A suggested standard for validation of LC-MS/MS based analytical series in diagnostic laboratories

Highlights

• •Policies applied to confirm LC-MS/MS-based results differ between laboratories.
• •We suggest a systematic approach for validation of individual diagnostic series.
• •An individual series validation plan can be established with a checklist.

1. Introduction

Method validation is addressed very intensively by the community of bio-analysts. However, since in particular LC-MS-methods are rather “volatile” in performance from day to day, validation of an actual analytical run or batch is of great importance too. Whatsoever, there seems to be rather limited scientific focus on this topic so far.

Evaluation of a measurement procedure as the final step of method development aims to characterize what a method is able to achieve; method validation is the test if then pre-defined performance requirements can accordingly be met with the method or not. Validation of a series – in our concept – aims to assess what the method actually has achieved. Based on pre-defined pass criteria – which should be essential part of a method description [1] -, measurement results are cleared for application in clinical decision-making (or are rejected) and compliance with performance requirements is confirmed. Meta-data obtained for individual samples in each run are assessed with respect to predefined pass criteria too, thus representing validation of a test result. This “three-level-terminology” of validation is visualized in Fig. 1.

Fig. 1.

Three-level concept of validation in clinical laboratory testing, with validation as the formal test if predefined specification and quality requirements are met.

We propose that series validation should be characterized as dynamic validation, an ongoing process that must effectively monitor method performance for the life cycle of the method (years in most clinical laboratories) under more challenging conditions than the initial validation. Factors contributing to the greater variance encountered in a series include highly variable LC-MS performance over the useful life of an instrument, use of multiple LC-MS instruments for the same method, multiple sample preparation analysts, periodic lot changes of reagents, calibrators, mobile phases, LC columns, sample preparation media, consumables, and internal standards, as well as the high complexity of matrix effect present in thousands of samples collected from patients with acute and chronic illness. In contrast, the initial validation of a measurement procedure as the final step of method development is typically performed by a limited number of the most highly skilled analysts in a laboratory, using <100 different sources of matrix (patient samples) to assess matrix effect and accuracy, and executed over days to weeks, rather than months to years. Viewed in this light, we can see why dynamic validation of a series may require processes and acceptance criteria more extensive and rigorous than the initial validation of method performance.

A substantial set of meta-data-based performance features and figures of merit for LC-MS/MS based quantitative analyses has been described (summarized in CLSI C62A [2]) – representing potential components of series validation protocols. However, to what extend such features are actually addressed in the validation of analytical series in diagnostic testing is highly heterogenous from lab to lab today. Notably, MS-kit manufacturers (e.g., Chromsystems, Recipe) give so far minimal guidance for meta-data-based validation of runs.

We here suggest a set of 32 generic criteria that can be covered by quality assurance policy and series validation rules in diagnostic laboratories applying LC-MS/MS.

We report these features as a checklist (topic addressed yes / no) (Table 1). We herein do not suggest actual numerical thresholds or quantitative pass criteria, but we merely list features and figures of merit to be assessed. (e.g., we mention “inaccuracy” as feature, not as e.g. “±10% acceptable”). This means that an individual QA/validation plan can be developed based on the suggestions made in this article, addressing the particular analytical and clinical requirements of a specific measurand.

Table 1.

Recommendations for items to include in series validation standard operating procedures (SOPs) and series checklist (procedural documentation associated with analytical records for each series). We use “series” here, other language for a series includes batch or run. For this discussion, a series is defined as a discrete set of samples including blanks, calibrators, QC and unknown (patient) samples extracted together by an analyst with documented competency using a validated protocol and injected sequentially on an LC-MS/MS instrument.

