- Top of page
Preterm birth (PTB), especially early PTB (<34 weeks of gestation), is a major cause of neonatal mortality and contributes to 75% of neonatal morbidity and 70% of neonatal mortality in industrialised countries. Despite advances in diagnostics, the exact etiology of PTB remains elusive. PTB can result from spontaneous preterm labour (40–45%), can follow preterm prelabour rupture of membranes (25–30%) or, finally, can be iatrogenic for medical/obstetric reasons (30–35%). There is need of an ideal predictor of PTB so as to allow for specific treatment and in utero transfer of at-risk pregnancies to an appropriate tertiary referral centre well in advance, thereby sparing unwarranted interventions, psychological stress and unnecessary tocolysis in those not identified to be at risk.
To have a significant impact on management, clinical tests should predict high-risk women for PTB early in pregnancy. Various biochemical and biophysical markers have been studied for the prediction of PTB. Only a documented past history of PTB and short cervical length (<15 mm) in the current pregnancy have been established as predictors of PTB. Whereas in animal studies a drop in progesterone, either in absolute terms or relative to the estrogen concentration, has been integral to the initiation of parturition at term, in humans functional progesterone withdrawal has been implicated in premature delivery.[4, 5] A study by Lachelin et al. suggested that a low concentration of free, unbound progesterone, as measured in the saliva, may be predictive of spontaneous preterm labour in women at risk of PTB. There have been conflicting reports of whether or not progesterone administration can delay the onset of PTB, although progesterone deficiency may stimulate or augment the proinflammatory mediators implicated in PTB.[7-11] Nonetheless, an increasing body of evidence has suggested that progesterone prophylaxis may delay labour and improve outcomes in women at risk of PTB.[3, 12-14] The American College of Obstetricians and Gynaecologists recommended progesterone treatment for the prevention of PTB for all ‘at risk’ gravidas, but blanket progesterone therapy would result in a marginal (0.3%) improvement in PTB rate. A meta-analysis has advocated progesterone as the treatment of PTB, but has recommended further research for the improved identification of candidate women for progesterone treatment.
The present study was designed to evaluate salivary progesterone as a biochemical marker for the prediction of early PTB in asymptomatic women who are at risk of PTB.
As measurements in saliva provide an estimate of the biologically active fraction of plasma hormones such as progesterone, and because saliva samples are readily collected and stored, if found useful the measurement of salivary progesterone can be readily introduced in clinical practice. In this study, we have also compared salivary progesterone concentration with the measurement of sonographic cervical length, which is currently a gold standard test for the prediction of PTB.
- Top of page
Asymptomatic women (n = 90) with a singleton pregnancy at 24–28 weeks of gestation, and with a history of previous PTB, preterm prelabour rupture of membranes, or late spontaneous miscarriage (at 20–28 weeks of gestation), attending a routine antenatal check-up in GTBH were included in the study. Prior to the study, the sample size was statistically calculated based on prevalence rate. A sample size of 80 women was adequate to detect a 25% overall recurrence (ranging from 10 to 40%) of PTB in women with a prior history of one or more PTBs, with a 10% absolute error and taking the alpha error as 5%. The sample size was inflated to allow for 20% follow-up loss. We therefore aimed for a sample size of 96 women, but were only able to perform the tests on 90 women because of time and money constraints. The exclusion criteria were: multiple pregnancy; known congenital malformation of fetus/uterus; antepartum haemorrhage; obstetric complications requiring iatrogenic PTB; non-reassuring fetal heart rate status; fetal growth restriction; the following medical disorders—hypertensive disorder of pregnancy, gestational diabetes, renal disease, cardiac disease, haematologic disorders, chronic liver disease, vaginal infections; cervical cerclage (current or planned); medication known to affect hormone concentration (such as antidepressants, corticosteroids or progesterone therapy); tocolytic therapy; any addiction; smoking; and the presence of any oral condition that would interfere with saliva collection, e.g. bleeding gums.
Following consent, women attended the clinic after overnight fasting and saliva samples were collected after rinsing the mouth with water 10 minutes before sample collection. Unstimulated passive drool samples (with a minimum of three) were collected in a sterile wide-mouthed glass or plastic container within 2 hours, at an interval of 30 minutes, and were then pooled together. The sample was refrigerated within 30 minutes and frozen, at or below −20°C until analysis.
