Development and validation of small molecule analytes by liquid chromatography-tandem mass spectrometry

Background

The application of liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) represents one of the most compelling opportunities for advancements in human health through the combination of reference measurement procedure capabilities, broad chemical coverage, and a rich history in support of drug development from the 1990s onward. Clinical application of these technologies has begun to gather pace in many laboratories, with diverse applications ranging from expanded newborn screening to identification of emerging toxicants. The promise of these technologies is vast and the need is palpable; however, the journey can be exacting. Perhaps somewhat unique among analytical techniques, LC-MS/MS assay development and validation requires significant knowledge of a number of specialties: a mastery of chemistry (sample preparation, chromatography, ionization), physics (ion manipulation), engineering principles (automation, order of experiments, programming), and mathematics (data reduction and interpretation) applied to questions of a biological origin (normal, disease, metabolism).

Content

This chapter provides a stepwise roadmap for systematically developing and validating an LC-MS/MS assay for small molecule analytes. Starting from first principles (i.e., salt correction in gravimetric weighing), each component of the LC-MS/MS assay (mass spectrometer tuning, ionization enhancements, chromatography, extraction) is detailed with best-practice experiments for development and data reduction techniques to fulfill performance goals. After refinement of each component of the assay, prevalidation experiments are described to enable efficient execution of validation. Finally, an array of validation guidance documents is reduced to a coherent process for burden of proof.

The analysis of small molecule analytes (atomic mass < 1000 amu) using mass spectrometric techniques represents the oldest and most diverse application of the technology within the clinical field. Forensic and clinical toxicology assays were published and deployed in the 1970s using gas chromatography–mass spectrometry (GC-MS) ^, ; notably, many of the same analytes are still measured by GC-MS today.

The development of atmospheric pressure ionization interfaces (see Chapter 20 ) enabled the genesis of metabolomics workflows and the application of newborn screening using blood spots, which represents the next real explosion in clinical chemistry from the early 1990s. Continuing advances in mass spectrometer sensitivity enabled the application of liquid chromatography–tandem mass spectrometry (LC-MS/MS) for targeted metabolite analysis such as the measurement of steroid hormones and biogenic amines in the early 2000s. During the same time frame, therapeutic drug monitoring applications became de rigueur, with analysis of immunosuppressants driving uptake of the technology in the hospital setting to enable preoperative and postoperative monitoring.

Advancements in high-performance liquid chromatography (HPLC) hardware (late 2000s) and mandates for measurement of drugs of abuse has led to highly multiplexed measurement of drug classes such as benzodiazepines and opiates within a single injection from a common sample preparation protocol in urine. ^, Combinations of HPLC, GC, and ionization mode, often with accurate mass time-of-flight technology, are currently being extensively used in untargeted metabolomics studies to generate new combinations of metabolites that may uncover disease phenotypes with the promise of precision medicine. This avenue of potential new clinical assays bodes well for the advancement of MS technologies in small molecule analysis, and the industry is poised to translate these into clinical practice. Finally, the real promise of MS technologies in providing confidence in measurement is being realized through a number of international standardization and harmonization efforts involving the generation of certified reference materials and reference method procedures. ^,

The use of isotope dilution and external calibration curves enables exceedingly low error and imprecision in the accurate measurement of analytes within complex mixtures and provides an opportunity for cross-platform harmonization and improved patient care.

As noted earlier, many potential technologies can be applied to small molecule analysis in clinical practice; however, the most contemporary application is highlighted in Fig. 23.1 and will be the predominant foundation for discussion within this chapter. The goals of this chapter are to provide guidance for systematically developing and validating an LC-MS/MS assay from first principles to enable clinically relevant measurement in the support of patient care. There have been a number of well-informed examples of when MS assays fail. The tools and considerations for determining and addressing such challenges are covered in this chapter. We have taken the liberty of describing, to our knowledge, the best practices extant in the literature and from our own experiences to underpin these processes and concepts here.

The LC-MS/MS development process is often a very personal journey with bias toward certain known and mastered technologies. In our experience, it is often beneficial to allow the analytes to dictate their own optimal conditions, determined through highly parallel experiments in selection of mobile phase, stationary phase, column dimensions, and sample preparation. For completely new analytes or classes, it is recommended that development flows from the mass spectrometer back toward the sample preparation in a systematic manner to evaluate optimal conditions and a compendium of solutions as go-to next steps in the event of developmental failure. Burden of proof mandates a thorough evaluation of method performance through prescriptive validation experiments with predefined acceptance criteria. Currently there is limited guidance that describes in detail the “what to do and how to do it” for clinical assay validation. ^, This chapter synthesizes a number of validation guidance documents into a coherent process for consideration. Notably, these processes are adhered to for each assay in our laboratories to provide high-quality results.

Calibration standard and matrix

The foundation of LC-MS/MS assays relies on well-characterized standard materials for generation of a calibration curve of a known analyte quantity. The definition of well-characterized is as follows: the structure and molecular composition of a standard material should be clearly listed, with any salt of crystallization described in detail. When dissolved into solution, the standard will dissociate into the free acid-base form; thus the salt content must be accounted for in concentration assignment. Further, the composition of all other impurities must be considered and subtracted from the final assignment of stock materials concentration.

Practically, the numbers of reference standards are limited relative to the clinical measurement landscape. Thus particular attention should be paid to the certificate of analysis provided with the standard materials. At a minimum, additional details, including the determination of purity of analysis by HPLC, should be known and requested as a chromatogram, together with water content through Karl-Fischer analysis. In the absence of these documents, it is recommended that two individual sources of standard materials be acquired for developmental and validation efforts. Note, determination that the alternative source of standard is of a different provenance is critical, as alternative vendors may re-label standards from a common preparation source. Finally, storage conditions should be clearly listed (and followed) in standard operating procedures (SOPs), together with expiry date (proved and adhered to) under that condition. Reassessment of standard materials purity is recommended where materials are repeatedly accessed, because of the hygroscopic nature of most standards, unless particular care is taken to control for this eventuality (i.e., storage under argon).

These standard materials are added using gravimetric techniques to an analyte-free matrix representative of the specimen type being assayed. Best manufacturing practice of calibrators requires the overall organic composition to be less than 5% by volume. Further, the hierarchy of calibration matrices clearly stipulates that the calibration matrix should be identical to the specimen being measured; one could argue that this is true even to the choice of counter-ion (Li versus Na heparin). Superficially this would appear to be facile, certainly for exogenous drug assays in blood-based matrices; however, two particular types of clinical assay are confounding. Firstly, endogenous analyte assays (e.g., steroid hormones) cannot be prepared in a true blank matrix. Although various quantitative approaches may be used in small molecule analysis, such as standard addition or internal standard (IS) signal matching, more commonly a surrogate matrix is used. Typically, this matrix is charcoal stripped or manufactured from standard chemicals to be physiologically similar to the patient sample (isotonic and containing protein for cerebrospinal fluid [CSF] or synthetic urine). Second, the current trend toward high-density LC-MS/MS assays for drugs of abuse spanning several orders of magnitude in measurement interval (analytical measurement range [AMR]) precludes the ability to combine stocks without failing the aforementioned 5% solvent rule. Thus solvent-based calibration systems are often used, although solvent-based calibration is the lowest desirable calibrator matrix defined in guidance. As will be described later, defined validation experiments demonstrating the comparable nature of calibrators and patient specimens is required for confidence in measurement, particularly in these situations.

