Volume 60, Issue 6

This tutorial gives an introduction into statistical methods for diagnostic medicine. The validity of a diagnostic test can be assessed using sensitivity and specificity which are defined for a binary diagnostic test with known reference or gold standard. As an example we use Procalcitonin with a cut off value ≥ 0.5 g/L as a test and Sepsis-2 criteria as a reference standard for the diagnosis of sepsis. Next likelihood ratios are introduced which combine the information given by sensitivity and specificity. For these measures the construction of confidence intervals is demonstrated. Then, we introduce predictive values using Bayes’ theorem. Predictive values are sometimes difficult to communicate. This can be improved using natural frequencies which are applied to our example. Procalcitonin is actually a continuous biomarker, hence we introduce the use of receiver operator curves (ROC) and the area under the curve (AUC). Finally we discuss sample size estimation for diagnostic studies. In order to show how to apply these concepts in practice we explain how to use the freely available software R.

Immunoassays are currently the methods of choice for the measurement of a large panel of complex and heterogenous molecules owing to full automation, short turnaround time, high specificity and sensitivity. Despite remarkable performances, immunoassays are prone to several types of interferences that may lead to harmful consequences for the patient (e.g., prescription of an inadequate treatment, delayed diagnosis, unnecessary invasive investigations). A systematic search is only performed for some interferences because of its impracticality in clinical laboratories as it would notably impact budget, turnaround time, and human resources. Therefore, a case-by-case approach is generally preferred when facing an aberrant result. Hereby, we review the current knowledge on immunoassay interferences and present an algorithm for interference workup in clinical laboratories, from suspecting their presence to using the appropriate tests to identify them. We propose an approach to rationalize the attitude of laboratory specialists when faced with a potential interference and emphasize the importance of their collaboration with clinicians and manufacturers to ensure future improvements.

Disruptive imaging and laboratory technologies can improve clinical decision processes and outcomes in oncology. However, certain obstacles must be overcome before these technologies can be fully implemented as part of the standard for care. An integrative diagnostic approach represents a unique opportunity to unleash the full diagnostic potential and paves the way towards personalized cancer diagnostics. To meet this demand, an interdisciplinary Task Force of the EFLM was initiated as a consequence of an EFLM/ESR during the CELME 2019 meeting in order to evaluate the clinical value of CNAPS/CTC (circulating nucleic acids in plasma and serum/circulating tumor cells) in early detection of cancer. Here, an overview of current disruptive techniques, their clinical implications and potential value of an integrative diagnostic approach is provided. Furthermore, requirements such as the establishment of diagnostic tumor boards, development of adequate software solutions and a change of mindset towards a new generation of diagnosticians providing actionable health information are presented. This development has the potential to elevate the position and clinical recognition of diagnosticians.

