Physiological background (what is CRP, and what does it do?)

CRP is a protein synthesised and secreted by the liver in response to inflammatory cytokines, particularly IL-6. It is a member of the pentraxin family of proteins being composed of five identical subunits arranged in a cyclic pentameter shape. These are non-covalently bonded and have a total molecular weight of 118 000 daltons.11 12

Although CRP is often used clinically as a marker for inflammation, its function is as a key component of the innate immune response to inflammation and infection. It plays this role by mediating phagocytosis, activating the complement cascade and by producing further inflammatory cytokines. CRP forms complexes by binding to a range of molecules, such as polysaccharides and peptopolysaccharides present on bacteria (for instance, Streptococcus pneumoniae, Haemophilus influenzae), parasites and fungi.5 CRP also binds to components of damaged host cells (both membrane and intracellular components).12 These CRP–ligand complexes bind to Fc receptors on phagocytic cells (neutrophils, macrophages, etc) mediating phagocytosis of pathogens opsonised by CRP and stimulating a pro-inflammatory response via the production of further cytokines (IL-1, tumour necrosis factor α). CRP–ligand complexes are also able to bind to C1q, the first component of the complement cascade, which again triggers phagocytic activity.12

Technological background (how is CRP measured?)

Measurement of CRP relies on the formation of complexes between CRP and anti-CRP antibodies. Quantitative measurement of CRP involves mixing serum with reagent containing anti-CRP monoclonal antibodies. These anti-CRP antibodies are either marked (using an enzyme or fluorescent tracer) or unmarked, depending on the technique used. The resulting complexes can then be measured accordingly.

In the enzyme-linked multiplied immunoassay, a substrate is added that reacts with the enzyme to produce a colour. This can then be measured by a spectrophotometer (handheld or automated). Immunofluorescence uses a fluorescent tracer, which is measured using a fluorescence microscope. Nephelometry uses test tubes that contain a fixed amount of anti-CRP antibody. To these are added human serum leading to formation of CRP–ligand complexes. A beam of infrared light is then passed through the tube. The amount of scatter of this light is measured using a nephelometer, turbidimeter or automated analyser. This is proportional to the amount of CRP–ligand complexes. Nephelometry is the most widely used technique, takes only 10 min using a fully automated analyser and has an analytical sensitivity of about 0.04 mg/l.11,–,13

In normal individuals without infection, serum CRP concentration is <10 mg/l. Generally, CRP is highly elevated by serious bacterial infections (SBIs) (any potentially life-threatening bacterial infection, such as meningitis or pneumonia), up to 150–350 mg/l, but less highly elevated by acute viral infection, to 20–40 mg/l. However, this is not absolute: some viral infections can cause elevations >100 mg/l (eg, adenovirus, mumps and measles).14 Furthermore, the elevation of CRP will also depend on what point during the course of infection the sample is taken. If taken very early, even in SBI, it may only be very slightly raised due to the 4–6 h time lag between inflammation and CRP synthesis.

Indications and limitations

In children with fever without focus, does CRP identify those with SBIs?

One of the difficulties in treating children who present with fever, but no symptoms or signs localising the infection, is to identify those with SBI. CRP has been proposed as a potential early discriminator.

One previous review was found that examined the evidence available.15 This review identified four exploratory prospective cohort studies and one retrospective case review that were relevant. The same four prospective cohort studies were included in a review published as part of National Institute for Health and Clinical Excellence (NICE) guidelines on Feverish Illness in Children,16 along with six other studies (table 1). These studies looked back at cases with knowledge of the outcome (laboratory or radiographically proven bacterial or non-bacterial infection) and then looked at single CRP measurements taken. Despite the existence of a number of different studies, interpretation of the results for clinical practice is difficult. A variety of CRP cutoff values were used in the studies (ranging from 10 to 70 mg/l), and the incidence of SBI was low, making direct comparisons difficult. Sensitivity of CRP for identifying SBI ranged from 63% to 95%, and specificity ranged from 40% to 91%.16 One study looked at whether the time from onset of fever at which CRP was measured affected its sensitivity or specificity. They found no significant difference in sensitivity or specificity between CRP values collected before or after 12 h from the onset of fever.17

Table 1 reports relative risks for each study. Here the relative risk is the ratio of the probability of SBI occurring in the group with CRP value above the chosen cutoff versus the group with CRP below the cutoff. Thus—for example, in the study by Isaacman,17 those with a CRP >44 mg/l are five times more likely to have a SBI than those with a CRP <44 mg/l.