Internal standard selection and utility

The “killer-app” of LC-MS/MS assays is undoubtedly the intrinsic ability to control measurement error and imprecision through the use of isotopically labeled ISs. Optimal ISs are a stable (nonradioactive) labeled version of the analytes and thus a physicochemical mimic, usually enriched with deuterons ( ² H), nitrogen ( ¹⁵ N), or carbon ( ¹³ C). Alternatively, an analog IS structurally dissimilar to the analyte of interest may be used. For quantitative diagnostic applications of MS, analog ISs are strongly discouraged. The selection of an IS is often constrained by availability and cost; however, if at all possible, a stable labeled version of the analytes should be used in clinical assays, with preference given to ¹³ C- and ¹⁵ N-labeled forms.

The nature of enrichment should follow four basic rules ( Table 23.1 ); first, ensure incorporation of at least three heavy atoms, unless the analyte contains chlorine or bromine. Ideal ISs for these analytes contain ¹³ C ₅ or ² H ₅ as a minimum as the +2 isotopes of chlorine and bromine in combination with naturally occurring ¹³ C in the analyte may contribute to the measurement of the IS. Second, the IS should not contain excess deuterons (more than ² H ₆ , often generated from historical GC-MS workflows) because this leads to chromatographic resolution of analyte and IS. Use of these ISs requires additional sample preparation constraints and validation studies to confirm that the IS controls for matrix effects. Third, the IS material should not contain excessive unlabeled analyte; this reduces the amount of IS that can be added without contribution to the analyte, leading to assay imprecision and inaccuracy. Finally, interrogation of the position of deuterons in the IS is necessary. De-deuteration can occur in sample preparation (acid hydrolysis, solid-phase extraction [SPE]) and the LC-MS interface, resulting in excessive variability in measurement.

Practical use of an IS is both quantitative (isotope dilution) and qualitative (recovery, matrix effects, chromatographic variance, and mass spectrometer sensitivity drift). For quantitative use of the IS (see Fig. 23.1 , step 2), the IS is added to all samples except double-blank samples (no analyte or IS) at the same concentration and at a reproducible volume greater than 25 μL unless high-quality robotics are used. The IS should account for extraction, injection, and ionization variance throughout the LC-MS/MS experiment (analog IS or chemically similar/nonisotopically labeled compounds used as the IS usually fail these criteria) and must be mass resolved as a discrete transition (s).

The calibration curve is generated by determining the concentration ratio of analyte to IS on the x -axis, albeit the IS value is usually set to 1, so the x -axis describes the concentration of analyte. Notably, only nonzero calibrators are used. The variable y -axis is defined as the integrated peak area ratio of analyte divided by IS. Assignment of unknown samples is thus extrapolated by measuring the peak area ratio (y) of analyte to IS and solving for the analyte concentration using a generic calibration curve equation ( x = y /slope − intercept).

Although often underappreciated, the qualitative use of ISs is highly informative. Analysis of IS peak area trends across a batch is used to monitor under-recovery, pipetting errors in preparation, or matrix effects (IS response between 50 and 150% of calibrator/quality control [QC] mean, predicating a decision for repeat sample analysis) or determining deleterious instrument drift resulting in corrective action (instrument clean). An appropriate IS may be used as a surrogate for the analyte in recovery studies in sample preparation development or, perhaps more importantly, to define the absolute retention properties of the analyte, informing the data reviewer of analyte peak shape and retention time.

When considering the nature and number of ISs to use in a high-density multi-analyte LC-MS/MS experiment (>15 compounds), the answer is rather trite—as many as possible and isotopically labeled appropriately. In appraising the value of the IS and the almost unique ability of MS to incorporate external calibration and internal standardization, one should aspire to leverage these strengths in LC-MS/MS assays deployed in the care of patients.

Mass spectrometry source setup and tuning

The first step in the practical development process involves interrogation of the analyte of measure, specifically the potential to ionize functional groups in solution. Consideration should be given to the upstream process constraints that the interface modality engenders ( Table 23.2 ; Chapter 20 ). Analytes containing amines or amide bond(s) will commonly form positive pseudomolecular protonated ions (biogenic amines), whereas analytes containing carboxylic acids (e.g., free fatty acids) or an electronegative ion (e.g., thyroxine and carboxylic acid moieties) will often form negative pseudomolecular deprotonated ions. Analytes containing these motifs are candidates for electrospray ionization (ESI). Analytes that are neutral in solution require the inclusion of adduct ions for electrospray or require gas phase ionization, as provided by atmospheric pressure chemical ionization (APCI) interfaces. Small molecules almost exclusively form singly charged ions; thus knowledge of the molecular weight of the analyte and ionization mode provides a good starting point to predict the expected precursor ion mass-to-charge ratio (m/z).

Two instrument configurations are shown in Fig. 23.2 , postcolumn infusion (PCI; A) and flow injection analysis (FIA; B). The PCI setup will be used for other developmental efforts, most notably matrix effect determination, ^, and is a valuable technique for any method development process. The PCI setup involves delivery of a mobile phase to a mixing tee (T), where a concentrated solution of analyte(s) at approximately 1 μg/mL in mobile phase (or methanol) is added at 5 to 10 μL/min.

The mixed eluent flow is introduced into the interface for rapid determination of ionization mode (ESI/APCI, positive or negative ion mode) and source parameters under conditions of solvent load. Initial determination of precursor ion(s) may be performed without the makeup flow from the chromatographic pumps for ESI to reduce chemical noise. ACPI requires solvent flow (>200 μL/min) to stabilize discharge current ( Chapter 20 and Table 23.2 ). As the analyte is predetermined, the expected molecular ion’s m/z ratio can be quickly calculated. Pseudomolecular ions will have an expected m/z of the analytes of molecular weight +1 for positive ionization mode and molecular weight −1 for negative ion mode.

The FIA setup differs by the removal of the HPLC column and T, incorporating injection of analyte solution prepared in a mobile phase to enable signal to noise (S/N)-based optimization or to remove the influence of solvent ions from the determination of precursor ions. PCI evaluation using ESI in positive ion mode for ² H ₄ -hydroxymidazolam is shown in Fig. 23.3 ; notably, the proposed precursor ion ( m/z = 346) is difficult to discern in this example. Modification of the default interface parameters (heat, orifice voltage, gas flow, etc.) is often required to generate appropriate selection of the precursor ion. Alternatively, shutting off the HPLC solvent flow (infusion pump only) or performing a secondary infusion of solvent only may enable precursor ions to be identified and differentiated from solvent ions by comparison of mass spectra.

