Abstract
OBJECTIVES:
Biomarkers with adequate precision to rule-in/rule-out sepsis in children represent an unmet need. SeptiCyte RAPID is a point-of-care, U.S. Food and Drug Administration-cleared, host gene expression biomarker test for the determination of sepsis probability (reported in 1 hr as a SeptiScore) using a small-volume peripheral blood sample. We have previously reported on SeptiCyte RAPID performance in adults. Here we report a retrospective, noninterventional clinical validation study of SeptiCyte RAPID, as a sepsis diagnostic adjunct in children.
DESIGN, SETTING, AND PATIENTS:
Critically ill children (n = 284), 1 month to 18 years old, admitted to PICUs, with or without suspected or documented infection, but all of whom met criteria for systemic inflammatory response syndrome (SIRS) and were enrolled at multiple clinical sites in the United States and Australia in four separate studies.
INTERVENTIONS:
None; SeptiCyte RAPID results were not used to guide patient treatment.
MEASUREMENTS AND MAIN RESULTS:
A combination of clinical adjudication and an independent panel of three physicians (Retrospective Physician Diagnosis [RPD]), blinded to SeptiCyte RAPID results, were used to retrospectively categorize patients into sepsis vs. noninfectious SIRS categories. Diagnostic performance of SeptiCyte RAPID for differentiating sepsis and noninfectious SIRS compared with RPD was measured using area under the curve (AUC), likelihood ratio, and a monotonic increase in probability of sepsis and increasing SeptiScores. Binomial logistic regression analysis was used to integrate clinical parameters in a model based on SeptiCyte RAPID. Higher SeptiScores correlated with higher probability of sepsis. SeptiCyte RAPID AUC for identifying sepsis in children ranged from 0.80 to 0.83. A model additionally integrating lactate, platelet number, mean arterial pressure, Paco2, mechanical ventilation, weight, age, and leukocyte count increased SeptiCyte RAPID performance (AUC, 0.86; 95% CI, 0.85–0.87).
CONCLUSIONS:
SeptiCyte RAPID demonstrated similar performance characteristics in children to that previously reported for adults. Additional real world effectiveness studies are warranted to determine real-time clinical utility, patient benefits, and health economic impact.
Keywords: biomarkers, child, human, point-of-care testing, sepsis
RESEARCH IN CONTEXT.
Pediatric sepsis diagnosis lacks rapid, reliable biomarkers; prior host gene expression studies were limited by small cohorts, research-only platforms, and validation of commercially available tests. Timely differentiation is crucial for improving patient outcomes, meeting sepsis bundle requirements and reducing inappropriate antimicrobial use.
This multicenter clinical validation study demonstrates that a point-of-care peripheral blood gene expression assay discriminates sepsis from noninfectious systemic inflammatory response syndrome (SIRS) in critically ill children with good accuracy (area under the curve, ~0.80–0.83) and performs similar to that previously demonstrated for adults.
Host response gene expression diagnostics may complement clinical assessment in pediatric sepsis. Broader validation and effectiveness studies are needed to define real-time clinical impact, stewardship benefits, and cost-effectiveness in PICU care.
AT THE BEDSIDE.
A peripheral blood gene expression test provides a sepsis probability within ~1 hour, supporting early bedside differentiation of sepsis from noninfectious SIRS in critically ill children.
Higher SeptiScores correlate with an increasing likelihood of sepsis, helping clinicians contextualize diagnostic uncertainty and prioritize sepsis bundles vs. alternative evaluations at the bedside.
When combined with readily available clinical data (e.g., lactate, platelets, mean arterial pressure, ventilation status), diagnostic performance improves, reinforcing the role of SeptiCyte RAPID as a bedside adjunct to clinical judgment.
Pediatric sepsis is a major cause of morbidity and mortality worldwide, with approximately half of all global sepsis cases occurring in children, with the heaviest burden among those under 5 years old (1–7). Pediatric sepsis was recently redefined as a life-threatening organ dysfunction in a child with documented or suspected infection and is operationalized by a Phoenix Sepsis Score (PSS) of at least 2 points, based on organ dysfunction measures (8).
One notable difference between the 2024 Phoenix sepsis definition and the 2005 International Pediatric Sepsis Consensus Conference definition (9, 10) is that systemic inflammatory response syndrome (SIRS) criteria are no longer included (11). However, SIRS elements may trigger diagnostic and therapeutic decisions and are included in a recently developed Quick Pediatric Septic Shock Screening Score (12). Optimizing sepsis outcomes is correlated to early identification, resuscitation, and management (1–3, 13) with simultaneous antimicrobial stewardship (14). However, specific biomarkers quantifying a dysregulated host response driving organ dysfunction (adult Sepsis-3 definition) are lacking, and the discrimination of infectious vs. noninfectious origins of SIRS in both children and adults is often challenging. Up to a third of adult patients initially suspected of sepsis have ultimately been shown to have a noninfectious illness (15). Clinical trials have ascertained that a significant proportion of children with sepsis exhibit no causative organism (bacterial or viral) for infection (4, 16). Therefore, there remains a need for point-of-care testing in children for timely identification of infection and sepsis.
We have previously reported on the development, validation, and regulatory clearance of a point-of-care, host-immune response, and adult sepsis test (SeptiCyte RAPID; Figs. SDC1 and 2; Table SDC1, https://links.lww.com/PCC/C772) (17–19) and demonstrated good diagnostic performance of a kit version in children (SeptiCyte LAB) (20). SeptiCyte RAPID uses a small blood sample, a disposable cartridge (Biocartis Idylla, Mechelen, Belgium), and provides a likelihood of sepsis on a scale of 0–15 (SeptiScore), which is subdivided into four “Bands” with a 1-hour turnaround (Fig. SDC1, https://links.lww.com/PCC/C772). While a number of studies have reported promising results examining host gene expression sepsis biomarkers in pediatric research settings (21–23), there is a lack of validation studies of commercially available bedside sepsis tests.
Here, we report on the results of a clinical validation study of SeptiCyte RAPID in critically ill children. The investigators hypothesized that the diagnostic performance characteristics of SeptiCyte RAPID for determining likelihood of sepsis among children initially suspected with infection would be similar to those previously reported for adults (18, 19).