Objectives Medical laboratory performance is a relative concept, as are quality and safety in medicine. Therefore, repetitive benchmarking appears to be essential for sustainable improvement in health care. The general idea in this approach is to establish a reference level, upon which improvement may be strived for and quantified. While the laboratory community traditionally is highly aware of the need for laboratory performance and public scrutiny is more intense than ever due to the SARS-CoV-2 pandemic, few initiatives span the globe. The aim of this study was to establish a good practice approach towards benchmarking on a high abstraction level for three key dimensions of medical laboratory performance, generate a tentative snapshot of the current state of the art in the region of Europe, Middle East, and Africa (EMEA), and thus set the stage for global follow-up studies. Methods The questionnaire used and previously published in this initiative consisted of 50 items, roughly half relating to laboratory operations in general with the other half addressing more specific topics. An international sample of laboratories from EMEA was approached to elicit high fidelity responses with the help of trained professionals. Individual item results were analyzed using standard descriptive statistics. Dimensional reduction of specific items was performed using exploratory factor analysis and assessed with confirmatory factor analysis, resulting in individual laboratory scores for the three subscales of “Operational performance”, “Integrated clinical care performance”, and “Financial sustainability”. Results Altogether, 773 laboratories participated in the survey, of which 484 were government hospital laboratories, 129 private hospital laboratories, 146 commercial laboratories, and 14 were other types of laboratories (e.g. research laboratories). Respondents indicated the need for digitalization (e.g. use of IT for order management, auto-validation), automation (e.g. pre-analytics, automated sample transportation), and establishment of formal quality management systems (e.g. ISO 15189, ISO 9001) as well as sustainably embedding them in the fabric of laboratory operations. Considerable room for growth also exists for services provided to physicians, such as “Diagnostic pathways guidance”, “Proactive consultation on complex cases”, and “Real time decision support” which were provided by less than two thirds of laboratories. Concordantly, the most important kind of turn-around time (TAT) for clinicians, sample-to-result TAT, was monitored by only 40% of respondents. Conclusions Altogether, the need for stronger integration of laboratories into the clinical care process became apparent and should be a main trajectory of future laboratory management. Factor analysis confirmed the theoretical constructs of the questionnaire design phase, resulting in a reasonably valid tool for further benchmarking activities on the three aimed-for key dimensions.

Objectives Neurofilament-light (NfL), glial fibrillary acidic protein (GFAP) and contactin-1 (CNTN1) are blood-based biomarkers that could contribute to monitoring and prediction of disease and treatment outcomes in neurological diseases. Pre-analytical sample handling might affect results, which could be disease-dependent. We tested common handling variations in serum of volunteers as well as in a defined group of patients with multiple sclerosis (pwMS). Methods Sample sets from 5 pwMS and 5 volunteers at the outpatient clinic were collected per experiment. We investigated the effect of the following variables: collection tube type, delayed centrifugation, centrifugation temperature, delayed storage after centrifugation and freeze-thawing. NfL and GFAP were measured by Simoa and CNTN1 by Luminex. A median recovery of 90–110% was considered stable. Results For most pre-analytical variables, serum NfL and CNTN1 levels remained unaffected. In the total group, NfL levels increased (121%) after 6 h of delay at 2–8 °C until centrifugation, while no significant changes were observed after 24 h delay at room temperature (RT). In pwMS specifically, CNTN1 levels increased from additional freeze-thaw cycles number 2 to 4 (111%–141%), whereas volunteer levels remained stable. GFAP showed good stability for all pre-analytical variables. Conclusions Overall, the serum biomarkers tested were relatively unaffected by variations in sample handling. For serum NfL, we recommend storage at RT before centrifugation at 2–8 °C up to 6 h or at RT up to 24 h. For serum CNTN1, we advise a maximum of two freeze-thaw cycles. Our results confirm and expand on recently launched consensus standardized operating procedures.

Objectives We compare measurement uncertainty (MU) calculations to real patient result variation observed by physicians using as our model anion gap (AGAP) sequentially measured on two different instrument types. An approach for discretely quantifying the pre-analytical contributions and validating AGAP MU estimates for interpretation of patient results is proposed. Methods AGAP was calculated from sodium, chloride, and bicarbonate reported from chemistry or blood gas analyzers which employ different methodologies and specimen types. AGAP MU was calculated using a top-down approach both assuming no correlation between measurands and alternatively, including consideration of measurand correlation. MU-derived reference change values (RCV) were calculated between chemistry and blood gas analyzers results. Observational paired AGAP data (n=39,626 subjects) was obtained from retrospectively analyzed specimens from five urban tertiary care hospitals in Calgary, Alberta, Canada. Results The MU derived AGAP RCV for paired specimen data by the two platforms was 5.2–6.1 mmol/L assuming no correlation and 2.6–3.1 mmol/L assuming correlation. From the paired chemistry and blood gas data, total observed variation on a reported AGAP has a 95% confidence interval of ±6.0 mmol/L. When the MU-derived RCV assuming correlation is directly compared against the observed distribution of patient results, we obtained a pre-analytical variation contribution of 2.9–3.5 mmol/L to the AGAP observed variation. In contrast, assuming no correlation leads to a negligible pre-analytical contribution (<1.0 mmol/L). Conclusions MU estimates assuming no correlation are more representative of the total variation seen in real patient data. We present a pragmatic approach for validating an MU calculation to inform clinical decisions and determine the pre-analytical contribution to MU in this system.