Interface ion chemistry may be confounding ( Table 23.3 ). Scanning the first quadrupole at ±80 amu of the expected intact precursor ion ([M+H] ⁺ or [M−H] ⁻ for positive or negative ionization, respectively) to observe the formation of adducts (i.e., ammonium for immunosuppressants) or thermal losses (i.e., water losses for steroids) is recommended, as is scanning slowly (1 second per 100 amu scan range). As a last resort, increase the concentration of the infusion solution or, alternatively, use FIA, which facilitates background subtraction of interfering ions, a functionality for all MS software systems, in order to ascertain the appropriate precursor ion characteristics. The influence of interface temperature on ion population can be performed in PCI mode, allowing 1 minute of equilibration for each interface temperature step of 100 °C before determining precursor ions formed.

After confirmation of precursor ion(s), determination of product ions from each individual precursor is performed in PCI mode unless confounded by solvent ions (FIA mode). A precursor ion is selected to traverse the first quadrupole and accelerated into a collision gas in the second quadrupole with a collision energy (CE) to induce collisional activated dissociation, generating product ions. The product ion population is then determined by scanning across a mass range with the third quadrupole. As noted earlier, scanning is performed slowly from m/z of 50 to precursor + 50 amu (1 second/100 amu) and the CE are stepped while summing the product ion spectra or the CE is ramped from 0 to 100 eV, noting the product ions formed as the CE changes. Fig. 23.4 shows product ions for ² H ₄ -hydroxymidazolam at 20, 40, and 60 eV, together with an extracted ion chromatogram for four precursor-to-product ion pairs (transitions).

Two things should be considered when selecting transitions. First, facile neutral losses of water, ammonia, and carbon dioxide (see Fig. 23.4A , low CE) should be avoided if at all possible. Transitions incorporating these losses are ubiquitous and thus not selective. Using these losses places a significant burden on chromatography and sample preparation to ensure measurement specificity of the LC-MS/MS method. Secondly, low-mass product ions generated at high CE (see Fig. 23.4C , high CE > 60 eV) retain limited structural content relative to the precursor ion and may have limited structural specificity (putative product ions of different analytes). Product ions incorporating 30 to 70% of the precursor ion mass are desirable to balance these two factors if observed (see Fig. 23.4B ).

Performing product ion CE ramp scans to generate a total ion current (TIC) plot and extracting product ions facilitates the generation of CE versus yield plots (CE voltage versus intensity as counts per scan; see Fig. 23.4D ). Optimal CEs are readily interpolated. In the absence of multiple product ions, maintaining the same neutral loss but with a discrete CE is recommended for transition ratio analysis. The CE offset should be sufficient to generate alternative transitions at 50 to 75% yield (ideally n = 4 transitions). A number of clinically relevant analyte measures follow this concept, notably tramadol and cortisol, by offsetting the precursor and/or product ions by 0.001 to 0.1 amu (for unit mass resolving instruments to generate a distinct “measured” transition) and selecting a CE for a readily observable difference. In the example shown, a second transition for ² H ₄ -hydroxymidazolam at m/z 346.1 to 203.1 and CE approximately 25 eV would yield a 1:2 response ratio versus optimal CE (∼36 eV). Initially, more than four transitions per analyte should be taken forward to further development with the final goal of two or more transitions per analyte in clinical sample analysis (transition ratio monitoring) to enable purity assignment and therefore confidence in the quality of reported results.

Analyzing as many transitions as possible is a good practice; redundant or unacceptable transitions will be systematically removed during HPLC and sample preparation development. Finally, the product ions generated for the unlabeled analyte ideally should be structurally identical to those selected for the matched stable isotope-labeled IS with the same CE voltage for each transition. As long as the precursor ion is differentiated between analyte and labeled IS, product ions can be identical because most modern mass spectrometers have eradicated crosstalk (the detection of precursor ions retained in the collision cell from a previously analyzed molecule, signal from analyte in the IS transition, or vice versa). An example would be the measurement of testosterone and its ³ C ₁₃ IS. The position of labeling of C ₁₃ atoms is on the A ring and thus is retained in the predominant product ions formed. ^, Selected transitions are generally m/z 289 to 97 or 289 to 109 for testosterone and m/z 292 to 100 or 292 to 112 for the IS with CEs of approximately 35 and approximately 38 eV for each matched pair of transitions.

An initial determination of acceptable interface conditions should be briefly determined. This can be performed in either PCI mode or by FIA. Fig. 23.5A shows the modification of declustering potential, gas flows, and ion spray voltages modified during PCI. Alternatively, methods can be built and the provisional source settings assessed by FIA (see Fig. 23.5B ). These are not final settings; without a sufficient chromatographic separation, noise or artificial signal cannot be discriminated from the signal generated solely by the analyte of interest.

Solvent chemistry optimization

After selection of preliminary interface conditions and relevant transitions, the next step in the development process involves determination of the appropriate volatile solvent chemistry to introduce the analytes into the interface. Using a four-solvent (quaternary) pump and the PCI modality from Fig. 23.2 , a programmed solvent sequence can be used to modify solvent chemistries and evaluate their influence on sensitivity in selected reaction monitoring mode (to reduce the observed influence of chemical noise). Fig. 23.6A demonstrates simple programming of solvent changes involving acetonitrile, water, methanol, and 100 mmol/L ammonium formate for the measurement of reverse triiodothyronine (rT ₃ ) in negative ion ESI mode. Simple observations of response changes enable selection of potential HPLC solvent chemistries. In the example shown, rT ₃ demonstrates improved response for methanol over acetonitrile and either no or minimal counter ion (formate). Any variety of solvents, buffers, or other mobile phase additives can be used in this screening mode.

Alternatively, Fig. 23.6B and C demonstrate FIA-based injections of atenolol in positive ESI mode whereby atenolol is spiked at the same amount into individual wells containing mixtures of solvents, buffers, acids, or bases, with the same final liquid volume in each independent well and thus the same concentration of atenolol in each well. Each well contains a unique combination of solvent chemistries and the influence of solvent chemistry on ionization is determined as the peak height of each injection (there is no column and thus limited mixing between the injected solution and carrier solvent). As shown in Fig. 23.6 , a 10-fold difference in response is observed when atenolol is injected in methanol and trifluoroacetic acid versus acetonitrile and formic acid when both solvents have the same composition of ammonium formate. This experimental setup enables faster screening of a broader array of potential solvent chemistries than repeatedly changing and priming new solvents in LC-PCI mode. In this manner, 96 individual samples are prepared and assayed and the data are reduced in approximately 3 hours in the previous reference. Finally, assessment of optimal chemistries in FIA mode can be used to determine which analyte has the lowest response (ionization and MS/MS transmission efficiency) relative to the desired lower limit of measurement interval (LLMI, otherwise called the lower limit of quantification [LLOQ]). When developing methods with multiple analytes, it is best to focus on optimizing sensitivity for the least sensitive analyte in terms of on-column detection requirements (analytical sensitivity) and circulating levels required to measure for clinical care, if known.

