Quantification of microdialysis related variability in humans

Objective: Target-site concentrations obtained via the catheter-based minimally invasive microdialysis technique often exhibit high variability. Catheter calibration is commonly performed via retrodialysis, in which a transformation factor, termed relative recovery (RR), is determined. Leveraging RR values from a rich data set of a very large clinical microdialysis study, promised to contribute critical insight into the origin of the reportedly high target-site variability. The present work aimed (i) to quantify and explain variability in RR associated with the patient (including non-obese vs. obese) and the catheter, and (ii) to derive recommendations on the design of future clinical microdialysis studies. Methods: A prospective, ageand sex-matched parallel group, single-centre trial in non-obese and obese patients (BMI=18.7-86.9 kg/m(2)) was performed. 1-3 RR values were obtained in the interstitial fluid of the subcutaneous fat tissue in one catheter per upper arm of 120 patients via the retrodialysis method (n(RR)=1008) for a panel of drugs (linezolid, meropenem, tigecycline, cefazolin, fosfomycin, piperacillin and acetaminophen). A linear mixed-effects model was developed to quantify the different types of variability in RR and to explore the association between RR and patient body size descriptors. Results: Estimated RR was highest for acetaminophen (69.7%, 95%CI=65.0% to 74.3%) and lowest for piperacillin (40.4%, 95%CI=34.6% to 46.0%). The linear mixed-effects modelling analysis showed that variability associated with the patient (sigma=15.9%) was the largest contributor (46.7%) to overall variability, whereas the contribution of variability linked to the catheter (sigma=5.55%) was similar to 1/6 (16.8%). The relative contribution of residual unexplained variability (sigma=12.0%, including intracatheter variability) was similar to 1/3 (36.4%). The limits of agreement of repeated RR determinations in a single catheter ranged from 0.694-1.64-fold (linezolid) to 0.510-3.02-fold (cefazolin). Calculated fat mass affected RR, explaining the observed lower RR in obese (Delta RRmean=-29.7% relative reduction) versus non-obese patients (p<0.001); yet only 15.8% of interindividual variability was explained by this effect. No difference in RR was found between catheters implanted into the left or right arm (p=0.732). Conclusions: Three recommendations for clinical microdialysis trial design were derived: 1) High interindividual variability underscored the necessity of measuring individual RR per patient. 2) The low relative contribution of intercatheter variability to overall variability indicated that measuring RR with a single catheter per patient is sufficient for reliable catheter calibration. 3) The wide limits of agreement from multiple RR in the same catheter implied an uncertainty of a factor of two in target-site drug concentration estimation necessitating to perform catheter calibration (retrodialysis sampling) multiple times per patient. To allow routine clinical use of microdialysis, research efforts should aim at further understanding and minimising the method-related variability. Optimised study designs in clinical trials will ultimately yield more informative microdialysis data and increase our understanding of this valuable sampling technique to derive target-site drug exposure.