LC-HRMS has nowadays become the “gold standard” for identification of xenobiotics in clinical and forensic toxicology providing both high selectivity and specificity. Sample preparation and introduction nevertheless remain a key step and should be simple, universal, reproducible and sensitive for a broad range of analytes. Two types of sample introduction were tested for drugs spiked in biological samples: on-line extraction by turbulent flow chromatography (TurboFlow®) versus direct injection. Then, parameters of protein precipitation such as organic solvent, sample/organic solvent ratio and evaporation after precipitation were evaluated. In all, 62 compounds (drugs of abuse, pharmaceuticals and metabolites) were selected according to several criteria such as polarity, log P, retention time, occurrence in drug screening and divided in two groups regarding their therapeutic (A: n = 20) or regulatory (B: n = 42) concentrations. Plasma, blood and urine samples were prepared by spiking different concentrations (0.25 to 1000 μg/L, n = 6) into blank matrices. TurboFlow® and LC analyses were performed using a Transcend II TLX-1 and Vanquish systems. Data were acquired on an Exploris 120 Orbitrap Mass spectrometer and processed with TraceFinder 5.1 software (ThermoFisher Scientific®, San Jose, USA). The analytical workflow was based on previous publication (Helfer, J Chrom B, 2017, 1043, 138–149) with minor modifications: Phenyl replaced by Fluoro Turboflow column. 400 μL of sample was mixed with 600 μL of acetonitrile for precipitation. After shaking and centrifugation, the supernatant was injected. Several parameters were assessed for protein precipitation: – Sample/organic solvent ratio: 1:1,5; 1:2 and 1:3; – Organic solvent: acetonitrile, methanol, acetonitrile plus 0.1% formic acid, acetonitrile:methanol (50:50, v:v), methanol plus 0.2 M ZnSO4 (80:20, v:v) and acetonitrile with 80 mg QuEChERS extraction salts; – Concentration: direct injection or evaporation after precipitation. The performances were evaluated regarding the limits of identification (LOI), reproducibility at the respective LOI of each compound and linearity. Reproducibility gave satisfactory results for 59 and 56 compounds in direct injection and in TurboFlow® in all matrices (CV < 15%). Linearity was acceptable in both injection modes with no exception (R2 > 0.95). Finally, sensitivity appeared to be the most differentiating factor, direct injection showing lower limits for 29 (plasma), 31 (blood) and 28 (urine) compounds. For example, identification was possible at 10 μg/L in plasma for gabapentin, metformin and baclofen in direct injection and at 100 μg/L in TurboFlow®. For protein precipitation, 1:1.5 and 1:2 ratios were statistically equivalent in term of sensitivity regardless of the matrix, whereas 1:3 ratio showed an average decrease of 48%, 35% and 22% of pic areas in plasma, blood and urine samples respectively. The LOI obtained with different organic solvent were identical except for few compounds. Acetonitrile provided lower LOI for sufentanyl (0.25 vs. 1 μg/L for other solvents) whereas methanol in urine showed better results for zopiclone (1 μg/L vs. 5 μg/L). Finally, the addition of QuEChERS salt and evaporation after precipitation doubled at least the average signal intensity for all drugs. Despite some benefits such as large injection volume and automation, the sensitivity of the method including TurboFlow® extraction did not show improvements compared with direct injection. Indeed, small hydrophilic molecules had generally less retention on TurboFlow® column resulting in lower recovery and signal suppression can occur due to many matrix components. The addition of formic acid for protein precipitation does not improve the sensitivity, neither ZnSO4 nor mix of solvents. Simple and providing adequate sensitivity, the use of acetonitrile as organic solvent for protein precipitation with a ratio of 1:2 was chosen for the 3 matrices for routine. The addition of QuEChERS salts and evaporation after precipitation could be dedicated to specific cases where very low LOI has to be reached.