Objectives Ventilator management in prehospital settings using end-tidal CO 2 can lead to inappropriate ventilation in the absence of point of care blood gas (POCBG) measurements. Implementation of POCBG testing in helicopter Emergency Medical Services (HEMS) is limited in part because of concern for preanalytical and analytical errors due to altitude, vibration, and other associated environmental factors and due to insufficient documentation of implementation challenges. Methods We performed accuracy and precision verification studies using standard materials tested pre-, in-, and post-flight (n=10) in a large HEMS agency. Quality assurance error log data were extracted and summarized for common POCBG errors during the first 31 months of use and air medical transport personnel were surveyed regarding POCBG use (n=63). Results No clinically significant differences were found between pre-, in-, and post-flight blood gas measurements. Error log data demonstrated a reduction in device errors over time. Survey participants found troubleshooting device errors and learning new clinical processes to be the largest barriers to implementation. Continued challenges for participants coincided with error log data including temperature and sampling difficulties. Survey participants indicated that POCBG testing improved patient management. Conclusions POCBG testing does not appear to be compromised by the HEMS environment. Temperature excursions can be reduced by use of insulated transport bags with heating and cooling packs. Availability of POCBG results in air medical transport appeared to improve ventilator management, increase recognition of ventilation-perfusion mismatch, and improve patient tolerance of ventilation.

Objectives Estimated glomerular filtration rate (eGFR) can be calculated using serum/plasma creatinine measured with automated chemistry analyzers. It is unclear whether eGFR can be calculated using creatinine values measured in whole blood (WB creatinine). The aim of this study is to determine the comparability between the eGFR calculated using WB creatinine and plasma creatinine. Methods Blood samples from 1,073 patients presented to the emergency department (ED), perioperative areas, intensive care unit (ICU) or nuclear medicine were used to determine the accuracy of WB creatinine. For each sample, WB creatinine was first measured with Radiometer ABL827 FLEX blood gas analyzer, then plasma creatinine was measured with Roche Cobas702 chemistry analyzer after samples were centrifuged. In a subset of 247 samples with the information of age and sex, whole blood eGFR (WB eGFR) and plasma eGFR were calculated using WB creatinine and plasma creatinine and the 2021 chronic kidney disease epidemiology collaboration (CKD-EPI) creatinine equation, respectively. Results WB creatinine correlated with plasma creatinine linearly with a slope of 1.06 and an intercept of −0.01. The coefficient of determination (R 2 ) was 0.99. WB eGFR correlated with plasma eGFR linearly with a slope of 0.95, intercept of −1.63, and R 2 of 0.97. Comparing to plasma eGFR, the sensitivity and specificity for WB eGFR to identify those with high risk (eGFR<30 mL/min/1.73 m 2 ) and low risk (eGFR>45 mL/min/1.73 m 2 ) for kidney injuries was 100 and 92.2%, respectively. The overall concordance in classifying the four stages of kidney damage between WB eGFR and plasma eGFR was 87.9%. Conclusions WB creatinine measured with Radiometer ABL827 Flex can be used to calculate eGFR using the 2021 CKD-EPI creatinine equation. The sensitivity and specificity for WB eGFR to identify those with high and low risks for potential kidney injuries are acceptable in patients needing rapid assessment of their kidney functions.