Estimation of salivary progesterone
Each frozen sample was thawed and centrifuged at 1502 g for 5 minutes. The clear colourless supernatant was used for the quantitative assessment of salivary progesterone using an ELISA kit and following the manufacturer's protocol (De Meditec, gmbh, lise-meitner-strasse2, 24145 Kiel, Germany). The lowest analytical detectable concentration of progesterone that could be distinguished from the zero calibrator was 5 pg/ml at the 2SD confidence interval. The range of the assay was 0–5000 pg/ml. The intra-assay variation was determined by 20 replicate measurements of three saliva samples within one run. The within-assay coefficient of variation was 5.96, 6.44 and 9.63%. The interassay variation was determined by duplicate measurements of three saliva samples over 10 days. The coefficient of variation was 8.6, 9.4 and 10.1%.
Determination of cervical length
The women afterwards underwent a transvaginal scan for cervical length. During the TVS scan, the woman was laid in the dorsal position. The transvaginal transducer (LOGI Q 500MD 7 MHz) probe was introduced in the anterior fornix and the cervical length was measured three times from the external os to the internal os, and the shortest measurement was noted for the purpose of the study. The sum of two separate measurements was taken if the cervix was angulated.
Both tests, salivary progesterone and TVS cervical length, were repeated at a second visit after 3–4 weeks. All women recruited were seen at regular visits by their primary doctor until delivery. None of the women received progesterone supplemental therapy during the study. Those who were admitted because of preterm labour (PTL) pains received betamethasone coverage and antibiotics if relevant, but no tocolytic therapy was given. After delivery, women's records were checked for any hospitalisation, intervention, and treatment given by their primary doctor during the antenatal period. Delivery details, including gestational age, mode of delivery, neonatal outcome and admittance to a neonatal intensive care unit (NICU), were noted. Mothers and neonates were followed until the time of discharge from the hospital. Based on delivery outcome, women were classified as early PTB (delivering at <34 weeks of gestation), late PTB (delivering at 34–36+6 weeks of gestation) and term birth (≥37 weeks of gestation).
Statistical software spss 17.0 and stata 10.0 were used for the statistical analysis. An unpaired Student's t-test and a chi-square test or Fisher's exact test were used to compare clinical variables (age, socio-economic status, residential area, educational status, gravidity, parity and history of abortion, including maternal weight, height and bodymass index [BMI]). The change in progesterone concentration between the first and the second visit was examined with a Student's paired t-test. Receiver operating characteristic (ROC) curves were used to determine the critical values of salivary progesterone and cervical length. The area under the curve (AUC) was compared using stata. The cut-off values for salivary progesterone concentration and cervical length were decided at the points where the sums of sensitivity and specificity were a maximum. The 95% confidence interval of sensitivity, specificity, and negative and positive predictive value was calculated using the exact binomial method in stata. A one-way anova followed by Tukey's test at a 5% level of significance was applied to compare first-visit salivary progesterone and TVS cervical length, and the changes in these values between two visits across the three groups, i.e. early and late PTB and term birth. The Pearson correlation was used to explore relationships between cervical length and salivary progesterone. P < 0.05 was taken as the level of significance.
- Top of page
Approximately 4000 women were screened from the antenatal outpatient department (OPD) over a period of 1 year, and 160 women met the criteria for study entry. Of these, six declined to participate in the study, 32 were excluded because they developed medical or obstetrical complications after recruitment, 25 women were lost to follow-up, three had vaginal infection, two underwent preterm caesarean section because of fetal distress, one had to undergo a cerclage procedure and one was excluded because of bleeding gums. Ninety women completed the study, of which 38 delivered as PTB: thus the recurrence rate of PTB was 42.2%.
The sociodemographic factors of maternal age, socio-economic status, residential area, education status, gravidity, parity and number of previous miscarriages were not found to have any significant association with PTB (combined early and late) in the present study (Table 1). The average maternal weight of 50.1 kg (SD 5.7 kg; P < 0.001) and BMI of 21.3 kg/m2 (SD 3.8 kg/m2; P = 0.024) were significantly lower in women with PTB in comparison with women delivering at term (Table 1). Maternal weight and BMI values were also lower for women with early PTB compared with women with PTB (Table 2); however, there was not much change in the result of the biochemical markers after adjusted analysis, taking age, socio-economic status, weight and BMI as plausible confounding factors (OR 1.28, 95% CI 1.13–1.46; adjusted OR 1.26, 95% CI 1.10–1.44).
