In toxicological analysis, a major evolution has been made through the combination of increasingly sensitive and specific methods and the use of microsamples [36]. Among the microsampling systems, DBS have been used in different fields of biology since the 1960s [37]. In toxicology, DBS were successfully applied to detect illicit drugs including natural and synthetic cannabinoids [26,38], opiates [39], 3,4-methylenedioxymethamphetamine (MDMA) [40], γ-hydroxybutyric acid (GHB) [41] and cocaine [42]. Despite the many advantages of DBS, their main limit remains in the uncertainty of the sampling volume due to hematocrit value variations. To overcome this drawback, in a study involving healthy volunteers, the authors did not consider the hematocrit value as it was within the physiological norms for all participants [42]. On the other hand, a study on the quantification of MDMA does not report this issue [40], while a 2010 paper lists all of the limits of using DBS for quantitative studies, especially the uncertain volume of the sample [43]. The development of a new microsampling system, VAMS, which is not, or less, affected by hematocrit values, has allowed for the development of quantitative methods. These methods have been rapidly applied in different fields, especially in therapeutic drug monitoring [30,31,32]. In toxicology, several publications report the quantification of various drugs from the VAMS system, focusing on the quantification of one drug and its metabolites such as cocaine [42], or on a family of compounds such as the cathinones [35]. In 2021, a publication reported on a toxicological screening using the VAMS system but was limited to the identification and quantification of 18 compounds [34]. The use of the LC-MS/MS approach in these different studies may explain the limited number of identified molecules. Furthermore, it is well known that LC-MS/MS does not allow for such extensive detection as when LC-HR/MS is used. In this work, the use of LC-HR/MS allowed for the detection of over 1500 molecules, compared to a reference library, upon the analysis of a 20 µL extract of whole blood collected using the Mitra^TM device. This study is the first to combine HR/MS and VAMS microsampling for extended toxicological screening purposes. Indeed, only Joye et al. [44] have used Orbitrap™ High Resolution for extended screening from whole blood DBS in clinical and forensic toxicology.

In the framework of toxicology, identification is a key challenge and represents a bottleneck for a screening procedure. Furthermore, this task is much more challenging if a single analysis is possible due to the matrices’ availability, as is the case of screening using VAMS, which has to deal with a one-shot analysis. In this context, to ensure a reliable toxicological screening using the LC-HR/MS-Mitra^TM devices combination, the proper extraction of molecules from the tips is a fundamental step of the pre-analytical process. To optimize this step, we investigated different solvent mixtures and conditions to efficiently extract 90 molecules, selected as the most commonly detected drugs in clinical toxicological screening. In the literature, methanol is frequently reported to extract molecules from various matrices collected from Mitra^TM devices [25,26,27,28,29,30,31,32,33,34]. In some cases, an aqueous solution or buffer alone was used but supplemented with solid-liquid extraction [42], microextraction [35] or online extraction [39]. Our study highlighted that the use of an aqueous solution alone or a mixture of organic solvents alone does not lead to the highest extraction yields. A good compromise was obtained using a mixture of 30% organic phase and 70% water containing 1% formic acid. This mixture was adapted from a publication by Mandrioni et al. [33] in which we replaced 1% phosphoric acid with 1% formic acid. The use of an aqueous solution without any organic component allows the extraction of water-soluble compounds, such as metformin or baclofen, but is not optimal for extracting highly lipophilic molecules such as irbesartan. For these molecules, the use of an organic phase alone for extraction should, theoretically, result in better extraction yields compared to our mixture. However, the optimization step of our work has demonstrated that this is not really the case. One of the reasons could be related to the precipitation of hemoglobin by the solvents, which causes sequestration of the molecules within the precipitate; thus reducing extraction yield. On the other hand, it has already been reported that the addition of an aqueous component to the solvents can facilitate the extraction of some types of molecules [45], strongly supporting the results of this work.

Many publications use a combination of agitation and sonication steps to extract molecules from micro samplings. We also used these sample preparation steps, with the times optimized at 30 min for sonication and agitation. These times are much longer than those reported in the literature, in the order of a few minutes for the two steps [33,35,39,41,42]. It is well known that the optimization of extraction conditions for DBS or VAMS is complex and depends on the molecules to extract [46]. During a toxicological screening, the number of molecules to extract is potentially very high, with different chemical structures and physicochemical properties. As reported by Joye et al. [44], the diversity of the structures and properties of the molecules requires the use of the most versatile extraction process. The optimization of the extraction conditions is, therefore, a compromise, aimed at extracting molecules with the highest efficiency. In our conditions, 30 min of sonication and agitation were optimal to achieve the best yields for all the investigated molecules. It may, nevertheless, be regarded as a bottleneck point of the pre-analytical step, longer than those reported in the literature. Another limitation of our study includes the number of investigated molecules for the optimization of extraction conditions. Indeed, only ten molecules were assessed, but they were chosen adequately as they cover a broad array of the physicochemical properties of both licit and illicit drugs. Despite these limitations, the extraction yields obtained from the 90 compounds selected to establish the identification limits were all very satisfactory, with yields higher than 80% for all of the molecules identified in the 90 compounds tested. These yields are comparable to those reported in the literature [38,39,40,42]. Finally, in the clinical part of our study, the extraction yields calculated on the 119 identified compounds further confirm the excellent extraction yields, which are similar to those obtained from the 90 selected compounds.

