Review Article Published on October 14, 2021

 

Role of  Biomarkers in Neonatal Sepsis. Are we in search of the Holy Grail?

P Madhava Chandran Naira, Anila V Panackalb

 

a. KIMS & Ananthapuri Hospitals, TVM, Kerala, India; b. Senior Registrar, KIMS Healthcare, Trivandrum

 

 

Abstract

Neonatal sepsis continues to be a common and significant health care burden, especially in very-low-birth-weight infants (<1500 g) with high mortality. The symptoms and signs are quite nonspecific resulting in under or over treatment with unnecessary antibiotics. The traditional markers like septic screen and hematological scores (total WBC count, Absolute Neutrophil count, IT ratio, peripheral smear), are a far cry from ideal. The gold standard blood culture including the Bactec is not full proof and time consuming. Acute phase reactant like C-Reactive Protein has good specificity, but poor sensitivity. Procalcitonin is expensive. The new biomarkers like Serum Amyloid A, Cytokines and chemokines like IL-6,IL-8, Tumour Necrosis Factor alpha, Cell surface markers like nCD64, the ‘Omics’ like Genomics, Metabolomics and Proteomics and the new biophysical markers like heart  rate variability and core temperature differences, and the role of artificial intelligence (AI)in early diagnosis of sepsis are being discussed. The ultimate search for an ideal biomarker with 100% sensitivity and specificity is still going on, so that ultimately neonatal sepsis can be diagnosed before symptoms manifest resulting in early treatment and good prognosis.

 

Neonatal sepsis continues to be a common and significant health care burden, especially in very-low-birth-weight infants (<1500 g). It is still one of the most important causes of neonatal morbidity and mortality. The incidence of Neonatal sepsis in India is approximately 30/1000 live births (NNPD- National Neonatal Perinatal Database). Signs and symptoms of neonatal sepsis are quite subtle and non-specific and hence diagnosis and treatment are difficult and challenging. These include fever or hypothermia, respiratory distress including cyanosis and apnea, feeding difficulties, lethargy or irritability, hypotonia, seizures, bulging fontanel, poor perfusion, bleeding problems, abdominal distention, hepatomegaly, unexplained jaundice, or more importantly, “just not looking right”.1 Rapid clinical deterioration ensues unless prompt antibiotic management is started in neonates with sepsis. Early and reliable identification or exclusion of sepsis is important, so that unnecessary or prolonged use of antibiotics can be avoided, and improved outcome can be guaranteed.1

Why do we need biomarkers?

  1. Neonatal sepsis presents with subtle signs and symptoms and deteriorate very fast. Hence early diagnostic markers are of paramount importance.              
  2. Many conditions like inborn errors of metabolism mimic sepsis
  3. Traditional tests like Blood culture takes time and positivity is less.
  4. Peculiarities of the fragile neonatal immune system
  5. Broad-spectrum antibiotics and invasive fungal infections
  6. Multi- drug resistance
  7. Increasing hospital stay / cost
  8. Unwanted treatment and risk of side effects
  9. Early unnecessary antibiotic use and association with asthma, inflammatory
  10. Bowel disease, obesity etc

What is an ideal biomarker

1.   High sensitivity: able to detect infection when it is present(100%)

2.   High specificity: ability to rule out sepsis when it is not present (>85%)

3.   High predictive value: Likelihood that the test accurately predicts presence or absence of sepsis (approaching 100%) in order to minimize unnecessary use of antibiotics in false positive cases.

4.   Discriminate etiology of sepsis: identify cause as viral, bacterial, fungal

5.   Timely results: Quick laboratory turn over time (necessary in early sepsis diagnosis)

Old and new biomarkers of Neonatal sepsis (NS)

Cultures: of blood, urine, or cerebrospinal fluid are usually the gold standard. But the turnaround time is takes 48-72 hours for the result. Sensitivity of blood cultures to detect sepsis in neonates is high when 1ml is inoculated and the infant has a bacteraemia concentration of at-least 4- colony -forming -units.2 Prior antibiotic treatment and low volume of collection can give false-negative reports. Contamination can give false positive results. The new BACTEC culture system can give results within 24 hours, even with low volumes of blood and a low colony count.