Table 1. Comparison of the sociodemographic characteristics in the two study groups
|Characteristics||Group I (n = 38, preterm) Number (%) || Group II (n = 52, term) Number (%) || P |
|Age (years)|| || ||0.550|
| <20||0 (0)||1 (1.9)|| |
| 20–30||36 (94.7)||49 (94.2)|
| 30–40||2 (5.2)||3 (5.7)|
| Socio-economic status || || ||0.195|
| Upper||0 (0)||0 (0)|| |
| Upper middle||2 (5.2)||6 (11.5)|
| Lower middle||7 (18.4)||16 (30.7)|
| Upper lower||28 (73.6)||30 (57.6)|
| Lower||1 (2.6)||0 (0)|
| Residential area || || ||0.932|
| Rural||3 (7.8)||4 (7.6)|| |
| Urban||28 (73.6)||40 (76.9)|
| Slum||7 (18.4)||8 (15.3)|
| Education status || || ||0.881|
| 0–5 years||15 (39.4)||23 (44.2)|| |
| 6–12 years||21 (55.2)||26 (50)|
| Graduation||2 (5.2)||3 (5.7)|
| Gravida || || ||0.538|
| 2||25 (65.7)||36 (69.2)|| |
| 3||10 (26.3)||13 (25)|
| ≥4||3 (7.8)||3 (5.7)|
| Parity || || ||0.818|
| 1||31 (81.5)||45 (86.5)|| |
| 2||4 (10.5)||5 (9.6)|
| ≥3||3 (7.8)||2 (3.8)|
| History of miscarriage || || ||0.958|
| 0||34 (89.4)||39 (75)|| |
| 1||4 (10.5)||12 (23.0)|
| 2||0 (0)||1 (1.9)|
| Weight (kg) ||50.1 (5.7)a||55 (5.4)a||<0.001|
| Body mass index (kg/m2) ||21.3 (3.8)a||23.2 (3.6)a||0.024|
Table 2. Comparison of maternal weight and body mass index (BMI) in the early preterm and preterm groups
| || PTB <34 weeks of gestation (n = 18) Mean (95% CI) || PTB <37 weeks of gestation (n = 38) Mean (95% CI) |
|Weight (kg)||49.7 (46.69–52.75)||50.1 (48.2–52.0)|
|BMI (kg/m2)||21.6 (20.3–22.9)||21.3 (20.1–22.6)|
The mean period of gestation at the time of delivery in group I was 34.0 (2.04) weeks and in group II was 38.2 (1.0) weeks. In group I, 52.6% of women delivered between 34 and 36+6 weeks of gestation, and 42.1% delivered between 32 and 33+6 weeks of gestation. The mean birthweight (P < 0.001) of neonates was significantly less in the preterm group, and as expected they stayed longer in the NICU (P = 0.006).
The concentrations of salivary progesterone measured in women with early and late PTB were significantly lower at both visits (3074 pg/ml at 24–28 weeks of gestation and 2400 pg/ml at 28–32 weeks of gestation) compared with women with term birth (4715 pg/ml at visit I and 4796 pg/ml at visit II, respectively). The salivary progesterone concentration significantly decreased between the preterm and term groups at each visit (P < 0.001; Table 3). In women who delivered prematurely, a 21.9% decrease in salivary progesterone concentration was observed between the two visits (Table 3).
Table 3. Mean salivary progesterone concentration in the two study groups at two visits
|Salivary progesterone (pg/ml)||Group I (preterm) (n = 38) Mean (SD) || Group II (term) (n = 52) Mean (SD) || P |
|Visit I (at 24–28 weeks of gestation)||3074.4 (1279.6)||4715.7 (1437.1)||<0.001|
|Visit II (at 28–32 weeks of gestation)||2400 (1313)||4796.6 (1468.9)||<0.001|
The concentration of salivary progesterone in the early preterm group was significantly lower at both visits compared with the term group, and decreased significantly between the two visits (Tables 4 and 5). On the contrary, women who delivered at term had an increase in salivary progesterone concentration between the two visits (Table 4).