Liquid chromatography development

The true complexity of clinical sample analysis with LC-MS/MS technologies arises in the interplay between selectivity and sensitivity. The next challenge is to evaluate and optimize the chemical purification methods of HPLC and sample preparation independently and then in combination, together with further streamlining of MS/MS transitions for validation. The scale and combinations of technologies that could be used is staggering; however, the techniques to optimize are predictable.

There is no singular aspect of the LC-MS/MS method that has more myth and mystery associated with it than the HPLC step ( Table 23.4 ). Certain myths are that HPLC separations are robust (HPLC systems leak, usually just after you have left the laboratory), columns should last longer than 1000 injections (after 500 the column has paid its way, it is a consumable, albeit an expensive one), smaller particles lead to better efficiency (they also lead to higher operating pressure), and isocratic separations are faster (theoretically, although at the possible costs of retention time variations and strongly bound materials eluting a number of samples later, inducing false-positive results or ion suppression). Chapter 19 incorporates significant detail on the principles of HPLC; the practicality of development is covered here.

The first step in the HPLC experiment is interrogation of the molecular features of the analytes. Table 23.5 provides details for consideration of the modality of HPLC separation based on analyte polarity (charge) and knowledge from the previous solvent chemistry screen. Many highly charged small molecule analytes are amenable to reverse-phase HPLC (RPLC), the separation technique that is most understood with the broadest range of commercially available stationary phase chemistries. The inclusion of ion pairing reagents to effect retention control or the retentive capacity (k′) of the earliest eluting analytes can be helpful for the most polar of small molecules chromatographed in reverse-phase mode. Although significantly less understood, hydrophilic interaction liquid chromatography (HILIC) is an excellent separation modality for clinically relevant metabolites, and a paper by the authors provides further guidance on the empirical nature of the HILIC development process. For analytes for which the optimal solvent for ionization is acetonitrile and the analytes are highly polar (amino acids, nicotine and its metabolite, cotinine, etc.), an exploration of HILIC may be desirable. For the sake of brevity, this chapter shall focus on RPLC developmental processes and its functional use, although many experimental considerations hold true for HILIC.

The primary goals for HPLC separation are to resolve analytes from interferences and ion suppression–inducing compounds (simplifying the composition introduced into the interface) and facilitate reproducible retention properties (retention time, baseline noise, and peak shape) among calibrators, QCs, and specimens. The first step in the developmental process is to explore classical chromatographic parameters (selectivity, retentive capacity, and peak shape) by screening combinations of stationary phases and the preferential solvent compositions determined previously. The injection solution should be compatible with RPLC (50 to 100% methanol, substitute with acetonitrile for HILIC) and provide an S/N ratio greater than 100, usually generated at 100× the assay LLMI with a 10-μL injection. The use of high amounts of methanol in the injection solvent may be counterintuitive. However, the reasoning is that adsorptive loss of compounds to the walls of the injection vessel often can be mediated with organic solvent and a column that can properly retain the analytes of interest in a high organic environment typically denotes good retentive capacity. Two additional samples are injected, a blank diluent sample (always run a negative control in all experiments) and, if the assay is to be performed in a blood-based matrix, protein precipitated (PPT) matrix at a minimum 3:1 solvent-to-sample ratio with IS included in the precipitating solvent, to evaluate matrix-induced retention variance and response changes. The inclusion of the blank sample ensures chromatographic system cleanliness (absence of analytes when screening new columns); the precipitated sample is used to monitor the retention properties of known interfering species, namely phospholipids, and to assess the variation in retention time of the analyte in a complex sample.

Phospholipids are well known for the ability to modify ionization cross-section of the analyte in ESI. To identify the retention properties of phospholipids, additional transitions are added in positive ion mode for two lysoglycerophosphatidyl cholines (1-P-2-OH: 1-palmitoyl-2-OH-sn-glycero-phosphocholine [18:2], m/z 496 to 184 and 1-S-2-OH: 1-stearoyl-2-OH-sn-glycero-phosphocholine [18:0], m/z 524 to 184) and two glycerophosphatidyl cholines (1-La-2-Li: 1-lauroyl-2-linoleoyl-sn-glycero-phosphocholine [12:0, 18:2], m/z 702 to 184 and 1-Li-2-A: 1-linoleoyl-2-arachidonoyl-sn-glycero-phosphocholine [18:2, 20:4] m/z 806 to 184), all using CEs of approximately 10 eV. These phospholipids comprise early and late eluting forms of the abundant matrix effect–causing species for both classes and will require resolution through either chromatographic separation or sample preparation. ^, When analytes are being assayed in negative ion mode, a separate acquisition of the precipitated sample in positive ion mode (monitoring phospholipids) is recommended to determine the elution properties of phospholipids relative to the analytes (or use polarity switching) as additional information during the screen. The measurement of phospholipids is unnecessary for urine-based assays; salts and metabolites are the primary matrix effect concerns for urine.

If research has indicated the probability of circulating isomers of the analyte of interest, that material should be sourced if available and included in the test solutions described herein. To discriminate the interfering peak from the analyte of interest, inclusion of the IS provides the expected retention time of the analyte. In the presence of a single isobar, three peaks should be observed for an appropriate HPLC separation. Two peaks that co-elute represent the analyte and its isotopically labeled IS; the third (isomer) is fully resolved from the former two.

It is recommended to screen different stationary phases using 50- × 2-mm columns with 5-μm particles because of the many different phases available. The goal is to screen the chromatographic characteristics of the phase and not change particle size or column dimensions (additional variables), which could confound conclusions. Fig. 23.7A indicates an example screening gradient; note that HPLC system dead-volume, maximum HPLC operating pressure (<80% pressure maximum recommended), and maximum interface eluent flow compatibility should be considered when setting the time or flow rate for the four steps in the gradient screen. The scouting gradient at 1 mL/min comprises injection focus at 100% A (30 seconds, 1.5 system volumes), gradient separation to 100% B (100 seconds), column clean at 100% B (50 seconds, 2.5 system volumes), and column reconditioning to 100% A (60 seconds, 3 system volumes) over 4 minutes with data acquisition for 3 minutes initiated when the gradient starts (30 seconds).

