Where can I find this paper?
What is this paper about (what is the research question)?
How good is procalcitonin at predicting the presence of serious bacterial infection in children – and in particular, those aged ≤ 3months of age?
Summary of the Paper
Design: Retrospective multicentre diagnostic trial.
Objective: to assess the value of procalcitonin in diagnosing serious bacterial illness and in particular invasive bacterial infections in well-appearing infants with “Fever Without Source” (known as pyrexia of unknown origin in the UK)
Reference standard: serious bacterial infection, defined as isolation of pathogen from blood, CSF, urine or stool, or invasive bacterial infection – isolation of bacterial pathogen from blood or CSF.
Test of interest: procalcitonin measured using quantitative immunochemiluminesence assay.
Population: Retrospective inclusion of cases from 5 Spanish and 2 Italian PEDs
- Inclusion: infants ❤ months of age with fever of unknown origin for whom procalcitonin was measured and blood culture performed.
- Exclusion: patients in whom source of fever could be identified, patients classed as appearing unwell at presentation, patients who were afebrile without thermometer confirmation of fever prior to attendance, patients in whom procalcitonin was not measured, not recorded or blood culture not obtained.
Results: 1531 infants ❤ months of age with fever of unknown origin identified over 3 year period (exact recruitment period variable by centre) across three centres. 1112 included for analysis.
26% were diagnosed with serious bacterial infection, 2.1% with invasive bacterial infection.
Multivariate logistic regression analysis was performed for risk factor variables. Only PCT ≥ 0.5ng/mL was an independent risk factor for invasive bacterial infection: OR 21.69 (95% CI 7.93-59.28), with an area under the curve for SBI 0.739 (95% CI 0.702-0.776) and for IBI 0.825 (95% CI 0.698-0.952).
Among young infants with fever without source, procalcitonin is a better marker than CRP for identifying patients with invasive bacterial infection and also seems to be the best marker for ruling out invasive bacterial infection.
On the study design
We have discussed previously the issues of retrospective design. There are issues here about the inclusion and exclusion criteria; many are reliant on subjective variables such as “appearing well”. Arguably this has inherent reliability problems and we could argue the accuracy of physician assessment is dependent on exposure to paediatric patients and experience levels.
Inclusion of patients for whom both procalcitonin and blood culture were performed seems to instinctively select out a population for whom there are red flag signs (or symptoms) of underlying illness, even though the subjects were considered to appear “well” and no source of fever was apparent on examination.
The issue of reference standard is important too. We know that some children are undoubtably septic but no organism is definitively cultured, despite best efforts. Absence of diagnostic culture does not equate to absence of serious or invasive bacterial infection.
What were the results and what does this mean?
The results here are a little complicated, so bear with me.
It’s difficult to even produce a 2×2 table; the raw data doesn’t seem to be there. Procalcitonin <0.5ng/mL has a negative LR of 0.66 (95% CI 0.59-0.73) which implies that the odds of a patient with a procalcitonin level <0.5ng/mL having underlying serious bacterial illness is 0.66 times the odds that the patient with a procalcitonin above this level has underlying serious bacterial illness. The LR of a negative procalcitonin (<0.5ng/mL) for invasive bacterial illness is even more compelling: 0.25 (95% CI 0.12-0.55).
As we have seen on other papers, the confidence intervals are quite wide.
As for the predictive power of a raised procalcitonin, where the level was ≥2.0ng/mL the LR was 7.12 for serious bacterial illness (95% CI 4.52-11.21) and 11.14 for invasive bacterial illness (95% CI 7.81-15.89). So the odds that a patient with a raised procalcitonin has underlying serious bacterial illness is 7.12 times the odds that a patient with a raised calcitonin doesn’t have serious bacterial illness, and for invasive bacterial illness the odds of disease presence are 11.14 times the odds that there is no disease when the procalcitonin is ≥2.0ng/mL
What can we take from this paper into clinical practice?
Procalcitonin seems to be a useful indicator that underlying bacterial illness might be present (≥2.0ng/mL) or absent (<0.5ng/mL).
Looking at the area under the ROC curves, it doesn’t appear that procalcitonin is quite good enough to definitively rule in or out the presence of underlying bacterial illness. The retrospective design and tight inclusion criteria also limit the extrapolation of these results to the undifferentiated PED population.
More questions to ask
- How useful would procalcitonin be when tested prospectively in an undifferentiated PED population – including those who appear “sick”?
- What do we clinically do with patients whose procalcitonin falls between these two cutoff points?
- How would the results be affected by a standardised investigative approach to infants with febrile illness without clear source?
Follow us on twitter: @PEMLit