A systematic review of studies comparing the diagnostic accuracy of raised CRP with microbiological confirmation of SBI published pooled estimates from six studies.18 These studies used a range of cutoff values for CRP (from 20 to 70 mg/l), although three of the six used the same cutoff value of 40 mg/l. The pooled estimate for sensitivity was 0.77 (95% confidence interval (CI) 0.68 to 0.83), specificity 0.79 (95% CI 0.74 to 0.83), positive likelihood ratio 3.64 (95% CI 2.99 to 4.43) and negative likelihood ratio 0.29 (95% CI 0.22 to 0.40).18

None of the reported results suggest that CRP is sensitive or specific enough to accurately identify all children with SBI who present with fever without focus. The NICE guidelines on Feverish Illness in Children16 recommend the use of CRP in conjunction with full blood count, blood cultures and urine testing (plus other investigations as indicated) in any child with fever <3 months old, and all children >3 months old who present with red or amber features. From the available research, these seem to be appropriate recommendations for the use of CRP in these patients. CRP is useful, but no specific cutoff level can be recommended. High CRP does suggest the presence of SBI but must be used together with other investigations to inform clinical decision-making on a case-by-case basis.

In a child with fever with a focus, does CRP help in distinguishing viral and bacterial aetiologies?

Pneumonia

It is notoriously difficult to distinguish bacterial from viral causes of pneumonia in children. Identifying the causative pathogen from cultures, viral antibody titres or PCR techniques is time-consuming, and treatment cannot be delayed for the results. In a child with pneumonia, does CRP help in distinguishing viral and bacterial aetiologies?

A meta-analysis of eight studies included a total of 1230 children with pneumonia.19 This population had an incidence of bacterial pneumonia of 41%. The studies included used CRP cutoffs ranging from 35 to 60 mg/l. An odds ratio (OR) for bacterial pneumonia given a CRP above the chosen cutoff value was calculated for each study. The outcomes varied significantly between different studies. Combining all eight studies gave an overall OR of 2.58 (95% CI 1.20 to 5.55). A CRP concentration >35–60 mg/l occurred significantly more often in children with bacterial pneumonia compared with nonbacterial pneumonia. The meta-analysis calculates that a child with clinical and radiographic evidence of pneumonia and a CRP exceeding 40–60 mg/l has a 64% probability of having bacterial pneumonia (the positive predictive value).19 Therefore, a serum CRP raised >40–60 mg/l is weakly predictive of bacterial pneumonia in children. However, although a lower CRP suggests a bacterial cause is unlikely, this association is weaker and is not in itself a reason to withhold antibiotics.

Meningitis

Serum CRP was able to distinguish bacterial meningitis with a negative Gram stain from viral meningitis on admission with a high sensitivity (96%), specificity (93%) and negative predictive value (99%).20

Gastroenteritis

The recent NICE guidelines on diarrhoea and vomiting in children under five found a lack of good-quality studies to evaluate the ability of laboratory tests to distinguish between bacterial and viral gastroenteritis. However, evidence from three studies suggested that raised CRP levels had a high diagnostic ability in detecting bacterial causes (the studies used different cutoff values for CRP, ranging from 12 to 95 mg/l), and CRP is suggested in immunocompromised children with diarrhoea.21

Septic arthritis of the hip

Distinguishing septic arthritis from transient synovitis of the hip in children can be challenging. A CRP level of >20 mg/l was a strong independent risk factor for septic arthritis of the hip.22 The OR for septic arthritis in patients with a CRP >20 mg/l versus those with a CRP <20 mg/l was 14.5 (95% CI 3.2 to 64.9).22

How should CRP be used in neonates with suspected sepsis?

Neonatal sepsis affects between 1 and 21 infants per 1000 live births worldwide. Untreated, it has mortality as high as 30–69%.13 Clinical diagnosis is difficult because of presentation with very non-specific signs and symptoms. Because of the high mortality, any test used to aid diagnosis at presentation must identify all infants with sepsis (high sensitivity), even at the expense of overdiagnosis (low specificity). Can CRP fulfil this role, and if not, how should it best be used in neonatal sepsis?

Healthy full- and preterm infants have a CRP level of 2–5 mg/l. The upper limit of normal has been established as 10 mg/l.13 Any elevation in CRP represents endogenous synthesis (as CRP crosses the placenta only in very low amounts).14 CRP in infants must also be interpreted carefully given that there can be a poor CRP response on the first day of postnatal life in full-term infants, and for longer in preterm infants.14 There is a reported 3-day rise in CRP due to the stress of delivery.14

“Ruling in” sepsis

In neonates with suspected sepsis, does CRP identify all neonates with sepsis?