#	area	item	Comment, alternatives, exceptions, examples	Include in series checklists?
1	CAL	There is a conclusive policy defined with respect to full calibration (at least 5 non-zero, matrix matched calibrators) in every series that characterizes the measuring range with verification of the LLOQ / ULOQ. See comments and #2 & #3 for alternatives.	A. If a full calibration protocol is NOT used there should be detailed guidance for application of and acceptance criteria for at least minimum calibration (see #2). B. If the full calibration protocol is NOT used, validation data for use of fewer calibrators should be available for review. C. If a full calibration is NOT performed, series acceptance criteria should include allowed recovery actions for calibration failure such as referral to secondary review; partial reporting based on QC results and calibration review; series entire or partial repeat; adding a new full or minimum calibration to the existing series, etc. D. See #7E re calibration stability within a series.	NO
2	CAL	If full calibration (see #1) in every series is NOT used, there is a conclusive policy for either full calibration at defined intervals and/or a minimum calibration function (at least 3 matrix-matched calibrators including the LLoQ and ULoQ) at defined intervals. See comments and #3 for alternatives.	A. If a full or minimum calibration is NOT performed with each series – there should be detailed guidance for applying an alternate calibration protocol to the series (i.e. # and concentration of calibrators; use of historical curve(s) OR additive calibrations OR update to historical calibration using < 3 calibrators; required frequency of minimum / full calibrations, etc. – (T1) B. For robustness – calibration frequency should be defined as to both time interval (e.g. daily, weekly, every 6 months, etc.) and series scheduling (e.g. every series; no more than X unknowns or series between calibrations; full or minimum calibration at defined intervals with intermittent series of patient, blank and QC samples that do NOT include calibrators; after instrument maintenance; etc.). C. Series acceptance criteria should include evaluation of the alternate calibration as well as recovery actions for calibration failure – see #1D above. D. Data validating calibration at defined intervals other than every series and/or the use of an alternate rather than full or minimum calibration should be available for review. E. If matrix matched calibrators are not used – matrix effect validation studies should be available for review documenting trueness across the full measurement range using calibrators made in the alternative matrix (T2, T3). F. Regarding LLoQ and ULoQ verification in each series – see #3.	YES -There is an acceptable calibration function for the series as defined on the checklist and/or in an SOP reference
3	CAL	Only results within the concentration range between the lowest (LLoQ) and the highest (ULoQ) concentration calibration sample (analytical measurement range or AMR) are reported	A.For Dilutions – see #29. B. If an LLoQ and ULoQ calibrator are NOT included in each series, alternate protocols and metrics to verify adequate signal at the LLoQ and acceptable linearity between the LLoQ and ULoQ for each series are included in the series acceptance criteria. C. If the alternate LLoQ/ULoQ verification fails, there is guidance for recovery – see #1D above. D. If LLoQ/ULoQ calibrators are not included in each series, there should be data with sufficient statistical power available for review that verifies acceptable accuracy and precision at and between the LLoQ and ULoQ during the entire defined time interval for reporting patient results between full or minimum calibrations (T4).	YES, the AMR was verified for the series as defined on the checklist or in an SOP reference
4	CAL	Predefined pass criteria for the signal intensity of the lowest calibration sample (representing the LLoQ) are met.	For robustness, include both minimum signal-to-noise and minimum peak area as LLoQ acceptance criteria.	YES
5	CAL	Predefined pass criteria for slope, intercept and coefficient of determination (R2) for the calibration function are met.	A. Internal data or published references validating the pass criteria for slope, intercept and coefficient of determination for the calibration function should be available for review (T2). B. Initial validation studies are an excellent internal data source for deriving statistically sound target values and acceptable variance of the slope, intercept and coefficient of determination (R2) for a specific method in the laboratory of record.	YES
6	CAL	There are pass criteria for the tolerable deviation of back-calculated calibrator samples from expected values.	Typical pass criteria are +/- 15% deviation of residuals except at the LLoQ (+/- 20%). If less stringent criteria are used, internal data or references validating such practice should be available for review (T2).	YES
