Background: Inter-laboratory proficiency schemes are widely used to control the performance of clinical and forensic toxicology laboratories. In 2016 the Laboratory of Environmental Hygiene and Forensic Toxicology - Venice (Italy) initiated an inter-laboratory proficiency test of blood-alcohol analysis. The number of participating laboratories gradually increased from 26 to 36. Furthermore, a few clinical laboratories were included if gas chromatographic (GC) methods were used for blood alcohol analysis. Procedure: Whole blood was obtained from the Blood Transfusion Centre of the Venice Hospital and a mixture of sodium fluoride and potassium oxalate was added as a preservative and anticoagulant, respectively. Aliquots of the blood were spiked with certified pure ethanol to obtain target blood-alcohol concentrations (BACs) ranging from 0 to 5.0 g/L. Two blood samples (4 mL each) were included in each shipment to the participating laboratories. The laboratories were asked to provide information about number of replicate BAC determinations they made, the types of ethanol reference standards used, and inherent measurement uncertainty. The aim of the testing was to obtain a mean consensus value for the target BAC and to assess inter-laboratory imprecision. All procedures for registration and submission of results were done on-line. A confidential report and statistical evaluations were returned to the participants one week later. Analytical methods: All participants used head-space GC (HS-GC) for the analysis of ethanol in blood. More than 85% of participants used HS-GC with flame-ionization detection, whereas the others used mass spectrometry (MS) as a detector. More than 40% of the participating laboratories kept the blood samples frozen (-20 degrees C) prior to analysis, whereas the others used refrigeration (+4 degrees C). The preliminary validation tests showed that there were no statistically significant differences between BAC in frozen or refrigerated samples for a period of 20 days. Results and conclusion: The statistical evaluation of results was done using an iterative procedure known as Algorithm A (ISO 13528:2015, C.3.1). This provides robust estimates for mean and standard deviation between laboratories and these were used as consensus values. More than 85% of participants provided satisfactory results (z-score amp;lt;1) and 94% of laboratories were within z-score amp;lt;2, based on five control samples. When a blood sample without any alcohol (blank) was sent for analysis, laboratories reported this as zero, 0.00 g/L, below limit of detection (LOD) or not detected. Some type of consensus should be reached for reporting blank samples. (C) 2018 Elsevier B.V. All rights reserved.