Hematological indices3-5

Total Leukocyte Count (TLC):

WBC count has poor predictive value in diagnosis of Early onset neonatal sepsis (EONS). Moreover, in EONS, diagnostic accuracy of neutrophil indices is more reliable if obtained after 6–12 hours, thus delaying the diagnosis. Leukopenia is more important than leucocytosis in very low birth weight babies. Severe sepsis usually manifests as leukopenia. Morphological or degenerative changes in neutrophils, such as vacuolisation, Döhle bodies, intracellular bacteria, and toxic granules in peripheral smear are also helpful but not specific.

Absolute Neutrophil Count (ANC)

Blood ANC varies in a neonate with postnatal age and gestation with the lower limit being <1,800/µL at birth, <7,800/µL at 12–14 hours of age and falling again to <1,800/µL at 72hours. We usually follow Manroe charts for term babies and Mouzinho charts for preterm babies (Figures 1 & 2).

Figure 1. Manroe chart for total neutrophil count range in term babies

Figure 2. Mouzinho chart for total neutrophil count range in preterm babies

I/T ratio (Immature to Total Leukocyte Ratio):

It is a sensitive indicator of sepsis. Values >0.27 in term and >0.22 in preterm neonates are significant. It has a high sensitivity of 90% and negative predictive value of 98%.

Micro Erythrocyte Sedimentation Rate (Micro ESR):

Traditionally used for septic screen but lacks sensitivity. It takes a few days for ESR to rise and the value varies significantly within the first  few days of life. Furthermore, the levels take a significant amount of time to return to normal again.

Platelets:4,5

Thrombocytopeniais a late marker of sepsis and NEC. Persistent low platelet counts are seen in fungal sepsis.

New: Increased mean platelet volume (>8.6 fL) has been studied recently as a marker of EOS and a predictor for mortality, especially in preterm neonates with a sensitivity of 97.14% and a specificity of nearly 100% and hence a promising early marker of EOS.

False positive: bronchopulmonary dysplasia; respiratory distress

Red blood cell distribution width ( RDW):4-6

In neonates, the normal RDW is 15.5% to 20% and as high as 23% can be considered normal in the preterm population. Elevated red cell–distribution width can predict mortality in EOS or LOS. RDW >16.35% was found to predict mortality with 70% sensitivity and 66.1% specificity in a population of 500 term infants diagnosed with EOS or LOS4

HSS- HEMATOLOGICAL SCORING SYSTEM4-6

Combining hematological indices with clinical findings and other biomarkers has been shown to improve diagnostic accuracy further, especially in cases of culture-negative sepsis. Sepsis is probable with a HSS score ≥3 (Table 2). This test has a high sensitivity of 96%, but a low positive predictive value of 31%.Neutrophil- Lymphocyte ratio (NLR):7

In neonatal sepsis there is an increase in neutrophil count as an immediate response accompanied by decrease in lymphocytes and monocytes, and hence an increase in neutrophil-lymphocyte ratio during sepsis. This is a marker of inflammation and along with CRP was found to be an easy, inexpensive rapid marker of late onset sepsis with no extra lab costs. NLR is now used as a prognostic marker of COVID-19 infection in assessing risk for mortality. Research on NLR in neonatal sepsis is limited and most studies are from adult population.

G-CSF (Granulocyte Colony Stimulating Factor), a mediator produced by the bone marrow for facilitating the proliferation and differentiation of neutrophils, proposed as a reliable infection marker for early diagnosis.6 Based on a cut off 200 pg/ml, it has a high sensitivity (95%) and negative predictive value (99%) for predicting early neonatal bacterial and fungal infections.