7	CAL, QC	There is a conclusive structure and detailed instructions (SOP or bench job aid [preferred]) regarding sequence and timing for sample preparation and instrument analysis within and between series for all sample types (blanks, calibrators, QC, unknowns). Parameters to consider include native and extracted sample stability, LC and MS/MS system robustness (e.g. stability of LC Rts, peak shape, MS/MS raw signal), potential for cross-contamination during sample preparation and instrument analysis.	For example: A. SST followed by full calibration in the morning; multiple series extraction with QC samples but no calibrators during the day, with intermittent stand-by times (LC flow-off); shut-down in the evening (LC washout/ion source bake-off protocols as needed followed by LC flow off, ion source off) OR B. SST followed by series of different assays; each with minimal calibration, blanks, and QC; run overnight continuously with LC washout/re-equilibration after each series as needed; followed by shut-down (LC washout/ion source bake-off protocols as needed followed by LC flow off, ion source off) OR many alternatives. C. Maximum series size should be defined: i.e. maximum # of total samples [calibrators + unknown samples + QC + blanks] that can be extracted together and injected sequentially during a defined time interval. See also 11B-D. D. The series extraction and injection order – e.g. zero or double blank 1st, blank with internal standard 2nd, then calibration curve, 1st QC, unknowns, 2nd QC, calibration curve – should be defined. There is documentation of the order for sample preparation and injection for each series (see also #8, 11 for QC sequencing). E. If a duplicate set of calibration samples are NOT extracted or injected (extract once, inject twice) at both the beginning and end of the series, data should be available for review that validates acceptable variance for native matrix sample replicates at the LLoQ and ULoQ that were processed at the beginning and end of a maximum size series. F. Related to stability within a series – see #16 re acceptance criteria for IS peak area consistency throughout the series. G. For reference measurement procedures more stringent protocols to reduce variance may apply, e.g. full calibration extracted at the beginning and at the end of a run, averaging of calibration functions, double injection of samples extracted in duplicate.	RECORD: # SAMPLES EXTRACTED IN THE SERIES, exceptions to sample preparation and injection timing rules (with approval), series records should include sample preparation & injection order.
8	QCs	QC samples in three concentration ranges are generally analyzed in each series, at least one in the highest and one in the lowest 20-25% of the calibration range. All QC samples in the series meet predefined pass criteria (relative deviation from target or multi-rule QC acceptance schema).	A. For robustness of methods with a wider AMR (e.g. >2 orders of magnitude) OR with multiple medical decision points dispersed across the AMR, use of more than 2 QC materials with concentrations that can verify accuracy in concentration ranges of clinical concern is recommended. For methods with narrow AMR 2 QC samples may be appropriate. B. For frequency and run order of QC samples within a series, see #11. C. For a discussion of multi-rule QC acceptance schema commonly used in clinical laboratories– see (T5).	YES
9	QC	Calibration samples and QC materials are manufactured independently (accuracy check).	A. Optimal procedure is preparation of QC and calibration materials by separate entities, e.g. one in-house and one commercial or by two distinct commercial sources. B. Alternative accuracy precautions when QC and calibrators are prepared together (in decreasing order of stringency) include: use of primary standards from different vendors; use of different lots of primary standard from the same vendor; preparation of QC and calibrators by different individuals using the same lot; separate weighing/pipetting events to create QC versus calibrator primary standard stock solutions from the same lot by the same individual.	NO
10	QCs	Target concentrations of QCs are verified externally, whenever possible.	A. Use of QC target values derived only from in-house or commercial source spiking of primary standard into blank matrix, WITHOUT a third-party analytical verification of accuracy, is deprecated. B. Optimally, commercial or in-house QC material concentrations are verified using a reference measurement procedure (RMP-often not available or too expensive). C. More feasible alternatives to 9B for verification or reassignment of QC target values (with decreasing degrees of stringency) include: parallel, in-house testing against a certified reference material as well as the current QC lot; parallel in-house testing of proficiency material that is value assigned with an RMP; external testing of QC material by a laboratory/LC-MS method with certified traceability to an RMP (e.g. CDC HoST certification); external testing of QC material by the laboratory/LC-MS/MS method used as a comparator in the validation of the current in-house LC-MS/MS method.	NO
11	QCs	A QC sample is analyzed at the start and at the end of each batch of unknowns and only results bracketed by accepted QCs are reported.	A. QC policies should distinguish between separate extractions of QC materials and re-injection of the same QC extract within the series. B. Defining the maximum # of samples allowed for a series (see 7C) should include criteria for the minimum # of replicates at each QC concentration and the order of QC replicates in the series, e.g. the maximum number of unknowns allowed between QC replicates; the minimum number of QC concentrations and replicates per 96 well plate; defined positions for QC in each 96-well plate of a multi-plate series in order to validate correct plate placement/ identification. C. Stability during extraction as well as injection should be addressed for 7C and 11B. D. Justification and validation data (if available) for the stated QC frequency and order in the series should be available for review.	NO, should be addressed by a single “Acceptable QC” checkbox. See #8.