Objectives We systematically investigated normally or subclinically increased thyroid stimulating hormone (TSH) values associated with unexpectedly increased thyroxine (FT4) and free triiodothyronine (FT3) in findings of patients without any thyroid disease. Moreover, we looked for alternatives to overcome such states with an improved diagnostic procedure and to investigate the pathogenetic background of the respective patients. Methods Samples with TSH concentrations within the range of 0.4–10 mU/L combined with increased concentrations of FT4 (n=120; Cobas, Roche) were collected over a period of around six years. Cobas FT4 results were compared with measurements from Liaison (DiaSorin) and Architect (Abbott) FT4 assays. For further validation all samples were measured for total thyroxine (TT4) (Cobas, Roche). Finally, FT3 and TT3 as complementary parameters were measured in samples with leftover material. To overcome potential analytical disturbances from stimulating heterophilic antibodies, we used heterophilic blocking tubes (HBTs). Results From the 120 samples with increased FT4 concentrations by Cobas, 51/120 were also increased by Liaison, and 26/120 by Architect. However, the measurement of TT4 indicated only n=10/120 increased values. The number of increased FT3 (n=71) measurements was higher in Architect>Cobas>Liaison (28>27>9). TT3 levels of 70/71 samples were within the reference interval. HBTs were inappropriate to reduce unspecific immunoreactivity in our samples. No clear pathogenetic background could be elucidated in the anamnesis of individual patients. Conclusions To overcome dubious constellations of TSH, FT4, and FT3, it is helpful to measure TT4 and TT3 for control or to use an immunoassay with an alternative assay design for the respective parameters.

Objectives The observation of numerous new structural defects in hemoglobin (Hb) has often been linked to the evolution and development of device technologies used for the separation and quantification of hemoglobin components. However, the increased use of preventive tests for hemoglobinopathies and separative methods to quantify glycated hemoglobin (HbA 1c ) also contributed to these observations, as demonstrated by the case described here. Our aim is to emphasize that different separative method can provide more useful information in patient management. Methods A 64-year-old diabetic woman of Moroccan descent was examined in the context of HbA 1c monitoring. The test was performed using high performance liquid chromatography (HPLC) and capillary electrophoresis (CE) systems. Molecular characterization was performed by direct sequencing of the β and α globin genes. Results The two methods used showed the presence of an anomalous fraction identified as HbS, already observed previously, but only through CE it was possible to observe the presence of another variant and its hybrid components. Direct sequencing of β and α globin genes confirmed heterozygous HbS [ β 6 (A3) Glu→Val; HBB: c.20A>T] and allowed to identify a mutation on the α 2 , [ α 114 (GH2) Pro→Leu gene; HBA 2 : c.344C>T] corresponding to the rare Hb Nouakchott variant. Conclusions The two Hb variants highlighted by the EC and the molecular characterization therefore allowed adequate advice, the correct assessment of HbA 1c and metabolic status and therefore better clinical management of the patient. The availability of different instruments in the same laboratory, confirming situations of diagnostic uncertainty, represents a valuable opportunity that should be encouraged.

Objectives Cholestasis and elevated serum bile 1 acid levels are common in critically ill patients. This study aims to define the specific pattern of bile acids associated with acute respiratory distress syndrome (ARDS) and the changes in pattern over time. Methods Prospective observational study. Serum samples of 70 ARDS patients were analyzed for primary bile acids (cholic acid, chenodeoxycholic acid) and secondary bile acids (deoxycholic acid, litocholic acid, and ursodeoxycholic acid) as well as their glycine and taurine glycation products. Results Primary bile acid levels increased from day zero to day five by almost 50% (p<0.05). This change bases on a statistically significant increase in all primary bile acids between day 0 and day 5 (cholic acid [CA] p=0.001, taurocholic acid [TCA] p=0.004, glycocholic acid [GCA] p<0.001, chenodeoxycholic acid [CDCA] p=0.036, taurochenodeoxycholic acid [TCDCA] p<0.001, glycochenodeoxycholic acid [GCDCA] p<0.001). Secondary bile acids showed predominantly decreased levels on day 0 compared to the control group and remained stable throughout the study period; the differences between day zero and day five were not statistically significant. Non-survivors exhibited significantly higher levels of TCDCA on day 5 (p<0.05) than survivors. This value was also independently associated with survival in a logistic regression model with an odds ratio of 2.24 (95% CI 0.53–9.46). Conclusions The individual bile acid profile of this ARDS patient cohort is unique compared to other disease states. The combination of changes in individual bile acids reflects a shift toward the acidic pathway of bile acid synthesis. Our results support the concept of ARDS-specific plasma levels of bile acids in a specific pattern as an adaptive response mechanism.