Clotting factors: Septic neonates are also prone to develop hemorrhagic and thrombotic complications. Circulating thrombin-antithrombin III complex, plasminogen activator inhibitor-1, plasminogen tissue activator, fibrinogen, and D-dimer concentrations are significantly raised in infected infants compared with non-infected patients.

Limitation: However, sick preterm newborns with respiratory distress syndrome also have deranged coagulation and fibrinolysis.

Acute Phase Reactants

Acute phase proteins are produced principally by the liver as part of an immediate inflammatory response to infection or tissue injury.

C-Reactive Protein8-10

Acute phase reactant synthesized by liver. It has a half- life of 24–48 hours. It takes 10–12 hours for CRP to change significantly after onset of infection. It is a late rising marker and hence of limited use in diagnosis of Neonatal sepsis and to start antibiotics (30 to 60% sensitivity only).

Serial determination of CRP 24–48 h after onset of symptoms increases its sensitivity (by 82% and 84%, respectively). Serial CRP measurement is helpful in monitoring response to treatment and help clinicians guide duration of antibiotic therapy. Specificity & positive predictive value is 93–100%. Thus, CRP is a “specific” but “late” marker of neonatal infection.

CRP is useful as a negative predictor of neonatal sepsis- If next 48 hours serial CRP measurements are negative and baby is stable, it correlates strongly with absence of infection thereby guiding safe discontinuation of antibiotic therapy.

Limitations: CRP sensitivity is low during the early phase of sepsis. It takes 10–12 hours to change significantly after onset of infection and hence it is not an ideal marker. False elevation of CRP can occur in a variety of non-infectious conditions like MAS, traumatic or ischemic tissue injuries, hemolysis or histologic chorioamnionitis, post-surgery and recent vaccinations.

In Pre-terms, high sensitivity assays of CRP (hs CRP)- can detect lower grade of inflammation. CRP remains a useful negative predictor of NS, and serial levels are valuable in helping clinicians monitor response to treatment. Using CRP in combination with other biomarkers, particularly those that rise early during infection, such as nCD64, IL6, or IL8 is beneficial and serial measurements of CRP 24 hours apart if negative can rule out sepsis and stopantibiotics9

Procalcitonin (Pct)8-10

Pct is an acute phase reactant (prohormone of calcitonin) produced by hepatocytes & macrophages. Serum concentration of Pct begin to rise 2-4 hours after exposure to bacterial endotoxin, peak at 6 to 8 hours & remain elevated for at least 24 to 48 hours, allowing for a relatively wide window for detection. Half-life is about 25–30 hours. Pct values are not affected by gestational age. Sensitivity & Specificity 83–100% & 70–100% respectively. In bacterial and fungal infections, it is elevated.

Pct in EONS has a sensitivity of 92%, specificity of 97%, positive predictive value of 94%, and negative predictive value of 96%. Pct is one of the best markers for EOS (in combination with others) and guide the duration of antibiotics.

Limitations: Pct is increased in fetal distress, & in infants born to mothers with chorioamnionitis (maternal GBS colonization & PROM>18 hours), neonatal hypoxia (HIE), RDS, Pneumothorax, and intracranial bleeding. Specific nomograms are needed for neonates and is expensive.

In summary, Pct as an intermediate- to early-rising inflammatory marker, but its use in detection and early diagnosis of NS is limited.  Pct is not sufficiently sensitive to act as a sole indicator to initiate empiric antibiotic therapy but can be useful when combined with other biomarkers.  Pct is also useful to guide the duration of antibiotic therapy.

New biomarkers for neonatal sepsis:9-11

Serum amyloid A (SAA)

Serum amyloid A (SAA) is an early acute phase reactant and is produced in the liver as apolipoprotein (Apo) SAA in response to infection and inflammation. Estimation of SAA levels in cord blood samples at birth, could be helpful in diagnosis and initiation of antibiotic therapy in EOS and in prognosticating mortality. It has a sensitivity and specificity of 96%and automated rapid testis available.