Studies have investigated the role of CRP values taken at presentation of suspected neonatal sepsis. The reported sensitivity of a raised CRP for diagnosing neonatal sepsis ranges from 35% to 94% and specificity from 60% to 96%.13 This variation reflects differences in study design. At presentation, CRP may remain normal despite the presence of sepsis because of the time lag in CRP rising. Furthermore, a raised CRP is not diagnostic for sepsis because it may be raised for other reasons (prolonged rupture of membranes, maternal fever during labour, fetal distress, perinatal asphyxia or shock, periventricular and intraventricular haemorrhages, pneumothoraces, and meconium aspiration pneumonitis).13 In conjunction with a suggestive clinical picture, and other blood tests (particularly white cell count and differential), CRP may aid in diagnosis of neonatal sepsis. However, it is not sensitive enough to be relied upon to identify all cases of neonatal sepsis, and therefore, antibiotics should not be withheld based on a low CRP in the presence of a clinical picture suggestive of sepsis.

“Ruling out” sepsis

In neonates with suspected sepsis, who have started antibiotics, do serial CRP measurements identify all neonates who do not have sepsis?

A number of studies have evaluated the role of serial CRP measurements in cases of suspected neonatal sepsis. Serial CRP values taken 24–48 h after the onset of symptoms have an improved sensitivity and specificity when compared with single CRP values at presentation. Reported sensitivities range from 78.9% to 98% and specificity from 84% to 97%. Two consecutive CRP levels <10 mg/l 24 h apart, 8–48 h after presentation, have a negative predictive value for sepsis of 99%.13 23 Thus, two serial CRP levels <10 mg/l, at least 24 h apart, can accurately identify infants who are unlikely to be infected or have a resolved infection (providing there is no other reason for CRP to remain low, such as in preterm infants). At 48 h after onset of symptoms, with at least two normal CRP values, and negative blood cultures, antibiotics can be stopped.

Serial CRP levels that fail to decrease, or that continue to rise after 48 h of antibiotic therapy, suggest treatment failure (such as fungal infection) and indicate the need for further clinical review and microbiological investigation.14 25

Topics for further research

Procalcitonin may outperform CRP in the identification of SBI in children with fever without focus.16 Further studies are required to establish this. The NICE guidelines on Feverish Illness in Children16 published a health economic evaluation for the use of CRP versus procalcitonin and concludes that CRP is both less costly and more effective than procalcitonin, but only when making certain assumptions (prevalence of SBI <27% in those children tested). They feel that further investigation is needed into the health economics of procalcitonin versus CRP.16

Procalcitonin may be more effective than CRP in early diagnosis of neonatal sepsis and in monitoring response to treatment.25 Further investigation is required.

Evidence in adults suggests that serial CRP measurements are a useful marker for treatment failure or for the development of infectious complications.11 Do serial CRP measurements have the same value in children with pneumonia?

Clinical bottom line

CRP levels begin to increase 4–6 h after an inflammatory trigger and peak at 36–50 h. Levels decrease rapidly with the resolution of inflammation. CRP levels that fail to decrease, or continue to rise, after 48 h of antibiotic therapy suggest treatment failure.

In children presenting with fever without focus, a single CRP level is not sensitive or specific enough to identify all children with SBI. However, a raised CRP does suggest SBI, and it should continue to be used in combination with other investigations as outlined in the NICE guidelines on Feverish Illness in Children.16

In children with pneumonia, a single CRP value >40–60 mg/l is weakly predictive of bacterial pneumonia. A CRP value below this cutoff should not in itself be a reason to withhold antibiotics.

In a child with a limp, a CRP level of >20 mg/l suggests septic arthritis of the hip rather than transient synovitis.

In infants with suspected neonatal sepsis, a single CRP level at presentation is not useful alone in either diagnosing (if raised) or excluding sepsis (if normal). If raised, it may aid a diagnosis of neonatal sepsis in conjunction with clinical findings and other investigations.

In infants being treated for suspected neonatal sepsis, serial CRP levels are useful in excluding that diagnosis if two CRP measurements 24 h apart are <10 mg/l. Serial CRP levels that fail to decrease or that increase after 48 h suggest treatment failure.