key: cord-0761902-vyo27mp1 authors: Reckers, Andrew; Wu, Alan H B; Ong, Chui Mei; Gandhi, Monica; Metcalfe, John; Gerona, Roy title: A combined assay for quantifying remdesivir and its metabolite, along with dexamethasone, in serum date: 2021-04-17 journal: J Antimicrob Chemother DOI: 10.1093/jac/dkab094 sha: 93211aca60dfa4f379c4b8ea38a656a4d9b20161 doc_id: 761902 cord_uid: vyo27mp1 BACKGROUND: As global confirmed cases and deaths from coronavirus disease 2019 (COVID-19) surpass 100 and 2.2 million, respectively, quantifying the effects of the widespread treatment of remdesivir (GS-5734, Veklury) and the steroid dexamethasone is becoming increasingly important. Limited pharmacokinetic studies indicate that remdesivir concentrations in serum decrease quickly after dosing, so its primary serum metabolite GS-441524 may have more analytical utility. OBJECTIVES: We developed and validated a method to quantify remdesivir, its metabolite GS-441524 and dexamethasone in human serum. METHODS: We used LC-MS/MS and applied the method to 23 serum samples from seven patients with severe COVID-19. RESULTS: The method has limits of detection of 0.0375 ng/mL for remdesivir, 0.375 ng/mL for GS-441524 and 3.75 ng/mL for dexamethasone. We found low intra-patient variability, but significant inter-patient variability, in remdesivir, GS-441524 and dexamethasone levels. CONCLUSIONS: The significant inter-patient variability highlights the importance of therapeutic drug monitoring of COVID-19 patients and possible dose adjustment to achieve efficacy. Coronavirus disease 2019 (COVID- 19) , which is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has resulted in the death of over 2.2 million people worldwide as of 15 October 2020. 1 Two of the most promising treatments for reducing morbidity and mortality are the antiviral drug remdesivir (GS-5734, Veklury) and the steroid dexamethasone. Remdesivir is an adenosine nucleotide prodrug of the monophosphate GS-441524, the primary metabolite measured in human serum. 2, 3 In host cells, GS-441524 is phosphorylated into the active triphosphate metabolite, which inhibits the RNA polymerase activity of coronaviruses. While remdesivir has not been proven to significantly reduce mortality, it can reduce the duration of hospital stay in severe cases of COVID-19. [4] [5] [6] Quantifying GS-441524 is useful for understanding the pharmacokinetics of remdesivir, as previous methods have shown that remdesivir is rapidly metabolized within 24 h to levels below the limit of quantification. 7, 8 Dexamethasone, a glucocorticoid, is the first COVID-19 treatment found to reduce mortality in clinical trials. 9 Dexamethasone reduces the transcription of several inflammatory agents, potentially decreasing the severity of the innate inflammatory pathways that can lead to organ failure and death. 10 Current NIH guidelines recommend the use of both remdesivir and dexamethasone in severe COVID-19, making their simultaneous quantification important. 11 Low levels of remdesivir due to drug-drug interactions or pharmacogenomic variation may partially explain the mixed clinical efficacy of the drug in various trials. 12, 13 Thus, in an attempt to improve the breadth and analytic sensitivity of published methods for therapeutic drug monitoring, 14 we developed and validated an LC-MS/MS assay that simultaneously quantifies remdesivir, its primary metabolite GS-441524 and dexamethasone. We purchased reference standards for remdesivir and GS-441524 (both 98% purity) from Aobious, for dexamethasone (98% purity) from Cayman V C The Author(s) 2021. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For permissions, please email: journals.permissions@oup.com. Chemical and for the internal standard dapivirine-d11 (96% chemical purity, <98% isotopic purity) from Santa Cruz Biotechnology. All solvents were HPLC-grade. We purchased water from Aqua Solutions, Inc., acetonitrile (ACN) and methanol (MeOH) from Honeywell Burdick and Jackson, and DMSO from Fisher Scientific. The drug-free serum we used is a product from UTAK Laboratories. Remdesivir, dexamethasone and dapivirine-d11 were prepared at 1 mg/mL in MeOH and GS-441524 was prepared at 1 mg/mL in DMSO. All stock solutions were stored at #80 C. Intermediate mixes were prepared with 1:1 MeOH: We used a protein precipitation method followed by evaporation. Similar methods have been published elsewhere. [14] [15] [16] [17] [18] To 50 lL of serum we added 100 lL of 7.5 lg/mL dapivirine-d11 in 1:1 MeOH: H 2 O (v/v). We precipitated the proteins by adding 600 lL of cold 1:1 MeOH:ACN followed by vortex mixing and centrifugation (2800 g for 10 min). We evaporated the supernatant and reconstituted it in the same volume of H 2 O. We injected 10 lL of extract into an LC-MS/MS system (Agilent LC 1260-AB Sciex API 5500, Agilent Technologies, Santa Cruz, CA, USA and AB Sciex, Foster City, CA, USA) and used positive electrospray ionization in multiple reaction monitoring mode. We separated the analytes using an Agilent Poroshell 120 EC-C18 column (3 % 50 mm, 2.7 lm particle size) with gradient