MATERIALS AND METHODS
Human Participants
This report summarizes four prospective, descriptive cohort investigations, two each conducted at Queensland Children’s Hospital (Brisbane, VIC, Australia) and Seattle Children’s Hospital (Seattle, WA). Details of human subjects review and approval by institutional review boards for each study and site are provided in the Supplemental Digital Content (https://links.lww.com/PCC/C772) and have been previously reported (20, 23, 24). Procedures were followed in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1975. Children were enrolled following documented, informed parental or legal custodian(s) consent.
Study Descriptions and Cohorts
The performance of SeptiCyte RAPID was assessed utilizing a total of 284 banked clinical samples, collected in PAXgene Blood RNA tubes, from participants between 1 months and 18 years old enrolled in four observational clinical trials:
-
1]
RAPIDS (Rapid Acute Paediatric Infection Diagnosis in Suspected Sepsis) (ANZCTR [Australia and New Zealand Clinical Trials Registry] registration number: ACTRN1262000134092).
-
2]
Gene Expression Objective Definition of Early Sepsis In Children (GEODESIC).
-
3]
NITRIC (NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congential heart defects) (ANZCTR registration number: ACTRN12617000821392).
-
4]
GAPPSS (Genotypes And Phenotypes in Pediatric SIRS and Sepsis) trial (ClinicalTrials.gov identifier: NCT02728401).
Patient blood samples from the RAPIDS and GEODESIC studies were obtained from ICU patients with either sepsis or SIRS with an initial suspicion of infection on day 1 of admission. Samples from the NITRIC and GAPPSS studies derived from post-surgical (on day 1) ICU patients. These children were deemed to have noninfectious SIRS (active infection would have precluded surgery) and did not undergo retrospective adjudication.
Exclusion criteria were age older than 18 years or younger than 28 days on the day of enrollment; no clinical cultures or serologic or virology tests ordered when infection was suspected; expectation of nonsurvival during the hospital stay; or “ward of the state” status (GEODESIC and GAPPSS only). Additional exclusions were based on nonavailability of blood samples or SeptiCyte RAPID results. Patients in the NITRIC and GAPPSS studies either had negative microbiology results or did not have microbiology testing performed. Patients in the GEODESIC and RAPIDS studies had microbiology testing performed as directed by the attending physician to include cultures and polymerase chain reaction (PCR) testing.
Definitions and Comparator Method
Full details of the comparator method are provided in the Supplemental Digital Content (https://links.lww.com/PCC/C772). All patients were enrolled before the introduction of the Phoenix Pediatric Sepsis Definition (8), and accordingly, sepsis diagnosis was based on the definitions formulated at the International Pediatric Sepsis Consensus Conference (9). Currently, there is a lack of a reliable gold standard comparator for sepsis. Given the higher frequency of culture-negative infections and positive respiratory virus tests, these challenges are pronounced in children.
For each enrolled patient in the RAPIDS and GEODESIC trials, retrospective diagnostic adjudication was performed. This consisted of a combination of two previously described methods, both used to decrease the degree of uncertainty obtained when using a single method.
The first method, which we have previously used for adults, focused on Retrospective Physician Diagnosis (RPD) (17). This method employed a panel of three independent physicians provided with relevant clinical and laboratory data. Based on the “Forced” algorithm of the RPD process, these physicians were required to indicate a diagnosis of either “sepsis” or “SIRS” (an “indeterminate” assignment was not allowed).
The second method, described by Schlapbach et al (23), required that two independent onsite physicians adjudicate infection status (definite bacterial, definite viral, probable bacterial, probable viral, combined bacterial and viral, noninfectious, or none/unknown). Adjudication of the clinical phenotype required unanimous agreement of both assessors; any disagreement was resolved together with a third senior assessor to ensure robust adjudication.
The final “sepsis” group (n = 141) included cases where both methods concluded a sepsis diagnosis (forced RPD identified sepsis AND site assessment identified definite/probable/combined infection). The final “SIRS” group included cases where both methods agreed that SIRS was operative or an infection was not operative (forced RPD identified SIRS AND site assessment identified noninfectious or none/unknown; n = 50). NITRIC and GAPPSS patients were directly added into the final SIRS group (n = 57). For 36 patients, there was discordance between RPD and infection status determination (indeterminate). The patient numbers for each group are therefore: sepsis (n = 141), SIRS (n = 107), and indeterminate (n = 36).
Indeterminates were initially excluded from the analyses. However, to demonstrate “real life” performance, a separate analysis was performed where indeterminates were randomly assigned either the sepsis or SIRS group using pre-set sepsis prevalence rates of 10% and 22% (4, 25). Therefore, the final analyses included: 1) a total of 248 patients with a consensus diagnosis (141 + 107 = 248) and 2) a total of 284 patients (248 + 36 = 284) with indeterminates included.
SeptiCyte RAPID
SeptiCyte RAPID is a host gene expression assay designed as an adjunct to diagnose sepsis. This biomarker has regulatory clearances for use in adults in the United States (day 1 ICU), Europe, and Australia, and for children (1 mo to 18 yr old) in Australia, and was originally validated in adult ICU patients (4, 17). Further independent validation has shown consistent performance in adults with COVID-19 and following abdominal surgery (26, 27). SeptiCyte RAPID is conducted on the Idylla hardware platform with test specific software (Biocartis NV) and requires 0.90 mL of PAXgene-stabilized blood (0.24 mL of whole blood) or 0.24 mL of anti-coagulated whole blood (EDTA) without further processing. A customized cartridge performs all assay steps including sample extraction/purification and reverse transcription-quantitative PCR (RT-qPCR) for relative quantification of PLAC8 (Placenta Associated 8) and PLA2G7 (Phospholipase A2 Group VII) messenger RNA targets (Table SDC1, https://links.lww.com/PCC/C772). Known biological roles include neutrophil degranulation in response to infection and regulation of platelet activation for PLAC8 and PLA2G7, respectively (for further detail, see Table SDC1, https://links.lww.com/PCC/C772). Test results (SeptiScore, range, 0–15; Figure SDC1, https://links.lww.com/PCC/C772) are calculated as the difference between the RT-qPCR quantitative cycle values for PLA2G7 and PLAC8. The test requires a hands-on time of ~2 minutes and a turnaround time of ~1 hour.