12	QCs	There is a conclusive, clinically-based rationale / policy for the accepted inaccuracy and imprecision observed for QC sample results based on the intended use of the measurement method	See T5.	NO
13	QC	A longitudinal control chart of QC results is kept. Imprecision and inaccuracy (and/or total error) are assessed at specified intervals with predefined pass criteria for bias and %CV.	A. If the same method is used on multiple instruments, a strategy is defined for maintaining acceptable bias between instruments, e.g. use of a single QC range for all instruments AND / OR a protocol for between instrument comparison of patient samples that addresses known instrument variance and process vulnerabilities (T7).	YES, prompt to manually post results to QC charts if not electronic
14	QC	Authentic native matrix material (individual or pooled patient samples) at an appropriate concentration(s) is analyzed longitudinally with acceptable variance.	A. This recommendation is included to address within or between series drift or episodic inaccuracy that is not reliably detected by QC or SST samples OR use of patient samples as a less expensive alternative to QC material e.g. when matrix effect studies show significant differences between native matrix (patient) samples and QC materials; if a native matrix QC material is not available near a clinically significant medical decision point such as the LLoQ; if there is borderline acceptable LC resolution between a measurand and a common interfering peak that may degrade within the series. B. Design the protocol and acceptance criteria to address the variance of concern e.g. repeat of a patient sample from a previous batch, multiple extractions of a patient sample pool distributed throughout the series or replicate re-injections of single patient sample extraction at the beginning, middle and end of the series.	YES
15	IS	For the individual sample there are rules for the acceptable deviation of IS peak area from a reference IS peak area.	A. To assess differential matrix effects, errors in I.S. pipetting, and LC injection variance that may degrade accuracy and LLoQ (T2, T6) B. Differences exist for definition / derivation of the reference IS peak area – fixed historical, rolling average, calculated from calibrators (and QC) in each batch, calculated from all samples in the batch. Any strategy may be acceptable, so long as there is validation data to support the practice. C. Differences exist for the allowable deviation of IS peak areas, statistical derivation (% CV) from validation or production data is a robust means to establish this parameter.	YES
16	IS	There are predefined pass criteria for the tolerable variation (% CV) of IS areas over the entire run. **	A. Reproducibility of IS peak areas is related to the variance of matrix effect between samples and the precision of IS pipetting, LC autosampler injection and mass analysis. A warning threshold, as well as a rejection threshold for % CV of a series IS peak areas may be useful to detect gradual degradation of pipetting systems or instrument performance. Variance of matrix effect between samples should be addressed in method development and characterized during validation (T2, T6). B. Graphical presentation of IS peak areas for the series versus injection order is useful for troubleshooting and rapid detection of missed IS peak integration. C. LC-MS vendor software may not calculate %CV for IS peak areas, limiting application.	YES
17	BR (branch-ing ratios)	There are rules for the assessment of a branching ratio (ratio of two recorded mass transitions of a measurand) applied to patients’ samples.	If applicable (at least two transitions of sufficient intensity must be observed). BR may also be described as ion ratio, MRM ratio, qualifier/quantifier ratio.	NO
18	BR	The variation of the branching ratios observed for the entire run is assessed with pre-defined pass criteria (% CV). **	A. The absolute value for BR and the %CV may drift as the MSMS hardware becomes contaminated with extracted sample residue. Collecting internal data over the long term (months) may be necessary to derive realistic pass criteria (%CV stringent enough to detect interferents and the need for MSMS maintenance but also sufficiently broad to prevent a high rate of false rejections) (T8-T11). B. LC-MS vendor software may not calculate BR %CV, limiting application.	YES
19	BR	Branching ratios observed for individual patient and QC samples are compared with pre-defined pass criteria to a reference branching ratio.	A. To confirm analyte identity, detect interferents, and control for systematic differences between calibrator and unknown matrices. B. Differences exist for definition / derivation of the reference BR – fixed historical, calculated from calibrators (and QC) in each series, calculated from all series samples having a measurand concentration within the AMR. Any strategy may be acceptable, so long as there is validation data to support the practice. C. Differences exist for the allowable deviation of BR. There is consensus guidance (CLSI C50, C62), publications from non-clinical (WADA, CRO, pharma) practice and recent discussion of a clinical big data analysis and a survey of current clinical practice (T2, T6, T8-T11).	YES
20	CHRO	There are acceptance rules for the retention time (Rt) of the measurand for individual patient’s samples (time window) compared to calibrators.	A. To confirm measurand identity, detect possible interferents. B. See consensus guidance (CLSI C62) for recommended windows for absolute Rt(T2).	YES
