Use of Internal Standards (IS) has been recommended in LC-MS/MS and LC-MS applications for quantitating drug concentrations in regulated bioanalysis in accordance with the 2022 FDA “M10 Bioanalytical Method Validation and Study Sample Analysis” Guidance (2022 FDA M10 BMV Guidance) [1]. Due to the complexity of biological samples, sample processing and instruments, a certain level of IS response variability (ISV) within a run or across runs in a study is expected. Following the 2019 publication of FDA's Q&A guidance “Evaluation of Internal Standard Responses During Chromatographic Bioanalysis: Questions and Answers” to provide the Agency's current thinking on ISV and the potential impact on the accuracy of the resultant measurements [2], the topic continues to be discussed among industry, academia and regulators. Although there is no global consensus on the acceptance criteria for ISV, the bioanalytical community agrees that the root cause of ISV and its effect on data accuracy should be investigated when ISV occurs in a bioanalytical study. In this commentary, we share our thoughts on this topic based on reviewing cases reported by the bioanalytical community and our research on validation and implementation of bioanalytical methods. The discussion will focus on the pattern of ISV informing the root cause investigation and remediation; adoption of the concept of parallelism to evaluate the trackability of an IS for an analyte in study samples during sample analysis utilizing a validated method; and practical considerations to consider when addressing ISV in bioanalytical studies.
1. Recommendations on the use of IS in 2022 FDA M10 BMV guidance
An IS is defined as “A structurally similar analogue or stable isotope labeled compound added to calibration standards, QCs, and study samples at a known and constant concentration to facilitate quantification of the target analyte”. Evaluation of analytical method selectivity for the IS is generally recommended. When MS detection is used, the use of the stable isotope-labeled analyte as the IS (SIL-IS) is recommended whenever possible, and the SIL-IS should be of high isotope purity and no isotope exchange reactions should occur.
2. Pattern of IS variability
The 2022 FDA M10 BMV guidance recommends that during study sample analysis, “The IS responses of the study samples should be monitored to determine whether there is systemic IS variability.” [1]. A number of review and research papers [3–10] have summarized cases of ISV, trouble shooting, remediation and even proposed decision trees [3,6,9] based on their SOP specifications on acceptance criteria and repeated analysis. The decision trees are a practical guide for analysts addressing ISV during their daily implementation of bioanalytical methods. In addition, we believe that it is beneficial to carefully examine the ‘pattern’ of ISV before study samples are reanalyzed.
Random ISV across the whole bioanalytical batch or study may be caused by instrument malfunction ([4] case 2), poor quality of lab supplies ([4] case 5), lack of homogeneity of processed samples ([4] case 10) and analysts' operational errors ([4] case 1).
Decreased IS response with increasing analyte concentration may be an indication of ionization suppression/competition ([7,11] case 2, [12]).
Systematical difference in IS response between CC/QC (Calibrators and Quality Control samples) and study samples may be caused by endogenous components (e.g., related to disease status or special population) in study samples causing matrix effect ([13] case 1) or interference ([5]), drug stabilizer added during study sample collection ([8] case 1), stability issue ([6] case 2), different anticoagulants used for plasma preparation ([10] case 2), or plasticizer in commercially available plasma/serum ([5] case 2).
Abnormal IS response observed in a few subjects may be related to the subject's underlying health conditions ([5] case 1, [14]) or concurrently administrated medication ([13] cases 2 and 3).
Different IS response between pre- and post-dose study samples may be attributed to drug metabolites ([13] case 4, [15]) or dosing vehicles [16].
Unexpected ISV during certain time frame of a bioanalytical study may be related to change of lab supplies ([4] case 5) and/or ‘charging’ of mass spectrometer ([4] case 3).
Abnormal IS response at regular intervals regardless of the nature of subject samples might be caused by malfunction of one or a few channels of the liquid handling system or multichannel pipette used for sample processing ([10] case 5).
ISV issues in reinjected samples may be an indication of lack of homogeneity of reconstituted samples ([6] case 4).
