Neonatal sepsis accounts for 15% of all neonatal deaths. Early detection enables prompt administration of antibiotic treatment, reducing morbidity and mortality. This study aims to review the sensitivity and specificity of procalcitonin in diagnosing microbiologically-proven sepsis in neonates to determine the optimal procalcitonin cut-off value for use in clinical practice.
Medline, EMBASE and PubMed were searched on 3 May 2023 for original studies in symptomatic neonates in whom both blood culture and procalcitonin levels were taken, and a procalcitonin cut-off with either sensitivity or specificity reported. Studies that included asymptomatic or culture-negative neonates in the proven sepsis group were excluded. Risk of bias was assessed using the Qualitative Assessment of Diagnostic Accuracy Studies 2 tool.
Nineteen original studies enrolling a total of 1920 symptomatic neonates (721 with proven sepsis) were included. Six studies used a procalcitonin cut-off of 0.5 ng/mL and found a sensitivity of 87% to 100% and specificity of 17% to 89%. Nine studies evaluated higher procalcitonin cut-off values between 0.99 ng/mL and 2 ng/mL, which were 67% to 98% sensitive and 41% to 89% specific. All other procalcitonin thresholds were neither sensitive nor specific. Meta-analysis was not performed because of high risk of bias within the identified studies.
This review found that procalcitonin was highly sensitive (87% to 100%) at a cut-off value of 0.5 ng/mL, although specificity varied greatly across all cut-off values reviewed. The variation in diagnostic accuracy between studies suggests that procalcitonin may be useful to guide antibiotic cessation but should not be used alone as a diagnostic marker for neonatal sepsis.
Neonatal sepsis is a leading cause of infant morbidity and mortality, accounting for 3.9 million infections annually and 15% of all neonatal deaths.1 Early diagnosis and rapid treatment with appropriate antimicrobial therapy are essential to address this. The clinical diagnosis of sepsis presents a major challenge in neonates because of nonspecific clinical symptoms and signs that are common to other neonatal conditions, including respiratory distress syndrome and viral infections.1 Blood culture remains the gold standard for diagnosing sepsis however has a low sensitivity and requires over 24 hours to return a result.2 Therefore, there is significant interest in biomarkers such as procalcitonin (PCT) to assist in the rapid and accurate identification of neonatal sepsis.
PCT is a 116 amino acid prohormone encoded by the CALC 1 gene and is produced in high quantities during bacterial infection.3 Lipopolysaccharide and other proinflammatory mediators trigger upregulation of CALC 1 gene expression in several nonendocrine tissues during bacterial infection.3 This leads to high levels of PCT production and release into the bloodstream, which can be detected within 2 to 4 hours of lipopolysaccharide exposure, with peak levels occurring within 24 hours.4,5 Elevation in serum PCT is not seen to this extent during viral infection as high levels of interferon-γ suppress the upregulation of CALC 1 gene expression.6 Therefore, PCT is an ideal biomarker to potentially distinguish viral and bacterial infections as well as identify uninfected neonates to improve antimicrobial rationalization.7 This study aims to review the sensitivity, specificity, and overall accuracy of PCT at different cut-off values as a diagnostic tool for microbiologically-proven bacterial sepsis in neonates with clinical features concerning for sepsis.
Methods
This systematic review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The MEDLINE (1946 to present), Embase (1947 to present) and PubMed (1996 to present) databases were searched on 3 May 2023. The search strategy was developed by 2 authors (R.B., A.G.) in consultation with an expert medical librarian. Search terms included keywords calcitonin, procalcitonin, sepsis, bacterial infection, urosepsis, septic, bacter?emia*, severe-infection*, severe-bacterial-infection*, newborn*, baby, neonat*, and infan* (Fig 1). Search results were independently screened by title and abstract by 2 authors (R.B., K.T.), and full texts of all relevant articles were reviewed by these authors. Any disagreements were resolved in discussion with a third author (A.G.).
All original studies that included neonates with clinical features suggestive of sepsis in whom blood culture and/or pan-bacterial 16S ribosomal ribonucleic acid polymerase chain reaction (16S rRNA PCR) and PCT levels were performed, and reported the sensitivity and specificity of a PCT cut-off value or included sufficient data for these to be calculated were included. Studies that enrolled blood culture and/or pan-bacterial PCR negative neonates in the proven sepsis group, had a sample size of fewer than 10 neonates or enrolled asymptomatic neonates in the control group were excluded. The latter, because the clinical utility of PCT is to identify or rule out bacterial sepsis in symptomatic rather than asymptomatic neonates (Fig 2).
Flow diagram of included and excluded articles. Adapted From: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. doi: 10.1136/bmj.n71.
Flow diagram of included and excluded articles. Adapted From: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. doi: 10.1136/bmj.n71.
