Influence of cut-off value on prevalence of short cervical length

Abstract

Objective: To assess the distribution of cervical length (CL) in a large cohort of asymptomatic low-risk women with singleton pregnancy and no previous preterm birth and to explain the low prevalence of short CL ≤ 30 mm in this cohort. Methods: This was a secondary analysis of a multicenter cohort study with an embedded randomized controlled trial (Triple P trial; NTR-2078) on the prevention of preterm birth with progesterone. In the cohort study, CL was measured in asymptomatic low-risk women with singleton pregnancy to investigate its predictive capacity to identify those at increased risk for preterm birth. A short CL was defined by a cut-off value of ≤ 30 mm, based on existing literature. Women with a short CL were subsequently included in a randomized controlled trial evaluating the effect of progesterone, compared with placebo, on preterm birth. In total, 57 centers and 20 234 women participated in the study. Normal distributions for CL were simulated based on the mean and SD of the original data. The distribution of CL was assessed for each individual center and measurements were compared between levels of care: primary (29 ultrasound centers), secondary (21 general hospitals) and tertiary (seven university medical centers) care institutions. Comparison was also performed between centers with low, intermediate and high volume of CL measurements. CL distributions before (n = 12 284 women) and after (n = 7950 women) a national symposium, at which the prevalence of short CL measurements was addressed publicly, were analyzed. Results: Between November 2009 and August 2013, 20 234 women had CL measurements, of whom 367 (1.8%) had a short CL. Mean ± SD CL was 44.2 ± 7.8 mm. A ‘dip’ in the distribution of CL measurements between 20 and 30 mm was observed, defined by a ratio of <50% when comparing the number of measurements in observed and simulated normal distributions. The dip was present in 89% of participating centers. All centers showed a dip in the distribution of measurements ≤ 30 mm when analyzed according to the level of care and volume of measurements. A significant difference was found when comparing the distribution before and after publicly addressing the low prevalence of short CL (1.7% vs 2.0% of measurements were ≤ 30 mm, respectively; P <0.001). Conclusions: A cut-off value of 30 mm for CL was used to include women in a randomized clinical trial that was embedded in a cohort study. We suggest that the use of a predefined cut-off value for a short cervix influences the distribution of the CL measurements. Since the measurement is not blinded, preference of assessors for the control or intervention arms may have introduced selection bias. This might have resulted in fewer measurements around the cut-off value. Other trials using similar designs could benefit from this observation and take precautions to avoid selection bias.