21	CHRO	The accepted retention time difference between measurand and IS is specified (relative retention time, RRt).	A. See consensus guidance (CLSI C50, C62) for recommended windows for RRt (T2). B. RRt may be more robust (fewer false rejections) if there is Rt drift during a series. However, also using absolute Rt acceptance criteria can be useful to detect LC column degradation, overpressure and leaks.	YES
22	CHRO	There are acceptance criteria for the variation of the retention time observed for the measurand in the entire run (%CV). **	A. Primarily an indicator of LC pump, plumbing, and column status. B. See a recent discussion comparing longitudinal Rt variance between instruments / methods from clinical laboratory production “big” data (T8). C. LC-MS vendor software may not calculate series Rt %CV, limiting application.	YES
23	CHRO	There are acceptance rules for the peak shape observed in individual unknowns.	A. In addition to visual review, multiple metrics exist to characterize peak shape: e.g. area-to-height ratio, skewness, half peak width value, peak width at baseline. B. Calibrator samples serve as the reference peak shape for comparison of measurand and IS peak shapes in patient samples	YES
24	CHRO	A carry-over test is performed within each run with defined acceptance criteria	e.g., zero sample following highest calibration sample. Acceptance criteria depend on the selected peak detection threshold (e.g., no peak detected; or percentage of the high sample, or a maximum peak area in the zero sample, or and/or a maximum S:N in the zero sample. CLSI C62: peak area in the blank must be <20% of the LLoQ peak area (T2).	YES
25	CHRO	A zero or double blank sample (mobile phase at starting gradient conditions) is assessed in each run.	Acceptance criteria should be established for maximum signal (S:N and/or peak area) in the zero sample at the Rts for the measurand and IS.	YES
26	ION-isation	A zero sample (mobile phase) spiked with IS in an identical way as patient’s samples is assessed in each run with defined pass criteria for the ratio of IS area observed in this sample compared to matrix based samples in the run (ideally following the highest calibrator to also exclude carry over, see above)	A. To evaluate matrix effect. B. To check for signal in the measurand trace from contamination of the working IS solution. C. A single zero sample (mobile phase or extracted blank matrix) injected as the first sample, re-injected after the high calibrator and re-injected as the last sample in the batch permits assessment of LC system contamination early in the series, a lack of carryover from the high calibrator, and lack of system contamination at the end of the series. Contamination of the blank vial itself from carryover is possible.	YES
27	GENeral	At least one sample (with a measurand concentration in the mid measuring range) is injected several times (≥3) in the run to assess reproducibility of observed results with specified pass criteria	E.g., use of a patient pool sample (See #14) or re-injection of a calibrator samples	YES
28	GEN	Background (baseline) noise is assessed in each run with pre-defined acceptance criteria	e.g., defined maximum counts per second, maximum baseline drift over a time interval	YES
29	GEN	There is a conclusive policy for dilution of samples with concentrations >ULoQ or alternate reporting rules (e.g. do not dilute, report as >ULoQ)	There should be definition of the dilution matrix or diluent, pipettors to use, default dilution factor, default volumes for diluent / sample, and the maximum validated dilution factor. Alternative dilution volumes (different from default) used for individual samples should be documented in the series records.	YES
30	GEN	There is a conclusive policy for performing reagent, IS, and calibrator lot to lot comparisons.	A. Lot to lot validation protocols should include number and characteristics of samples to test and acceptance criteria. Lot tracking using barcode identifiers and a database facilitates troubleshooting and regulatory review. B. All lot information for materials (calibrator, reagents, IS, consumables as needed) used in the series including expiration dates should be included in series records.	YES – lots used for the series
31	GEN	Validation of series or result despite violation of a rule: evaluation and documentation is regulated.	e.g., Acceptance of any deviation to include: a) review by technical lead, supervisor, director b) written justification for the acceptance. C) signature, date & time of the reviewer accepting the deviation	YES
32	** GEN	A protocol is defined for between series (longitudinal) monitoring of selected raw LC-MS data / metadata parameters (LLoQ peak S:N or peak area; IS peak areas; BRs; Rts, etc.) at defined intervals or as needed for troubleshooting, similar to review of longitudinal QC data (see #13).	A. Longitudinal monitoring of LC-MS instrument parameters / raw data / metadata can provide early warning signs (prior to shifts or trends in QC results) of service required / maintenance due for instruments; OR method problems related to sample preparation, reagents or consumables; OR unacceptable variance between instruments (T8-T11). B. Commercial (Indigo Bioautomation ASCENT) and open source (MSStatsQC; Univ of Washington Skyline/Panorama, ) software applications with real time automation of this functionality are available. C. Series by series export from LC-MS data archives and analysis with Excel or R is feasible for occasional troubleshooting but very labor intensive for routine monitoring.	NO