Data, including patient demographics, study characteristics, microbiology results and PCT assay, sensitivity, specificity and cut-off value, were extracted by a single author (R.B.) from articles that met the inclusion criteria. The specificity and sensitivity were calculated where sufficient information was provided8–12 or authors contacted to further clarify eligibility.13 Risk of bias assessment was performed by 1 author (R.B.) using the Qualitative Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) on included studies14 (Fig 3). As per the Cochrane handbook for systematic reviews, a meta-analysis was not performed because of the high risk of bias of the included studies.15
Results
Overall, the search yielded 562 MEDLINE, 974 Embase, and 201 PubMed records. After removal of duplicates, 1061 articles remained, of which 69 full texts were reviewed with 18 studies meeting our inclusion criteria. The authors of 2 further papers were contacted to clarify eligibility,13,16 and 1 response was received13 resulting in a total of 19 original studies included in this review. The final studies included 9 cross-sectional studies,8,10,13,17–22 7 prospective cohort studies,9,11,12,23–26 1 retrospective case-control study,27 1 prospective case-control study,28 and 1 retrospective cohort study29 (Tables 1 and 2). One study included only preterm neonates,27 another only term neonates,13 with the remaining 17 studies including both. Also, 11 studies included neonates with both early-onset sepsis (EOS) and late-onset sepsis (LOS), 1 included only the former,9 whereas another 4 studies included only the latter11,22,24,29 (3 did not specify the onset of sepsis).10,20,25
Overview of Extracted Data
| Citation | Population Size (% BC+/ PCR+), n (%) | PCT Cut-off (ng/mL) | Sens, % | Spec, % | Risk of Biasa |
|---|---|---|---|---|---|
| Iskandar et al (2019)18 | 51 (69) | 0.16 | 69 | 63 | At risk |
| Rashwan et al (2018)28 | 168 (61) | 0.39 | 97 | 100 | At risk |
| Habib et al (2021)17 | 171 (50) | 0.5 | 98 | 71 | At risk |
| Yu et al (2020)12 | 60 (52) | 0.5 | 87 | 17 | Low risk |
| Rohsiswatmo et al (2020)22 | 52 (33) | 0.5 | 100 | 37 | At risk |
| Morad et al (2020)20 | 50 (82) | 0.5 | 98 | 89 | At risk |
| Sucilathangam et al (2012)23 | 50 (28) | 0.5 | 93 | 75 | At risk |
| Lubis (2021)19 | 30 (40) | 0.5 | 93 | 81 | At risk |
| López et al (2006)24 | 100 (61) | 0.59 | 81 | 81 | At risk |
| Fendler et al (2007)29 | 78 (81) | 0.99 | 98 | 89 | At risk |
| Blommendahl et al (2002)25 | 169 (8) | 1.0 | 77 | 62 | At risk |
| Adib et al (2011)8 | 69 (29) | 1.1 | 70 | 41 | At risk |
| Berka et al (2021)27 | 285 (23) | 1.2 | 79 | 79 | At risk |
| Al-Zahrani et al (2015)9 | 100 (34) | 1.7 | 85 | 50 | At risk |
| Shivasharana et al (2022)11 | 128 (33) | 2.0 | 79 | 51 | At risk |
| Sakha et al (2008)10 | 117 (23) | 2.0 | 67 | 50 | At risk |
| Boo et al (2008)26 | 87 (21) | 2.0 | 89 | 65 | At risk |
| Omar et al (2019)21 | 60 (20) | 2.0 | 67 | 67 | At risk |
| Lubis (2021)19 | 35 (43) | 2.5 | 78 | 62 | At risk |
| Kamiab et al (2019)13 | 60 (23) | 23 | 79 | 50 | At risk |
| Citation | Population Size (% BC+/ PCR+), n (%) | PCT Cut-off (ng/mL) | Sens, % | Spec, % | Risk of Biasa |
|---|---|---|---|---|---|
| Iskandar et al (2019)18 | 51 (69) | 0.16 | 69 | 63 | At risk |
| Rashwan et al (2018)28 | 168 (61) | 0.39 | 97 | 100 | At risk |
| Habib et al (2021)17 | 171 (50) | 0.5 | 98 | 71 | At risk |
| Yu et al (2020)12 | 60 (52) | 0.5 | 87 | 17 | Low risk |
| Rohsiswatmo et al (2020)22 | 52 (33) | 0.5 | 100 | 37 | At risk |
| Morad et al (2020)20 | 50 (82) | 0.5 | 98 | 89 | At risk |
| Sucilathangam et al (2012)23 | 50 (28) | 0.5 | 93 | 75 | At risk |
| Lubis (2021)19 | 30 (40) | 0.5 | 93 | 81 | At risk |
| López et al (2006)24 | 100 (61) | 0.59 | 81 | 81 | At risk |
| Fendler et al (2007)29 | 78 (81) | 0.99 | 98 | 89 | At risk |
| Blommendahl et al (2002)25 | 169 (8) | 1.0 | 77 | 62 | At risk |
| Adib et al (2011)8 | 69 (29) | 1.1 | 70 | 41 | At risk |
| Berka et al (2021)27 | 285 (23) | 1.2 | 79 | 79 | At risk |
| Al-Zahrani et al (2015)9 | 100 (34) | 1.7 | 85 | 50 | At risk |
| Shivasharana et al (2022)11 | 128 (33) | 2.0 | 79 | 51 | At risk |
| Sakha et al (2008)10 | 117 (23) | 2.0 | 67 | 50 | At risk |
| Boo et al (2008)26 | 87 (21) | 2.0 | 89 | 65 | At risk |
| Omar et al (2019)21 | 60 (20) | 2.0 | 67 | 67 | At risk |
| Lubis (2021)19 | 35 (43) | 2.5 | 78 | 62 | At risk |
| Kamiab et al (2019)13 | 60 (23) | 23 | 79 | 50 | At risk |
BC+, blood culture positive; PCR+, bacterial 16S ribosomal DNA polymerase chain reaction positive; sens, sensitivity; spec, specificity.
Risk of bias assessment performed using the QUADAS-2.