elution. H 2 O and ACN were mobile phase A and B, respectively (Table 1) . Following the 14 calibration points, we injected two blanks and then low (0.6 ng/mL remdesivir, 12 ng/mL GS-441524 and 24 ng/mL dexamethasone), mid (12 ng/mL remdesivir, 120 ng/mL GS-441524 and 120 ng/mL dexamethasone) and high (120 ng/mL remdesivir, 1200 ng/mL GS-441524 and 1200 ng/mL dexamethasone) quality controls (QCs). Passing runs consisted of three sets of QCs within 15% accuracy and 15% precision, run at the beginning, middle and end of a sample batch. We analysed the data using AB Sciex Analyst 1.6.3 and AB Sciex MultiQuant 2.1. We used the two most abundant fragment ion transitions and retention time to confirm peak identity. The area of the reference drugs over the area of dapivirine-d11 was used for quantification. We used 1/x weighted linear regression for the calibration curves. We evaluated the linearity, sensitivity, precision, accuracy, matrix effect, recovery, dilution effect, injection repeatability, carry-over and specificity of the method. We also tested the stability of unextracted samples using this method. On three separate days, we ran a calibration curve (0.015-135 ng/mL for remdesivir, 0.15-1350 ng/mL for GS-441524 and 0.15-1350 ng/mL for dexamethasone). Passing criteria included an r value 0.95 and 75% of calibration points within ±20% accuracy. The limit of detection (LOD) was defined as the lowest concentration at which the signal to noise ratio was 3. The lower limit of quantification (LLOQ) was the lowest point with a signal to noise ratio 10 that maintained an r value 0.95. We tested precision and accuracy by thrice preparing and running five replicates of spiked drug-free serum at the low, mid and high QC levels alongside a calibration curve. We assessed the intra-and inter-day imprecision via coefficient of variation (CV) and the intra-and inter-day accuracy via relative error (RE). We tested the matrix effect by spiking low, mid and high QC levels into H 2 O and comparing the concentrations with the same levels spiked into drug-free serum and then extracted. We tested recovery by spiking low, mid and high QC levels into extracted matrix and comparing the concentrations with the same levels spiked into drug-free serum and then extracted. Due to the high matrix enhancement effect observed in remdesivir, we also analysed the matrix effect of remdesivir in five replicates of six patient samples that did not have detectable levels of remdesivir. We spiked remdesivir into the six patient samples and the drug-free serum at the three QC levels and compared the precision of the matrix effect within the replicates and between the seven serum types. Due to the low quantitative range of remdesivir in the assay, we tested the effect of diluting samples with concentrations above the upper limit of quantification (ULOQ). We extracted five replicates of samples spiked 20fold higher than the three QC levels and then diluted these samples 20-fold with extracted matrix blank. We assessed the precision (CV) and accuracy (RE) of these 15 samples to determine the dilution effect. We tested injection repeatability by injecting a vial of low, mid and high QC levels five times in a row and assessing the CV. We tested carry-over by injecting the high QC once, twice and then thrice, each time preceded and followed by three matrix blank injections. For specificity, we ran drug-free serum spiked with 20 common cold, flu and antiviral drugs at low and high levels of the drug's therapeutic range, or the therapeutic range for a similar drug if no published therapeutic range was found (Table 2) . A peak was considered to interfere with the target analyte if it fell within 0.2 min of the established retention time. We also analysed specificity of the patient samples obtained. We assessed the stability of unextracted samples at room temperature, in an ice bath (kept at approximately 2 C) and after three freeze-thaw cycles. For all three treatments, GS-441524 and dexamethasone were run at the low, mid and high QC levels. Remdesivir was run at elevated levels (2000, 1000 and 200 ng/mL) to replicate the higher remdesivir levels in patients immediately following infusion and then samples were diluted post-extraction with extracted matrix. All levels were run in triplicate and analytes were considered stable if the final timepoint concentration was within 20% of the original concentration. For the room temperature and ice bath experiments, we spiked drug-free serum with the analytes, capped the tubes and placed them in the appropriate storage area. We extracted at 0, 1, 2, 4, 6 and 8 h. In addition to the 8 h stability experiments, we also conducted a room temperature stability experiment of 48 h. We spiked remdesivir at the higher concentrations described earlier and spiked GS-441524 and dexamethasone at 20-fold higher than the QC levels. Following extraction at 0, 8, 24, 36 and 48 h, we diluted the samples 20fold with extracted matrix. For freeze-thaw stability, samples were spiked as described above, frozen at #80 C and then thrice removed, thawed and