Historically, DBSs have been the flagbearer for microsampling. The concept for the preservation of dried human biological samples as a spot is attributed to Ivar Christian Bang in 1913 ^[1][2][39,40]. Typically, these microsamples can be collected via a skin prick with a lancet or created by transferring from a phlebotomy tube with a micro-pipette ^[3][4][30,41]. Currently, such samples are collected on a specialised filter paper, which can contain anywhere from 15 to 50 µL of blood, and take approximately three to four hours to completely dry at room temperature ^[1][39]. This can make DBSs amenable to collections where cold-chain shipping protocols may not be feasible. However, it should be noted that below-zero temperatures (−20 °C, −80 °C) are still recommended for the long-term storage of DBSs ^[1][3][4][30,39,41].

DBS samples have been evaluated in mass spectrometry (MS)-based metabolic phenotyping and have been used in applications of cancer diagnostics ^[5][42], cancer treatments ^[6][43], air pollution ^[7][44], drug discovery ^[8][45], acidemia ^[9][46], and pyruvate kinase deficiency ^[10][47]. Despite this wide range of applications of DBS samples, there still remain concerns around their volumetric accuracy, particularly in metabolic phenotyping, where the standardisation of sample volume is required. For example, the uneven spread of collected blood across the spot can introduce metabolite variation due to differences in viscosity/haematocrit (hct %; the fraction of blood made up of red blood cells). This has been observed in DBS studies using liquid scintillation analysis, where sub-punches from an individual sample were taken and the number of metabolites measured varied vastly from sample to sample ^[11][48]. DBS collections are remarkably susceptible to variation based on an individual’s haematocrit. Varying haematocrit levels are of particular concern due to its propensity to undergo rapid changes in the body. This is a widely known phenomenon in DBS microsampling and is commonly referred to as the ‘haematocrit effect’ ^[12][49], where external factors such as dehydration, polycythaemia, anaemia, overhydration, kidney failure, or chronic inflammatory conditions can cause plasma volume perturbations. Additionally, pregnancy may also cause slightly decreased hct due to an increase in blood volume ^[3][13][30,50]. For reference, the normal range of hct is 36–48% for women and 42–52% for men ^[14][51]. Put simply, increases in hct beyond the normal range affect blood viscosity and therefore reduce the spread of the blood spotted on the carrier material, whereas decreases in hct (i.e., reduced viscosity) can create greater spread ^[15][52]. As such, the sample’s spotted area has a linear, inverse relationship with hct ^[12][13][49,50].

In MS-based studies, normalisation techniques have been employed in order to address the ‘haematocrit effect’ ^[16][53]. However, it remains one of the most prevalent challenges faced for translating the wide use of microsamples in metabolic phenotyping workflows. To date, studies have successfully performed DBS haematocrit normalisation through potassium content ^[17][54], using near-infrared (NIR) spectroscopy ^[18][55], haemoglobin measurement using non-contact diffuse reflectance spectroscopy ^[19][56], and using wax barriers on DBSs ^[20][57]. Despite the development of these normalisation techniques for DBSs, they are yet to be widely adopted in metabolic phenotyping workflows, with many literature examples not implementing a normalisation step in their protocol descriptions ^{[6][7][9][10][18][21][22][23]}[43,44,46,47,55,58,59,60]. Only two studies have openly reported hct normalisation of their DBS samples as part of their metabolic phenotyping workflow. The first was by Koulman et al., who analysed infant heel-prick DBS samples, utilising a volumetric and hct-independent Liquid chromatography–mass spectrometry (LC–MS) method ^[24][61]. This was performed by relatively expressing the extracted lipid intensity of a given DBS sub-punch to its summed intensity. Interestingly, this method revealed that lipid profiles in DBSs showed comparable or better precision to plasma and whole blood samples, which the authors propose could be attributed to the halting of the oxidative process in dried samples compared to traditional venous whole blood and plasma samples ^[24][61]. Another study successfully employed an automated haematology analyser for the analysis of hct in serum to normalise steroid concentrations obtained from traditional DBSs ^[25][36]. This correction, by Salamin et al., used the following equation ^[25][36]:

