Analyte losses can happen at any stage before, during, and after sample preparation (Table 1). Identifying the source(s) of analyte losses is required before sample preparation conditions can be optimized to increase analyte recovery. However, chasing the source of analyte losses can be a cumbersome or sometimes an impossible task due to the complexity and overlap between these sources at the various stages of sample preparation. Therefore, we proposed a scheme to systematically eliminate irrelevant sources and narrow down or pool the remaining sources until the source(s) responsible for analyte loss are identified. This scheme is divided into tiers, which prioritizes the identification and optimization of the various sources of analyte loss based on the simplicity of identifying these sources and how common they are.

First, various sets of QC standards are prepared, which differ in one source of analyte loss, or, when not possible, a group of analyte losses pooled together (Table 2). These standards are designed to exclude source(s) of analyte losses of interest one at a time. Then, various recoveries are calculated as ratios of analyte concentrations measured in these standards (Table 3). Common sources of analyte losses existing in both standards used to calculate a particular ratio/recovery are canceled out, leaving behind only the additional source(s) of losses in one standard vs. the other.

Table 3.

Examples for various sources of analyte losses identified at different stages of the protocol using standards prepared at nominal concentration of 100 ng/mL.

QC Standards		Ratios
Name	Conc	Name	Calculation	Value
Tier-1: Overall, Matrix Effect, (Evaporation & Reconstitution), & Remaining Recoveries
Pre-i: Pre-Extraction-Overall	50	Overall Recovery	(Pre-i) vs. Neat	50.0%
During-ii: ACN-1st-Immediate	93	Matrix Effect	(Post-ii) vs. Neat	95.0%
Post-ii: Post-Extraction-Immediate	95	Evap & Recon Recovery	(During-ii) vs. (Post-ii)	97.9%
Neat	100	Remaining Recovery	(Pre-i) vs. (During-ii)	53.8%
Tier-2: Break down of Individual Components of Low Remaining Recovery
[A]: During & Post-Extraction time-dependent instabilities
During-i: ACN-1st-2 h	93	During-Extraction-Time-Dependent Instability	(During-i) vs. (During-ii)	100.0%
Post-i: Post-Extraction-48 h	60	Post-Time-Dependent Instability in Matrix	(Post-i) vs. (Post-ii)	63.2%
Post-i-S: Post-Extraction-48 h-Sample Solvent	95	Post-Time-Dependent Instability in Solvent	(Post-i-S) vs. Neat	95.0%
[B]: Extraction Efficiency
During-i-Acc: ACN-1st-Accelerated	93	Combined Pre- & During-Extraction-Immediate	(Pre-ii-Acc) vs. (During-i-Acc)	77.4%
Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated	72
During-i-Acc-I: ACN-1st-Accelerated-Inactive Matrix	90	Extraction Efficiency ~ Combined Immediate Pre- & During-Extraction-Inactive	(Pre-ii-Acc-I) vs. (During-i-Acc-I)	83.3%
Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix	75

For example, the standard “Post-ii: Post-Extraction-Immediate” has only one source of analyte losses, which is matrix effect. While the analyte in the standard “During-ii: ACN-1^st-Immediate” undergoes evaporation & reconstitution in addition to the matrix effect losses. Therefore, when the ratio of analyte concentration between these two standards is calculated, the matrix effect is canceled out, while only the “Evaporation & Reconstitution” loss is retained. Therefore, we call this ratio the “Evap & Recon” recovery.

3.2. Tier-1: Overall, Matrix Effect, (Evaporation & Reconstitution), & Remaining Recoveries

Fig. 1 summarizes all standards and recoveries used in Tier-1 to identify sources of analyte losses related to overall recovery, matrix effect, evaporation and reconstitution, and all other remaining sources lumped together. Strategies to optimize bioanalytical conditions and improve recovery are also presented.

3.2.1. Overall Recovery

$$\text{Overall Recovery}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Overall}}{\mathbf{\text{Neat}}}$$

Overall recovery accounts for the sum of all losses before, during, and after samples are extracted until the time they are analyzed by the LC-MS/MS system. If overall recovery is > 85-90%, this indicates acceptable overall losses during bioanalysis. Therefore, no further optimization of any conditions is needed before proceeding to method validation.

However, if this recovery is <85-90%, the source(s) of the analyte losses must be identified to allow optimization of the relevant conditions, in an efficient and systematic manner.