SeptiCyte RAPID Performance Measures
We determined the performance of SeptiCyte RAPID by comparing patient SeptiScores (1–15) and bands (1–4), to retrospective clinician adjudication. Further information on SeptiCyte RAPID bands is provided in Figure SDC1 (https://links.lww.com/PCC/C772). The aim was to test two hypotheses: 1) a monotonic increase in SeptiScore will correlate with increasing likelihood of sepsis and 2) nonoverlapping 80% CIs between bands 1 and 3, and bands 2 and 4 will be observed, as previously demonstrated for adults. For children, we performed this comparison using three different patient subgroups: 1) exclusion of indeterminates, 2) inclusion of indeterminates at 10% sepsis prevalence, and 3) inclusion of indeterminates at 22% sepsis prevalence. We also calculated likelihood ratios (LRs) and area under the receiver operating characteristic (ROC) curve (AUC) as described in the Statistical Analyses section below.
Statistical Analyses
Conventional Statistical Tests
p values for two-group comparisons were calculated with Welch’s two-sample t test as implemented in the R “stats” package (R Core Team, Vienna, Austria, 2021) unless otherwise noted.
Positive LRs were calculated by dividing test sensitivity by 1–specificity. While actual LRs are a continuum of SeptiScores between 0 and 15, we calculated LRs for each SeptiCyte RAPID band when using different diagnostic methods (consensus and forced) and for two sepsis prevalences (10% and 22%) for indeterminate cases.
Receiver Operating Characteristic Curve Analysis
ROC curve analysis with calculation of AUC values was performed using the pROC package in R (28) and CIs for AUC were calculated by DeLong method (29) as implemented in pROC (30).
Binomial Logistic Regression
A permutation analysis of 14 variables (SeptiScore, mean arterial pressure [MAP], leukocyte count [WBC], platelet count, heart rate [HR], creatinine, lactate, age, weight, Glasgow Coma Scale [GCS] score, mechanical ventilation [MV], baseline Pediatric Logistic Organ Dysfunction [PELOD]-2 (31), Paco2, and PSS) (8), alone or in any combination, was conducted, resulting in 16,383 unique logistic regression models and 100 iterations (1,638,300 models). Multicollinearity due to variables such as age and weight or the PELOD-2 score and components thereof (Paco2, MV, GCS, MAP, and WBC) was ignored to showcase the full spectrum of combinations possible and improve the generalizability of the models in low-resource settings where the PELOD-2 score might not be calculated when treating pediatric patients with suspected sepsis. Performance was compared with a consensus diagnosis for 248 patients (sepsis = 141; SIRS = 107). Indeterminates (n = 36) were excluded. These variables were selected because they are readily available to clinicians within the first 3 hours of hospital presentation and missing data were less than ~30%. This missing value cutoff was chosen based on a balance between clinical utility and model accuracy.
Missing values (HR n = 5, creatinine n = 29, platelets n = 46, WBC n = 47, Paco2 n = 77, lactate n = 84, MV n = 2, MAP n = 13, baseline PELOD-2 n = 7, GCS n = 16, temperature [minimum] n = 41, temperature [maximum] n = 38, and weight n = 2) were imputed using a multivariable imputation algorithm (32–34) using 100 iterations. If only one temperature value was recorded (in some studies) it was substituted as both the minimum and maximum value measured within the first 24 hours of admission.
RESULTS
Patient Demographic and Clinical Data
The majority of research participants presented through the emergency department (40.5%) or were transferred from a surgical suit (24.6%). Demographic and clinical data for the four combined studies are presented in Table 1. Further patient details including microbiology results are provided in the Supplemental Digital Content (Tables SDC2 and 3, https://links.lww.com/PCC/C772). Figure 1 provides a flow diagram for the collective study cohort.
TABLE 1.
Patient Demographic and Clinical Data
| Demographic and Clinical (Categorical) | Sepsis (n; n = 141) | Systemic Inflammatory Response Syndrome (n; n = 107) | Indeterminate (n; n = 36) |
|---|---|---|---|
| Sexa | |||
| Female | 72 | 50 | 16 |
| Male | 69 | 56 | 20 |
| Age | |||
| Infant-toddler (1–23 mo) | 39 | 46 | 12 |
| Child (2–11 yr) | 57 | 32 | 11 |
| Adolescent (12–18 yr) | 45 | 29 | 13 |
| Race/ethnicity | |||
| American Indian and Alaska Native | 0 | 3 | 0 |
| Asian | 8 | 5 | 1 |
| Black | 3 | 2 | 0 |
| Hispanic | 15 | 13 | 2 |
| Mixed or other | 16 | 9 | 7 |
| Pacific Islanders and aboriginal | 0 | 6 | 0 |
| Unknown | 56 | 13 | 20 |
| White (Caucasian) | 43 | 56 | 6 |
| Site | |||
| Queensland Children’s Hospital (Rapid Acute Paediatric Infection Diagnosis in Suspected Sepsis [RAPIDS] and NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congenital heart defects [NITRIC] trials) | 77 | 50 | 32 |
| Seattle Children’s Hospital (Gene Expression Objective Definition of Early Sepsis In Children [GEODESIC] and Genotypes and Phenotypes in Pediatric SIRS and Sepsis [GAPPSS] trials) | 64 | 57 | 4 |
| Source of admission | |||
| Emergency department | 75 | 22 | 18 |
| Outside hospital | 14 | 11 | 0 |
| Surgical suite | 8 | 62 | 0 |
| Transfer from another hospital (not ICU) | 18 | 1 | 10 |
| Unknown | 12 | 9 | 6 |
| Ward | 14 | 2 | 2 |
| Death | |||
| No | 138 | 100 | 34 |
| Yes | 3 | 6 | 2 |
| Invasive mechanical ventilationb | |||
| No | 88 | 28 | 23 |
| Yes | 52 | 78 | 13 |
| Unknown | 1 | 1 | 0 |
| Microbiology | |||
| Bacterial | 53 | 1 | 7 |
| Bacterial/fungal | 2 | 0 | 0 |
| Viral | 44 | 1 | 5 |
| Viral/bacterial | 18 | 0 | 4 |
| Viral/fungal | 1 | 0 | 0 |
| Fungal | 0 | 0 | 1 |
Sex data unavailable for one participant.