Table 4. Comparison of the salivary progesterone (SP) concentration and the TVS cervical length (CL) between two visits in the early preterm, late preterm and term groups
|Parameters||Groups||Number of subjects||Mean (SD) Visit I (at 24–28 weeks of gestation) || Mean (SD) Visit II (at 28–32 weeks of gestation) || P ||% change (95% CI)|
|Salivary progesterone (pg/mL)||Group Ia (early preterm)||18||2587.7 (1193.9)||1775.8 (968.1)||0.002||29.6 (17.8–41.4)|
|Group Ib (late preterm)||20||3512.5 (1219.3)||2962.5 (1347.7)||0.024||14.9 (2.69–27.1)|
|Group II (term)||52||4715.7 (1437.1)||4796.6 (1468.9)||0.429||–2.7 (from −8.0 to 2.5)|
|TVS CL (mm)||Group Ia (early preterm)||18||25.5 (4.8)||22.09 (5.14)||<0.001||13.4 (7.14–19.7)|
|Group Ib (late preterm)||20||26.92 (5.60)||24.31 (5.09)||<0.001||9.5 (7.1–11.8)|
|Group II (term)||52||30.8 (4.09)||28.2 (4.3)||<0.001||8.4 (6.2–10.6)|
Table 5. Multiple comparisons of salivary progesterone concentration between the early preterm, late preterm and term groups
|Salivary progesterone||Visit I (at 24–28 weeks of gestation)||Visit II (at 28–32 weeks of gestation)|
|Mean difference (95% CI) early preterm versus term (P value)|| || |
|Mean difference (95% CI) late preterm versus term (P value)|| || |
|Mean difference (95% CI) early versus late preterm (P value)|| || |
An ROC curve was used to calculate a critical salivary progesterone concentration below which significant number of women went into PTL and had PTBs. Apart from the single critical cut-off value of salivary progesterone for predicting early PTB, which was calculated for all the ‘at risk’ women in the present study (Figure 1), a separate critical value for each visit was also calculated (Figures 2 and 3; Table 6). When salivary progesterone concentration was used for predicting early PTB, using the values of both visits, a cut-off of 2575 pg/ml was calculated, which had sensitivity, specificity, and positive and negative predictive values of 83% (95% CI 58.6–96.4%), 86% (95% CI 75.9–93.1%), 60% (95% CI 38.6–78.8%) and 95% (95% CI 87.1–99.0%), respectively (Figure 1). Also, when all values from both visits were considered, a salivary progesterone level below 3625 pg/ml had the sensitivity and specificity of 81% (95% CI 65.7–92.3%) and 77% (95% CI 63.2–87.5), respectively, in predicting any (early and late) PTB. The positive and negative predictive values for predicting PTB were 72% (95% CI 56.3–84.6) and 85% (95% CI 72.7–93.8), respectively (Figure 2; Table 6).
Table 6. Predictive accuracy of TVS cervical length and salivary progesterone concentration for PTB <37 weeks of gestation
| ||Critical value||Sensitivity%(CI)||Specificity%(CI)||PPV %(CI)||NPV %(CI)||AUC (95% CI)|
| 24-32 weeks |
|TVS cervical length (mm)||27|| || ||69 (57.9–83.6)|| || |
|Salivary progesterone (pg/ml)||3625|| || ||72 (56.3–84.6)|| || |
| 24–28 weeks |
| TVS cervical length (mm)||28|| || || || || |
|Salivary progesterone (pg/ml)||3950|| || || || || |
| 28–32 weeks |
| TVS cervical length (mm)||26|| || || || || |
|Salivary progesterone (pg/ml)||3412|| || || || || |
Figure 1. ROC curve for the estimation of a critical cut-off value of salivary progesterone and TVS cervical length in the prediction of PTB at <34 weeks of gestation; AUC of SP (0.893, 95% CI 0.808–0.977); AUC of CL (0.770, 95% CI 0.652–0.887).
Download figure to PowerPoint
Figure 2. ROC curve for the estimation of a single critical cut-off value of salivary progesterone in the prediction of PTB at <37 weeks of gestation (AUC 0.857, 95% CI 0.778–0.937).