Objectives The National Institute for Health and Care Excellence recommends faecal calprotectin (f-cal) to help differentiate inflammatory bowel diseases from irritable bowel syndrome. Faecal samples for calprotectin have historically been collected at home by patients into screw-top pots and sent to laboratories where calprotectin is extracted and analysed. Faecal haemoglobin (f-Hb) samples are collected at home into specific collection devices containing stabilising buffer. We evaluated the OC-FCa method for f-cal, developed by Eiken Chemical Co., Ltd. (Japan) that uses the same collection device and analyser as f-Hb. Methods OC-FCa was assessed for limit of blank (LOB), limit of detection (LOD), limit of quantification (LOQ), within and between-run imprecision, linearity, prozone, recovery and carryover. A method comparison against the BÜHLMANN fCAL ® turbo (BÜHLMANN Laboratories AG, Switzerland) was performed using patient samples and EQA. Results The LOB was 3 µg calprotectin/g faeces (µg/g), LOD 8 μg/g and LOQ 20 μg/g. Within and between-run imprecision was <5%; linearity was good (R 2  > 0.99); prozone was appropriately detected; recovery was 99.6%; no observed carryover. OC-FCa showed a strong positive bias compared with BÜHLMANN fCAL ® turbo (Z=−5.3587, p < 0.001). When categorised using our local pathway, which interprets calprotectin concentrations and need for further investigation, Cohen’s Kappa demonstrates substantial agreement at <50 μg/g (κ=0.80) and >150 μg/g (κ=0.63) and fair agreement (κ=0.22) in the borderline category 50–150 μg/g. Conclusions The OC-FCa method performed well in the evaluation. With the lack of standardisation for f-cal a clinical study is required to evaluate the positive bias and establish suitable cut-off levels.

Objectives Our aim was to derive reference intervals for all Sysmex XN hematology analyzer parameters. The rationale behind the study was the lack of reference intervals for the XN analyzer cell population data (CPD) and functional parameters. Methods Fresh fasting blood samples from 18,484 participants in the Dutch Lifelines study were analyzed using two automated XN analyzers. Structured health questionnaire data were used to select a subgroup of 15,803 apparently healthy individuals for inclusion in the reference population. The Latent Abnormal Values Exclusion (LAVE) approach was used to reduce the influence of latent diseases in the reference population on the resulting reference intervals. We applied analysis of variance to judge the need for partitioning of the reference intervals by sex or age. Results We report reference intervals for 105 XN analyzer hematological parameters with and without applying LAVE. Sex-related partitioning was required for red blood cells, (RBC, RBC-O), hemoglobin (HGB, HGB-O), hematocrit (HCT), mean corpuscular hemoglobin concentration (MCHC), reticulocyte production index (RPI), and side scattered light intensity of the red blood cell population in the RET channel (RBC-Z). Partitioning for age was not warranted. Body mass index (BMI) and smoking had moderate influence on a minority of the parameters. Conclusions We provide reference intervals for all Sysmex XN analyzer routine, CPD and functional parameters, using a direct approach in a large cohort in the Netherlands.