returned to the freezer. To control for stability at #80 C, split samples were stored at #80 C in parallel with the freeze-thaw experiment and sampled at each thaw cycle (0, 1, 2 and 4 h). We obtained 26 remnant serum samples for up to 6 days from seven hospitalized patients seen at Zuckerberg San Francisco General Hospital between 23 September 2020 and 30 September 2020. These patients were confirmed to be positive for SARS-CoV-2 using routine molecular diagnostic methods. As part of their medical care, they were treated with remdesivir, dexamethasone or both. The Institutional Review Board (IRB) of the University of California, San Francisco, approved the use of remnant samples and review of medical and pharmacy records without consent. The samples were de-identified prior to delivery to the testing laboratory. As a condition of the IRB approval, results of the drug levels were not made known to either the patient or attending medical staff. COVID-19 restrictions meant that we did not always have access to patient samples immediately following sample collection. Samples were stored for 2-5 days in the refrigerator and then for up to a month at #20 C before being extracted. This delay in receiving samples meant that we did not implement a sample inactivation procedure, as we would not have been able to inactivate immediately following sample collection. We developed a 10 min method to quantitatively analyse remdesivir (GS-5734), its primary plasma metabolite GS-441524 and dexamethasone in serum using LC-MS/MS. For the three target analytes and the chosen internal standard (dapivirine-d11), we analysed two mass spectral transitions ( Figure 1 ). We used the most abundant transition as the quantifier ion and the other transition as the qualifier ion (Table 3) . We validated the method using a 13-point calibration curve and three QC levels spanning the low, mid and high ends of the linear range. For remdesivir, the LOD and LLOQ were 0.0375 ng/mL and the ULOQ was 135 ng/mL. For GS-441524, the LOD and LLOQ were 0.375 ng/mL and the ULOQ was 1350 ng/mL. For dexamethasone, the LOD and LLOQ were 3.75 ng/mL and the ULOQ was 1350 ng/mL. The average linearity coefficients of determination for remdesivir, GS-441524 and dexamethasone were 0.998, 0.996 and 0.997, respectively. In assessing precision, all three analytes had average within-and between-run CVs <8%. Precision was lower at the low QC level than at the mid and high QC levels. In assessing accuracy, all three analytes had average within-and between-run REs <14%. See Table 4 . The matrix greatly enhanced the remdesivir signal. At low, mid and high QC levels, the matrix effect was 601.7%, 787.5% and 634.5%, respectively. The CV at each level was <12%. The matrix effect was less pronounced for GS-441524 and dexamethasone. The matrix effect of GS-441524 at the low, mid and high QC levels was 2.1%, #2.2% and #2.4%, respectively. The CV at each level was <4%. The matrix effect of dexamethasone at the low, mid and high QC levels was 14.7%, #10.6% and #15.8%, respectively. The CV at each level was <12% ( Table 5 ). The precision of the remdesivir matrix effect was similar within the five replicates and between the six patient samples and drugfree serum sample. The CV at the low, mid and high QC levels was 3.0%, 4.5% and 2.0%, respectively. The precision of the five replicates for each QC level was <8% for all seven serum types ( Table 6 ). The recovery of remdesivir at the low, mid and high QC levels was 86.7%, 87.1% and 85.6%, respectively. The CV at each level was <6%. The recovery of GS-441524 at the low, mid and high QC levels was 85.3%, 92.7% 88.9%, respectively. The CV at each level was <5%. The recovery of dexamethasone at the low, mid and high QC levels was 88.4%, 86.4% and 82.1%, respectively. The CV at each level was <8% (Table 5 ). Dilution did not significantly affect the method precision and accuracy, as CV across the three QC levels was <10% and RE was <12% (Table 7) . The CV for each QC level was <4% for all analytes. No signal above the threshold signal-to-noise ratio of three was detected in any of the matrix blank samples for GS-441524 or dexamethasone. The remdesivir signal was observed, but decreased after one matrix blank injection. No interfering signal for remdesivir, GS-441524 or dexamethasone was detected at the correct retention time after running low and high concentrations of 20 other common cold, flu and antiviral drugs. Remdesivir, GS-441524 and dexamethasone were quantified in 23 samples obtained from seven COVID-19 patients. Remdesivir was given in six of the seven patients (Table 11) . We did not observe remdesivir or GS-441524 in samples of the patient that did not receive remdesivir (3A-3E) as well as in a sample collected from a patient prior to remdesivir administration (7A). Dexamethasone was given in five of seven patients. We did not observe dexamethasone in the two patients that were not administered dexamethasone (5 and 6), verifying the specificity of our method in clinical samples. After their administration remdesivir and dexamethasone were