Limitation: Hepatic and nutritional status may affect the values- limiting SAA’s use in LOS

Presepsin:

Recently recognized, potential biomarker for diagnosis of neonatal sepsis. Results are obtained within 15 minutes with blood specimens of only 100 µL. Availability of automated testing using chemiluminescence is present.

Limitation: Identification of reliable reference cut offs needed.

Chemokines and cytokines9-11

Among the various cytokines released by the immature neonatal immune system, the major ones are  interleukin-1β (IL-Iβ), interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-2 soluble receptor (SIL2R), & tumor necrosis factor α (TNF-α) .They rise early in response to bacterial infection even before onset of symptoms. Since they do not cross placental barrier, elevations found in umbilical cord blood can predict the possibility of infant going to develop sepsis in first few hours of life. 

IL-6: IL-6 is a proinflammatory marker synthesized by mononuclear cells, chorion, amnion, and trophoblastic cells, and proves to be an early marker in sepsis. This protein triggers the production of CRP and thus its levels elevate prior to CRP. Half-life is very short. Sensitivity falls by 24-48 hours. Hence IL-6 is an early, sensitive marker of neonatal infection. Diagnostic accuracy improved by combining IL-6 (early and sensitive) & CRP (late and specific) in first 48 hours of presumed clinical sepsis.

TNF-α  (Tumour Necrosis factor-alpha) :Proinflammatory cytokine

Concentrations are significantly higher in infected versus non-infected newborns. Diagnostic accuracy is equal to Pct. Sensitivity & specificity increases to 60% &100% respectively when TNF-α & IL-6 levels are combined. TNF- alpha stimulates IL-6 production, the combination of IL-6, TNFα, and CRP has a sensitivity and negative predictive value of approximately ≥90% for diagnosing EOS.

IL-8:A pro-inflammatory cytokine produced by monocytes, macrophages and endothelial cells with kinetics like IL-6.IL-8mediates activation & chemotaxis of neutrophils and its level rises and falls within 4 hours of infection; it also has a sensitivity of 90% and has a varied specificity between 75–100%. It is a marker for sepsis and associated with severity of infection.

Limitation: Requires specialized equipment like as in a research-laboratory setting. Turnaround time is around 6 hours.

Cell surface markers11-13

Cell surface antigens are seen on blood cells and can be detected by flow cytometry. It requires only a very low volume (0.05 ml) of whole blood.  Neutrophil cluster of differentiation (CD)CD11β & CD64 are reliable markers for detecting EOS & LOS with high sensitivity & specificity. Their expression increases within minutes following exposure to bacterial products.

CD64

CD64 is a high affinity antibody receptor which binds to the Fc region of the immunoglobulins that increase in infection. It is expressed at a very low level on surface of neutrophils in the absence of an infection. Expression of CD64 on activated neutrophils markedly increases after an episode of bacterial infection.9

The sensitivity of CD64 in diagnosing EOS is 80% and negative predictive value is 89%. Combined with CRP and IL, its sensitivity may reach 100%. It is an early marker since levels rise within 1–6 hours of bacterial invasion, and remain elevated for >24 hours, hence a promising target for early detection of Neonatal sepsis. Addition of IL-6 or CRP to CD64 levels further enhances its sensitivity& negative predictive value to 100%.9,10 With CD64 use, clinicians can discontinue antibiotics within 24 hours in non-infected newborns. In summary, nCD64 is of limited use as a sole marker of NS, but when combined with CRP can be useful in guiding the decision to continue antibiotics beyond 36–48 hours.

CD163: in EOS with sensitivity up to 100%, can differentiate between infectious and non-infectious conditions

CD11β: This cell surface marker isα-subunit of the β2-integrin adhesion molecule, involved in neutrophil adhesion, diapedesis, and phagocytosis. Detectable within 5 min in response to bacterial infection with sensitivity & specificity as high as 96–100% and 100%.

sTREM1 (soluble Triggering Receptor Expressed on Myeloid Cells-1)

Pilot study shows this to be a potential biomarker for NS, with sensitivity of 70%–100% and specificity 71%–100% and a predictor of septic shock and death. Urinary sTREM1 levels have also been studied.