Ventilator data unavailable for two participants.
Figure 1.
Enrollment flow diagram for pediatric patients included in this study. A total of 284 critically ill pediatric patients, all with systemic inflammatory response syndrome (SIRS) and some with documented or suspected infection, were gleaned from the Rapid Acute Paediatric Infection Diagnosis in suspected Sepsis (RAPIDS), NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congenital heart defects (NITRIC), Genotypes And Phenotypes in Pediatric SIRS and Sepsis (GAPPSS), and Gene Expression Objective Definition of Early Sepsis In Children (GEODESIC) studies and included in the current investigation. The RAPIDS and GEODESIC trials included sepsis and SIRS patients, whereas the NITRIC and GAPPSS trial patients included only SIRS patients (Table 1; for details of each cohort, see Tables SDC2 and 3, https://links.lww.com/PCC/C772). ND = SeptiCyte RAPID result not determined.
For the RAPIDS and GEODESIC trials, a total of 141 sepsis patients, 107 SIRS patients and 36 indeterminate patients were included with a broad range of ages, race/ethnicities, and admission sources. Microbiology results included bacterial, viral, fungal, and mixed infection isolates. In the sepsis subgroup, 83.7% and 22.2% of cases had positive microbiology results (positive culture and/or PCR) and positive blood cultures, respectively. In all subgroups, the most common pathogen identified, using any type of testing, was bacterial (21.5%) followed by viral (17.6%). For positive blood cultures (n = 35), the most common pathogens identified were Escherichia coli (n = 7) and Staphylococcus aureus (n = 7). For all subgroups, 38% of patients required invasive MV, and mortality was 2.0%. Median admission (baseline) organ dysfunction using PELOD-2 for the sepsis patients (n = 138, missing data for three patients) was 2.5 (IQR, 0.25–6).
Additional SIRS patients (n = 57) were enrolled through the NITRIC and GAPPSS trials. This patient subgroup consisted of a range of race/ethnicities although the predominant race/ethnicity was White/Caucasian (62.0% and 53.9% for NITRIC and GAPPSS, respectively). The NITRIC trial consisted entirely of patients 1–23 months old, and the GAPPSS trial predominantly consisted of children 2–11 years old (62%). There was an overall low mortality rate (< 2%), and all patients received invasive MV as part of post-surgical support.
For the entire patient population (n = 284), 49.6%, 37.7%, and 12.7% were classified as “sepsis,” “SIRS,” and “indeterminate,” respectively. The predominant race/ethnicity, if recorded, was White/Caucasian with a higher proportion in the SIRS subgroup (52%). The largest proportion of sepsis patients were admitted through the emergency department (53.2% of 141), and the largest proportion of SIRS patients (57.9% of 107) were admitted through a surgical theater. Mortality was higher in the SIRS subgroup (7%) compared with the sepsis subgroup (2%). A high percentage of SIRS patients received invasive MV (73% compared with 37% for sepsis) since all NITRIC and GAPPSS patients received postoperative MV.
SeptiCyte RAPID Performance
As for adults, we observed in children a monotonic increase in the probability of sepsis with increasing SeptiScore without or with inclusion of indeterminate patients. Without the inclusion of indeterminates, sepsis probability for bands 1–4 were 10.4%, 37.7%, 70%, and 82.2%, respectively (Fig. 2A). With inclusion of indeterminate patients at 10% sepsis, probability for bands 1–4 were 10.54%, 32.44%, 62.3%, and 76.73%, respectively (Fig. 2B). With inclusion of indeterminate patients at 22%, sepsis probability for bands 1–4 were 12.11%, 34.65%, 63.75%, and 77.58%, respectively (Fig. 2C). There were also nonoverlapping CIs for bands 1 vs. 3 and 2 vs. 4 without, or with, inclusion of indeterminate patients. AUC performance for SeptiCyte RAPID when indeterminates were excluded was 0.83 (AUC; 95% CI, 0.78–0.88), and 0.80 when indeterminates were included at sepsis prevalences of both 10% (AUC; 95% CI, 0.79–0.82) and 22% (AUC; 95% CI, 0.78–0.82).
Figure 2.
Dot and whisker plots of SeptiCyte RAPID performance based on probability and band. Each figure shows the distribution of sepsis: systemic inflammatory response syndrome (SIRS) patients based on the four SeptiCyte RAPID bands (1–4, x-axis) vs. the retrospective probability of sepsis (y-axis). The percentage of sepsis patients in each band is shown alongside each dot and the number of sepsis: SIRS patients in each band is indicated above each whisker. Whiskers represent 80% CIs. A, Excluded indeterminates (n = 36). B and C, Included indeterminates distributed across bands at two different “prevalences” of sepsis (10% and 22%). B, Includes all patients (n = 284) where 10% of indeterminates were assigned sepsis and the remaining as SIRS using 1000 iterations to simulate a sepsis prevalence of 10% in the indeterminate patients. C, Includes all patients (n = 284) where 22% of indeterminates were assigned sepsis and the remaining as SIRS using 1000 iterations to simulate a sepsis prevalence of 22% in the indeterminate patients.
For forced diagnosis, the LRs ranged from 0.11/0.13 for band 1 to 3.17/3.14 for band 4 at 10% and 22% prevalences, respectively. The consensus LRs observed for children for band 1 (0.09) and band 4 (3.51) are lower than that previously reported for adults (0.15–6.05; Table 2). Using Fagan’s nomograms, the consensus LRs for each band, and pre-test probabilities of 10% and 22% (prevalences), we calculated post-test probabilities of sepsis. For a pre-test probability of 10%, the post-test probabilities for bands 1, 2, 3, and 4 were 1.0%, 4.9%, 16.5%, and 28.1%. For a pre-test probability of 22%, the post-test probabilities for bands 1, 2, 3, and 4 were 2.5%, 11.5%, 33.4%, and 49.7% (Fig. SDC5, A and B, https://links.lww.com/PCC/C772).
