Fever is a very common cause of paediatric consultation. Physical examination cannot identify the focus of the infection in many patients, and this problem becomes more accentuated the younger the child. Although the majority of children present minor infections, it is important to identify those with serious bacterial infection (SBI) in order to start antibiotic treatment early.1 2 The risk of SBI varies with age; in infants who present with fever, it develops in 8% to 14% of those under 1 month of age, in 5% to 9% of those between 1 and 3 months of age, and in 3% to 15% of those over 3 months of age.2^–5

Traditionally, the management of fever in infants under 2 months of age included the use of complementary investigations and the admission of all patients to hospital.6 Later, the tendency was to select patients with a low risk of SBI by using evaluation scales (Children's Hospital of Philadelphia, Philadelphia, USA; Dagan, Rochester, USA) and to manage them as outpatients.3 4 7^–9 The Rochester scale includes a good general appearance in a previously healthy child, the absence of focal infection, and certain laboratory values (leucocyte count of 5000–15 000, ⩽1500 band neutrophils, urinalysis with ⩽10 leucocytes per high power field, and ⩽5000 leucocytes per high power field in faeces in patients with diarrhoea) as indicators of good prognosis; the negative predictive value of this scale was 98.9% for SBI and 99.5% for bacteraemia.9 Subsequent studies have suggested that markers such as C reactive protein (CRP) and, more recently, procalcitonin (PCT) could perhaps be more useful than the above indicators for predicting SBI.10^–14

CRP is an acute-phase reactant released by the liver in response to an inflammatory stimulus. Synthesis begins 4–6 h after the onset of inflammation and doubles every 8 h, peaking between 36 and 50 h later.12 15 PCT is a 116 amino acid peptide and is one of the precursors of calcitonin. It appears to be released by the liver and by mononuclear cells on stimulation by lipopolysaccharides and cytokines released during inflammation.14 16 PCT can be detected in plasma 2–3 h after an injection of endotoxin and peaks around 12 h after the onset of infection.17 18

The objective of this study was to determine the ability of CRP and PCT to predict SBI in febrile infants under 3 months of age and to compare them with the Rochester criteria.

MATERIAL AND METHODS

This was a retrospective study that included all consecutive infants between 4 and 90 days of age seen in the emergency department for fever (rectal temperature >38°C), in whom a detailed history and physical examination did not reveal a focus of infection, and in whom a blood test was performed. The study was performed in the paediatric emergency department of Donostia Hospital (San Sebastian, Spain) between January 2004 and December 2006. The exclusion criteria were lack of blood test, fever of more than 7 days’ duration, antibiotic therapy in the 48 h prior to diagnosis, and the presence of any type of immunodeficiency.

The SBI group included infants with: (1) microbiologically confirmed bacteraemia; (2) bacterial meningitis diagnosed by positive cerebrospinal fluid (CSF) culture; (3) sepsis, established according to the criteria defined by Levy et al19 including documented or suspected infection and findings of inflammation such as haemodynamic instability, tissue perfusion alteration and indications of organ dysfunction; (4) urinary tract infection confirmed by a positive urine culture (any number of colonies by suprapubic aspiration or >10 000 colony forming units in urine collected via a catheter); (5) pneumonia indicated by an infiltrate on chest x ray; (6) bacterial gastroenteritis confirmed by a positive stool culture; or (7) cellulitis with a suggestive physical examination. Infants with negative cultures or with improvement despite no antibiotic treatment were included in the non-SBI group. The subgroup of more invasive bacterial infection included cases of bacteraemia, sepsis, and bacterial meningitis.

Demographic, personal, clinical (degree and duration of fever), physical examination, and laboratory (leucocyte count, neutrophils, CRP, and semiquantitative PCT; PCT-Q) data were recorded. Two subgroups of infants were defined according to duration of fever greater or less than 12 h.

PCT was measured using a semiquantitative immunochromatographic assay (Brahms PCT-Q, Brahms, Hennigsdorf, Germany) with measurements of <0.5 ng/ml, 0.5 ng/ml, 0.5–2 ng/ml, 2 ng/ml, 2–10 ng/ml, 10 ng/ml and >10 ng/ml. CRP measurement was performed using an immunoturbidimetric assay (Tina quant, Roche Diagnostics, Mannheim, Germany). The leucocyte count was measured using an automatic cell counter (Sysmex XE 2100, Kobe, Japan).

The discriminatory power of the following variables was analysed: demographic data, clinical data, leucocyte count, neutrophil count, CRP, and PCT. The appropriate statistical tests were used for this purpose: student t test for comparing means of normally distributed parameters and χ² test when comparing the distribution of categorical variables between SBI and non-SBI. For the receiver operating characteristic (ROC) curve analysis, the three parameters (PCT, CRP, leukocyte count) were grouped into seven categories (PCT), six categories (CRP) and 11 categories (leucocyte count). We calculated the area under the curve (AUC) and the best cut-off points for each laboratory parameter and studied the values of sensitivity, specificity, positive predictive value, negative predictive value (NPV), and likelihood ratio. Finally, a multiple logistic regression model was used to determine the intrinsic predictive value of each factor studied.