Summary of Study Characteristics and Extracted Data
| Citation | Study Group | Study Type | Country | PCT Cut-off, Sens and Spec for BC Proven Bacterial Sepsis |
|---|---|---|---|---|
| Shivasharana et al (2022)11 | 128 neonates PNA > 72 hr with suspected sepsis of whom 42 (33%) were BC+. Exclusion criteria: admission for surgical issue, life-threatening congenital abnormality, extra-mural babies, hypoxic-ischaemic encephalopathy stage III. | Single center prospective cohort study | India | PCT > 2 ng/mL (LOS): • sens 79%d, spec 51%d |
| Berka et al (2021)27 | 285 very preterm neonates (GA ≤ 32 w) PNA > 72 hr with clinically suspected sepsis of whom 66 (23%) BC+ with: • 49 (74%) gram positive: 36% S. aureus, 30% CoNS; • 17 (26%) Gram negative: 20% E. coli. Exclusion criteria: incomplete laboratory data, unknown outcomes, possible SIRS, other identified cause of clinical worsening. | Single center retrospective case-control study | Czech Republic | PCT > 1.2 ng/mL (LOS): • sens 79%, spec 79%; • AUC: 0.90; • overall accuracy 79% |
| Habib et al (2021)17 | 171 neonates (median GA 37 w) with clinically suspected sepsis of whom 86 (50%) were BC+. Of these, 102 (60%) with EOS, 69 (40%) with LOS. Exclusion criteria: antibiotics before BC collection, congenital disorders, respiratory distress syndrome, metabolic disorders. | Single center prospective cross-sectional study | Pakistan | PCT ≥ 0.5ng/mL: All neonates (PNA ≤ 28 d); • sens 98%, spec 71%; • AUC: 0.87; • overall accuracy 84%; EOS (PNA ≤ 7 d); • sens 98%, spec 65%; • overall accuracy 82%; LOS (PNA > 7 d); • sens 97%, spec 78%; • overall accuracy 87% |
| Lubis (2021)19 | 65 neonates (41 preterm) with clinically suspected sepsis of whom 27 (42%) BC+ with: • 6 (22%) gram positive: 7% M. luteus; • 21 (78%) gram negative: 33% Klebsiella spp., 26% Enterobacter spp. 11% A. baumannii; of the 65 neonates, 35 (54%) with EOS, 30 (46%) with LOS. Exclusion criteria: immunodeficiency, hematological disorders, receiving immunomodulatory drugs. | Single center prospective cross-sectional study | Indonesia | PCT > 0.5 ng/mL (LOS): • sens 93%, spec 81%; • AUC: 0.89; PCT > 2.5 ng/mL (EOS): • sens 78%, spec 62%; • AUC: 0.68 |
| Yu et al (2020)12 | 60 neonates (mean GA 37.4±3.4 w) with clinically suspected sepsis of whom 31 (52%) were BC+ with: • 18 (58%) gram positive: 32% S. aureus; 23% E. faecalis; • 13 (42%) gram negative: 29% P. aeruginosa; of the 60 neonates 19 (32%) had EOS, 41 (68%) had LOS. Exclusion criteria not detailed. | Single center prospective cohort study | China | PCT ≥ 0.5 ng/mL: • sens 87%d, spec 17%d |
| Rohsiswatmo et al (2020)22 | 52 neonates PNA > 72 hr (46 preterm) with clinically suspected sepsis of whom 17 (33%) were BC+ with: • 41% Acinetobacter, 23% K. pneumoniae. Exclusion criteria: multiple congenital abnormalities, postsurgical, birth weight < 1000 g | Single center prospective cross-sectional study | Indonesia | PCT ≥ 0.5 ng/mL (LOS): • sens 100%, spec 37%; • AUC: 0.686 |
| Morad et al (2020)20 | 50 neonates (19 preterm) with clinically suspected sepsis of whom 31/50 (62%) were BC+ and 39/50 (78%) were 16S rDNA PCR+a with: • 6 (19%) gram positive: 10% CoNS, 7% GBS; • 24 (77%) gram negative: 61% Klebsiella spp., 10% E. coli; • 1 (3%) gungal: Candida albicans. Exclusion criteria: major congenital abnormalities, Apgar score <7, antibiotic therapy before start of study. | 2 center prospective cross-sectional study | Egypt | PCT ≥ 0.5 ng/mL: • sens 98%, spec 89%; • overall accuracy 96% |
| Omar et al (2019)21 | 60 neonates (mean GA 34±4.6 w) with suspected sepsis of whom 12 (20%) were BC+. Of these, 17 (28%) with EOS, 43 (72%) with LOS. Exclusion criteria not detailed | Single center prospective cross-sectional study | Malaysia | PCT > 2 ng/mL: • sens 67%, spec 67%; • AUC: 0.71 |
| Kamiab et al (2019)13 | 60 term neonates with clinically suspected sepsis of whom 14 (23%) were BC+. Of these, 34 (57%) with EOS, 26 (43%) with LOS. Exclusion criteria: neonatal death, left hospital before results received, disease other than sepsis during hospitalization. | Single center prospective cross-sectional study | Iran | PCT > 23 ng/mL: • sens 79%, spec 50% |
| Iskandar et al (2019)18 | 51 neonates (27 preterm) who fulfill SIRS criteria of whom 35 (69%) were BC+ with: • 11 (31%) gram positive: 31% CoNS; • 22 (63%) gram negative: 23% K. pneumoniae, 23% Enterobacter gergoviae; • 2 (6%) gungal: yeast-like fungi. Of the 35 BC+ neonates, 11 (31%) with EOS, 24 (69%) with LOS. Exclusion criteria not detailed. | Single center prospective cross-sectional study | Indonesia | PCT > 0.16 ng/mL: • sens 69%, spec 63%; • overall accuracy 67% |