3.2.2. Matrix Effect

$$\text{Matrix Effect}=\frac{\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{Post}-\text{Extraction}-\text{Immediate}}{\mathbf{\text{Neat}}}$$

Matrix effect is caused by the endogenous matrix components that are co-extracted and co-eluted with the analyte. These components more often suppress, but can also enhance, the ionization of the analytes of interest in the MS source. If matrix effect recovery is < 85-90%, bioanalysis conditions should be optimized to decrease the matrix effect to acceptable and consistent levels. Optimization strategies include (i) changing extraction conditions to exclude the interfering matrix endogenous components; (ii) further dilution of samples, pre- and/or post-extraction to decrease the concentration; therefore, the interference of these endogenous components; and (iii) changing chromatography conditions to separate these endogenous components from the analytes of interest; therefore, avoid their interference in the MS source.

If matrix effect recovery is > 85-90%, this means the source of analyte loss causing low overall recovery is due to something else, which needs to be identified.

3.2.3. Evaporation & Reconstitution

$$\text{Evap}\&\text{Recon}=\frac{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Immediate}}{\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{Post}-\text{Extraction}-\text{Immediate}}$$

Analytes often require concentration after sample extraction to increase detection limits. This is accomplished by sample volume reduction or complete evaporation using vacuum with or without heating. The sample residue is then reconstituted with a solvent often composed of a mixture of organic solvent and water or buffer, which are compatible with the LC organic and aqueous mobile phases. Analyte losses can take place during this process due to analyte degradation, nonspecific binding to the vial walls, and/or inability of the reconstitution solvent to completely dissolve the analyte in the dried sample residue.

If evaporation & reconstitution recovery is < 85-90%, bioanalysis conditions should be optimized to decrease the losses associated with this step to acceptable and consistent levels. Optimization strategies include (i) adjusting evaporation temperature and/or duration; (ii) finding alternative evaporation methods such as nitrogen blowdown; (iii) changing or treating the vial containing the sample to decrease nonspecific binding; and (iv) adjusting the composition of the reconstitution solvent to increase analyte solubility and/or overcome nonspecific binding.

If this recovery is > 85-90%, this means the source of analyte loss causing low overall recovery is due to something else, which needs to be identified.

3.2.4. Remaining Sources of Analyte Loss

$$\text{Remaining}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Overal}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Immediate}}$$

If both matrix effect and “Evap & Recon” recoveries were > 85-90% and yet, overall recovery was < 85-90%, this indicates other sources are responsible for analyte loss. At this point, these sources are lumped together into the “remaining” recovery, and their breakdown into individual components is addressed in Tier 2.

3.3. Tier-2: Break down of Individual Components of Low Remaining Recovery

In Tier 1, the contributions of matrix effect and “Evap & Recon” recoveries to the overall recovery were isolated, quantified, and minimized by optimization. If these two steps were found to be not responsible for the low overall recovery, the remaining sources of analyte losses can be divided into: [A]-During- and/or Post-Extraction time-dependent instabilities, [B] Immediate Pre- & During-Extraction losses, and [C] Biological Pre-Extraction time-dependent instability.

3.3.1. Tier-2 [A]: During- and/or Post-Extraction Time-Dependent Instabilities

Fig. 2 summarizes all standards and recoveries used in Tier-2-Stage [A] to quantify analyte losses due to time-dependent instabilities during- and/or post-extraction as well as strategies to optimize bioanalytical conditions to improve recovery.

Fig. 2.

QC strands and recovery ratios used in Tier-2-Stage [A] to quantify analyte losses due to time-dependent instabilities during-and/or post-extraction.

3.3.1.1. During ACN-Time-Dependent Instability

$$\mathbf{\text{During}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Time}}\mathbf{-}\mathbf{\text{Dependent Instability}}=\frac{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{:}\text{ACN}-1^{\text{st}}-2\text{h}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Immediate}}$$

Typically, the addition of ACN inhibits the metabolic activity of the biological matrix. However, in occasions, ACN does not completely inactivates the metabolic activity of the matrix, with some residual metabolic activity remaining, which can result in the degradation of the analyte in the presence of ACN. Analyte degradation in this case will increase over time after adding ACN until the time the sample is transferred for evaporation. Also, analyte degradation at this stage could be a chemical process independent of any matrix components. Either way,the analyte may continue to degrade during evaporation, which is accounted for, in the “Evaporation & Reconstitution” recovery calculated in Tier 1.

