(P-080) Validation of Self-reported Medication Use Applying Untargeted Mass Spectrometry-based Metabolomics Techniques in the Rhineland Study

Background: Quality of self-reported drug use is often questioned because it may be prone to several forms of bias. Secondary data may not adequately represent actual drug intake, and validation of self-reported drug use with secondary medication data has given conflicting results. Drug metabolites measured in plasma can give a better, unbiased, estimate of actual drug intake.

Objectives: To assess the validity of self-reported regular drug use with drug metabolites measured in plasma by using an untargeted mass spectrometry-based approach.

Methods: This study was based on the Rhineland Study, an ongoing population-based cohort study in Bonn, Germany. Participants (aged ≥30 years) undergo in-depth phenotyping, including assessment of drug use and untargeted metabolomic profiling in plasma. For all detected drug metabolites, we assessed their suitability to reflect regular drug intake based on pharmacokinetics. We only included drugs with ≥15 self-reported users. We assessed the concordance between self-reported drug use and measured metabolites by calculating Cohen’s kappa (𝜅) with 95% confidence intervals (CI). Additionally, we performed the same analyses stratified for sex and age ( <65 versus ≥65 years).

Results: Analyses were based on 4,386 participants of whom metabolomics and self-reported drug use data were available (mean age 55.0 years, standard deviation 14.0 years; 56.1% women). Out of 13 analysed drugs, 11 showed an almost perfect agreement (𝜅>0.8). For sitagliptin agreement was only slightly lower with 𝜅=0.80 (46 users; CI: 0.71-0.90). For hydrochlorothiazide agreement was only 0.61 (355 users; CI: 0.56-0.66), as for 187 self-reported users we did not detect metabolites. A possible explanation for the latter may be that those participants decided not to use their diuretic when going out for the day. Eight drugs were frequently enough used in both sexes to allow for sex stratified analyses. For 5 of those, concordance rates were higher in women compared to men, yet we found only a significant difference for allopurinol with higher concordance in men (women: 15 users; 𝜅=0.68; CI: 0.52-0.83; men: 77 users; 𝜅=0.89; CI: 0.84-0.95). Age stratified analyses could be performed on 9 drugs. For 7 of those, concordance rates were higher for individuals aged ≥65 years compared to individuals aged <65 years. However, in all instances the CIs of the 𝜅-values overlapped, indicating no significant age-related differences.

Conclusions: The high concordance rates between self-reported drug use and metabolites measured in plasma, supports that self-reported drug use is a reliable and accurate method to assess drug intake.