Limitations: Cost, need for sophisticated equipment &processing time are barriers to the use of these markers in clinical practice. It is at an experimental level only as of now.

Genomics12-15

Detection of specific genomes of microbes by DNA sequencing- eg; PCR

(Multiplex PCR Assay): When 16S RNA–subunit PCR was evaluated alongside blood culture in a prospective clinical trial, sensitivity, specificity, negative predictive value (NPV), and positive predictive value of 100%, 95.4%, 77.2%, and 100%, respectively, were reported. In contrast to blood cultures that take up to 3-5 days to grow, results from 16S rRNA–subunit PCR are available within hours. Multiplex PCR assay (NeoSepID) tailored to detect the eight most common pathogens, can give results in 4 hours. Species are identified with specific probes. It helps to stop antibiotics at an earlier time point-or narrow down the spectrum of antibiotic.

Limitation: non-availability of the entire genome and lack of pure isolates for antibiotic-susceptibility testing.

Metabolomics13-15

Metabolites or metabolic pathways produced by the microorganisms in response to sepsis are used in diagnosis of sepsis. Spectrometric analysis (such as magnetic resonance spectrometry, nuclear magnetic resonance, and gas-chromatography mass spectrometry) is done. Wide scope is there in the future to enable the identification of more biomarkers and better diagnostic accuracy for EOS.

Eg; MALDI-TOF-Matrix assisted laser desorption/ionization time-of-flight mass spectrometry (MS) on cord-blood samples, identify a precocious haptoglobin “switch-on” pattern for the diagnosis of EOS

Proteomics

This involves the use of proteins produced by the immune system’s defenses in response to infection, for diagnosing sepsis. They include Damage associated proteins like S-100 (calgranulin), heat shock proteins, altered matrix proteins (MMP-8), IL-8 mRNA. They are produced by the fetus in response to inflammation and have a protective role. Definitive and rapid diagnosis of neonatal sepsis is possible.

More markers in the pipeline: Interferon gamma-induced Pr 10, Integrin alpha M, inter-alpha inhibitor proteins (IAIP), MCP-1,PSP, alpha-1-acid glycoprotein,  Neopterin, Visfatin and Resistin, ischemia-modified albumin, hepcidin, pentraxin 3 (PTX3), angiopoietins, suPAR etc

Inter α Inhibitor Proteins (IAIP)

The inter alpha inhibitor family of proteins (IAIP) are serine protease inhibitors, secreted by liver. They provide protection from increased protease activity associated with systemic immune system activation that accompanies sepsis and inflammation and is involved in extracellular matrix stabilization, inflammation, wound healing, anti-inflammatory & regulatory role in infection. IAIP levels are significantly lower in septic neonates. Sensitivity 89.5%, specificity 99%, a positive predictive value 95% & negative predictive value of  98%.

Artificial intelligence (AI) and Machine learning in Sepsis16

AI is an evolving field in Medicine. Processing electronic clinical notes and by machine learning, dedicated algorithms are made which can predict early the chance of sepsis, impact with antibiotics and prognosis.

The SERA algorithm: The sepsis early risk assessment algorithm are 2 interlinked algorithms of which one is a diagnostic algorithm (shows if patient has sepsis at the time of consultation -AUC sensitivity 0.94 & specificity -0.85 and PPV0.85 ) and the second an early prediction algorithm (determine the patient’s risk of having sepsis in the next 4,6,12,24 and 48 hours. SERA algorithm predicts whether patient has high risk of sepsis before being diagnosed and the area under the curve (AUC) increases to >0.94, 12 hours before sepsis. Thus, machine learning models can identify sepsis in NICU hours prior to clinical recognition.

Biosensors: Nano material based electro-chemical biosensors make early detection platforms for different biomarkers.