The study was approved by the Clinical Research Ethics Committee of Gipuzkoa, Spain (CEIC Gipuzkoa).

RESULTS

During the study period, 375 infants less than 3 months of age were seen for fever of unknown origin. Twenty-eight (7.5%) patients were excluded (the attending doctor chose not to perform a blood test in 25 and three had received antibiotic therapy in 48 h prior to diagnosis); the data studied therefore came from 347 patients (151 girls, 196 boys) with a mean (SD) age of 47 (24) days.

The mean duration of fever at the time of diagnosis was 15 h; 258 (74%) patients had had fever for 12 h or less. Blood cultures were obtained in 330 (95%) patients, urine cultures in 333 (96%), CSF cultures in 170 (49%), leucocyte and neutrophil counts in 342 (99%), CRP in 339 (98%), and PCT in 320 (92%).

SBI was diagnosed in 82 (23.6%) patients: 69 urinary tract infections (4 with bacteraemia), 5 occult bacteraemias, 2 cases of cellulitis (1 with bacteraemia), 4 cases of sepsis (2 with bacteraemia), 1 acute bacterial gastroenteritis (with bacteraemia), and 1 pneumonia. Escherichia coli was responsible for 86% of the urinary tract infections and the most common organisms found in the occult bacteraemias were Streptococcus agalactiae B, Streptococcus pneumoniae, and Gram-negative bacilli. A positive reaction of the urine dipstick was found in 94% of the patients with urinary tract infection. The non-SBI group comprised 265 patients: 74 with confirmed viral infections (viral meningitis, viral gastroenteritis, or respiratory tract infections) and 191 with infections of probable viral aetiology (negative cultures and spontaneous resolution of the condition).

The demographic, clinical, and laboratory data were compared in the two groups using univariate analysis; values were only significantly higher in the SBI group in the case of plasma leucocyte and neutrophil counts, CRP, and PCT (table 1).

The AUC for PCT was 0.77 (95% CI 0.72 to 0.81), similar to the value obtained for CRP (0.79, 95% CI 0.75 to 0.84); both were significantly higher than the total leucocyte count (0.67, 95% CI 0.63 to 0.73) (fig 1). The sensitivity, specificity, predictive values, and likelihood ratios were determined for the cut-off points most often recommended in the literature. The simultaneous application of the three tests led to an increase in the sensitivity and NPV with respect to previous data, but at the expense of a lower specificity (table 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Receiver operating characteristic curves for procalcitonin (PCT), C reactive protein (CRP), and leucocyte count (WCC).

The ability to rule out SBI was considerably enhanced when CRP less than 30 mg/l, PCT less than 0.5 ng/ml, and leucocyte count between 5000 and 15 000 occurred simultaneously in a previously healthy infant with a good general state and a negative urine dipstick result: NPV of 96% for SBI and 100% for sepsis and bacteraemia (table 2).

In the infants with more invasive bacterial diseases (sepsis, bacteraemia), the diagnostic value of PCT using a cut-off point of 0.5 (AUC 0.84, 95% CI 0.79 to 0.88) was higher than that of CRP using a cut-off point of 30 (AUC 0.68, 95% CI 0.63 to 0.73) (table 2).

Multivariate logistic regression was performed with the variables that were significant in the univariate analysis and it was found that the leucocyte count, CRP, and PCT had intrinsic predictive value for SBI (table 3).

In infants with fever of 12 h or less (n = 258), the three laboratory parameters were also significantly different in the two groups, with a trend towards greater predictive value for PCT and lower predictive value for CRP compared with the overall group (table 4).

DISCUSSION

Our study shows that, unlike other criteria included in the various scales available, the leucocyte count, and with greater diagnostic value, CRP and PCT-Q have intrinsic and independent predictive value for the diagnosis of SBI in infants between 4 and 90 days of age with fever of unknown origin.