If analyte instability was detected in the presence of ACN (recovery < 85-90%), bioanalysis conditions should be optimized to minimize this loss. Optimization strategies include (i) speeding this extraction step by using shorter incubation time with ACN; (ii) further suppressing the residual metabolic reaction by adjust the pH; and (iii) using a different organic solvent or a completely different extraction method such as solid phase extraction.

If During-ACN recovery is > 85-90%, this means the source of analyte loss causing the low overall recovery is due to something else, which needs to be identified.

3.3.1.2. Post-Extraction-Time-Dependent Instability

$$\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Time}}\mathbf{-}\mathbf{\text{Dependent Instability in Matrix}}=\frac{\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{i}}\mathbf{:}\text{Post}-\text{Extraction}-48\text{h}}{\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{ii}}\mathbf{:}\text{Post}-\text{Extraction}-\text{Immediate}}$$

$$\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Time}}\mathbf{-}\mathbf{\text{Dependent Instability in Solvent}}=\frac{\mathbf{\text{Post}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{S}}\mathbf{:}\text{Post}-\text{Extraction}-48\text{h}-\text{Sample Solvent}}{\mathbf{\text{Neat}}}$$

Typically, the matrix is rendered metabolically inactive after ACN extraction. However, some residual metabolic activity of the matrix components may remain after extraction, which can degrade the analyte over time until the sample is injected. In addition, analyte instability post-extraction could be due to complexation with some non-enzymatic components of the matrix or could be a chemical process, completely independent of any matrix components.

Whether matrix-mediated or a chemical process, if post-extraction stability is < 85-90%, the analyte should be stabilized. Stabilization strategies include (i) Shortening the storage time between sample preparation and sample injection; (ii) decreasing storage and/or autosampler temperature; (iii) adjusting the reconstitution solvent composition and/or increasing its volume; and (iv) using a different extraction solvent or a completely different extraction method.

3.3.2. Tier-2 [B]: Extraction Efficiency

Extraction efficiency reflects the capability of ACN to strip the analyte from the matrix components. Extraction efficiency can be quantified from the immediate losses from the matrix after ACN is added. In addition, even though unlikely, these immediate losses may be due to the insolubility or fast degradation in the presence of both the matrix and ACN. Similarly, immediate losses can also take place pre-extraction, i.e., before adding ACN. This could be due to fast metabolic instability and/or irreversible binding to the matrix macromolecules. In addition, even though unlikely, nonspecific binding to the vail walls or analyte precipitation in the presence of the matrix and before adding ACN can still occur.

Immediate losses pre-extraction (before adding ACN) cannot be directly differentiated from immediate losses during-extraction (after adding ACN) because, by definition, both losses take place immediately after spiking the analyte and after adding ACN, respectively. Therefore, no standards can be designed to individually isolate the two components, and instead we lump them together. However, differentiating these two components can be accomplished indirectly using a pre-inactivated matrix as explained below.

Fig. 3 summarizes all standards and recoveries used in Tier-2-Stage [B] to quantify analyte losses due to low extraction efficiency during-extraction. Fig. 4 summarizes strategies to optimize bioanalytical conditions to improve recovery.

Fig. 3.

QC strands and recovery ratios used in Tier-2-Stage [B] to quantify analyte losses due to low extraction efficiency during-extraction.

3.3.2.1. Combined Pre- & During-Extraction-Immediate Recovery

$$\mathbf{\text{Combined Pre}}\mathbf{-}\mathbf{\&}\mathbf{\text{During}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Immediate}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Accelerated}}$$

This recovery lumps the immediate analyte losses pre- (before adding ACN) and during-extraction (after adding ACN). If this recovery is > 85-90%, this indicates that extraction efficiency is acceptable, and that the analyte does not undergo fast pre-extraction degradation by the matrix. Therefore, we can move to the last stage of Tier 2 to examine the biological stability of the analyte in the matrix. However, if this recovery is < 85-90%, the source of analyte loss should be differentiated between pre- vs. during-extraction before the relevant extraction conditions can be optimized.