TABLE 2.
Sepsis Likelihood Ratios for SeptiCyte RAPID Bands Based on Prevalence and Diagnostic Method
| SeptiCyte RAPID Interpretation Band | Forced Diagnosis | Consensus Diagnosis | ||||
|---|---|---|---|---|---|---|
| Forced Diagnosis Sepsis Prevalence (10%/22%) | Sepsis LR | Consensus Diagnosis Sepsis Prevalence | Sepsis LR | |||
| SIRS | Sepsis | SIRS | Sepsis | |||
| Band 4 (7.4–15; high risk of sepsis) | 23%/22% | 77%/78% | 3.17/3.14 | 18% | 82% | 3.51 |
| Band 3 (6.2–7.3) | 38%/36% | 62%/64% | 1.60/1.60 | 30% | 70% | 1.78 |
| Band 2 (5.0–6.1) | 68%/65% | 32%/35% | 0.46/0.48 | 62% | 38% | 0.46 |
| Band 1 (0–4.9; low risk of sepsis) | 89%/88% | 11%/12% | 0.11/0.13 | 90% | 10% | 0.09 |
LR = likelihood ratio, SIRS = systemic inflammatory response syndrome.
Calculated sepsis LRs for each SeptiCyte RAPID band (left-hand column, bands 1–4) based on the diagnostic method used (Forced or Consensus, first row) and at different sepsis prevalences for Forced Diagnosis (10%, 22%). Figures in each row under the headings “SIRS” or “sepsis” represent the percentage of each patient subgroup for the four bands. By example, for Forced diagnosis and band 4 results, the percentage of sepsis patients at sepsis prevalences of 10% and 22% was 77% and 78%, respectively. That is, ~78% of patients with a band 4 SeptiScore were retrospectively, and forced, diagnosed with sepsis. The best performing sepsis LRs ranged from 0.09 to 3.51 for bands 1 and 4, respectively, using consensus diagnosis. For comparison, consensus sepsis diagnosis LRs for bands 1–4 in adults were 0.15, 0.38, 1.07, and 6.05.
For sepsis patients with positive blood cultures, 68.4% had SeptiScores in band 4 (range, 7.6–11.4). All patients with positive blood severe acute respiratory syndrome coronavirus 2 PCR results (n = 6) received a consensus and forced RPD diagnosis of sepsis, and SeptiScores in band 4 (high probability of sepsis) ranged from 8.6 to 13.4.
For children, SeptiCyte RAPID bands 2 and 3 included 37.7% and 70.0% of patients with sepsis, respectively. For both adults and children, SeptiCyte RAPID bands 1 and 4 contained similar percentages of sepsis patients (Fig. SDC2, A and B, https://links.lww.com/PCC/C772).
We found no statistical significance in SeptiCyte RAPID performance between the U.S. cohort and the Australian patient cohorts (Fig. SDC3, https://links.lww.com/PCC/C772). Further, we demonstrate consistent SeptiCyte RAPID performance across different pediatric age groups (adolescent, child, and infant; Fig. SDC4, https://links.lww.com/PCC/C772).
Performance of SeptiCyte RAPID in Combination With Other Clinical Parameters
We performed a binomial logistic regression analysis using a number of clinical and laboratory variables to determine SeptiCyte RAPID diagnostic performance when used alone, and whether other variables could improve diagnostic performance. These analyses showed that SeptiCyte RAPID used alone outperformed other clinical parameters (AUC, 0.83; 95% CI, 0.83–0.83). This was true whether the clinical variables were examined alone (AUCs, 0.51–0.63) or in combination (AUC up to 0.75; Fig. 3). Some variables augmented SeptiCyte RAPID by improving AUC up to 0.87. In models where multicollinear parameters were removed but SeptiScore was retained in combination with either baseline PELOD-2, HR, and weight or with PELOD-2, HR, and age, the AUC for distinguishing sepsis from SIRS was 0.84.
Figure 3.
Performance of clinical and laboratory parameters alone, in combination, and with or without SeptiCyte RAPID. Plot of results of a binomial logistic regression analysis using SeptiCyte RAPID with, or without, an additional 13 clinical parameters. The x-axis shows performance (area under the curve [AUC]) of various models alone or containing any combination of 13 clinical parameters (black distribution), and various models for SeptiScore alone or containing any combination of 13 clinical parameters (red distribution on right). Any model containing SeptiScore had higher AUCs relative to combinatorial models not using SeptiScore or individual biomarkers by themselves (vertical line or in the table). MAP = mean arterial pressure, PELOD 2 = Pediatric Logistic Organ Dysfunction 2.
Two models generating the highest SeptiCyte RAPID performance of AUC = 0.86 with minimum parameters (n = 6) were: 1) SeptiScore, lactate, platelets, MAP, Paco2, and MV (95% CI, 0.85–0.86) and 2) SeptiScore, lactate, WBC, MAP, Paco2, and MV (95% CI, 0.85–0.86).
DISCUSSION
A number of host response-based sepsis biomarkers have recently become commercially available, but validation studies in pediatric cohorts across different settings are lacking. Here, we report the results of a clinical validation study of SeptiCyte RAPID in critically ill children. Our results demonstrate that SeptiCyte RAPID has very similar performance characteristics in children to that previously reported for adults (4, 16, 17). SeptiCyte RAPID outperformed combinations of clinical and conventional laboratory parameters. We also identified additive clinical and laboratory parameters that when included with SeptiCyte RAPID may further aid clinicians in identifying children with infection in the context of SIRS.
We have previously reported an AUC of 0.99 for SeptiCyte LAB, a four-gene, kit-based version of SeptiCyte RAPID that includes the assessment of PLA2G7 and PLAC8, for differentiating post-cardiac surgery children with SIRS from those retrospectively determined to have definite/probable bacterial sepsis (18). In the present study, involving a more heterogeneous patient population spanning cohorts from two continents and using SeptiCyte RAPID, the AUC for differentiating SIRS from sepsis was 0.83, and there were subtle differences compared with those reported for adults, including differences in the sepsis-to-SIRS patient proportions in bands 2 and 3. These differences may be due to differing patient numbers and sepsis prevalence. The observed differences in this study are also borne out in calculated LRs for SeptiCyte RAPID bands using consensus diagnosis. Children with band 1 and band 4 SeptiScores had lower LRs compared with adults, and children with band 3 SeptiScores had a higher LR compared with adults. These differences may lead to differences in clinical utility between adults and children, including the ability to rule-in and rule-out sepsis.