Modern HPLC systems have significantly reduced dead volumes (<150 μL); however, large injector loop volumes (>20 μL) and poorly made tubing fittings will add unwanted gradient delay and should be avoided and corrected. Further, it is good practice to divert solvent away from the ion source before and after the acquisition window to maintain interface and MS ion optics cleanliness; the first and last 30 seconds are diverted to waste during the scouting gradient at the flow rates listed. Screening multiple columns requires variety for serendipity; recommended RPLC screening functionalities are C18 (two manufacturers), C8, pentafluoro phenyl (PFP), phenyl hexyl, biphenyl, cyano, and amide, among others.

Further consideration of the gradient is warranted; the 1% per second change in elution solvent over 100 seconds provides an approximation to the elution composition when analytes elute. When a peak elutes at 50 seconds, the pump is delivering 50% B; however, analyte retention time is affected by gradient delay and the HPLC system dead volume. With the conditions listed using standard HPLC hardware, the gradient delay is approximately 20 seconds, and thus the analyte is eluting from the HPLC column at approximately 30% organic composition. Understanding the interplay of observed retention time and dwell volume can enable the reduction of HPLC method cycle time by truncation of unnecessary steps. Fig. 23.7B , indicates an accelerated gradient profile (0 to 100% over 100 seconds). This is the intermediate goal, but column screening must be performed before optimization occurs.

The previous solvent chemistry screen (development phase 1) should define the aqueous and organic solvent compositions applied to the RPLC screen. Because of the variety of aqueous-soluble buffers, three aqueous solvents and one organic solvent are routinely used. The three aqueous compositions should be materially different to ensure retention of some degree of variety. Note the two solvents in Fig. 23.6 (B and C) with trifluoroacetic acid versus formic acid. The experiment is performed (on a binary mixing pump) as follows: inject blank solvent twice (ignore the first such so the system is dynamically equilibrated), then neat analyte in 50 to 100% organic, and then the 3: 1 organic-to-matrix precipitated sample. This precipitated sample is used for phospholipid retention determination and gross IS variance relative to neat analyte solutions. This sample should be centrifuged and the supernatant transferred and injected to prevent the loading of precipitated proteins onto the head of the column, inducing increased pressure. After data acquisition, remove the first column, replace with the second column, and repeat the sample sequence. After all columns have been screened with the first set of solvents, the experiment is repeated with the second set of optimal solvents. It is more efficient to replace an HPLC column than to swap or purge mobile phases in screening HPLC modalities.

If the chosen eluting solvent is acetonitrile, 2 to 4 column volumes (1 mL for 50- × 2-mm column) of methanol should be applied to the column after the precipitated sample to remove phospholipids for the next time the column is used. Use of a quaternary pump allows three aqueous solvents to be screened between switching columns (12 injections sequentially) or the automatic addition of methanol after injection of the PPT sample to clean an RPLC column off after addition of phospholipid-rich precipitated matrix.

Fig. 23.8A describes data generated using a methanol-to-water gradient screening for a 10-steroid profile. Because of knowledge of isobaric interferences (structural isomers, no specific transitions for analytes are generated) the gradient pitch was reduced from 1 to 0.25% B per second (0 to 100% over 6.67 minutes). This example is included as a reminder that the HPLC screening stage should incorporate resolution from known interferents spiked into the analyte sample for determination of appropriate stationary phase and solvent chemistries that enable selective MS/MS measurement of the target analyte. Reduction of the data is shown in Table 23.6 . Amino, Cyanopropyl, BetaMax Base, and Hypercarb columns are qualitatively poor and excluded. Phenyl-Hexyl, PFP, Fluophase RP, and particularly C8 are excluded because of a lack of resolution of known isobars (C8 only has five peaks, peak counting is an additional tool for qualitative assessment of isobar resolution, as noted earlier) or peak fronting (PFP). Of the two C18 columns, the second demonstrates superior resolution of isobars over the first brand; however, when comparing to the BetaMax Acid separation, it is clear that two isomers (11-desoxycortisol and 17-hydroxyprogesterone, pink) are resolved but coelute on C18 (1). Thus Betamax Acid and C18 (2) would be good HPLC column functionalities for further development. This particular example demonstrates a complicated data set reduction and highlights some of the concepts of qualitative assessment of chromatographic data.

A secondary qualitative assessment is shown in Fig. 23.8B whereby imatinib (Gleevec; 35 ng/mL, 350 pg on column) was injected, followed by a precipitated sample to monitor the presence of phospholipids; chromatograms were overlaid to assess for selectivity. The monitoring of phospholipids was performed by analysis of the expected [M+H] ⁺ of highly abundant phospholipid species, including both lysophospholipids (1-acyl-2-OH glycerophosphocholines) and diacyl species (1-acyl-2-acyl glycerophospholipids). Owing to the additional fatty acid chain in the diacyl species, two distinct retention regions may be observed in reverse-phase separations. The ability to effectively resolve phospholipids chromatographically provides flexibility in sample preparation approaches.

Quantitative assessment of chromatographic performance ( Chapter 19 ) is routinely performed after peak integration. Commonly reviewed features include analyte peak height, retention time, peak width, peak asymmetry, and selectivity (between isobars), notably, the S/N ratio is less relevant here. Often, a first-pass screen of columns and solvents (see Fig. 23.8 ) leads to a second-tier screen of a particular functional modality as similar stationary phases from the same or different manufacturers may lead to dramatic changes in performance because of proprietary functionality and secondary chromatographic retention characteristics related to composition of the silica bed. For example, not all C18 columns are the same because various endcapping techniques, carbon load, formation, and abundance of residual silanols can induce the subtle retention variations a separation may require. Stationary phases exhibiting poor retentive or elution characteristics are abandoned, and additional energy is focused on the class of stationary phase that demonstrates preliminary success.

Fig. 23.9A indicates data reduction for temazepam from a benzodiazepine mixture; note that temazepam demonstrated the lowest transmission efficiency with the lowest desired LLMI, and thus was the key analyte of focus in optimization of sensitivity. First-tier HPLC screening indicated C18 was marginally better overall; however, Phenyl-Hexyl and fluorinated PFP phases also demonstrated acceptable performance and so were carried forward into the second tier. Review of the tabulated data indicates a number of key results that will enable appropriate selection of primary, secondary, and tertiary column (and solvent chemistry) choices. The choice of columns is rationalized by reduction of data for peak height (a measure of response), retention time (measure of acceptable retention), peak width (to assess broad peaks; compressed peaks generate greater signal above noise), and peak asymmetry (a measure of fronting or tailing).