Objectives Cardiac troponin (cTn) is the biochemical gold standard for diagnosing myocardial infarction (MI). We compared the Siemens ADVIA Centaur High-Sensitivity (hs-cTnI) assay with the Siemens Ultra assay (cTnI-U). Methods Over 3 months cTnI-U and hs-cTnI were measured simultaneously at Herlev-Gentofte Hospital. Acute myocardial injury was diagnosed using the 4th universal definition. Disputed cases were adjudicated using clinical data. We compared diagnostic accuracy using area under the curve (AUC) of the receiver operating characteristic. Outliers in between-assay differences were defined as a factor-5 difference and ≥1 measurement >40 ng/L. Patients with outlier differences were invited for re-sampling and tested with serial dilution and heterophilic blocking tubes. Results From the 18th January to the 20th April 2019, 4,369 samples on 2,658 patients were included. cTnI-U measured higher concentrations than hs-cTnI (mean 23%, −52–213%), resulting in a higher frequency of acute myocardial injury, 255 (9.6%) vs. 203 (7.6%), p<0.001. This remained significant after adjudication, 212 vs 197, p<0.001. AUC for the prediction of MI for was 0.963 for cTnI-U and 0.959 for hs-cTnI, p=0.001. Outlier differences were seen in 35 (1.2%) patients, primarily with elevated hs-cTnI (n=33, 94%). On two re-samplings (median 144 and 297 days since inclusion), 16 of 20 (80%) and 11 of 11 had sustained elevation of hs-cTnI. The samples showed no signs of heterophilic antibodies. Conclusions Using hs-cTnI resulted in a subset of patients with large, discrepant elevations in concentration. These patients still had elevated hs-cTnI 6–10 months post admission but no heterophilic antibodies.

Objectives We provide here an updated analysis of an ongoing serosurveillance study, presenting data on the effect of a third dose of Pfizer/BioNTech BNT162b2 vaccine on serum anti-SARS-CoV-2 IgG antibodies. Methods We tested baseline SARS-CoV-2 seronegative healthcare workers undergoing primary vaccination with the mRNA-based COVID-19 Comirnaty vaccine, followed by administration of homologous vaccine booster (third dose). Venous blood was collected before either dose of primary vaccination, at 1, 3 and 6 months afterwards, as well as before and 1 month after receiving the vaccine booster. The serum concentration of anti-SARS-CoV-2 IgG was assayed with DiaSorin Trimeric spike IgG immunoassay. Results The final study population included 53 SARS-CoV-2 seronegative healthcare workers (median age 46 years; 60% females). A first peak of anti-SARS-CoV-2 spike trimeric IgG values was reached 1 month after completing primary vaccination, after which the levels gradually declined until before receiving the vaccine booster. A second peak of anti-SARS-CoV-2 spike trimeric IgG concentration was observed 1 month after receiving the vaccine booster dose (8,700 kBAU/L), which was 39-fold higher than before receiving the vaccine booster (221 kBAU/L; p<0.001), but was also nearly threefold higher compared to values seen at the first peak (2,990 kBAU/L; p<0.001). The rate of subjects with protective anti-SARS-CoV-2 spike trimeric IgG values (i.e., >264 kBAU/L) increased from 47.2% to 100% after 1 month from vaccine booster. Conclusions These results support current policies fostering COVID-19 vaccine boosters to reinforce humoral immunity against SARS-CoV-2.