Figure 3 illustrates that differentiating SIRS and sepsis using clinical and commonly used laboratory parameters alone has limited diagnostic performance (maximum AUC, 0.75), whereas SeptiCyte RAPID alone had an AUC of 0.83 and a maximum AUC of 0.87 when combined with parameters such as lactate, platelet count, leukocyte count, MAP, and need for MV. These results suggest that select clinical parameters can be used to augment the diagnostic performance of SeptiCyte RAPID in children.
There are limitations to this study. First, the total number of children enrolled is relatively small (n = 284) and restricted to patients in the ICU, which may limit the generalizability of the results. However, this, in part, may be offset by enrollment of children from different countries, with diverse ethnicities and comorbidities. Second, we restricted enrollment to children 1 month to 18 years old because of the unknown effect of immune development in neonates on a host response test. Results from this study should therefore not be extended to neonates. Third, the enrolled SIRS population is likely to be less heterogeneous than that encountered in usual clinical practice because of the inclusion of a high percentage of post-surgical patients, which may lead to biased results. We also used the 2005 pediatric sepsis definition (9), which includes organ dysfunction only in the definitions of “severe sepsis” and “septic shock.” However, in the early stages of sepsis, organ dysfunction (the sine qua non for sepsis under the pediatric Phoenix sepsis definition) may not yet be evident or easily detected. Identification of sepsis in its early stages where treatment could be most effective (i.e., before organ damage is extensive) remains a challenge but important opportunity. Finally, it is recognized that further prospective validation studies will be required to understand the effects of patient sample timing and clinical effectiveness.
The clinical utility of SeptiCyte RAPID lies in providing fast, earlier, and more accurate identification of the cause of presenting clinical signs (e.g., SIRS) in patients initially suspected of having infection. For adult patients initially suspected with sepsis, up to 25% were shown retrospectively to have a “sepsis mimic” (35). Such patients, identified by a low SeptiScore, require a different diagnostic and therapeutic workup compared with those with infection. Based on low post-test probability of sepsis for SeptiCyte RAPID band 1 results, clinical utility in children may therefore lie in the ability to rule-out sepsis and promote antibiotic stewardship. In this instance, clinicians may consider an alternate diagnosis and/or consider limiting patient antibiotic exposure. For patients with high SeptiScores, utility may lie in diagnostic stewardship. That is, patients with a higher likelihood of sepsis will require a “sepsis bundle” within a prescribed timeframe. Further, additional efforts at obtaining blood cultures could be made in such patients, since we have previously shown in adults that high SeptiScores correlate with a higher likelihood of blood culture positivity.
CONCLUSIONS
SeptiCyte RAPID can be used in conjunction with other available clinical and laboratory parameters, including measures of early organ dysfunction, to guide earlier and more accurate identification of children with sepsis. Further prospective validation studies will be required to understand the effects of patient sample timing and to determine where clinical utility lies.
ACKNOWLEDGMENTS
We gratefully acknowledge Deana Rich’s valuable contribution to this study as a Research Coordinator who managed patient enrollment and data collection during the Gene Expression Objective Definition of Early Sepsis In Children (GEODESIC) clinical trial. We also gratefully acknowledge Carolyn Pardo who coordinated patient sample collection and generated and collated SeptiCyte RAPID results for the Rapid acute Paediatric Infection Diagnosis in suspected Sepsis (RAPIDS) and NITric oxide during cardiopulmonary bypass to improve Recovery in Infants with Congenital heart defects (NITRIC) trials.
Supplementary Material
Footnotes
Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/pccmjournal).
RAPIDS (Rapid Acute Paediatric Infection Diagnosis in Suspected Sepsis) was supported by the Australian Government Medical Research Future Fund Genomic Health Futures Mission Project Grant GHFMPACI000004, with further support from the Children’s Hospital Foundation Queensland Program Grant 50335; the Australian Government Medical Research Future Fund administered by Brisbane Diamantina Health Partners (MRFF9100000); Emergency Medicine Foundation EMPJ-378R27-2017-SNE; Gold Coast Hospital Foundation RGS2017-LG0020; Far North Queensland Foundation; Townsville Hospital and Health Services SERTA (Study, Education and Research Trust Account) Grant; Australian Infectious Diseases Research Centre; and Swiss Personalized Health Technologies (PHRT-273). NITRIC (NITric oxide during cardiopulumonary bypass to improve Recovery in Infants with Congenital heart defects) was supported by grants from the Medical Research Future Fund (ARGCHDG000036), Australia, and grant GNT1140322 from the National Health and Medical Research Council; grants from the HeartKids Foundation, Children’s Hospital Foundation, Perth Children’s Hospital Foundation, Intensive Care Foundation, and Green Lane Research and Educational Fund. The Gene Expression Objective Definition of Early Sepsis In Children (GEODESIC) study was funded in part through U.S. government support under Contract HHSO 100201900023C awarded by the Biomedical Advanced Research and Development Authority, part of the Office of the Assistant Secretary for Preparedness and Response within the U.S. Department of Health and Human Services. The GAPPSS (Genotypes and Phenotypes in Pediatric SIRS and Sepsis) study was funded by Immunexpress.
