Results
The P5 trial screened 8168 women, of whom 7857 were singleton and 1081 had CL ≤30mm. In a CL distribution curve including only singleton pregnancies, 1081 women corresponds to 13.7% of total. For this study, we excluded 310 twins, 14 women without CL data and 3 women in progesterone group without gestational age at birth. We included 430 singleton women with CL ≤30 mm randomized to progesterone alone and we projected this group to correspond to 13.7% of CL ≤30mm for our analysis. To complete our final sample and reach the complementary 86.3% of CL >30mm, we randomly selected 2709 singleton women with CL >30 mm, comprising a total of 3139 women (Figure S1).
Among women with CL ≤30mm receiving progesterone, compliance was 82%. Regarding obstetric history, 46.2% (1449) of our sample were nulliparous, 10.1% (318) had at least one previous PTB and 24.4% had a previous abortion. The prevalence of PTB at <37 weeks was 14.43%: sPTB at <37 weeks was found in 7.1% (223/3139); and sPTB at <37 weeks in women with CL ≤30mm receiving progesterone was 16.7% (72/430). Of all 223 women who had a sPTB, 32.3% (72/223) had a CL ≤30mm. Sociodemographic information is listed in Table 1.
Logistic univariate regression analysis for PTB at <37 weeks identified the following risk factors: low body mass index (BMI ≤ 18.5) (OR = 1.95, 95%CI = 1.05–3.43,); hypertension (OR 2.15, 1.5–3.02); endocrinopathies (OR = 1.73, 1.27–2.33); previous PTB (OR = 2.51, 1.88–3.32); previous abortion (OR = 1.43, 1.15–1.78); cervical length ≤30mm (CL 25-≤ 30mm OR 2.10, 1.47 - 2.95; CL 20-25mm OR 2.55, 1.71 - 3.72; CL 15-20mm OR 3.33, 1.74 - 6.11; CL 10-15 mm OR = 6.40, 2.53–5.99, and CL ≤10mm OR 11.17, 4.37–30.55); funneling at measurement (OR = 5.03, 3.36–7.49); and sludge at measurement (OR = 3.50, 2.24–5.39). Considering only sPTB at <37 weeks, these factors presented an even higher association except for comorbidities and low BMI. A comparison between sPTB at < 34 weeks and ≥ 34 weeks illustrates that there is a robust association among risk factors and sPTB<34 weeks, highlighting CL≤10mm (OR 44.9, 15.45–125.87) and 10–15mm (OR13.32, 2.98–43.09), funneling at measurement (OR 10.22, 5.57–17.95) and sludge at measurement (OR = 5.61, 2.63–10.86) (Table 2).
A multivariate logistic regression analysis also identified an association between CL ≤30mm and PTB (CL 25-≤30mm ORa 1.80, 1.23-2.63; CL 20-25mm ORa 1.93, 1.22-3.06; CL 10-20mm ORa 3.04, 1.54-5.71, and CL ≤10mm ORa 3.82, 1.12-13.06). The ORa for cervical length <30mm increased when considered only sPTB <37 (CL 25-≤30mm ORa 2.2, 1.35-3.57; CL 20-25mm ORa 2.07, 1.14-3.76; CL 10-20mm ORa 4.59, 2.12-9.94, and CL≤10mm ORa 6.71, 1.79-25.27). For sPTB<34, there was an association with CL ≤25 mm (Table S2). We also performed a multivariate analysis for cervical length and PTB <37, sPTB <37 and sPTB<34 weeks with adjusted odds ratios for BMI, comorbidities, obstetrical history, funneling and sludge and the association between CL<30mm and PTB and sPTB<37 was also significant. Again, moderate sPTB (sPTB<34) where associated with CL ≤25mm (Table S3).
We identified an inverse association between CL and sPTB at <37 weeks (OR = 7.84, 5.5–11.1). The ROC curve analysis to predict PTB at < 37 weeks and sPTB at <37 weeks showed low performance, with area under the curve (AUC) of 0.598 (0.57–0.63) and 0.643 (0.60–0.68), respectively. For sPTB at <34 weeks and sPTB at <32 weeks the ROC curve presented a moderate performance with AUC of 0.665 (0.59–0.74) and 0.718 (0.62–0.81), respectively; and for sPTB at <28 weeks the ROC curve demonstrated good performance, with AUC of 0.820 (0.63–0.95) (Table S4 and Figure 1).
The best cutoff point to predict PTB at <37 weeks was 31.75 mm, with 31.3% sensitivity and 84.4% specificity. To predict sPTB at <37 weeks the best cutoff point was 31.75mm, with 37.2% sensitivity and 84.3% specificity. TVU provided good prognostic results combining: AUC (0.82), high sensitivity (73.7%) and acceptable specificity (91.3%) rates for sPTB at <28 weeks’ gestation (Table S4). The best cutoff points to predict sPTB at <34, <32 and <28 weeks were 28.05, 28.05 and 26.55 mm, respectively.
Kaplan-Meyer survival analysis demonstrated an association between extremely severe, severe, moderate and late PTB and CL ≤25 mm, and an association between CL of 25–30mm and late PTB (p<0.001) (Figure 2). The number needed to screen (NNS) to detect one true positive sPTB <34 weeks in women with CL ≤25mm is 121. To prevent one sPTB <34 weeks among women with CL ≤25mm, the number needed to treat (NNT) with vaginal progesterone prophylaxis is 1812. Assuming that all women with CL ≤25mm are treated with vaginal progesterone, we estimated that the number of TVU necessary to identify 18 women with CL ≤25mm and prevent one sPTB <34 weeks is 248.