Biobanking of patient samples plays a critical role in advancing medical research, enabling the development of personalized therapies and improving diagnostic tools. However, obtaining blood samples from polytrauma patients, particularly during emergency room visits, can be challenging due to the urgency of treatment and potential difficulty in accessing veins. As part of a biobank managing routine sample collection, we have become increasingly aware of visual differences in samples, highlighting the importance of careful monitoring for potential contaminants that may affect downstream analyses. The goal of this study was to visually inspect routinely collected serum biobank samples and analyse them using spectrophotometry for hemolysis, lipemia, and bilirubin contaminants. The focus was to identify contaminated samples and determine if these contaminants could be linked to patient factors. Additionally, the study aimed to assess whether contaminants affect the use of these samples in downstream analyses, like EV and EV-miRNA studies.

For serum sample quality control, we chose optical absorption analysis, a fast method that does not require complex equipment or specialized personnel. It has been previously used to detect hemolysis, icterus (bilirubin), and lipemia interference in serum and plasma samples and several modern biochemistry analysers have been developed for clinical use [9].

Hemolysis is well known the most common pre-analytical error affecting biochemical analyses of liquid biopsy samples. Hemolysis was the most common quality issue, occurring in 31.8% of our patients’ samples. Most hemolytic samples were from the ER, consistent with literature reporting that 31% of ER samples are hemolytic [10]. In trauma patients several exogenous factors, like transfusion via an intraosseous administration [11], transfusion of uncross-matched red blood cell units [12], the use of veno-venous extracorporeal membrane oxygenation are discussed to be associated with the development of hemolysis [13]. Additionally trauma itself could lead to the hemolysis in these patients for example due to damaged pulmonary epithelium after severe chest trauma [14]. The literature shows that more critically injured patients, non-survivors, and those with longer ventilator use have higher hemolysis levels, with these markers also linked to increased red blood cell transfusions [15]. We did not find any association of hemolysis with the ISS, with the amount of transfused red blood cells or with the changes in hemoglobin levels in our trauma patients. In addition to pathophysiological changes, factors related to blood sample collection and processing can also contribute to hemolysis [16, 17]. Although the sampling, transport, and processing steps in our serum/plasma NTF biobank are standardized and performed by trained, experienced personnel to prevent in vitro hemolysis [18], it cannot be entirely excluded due to the complexities of sample collection from trauma patients. These complexities arise from factors such as the frequent use of intravenous catheters, prolonged tourniquet times during blood withdrawal in the emergency room, and circulatory depression in trauma patients.

Hemolysis is known to affect laboratory testing by releasing intracellular from damaged erythrocytes [19]. With the focus on the EV biobank we aimed to determine if hemolysis in serum samples also impacts EV (EV-miRNA) analyses. We compared miR-16-5p expression in EVs isolated from hemolytic, lipemic and non-affected serum samples. This miRNA was chosen due to its high expression in red blood cells and its plasma/serum concentration being influenced by RBCs contamination [20, 21]. Our findings demonstrated, that miR-16-5p was significantly increased in EVs from hemolytic samples (Fig. 4) compared to lipemic and control samples, suggesting hemolysis affects both serum miRNA and EVs miRNA measurements. These findings indicate that miRNA-16-5p should not be used as a normalization miRNA in plasma/serum studies [22, 23].

Lipemia can interfere with serum/plasma biochemical analyses by altering light scattering, uneven distribution of analytes between aqueous and lipid phases, or interacting with assay reagents [9, 24, 25]. Lipemia was detected in one or more samples collected during 10 days after trauma in 15.9% of our patients. The most common cause of lipemia is non-fasting sample collection, which is unavoidable in polytrauma patients in the ER [2] or those receiving lipid-rich parenteral nutrition [26]. In our patient cohort, no link between parenteral nutrition and lipemia was found. However, this should be interpreted with caution due to the small number of patients on parenteral nutrition, the high number of fasted patients, and the misalignment between blood sample collection and the nutrition schedule. Lipemia can also result from medications like propofol, commonly used in trauma patients. Propofol-associated hypertriglyceridemia is frequent in mechanically ventilated ICU patients [27], and it has been linked to life-threatening propofol infusion syndrome in about 4% of trauma patients at level 1 trauma centers [28]. In the present patient collective, most received propofol anaesthesia on the first day after trauma, which coincide with the highest percentage of lipemic samples. This suggests a potential association between propofol use and the appearance of the turbid lipemic samples, warranting further investigation. Additionally, comparing TAG levels between lipemic and non-lipemic polytrauma samples revealed that TAG concentration at ER was almost two-fold higher in the lipemic group. The literature underscores the significance of circulating lipid analytes as prognostic markers in polytrauma patients [29, 30]. Our findings of a significant presence of lipemic samples in the biobank, potentially linked to propofol or parental nutrition, emphasize the need for awareness of this contaminant in studies focusing on circulating lipids analytes.