| Rashwan et al (2018)28 | 168 neonates (mean GA 38±0.6 w) with clinical signs of sepsis of whom 102 (61%) were BC+ with: • 24 (24%) gram positive: 15% S. aureus, 9% Enterococcus spp.; • 78 (76%) gram negative: 28% Klebsiella spp., 9% E. coli. Of the 102 BC+ neonates 47 (46%) with EOS, 55 (54%) with LOS. Exclusion criteria: congenital malformations, prematurity, Apgar < 7, antibiotic therapy before sepsis suspected. | Single center prospective case-control study | Egypt | PCT > 0.39 ng/mL: • sens 97%, spec 100% |
| Al-Zahrani et al (2015)9 | 100 neonates (20 preterm) with clinically suspected EOS of whom 25/100 (25%) were BC+ and 32/100 (32%) 16S rDNA PCR+b with: • 7 (28%) gram positive: 12% GBS, 8% CoNS; • 16 (64%): gram negative: 28% Klebsiella spp., 20% E. coli, 8% Enterobacter spp.; • 2 (8%): unspecified. Exclusion criteria: congenital malformation, TORCH infection. | Single center prospective cohort study | Saudi Arabia | PCT > 1.7 ng/mL (EOS): • sens 85%d, spec 50%d |
| Sucilathangam et al (2012)23 | 50 neonates (22 preterm) with clinically suspected sepsis of whom 14 (28%) were BC+ with: • 9 (64%) gram positive: 35% Acinetobacter, 14% S. aureus, 14% CoNS; • 5 (36%): gram negative: 21% Klebsiella spp. Of the 50 neonates: 29 (58%) with EOS, 21 (42%) with LOS. Exclusion criteria: antibiotics within 72hrs of symptoms, birth asphyxia, aspiration syndromes, congenital abnormalities, inborn errors of metabolism. | Single center prospective cohort study | India | PCT ≥ 0.5 ng/mL: • sens 93%, spec 75% |
| Adib et al (2011)8 | 69 neonates with clinically suspected sepsis of whom 20 (29%) were BC+ with: • 15 (75%) gram positive: 30% S. aureus, 30% CoNS, 15% GAS; • 5 (25%): gram negative: 15% E. coli. Of the 20 BC+ neonates: 13 (65%) with EOS, 7 (35%) with LOS. Exclusion criteria not detailed. | 2 center prospective cross- sectional study | Iran | PCT > 1.1ng/mL: • sens 70%d, spec 41%d |
| Sakha et al (2008)10 | 117 neonates (GA > 35 w) of whom 27 (23%) were BC+. Exclusion criteria: congenital abnormality, GA < 34 w, suspected hemorrhage, birth asphyxia. | Single center retrospective cross-sectional study | Iran | PCT > 2.0 ng/mL: • sens 67%d, spec 50%d; PCT > 1.36 ng/mL: • sens 56%, Sspec 63%; • AUC: 0.61 |
| Boo et al (2008)26 | 87 neonates (60 preterm) of whom 18 (21%) were BC+ with: • 12 (67%) gram positive: 44% CoNS, 11% S. aureus; • 6 (33%) Gram negative: 11% Acinetobacter spp., 11% Klebsiella spp. Of the 18 BC+ neonates: 2 (11%) EOS, 16 (89%) LOS. Exclusion criteria: prior antibiotics within 72 hr. | Single center prospective cohort study | Malaysia | PCT ≥ 2 ng/mL: All neonates; • sens 89%, spec 65%; LOS (PNA < 48 hr); • sens 88%, spec 77% |
| Fendler et al (2007)29 | 78 episodes of LOS among 73 neonates PNA > 48 hr (median GA 29 w) of whom 63 (81%) were BC+ with: • 47 (75%) gram positive; • 16 (25%) gram negative. Exclusion criteria: prior sepsis within 7 d. | Single center retrospective cohort study | Poland | PCT > 0.99 ng/mL (LOS): • sens 98%, spec 89%; • AUC: 0.93 |
| López et al (2006)24 | 100 neonates PNA > 4 d (med GA 30w) with clinically suspected sepsis of whom 61 (61%) were BC+ with: • 28 (46%) gram positive: 36% CoNS, 10% Entercoccus spp.c • 30 (49%) gram negative: 25% K pneumoniaec, 11% E.coli, 10% E. cloacae; • 2 (3%) Fungal: 3% Candida spp.; • 1 (2%) Mixed infection by Enterococcus and K. pneumoniae. Exclusion criteria: suspected vertical transmission. | Multi center prospective cohort study | Spain | PCT > 0.59 ng/mL: • sens 81%, spec 81%; • AUC: 0.78 |
| Blommendahl et al (2002)25 | 169 neonates (median GA 264 d) of whom 13 (8%) were BC+ with: • 13 (100%) gram positive: 46% GBS, 23% S. aureus, 15% CoNS. Exclusion criteria: prior antibiotics (including maternal antibiotic treatment). | Single center prospective cohort study | Finland | PCT > 1 ng/mL: • sens 77%, spec 62% |
| Citation | Study Group | Study Type | Country | PCT Cut-off, Sens and Spec for BC Proven Bacterial Sepsis |
|---|---|---|---|---|
| Shivasharana et al (2022)11 | 128 neonates PNA > 72 hr with suspected sepsis of whom 42 (33%) were BC+. Exclusion criteria: admission for surgical issue, life-threatening congenital abnormality, extra-mural babies, hypoxic-ischaemic encephalopathy stage III. | Single center prospective cohort study | India | PCT > 2 ng/mL (LOS): • sens 79%d, spec 51%d |