Biophysical markers17

Heart rate-variability monitoring is a sensitive clinical method to alert the clinician early in neonatal sepsis. HeRO score is a measure obtained from continuous analysis of electrocardiography for changes in heart-rate variability. A large multicenter trial with continuous analysis of ECG for changes in heart rate in sepsis showed statistically significant decrease in mortality.

Limitation: It is not specific. False positives are seen in surgery, non-infection related deteriorations in respiratory status, unknown causes. At present, the HeRO score remains a marker for a clinician to consider NS, rather than an absolute indication to start treatment.

RALIS is a computerized-algorithm device that relies on heart rate, respiratory rate, core temperature, body weight, number of desaturation events (<85%), and number of brady cardiac events to be documented (heart rate <100 beats per minute) which helps in detection of sepsis. A Pilot study had shown a sensitivity of 95.8% and specificity of 77.3%. Changes in the RALIS score were noted on average of 33 hours before the onset of clinical symptoms, prompting early diagnosis and treatment. The high NPV of the RALIS score helps to with hold or shorten the duration of antibiotic therapy with greater confidence but larger prospective trials are needed.

Core-peripheral Temperature Gradient (Continuous Measurement)

In a recent study, a core- peripheral temperature gradient >2°C sustained for at ≥4 hours were able to predict NS with 83% sensitivity and a NPV of 94%. Though a great potential for early sepsis detection, validation is required.

Neonatal sepsis calculator (Figure 3)

It is an interactive calculator developed to assess the risk of early onset sepsis for infants ≥ 34 weeks gestation at birth. When the necessary inputs are given it gives the probability of EOS per 1000 babies considering also maternal risk factors along with infant presentation. With its widespread adoption, there has been a threefold decrease in lab tests for sepsis including blood culture. There has been a decrease in empiric antibiotic use in the first 24 hours of life, from 5% to 2.6% and significant positive impact on mother–infant bonding and breastfeeding seen.

Figure 3. Neonatal sepsis risk calculator

Conclusion

An ideal biomarker satisfying all the needed criteria is still evading. New biomarkers need sophisticated technology and presently utility of most of them are confined to research labs. Cost of diagnosis and accessibility is an important limiting factor in low- and middle-income countries.

We must consider the cost of diagnosis versus cost of prolonged hospital stays, mortality, and morbidity when selecting a biomarker. A biomarker which uses minimal volume of blood is preferred especially in pre-terms

Early initiation of treatment is crucial and hence a biomarker with least turnaround time and which gives a precise diagnosis should be chosen

The trends seem to be moving away from reliance on conventional markers, such as CBC and CRP, to advanced monitoring algorithms and population-based risk calculators. There is no single biomarker that can confirm or refute NS with 100% certainty, so we must continue to evaluate accessible markers alongside clinical history and examination findings. The search for “the Holy Grail” continues!