The management of febrile infants under 3 months of age is based on clinical scales that aim to define a group at low risk of presenting SBI and thus to reduce unnecessary hospital admissions and use of antibiotics.3^–5 7^–9 Several studies have investigated the predictive factors of SBI in this age range but few have included CRP and, in particular, PCT for this diagnosis after the neonatal period. Gajdos et al20 studied the predictive factors of SBI in a group of febrile infants under 3 months of age and found that, of the factors studied, only an elevated leucocyte count with >50% neutrophils and a CRP value >20 mg/l were able to predict SBI, with an NPV of 93%. Maniaci et al,21 in a recent study focused on PCT in febrile infants aged <90 days old, demonstrated that PCT may be a useful marker for SBI in young febrile infants. The authors reported a ROC AUC of 0.76 for definite and possible SBI, that is, better than that of leucocyte or neutrophil count. The optimal cut-off value used was 0.12, lower than in other studies with a sensitivity of 95.2% and NPV of 96.1% but at the expense of a lower specificity. In our study, the combination of leucocyte count, CRP, and PCT predicted the absence of SBI with an NPV of 93%. The ability to exclude SBI increased when a CRP value <30 mg/l, PCT <0.5 ng/ml, and leucocyte count between 5000 and 15 000 were found simultaneously in a previously healthy infant with a good general state and a negative urine dipstick, giving an NPV of 96% for SBI and 100% for sepsis and bacteraemia.

In recent years, both CRP and subsequently PCT have been studied by a number of authors as an aid to diagnosis of bacterial infection.11 13 20^–29 Few of these studies have been performed in emergency departments, looking at markers of SBI in infants with fever of unknown origin,11 20^–27 and fewer still have been conducted in the subgroup of infants under 3 months of age.20^–22

Pulliam et al11 demonstrated that CRP was better able to predict SBI in a group of febrile children between 1 and 36 months of age than leucocyte count and neutrophil count. In contrast, Isaacman23 found that the three laboratory parameters were of similar value in detecting febrile children between 3 and 36 months with occult bacterial infection, possibly due to the low incidence of such infection in their sample.

With respect to PCT, some studies with a similar design to ours concluded that PCT and CRP were of similar predictive value in the diagnosis of SBI in infants with fever of unknown origin and that both were of greater value than the leucocyte count.24 25 27 Andreola et al,27 in a recent study performed on a group of patients under 36 months of age in the emergency department of a tertiary hospital, demonstrated that PCT and CRP were the only factors predictive of SBI, with a similar AUC for each parameter. On studying only infants aged less than 3 months, they did not find differences in the AUCs for PCT and CRP with respect to the overall group of patients. In contrast, other studies have reported a greater diagnostic yield of PCT compared with CRP in the detection of invasive bacterial infections and in differentiating these infections from localised or viral infections.13 28 29 Likewise, in some studies, PCT has been shown to be better at distinguishing between bacterial and viral infections,13 a finding that could be due to the higher prevalence of more invasive bacterial diseases (bacterial meningitis, sepsis, bacteraemia) in these studies, as PCT has been shown to have greater diagnostic value than CRP in patients with more serious bacterial diseases such as sepsis and meningitis.13 14 18 27^–33 In addition, it has been observed that PCT appears to be more useful as a marker of the severity of infection.30^–33

In our study, PCT and CRP had similar predictive values for SBI. On considering infants with more invasive bacterial infections (sepsis, bacteraemia, meningitis), a trend to greater diagnostic value of PCT was observed, demonstrating the greater value of this marker compared with CRP and leucocyte count for the diagnosis of more severe infections, as reported in other studies.

The incidence of SBI in our study was 23.6%, somewhat higher than the incidence reported in the general literature, but similar to the figure obtained in studies of similar characteristics to ours.20 22 24 25 27 As found in other studies, the most common SBI was urinary tract infection (84%) and the prevalence of occult bacteraemia was 1.4%.5 10 20^–22 34 Taking into account that the most common bacterial infection in this age group is urinary tract infection, and that the majority are detected by a positive urine dipstick reaction, it would appear important to find a marker with higher diagnostic value for more severe bacterial infections, particularly in infants with fever of unknown origin in whom the urine dipstick is negative.

In infants with fever of 12 h of duration or less, we observed a trend towards improvement in the predictive value of PCT compared with CRP although this trend was smaller than has been reported in other studies due, in part, to the higher percentage of infants in our study with fever of short duration (74%).27^–29

Although this was a retrospective study, we included all infants seen in the emergency department in the study period. Only 7.5% of the infants were excluded, due mainly to the fact that the good general state of these infants led the doctor to consider blood tests unnecessary. The data from the history and physical examination were obtained from the clinical records, under conditions of normal clinical practice. The corresponding cultures were not performed in a small percentage of the infants included in the study, particularly in infants between 2 and 3 months of age with a good general state. Although a specifically designed prospective study would place greater emphasis on an accurate history and physical examination, the results would be less applicable to the general population. The measurement of PCT was performed using a semiquantitative method. We do not believe that this significantly affected the results as good correlation with the quantitative method has been demonstrated in previous studies,25 26 29 but a second generation assay for the measurement of PCT has shown recently to be more sensitive than the previous ones.35

CONCLUSION

PCT-Q, CRP, and the leucocyte count are predictive of SBI in infants under 3 months of age with fever of unknown origin. The diagnostic value of PCT is greater in the case of more invasive bacterial infections.