| Berka et al (2021)27 | 285 very preterm neonates (GA ≤ 32 w) PNA > 72 hr with clinically suspected sepsis of whom 66 (23%) BC+ with: • 49 (74%) gram positive: 36% S. aureus, 30% CoNS; • 17 (26%) Gram negative: 20% E. coli. Exclusion criteria: incomplete laboratory data, unknown outcomes, possible SIRS, other identified cause of clinical worsening. | Single center retrospective case-control study | Czech Republic | PCT > 1.2 ng/mL (LOS): • sens 79%, spec 79%; • AUC: 0.90; • overall accuracy 79% |
| Habib et al (2021)17 | 171 neonates (median GA 37 w) with clinically suspected sepsis of whom 86 (50%) were BC+. Of these, 102 (60%) with EOS, 69 (40%) with LOS. Exclusion criteria: antibiotics before BC collection, congenital disorders, respiratory distress syndrome, metabolic disorders. | Single center prospective cross-sectional study | Pakistan | PCT ≥ 0.5ng/mL: All neonates (PNA ≤ 28 d); • sens 98%, spec 71%; • AUC: 0.87; • overall accuracy 84%; EOS (PNA ≤ 7 d); • sens 98%, spec 65%; • overall accuracy 82%; LOS (PNA > 7 d); • sens 97%, spec 78%; • overall accuracy 87% |
| Lubis (2021)19 | 65 neonates (41 preterm) with clinically suspected sepsis of whom 27 (42%) BC+ with: • 6 (22%) gram positive: 7% M. luteus; • 21 (78%) gram negative: 33% Klebsiella spp., 26% Enterobacter spp. 11% A. baumannii; of the 65 neonates, 35 (54%) with EOS, 30 (46%) with LOS. Exclusion criteria: immunodeficiency, hematological disorders, receiving immunomodulatory drugs. | Single center prospective cross-sectional study | Indonesia | PCT > 0.5 ng/mL (LOS): • sens 93%, spec 81%; • AUC: 0.89; PCT > 2.5 ng/mL (EOS): • sens 78%, spec 62%; • AUC: 0.68 |
| Yu et al (2020)12 | 60 neonates (mean GA 37.4±3.4 w) with clinically suspected sepsis of whom 31 (52%) were BC+ with: • 18 (58%) gram positive: 32% S. aureus; 23% E. faecalis; • 13 (42%) gram negative: 29% P. aeruginosa; of the 60 neonates 19 (32%) had EOS, 41 (68%) had LOS. Exclusion criteria not detailed. | Single center prospective cohort study | China | PCT ≥ 0.5 ng/mL: • sens 87%d, spec 17%d |
| Rohsiswatmo et al (2020)22 | 52 neonates PNA > 72 hr (46 preterm) with clinically suspected sepsis of whom 17 (33%) were BC+ with: • 41% Acinetobacter, 23% K. pneumoniae. Exclusion criteria: multiple congenital abnormalities, postsurgical, birth weight < 1000 g | Single center prospective cross-sectional study | Indonesia | PCT ≥ 0.5 ng/mL (LOS): • sens 100%, spec 37%; • AUC: 0.686 |
| Morad et al (2020)20 | 50 neonates (19 preterm) with clinically suspected sepsis of whom 31/50 (62%) were BC+ and 39/50 (78%) were 16S rDNA PCR+a with: • 6 (19%) gram positive: 10% CoNS, 7% GBS; • 24 (77%) gram negative: 61% Klebsiella spp., 10% E. coli; • 1 (3%) gungal: Candida albicans. Exclusion criteria: major congenital abnormalities, Apgar score <7, antibiotic therapy before start of study. | 2 center prospective cross-sectional study | Egypt | PCT ≥ 0.5 ng/mL: • sens 98%, spec 89%; • overall accuracy 96% |
| Omar et al (2019)21 | 60 neonates (mean GA 34±4.6 w) with suspected sepsis of whom 12 (20%) were BC+. Of these, 17 (28%) with EOS, 43 (72%) with LOS. Exclusion criteria not detailed | Single center prospective cross-sectional study | Malaysia | PCT > 2 ng/mL: • sens 67%, spec 67%; • AUC: 0.71 |
| Kamiab et al (2019)13 | 60 term neonates with clinically suspected sepsis of whom 14 (23%) were BC+. Of these, 34 (57%) with EOS, 26 (43%) with LOS. Exclusion criteria: neonatal death, left hospital before results received, disease other than sepsis during hospitalization. | Single center prospective cross-sectional study | Iran | PCT > 23 ng/mL: • sens 79%, spec 50% |
| Iskandar et al (2019)18 | 51 neonates (27 preterm) who fulfill SIRS criteria of whom 35 (69%) were BC+ with: • 11 (31%) gram positive: 31% CoNS; • 22 (63%) gram negative: 23% K. pneumoniae, 23% Enterobacter gergoviae; • 2 (6%) gungal: yeast-like fungi. Of the 35 BC+ neonates, 11 (31%) with EOS, 24 (69%) with LOS. Exclusion criteria not detailed. | Single center prospective cross-sectional study | Indonesia | PCT > 0.16 ng/mL: • sens 69%, spec 63%; • overall accuracy 67% |
| Rashwan et al (2018)28 | 168 neonates (mean GA 38±0.6 w) with clinical signs of sepsis of whom 102 (61%) were BC+ with: • 24 (24%) gram positive: 15% S. aureus, 9% Enterococcus spp.; • 78 (76%) gram negative: 28% Klebsiella spp., 9% E. coli. Of the 102 BC+ neonates 47 (46%) with EOS, 55 (54%) with LOS. Exclusion criteria: congenital malformations, prematurity, Apgar < 7, antibiotic therapy before sepsis suspected. | Single center prospective case-control study | Egypt | PCT > 0.39 ng/mL: • sens 97%, spec 100% |