The patterns themselves, along with prior knowledge on the analytes, matrices, sample processing methods and instruments, may provide insightful ‘hints’ into the root cause of ISV, its impact on data accuracy, and guide the user with remediation approaches. For example, thoroughly examining experimental notes led to identifying sample processing supplies causing an abnormal IS response during a certain period of the study ([4] case 5); dilution with the control matrix was typically used when there was a systematic difference between the IS response in CC/QC and study samples to investigate the impact of matrix effect on sample accuracy [3,13]; changing calibrators and/or IS concentrations, optimizing MS parameters or switching the ion source from ESI to APCI was used to address ionization suppression/competition between analyte and IS [11,12]; a variable IS response caused by interfering endogenous substances in a subset of study subjects [14] or drug metabolite(s) in post-dose samples ([13] case 4, [15]) often necessitated the redevelopment of the method.
3. Adoption of parallelism approach to evaluate the trackability of IS
The trackability of the IS for the analyte is a key factor that affects the overall performance of LC-MS/MS assays [3]. In many cases of abnormal IS response caused by the matrix effect mentioned above, the ‘right’ ISs were able to efficiently track the analytes, so the data accuracy was not impacted, although sample processing and/or LC methods were later optimized to remedy the IS issues in most of the reports (e.g., [5] case 2 [6], case 3 [13], case 1). The outcome is not surprising because SIL-ISs and corresponding analytes are generally very close in terms of physical–chemical characteristics. However, caution should still be exercised when interpreting the impact of ISV on data accuracy even when deuterated ISs are used. It is well known that some deuterated ISs have a slight shift in chromatographic retention time compared with the non-deuterated analyte, and the slight difference may translate to suboptimal tracking if the sample processing method is not optimized [3]. For instance, case 12 ref [4] reported a difference in deuterated IS response between originally injected and reinjected samples. Although the root cause was not investigated because samples were not available, the authors speculated that the differences were caused by suboptimal tracking as the IS and analytes peaks were not overlapping. Jemal M et al., reported similar findings in urine samples from a study subject where the normalized matrix factor for mevalonic acid was only around 50% when using deuterated IS mevalonic acid-d7 [17]. Similar observations have also been reported for omeprazole using omeprazole-d3 as the IS [18].
During method development, the performance of IS should be evaluated by recovery, matrix effect, and importantly by comparing incurred sample reanalysis results [3,19]. Theoretically, the IS-normalized matrix factor should be close to 100% across low, middle and high concentrations in all lots of plasma (or other biological matrixes) during method development and validation [3]. Of note, it is not feasible to evaluate the trackability of IS in study samples by IS-normalized matrix factor because the concentration of analyte is unknown in study samples.
In comprehensive reviews [3,13], the parallelism approach (dilution or standard addition) has been proposed to investigate ISV due to the matrix effect and/or specificity issues. The efforts had focused on achieving similar IS response in study samples to those of calibrators and QCs, and the repeated analytical results after dilution or standard addition being consistent with the original values (e.g., within ±20%). Parallelism is defined as a parallel relationship between the calibration curve and serially diluted study samples to detect any influence of dilution on analyte measurement. Non-parallelism usually is caused by the lack of trackability of IS for the analyte due to the matrix effect and/or the presence of interfering substances. The benefit of parallelism is that the approach directly investigates the in-study samples which are different from blank matrices and are not available during method development and validation.
Though the IS response of aberrant samples could be similar to those of calibrators and QCs after dilution or standard addition, there might still be concerns with data accuracy if the IS doesn't track the analyte well for matrix effect. The ultimate goal of parallelism experiments should be to demonstrate the trackability of IS.
The central hypothesis in the use of an IS was that the IS should be able to track the analyte when there is matrix effect, and the analyte to IS signal ratio should remain unchanged when the concentration of chemical components causing matrix effect decreased (dilution of matrix or standard addition of analyte/IS) in study samples. To evaluate this hypothesis, we developed and validated non-chromatographic MS/MS (Refs [20,21] for instrument details) methods for drugs A and B in human plasma using d4-A and d5-B as ISs, respectively. Drugs A and B have similar therapeutic indication and physico-chemical properties (i.e., LogP and solubility). As can be seen from Supplemental Figure S1, the IS response of d4-A and d5-B was suppressed in lipemia plasma samples. A twofold dilution of the lipemic plasma with normal plasma samples significantly diminished the impact of lipemia on IS signals, and the IS signals in 4x, 8x and 16x diluted lipemic plasma samples were similar as those in normal plasma samples. Accuracy of LLOQ QC and HQC samples were acceptable in serially diluted lipemic samples. In serially diluted lipemia plasma, the ratios of A/d4-A and B/d5-B were consistently independent of the dilution factor, which was an indication of good trackability of d4-A to drug A, and d5-B to drug B.
