Results

Participant characteristics

A total of 334 women were potentially eligible, of whom 19 (5.7%) were excluded as outliers; details on outliers are presented in Appendix S1. Of the 315 women included in the development cohort, most were white (93%), had a normal BMI or were moderately overweight, and delivered at full term (mean gestation 40.4 weeks). Most babies did not require admission to the neonatal unit (98.1%). On average, overweight women (≥25 kg/m2) had a significantly higher mean arterial blood pressure and gave birth to heavier babies than women with a normal BMI (<25 kg/m2). Fewer than 2% of data were missing according to any characteristic. Data on participant characteristics are summarised in Table 1.

Reference interval development

Trimester-specific reference intervals for CRP are presented in Table 2.
CRP was not associated with maternal age, mean arterial pressure or ethnicity, but BMI was significantly log-linearly associated with CRP in all trimesters (see Figure 1 and Appendix S2). CRP was marginally higher in women pregnant with male babies than female babies, but this finding was only significant in the second trimester. The statistical conditions for partitioning the reference intervals for CRP by BMI were either met or were marginal in all trimesters, using a cut-off of 25 kg/m2, so an overall reference interval was used. BMI-specific reference intervals are presented in Table 3 and Figure 2.

Diagnostic accuracy

The characteristics of women in the development and evaluation cohorts were similar (see Table 1). Of the 50 women in the evaluation cohort, 15 (30%) had chorioamnionitis and 35 (70%) did not have any infection. The median CRP values among women with and without infection were 30.0 mg/L and 8.1 mg/L, respectively (Mann-Whitney test pdiff<0.001).
The overall diagnostic accuracy of CRP for infection was significantly higher when using a pregnancy-specific upper reference limit of 18 mg/L, compared to 7 mg/L (p=0.002) (see Figure 3). The sensitivity and specificity using the pregnancy-specific reference limit were 80% and 86%, respectively. Summary statistics on diagnostic accuracy for both reference limits are presented in Table 4.