| Al-Zahrani et al (2015)9 | 100 neonates (20 preterm) with clinically suspected EOS of whom 25/100 (25%) were BC+ and 32/100 (32%) 16S rDNA PCR+b with: • 7 (28%) gram positive: 12% GBS, 8% CoNS; • 16 (64%): gram negative: 28% Klebsiella spp., 20% E. coli, 8% Enterobacter spp.; • 2 (8%): unspecified. Exclusion criteria: congenital malformation, TORCH infection. | Single center prospective cohort study | Saudi Arabia | PCT > 1.7 ng/mL (EOS): • sens 85%d, spec 50%d |
| Sucilathangam et al (2012)23 | 50 neonates (22 preterm) with clinically suspected sepsis of whom 14 (28%) were BC+ with: • 9 (64%) gram positive: 35% Acinetobacter, 14% S. aureus, 14% CoNS; • 5 (36%): gram negative: 21% Klebsiella spp. Of the 50 neonates: 29 (58%) with EOS, 21 (42%) with LOS. Exclusion criteria: antibiotics within 72hrs of symptoms, birth asphyxia, aspiration syndromes, congenital abnormalities, inborn errors of metabolism. | Single center prospective cohort study | India | PCT ≥ 0.5 ng/mL: • sens 93%, spec 75% |
| Adib et al (2011)8 | 69 neonates with clinically suspected sepsis of whom 20 (29%) were BC+ with: • 15 (75%) gram positive: 30% S. aureus, 30% CoNS, 15% GAS; • 5 (25%): gram negative: 15% E. coli. Of the 20 BC+ neonates: 13 (65%) with EOS, 7 (35%) with LOS. Exclusion criteria not detailed. | 2 center prospective cross- sectional study | Iran | PCT > 1.1ng/mL: • sens 70%d, spec 41%d |
| Sakha et al (2008)10 | 117 neonates (GA > 35 w) of whom 27 (23%) were BC+. Exclusion criteria: congenital abnormality, GA < 34 w, suspected hemorrhage, birth asphyxia. | Single center retrospective cross-sectional study | Iran | PCT > 2.0 ng/mL: • sens 67%d, spec 50%d; PCT > 1.36 ng/mL: • sens 56%, Sspec 63%; • AUC: 0.61 |
| Boo et al (2008)26 | 87 neonates (60 preterm) of whom 18 (21%) were BC+ with: • 12 (67%) gram positive: 44% CoNS, 11% S. aureus; • 6 (33%) Gram negative: 11% Acinetobacter spp., 11% Klebsiella spp. Of the 18 BC+ neonates: 2 (11%) EOS, 16 (89%) LOS. Exclusion criteria: prior antibiotics within 72 hr. | Single center prospective cohort study | Malaysia | PCT ≥ 2 ng/mL: All neonates; • sens 89%, spec 65%; LOS (PNA < 48 hr); • sens 88%, spec 77% |
| Fendler et al (2007)29 | 78 episodes of LOS among 73 neonates PNA > 48 hr (median GA 29 w) of whom 63 (81%) were BC+ with: • 47 (75%) gram positive; • 16 (25%) gram negative. Exclusion criteria: prior sepsis within 7 d. | Single center retrospective cohort study | Poland | PCT > 0.99 ng/mL (LOS): • sens 98%, spec 89%; • AUC: 0.93 |
| López et al (2006)24 | 100 neonates PNA > 4 d (med GA 30w) with clinically suspected sepsis of whom 61 (61%) were BC+ with: • 28 (46%) gram positive: 36% CoNS, 10% Entercoccus spp.c • 30 (49%) gram negative: 25% K pneumoniaec, 11% E.coli, 10% E. cloacae; • 2 (3%) Fungal: 3% Candida spp.; • 1 (2%) Mixed infection by Enterococcus and K. pneumoniae. Exclusion criteria: suspected vertical transmission. | Multi center prospective cohort study | Spain | PCT > 0.59 ng/mL: • sens 81%, spec 81%; • AUC: 0.78 |
| Blommendahl et al (2002)25 | 169 neonates (median GA 264 d) of whom 13 (8%) were BC+ with: • 13 (100%) gram positive: 46% GBS, 23% S. aureus, 15% CoNS. Exclusion criteria: prior antibiotics (including maternal antibiotic treatment). | Single center prospective cohort study | Finland | PCT > 1 ng/mL: • sens 77%, spec 62% |
AUC, area under receiver operating characteristic curve; BC, blood culture; BC+, blood culture positive; d, days; GA, gestational age; GAS, Group A Streptococcus; GBS, Group B Streptococcus; MRSA, methicillin-resistant Staphylococcus aureus; PNA, post-natal age; sens, sensitivity; SIRS, systemic inflammatory response syndrome; spec, specificity; TORCH, toxoplasmosis, syphilis, hepatitis B, rubella, cytomegalovirus, herpes simplex virus; w, weeks.
Total of 41 neonates either BC+ and/or 16S PCR+.
Total of 34 neonates either BC+ and/or 16S rDNA PCR+.
1 neonate with mixed infection by K. pneumoniae and Enterococcus spp.
Calculated.
All included neonates were clinically suspected of having sepsis, however specific inclusion criteria were detailed in only 12 studies. Most commonly this included temperature instability, apnoea, tachypnoea, or tachycardia (9 studies), feeding intolerance, bradycardia or other signs of respiratory distress (8 studies), and lethargy or poor peripheral perfusion (7 studies). Neonates were commonly excluded because of congenital malformation (8 studies) or prior antibiotic administration within 72 hours (5 studies).