Our extraction method and the analysis using LC-HR/MS enabled us to achieve identification limits lower than 12.5 ng·mL⁻¹ for 82.2% of the 90 tested molecules. These limits are significantly higher than those reported by Mestad et al. [35] using Mitra^TM devices and by Joye et al. [44] using DBS. In these two publications, the detection limits were lower than 5 ng·mL⁻¹ for the majority of the detected compounds. These values were established for the detection of drugs in cases of drivers under influence [35] or in a forensic context [44]. In the context of our study, our identification limits seem suitable for a reliable screening in clinical toxicology. Nevertheless, with our extraction process, some compounds, such as tetrahydrocannabinol (THC) and its metabolites, including THC-COOH, are not detected. These results are in line with those reported in the study performed by Mestad et al. [35]. On the opposite side, Joye et al. [44] demonstrated that it is possible to detect only THC-COOH from DBS after a double extraction process, including liquid extraction with a non-polar solvent. The lack of detection in our study cannot be attributed to the analytic method. In our study and in the study by Joye et al. [44], a screening using LC-HR/MS was developed, while LC-MS/MS was used in the publication of Mestad et al. [35]. As reported by Joye et al. [44], the percentage of THC-COOH detection in the samples varies depending on the type of support used to carry out the DBS, due to varying matrix effects based on the type of support. The absence of THC-COOH detection in our study may, thus, be due to poor extraction yield from the Mitra^TM device. Although reported in some publications [35], the lack of detection of cannabinoids and their derivatives remains a limiting point of our new screening method that needs to be optimized. The optimization of the pre-analytical step to extract these compounds from the Mitra^TM device remains a key point. The mixture extraction that we used contains a large part of water, which is not optimized for the extraction of very lipophilic compounds such as cannabinoid derivatives. In the near future, we will, thus, consider a two-step extraction to collect both hydrophilic and hydrophobic compounds; thus, allowing us to cover the broad array of molecular diversity responsible for intoxication, as recommended by Joye et al. [44]. This new pre-analytical approach should improve this lack of detection of cannabinoids.

A clinical application allowed us to demonstrate the feasibility of our new screening method. Fingertip sampling on the Mitra^TM device was performed by the nursing team upon admission of the patient, without the involvement of the laboratory team. The Mitra^TM devices were stored at room temperature on a special rack and extracted between 6 and 72 h after sampling. The stability of 90 molecules was evaluated over 72 h and no significant variation was observed. Our results are consistent with those of several other publications indicating excellent stability of the molecules, even the most fragile ones, such as cocaine, zopiclone and cathinones, on DBS [26,34,42,44] and on Mitra^TM [26,33,34,35,44]. The majority of the molecules are stable for one week at room temperature but can remain stable for several months when stored at −20 °C in alternative matrices, as previously reported [29].

The mass precision provided using LC-HR/MS in full-scan data-dependent MS2 (ddMS2) mode enables a more accurate identification of molecules during the screening procedure performed both on classical matrices [17,47,48] or alternative matrices such as DBS [42,44,49,50,51]. In routine laboratory use within a clinical toxicology context, this detection method is applied to plasma screening after the precipitation of proteins with acetonitrile containing various internal standards. To perform the screening on Mitra^TM, we transposed our routine screening method to whole blood. In the 15 patients included in our study, we compared the number of molecules detected among three samples: plasma, whole blood and Mitra^TM. From 20 µL of each of these samples, we obtained complete concordance in the detection between plasma and whole blood. The percentage of recovery was 98% between the whole blood and Mitra^TM. This result confirms the finding of Protti et al. [26], which showed an excellent overlap in the number of compounds detected between plasma and VAMS or between plasma and DBS. In one patient, no drugs were detected suggesting the absence of false positives and the specificity of our method. In the remaining patients, we identified between 1 and 15 compounds from various therapeutic classes, with different structures and physical properties, including a natural compound, namely ricinine. The diversity of the compounds detected in our study is similar to that reported by Joye et al. [44] using DBS. In two patients, diazepam was not confirmed according to our criteria because the signal intensity was below 10⁶. This poor detection of diazepam has been previously reported by Mestad et al. [35]. The authors attribute this to non-optimized extraction conditions for diazepam due to its physicochemical properties. In our study, we observed the same phenomenon in our two patients, despite using extraction conditions that were completely different from those of Mestad et al. [35]. Moreover, the good detection limit that we obtained for diazepam in overloaded solutions (2.5 ng·mL⁻¹) leads us to, rather, consider another explanation for this poor detection. In this study, it must be emphasized that the same molecules are detected in peripheral blood and capillary blood collected on Mitra^TM devices, suggesting a similar distribution of molecules between the two sampling points, as reported in several studies [52,53]. However, a hypothesis of the different distribution of diazepam between capillary and peripheral blood may explain the low detection of diazepam on Mitra^TM devices. This hypothesis will have to be verified with complementary studies to investigate other different distributions concerning other compounds not tested in this study. Another limitation of our study concerns the clinical part. In order to limit the impact of this study on the management of patients during admission, it was only possible to collect one Mitra^TM device for each patient, preventing us from testing the reliability of our collection conditions. However, to the best of our knowledge, this study is the first to compare toxicological screening performed on usual matrices, i.e., a plasma sample, to whole blood collected using a Mitra^TM device in real-life hospital intoxication cases. Even if only a single Mitra^TM device could be collected for each patient, the reliability of our approach has been illustrated thanks to the number of investigated patients, i.e., 15 cases.