Download figure to PowerPoint
Figure 3. ROC curve for the estimation of the single critical cut-off value of TVS cervical length in the prediction of PTB at <37 weeks of gestation (AUC 0.771, 95% CI 0.674–0.868).
Download figure to PowerPoint
The mean TVS cervical length at first visit in the preterm group (groupI) was 25.8 mm (SD 6.4 mm), which was significantly shorter (P < 0.001) in comparison with the term group (group II): 31.08 mm (SD 4.21 mm). The cervical length had reduced by the second visit for both groups: 22.6 mm (SD 6.8 mm) in the preterm group and 28.2 mm (SD 4.3 mm) in the term group, respectively. This decrease in TVS cervical length was 12.3% (over a period of 4 weeks) in the preterm group. The mean TVS cervical length among term, and early and late preterm groups decreased significantly over visits; with a decrease of 13.4% (95% CI 7.14–19.7%; P = 0.001) in the early preterm group (Table 4).
Using an ROC curve, the critical cut-off measurement of 27 mm for TVS cervical length at or beyond 24 weeks of pregnancy had sensitivity, specificity, and a positive and negative predictive value of 77% (95% CI 52.3–93.5%), 68% (95% CI 56.0–78.6%), 37% (95% CI 22.4–55.2%) and 92% (95% CI 81.8–97.9%), respectively, for the prediction of early PTB (Figure 1). Similar to salivary progesterone, the critical value of cervical length at each visit and from both visits combined was considered. The critical cut-off measurement of 27.1 mm for TVS cervical length at or beyond 24 weeks of gestation had sensitivity, specificity, and positive and negative predictive values of 65% (95% CI, 48.6–80.4%), 79% (95% CI 65.3–88.9%), 69% (95% CI 57.9–83.6%) and 76% (95% CI 62.4–86.5%), respectively, for the prediction of any (early and late) PTB (Figure 3; Table 6). Low salivary progesterone alone could predict PTB beyond 24 weeks of gestation (Table 6).
When a direct comparison of both the tests was performed on the basis of ROC curves for any (early and late) PTB, it was found that the two AUCs did not differ significantly at <37 weeks of gestation (progesterone area under curve, AUC, 0.857, 95% CI 0.778–0.937; cervical length AUC, 0.771, 95% CI 0.674–0.868; P = 0.112; Figures 2 and 3). However, for early PTB at <34 weeks of gestation the AUC (0.893, 95% CI 0.808–0.977) of salivary progesterone was higher than the AUC of TVS cervical length (0.770, 95% CI 0.652–0.887) (P = 0.040; Figure 1). There was a positive association between the saliva progesterone concentration after log transformation and cervical length at the first (24–28 weeks of gestation; r = 0.371) and second visit (28–32 weeks of gestation, r = 0.508).
We assessed the combined predictive value of both tests, by dividing the women in four groups: both tests negative (salivary progesterone, SP > 3625 pg/mL; cervical length, CL > 27.1); both tests positive (SP < 3625 pg/mL; CL < 27.1); SP positive, CL negative (SP < 3625 pg/mL; CL > 27.1); and SP negative, CL positive (SP > 3625 pg/mL; CL < 27.1). On the basis of the cut-off obtained from the ROC curves of SP and CL (Table 6), the percentage of any PTB when both tests were positive, was 52.6% (20/38). When only CL was positive, the percentage of PTB was 65% (25/38). When only SP was positive, the percentage of PTB was 81% (31/38). When both the tests were negative the percentage of PTB was only 5% (2/38; Table 7).