If the ISs were swapped, i.e., d4-A was used as IS for drug B and d5-B was used as IS for drug A, the accuracy for drug B was not significantly impacted in undiluted, 2x, 4x and 8x diluted lipemia plasma samples (<20% difference). However, drug A concentrations were ∼50% overestimated in undiluted, 2x, 4x diluted lipemic plasma samples (Supplemental Figure S2) although the IS d5-B signal was close to normal with the 2x dilution (Supplemental Figure S1). The ratios of B/d4-A and A/d5-B were inconsistent (or even sporadic) in serially diluted lipemic plasma, which indicated poor trackability of d4-A to drug B, and d5-B to drug A. We observed similar pattern when swapping ISs for bupropion and its pharmacologically active metabolite hydroxy-bupropion, d9-bupropion and d6-hydroxy-bupropion, respectively. We acknowledged that this was a hypothetical case as drugs A and B have been on the market for decades, and d4-A and d5-B have been routinely used as IS for drugs A and B. However, the scenario may reflect real-world situation as the French National Agency for Medicines and Health Product Safety acknowledged having a case where the IS had a completely different structure of the analyte used during an inspection of a bioequivalence study [10].
We would like to point out that this parallelism approach investigating trackability of IS for analyte may be of practical value when: evaluating the trackability of analog IS if not using SIL-IS (e.g., when SIL-IS is not available at early development stage) for method development and implementation; optimizing sample processing, LC and other method parameters and assuring trackability of IS for analyte when implementing a validated method to ‘new’ sample matrix (e.g., new clinical subject population in terms of age, gender or disease status or even a new drug product formulation). We have implemented this approach for a genotoxic impurities' method [20] and an in vitro permeability test (IVPT) method [21] to ensure analytical data quality despite the lack of ‘perfectly’ matched control matrices and internal standards for each individual analyte.
4. Practical considerations
First, it is important to note that consistent IS response does not always translate to good accuracy. For example, in case 9 of [4], discrepancies were observed between original results and the re-analysis of a few samples with dilution, despite the relatively stable IS response. An investigation suggested that minor changes in sample pH after dilution caused a difference in recovery of the analyte and analog IS during sample extract by mixed cation exchange (MCX) solid phase extraction. Switching to deuterated IS was able to efficiently track the difference in MCX recovery thus improved the accuracy. This case indicates that stable IS responses cannot be automatically assumed as an indication of good accuracy unless they reflected the variations the analyte experienced.
The scenario of using one IS for two or more analytes can be complicated, especially when the analytes are structurally related such as drug and metabolite(s) [3]. Theoretically the IS cannot have almost identical physico-chemical characteristics with two or more analytes. As previously discussed, deuterated IS signal can be suppressed by co-eluting analyte [11,12]. In that case, if the IS signal is being suppressed by only one analyte that could indicate that the IS may not efficiently track other analytes.
Though some cases of abnormal IS response were reported to not impact data accuracy, it might affect sensitivity, especially samples in which the analytes concentration was around the LLOQ level. In addition, greater ISV in study samples than CC/QC samples might be an indication of non-optimized sample processing method. For example, we published a method for six analytes in IVPT receptor media [21]. The initial sample processing method was lyophilization of the receptor media followed by organic solvent extraction and the method was successfully validated. However, when the method was implemented for IVPT studies of different products, the ISV was considerably higher in study samples. The speculation was that, during for IVPT studies, excipients in test products and skin components might be dynamically solubilized in the receptor media and then the initial sample processing method may not be able to efficiently minimize their impact on mass spectrometry signals. Due to lack of ‘perfectly’ matched blank matrixes and IS for each individual analyte, we optimized the sample processing method to protein precipitation followed by dilution to better control the matrix effect and ISV was similar for study samples and CC/QC samples using the finalized method.