References to Table 1
T1	Rule GS, Rockwood AL. Improving quantitative precision and throughput by reducing calibrator use in liquid chromatography-tandem mass spectrometry. Anal Chim Acta. 2016;919:55-61.
T2	CLSI. Liquid Chromatography-Mass Spectrometry Methods; Approved Guideline. CLSI document C62- A. Wayne, PA: Clinical and Laboratory Standards Institute; 2014.
T3	Suhr AC, Vogeser M, Grimm SH. Isotope Inversion Experiment evaluating the suitability of calibration in surrogate matrix for quantification via LC-MS/MS-Exemplary application for a steroid multi-method. J Pharm Biomed Anal. 2016;124:309-318.
T4	Armbruster DA, Pry T. Limit of Blank, Limit of Detection and Limit of Quantitation. Clin Biochem. 2008; 29-Suppl(i):S49-S52.
T5	Westgard JO, Westgard SA. Quality control review: implementing a scientifically based quality control system. Ann Clin Biochem. 2016;53:32–50.
T6	Vogeser M, Seger C. Quality management in clinical application of mass spectrometry measurement systems. Clin Biochem. 2016;49(13-14):. 947-54.
T7	Calleja J. Parallel Processing and Maintaining Adequate Alignment between Instruments and Methods. Clin Biochem Rev. 2008; 29 Suppl (i):S71-77
T8	Zabell APR, Stone JS, Julian RK. Using Big Data for LC-MS/MS Quality Analysis. Clin Lab News; 1 May, 2017.
T9	Dogu E, Mohammad-Taheri S, Abbatiello SE et al. MSstatsQC: Longitudinal System Suitability Monitoring and Quality Control for Targeted Proteomic Experiments. Mol Cell Proteomics. 2017;16:1335–1347.
T10	Hayden J, Bachmann L. Questioning Quality Assurance in Clinical Mass Spectrometry. Clin Lab News;1 July, 2019.
T11	Berman MS, Beri J, Vagisha S et al. An Automated Pipeline to Monitor System Performance in Liquid Chromatography−Tandem Mass Spectrometry Proteomic Experiments. J Proteome Res. 2016;1:4763−4769

