Clinical evaluation of point-of-care testing for wide-range C-reactive protein (wr-CRP) in neonates with suspected sepsis

Introduction

Early-onset neonatal bacterial infection (infection with an onset within 72 h of birth) is a significant cause of mortality and morbidity in newborn babies. The National Institute for Health and Care Excellence (NICE) recommends that babies with suspected early-onset neonatal infection should be treated as quickly as possible.

C-reactive protein (CRP) is an acute phase reactant. Following the onset of an acute phase response, serum levels rise rapidly and extensively, peaking at 24–48 h. CRP is a valuable marker of infection and can support or refute the early diagnosis of neonatal sepsis.

Standard laboratory CRP testing requires specialised equipment and experienced personnel, and the results are available within 1–2 h at best.

Point-of-care testing (POCT) refers to any diagnostic test administered outside the central laboratory at or near the location of the patient. The aim of POCT is to cut down on any transport/processing delays and result in rapid feedback of the test results to medical decision-makers.

There are now CRP point-of-care tests available. These have many advantages as they can provide results within minutes and can be used by clinicians or nursing staff. They also require smaller aliquots of blood. Blood conservation is an extremely important consideration in neonates to reduce the incidence of iatrogenic anaemia.

The aim of this study was to evaluate the wide-range (wr)-CRP point-of-care test, using a QuikRead instrument, in a neonatal population and compare the results to our standard laboratory Roche/Hitachi Cobas c 701 analyser.

Materials and methods

Results

Evaluation of relationship between standard laboratory testing and QuikRead CRP

Simple (univariate) regression

A scatterplot graph with an appropriately fitted regression line is shown in Figure 1. The adjusted regression coefficient of determination (R²) was 0.98. This means that 98% of variation in the QuikRead CRP POCT is explained by the laboratory CRP test. The slope coefficient was 0.94 with a 95% confidence interval (CI) of 0.92–0.96. Therefore, when the laboratory CRP increased by 1.0 mg/L, the QuikRead CRP POCT closely correlated by an increase of 0.94 mg/L.

Figure 1:

Linear regression analysis between the laboratory CRP test and the QuikRead CRP POCT.

Bland-Altman plot

The absolute differences between the two CRP tests against their means is illustrated in Figure 2 in a Bland-Altman plot. This shows there is an increase in variability between the two tests as the QuikRead CRP levels increase. Clinically, a difference between the two tests becomes less relevant as the CRP levels increase. However, a difference between the two tests is relevant when the CRP is low, especially as we use cut-off values to support our diagnosis of sepsis.

Figure 2:

The Bland-Altman plot shows the differences plotted against the averages of the two sets of CRP values, obtained from the QuikRead CRP POCT and laboratory analyser.

ROC curve

In our study, we used a cut-off value of CRP >5 mg/L to define sepsis. This was applied to both the laboratory and QuikRead CRP POCT results. Shown in Figure 3 is an ROC curve which we used to show at what threshold the QuikRead CRP POCT is predictable of a CRP >5 mg/L in the laboratory. We found that if the QuikRead CRP POCT is ≥6.2 mg/L, the specificity is 100% with a sensitivity of 77%. This will still miss some cases, so we found that the optimal cut-off value of the QuikRead CRP POCT with a 100% sensitivity and 94% specificity is ≥3 mg/L. The area under the ROC curve is 0.99 with a 95% CI of 0.98–1.00 and a p-value of 0.0047, which implies that the QuikRead CRP POCT is a good diagnostic test.

Figure 3:

ROC curve for the QuikRead CRP POCT.

Sub-group analysis

We attempted to correct for clustering in the analysis and tested to see if there was an interaction between both of the CRP tests (laboratory and QuikRead), polycythaemia and gestational age.

Polycythaemia (haematocrit >50%)

CRP whole blood POCTs need to be calibrated so that they correlate with the corresponding serum/plasma level. The mean normal range of haematocrit in adults is 40%, whereas haematocrit values can range from 29 to 68% in paediatric samples [1]. Any deviations from a haematocrit of 40% need to be corrected using a haematocrit correction factor. In this study, the point-of-care QuikRead wr-CRP POCT automatically calibrates for haematocrit using its wide range of haematocrit corrections (15–75%). However, we tested for an association between a haematocrit of >50% and CRP values. We found no evidence that the QuikRead CRP and laboratory CRP are modified by a haematocrit of >50% (p-value 0.24).

Gestational age

We also considered whether gestational age is associated with a difference in CRP values. Research has shown that different CRP reference intervals are needed for healthy term and preterm neonates due to the strong positive effect of gestational age on the elevation [2]. We found no evidence that the QuikRead CRP and laboratory CRP are modified by gestational age (p-value 0.26).

Discussion

Early-onset neonatal bacterial infection (infection with an onset within 72 h of birth) is a significant cause of mortality and morbidity in newborn babies. If the diagnosis of neonatal sepsis is delayed, its clinical course can be rapidly fatal. CRP is a valuable marker of neonatal sepsis. Standard laboratory CRP results can take 1–2 h at best to become available and usually a minimum of 0.3 mL is required for sampling. Bedside CRP testing provides results rapidly and requires less blood sampling, both relevant factors in a neonatal population.

This was a diagnostic test evaluation study in which the point-of-care QuikRead wr-CRP test was compared to our standard laboratory testing in babies with suspected infection. This is the first report of its kind, where this specific POCT has been trialled in this setting.