Matrix effects, an inherent limitation for mass spectrometry, highlights that optimized sample processing and chromatographic separation methods are equally important as mass spectrometry parameters. As we previously presented in 2019 and 2020 Workshop on Recent Issues in Bioanalysis (WRIB) meetings [22,23], the abnormal IS response caused by lipemia, metal adducts, and other reasons can be efficiently remedied by changing of sample delivery solvents and sample processing methods (e.g., using solid supported liquid extraction). Changing sample preparation approaches to limit what is injected into the mass spectrometer not only can help to assure data quality, but also helps to reduce the ‘sample matrix burden’ to the mass spectrometer itself. Such optimization is particularly important when there is lack of a ‘perfectly’ matched blank sample matrix and a IS for each individual analyte as we described previously.
5. Internal standard is not just for bioanalysis
Evolving quality expectations drives the adoption of advanced technologies, which in turn help to better assure product quality. Historically, MS/MS based analytical methods were mainly used for bioanalysis in support of clinical studies due to their inherent high sensitivity and procedural specificity. In recent years, MS/MS methods have been used worldwide for trace impurities, especially genotoxic impurities such as nitrosamines. The recently finalized ICH Q2(R2) guideline has highlighted the importance of ISs in MS/MS methods and provided a framework for evaluating robustness and a range of parameters range for analytical procedures [24].
In a recently published FDA laboratory study investigating the mitigation approaches for nitrosamine drug substance-related impurities (NDSRIs) [25], we implemented the parallelism approach described above to demonstrate trackability of IS for N-nitrosobumetanide impurities and evaluated the in-study method performance in multiple formulations. In this study, we developed and validated a non-chromatographic-MS/MS method to quantitate N-nitrosobumetanide in bumetanide drug products using d5-N-nitrosobumetanide as the IS. There was limited prior knowledge of N-nitrosobumetanide as it was a recently identified impurity. Additionally, antioxidants and pH modifiers added to the in-house formulated bumetanide drug products and changes in formulation procedure made the in-study samples matrixes different from that using for method development and validation. For the parallelism evaluation, in-study samples were serially diluted 2, 4, 8, 16, 32, 64, 128, 256 and 512-times and analyzed with the original samples together in an analytical batch. Back calculated concentrations of the diluted samples based on N-nitrosobumetanide/d5-N-nitrosobumetanide ratios was almost identical with the original samples in selected formulations, regardless of dilution level. The data strongly suggested different matrix effects in different formulations, if any, were very well tracked by the IS. The data quality was further verified by comparison with data generated with an orthogonal LC-HRMS method.
Matrix effects are the ‘Achilles' heel’ of electrospray mass spectrometry methods, which are now widely used procedures for bioanalytical and analytical methods. The primary value of the internal standard is that the standard can help to address the negative impact that a matrix effect can have on data accuracy and, uniquely, can serve as a ‘probe’ for data quality from complex biomatrices. As such, IS has been highly recommended in for bioanalytical method development and increasingly been used in small molecule pharmaceutical analytical methods to enhance data accuracy. Based on the organized literature information presented on the root cause investigation of ISV and our shared practical considerations, we proposed the concept of parallelism to evaluate the trackability of IS based on deductive reasoning in this commentary. We hope this attempt may help move away initially from root cause investigation of ISV using reactive trouble shooting to proactive risk assessment and potential mitigation.
Supplementary Material
Acknowledgments
The authors thank W Li (Novartis), S Lowes (Q Squared Solutions), E Woolf (Merck), N Weng (Janssen, now GSK), W Jian (Janssen), FF Vazvaei-Smith (Merck) and L Li (Office of Bioequivalence, Office of Generic Drugs, FDA/CDER) for valuable discussions.
Supplemental material
Supplemental data for this article can be accessed at https://doi.org/10.1080/17576180.2024.2348939
Author contribution
J Zhang: project administration, investigation, methodology, formal analysis, writing. A Dasgupta: investigation. R Ayala: investigation. C Bonapace: investigation, supervision. S Kassim: project administration, supervision.PJ Faustino: project administration, supervision.
Disclaimer
This publication reflects the views of the authors and should not be construed to represent FDA's views or policies.
