MSACL 2019 EU Abstract

Modelling the Within- and Between-run Variation in Internal Standard Signal to More Accurately Assess Ion Suppression Effects

Edmund H Wilkes, Matthew J Whitlock, Emma L Williams
Department of Clinical Biochemistry, North West London Pathology, Charing Cross Hospital, Fulham Palace Road, London, W6 8RF

Warning: Undefined variable $headshot in /var/www/html/view_abstract/view_abstract_in_program.php on line 704

Edmund Wilkes (Presenter) Imperial College Healthcare NHS Trust

Presenter Bio: Ed completed his BSc in Biochemistry at the University of Warwick, before going on to complete his PhD in Biochemistry and post-doctoral work at Barts Cancer Institute, Queen Mary University of London. Ed then trained as a Clinical Biochemist at University College London Hospitals and the Royal London Hospital and is now based within the Endocrinology and Mass Spectrometry sections within North West London Pathology. His main interests lie in the clinical applications of mass spectrometry and data science.

Relevant Financial Disclosures (within past 24 months) Salary Ultromics Ltd Conflict mitigation NOT REQUIRED.

Background:
Ion suppression and enhancement are commonly encountered issues with the use of electrospray ionisation (ESI) techniques. Determining which samples suffer from these issues is critically important in a clinical environment in order to avoid factitiously inaccurate results. Despite this, determining what constitutes significant ion suppression or enhancement is marred by the fact that internal standard (IS) signal may vary linearly over the course of an analytical run due to normal variation in the ESI plume’s intensity. As such, typical approaches that employ fixed thresholds that ignore a sample’s position in an analytical run may lead to both false positives and negatives. To overcome this, we applied robust linear mixed-effects models to capture the within- and between-run variability in IS signal in order to generate parametric quality control (QC) ranges for routine use.

Methods:
Fifty-nine analytical runs of an in-house LC-MS/MS assay for 25-hydroxyvitamin D3 and D2 were retrospectively collected. Each run consisted of 85 samples (excluding calibrators and QC material). The variation in the percentage deviation of the internal standard area from the mean of the calibrators was modelled through the use of robust linear mixed-effects models in the R statistical computing environment.

Results:
Two models were fitted to the data, one for each compound measured in the assay (25-hydroxyvitamin D3 and D2). The models were fitted with the “robustlmm” R package and revealed that the majority of total variance could be attributed to sample-specific deviations in IS signal (64.3% and 70.8% for D3 and D2, respectively). The remainder of the variance could be attributed to differences in mean IS deviation between runs (35.7% and 29.2%). The models demonstrated a significant positive drift in IS signal over the analytical runs, with coefficients of 0.120 (95% CI: 0.096, 0.143) and 0.228 (0.199, 0.257) for D3 and D2, respectively.

Conclusions:
Here, we have described the application of robust linear mixed-effects models to extensively characterise the variation in IS signal both within- and between-runs in a routinely performed LC-MS/MS assay. These models allowed us to better capture the variation in IS signal as a function of sample position in the run. Taken together, this approach facilitates more robust QC assessment of ion suppression effects in routine clinical practice and therefore improves patient safety. Importantly, this approach is applicable to other MS-based assays that make use of ESI.