• 
• 

Three Part Question

[In a newborn infant with suspected sepsis] are [blood parameters (CRP/FBC) useful] in [the diagnosis or ruling out of sepsis, bacteraemia or meningitis].

Clinical Scenario

A newborn infant is born at term. Her mother had prolonged rupture of membranes but no other risks for sepsis and so the baby is observed on the postnatal ward. The baby is reviewed at 6 hours-of-age as she is febrile (38 degrees Celsius). The infant is otherwise clinically well. A blood cell count (for leukocyte count), c-reactive protein (CRP) and blood culture are performed. Despite getting blood the cannula was not correctly sited. The tests are performed urgently and in light of the normal blood cell count and C reactive protein the team would like to know if it is safe to not give antibiotics.

Search Strategy

Keywords: newborn, sepsis, meningitis, blood cell count, c reactive protein, diagnostic tests (routine).
These terms were inputted as MeSH headings (exploded). Further searches (title and abstract) were performed with similar truncated textwords. Searches were limited to reviews (keyword: reviews: review, systematic review, meta-analysis).

The Cochrane library, (2008 issue 4, 27 hits), Medline, (1950 to Nov Week 3 2008, 149 hits), EMBASE, (1980 to 2008 week 49, 259 hits) were searched.

Search Outcome

Only relevant reviews (systematic reviews and meta-analyses) were included (four hits). Bibliographic lists of these papers were searched and a further systematic review found. One paper did not include blood cell count or CRP in its systematic review and was therefore excluded.A total of four papers were therefore reviewed.

Relevant Paper(s)

Author, date and country	Patient group	Study type (level of evidence)	Outcomes	Key results	Study Weaknesses
Fowlie 1998 Scotland	Review of 52 articles	Systematic review (level 1b)	Positive blood, CSF, urine or deep tissue culture or x-ray changes supported by bacteriological growth from endotracheal tube aspirate)	Haematological indices: (Sensitivity: 1-90%, Specificity: 44-98%, PPV:1-91%, NPV: 49-99%, Positive LR: 0.75-20.4, Negative LR: 0.32-1.12). CRP: (Sensitivity: 29-90, Specificity: 41-100, PPV:13-100, NPV: 38-98, Positive LR: 0.82-12.6, Negative LR: 0.22-1.11)	Only 23 of the 52 papers appraised used to review individual test results. 22 papers used tests other than haematological indices, CRP or swabs. Pooling not performed due to heterogeneity
Da Silva 1995 Canada	Review of 16 articles.	Systematic review (level 1b)	Clinical condition compatible with SBI and a positive culture from blood, CSF, urine, other normally sterile sites or post-mortem histopathological diagnosis	Leukocyte count and ratios: (Sensitivity: 17-90%, Specificity: 31-100%, PPV: 41-100%, NPV: 53-96%). Quantitiave CRP assessment: (Sensitivity: 58-100%, Specificity: 67-94%, PPV: 23-82%, NPV: 86-100%)	Unable to pool data in light of heterogeneity.
Baraff 1992 USA	Review of 19 articles	Systematic review(level 1b)	Documented positive bacterial cultures of either blood, CSF, urine, stool, joint fluid or bone aspirate and cellulites with or without a positive culture.	“Low risk” infants8 (95% confidence intervals, CI): Probability of SBI - 2.6% (1.5-4.0), bacteremia - 2.0% (0.8-3.1) and meningitis - 1.0% (0.2-2.4). “high risk” infants (95% CI): Probability of SBI - 24.3% (18.2-31.4), bacteremia – 12.8 (7.3-19.9) and meningitis – 3.9% (1.7-7.0). “high risk” vs “low risk” (confidence intervals): Risk ratios for SBI – 9.3 (4.6-20.9), bacteremia – 9.8 (3.5-24.9), meningitis (1.7-23.3).	CRP not appraised in this study. Not all studies used white cell count. Combination of clinical picture and investigations to define “high” or “low” risk. Clinical findings could be variable between included papers therefore not reproducible. No discussion of heterogeneity despite pooling of data.
Klassen 1992 Canada	Reveiw of 10 articles	Systematic review (level 1b)	Blood, CSF, Stool & Urine culture.	Patients who fulfil Rochester criteria (95%CI): Negative likelihood ratio: 0.03, (0 to 0.23).	Only Outpatient facilities included. Both Clinical and laboratory assessment included. Only 2 out of the 10 were prospective and used the white cell count (within the Rochester criteria**) to rule out SBI

