Discussion

CRP in the healthy pregnant women in our study was substantially higher than in most non-pregnant adults. Using a pregnancy-specific RI improved the overall diagnostic accuracy for infection compared to the existing non-pregnant RI, conferring a particular improvement in excluding infection in healthy women without reducing the sensitivity materially. The RIs for CRP were similar in all trimesters with only a small reduction as pregnancy progressed.
The RIs reported in guidelines and large studies of CRP vary substantially, e.g. ≤7 mg/L in the UK,10 and ≤10 mg/L in the US and Australia.18, 19 The unmet need for an established pregnancy-specific RI creates diagnostic uncertainty, which risks clinicians disregarding elevated CRP results, or making unnecessary interventions (e.g. iatrogenic delivery). Our study supports using a relatively high upper reference limit for CRP in pregnancy, as this predicted infection with a much greater degree of accuracy than the existing guidance.
The value of using CRP for diagnosing chorioamnionitis through intact membranes is unclear, but there is a growing body of evidence supporting its use in pregnant women with preterm prolonged rupture of membranes. A 2018 meta-analysis reported pooled sensitivities and specificities of 69% and 77%, respectively, although the upper reference limits for CRP varied widely (3-18.7 mg/L).20 The results of our study are similar to those of a more recent review which reported sensitivities and specificities of 59% and 83%, respectively, using an upper reference limit of 20 mg/L.21 Studies of other blood biomarkers have reported better diagnostic accuracy when using pregnancy-specific reference intervals: Girling, et al. reported that aspartate transaminase, alanine transaminase, bilirubin and gamma glutamyl transferase were lower in pregnancy, and that using these pregnancy-specific RIs significantly improved disease detect rates.22
Adipose tissue is known to mount an immune response to overnutrition,23 which is likely to drive the elevated levels of CRP seen in the overweight women in this study, as CRP is not able to discriminate between infection and other causes of inflammation. The relationship between obesity and CRP is well documented in pregnant24, 25 and non-pregnant populations,26 but partitioned RIs for CRP are not routinely used in clinical practice. We propose that stratifying CRP results according to BMI would be challenging for routine use, but we have presented them for consideration. CRP in women pregnant with male babies was slightly higher than those with female babies, which was statistically significant at the 1% level in the second trimester only; we include this as it may be useful for consideration in future research.

Strengths and weaknesses

Ours is the largest study identified to have investigated CRP in pregnant women, and its associations with BMI and gestational age. Our study may include some women with undiagnosed gestational diabetes, as screening is not universal in the UK, but it is uncertain whether this is independently associated with CRP after adjustment for BMI.27 Most women in our study were white so we cannot comment on differences in CRP between ethnic groups.
Diagnostic accuracy studies in pregnancy are challenging without a gold standard for maternal infection. Our approach is consistent with previous studies on histological chorioamnionitis20and we have augmented this with microbiological and microscopic data. We acknowledge that the timing of some of the CRP samples (postnatal) may over-represent the antenatal CRP at the actual time of amniocentesis (such procedures are not usually performed in those with a very high CRP). However, within the limits of this pragmatic retrospective study, it should not preclude comparing the two reference limits evaluated here.

Conclusions

Concentrations of CRP are sufficiently elevated in pregnant women that pregnancy-specific RIs should be used to improve diagnostic accuracy. We have defined and evaluated a new RI for CRP in healthy pregnant women, with specific considerations for characteristics that may modify its limits.