The data shown in Table 23.7 are derived from the values shown in Fig. 23.9 and are rank-ordered by the characteristics detailed previously. Firstly, peak height was demonstrably better on the Hypersil Gold with Luna Phenyl-Hexyl, SB C18, and Extend C18 following in order. Peak height relates to sensitivity, theoretically enabling improved S/N ratio at the LLMI. The retention time of 1.7 minutes likely accounts for some of the observed increase in sensitivity as the organic eluent composition when temazepam elutes is greater using this media than the other columns screened. The organic composition during elution of this analyte (as calculated using retention time and dwell volume of the system) is approximately 60%. For effective chromatographic control (or the ability to manipulate retention time by gradient pitch/shape modifications), analyte elution in RPLC between 40 and 80% organic facilitates injection focusing and gradient shaping to resolve interferences. Thus analyte retention time in the panel should fall within this solvent composition window. Here again, the Hypersil Gold demonstrates the strongest retention, followed by the Atlantis T3 and Luna Phenyl-Hexyl (tied at third most retentive).

Evaluation of the peak width and, more importantly, peak asymmetry are key variables for consideration. Review of the data indicates the peak width is approximately 0.13 minutes (7.8 seconds) and the Fluophase PFP peak width is 0.077 minutes (4.6 seconds). These differences are relevant when complex separation is required (resolution from isobars); at this stage the detail is good information when selecting a secondary (backup column) to move forward. More importantly, peak asymmetry is a final deciding factor, calculated as (retention time − peak start)/(peak end − retention time). Perfectly Gaussian peaks have peak asymmetry (As) of 1. Peaks exhibiting asymmetry values between 0.8 and 1.5 are considered acceptable for further evaluation. The consideration of peak asymmetry is important for downstream data reduction because automated peak integration programs integrate Gaussian peaks more reproducibly than fronting (As >0.8) or excessively tailing (As >2) peaks. The rank ordering for As in Table 23.7 is presented as the absolute value of the difference between the measured As and the value of 1 (perfectly symmetrical) peak. The Fluophase RP achieves the highest ranking, followed by the Hypersil Gold and Max RP. The poorest asymmetry performers were the XDB-C18 and Luna C18 for substantial fronting and the Atlantis T3 for substantial tailing.

It is at this point that a single column is optimized, and the additional column ranking will be used again later when development dictates the need for a ready-made backup column in the event of manufacturing failure of lots or obsolescence of a particular stationary phase (clinical methods often survive for >10 years). Notably, the particular column characteristics were not chosen before initiation of this screening process. In these cases, it is the molecule’s preferential mobile phases (examined in solvent screening) and that molecule’s preference for a stationary phase (as deduced from empirical data) that selects the column chemistry of choice.

Additional high-performance liquid chromatography optimization

After selection of hardware configuration (solvents, stationary phases), the chromatographic program should remove superfluous steps to increase efficiency. The next stage in the development process involves reducing the cycle time of the HPLC method, as shown in Fig. 23.7B . This involves four steps:

• Increase percent of organic at injection

• Remove excess gradient time

• Reduce acquisition window/increase mass spectrometer bypass time

• Reduce or increase re-equilibration times

In Fig. 23.7 , the scouting gradient initiated at 0% organic. However, acceptable retention of the analyte at the head of the column may be achieved at a higher percent organic, facilitating the pass-through of poorly retained molecules in the injection slug (hopefully to waste by way of a bypass valve). Thus the starting solvent composition (percent solvent B) was increased to approximately 10% less than the solvent composition of the earliest eluting analyte. This also can be performed empirically by increasing the starting percent organic in 5% steps until breakthrough is observed (peak fronting).

Second, the gradient continues far after the latest peak of interest elutes; this additional gradient time does not provide any more useful data. It is thus removed by truncating the gradient to end at a solvent composition that is approximately 10% higher organic than the organic elution composition (percent solvent B) of the most retained analyte in the assay. The HPLC method is then immediately stepped to the washing conditions to remove strongly bound materials from the column. Here, the mass spectrometer acquisition window should be reduced to include just the analyte(s) of interest (with 6 to 10 seconds either side) and specifically exclude the materials eluting during the washing phase of the separation. A bypass valve should be used to divert the waste (i.e., unmeasured materials) away from the source of the mass spectrometer, preserving the cleanliness of the ion optics.

Finally, empirical evidence can be generated to shorten the re-equilibration time of the system back to the original starting conditions. Often, 3 to 5 column volumes are recommended (5 to 10 for HILIC), but this is not a steadfast rule. Reducing the amount of re-equilibration time at the end of the gradient and injecting the analyte of interest ( n = 3 per time reduction) should generate identical peak shape and retention time compared to longer equilibration steps in HPLC methods when optimizing this variable.

Adsorptive loss evaluation

After removal of unnecessary HPLC cycle time and reduction of the eluent window directed to the mass spectrometer, the possibility of analyte adsorption to containers and pipettes must be addressed. This is an exploratory experiment consisting of stressing the solubility of the analyte in various neat solutions against possible loss of the analyte to the walls of vessels and pipette tips. Equimolar solutions of the analyte (at some measurable concentration in the LC-MS/MS assay) are prepared in methanol, acetonitrile, and water, preferably in glass tubes. A moderate volume of these solutions is then transferred to three individual autosampler vials or wells of a 96-well plate (150 μL for a 200-μL well plate, 500 μL for a 1.0-mL well plate, etc.) in which the final sample will be stored for injection. One well is unmodified and serves as the control. Using a fresh pipette tip, perform a single aspiration and dispense event, then dispose of the pipette tip. This is then repeated 10 times, inducing the possibility of loss of the analyte to the walls of the pipette tip. The third sample is transferred to 10 unused vials or wells using the same pipette tip (do not dispose after the aspirate/dispense cycle).

Step 1: Prepare equimolar solution in methanol, acetonitrile, and water (all LC-MS–grade reagents), with three vials or wells per solution.

Step 2: Aliquot a fixed volume of each solution in triplicate to an autosampler vial or well of a 96-well plate. Note that the volume depends on maximum volume of container. Use appropriate judgment.

Step 3: Aspirate and dispense one sample of each solution using 10 different pipette tips, dispensing the liquid back to the origin vial or well.

Step 4: Transfer entire contents of a separate vial or well for each solution to an unused vial or well using the same pipette tip. Repeat for a total of 10 transfers.

Step 5: Aliquot the same fixed volume of an unspiked solution (neat water, methanol, and acetonitrile) to an autosampler vial or well of a 96-well plate as a negative control (contamination).

Four different specimens have been created from each of the three solutions ( n = 12 total). One set is the experimental baseline (interaction with a single pipette tip and single vessel), one set is completely blank solvent (contamination), one set repeatedly had interactions with fresh pipette tips, and one set has undergone repeated access to container surfaces. Injection of blank samples (first so there is no possibility of carryover) provides details on selectivity of containers—that is, do pipet tips and container walls generate interferences. Interferences generated from vessels require the sourcing of a different material or manufacturer and replication of the experiment. The comparison of adsorption in the different solvents across the different conditions to the single interaction (baseline) provides insight into the loss of analyte after sample preparation and can guide in the required final solution after sample preparation. Absolute losses greater than 10% should be eradicated through materials or chemistry changes (buffers, carrier solutions, etc.). These details can also inform relevant conditions for the preparation and storage of calibrators and QCs.