3.3.2.2. Extraction Efficiency

$$\mathbf{\text{Extraction Efficiency}}\mathbf{~}\mathbf{\text{Combined Pre}}\mathbf{-}\mathbf{\&}\mathbf{\text{During}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Immediate}}\mathbf{-}\mathbf{\text{Inactive Matrix}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}-\text{Inactive Matrix}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Accelerated}-\text{Inactive Matrix}}$$

Differentiation between Pre- (before adding ACN) vs. During-Extraction (after adding ACN) can be accomplished indirectly by replacing the active matrix with a pre-inactivated matrix to suppress the pre-extraction losses and isolate the losses after adding ACN.

The assumption is that matrix inactivation will extinguish the metabolic activity of the matrix; therefore, prevent any pre-extraction losses and yet will not cause enough changes in the matrix to change extraction efficiency. Therefore, given pre-extraction losses are eliminated by using an inactive matrix, only the extraction efficiency (immediate losses after adding ACN) will retain in this recovery calculation. We will call this recovery, “extraction efficiency”.

If this recovery is < 85-90%, this indicates low extraction efficiency, which warrants optimization of the extraction conditions. Optimization strategies include (i) increasing the volume of ACN and/or adjusting its pH; (ii) switching to a different organic solvent such as MeOH; and (iii) trying a completely different approach than protein precipitation.

Similar to the active matrix, if this recovery in the inactive matrix is > 85-90%, this indicates that extraction efficiency is acceptable, and that the immediate losses are rather due to fast metabolic instability pre-extraction, which will be addressed in the next and last Tier of this protocol.

3.3.3. Tier-2 [C]: Pre-Extraction Instability

After the analyte is spiked into the matrix, it may undergo degradation mediated by matrix components over time until the reaction is terminated by the addition of ACN. This analyte loss is quantified by comparing analyte concentration at multiple time points after incubation with the matrix for a pre-determined period of time, based on the experiment requirements, to the analyte concentration at time zero. Typically, a time-zero control is prepared by first adding ACN to a blank matrix to inhibit its metabolic activity before spiking with the analyte. However, this may not be always accurate. Different time-zero controls can be prepared in a way to exclude analyte losses not related to analyte degradation, and only include degradation losses. Otherwise, losses unrelated to analyte degradation may be included in the metabolic stability calculation leading to false conclusions about the analyte’s pre-extraction stability.

Primarily, the selection of the appropriate time-zero control depends on the magnitude of the immediate analyte losses after spiking the analyte into the matrix. As explained previously, immediate analyte losses can take place before (Pre-Extraction) vs. after (During-Extraction) ACN. Immediate Pre-Extraction losses are typically due to fast metabolic degradation, irreversible binding to the matrix macromolecules, and/or analyte insolubility or nonspecific binding to the vail walls in the presence of the matrix. While immediate During-Extraction losses are typically caused by issues related to ACN extraction efficiency. Differentiation between these two components is the key for the correct calculation of pre-extraction stability.

To address this issue, we propose three types of time-zero controls to be used in the quantification of pre-extraction stability (Fig. 5). The appropriate time-zero control should have the same type of immediate pre- and during-Extraction losses as the actual time-point standards, so that the ratio of the two exclude these losses and only retain pure active-matrix-mediated instability losses.

Fig. 5.

(a) QC strands and recovery ratios used in Tier-2-Stage [C] to quantify analyte losses due to pre-extraction stability. (b) Strategies to select the appropriate time-zero control for pre-extraction recovery (biological stability) calculation.

Method-1: Analytes without immediate analyte losses pre- or during-Extraction:

$$\mathbf{\text{Combined Pre}}\mathbf{-}\mathbf{\&}\mathbf{\text{During}}\mathbf{-}\mathbf{\text{Extraction}}\mathbf{-}\mathbf{\text{Immediate}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Accelerated}}$$

Analytes with this recovery >85-90% do not have any immediate losses pre-extraction (fast metabolism, insolubility, nonspecific binding, etc.) or after adding ACN (low extraction efficiency). Therefore, the typical time-zero control, mentioned above, (During-i-Acc: ACN-1^st-Accelerated) is used, where a blank matrix is treated with ACN first followed by analyte spiking. Both the control and the actual sample have the same losses except the additional pre-extraction losses in the sample. Therefore, the ratio of the two quantifies the pure pre-extraction stability in the biological matrix.