Drs. Zimmerman’s and Davis’s institutions received funding from Immunexpress. Dr. Zimmerman’s institution received funding from the National Institutes of Health (NIH) and the National Institute of Child Health and Human Development (NICHD); he received funding from Elsevier and royalties from Elsevier Publishing. Drs. Zimmerman and Yager received support for article research from the Biomedical Advanced Research and Development Authority (BARDA). Drs. Zimmerman, Cermelli, Gibbons, Yager, and Brandon disclosed off-label use of SeptiCyte RAPID. Dr. Cermelli’s institution received funding from the U.S. government BARDA, the Office of the Assistant Secretary for Preparedness and Response (ASPR), the U.S. Department of Health and Human Services (HHS; HHSO 100201900023C). Drs. Cermelli and Brandon received support for article research from the U.S. government through BARDA, ASPR, and HHS (HHSO 100201900023C). Drs. Cermelli, Killien, Navalkar, Yager, and Brandon received funding from Immunexpress. Dr. Killien received NIH funding for grants (NICHD; funds paid to the institution) and a consultant fee (Retrospective Physician Diagnosis [RPD] panel) from Immunexpress, ending October 2023. Dr. George’s institution received funding from Gold Coast Hospital Foundation. Drs. George’s, Gibbons’, and Coin’s institution received funding from the Australian Government Medical Research Future Fund. Dr. Gibbons’ institution received funding from the Children’s Hospital Foundation and the National Health and Medical Research Council (NHMRC). Dr. Irwin received honoraria from bioMérieux and Gilead, paid to his institution, for lectures. Dr. Navalkar disclosed work for hire. Dr. Yager’s institution received funding from BARDA. Dr. Coin received support for article research from the Australian Government Medical Research Future Fund. Dr. Zapf received a consultant fee (RPD panel) from Immunexpress, ending October 2023. Dr. Schlapbach received support from the Swiss National Science Foundation (grant number 342730_153158/1 and 320030_201060/1), the NOMIS foundation, and the Thomas and Doris Ammann Foundation. Dr. Bandon’s institution received funding from the Australian Government Medical Research Future Fund (GHFMPACI000004; MRFF9100000), Children’s Hospital Foundation Queensland (50335), Emergency Medicine Foundation (EMPJ-378R27-2017-SNE), Gold Coast Hospital Foundation (RGS2017-LG0020), Far North Queensland Foundation, Townsville Hospital and Health Services SERTA Grant, Australian Infectious Diseases Research Centre, and the Swiss Personalized Health Technologies (PHRT-273); he received support for article research from the Australian Government Medical Research Future Fund (GHFMPACI000004), Children’s Hospital Foundation Queensland (50335), the Australian Government Medical Research Future Fund (MRFF9100000), Emergency Medicine Foundation (EMPJ-378R27-2017-SNE), Gold Coast Hospital Foundation (RGS2017-LG0020), Far North Queensland Foundation, Townsville Hospital and Health Services SERTA Grant, Australian Infectious Diseases Research Centre, the Swiss Personalized Health Technologies (PHRT-273), the Medical Research Future Fund (ARGCHDG000036), the NHMRC (GNT1140322), the HeartKids Foundation, Children’s Hospital Foundation, Perth Children’s Hospital Foundation, Intensive Care Foundation, Green Lane Research and Educational Fund, and Immunexpress. The remaining authors have disclosed that they do not have any potential conflicts of interest.
Contributor Information
Jerry J. Zimmerman, Email: jerry.zimmerman@seattlechildrens.org.
Sainath Raman, Email: sainath.raman@uq.edu.au.
Warwick W. Butt, Email: warwick.butt@rch.org.au.
Silvia Cermelli, Email: silvia.c@immunexpress.com.
Roy F. Davis, Email: roy.d@immunexpress.com.
Elizabeth Y. Killien, Email: elizabeth.killien@seattlechildrens.org.
Shane George, Email: shane.george@uq.edu.au.
Kristen S. Gibbons, Email: k.gibbons@uq.edu.au.
Adam D. Irwin, Email: a.irwin@uq.edu.au.
Krupa A. Navalkar, Email: krupa.n@immunexpress.com.
Thomas D. Yager, Email: thomas.y@immunexpress.com.
Florian H. Zapf, Email: florian.zapf@gmx.ch.
Lachlan J. M. Coin, Email: lachlan.coin@unimelb.edu.au.
Luregn J. Schlapbach, Email: luregn.schlapbach@kispi.uzh.ch.
REFERENCES
- 1.Watson RS, Carrol ED, Carter MJ, et al. : The burden and contemporary epidemiology of sepsis in children. Lancet Child Adolesc Health 2024; 8:670–681 [DOI] [PubMed] [Google Scholar]
- 2.Carter MJ, Carrol ED, Ranjit S, et al. : Susceptibility to childhood sepsis, contemporary management, and future directions. Lancet Child Adolesc Health 2024; 8:682–694 [DOI] [PubMed] [Google Scholar]
- 3.Souza DC de, Paul R, Mozun R, et al. : Quality improvement programmes in paediatric sepsis from a global perspective. Lancet Child Adolesc Health 2024; 8:695–706 [DOI] [PubMed] [Google Scholar]
- 4.Weiss SL, Fitzgerald JC, Pappachan J, et al. ; Sepsis Prevalence, Outcomes, and Therapies (SPROUT) Study Investigators and Pediatric Acute Lung Injury and Sepsis Investigators (PALISI) Network: Global epidemiology of pediatric severe sepsis: The sepsis prevalence, outcomes, and therapies study. Am J Respir Crit Care Med 2015; 191:1147–1157 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Miranda M, Nadel S: Pediatric sepsis: A summary of current definitions and management recommendations. Curr Pediatr Rep 2023; 11:29–39 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fleischmann-Struzek C, Goldfarb DM, Schlattmann P, et al. : The global burden of paediatric and neonatal sepsis: A systematic review. Lancet Respir Med 2018; 6:223–230 [DOI] [PubMed] [Google Scholar]
- 7.Rudd KE, Johnson SC, Agesa KM, et al. : Global, regional, and national sepsis incidence and mortality, 1990–2017: Analysis for the Global Burden of Disease Study. Lancet 2020; 395:200–211 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Schlapbach LJ, Watson RS, Sorce LR, et al. ; Society of Critical Care Medicine Pediatric Sepsis Definition Task Force: International consensus criteria for pediatric sepsis and septic shock. JAMA 2024; 331:665–674 [Google Scholar]