Financial disclosure
The authors have no financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
Competing interests disclosure
The authors have no competing interests or relevant affiliations with any organization or entity with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
Writing disclosure
No writing assistance was utilized in the production of this manuscript.
References
- 1.FDA . M10 bioanalytical method validation and study sample analysis. 2022. https://www.fda.gov/media/162903/download
- 2.FDA/CDER . Evaluation of internal standard responses during chromatographic bioanalysis: questions and answers guidance for industry. 2019. https://www.fda.gov/media/130451/download
- 3.Fu Y, Barkley D, Li W, et al. Evaluation, identification and impact assessment of abnormal internal standard response variability in regulated LC-MS bioanalysis. Bioanalysis. 2020;12(8):545–559. doi: 10.4155/bio-2020-0058 [DOI] [PubMed] [Google Scholar]
- 4.Tan A, Hussain S, Musuku A, et al. Internal standard response variations during incurred sample analysis by LC-MS/MS: case by case trouble-shooting. J Chromatogr B Analyt Technol Biomed Life Sci. 2009;877(27):3201–3209. doi: 10.1016/j.jchromb.2009.08.019 [DOI] [PubMed] [Google Scholar]
- 5.Verhaeghe T. Systematic internal standard variability and issue resolution: two case studies. Bioanalysis. 2019;11(18):1685–1692. doi: 10.4155/bio-2019-0165 [DOI] [PubMed] [Google Scholar]
- 6.van de Merbel NC, Koster RA, Ohnmacht C. Very complex internal standard response variation in LC-MS/MS bioanalysis: root cause analysis and impact assessment. Bioanalysis. 2019;11(18):1693–1700. doi: 10.4155/bio-2019-0122 [DOI] [PubMed] [Google Scholar]
- 7.Wright MJ, Wheller R, Wallace G, et al. Internal standards in regulated bioanalysis: putting in place a decision-making process during method development. Bioanalysis. 2019;11(18):1701–1713. doi: 10.4155/bio-2019-0169 [DOI] [PubMed] [Google Scholar]
- 8.Buonarati MH, Schoener D. Investigations beyond standard operating procedure on internal standard response. Bioanalysis. 2019;11(18):1669–1678. doi: 10.4155/bio-2019-0187 [DOI] [PubMed] [Google Scholar]
- 9.White S, Adcock N, Elbast W, et al. European Bioanalysis Forum: recommendation for dealing with internal standard variability. Bioanalysis. 2014;6(20):2767–2774. doi: 10.4155/bio.14.221 [DOI] [PubMed] [Google Scholar]
- 10.Blaye OL. Variations in internal standard response: some thoughts and real-life cases. Bioanalysis. 2019;11(18):1715–1725. doi: 10.4155/bio-2019-0146 [DOI] [PubMed] [Google Scholar]
- 11.Jian W, Edom RW, Xu Y, et al. Potential bias and mitigations when using stable isotope labeled parent drug as internal standard for LC-MS/MS quantitation of metabolites. J Chromatogr B Analyt Technol Biomed Life Sci. 2010;878(31):3267–3276. doi: 10.1016/j.jchromb.2010.10.008 [DOI] [PubMed] [Google Scholar]
- 12.Sojo LE, Lum G, Chee P. Internal standard signal suppression by co-eluting analyte in isotope dilution LC-ESI-MS. Analyst. 2003;128(1):51–54. doi: 10.1039/b209521c [DOI] [PubMed] [Google Scholar]
- 13.JW E. Learning how to interpret ‘dangerous’ internal standard behaviors. Bioanalysis. 2019;11(18):1679–1684. doi: 10.4155/bio-2019-0005 [DOI] [PubMed] [Google Scholar]
- 14.de Vries R, Huang M, Bode N, et al. Bioanalysis of ibrutinib and its active metabolite in human plasma: selectivity issue, impact assessment and resolution. Bioanalysis. 2015;7(20):2713–2724. doi: 10.4155/bio.15.159 [DOI] [PubMed] [Google Scholar]