The circulating blood metabolome is a tightly controlled homeostatic system, where preanalytical variation (paper substrate, storage, packaging, transport) can unavoidably lead to inaccurate and possibly misleading results ^[30][64]. This is a current limitation in pre-analytical workflows for DBS microsamples, particularly when considering the inherently heterogenous nature of biological samples, where physiological conditions (i.e., hct) already contribute dynamic changes ^[31][65]. As such, efforts seeking to enhance accuracy, sensitivity, and specificity during the analytical phase are in vain if the technical aspects that underpin microsample workflows are not reproducible. Thus, a lack of reproducibility and accuracy is detrimental to achieving outcomes of Predictive, preventive, personalised, and participatory medicine (P4 medicine) using DBS samples in metabolic phenotyping pipelines ^[32][1].

Advancements in microsampling devices have allowed for improvements in blood collections by removing some of the inconsistencies experienced in DBS collections whilst maintaining the convenience of microsampling (Table 1). These devices can be classified into three broad classes: advanced dried samples; passive separation devices; and whole biofluid collectors, which will be discussed in the context of metabolic phenotyping below.

Advanced dried sample devices collect samples as a fluid and produce a dried sample either with a polymer tip (volumetric absorptive microsampling, “VAMS”—Neoteryx; Torrance, CA, USA) or carriers akin to DBS cards, such as the hemaPEN (Trajan; Melbourne, VIC, Australia), HemaXis DB10 (DBS System SA; Gland, Switzerland), Capitainer qDBS (quantitative dried blood spot) and B-Vanadate (Capitainer AB; Solna, Sweden), TASSO-M20 (HemoLink; Seattle, WA, USA), and HemaSpot HD and HF (Spot on Sciences; San Francisco, CA, USA) ^{[25][58][59][60][61][62][63][64][65][66][67]}[27,28,36,66,67,68,69,70,71,72,73]. These technologies improve upon standard DBS cards by providing accurate/volumetric aspiration and thus heamatocrit-independent sample collection over a wide range of microsample sizes (2.74–30 µL).

Passive separation devices allow for the in situ separation of whole blood into its sub-components. e.g., serum or plasma, which can then store the resultant product in liquid form, e.g., TASSO+ (HemoLink; Seattle, WA, USA); as dried serum, e.g., HemaSpot SE (Spot on Sciences; San Francisco, CA, USA), or as dried plasma, e.g., Tellimmune Plasma Separation Cards (Novilytic; West Lafayette, IN, USA). Passive separation devices represent an expanding area commercially, with many dried plasma spot (DPS) devices currently in development, including the DPS (Capitainer AB; Solna, Sweden), Book-Type DPS (Q2 Solutions; Morrisville, NC, USA), and the Hemaxis DX (DBS System SA; Gland, Switzerland) ^[48][49][50][74,75,76]. It is not known how these devices will perform in metabolic phenotyping. However, as they produce samples akin to those obtained from venous whole blood separations, which are commonly used in metabolic phenotyping, they warrant further investigation in the field.

Whole biofluid collectors are advanced devices with the ability to collect and produce samples as liquid samples without the need for cellulose material. These devices can collect sample volumes as small as 23 µL, e.g., the MSW2 (Shimadzu; Kyoto, Japan), and extend to 100 µL, e.g., the TAP (Yourbio Health; Medford, MA, USA), and up to 600 µL, e.g., the TASSO+ and TASSO-SST (HemoLink; Seattle, WA, USA) ^[37][68][69][10,77,78].

All three broad classes of advanced microsampling devices are well positioned for direct implementation in metabolic phenotyping workflows in clinical and epidemiology studies. For example, they are already commercially available and therefore have advanced manufacturing consistency; they have been designed to counter specific challenges in DBS and traditional microsampling workflows, including the ‘haematocrit effect’ ^[3][65][30,71], and have already achieved translation to non-metabolic phenotyping analytical chemistry protocols ^[58][70][27,79]. Despite these benefits, the translation of microsampling devices to metabolic phenotyping research is yet to be widely adopted by the field, as extensive studies investigating comparability to venipucture and metabolite stability are lacking ^[71][80]. This is important as validated microsampling methods that leverage the advances in device design have the potential to enhance metabolic phenotyping studies in clinical and epidemiology settings, facilitating greater sampling frequency and sample size, and thereby providing valuable gains in statistical power ^[71][72][80,81].