- 9.Goldstein B, Giroir B, Randolph A; International Consensus Conference on Pediatric Sepsis: International pediatric sepsis consensus conference: Definitions for sepsis and organ dysfunction in pediatrics. Pediatr Crit Care Med 2005; 6:2–8 [DOI] [PubMed] [Google Scholar]
- 10.Carlton EF, Perry-Eaddy MA, Prescott HC: Context and implications of the new pediatric sepsis criteria. JAMA 2024; 331:646–649 [DOI] [PubMed] [Google Scholar]
- 11.Jabornisky R, Kuppermann N, González-Dambrauskas S: Transitioning from SIRS to phoenix with the updated pediatric sepsis criteria—the difficult task of simplifying the complex. JAMA 2024; 331:650–651 [DOI] [PubMed] [Google Scholar]
- 12.Georgette N, Michelson K, Monuteaux M, et al. : Development of a new screening tool for pediatric septic shock. Ann Emerg Med 2024; 84:642–650 [DOI] [PubMed] [Google Scholar]
- 13.Weiss SL, Peters MJ, Alhazzani W, et al. : Surviving Sepsis Campaign international guidelines for the management of septic shock and sepsis-associated organ dysfunction in children. Pediatr Crit Care Med 2020; 21:e52–e106 [DOI] [PubMed] [Google Scholar]
- 14.Gerber JS, Jackson MA, Tamma PD, et al. ; COMMITTEE ON INFECTIOUS DISEASES, PEDIATRIC INFECTIOUS DISEASES SOCIETY: Antibiotic stewardship in pediatrics. Pediatrics 2021; 147:e2020040295 [Google Scholar]
- 15.Klouwenberg PMCK, Cremer OL, Vught LA van, et al. : Likelihood of infection in patients with presumed sepsis at the time of intensive care unit admission: A cohort study. Crit Care 2015; 19:319. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Salud D, Reeder RW, Banks RK, et al. ; Life After Pediatric Sepsis Evaluation (LAPSE) Investigators: Association of pathogen type with outcomes of children encountering community-acquired pediatric septic shock. Pediatr Crit Care Med 2022; 23:635–645 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Miller RRM, III, Lopansri BK, Burke JP, et al. : Validation of a host response assay, SeptiCyte LAB, for discriminating sepsis from systemic inflammatory response syndrome in the ICU. Am J Respir Crit Care Med 2018; 198:903–913 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Balk R, Esper AM, Martin GS, et al. : Validation of SeptiCyte RAPID to discriminate sepsis from non-infectious systemic inflammation. J Clin Med 2024; 13:1194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Balk R, Esper AM, Martin GS, et al. : Rapid and robust identification of sepsis using SeptiCyte RAPID in a heterogeneous patient population. J Clin Med 2024; 13:6044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Zimmerman JJ, Sullivan E, Yager TD, et al. : Diagnostic accuracy of a host gene expression signature that discriminates clinical severe sepsis syndrome and infection-negative systemic inflammation among critically ill children. Crit Care Med 2017; 45:e418–e425 [DOI] [PubMed] [Google Scholar]
- 21.Herberg JA, Kaforou M, Wright VJ, et al. ; IRIS Consortium: Diagnostic test accuracy of a 2-transcript host RNA signature for discriminating bacterial vs viral infection in febrile children. JAMA 2016; 316:835–845 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rautiainen L, Cirko A, Pavare J, et al. : Biomarker combinations in predicting sepsis in hospitalized children with fever. BMC Pediatr 2022; 22:272. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Schlapbach LJ, Ganesamoorthy D, Wilson C, et al. ; EUCLIDS consortium: Host gene expression signatures to identify infection type and organ dysfunction in children evaluated for sepsis: A multicentre cohort study. Lancet Child Adolesc Health 2024; 8:325–338 [DOI] [PubMed] [Google Scholar]
- 24.Schlapbach LJ, Gibbons KS, Horton SB, et al. ; NITRIC Study Group, the Australian and New Zealand Intensive Care Society Clinical Trials Group (ANZICS CTG), and the ANZICS Paediatric Study Group (PSG): Effect of nitric oxide via cardiopulmonary bypass on ventilator-free days in young children undergoing congenital heart disease surgery: The NITRIC randomized clinical trial. JAMA 2022; 328:38–47 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Gill PJ, Anwar MR, Thavam T, et al. ; Pediatric Research in Inpatient Setting (PRIS) Network: Identifying conditions with high prevalence, cost, and variation in cost in US Children’s Hospitals. JAMA Netw Open 2021; 4:e2117816. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Gravrand V, Mellot F, Ackermann F, et al. : Stratification of COVID-19 severity using SeptiCyte RAPID, a novel host immune response test. Viruses 2023; 15:419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.von der Forst M, Back L, Tourelle KM, et al. : SeptAsTERS—SeptiCyte® RAPID as assessment tool for early recognition of sepsis—a prospective observational study. Infection 2025; 53:953–965 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.R Core Team: R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. 2024. Available at: http://www.R-project.org/. Accessed September 1, 2024
- 29.Delong ER, Delong DM, Clarke-Pearson DL: Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 1988; 44:837–845 [PubMed] [Google Scholar]
- 30.Robin X, Turck N, Hainard A, et al. : pROC: An open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinf 2011; 12:77. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Leteurtre S, Duhamel A, Salleron J, et al. ; Groupe Francophone de Reanimation et d’Urgences: “PELOD-2: An update of the PEdiatric Logistic Organ Dysfunction score”. Crit Care Med 2013; 41:1761–1773 [DOI] [PubMed] [Google Scholar]
- 32.Honaker J, King G, Blackwell M: Amelia II: A program for missing data. J Stat Softw 2011; 45:1–47 [Google Scholar]
- 33.Langkamp DL, Lehman A, Lemeshow S: Techniques for handling missing data in secondary analyses of large surveys. Acad Pediatr 2010; 10:205–210 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Khan SS, Ahmad A, Mihailidis A: Bootstrapping and multiple imputation ensemble approaches for classification problems. J Intell Fuzzy Syst 2019; 37:7769–7783 [Google Scholar]
- 35.Lengquist M, Varadarajan A, Alestam S, et al. : Sepsis mimics among presumed sepsis patients at intensive care admission: A retrospective observational study. Infection 2024; 52:1041–1053 [DOI] [PMC free article] [PubMed] [Google Scholar]