References

  1. Shah BA, Padbury JF. Neonatal sepsis: an old problem with new insights. Virulence. 2014 Jan 1;5(1):170–8.
    [Pubmed] | [Crossref]
  2. Schelonka RL, Chai MK, Yoder BA, Hensley D, Brockett RM, Ascher DP. Volume of blood required to detect common neonatal pathogens. J Pediatr. 1996 Aug;129(2):275–8.
    [Pubmed] | [Crossref]
  3. Ng P. Diagnostic markers of infection in neonates. Arch Dis Child Fetal Neonatal Ed. 2004 May;89(3):F229–35.
    [Pubmed] | [Crossref]
  4. Stoll BJ, Hansen N, Fanaroff AA, Wright LL, Carlo WA, Ehrenkranz RA, et al. Late-onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics. 2002 Aug;110(2 Pt 1):285–91.
    [Pubmed] | [Crossref]
  5. Gandhi P, Kondekar S. A Review of the Different Haematological Parameters and Biomarkers Used for Diagnosis of Neonatal Sepsis. EMJ Hematol. 2019;7[1]:85-92.
    [Source]
  6. Makkar M, Gupta C, Pathak R, Garg S, Mahajan NC. Performance Evaluation of Hematologic Scoring System in Early Diagnosis of Neonatal Sepsis. J Clin Neonatol. 2013;2(1):25–9.
    [Pubmed] | [Crossref]
  7. Sumitro KR, Utomo MT, Widodo ADW. Neutrophil-to-Lymphocyte Ratio as an Alternative Marker of Neonatal Sepsis in Developing Countries. Oman Med J. 2021 Jan;36(1):e214.
    [Pubmed] | [Crossref]
  8. Polin RA, Newborn  the COFA. Management of Neonates With Suspected or Proven Early-Onset Bacterial Sepsis. Pediatrics. 2012 May 1;129(5):1006–15.
    [Crossref]
  9. Gilfillan M, Bhandari V. Neonatal sepsis biomarkers: where are we now?. RRN. 2019 Mar 14;9:9–20.
    [Crossref]
  10. Ng PC, Lam HS. Biomarkers for late-onset neonatal sepsis: cytokines and beyond. Clin Perinatol. 2010 Sep;37(3):599–610.
    [Pubmed] | [Crossref]
  11. Franz AR, Steinbach G, Kron M, Pohlandt F. Reduction of unnecessary antibiotic therapy in newborn infants using interleukin-8 and C-reactive protein as markers of bacterial infections. Pediatrics. 1999 Sep;104(3 Pt 1):447–53.
    [Pubmed] | [Crossref]
  12. Streimish I, Bizzarro M, Northrup V, Wang C, Renna S, Koval N, et al. Neutrophil CD64 with hematologic criteria for diagnosis of neonatal sepsis. Am J Perinatol. 2014 Jan;31(1):21–30.
    [Pubmed] | [Crossref]
  13. Shi J, Tang J, Chen D. Meta-analysis of diagnostic accuracy of neutrophil CD64 for neonatal sepsis. Ital J Pediatr. 2016 Jun 7;42:57.
    [Pubmed] | [Crossref]
  14. Fanos V, Van den Anker J, Noto A, Mussap M, Atzori L. Metabolomics in neonatology: fact or fiction? Semin Fetal Neonatal Med. 2013 Feb;18(1):3–12.
    [Pubmed] | [Crossref]
  15. van den Brand M, van den Dungen FAM, Bos MP, van Weissenbruch MM, van Furth AM, de Lange A, et al. Evaluation of a real-time PCR assay for detection and quantification of bacterial DNA directly in blood of preterm neonates with suspected late-onset sepsis. Crit Care. 2018 Apr 22;22(1):105.
    [Pubmed] | [Crossref]
  16. Wu M, Du X, Gu R, Wei J. Artificial Intelligence for Clinical Decision Support in Sepsis. Front Med (Lausanne). 2021;8:665464.
    [Pubmed] | [Crossref]
  17. Coggins SA, Weitkamp J-H, Grunwald L, Stark AR, Reese J, Walsh W, et al. Heart rate characteristic index monitoring for bloodstream infection in an NICU: a 3-year experience. Arch Dis Child Fetal Neonatal Ed. 2016 Jul;101(4):F329-332.
    [Pubmed] | [Crossref]
  18. Van Herk W, Stocker M, van Rossum AMC. Recognising early onset neonatal sepsis: an essential step in appropriate antimicrobial use. J Infect. 2016 Jul 5;72 Suppl:S77-82.
    [Pubmed] | [Crossref]
  19. Balayan S, Chauhan N, Chandra R, Kuchhal NK, Jain U. Recent advances in developing biosensing based platforms for neonatal sepsis. Biosens Bioelectron. 2020 Dec 1;169:112552.
    [Pubmed] | [Crossref]
  20. Takkar VP, Bhakoo ON, Narang A. Scoring system for the prediction of early neonatal infections. Indian Pediatr. 1974 Sep;11(9):597–600.
    [Pubmed]