$$\mathbf{\text{Metabolic Stability}}=\text{Total Pre}-\text{Extraction Instability}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-30\text{min}-\text{Accelerated}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Accelerated}}$$

Method-2: Analytes with immediate losses during-Extraction:

As mentioned earlier, if the above recovery is <85-90%, the contribution of pre- vs. during-extraction should be determined to be able to choose the appropriate time-zero control. Immediate losses after adding ACN can be indirectly isolated and quantified using this ratio:

$$\mathbf{\text{Extraction Efficiency}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}-\text{Inactive Matrix}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{ACN}-1^{\text{st}}-\text{Accelerated}-\text{Inactive Matrix}}$$

Analytes with this recovery < 85-90% have low extraction efficiency. Therefore, the time-zero control should also have low extraction efficiency to exclude it from the metabolic stability calculation. In this case, the “Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated” standard is used as the time-zero control, where a blank matrix is spiked with analyte first, vortexed, and then ACN is immediately added. This way immediate during-Extraction losses (low extraction efficiency) exist to the same extent in both the sample as well as the time-zero control, which will eliminate it from the metabolic stability recovery and only retain pure active-matrix-mediated instability losses.

$$\mathbf{\text{Metabolic Stability}}=\text{Pre}-\text{Extraction Time}-\text{Dependent Instability}:\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-30\text{min}-\text{Accelerated}}{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}}$$

Method-3: Analytes with immediate losses pre-Extraction:

Two pre-extraction standards are prepared in active and inactive matrices and extracted immediately after spiking. The two standards should have the same losses except the immediate pre-extraction matrix-dependent losses in the active standard. Therefore, immediate losses before adding ACN can be indirectly isolated and quantified using this ratio:

$$\mathbf{\text{Matrix}}\mathbf{-}\mathbf{\text{Dependent Immediate Degradation Pre}}\mathbf{-}\mathbf{\text{Extraction}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}}{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}-\text{Inactive Matrix}}$$

Analytes with this recovery < 85-90% undergo fast matrix-mediated degradation. Therefore, the time-zero control should not have this loss so that it retains in the metabolic stability calculation. In this case, the “Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix” standard is used as the time-zero control, where a pre-inactivated blank matrix is spiked with analyte first, vortexed, and then ACN is immediately added. This way, immediate pre-extraction loss (fast enzymatic degradation) does not exist in the time-zero control; therefore, is retained it in the pure active-matrix-mediated instability losses.

$$\text{Metabolic Stability}=\text{Uncorrected Total Pre}-\text{Extraction Instability}:\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Extraction}-30\text{min}-\text{Accelerated}}{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}-\text{Inactive Matrix}}$$

Because matrix effect may be different between the active vs. the inactive matrix, a correction factor should be applied to account for this difference.

$$\mathbf{\text{Correction Factor}}=\frac{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}:\text{ACN}-1^{\text{st}}-\text{Accelerated}-\text{Inactive Matrix}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}:\text{ACN}-1^{\text{st}}-\text{Accelerated}}$$

$$\mathbf{\text{Corrected Metabolic Stability}}=\mathbf{\text{Uncorrected Metabolic Stability}}\times \mathbf{\text{Correction Factor}}=\frac{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-30\text{min}-\text{Accelerated}}{\mathbf{\text{Pre}}\mathbf{-}\mathbf{\text{ii}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{:}\text{Pre}-\text{Extraction}-\text{Immediate}-\text{Accelerated}}\times \frac{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}\mathbf{-}\mathbf{\text{I}}:\text{ACN}-1^{\text{st}}-\text{Accelerated}-\text{Inactive Matrix}}{\mathbf{\text{During}}\mathbf{-}\mathbf{\text{i}}\mathbf{-}\mathbf{\text{Acc}}:\text{ACN}-1^{\text{st}}-\text{Accelerated}}$$

4. Discussion & Examples

Because of the very complex nature of many biological samples, efficient sample preparation is required to remove unwanted components and to selectively extract the compounds of interest. The FDA “Bioanalytical Method Validation” guideline requires recovery to be measured to ensure sample extraction is efficient and reproducible [2]. According to the guideline, recovery is determined by “comparing the analytical results of extracted samples with corresponding extracts of blanks spiked with the analyte post-extraction (i.e., to represent 100 percent recovery)”. Low overall analyte recovery is the net result of losses that can happen for multiple reasons at all stages of sample preparation and analysis pre-, during-, and post-extraction. Therefore, identifying the source(s) of analyte loss is required before sample preparation and analysis conditions can be optimized to minimize these losses.