Comment(s)

Neonatal sepsis carries a significant morbidity and mortality and therefore appropriate empirical treatment is often required. Haematological markers (CRP and blood cell count) are commonly employed in newborns with suspected sepsis. For a test to diagnose or rule out a serious bacterial infection it must have a sensitivity approaching 100% (i.e. not miss a case of sepsis) and/or have a negative predictive value approaching 100% (i.e. to exclude sepsis when the test is negative). This is particularly important in this case as missing a case of sepsis in this population group would have such significant implications. The tests should also be simple to perform, return quick results and be cheap. 1 3 6 3 papers did not solely look at newborns on a neonatal unit rather discussing infants of less than three months-of-age with a fever.2, 3 5 This is important when considering our question here as infants in the first 24 hours of life are more likely to have an infection (pre-test probability) and so sensitivity and negative predictive value become less useful (Da Silva et al1) and rather likelihood ratios become favourable (Klassen3 & Fowlie5). For example, our patient who was born following prolonged rupture of membranes and is febrile at 6 hours-of-age has a much higher pre-test probability than a 10 week-old-infant who is coryzal and febrile. Culture of blood, CSF, urine and stool were most commonly used as the outcome measure. The sensitivity of blood cultures is hard to quantify though one study, using post-mortem diagnosis, identified blood culture sensitivity as 82%.7 In terms of using blood cell count and CRP in isolation, none of the papers above would allow a clinician to safely rule out or diagnose serious bacterial infections in infants.1 2 3 5 None of the papers using likelihood ratios found values consistently outside of the range (0.1 to 10) necessary to alter pre-test probability.2 3 5 The papers using sensitivity and negative predictive value also did not find values consistently approaching 100% which we would require to rule out or diagnose SBI.1 5 Clinical assessment in combination with a number of tests was used in two papers.2 3 Klassen3 did report a likelihood ratio of 0.03 (0 to 0.23) in infants who fulfilled the “Rochester criteria”. The Rochester criteria defines infants as low risk of SBI if they are a previously healthy infant younger than 2 months of age without evidence of soft tissue or musculoskeletal infection who had white blood cell counts of 5-15,000mm-3, band form counts ≤1500mm-3, urinalysis ≤10 white blood cells/high power field (spun sediment), and stool ≤5 white blood cells (if diarrhoea is present).8 In this case it is difficult to assess the infant as previously health as she is only 6 hours old. A urinalysis would also need to be performed. Baraff2 also used the Rochester criteria but includes a lumbar puncture (LP) and suggests that in febrile infants (less than 3 months) who present from the community a complete blood count, blood culture, urinalysis and LP (if less than 8 weeks) should be performed. If the infant is clinically well with a normal white cell count, CSF and urinalysis then a stat dose of intramuscular ceftriaxome should be employed and the infant followed up in the community.2 This paper had several methodological flaws in its pooling of data and the alternative proposed offers little in the way of benefits as all these infants would require a lumbar puncture (irrespective of clinical picture) and so few infants would meet the criteria stipulated. The evidence therefore suggests that haematological parameters (blood cell count and/or CRP) can not be used, in isolation, to either diagnose or rule out serious bacterial infection in the newborn infant. The safest option would therefore be to continue antibiotics for a period until the gold standard (culture) returns as being negative.

Clinical Bottom Line

Blood cell count and CRP should not be used in making a decision to either commence or discontinue antibiotic therapy in a febrile infant (level 1b).

References

1. Fowlie PW, Schmidt B. Diagnostic tests for bacterial infection from birth to 90 days--a systematic review. Arch Dis Child Fetal Neonatal Ed 1998;78(2):F92-8
2. Da Silva O, Ohlsson A, Kenyon C Accuracy of leukocyte indices and C-reactive protein for diagnosis of neonatal sepsis: a critical review. Pediatr Infect Dis J 1995;14(5):362-6.
3. Baraff LJ, Oslund SA, Schriger DL, Stephen ML Probability of bacterial infections in febrile infants less than three months of age: a meta-analysis. Pediatr Infect Dis J 1992;11(4):257-64.
4. Klassen TP, Rowe PC. Selecting diagnostic tests to identify febrile infants less than 3 months of age as being at low risk for serious bacterial infection: a scientific overview. J Pediatr 1992;121(5 Pt 1):671-6.