After determination of the appropriate solution to prevent adsorptive loss, perform increasing injections from 10 to 100 μL in 10-μL increments (after placing a 100-μL loop on the autosampler) and monitor for the degradation in peak shape. Fig. 23.10 indicates increasing injection of neat methanol solutions of cortisol onto an RPLC separation where cortisol elutes at 25% acetonitrile in the gradient. A subtle retention time shift and substantial chromatographic peak fronting are observed at 20 and 30% acetonitrile, respectively. The previously mentioned experiment was then replicated to facilitate a balance between the adsorptive loss of cortisol to a polypropylene 96-deep well plate and the appropriate retention of the molecule on the stationary phase at higher injection volumes. In the particular case of cortisol, a 1:1 mixture of methanol to water enabled an 80-μL injection to be performed without losses to either tips or wells in the 96-well plate. Note, however, that 80 μL is not the final injection volume for the assay but the maximum injectable volume at which peak shape is unaffected. Maximum injection volumes are used to facilitate robust operations only when the mass spectrometer performance has degraded (sensitivity loss). Ideally, the typical assay injection volume should not exceed 80% of the maximum injectable amount to allow flexibility (source contaminated) and should be used as a prompt for corrective action (instrument maintenance).

On-column detection limits

An estimation of the on-column detection limit that can be achieved is required. Stock solutions at 100× the assay expected LLMI are serially diluted in 10-fold steps through 6 cycles, generating final solutions from 100 to 1/1000th of the assay LLMI, together with the blank diluent solution. For example, an assay has an expected LLMI of 5 ng/mL. Solutions of that analyte are prepared in the injection solvent (determined previously) at 500, 50, 5, 0.5, 0.05, and 0.005 ng/mL. These concentrations are then converted from nanograms per milliliter to picograms per microliter, such that the final concentrations are 500, 50, 5, 0.5, 0.05, and 0.005 pg/μL. These solutions are injected at a fixed volume (microliters), allowing for the calculation of amount of material on the column. For a 10-μL injection of each of the solutions, the analyte amount would be 5000, 500, 50, 5, 0.5, and 0.05 pg on-column.

After triplicate injection of these materials at a volume equal to 50% of the maximum injectable volume determined previously, data are reduced by peak response and reproducibility of peak characteristics (peak width, peak area). The lowest concentration observed in the on-column detection study (S/N ratio >20:1 and consistent peak width and peak area determined as coefficient of variation [CV] < 10%) can be extrapolated to deduce the on-column detection limit. Note that the value for the S/N ratio can be misleading because the region of noise selected for all peaks may exhibit differences between injections due to the scanning nature of the mass spectrometer. This determination of on-column detection limits via peak area reproducibility is critical to decision making in the sample preparation process.

In an ideal world, the on-column detection limit would enable analysis of a reasonable sample volume (<100 μL for blood-based matrices) allowing for a 20% loss during sample preparation (desirable, to be discussed), yet providing an extract that can be injected twice. Following on from the previous example with a 5-ng/mL LLMI, desirable on-column detection limit = maximum sample volume × concentration of LLMI.

Assuming for 80% extraction efficiency:

Allowing for duplicate injection and residual volume after the injection:

The determined on-column detection limit guides the sample extraction strategy. Using the earlier example, having a lower on-column detection limit of 1 pg on column would allow for a 100-fold dilution before sample injection. A protein precipitation (PPT) extraction or dilute-and-shoot workflow may well suffice. Should the on-column detection limit be only 500 pg, however, the assay would require some form of concentration (approximately fivefold) and quite likely a larger sample volume (>100 μL), particularly when considering the necessity of final extract volume needed for repeat injections.

Final development stage 2 considerations

Before moving into sample preparation evaluation, ensure that the HPLC system is set up correctly, with the shortest tubing lengths installed to reduce systemic dwell volume, incorporating stainless-steel tubing where appropriate (where pressures exceed 200 bar or 3000 psi). Internal diameters of 0.005 inches are preferred for most HPLC tubing. Polyetheretherketone tubing must be cut flush (all fittings checked), and stainless-steel tubing should be laser cut and polished by the manufacturer because hand cutting in the laboratory can create burrs at outlets. An appropriate sample loop should be installed based on maximum injection volume allowable. Use of an inline filter frit between the autosampler and HPLC column (0.5- to 2-μm frit filter) can help reduce particulates clogging the interstitial spaces of the HPLC column, which can be responsible for increased back pressures and necessitate column replacement. Ensure there are multiple lots of the optimal columns on hand, as well as multiple lots of high-purity solvents; analytical-grade chemicals and HPLC- or MS-grade solvents should be used exclusively ( Table 23.8 ).

Finally, an evaluation of transition dwell times to enable adequate points across a peak (peak sampling) should be performed. When determining the appropriate dwell time per transition, the following details should be considered. At the very minimum, 10 points should be used to generate the peak shape, with 15 to 30 being far more reproducible. This must be balanced against the individual transition dwell time. The detection event is an averaging of the ion counts across a certain time span (counts per second [cps]). Very fast scanning (<5-ms dwell time) can incorporate random variation because of electronic or chemical noise in detection, resulting in jagged peaks that may be difficult to integrate. To assess appropriate dwell times, use the peak with the smallest width at its lowest abundance (the LLMI) and calculate the peak width, but exclude any fronting or tailing in seconds and multiply by 1000 to convert to milliseconds. Determine the total intertransition delay (number of transitions × fixed intertransition delay time) in milliseconds. The delta of these values is the available scan time. An example using 20 transitions, a 3-second-wide peak, and a 5-ms inter-transition delay is shown in the following equations:

Available scan time is thus 2900 ms available to scan across the 3-second-wide peak. Dividing this again by the total number of transitions (20) equals the cycle time per transition:

To maintain a minimum of 10 scans across each peak, the time per transition is divided by 10 to generate the necessary scan time per transition:

Note here that the value of 10 scans per transition is the absolute minimum recommended. A more preferable scan rate would include a higher number of points (scanning faster) but may come at the cost of noisier data. A number of theoretical examples are described in Table 23.9 using 3-second-wide peaks (3000 ms) and a fixed intertransition delay to evacuate the collision cell (Q2) between transitions. For most modern instruments, the interchannel delay is approximately 1 to 5 ms.