Herein, we proposed a protocol to identify, isolate, and quantify the sources of analyte losses followed by strategies to correct the underlying causes to improve recovery. The protocol was designed in a top-down approach composed of tiers, to maximize productivity and avoid unnecessary testing. We have used protein precipitation using ACN followed by evaporation and reconstitution as a representative example of plasma sample extraction methods, but the protocol can be similarly applied to other extraction methods. Different sets of standards were prepared to allow the calculation of recoveries of interest at the various stages.

In Tier-1, overall recovery is determined followed by the exclusion or optimization of matrix effect and evaporation and reconstitution analyte losses. Any other remaining sources of analyte losses are lumped together followed by identification of individual components in Tier-2. The first stage of Tier-2 aims to exclude or optimizes analyte losses due to the instability of analyte during- and/or post-extraction. In the second stage, losses due to low extraction efficiency are excluded or optimized, followed by the third and last stage, which includes various approaches to quantify any pre-extraction instability.

Table 3 demonstrates the results of the application of this protocol in our laboratory to the extraction of analytes with different sources of analyte losses. In Tier 1, the analyte had low overall analyte recovery of 50%. Matrix effect and losses due evaporation and reconstitution were excluded. Therefore, unnecessary efforts to optimize conditions related to these sources of analyte loss were avoided. The analyte was then moved to the first stage of Tier-2 to identify the reason for the 53.8% recovery of the remaining sources of analyte losses using the same extraction method. We found that the analyte was not stable post-extraction and only 63% was recovered after the sample was extracted until it was injected in the LC-MS/MS system. Sample extraction conditions were then optimized as described earlier to stabilize the analyte after sample extraction. For this analyte, we found that adding 0.1% formic acid to the 1:1 ACN:H₂O reconstitution solvent improved post-extraction stability and overall recovery.

In another example, the cause of low overall recovery from Tier-1 was identified in the second stage of Tier-2 (Tier-2 [B]) as low extraction efficiency by ACN (83.3%). Accordingly, we tried a different extraction solvent and found that MeOH was more efficient in extracting the analyte from the matrix and improved the overall recovery.

Table 4 compares three methods to calculate the pre-extraction (metabolic/biological) stability of three analytes using different time-zero controls. Analyte-1 did not have any significant immediate losses pre-extraction (fast metabolism, insolubility, nonspecific binding, etc.) or after adding ACN (low extraction efficiency). Therefore, the typical time-zero control, (During-i-Acc) was used, where a blank matrix is treated with ACN first followed by analyte spiking. Both the control and the actual sample have the same losses except the additional pre-extraction losses in the sample. Therefore, the ratio of the two standards quantifies the pure pre-extraction stability in the biological matrix (95%).

Table 4.

Examples for the application of the three different methods for metabolic/biological/pre-extraction stability.

	Analyte-1	Analyte-2*	Analyte-3**
QC Standards	Concentration
Pre-i-Acc: Pre-Extraction-30 min-Accelerated	95	75	64.6
Pre-i-Acc-I: Pre-Extraction-30 min-Accelerated-Inactive Matrix	96	79	80
Pre-ii-Acc: Pre-Extraction-Immediate-Accelerated	96	76	76
Pre-ii-Acc-I: Pre-Extraction-Immediate-Accelerated-Inactive Matrix	94	79	87
During-i-Acc: ACN-1st-Accelerated	100	100	100
During-i-Acc-I: ACN-1st-Accelerated-Inactive Matrix	95	95	95
Ratios Used to determine the proper Time-zero Control
Combined Pre- & During-Extraction-Immediate = (Pre-ii-Acc) vs. (During-i-Acc)	96.0%	76.0%	76.0%
Extraction Efficiency =(Pre-ii-Acc-I) vs. (During-i-Acc-I)	98.9%	83.2%	91.6%
Matrix-Dependent Immediate Degradation Pre-Extraction = (Pre-ii-Acc) vs. (Pre-ii-Acc-I)	102.1%	96.2%	87.4%
Metabolic Stability Calculations
Method 1-Metabolic Stability = (Pre-i-Acc) vs. (During-i-Acc)	95.0%	75.0%	64.6%
Method 2-Metabolic Stability = (Pre-i-Acc) vs. (Pre-ii-Acc)	99.0%	94.9%	80.8%
Method 3-Metabolic Stability
Uncorrected Metabolic Stability = Pre-i-Acc) vs. (Pre-ii-Acc-I)	101.1%	94.9%	74.3%
Correction Factor=(During-i-Acc-I) vs. (During-i-Acc)	95.0%	95.0%	95.0%
Corrected Metabolic Stability=Uncorrected Metabolic Stability × Correction Factor	96.0%	90.2%	70.5%