Table 7. Predictive accuracy of a combination of salivary progesterone concentration and TVS cervical length for PTB <37 weeks of gestation in women demonstrating values lower than the calculated threshold for each test
| SP < 3625 pg/ml + CL < 27.1 mm ||Sensitivity% (n/n, 95% CI)||Specificity% (n/n, 95% CI)||PPV% (n/n, 95% CI)||NPV% (n/n, 95% CI)|
|Both positive||52 (20/38, 35.8–69.0)||86 (45/52, 74.2–94.4)||74 (20/27, 53.7–88.9)||71 (45/63, 58.6–82.1)|
- Top of page
Early PTB causes significant neonatal mortality and morbidity. There is a need to identify women at high risk for early PTB (<34 weeks of gestation), and ideally an accurate prediction should be possible in their first pregnancy. Multiple biochemical and biophysical markers have been tested, but an accurate predictor of early PTB remains elusive.[19-21]
The measurement of biomarkers in the saliva has the advantages of reflecting biologically active substances, non-invasive testing, and easy collection and storage. It is therefore more suitable for steroid hormone tests. In the preterm group, salivary progesterone concentration was lower in comparison with the term group, and decreased significantly from first to second visit. The decline in progesterone concentration was more significant in the early PTB group compared with the late PTB and term birth groups (Table 5). On the other hand, the salivary progesterone concentration increased in the term birth group from the first to the second visit (Table 4). The present study has confirmed the hypothesis that progesterone concentration is significantly lower in women who have spontaneous PTB than in those delivering after 37 weeks of gestation, and also that the spontaneous onset of PTL is preceded by a decrease in progesterone concentration. Our study has added strength to the concept that measurement of salivary steroids is a simple yet important method for the early identification of women at risk for PTL. The critical cut-off value of salivary progesterone was below 2575 pg/ml (sensitivity 83%; specificity 86%) at 24–32 weeks of gestation, the level at or below which 83% women were likely to have early PTB (Figure 1).
Delivery before 34 weeks of gestation is associated with greater neonatal morbidity than delivery between 34 and 37 weeks of gestation. Women who delivered before 34 weeks of gestation, at 34–37 weeks of gestation and at term (≥37 weeks of gestation) were therefore considered separately in some analyses in the present study. Only a few other studies available in the literature report salivary progesterone concentration as being associated with PTB. In a recent study from the UK, where serial salivary progesterone concentration was measured from 24 weeks of gestation in asymptomatic high-risk women, an 89% lower progesterone concentration was observed in women who delivered before 37 weeks of gestation compared with those who had term pregnancies. Furthermore, they reported that the concentration of salivary progesterone was relatively lower to start with in women with early PTB (<34 weeks of gestation), and did not rise as the gestational age progressed. In contrast, Bell found that the concentration of plasma progesterone in women with PTB was similar to that of women with term births. Many previous studies have reported on the salivary estriol/progesterone ratio, and found that it is the increase in estriol rather than the decrease in progesterone that is associated with PTB.[23, 24] To the best of our knowledge, this is the first study where mean salivary progesterone concentration and critical cut-off values have been calculated and shown to be a good predictor of PTB and early PTB.
It has been seen that cervical shortening changes start as early as 18–22 weeks of gestation in women about to have PTL.[25, 26] Our findings confirmed an inverse correlation between TVS length of cervix and frequency of PTB. Similar to salivary progesterone concentration, the greatest shortening in cervical length was observed in women who delivered before 34 weeks of gestation (Table 4).
In the present study we found a linear and positive correlation of salivary progesterone with TVS cervical length. This suggests that the change (increase or decrease) in TVS cervical length is reflected in salivary progesterone concentrations. This may infer that there are common mechanistic pathways to PTL. When the tests were compared, salivary progesterone was more predictive than cervical length for delivery at <34 weeks of gestation (Figure 1), but not at 37 weeks of gestation, which might indicate that low progesterone precedes the shortening of the cervix. A common mechanism could also be inferred by the lack of any additive effect of saliva progesterone and cervical length when the tests were combined (Table 7). Although no studies comparing the two directly are available in the literature a recent study has shown that as the TVS cervical length becomes less than 15 mm, progesterone gel supplementation becomes effective in preventing PTB.
Salivary progesterone estimation in a high-risk asymptomatic population can identify the subgroup of women who require and might gain maximum benefit from progesterone supplementation. Moreover, because of the ease of sample collection, the safety and simplicity of the procedure, the non-invasive nature of the procedure, the preservation of the patient's privacy, and the avoidance of any discomfort and psychological stress associated with a TVS procedure, the measurement of salivary progesterone might emerge as an excellent tool for the prediction of early PTB.
The present study emphasises the fact that low salivary progesterone (<2575 pg/mL) can be used as a predictor of early PTB, and can identify women at risk of early PTB. However, more studies are required on a larger sample size before it can be put to clinical use.