The identified studies included a total of 1920 symptomatic neonates, of whom 721 (38%) had proven sepsis as defined by either positive blood culture or positive pan-bacterial 16S rDNA PCR. A pathogen was reported in 507 of 721 (70%) episodes. For the 507 neonates in whom a pathogen was identified, a gram-positive bacterium was isolated in 247 (49%), a gram-negative bacterium in 253 (50%), and fungi in 7 (1%) cases. Of the 19 studies, only 1 study required 2 positive blood cultures for the diagnosis of sepsis due to a potential contaminant or low-virulence bacterium such as coagulase-negative staphylococci (CoNS).24
PCT Assays
Serum PCT levels were measured using 7 different assay techniques, each with different measurement ranges. The most common assays were immunoluminometric assay (range 0.1–500 ng/mL, 6 studies), enzyme linked immunosorbent assay (range 0.03–2.0 ng/mL, 2 studies and 0.03–200 ng/mL, 2 studies), chemiluminescence immunoassay (range 0.02–100 ng/mL, 3 studies) and electrochemiluminescence immunoassay (range 0.02–100 ng/mL, 2 studies).
PCT Cut-off Values
Studied PCT cut-off values ranged from 0.16 ng/mL to 23 ng/mL, with the most common cut-off value of 0.5 ng/mL used in 6 studies (Table 1).
PCT Cut-off <0.3 ng/mL
A small single center study found a low PCT cut-off of 0.16 ng/mL was neither sensitive nor specific (69% and 63%, respectively) for proven sepsis in 51 neonates (35 with positive blood culture).18 Notably, this study used a PCT assay with an upper limit of quantification of 2 ng/mL, a lower PCT level than recorded in most septic neonates, which may have influenced the diagnostic accuracy for identifying proven sepsis. This study had a high risk of bias for the reference and index tests.
PCT Cut-off 0.39 to 0.59 ng/mL
A PCT cut-off value of 0.5 ng/mL was investigated in 6 studies, including a total of 413 neonates and found to be highly sensitive (81% to 100%) but with variable specificity (17% to 89%). The largest of these studies enrolled 171 neonates (86 with positive blood culture) and found a high sensitivity of 98% and specificity of 71%, with overall diagnostic accuracy of 84%.17 Of note, subgroup analysis showed diagnostic accuracy was slightly higher in LOS (87%) compared to EOS (82%). These findings are consistent with a smaller study of 50 neonates that used either a positive 16S rDNA PCR or blood culture to define proven sepsis and found PCT to be highly sensitive (98%) and specific (89%) at this cut-off.20 However, a study of 60 neonates (31 blood culture-positive) found the same cut-off value to be poorly specific (17%) although still highly sensitive (87%) in diagnosing blood culture-positive neonates. In this study, the high proportion (56 of 60, 93%) of neonates with a positive 16S rRNA PCR may have accounted for the poor specificity of PCT.12 These findings were supported by 3 small single center studies that found a PCT cut-off of 0.5 ng/mL to be highly sensitive (93% to 100%) but with variable specificity (37% to 81%).19,22,23
Two studies in this group determined the optimal PCT cut-off value based on receiver-operating characteristic curves using the Youden’s index. A PCT cut-off of 0.39 ng/mL was both highly sensitive (97%) and specific (100%) for diagnosing bacterial sepsis in a large study of 168 neonates (61% with proven sepsis).28 However, with a slightly higher PCT value of 0.59 ng/mL, the sensitivity fell to 81% in a study of 100 neonates with LOS.24
Of the 8 studies, 4 had a high risk of bias for the index test, and 1 had a high risk of bias for the reference test using the QUANDAS-2 tool.
PCT Cut-off 0.99 to 2.0 ng/mL
Higher PCT cut-off values of 0.99 to 2.0 ng/mL were evaluated in 9 studies and were associated with reduced sensitivity (67% to 89%) and variable specificity (41% to 89%). Only 1 study found high diagnostic accuracy of PCT at this level: a retrospective cohort study of 78 episodes of neonatal sepsis that reported a 98% sensitivity and 89% specificity using a 0.99 ng/mL cut-off value.29 Notably, this study had an unusually high rate of positive blood cultures (63 of 78, 81%), suggesting included neonates may have been more unwell.
All other studies within this range of cut-off values had poor sensitivity. The largest of these included 285 preterm neonates (gestational age ≤ 32 weeks) with LOS (66 culture positive) and found PCT was 79% sensitive and specific using a calculated optimal cut-off of 1.2 ng/mL.27 Similarly, a cut-off value of 1.0 ng/mL had only moderate sensitivity and specificity (77% and 62%, respectively) among 169 neonates.25 In this study, only 13 of 169 (8%) neonates had a positive blood culture that may have contributed to the low specificity in this case. Similarly, a poor specificity (42%) and sensitivity (70%) of a PCT value of 1.1 ng/mL was reported in a multicenter prospective study of 69 neonates (20 with bacterial sepsis).8
A higher cut-off value of 2.0 ng/mL did not improve specificity (50% to 67%) for proven sepsis in 4 single center studies, including a total of 392 neonates (99 blood culture positive). As expected, sensitivity (67% to 89%) was also lower at this PCT cut-off.10,11,22,27
Of these 5 studies, 2 had high risk of bias for the reference test using the QUANDAS-2 tool.