- 15.Furlong M, Bessire A, Song W, et al. Use of high-resolution mass spectrometry to investigate a metabolite interference during liquid chromatography/tandem mass spectrometric quantification of a small molecule in toxicokinetic study samples. Rapid Commun Mass Spectrom. 2010;24(13):1902–1910. doi: 10.1002/rcm.4587 [DOI] [PubMed] [Google Scholar]
- 16.Jian W, Romm MV, Edom RW, et al. Evaluation of a high-throughput online solid phase extraction-tandem mass spectrometry system for in vivo bioanalytical studies. Anal Chem. 2011;83(21):8259–8266. doi: 10.1021/ac202017c [DOI] [PubMed] [Google Scholar]
- 17.Jemal M, Schuster A, Whigan DB. Liquid chromatography/tandem mass spectrometry methods for quantitation of mevalonic acid in human plasma and urine: method validation, demonstration of using a surrogate analyte, and demonstration of unacceptable matrix effect in spite of use of a stable isotope analog internal standard. Rapid Commun Mass Spectrom. 2003;17(15):1723–1734. doi: 10.1002/rcm.1112 [DOI] [PubMed] [Google Scholar]
- 18.Liu G, Ji QC, Arnold ME. Identifying, evaluating, and controlling bioanalytical risks resulting from nonuniform matrix ion suppression/enhancement and nonlinear liquid chromatography-mass spectrometry assay response. Anal Chem. 2010;82(23):9671–9677. doi: 10.1021/ac1013018 [DOI] [PubMed] [Google Scholar]
- 19.Fu Y, Li W, Picard F. Non-regulated LC-MS/MS bioanalysis in support of early drug development: a Novartis perspective. Bioanalysis. 2023;15(3):109–125. doi: 10.4155/bio-2022-0204 [DOI] [PubMed] [Google Scholar]
- 20.Zhang J, Selaya SD, Shakleya D, et al. Rapid quantitation of four nitrosamine impurities in angiotensin receptor blocker drug substances. J Pharm Sci. 2023; 112(5), 1246–1254 doi: 10.1016/j.xphs.2022.12.005 [DOI] [PubMed] [Google Scholar]
- 21.Zhang J, Yang Y, Ashraf M, et al. An advanced automation platform coupled with mass spectrometry for investigating in vitro human skin permeation of UV filters and excipients in sunscreen products. Rapid Commun Mass Spectrom. 2022;36(11):e9273. doi: 10.1002/rcm.9273 [DOI] [PubMed] [Google Scholar]
- 22.Booth B, Stevenson L, Pillutla R, et al. 2019 White Paper On Recent Issues in Bioanalysis: FDA BMV Guidance, ICH M10 BMV Guideline and Regulatory Inputs (Part 2 - Recommendations on 2018 FDA BMV Guidance, 2019 ICH M10 BMV Draft Guideline and Regulatory Agencies' Input on Bioanalysis, Biomarkers and Immunogenicity). Bioanalysis. 2019;11(23):2099–2132. doi: 10.4155/bio-2019-0270 [DOI] [PubMed] [Google Scholar]
- 23.Spitz S, Zhang Y, Fischer S, et al. 2020 White Paper on Recent Issues in Bioanalysis: BAV Guidance, CLSI H62, Biotherapeutics Stability, Parallelism Testing, CyTOF and Regulatory Feedback (Part 2A - Recommendations on Biotherapeutics Stability, PK LBA Regulated Bioanalysis, Biomarkers Assays, Cytometry Validation & Innovation Part 2B - Regulatory Agencies' Inputs on Bioanalysis, Biomarkers, Immunogenicity, Gene & Cell Therapy and Vaccine). Bioanalysis. 2021;13(5):295–361. doi: 10.4155/bio-2021-0005 [DOI] [PubMed] [Google Scholar]
- 24.ICH . Validation of analytical procedures Q2(R2). 2023. https://database.ich.org/sites/default/files/ICH_Q2-R2_Document_Step2_Guideline_2022_0324.pdf
- 25.Shakleya D, Asmelash B, Alayoubi A, et al. Bumetanide as a model NDSRI substrate: n-nitrosobumetanide impurity formation and its inhibition in bumetanide tablets. JPharm Sci. 2023; 112(12), 3075–3087. doi: 10.1016/j.xphs.2023.06.013 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.