^*Example on Analayte-2 is an experimental long-acting lipophilic fatty acid ester prodrug of Cabotegravir (M2CAB) [48].

^**Example on Analayte-3 is anticancer (breast cancer gene 1 (BRCT-BRCA1) inhibitor) experimental compound “BI-94: (4-(benzylsulfonyl)-7-(hydroxy(oxido)amino)-2,1,3-benzoxadiazole) [49].

In contrast, both analytes-2 and -3 had low combined immediate pre- & during-Extraction recovery (76.0%), but for two different reasons. Analyte-2 had relatively low extraction efficiency (83.2%), which could not be further improved, while analyte-3 underwent immediate degradation in the active matrix (87.4%). For analyte-2, the time-zero control should also include the same losses so that they are excluded from the actual incubation sample vs. time-zero concentrations ratio, and only pure pre-extraction stability losses are retained. In this case, the time-zero control (Pre-ii-Acc) is prepared by spiking the analyte first before immediately adding ACN so that the same analyte losses due to low ACN extraction efficiency take place in the time-zero control just like the actual time-point sample. This way, the analyte was identified to be actually metabolically stable (94.9%) using Method-2 rather than metabolically unstable (75.0%) using Method-1. In method-1, the typical time-zero control, (During-i-Acc) bypasses the low ACN extraction efficiency because the analyte is spiked after the sample is treated with ACN. Method-1 falsely included low extraction efficiency losses in the calculation, which leads to the wrong conclusion that the analyte is metabolically unstable in this matrix. This could trigger futile efforts to improve the overall recovery by trying to stabilize the analyte in the matrix pre-extraction, while the problem lies somewhere else due to low extraction efficiency during ACN extraction.

Analyte-3 had significant (87.4%) immediate losses pre-extraction (fast matrix-mediated degradation). Therefore, the time-zero control should not have this loss so that it retains in the metabolic stability calculation. In this case, the (Pre-ii-Acc-I) standard is used (Method-3) as the time-zero control, where a pre-inactivated blank matrix is spiked with analyte first and then ACN is immediately added. This way, immediate pre-extraction loss (fast enzymatic degradation) does not exist in the time-zero control; therefore, is retained it in the pure active-matrix-mediated instability calculation (74.3%). Differences in matrix between the active vs. the inactive matrix, if any, are then corrected for using the correction factor (95.0%), which brought the corrected metabolic stability to 70.5%. Method-1 would have exaggerated the metabolic instability of this analyte because it would have included the irrelevant extraction efficiency losses in the calculation. While method-2 would have underestimated the metabolic instability because the (Pre-ii-Acc) standard does undergo the same immediate losses pre-extraction as the sample. Therefore, the ratio of the two would have excluded this loss, when it should have been included as a part of the metabolic stability.

5. Summary

High and reproducible recovery of analytes from matrices is essential for valid bioanalytical methods designed for the quantification of drugs in biological matrices. Overall extraction recovery calculation detects the problem of low recovery but does not identify the sources of analyte losses during sample preparation. In this article we proposed a practical protocol to efficiently identify and quantify the sources of low analyte recovery. This allows the proper choice of strategies to optimize the relevant bioanalytical conditions to minimize analyte losses and improve overall recovery.

Highlights.

1. High and reproducible recovery of analytes is essential for the development of valid bioanalytical methods for the quantification of drugs in biological matrices.

2. We have developed a protocol to identify and quantify the various sources of analyte losses at all stages of sample collection, preparation, LC-MS/MS analysis, and storage before and during analysis.

3. Breaking down the individual components of overall recovery is essential to identify where analyte losses took place before bioanalytical recovery can be improved.

4. This allows the proper choice of strategies to optimize the relevant bioanalytical conditions to minimize analyte losses and improve overall recovery.