Objectives Evaluating anti-SARS-CoV-2 antibody levels is a current priority to drive immunization, as well as to predict when a vaccine booster dose may be required and for which priority groups. The aim of our study was to investigate the kinetics of anti-SARS-CoV-2 Spike S1 protein IgG (anti-S1 IgG) antibodies and neutralizing antibodies (NAbs) in an Italian cohort of healthcare workers (HCWs), following the Pfizer/BNT162b2 mRNA vaccine, over a period of up to six months after the second dose. Methods We enrolled 57 HCWs, without clinical history of COVID-19 infection. Fluoroenzyme-immunoassay was used for the quantitative anti-S1 IgG antibodies at different time points T1 (one month), T3 (three months) and T6 (six months) following the second vaccine shot. Simultaneously, a commercial surrogate virus neutralization test (sVNT) was used for the determination of NAbs, expressed as inhibition percentage (% IH). Results Median values of anti-S1 IgG antibodies decreased from T1 (1,452 BAU/mL) to T6 (104 BAU/mL) with a percent variation of 92.8% while the sVNT showed a percent variation of 34.3% for the same time frame. The decline in anti-S1 IgG antibodies from T1 to T6 was not accompanied by a loss of the neutralizing capacity of antibodies. In fact at T6 a neutralization percentage <20% IH was observed only in 3.51% of HCWs. Conclusions Our findings reveal that the decrease of anti-S1 IgG levels do not correspond in parallel to a decrease of NAbs over time, which highlights the necessity of using both assays to assess vaccination effectiveness.

Objectives The assessment of SARS-CoV-2 infections in children is still challenging, but essential for appropriate political decisions. The aim of this study was to investigate whether residual blood samples can be used for SARS-CoV-2 seroprevalence monitoring in pediatrics. Methods In this repeated cross-sectional cohort study, anonymous residual blood samples from pediatric patients aged 0–17 years were collected in three time-periods (Oct.–Nov. 2020, April 2021, and June–July 2021) and analyzed for SARS-CoV-2 Spike protein (anti-S) and nucleocapsid (anti-N) antibodies using commercial antibody assays. 28 reactive samples were used to compare antibody levels with a pseudotyped neutralization assay. The results were further compared to the official national COVID-19 surveillance data to calculate the number of unreported cases. Results In total, n=2,626 individual blood samples were analyzed. In this unvaccinated pediatric cohort anti-S and anti-N antibody seroprevalence increased over the three time periods (anti-S: 1.38–9.16%, and 14.59%; anti-N: 1.26%, to 6.19%, and 8.56%). Compared to the national surveillance data this leads to a 3.93–5.66-fold increase in the number of unreported cases. However, a correlation between the cumulative incidence of the individual provinces and our assigned data was found (r=0.74, p=0.0151). In addition, reactive samples with anti-S and anti-N and samples with only anti-S showed neutralization capabilities (11/14 and 8/14, respectively). Anti-S levels were not significantly different between age groups and sexes (all p>0.05). Conclusions The present study suggests that residual blood samples from routine laboratory chemistry could be included in the estimation of the total SARS-CoV-2 seroprevalence in children.

Objectives The diagnosis of sepsis is challenging, the need for sensitive and specific diagnostic and prognostic biomarkers has not been met. Soluble CD25 (sCD25) is a readily available biomarker reported to represent the severity of the disease. This study aimed to assess the association between sCD25 and mortality in patients with sepsis. Methods In total, 329 adult patients with sepsis were screened through a prospective, observational study. We investigated the severity scores and sCD25 levels at admission to the intensive care unit (ICU), defined by sepsis (sepsis-3). The prognostic value of sCD25 was assessed using receiver operating characteristic (ROC) curves and binary logistic regression models in predicting unfavourable outcome. The correlations between variables and severity of disease were analysed by Spearman correlation tests. Results After entering the ICU, the sCD25 level and sequential organ failure assessment (SOFA) score were significantly higher in nonsurvivors than in survivors. The prognostic values estimated by the ROC curves were 0.678 for sCD25 and 0.945 for SOFA score at ICU admission. sCD25 had a modest ability to predict poor outcome. Logistic regression showed that increased levels of sCD25 were independently associated with unfavourable outcome. Spearman correlation tests showed that sCD25 levels were positively correlated with disease severity. Conclusions In sepsis patients, increased sCD25 levels were independently associated with poor clinical outcomes. Further research is needed to improve the understanding of the pathophysiology of this relationship.