Thought that lipoprotein particles (similar by physical characteristics with EVs) could interfere with EV measurements in lipemic samples we compared EV isolates from control serum samples and serum samples contaminated with parenteral nutrition SmofKabiven. Our results showed nutrition contaminants significantly affected NTA measurements, with both mean particle size and concentration being significantly higher in the contaminated group. These results align with previous studies, showing that low-density lipoproteins mimic plasma-derived exosomes and that current EV-isolation methods cannot fully remove lipoprotein particles [31]. This means that lipemic samples should be avoided in studies involving EV particle-size and/or particle-concentration analyses [32].

The third main topic of the present analysis was the biobank serum samples with increased bilirubin concentration. Hyperbilirubinemia or icterus has been frequently reported in severely injured patients and was associated with poor outcome [33]. The trend observed in our study linking injury severity with increased bilirubin concentration is consistent with the literature [33]. Bilirubin, a catabolic product of hemoglobin [34], hemoglobin levels were compared between patients with and without elevated bilirubin, but no difference was found. However, in one patient with severe hemorrhagic shock and the need for > 20 RBC units, a significant increase in bilirubin was observed in all collected samples over 10 days. Interestingly, we observed the greatest increase in bilirubin concentrations between the ER and day one in both groups of patients, coinciding with the maximum hemoglobin decrease. A decrease in hemoglobin and increase in billirubin have been linked to bleeding, subarachnoid hemorrhage [35] and higher mortality in traumatic brain injury patients [34]. Patients with extreme hyperbilirubinemia (above 12 mg/dl), caused by infection, sepsis, or hypoxic hepatitis have shown high mortality rates (up to 76%) [36].

The presence of bilirubin in serum samples itself is not expected to directly interfere with EV measurements and in the current literature, is not associated with changes in the EV-miR profile, so far.

Overall, our results reveal a relatively high percentage of hemolysis, lipemia, and bilirubin-contaminated samples in our biobank, highlighting the importance of pre-analytical testing of serum samples to ensure the accuracy of downstream analyses. We recommend screening biobank samples before inclusion in research studies using simple adsorption spectrum measurements. A suggested algorithm for evaluating the spectrophotometric results is provided in supplementary Figure S2. Nevertheless, excluding biobank samples that are positive for hemolysis and/or lipemia could introduce a bias into the sample cohort, especially given the high percentage of such samples. Considering the limited number of patient samples and the potential bias introduced by excluding contaminated samples, an alternative approach could involve incorporating additional steps in the study design that would allow these samples to be retained without compromising the analysis. For example, for lipemic samples and EV particle NTA analysis, the ratio of lipemic non-EV particles could be estimated using flow cytometry with fluorescently labelled EV-epitope antibodies and/or fluorescent labelling of DNA/RNA in the EV cargo. Our findings and the approach we employed are not limited to trauma patient samples and could be extended to other patient biobanks. Furthermore, they underscore the lack of homogeneity in biobank samples, raising important questions about the feasibility of extrapolating data obtained from a representative cohort to the entire biobank.

This study has several limitations. While we focused on hemolysis, lipemia, and bilirubin, other contaminants, such as protein impurities, may also be present. Additionally, we analyzed only one miRNA, though other miRNAs could also impact EV results. Future research should thoroughly investigate the causes of contamination in biobank samples, as well as their effects on EVs, miRNAs, and other analytes measured in biobanks.