Our study found that the point-of-care QuikRead CRP results were in excellent agreement with the laboratory CRP values with Pearson’s correlation co-efficient of 0.94 (0.92–0.96). An increase in variability was discovered between the two tests as the QuikRead CRP levels increase; however, at low levels (which are arguably more important), there is less variation. An optimal cut-off of ≥3 mg/L with the point-of-care QuikRead wr-CRP POCT is recommended to reflect a laboratory CRP value of ≥5 mg/L. No correlation was found between haematocrit and the gestational age and the QuikRead or laboratory CRP values.

The QuikRead CRP POCT has been validated in the paediatric population. However, most of the studies are based in a primary care or paediatric emergency department setting. In the study by Esposito et al. [3], the QuikRead CRP POCT was compared with standard laboratory testing in children <14 years of age who arrived at their paediatric emergency department with respiratory infection. Studies by Nafar et al. [4], [5] also used the QuikRead CRP POCT in children with suspected bacterial pneumonia and gastroenteritis in the paediatric emergency department. In these studies, the QuikRead CRP POCT was found to be a useful predictor of infection and correlated well with standard laboratory testing.

There are very few studies in the literature that have used point-of-care CRP testing in a predominately neonatal setting. Much of the evidence available is also based on a predecessor version of the technology. Advances have been made that make this testing kit more applicable to the neonatal population. In a study by Diar et al. [6], a predecessor version of the QuikRead CRP POCT was used to analyse a neonatal population in Soweto, South Africa. Their conclusion was that it would be a useful screening test in cases where a clinical diagnosis of sepsis was in doubt. They found that a CRP cut-off level of 16.2 mg/L for the QuikRead CRP POCT minimised the possibility of a false-negative result. The QuikRead CRP POCT used in this study had a CRP detection limit of 8.0–160 mg/L. The median CRP concentrations were 6 mg/L and 9 mg/Lin the first 24 h after birth in septic preterm and term neonates, respectively [7]. One of the strengths of our study is that we used a QuikRead version with a wide measuring range (0.5–180 mg/L for whole blood). This is important in neonates where highly sensitive CRP assays that can reliably measure low CRP values are needed. It also had a wide range of automatic haematocrit correction (15–75%) leading to more reliable results.

There are other studies that have used alternative point-of-care CRP testing in a neonatal setting. Aydin et al. [8] used NycoCard CRP POCT and compared them with laboratory CRP tests in 63 babies. They concluded that the NycoCard CRP POCT is highly predictive. In a study by Zecca et al. [9] where both the Nycocard and QuikRead CRP POCTs were compared to standard laboratory testing, both bedside tests had good specificity (QuikRead 80.5%, NycoCard 83.3%) and sensitivity (QuikRead 97.2%, NycoCard 94.4%). The agreement of measurement with central laboratory values was high for both the bedside tests, without statistically significant differences between the methods. In similar findings to our study, the accuracy of the results of both bedside tests was found to be decreased when CRP concentrations were >100 mg/L.

Laboratory testing leading to phlebotomy losses during the first few weeks of life is one of the main causes for anaemia in extremely low birth weight (ELBW) infants. Point-of-care tests use minimal sample volumes and thus have potential for reducing the incidence of phlebotomy-associated anaemia in neonates. In our study, 0.1 mL of whole blood was required for the POCT compared to 0.3 mL whole blood needed as a minimum for standard laboratory testing. Madan et al. [10] introduced a point-of-care blood gas analyser in their NICU and found that the mean volume of red blood cell transfusions in ELBW infants decreased by 43% in the first 2 weeks of life.

Our study has some limitations. It is a small study of only 91 neonates and 126 paired samples. We have included infants of variable gestation and weight who had varying methods of venepuncture (mostly capillary) to determine the CRP values. We have attempted to correct for this clustering of data and found no interaction with gestational age or haematocrit values. However, there is limited power to look at the subgroup analysis, and some questions remain unanswered. We also defined a cut-off CRP level based on the data collected but have not yet tested this in practice. Therefore, a follow-up study is required to determine if this cut-off value is reliable.

The NICE recognises the benefits of point-of-care CRP testing and has already published guidance on its use. When managing pneumonia in adults based within a primary care setting, the NICE recommend that point-of-care CRP testing should be considered if a diagnosis is unclear after clinical assessment, and that antibiotics should be prescribed based on the result [11]. In September 2016, the NICE published a MedTech innovation briefing on the QuikRead CRP POCT for CRP testing in primary care [12]. Evidence from several studies (n=4874) showed that the QuikRead CRP POCT performs with similar accuracy to the standard laboratory CRP test in detecting pneumonia and results in reduced antibiotic prescribing rates compared with standard care. The NICE believes that “using this technology could contribute to fulfilling antibiotic stewardship programmes”.

The innovative aspects of using CRP POCT in the NICU are that it is quick, reliable and only requires 0.1 mL of blood. In addition to reducing the risk of iatrogenic anaemia secondary to blood sampling, serial point-of-care CRP measurements are likely to assist with earlier discontinuation of antibiotic therapy in support of antibiotic stewardship.

Conclusions

The point-of-care QuikRead wr-CRP test provides reliable results in a neonatal population when compared to standard laboratory testing and might be used as a bedside tool to aid decisions in the management of neonatal infection.