CAL = calibration, QC = quality control, IS = internal standard, BR = branching ratio, CHRO = chromatography. LLoQ = lower limit of quantitation, ULoQ = upper limit of quantitation, Rt = retention time, RRt = relative retention time

2. Discussion

LC-MS-based methods and instrument configurations are highly complex and offer a multitude of error possibilities [3], [4]. Compared to fully automated standard clinical chemistry systems, both laboratory developed and commercially available kit-based LC-MS methods have a significantly higher risk potential for clinical diagnostic applications. As part of risk management in this context, the series validation concept proposed in this article aims to avoid unfavorable analytical results and negative consequences for patient safety.

This suggested standard does obviously not represent a minimum standard and is beyond the efforts applied in many diagnostic laboratories at present. However, we aimed to maintain a sound balance between the realistically achievable allocation of resources and the risks of implementation of a highly complex technology for clinical decision-making in a laboratory.

The extra-effort to follow this suggested run validation standard is given but still easily practicable – in terms of additional injections, increased overall run time and work. The components of the suggested standard can be implemented in spreadsheet files and run semi-automated on standard report files of chromatography software solutions provided by the instrument vendors.

A major contributor to between laboratory heterogeneity in validation of series is the labor-intensive nature of these procedures – especially in multi-analyte methods. Software solutions for automating or semi-automating the metadata analysis component of series validation can be described as follows:

• 1.LC-MS vendor software: Standard chromatography software packages provide some tools for meta-data analysis (e.g., assessment of quantifier-qualifier-ratios, exception finding); however, there is no accepted common industry standard established so far in this area.
• 2.3rd party software for LC-MS: Highly sophisticated solutions for peak review, series validation and performance monitoring are commercially available from dedicated providers (e.g. Indigo; Veritomyx). An open source application – Skyline/Panorama from the MacCoss laboratory – was developed for proteomics but has been used successfully for peak and metadata review of a small molecule method [5]. In part these solutions for primary data handling are hosted on centralized servers.
• 3.In-house applications for metadata analysis: In this use case – peak review (and in some cases metadata parameter calculation) is performed with LC-MS vendor software but raw or processed data is exported to an in-house developed application for metadata calculations and/or finding exceptions. For example, see the use of an in-house Python program developed by the Univ. of Washington [6]. Use of Excel spreadsheets for this purpose is also in use.
• 4.Commercial middleware and LIS applications for metadata analysis: As for the in-house solutions described above, in this use case LC-MS vendor software is used for peak review/metadata parameter calculation but commercially available middleware or LIS auto-verification functionality is used to calculate, identify, prevent auto-verification, and present metadata exceptions for review. Several examples have been described with options for parsing out to either LC-MS vendor software, commercial middleware (Data Innovations Instrument Manager) or commercial LIS applications (SOFT Labs LLC) the separate steps of a) metadata exception calculation b) exception flagging c) exception presentation for review and d) application of auto-verification rules.

Leading laboratory accreditation schemes require that any reported patient result has to be trackable back to reagent lot, calibrator lot, calibration curve, and maintenance records of the instrument used to report the result. This requirement is not specific for mass spectrometry-based methods and thus not incorporated in the matrix suggested here. Tracking may have long-term quality impact on an organization level [7] but not specifically on a series level, which is the main focus of our work.

In the pharmaceutical field run acceptance criteria have been addressed previously [8]; however, in a very abstracted concept that is not evidently applicable to clinical laboratories. A less detailed suggestion for the implementation of quality assurance plans in diagnostic laboratories using LC-MS/MS has been published recently [9].

Fully automated and “closed” LC-MS-based analyzer systems will probably play an important role in the future of clinical chemistry. However, development and dissemination of such instruments still requires several years and will not cover all useful measurands. Consequently, more risk-prone laboratory developed tests or test kit implementations on very heterogeneous instrument setting will dominate in the clinical application of mass spectrometry for a rather long time. With this article – in summary – we suggest a guidance scheme of run validation of such not-industrialized LC-MS-based diagnostic tests for discussion and evaluation in the community of MS-users in clinical laboratories; this may include preliminary use of the checklist in method audits. We intend to systematically assess feedback to our article for ongoing development of this document in the MSACL community (msacl.org, Mass Spectrometry Application to the Clinical Laboratory) and in particular as letters to the Editor of Clinical Mass Spectrometry.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.