Tools and considerations

All LC-MS/MS assays have a requisite sample preparation component and because of the variety of measures performed, the greatest diversity of technologic solutions (see Chapter 21 ). Sample preparation provides the following attributes: (1) reduction of the sample complexity (selectivity for the analytes), (2) addition of ISs for improved control in measurement, and (3) concentration of analyte into an injectable volume. Sample preparation is ideally simple, reproducible, and orthogonal to the HPLC method (e.g., ion-exchange SPE with RPLC). The uncontrolled nature of clinical specimens places a significant burden on sample preparation to work in concert with HPLC separation and the MS transitions selected to add confidence in measurement. Therefore clinical methods should not focus on quantitative recovery (>95%), selective recovery is the end goal.

Before performing sample preparation development, a frame of reference is needed to ensure system performance and comparator—a system suitability test (SST) injection. Following on from adsorption studies, two solutions are prepared in bulk (store the remainder in a freezer or refrigerator) in large aliquots and placed in vials or wells in the autosampler for repeated injections. The first is the blank sample diluent and should be injected to ensure the analytical system is free of interferences (<20% analyte LLMI response and <5% IS response). The second is a neat solution of analytes at 100× LLMI, and the IS at 10× LLMI of the analytes. Where samples are expected to be concentrated or diluted by more than 10-fold, adjust the concentrations of the analyte and IS in the SST accordingly.

Before injecting samples, ensure the LC-MS/MS system is free of analyte or IS interferences at the expected retention times (two blank injections), inject the analyte SST solution in triplicate, then inject blanks again (2×) ensuring no appreciable system carryover is observed in the first blank (response <20% analyte LLMI and <5% IS). When performing provisional sample preparation optimization, the injected sample content can deleteriously affect instrument response. Repeated SST injections are monitored within-run to assess for instrument measurement drift (increases or losses for analyte/IS) that can confound assay recovery conclusions.

Three key experimental approaches used in evaluation of sample preparation include classic spike and recovery, determination of zones of suppression using PCI, and pre-extract and postextract spiking. The latter method is recommended and involves the individual determination of recovery, matrix effects, and total efficiency using three distinct samples. The first is a neat sample (A) that does not undergo extraction or contain sample matrix. This sample represents 100% recovery and zero matrix effects (the SST solution serves this purpose). The A sample should be fortified or diluted appropriately to correct for the expected extracted concentration for the sample preparation scheme. The second sample is spiked at 100× LLMI with analytes and 10× LLMI IS after extraction (B). The final concentration of this material should be equivalent to that of the A sample, although the detected response may be different. This specimen contains extracted matrix but no recovery losses for analyte or IS. The third sample is spiked at 100× LLMI for analytes before extraction and 10× LLMI IS after extraction (C). This specimen includes extracted matrix content and recovery losses with the opportunity to create a peak area ratio (IS added afterward) when ionization effects are confounding. The three calculations made from this sample set are:

This approach is much simpler to perform for exogenous drug analytes because true blank matrix is readily attainable. For endogenous biomarkers, a starting pool (as low as possible) that is overspiked to 100 × LLMI for analytes is used or the ISs are used to evaluate because they are unique to the testing specimen (if selected and chromatographed properly) and serve as a surrogate for the analyte where residual analyte concentrations are confounding (ISs replace analyte for the spike order samples 2 and 3 above). Consideration that stripped matrix may not contain the same binding proteins as intact matrix is important because sample preparation may be acceptable for charcoal-stripped but not for authentic matrices. When performing these studies, spiking of the low-level pool pre-extraction should be performed in bulk where the spike represents less than 5% of the total volume. Binding equilibration must be ensured before use to provide recovery calculations that are representative of the true nature of analytes in a sample before extraction. Absolute equilibration in binding is difficult to determine without a validated assay, and thus some assumptions concerning the equilibration must be made. Generally, extracting spiked samples immediately is discouraged; mixing for an hour or allowing a spiked sample to equilibrate in refrigerated conditions over a weekend is a good starting point.

The majority of sample preparation schemes used in clinical chemistry fall into one of five approaches ( Table 23.10 ): PPT, liquid-liquid extraction (LLE), supported liquid extraction (SLE), SPE, and online-direct injection (online). Several considerations are important in selection of the appropriate technique to start with; structural features (more specifically the uniqueness of structural features), the need to concentrate or dilute, and process efficiency.

The nature of the analytes is the first differentiator. For analytes that are charged in solution (carboxylic acids, tertiary amines, etc.) or highly polar, ion-exchange SPE can provide effective purification because of ionic or multimodal retention and the ability to use aggressive washing protocols to elute unwanted materials before elution of analytes. By contrast, LLE and SLE are preferred for neutral analytes to enable partition into the predominantly apolar solvent phase. Where analytes are thermally labile (acyl glucuronides), evaporation and reconstitution should be avoided; thus a “lossless” online approach or PPT injection is preferred over SLE, LLE, or SPE. The latter often creates diluted extracts relative to the starting sample volume, requiring concentration and usually a solvent exchange (evaporation/reconstitution) before injection. PPT protocols are generic, requiring little optimization but producing extracts with a significant amount of interferences, most notably phospholipids and salts, which lead to ionization effects. LLE protocols are often fine-tuned with just enough polar composition added to the extraction solvent to effect necessary recovery, are thus less generic, but provide the opportunity for cleaner extracts for subsequent analysis. In Table 23.10 , SPE is listed as selective (*) for ion-exchange modality only and hydrophobic SPE is predominantly a desalting and deproteinizing technique that affords limited orthogonal selectivity to the method.

High sample volume tests require considerations to the process efficiency. Dilute and direct inject represent a two-step addition process (sample + IS in diluent) and is readily automated in a 96-well plate—simple and fast. PPT is similar, usually including an additional post-precipitation transfer of supernatant to a fresh plate for injection. Multistep SPE and SLE can be performed with 96-well plate–based formats. LLE in tubes is inherently a poor candidate for liquid handlers, and in-plate LLE often suffers from well-to-well contamination (wicking). Multistep SPE and LLE are much lower throughput than SLE, PPT, or online, and preparative materials costs are worthy of consideration. Developmental experiments will be considered individually for each technique, although certain approaches will apply to multiple techniques.

Sample preparation development represents a two-step process. The first uses overspiked samples and comparison of SST with IS spiked before or after extraction to assess early success or failure of a probable extraction procedure. The second step involves generation of calibration curves in an appropriate matrix, together with analysis of patient samples for evaluation of interferences, spike/recovery, and dilutional mixing. Each preparative technique has step 2 as a common process. Discussion of first-pass developmental principles will be described for each technique first. The step 2 process will be described; however, if there are known interferents, such as isomers to the analyte of interest, that could be removed during sample preparation evaluation, consider adding to the screening protocol designed (phospholipids are just one example).

For final consideration, the MS/MS transitions that are selective have not yet been chosen; thus evaluation of baseline noise (S/N ratio) and the potential of interferences to extract and co-elute in a transition as the sample preparation motif is modified should be evaluated. Only after real human samples have been evaluated can the determination be made of which transitions (within the linear measurable range of the mass spectrometer) are of acceptable specificity.