PCT Cut-off >2 ng/ml
Neither specificity (50% to 62%) nor sensitivity (78% to 79%) increased at higher optimal PCT cut-off values of 2.5 ng/mL and 23 ng/mL calculated using receiver-operating characteristic curves. In 35 neonates (15 proven sepsis) with EOS, PCT was 78% sensitive and 62% specific at a PCT cut-off of 2.5 ng/mL.19 Sensitivity (79%) and specificity (50%) were similar at a much higher PCT cut-off value of 23 ng/mL in a study of 60 neonates (23% with proven sepsis).13 Notably, this was the only study to exclude preterm neonates, potentially suggesting a greater PCT response to infection in term neonates. These studies both had high risk of bias in the index test, with the latter also having high risk of bias in patient selection using the QUADAS-2 tool.
Discussion
This review found a PCT cut-off of 0.5 ng/mL to be highly sensitive (87% to 100%) but with variable specificity for diagnosing microbiologically-proven neonatal sepsis. Therefore, PCT may be useful in decision making regarding the cessation of antibiotics in neonates. This supports the recommendations in the American Academy of Pediatrics management of neonates born at ≥35 0/7 weeks’ gestation with suspected or proven early onset bacterial sepsis guideline, stating that consistently normal PCT values indicate the absence of sepsis, however elevated PCT does not confirm infection nor need for antibiotic administration. Notably, a PCT cut-off value is not recommended in these guidelines.30 Across all PCT cut-off values assessed, none were consistently specific for the diagnosis of neonatal sepsis. These findings are similar to studies of PCT in different patient cohorts, including burns and postoperative sepsis.31 In pediatric burns patients, Rosanova et al found PCT to be highly sensitive (91%) but poorly specific (18%) for differentiating sepsis from systemic inflammatory response syndrome at a cut-off value of 0.5 ng/mL.32
Factors contributing to the poor specificity of PCT in neonates include the mild elevation caused by several noninfectious conditions in neonates. In EOS, respiratory distress syndrome from meconium aspiration, birth asphyxia, or persistent pulmonary hypertension of the newborn may cause elevated PCT.20 Viral infections, such as varicella zoster virus and herpes simplex virus, among others, have been found to cause a mild PCT elevation to 1.5 ng/mL in previous studies,33 however none of the studies included in this review performed viral PCR testing. Also, the physiological PCT rise that occurs in the first 48 hours of life may represent a significant barrier to distinguishing true sepsis from other conditions in this age group.20
Up to 7.4% of neonates are administered antibiotics in the first 72 hours of life, despite culture-proven sepsis occurring in less than 0.1% of cases.7 A potential benefit of using PCT for diagnosing neonatal sepsis is reducing unnecessary antibiotic exposure in neonates which can increase the risk of subsequent necrotizing enterocolitis, invasive fungal disease, and all-cause mortality in preterm neonates.34 Preventing unnecessary antibiotic administration also reduces the risk of antimicrobial resistance and the burden on healthcare systems.7 However, PCT is not widely accessible, particularly in resource poor settings, where the high cost of specialized equipment and testing may be prohibitive. Despite this, a recent study from Egypt found that repeat PCT measurements reduced the length of stay in the NICU by 30%, outweighing the higher cost of PCT compared to C-reactive protein (CRP).35 In Australia, a single PCT test costs more than 3 times that of CRP, and almost equivalent to blood culture, bacterial subtype, and sensitivity, which provides more clinically useful information.36,37 However, the difference in cost between a hospital ward bed and a NICU bed is over 2000 Australian dollars per night, which may potentially justify the increased cost of PCT testing to enable early discharge from NICU.38 PCT can be a sensitive biomarker using a cut-off value of 0.5 ng/mL, however its overall diagnostic accuracy ranged from 67% to 92%, which may not be sufficient to justify the use of this test in resource-poor settings.
This review is limited by the absence of high-quality prospective studies restricting the ability to perform a meta-analysis of the data. Of the 19 included original studies, 18 did not clearly define the need for 2 blood cultures to confirm the diagnosis of proven sepsis when potential contaminants were detected. Other limitations are the small study cohorts, single or dual center design, and heterogenous inclusion criteria. Included studies also used 7 different PCT assays with 7 different quantification ranges, including 2 assays designed to detect PCT levels close to the physiological baseline level (<0.05 ng/mL).6 In neonates with abnormally high PCT levels during sepsis, significant differences in quantification between assays is likely to skew the results. Finally, a major limitation is that of the included original studies, only 702 (37%) neonates had a positive blood culture. However, it is possible that because of the low sensitivity of blood culture (40% to 83%) when insufficient blood volume is drawn, such as in unwell or low birth weight neonates, an elevated PCT in the absence of a positive culture may still indicate true sepsis.2,39
Conclusions
This review found that PCT was highly sensitive (87% to 100%) using a PCT cut-off value of 0.5 ng/mL, although specificity was variable at all cut-off values reviewed. In clinical practice, this suggests that PCT may be useful to aid decision making around ceasing rather than commencing antibiotics. The variation in diagnostic accuracy between studies and assay types suggests that PCT should not be used as the sole diagnostic marker for neonatal sepsis, particularly where there is clinical concern.
Acknowledgement
We thank Ms. Poh Chua, expert medical librarian, The Royal Children’s Hospital Melbourne, for her assistance with the literature search.
A/Prof Gwee conceptualized and designed the study, reviewed and revised the manuscript, provided supervision of the methodology, data analysis, and manuscript preparation; Dr Beaumont designed the study, performed the literature search; led data collection, analysis and interpretation; drafted the initial manuscript, reviewed and revised the manuscript, and collated resources; Dr Tang performed the literature search; and all authors approved the final manuscript for submission.
FUNDING: No external funding. A/Prof Gwee receives salary support from a National Health and Medical Research Council Investigator Grant.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no conflicts of interest to disclose.
Comments