Accuracy of salivary estriol testing compared to traditional risk factor assessment in predicting preterm birth

R. Phillip Heine, MD,a James A. McGregor, MD, CM,b and Vivian K. Dullien, PhDc Pittsburgh, Pennsylvania, Denver, Colorado, and Dublin, California
OBJECTIVE: The objective was to compare the predictive accuracy (percentage of correct vs incorrect predictions) of salivary estriol levels (SalEst; Biex, Inc, Dublin, Calif) with that of the modified Creasy score for predicting preterm labor followed by preterm delivery. STUDY DESIGN: A triple-blinded prospective trial was conducted at 8 US centers. RESULTS: Among 601 evaluable patients, serial salivary estriol testing correctly predicted the appropriate outcome 91% of the time and the Creasy scoring method correctly predicted the appropriate outcome 75% of the time (McNemar test P < .001). Among subjects with Creasy scores 10 (high-risk group, n = 152), use of salivary estriol testing correctly predicted the end point 87% of the time, compared with only 7.2% correctly predicted by modified Creasy scoring (McNemar test P < .001). CONCLUSION: Salivary estriol assessment was more accurate in predicting outcome than was modified Creasy scoring. (Am J Obstet Gynecol 1999;180:S214-8)

Key words: Estriol, prematurity, preterm birth, preterm labor, risk, saliva

Preterm birth is a leading cause of neonatal morbidity and of excess perinatal and pediatric costs. Preterm labor can be considered a multifactorial syndrome caused by multiple etiologic mechanisms.1-5 Traditional risk scoring systems assess risks of preterm birth on the basis of patients medical history, lifestyle, behavior, and demographic factors.6 The most commonly used risk scoring system6 was developed by Papiernik7 and modified by Creasy et al.8 This widely used risk scoring system detects fewer than half of women at risk for preterm birth and is less accurate in assessing women in their first pregnancy.6 Thus there is a compelling need for new clinical tools, including biologic markers, that can more accurately determine which pregnant women are at risk for preterm birth and point to possible pathophysiologic mechanisms. Knowledge of such mechanisms could lead to formulation of etiologically focused interventions, which could then be evaluated in controlled trials. Estriol first appears at 9 weeks gestation in a normal pregnancy, and plasma concentrations increase throughout gestation along with those of estrone and estradiol.2, 9, 10 Estriol is the most abundant estrogen in late human pregnancy.2 Approximately 3 to 5 weeks before parturition, concentrations of estriol surge, and they continue to increase until birth.11 Estriol plays roles in preparing uter-

ine tissues, possibly signaling parturition, in both human beings and great apes.2, 9, 12 A rise in estriol level, which can be monitored in saliva, has been shown to precede parturition, regardless of whether it occurs preterm, at term, or postterm.13, 14 During the past 2 decades saliva has achieved increasing recognition as a useful matrix for monitoring hormone levels and therapeutic drug concentrations.15 Saliva estriol levels reflect the concentrations of free, unconjugated estriol in the plasma.16 Advantages of saliva as a test matrix include ease of use, noninvasive and painless collection procedures, stability during transport, and a well-established enzyme immunoassay.15, 17 In this study we compared measures of predictive accuracy (percentage of correct predictions vs percentage of incorrect predictions) for serial salivary estriol (SalEst; Biex, Inc, Dublin, Calif) testing with modified Creasy antenatal risk factor assessment for predicting risk of preterm labor followed by preterm delivery before 37 weeks gestation in a large multicenter prospective controlled trial. Methods We conducted a prospective, longitudinal, triple-blind, multicenter study that compared the predictive abilities of serial estriol testing with those of traditional Creasy score testing. The trial was conducted between May 1995 and December 1996 at the antenatal clinics of 8 US medical centers. The study was approved by each participating institutions institutional review board. Subjects gave informed written consent in English, Spanish, or both. The study was triple blinded; neither care providers, laboratory personnel, nor subjects had access to test results.

From the University of Pittsburgh,a the University of Colorado Health Sciences Center,b and Biex, Inc.c Reprint requests: R. Phillip Heine, MD, University of Pittsburgh, Magee Womens Research Institute, 204 Craft Ave, Room 530, Pittsburgh, PA 15213. Copyright 1999 by Mosby, Inc. 0002-9378/99 $8.00 6/0/94434

S214

Volume 180, Number 1, Part 3 Am J Obstet Gynecol

Heine, McGregor, and Dullien S215

Samples were kept frozen at 20C and processed in batches after all subjects were delivered. Preterm labor and delivery was defined as spontaneous preterm labor with intact membranes that resulted in preterm birth within 72 hours before achievement of 37 weeks gestation. Gestational age was estimated by the last menstrual period, clinical parameters, ultrasonography, and numerous examinations. Most subjects had 1 ultrasonographic examination. If there was a >14-day difference between the age determined by last menstrual period and ultrasonographic evaluation, the ultrasonographic measurement and neonatal examination were used to determine gestational age. Gestational age determinations were made without knowledge of test results. Risk status for preterm labor was designated according to the Creasy score at study intake. A score 10 was considered to indicate high risk and a score <10 was considered to indicate low risk.2 Subjects were enrolled at 21 to 25 weeks gestation and followed up at biweekly intervals until delivery. A baseline examination included (1) assessment for signs and symptoms of preterm labor (abdominal pain, backache, flulike symptoms, pelvic pressure, cramping, change in vaginal discharge, vaginal bleeding, diarrhea); (2) evaluation of any contractions, rupture of membranes, or leakage of amniotic fluid; and (3) a digital vaginal examination to assess the position, consistency, length, and dilatation of the cervix. Demographic and obstetric data, including details both of previous pregnancies and of the current pregnancy, were obtained from the patients medical record. Subsequent clinic visits included questioning the subject for signs or symptoms of preterm labor and performing a cervical examination every 4 weeks, beginning at week 27. Changes in medication, hospitalization, emergency department visits, and perceived uterine contractions were also recorded. Inquiries intended to ensure that saliva samples taken at home were collected and submitted properly were also made at each visit. Women with a singleton pregnancy were considered eligible for study participation if their pregnancy was between 21 and 25 weeks gestation and they were 18 years old and gave informed consent. Exclusion criteria included symptoms of preterm labor, tocolytic therapy, presence of placenta previa, cervical changes necessitating cerclage, ruptured fetal membranes in the current pregnancy, preeclampsia, medications known to affect hormone levels, a planned cesarean delivery, the presence of major congenital abnormalities, intrauterine growth restriction, erythroblastosis fetalis or fetal death, the presence of an oral condition that would interfere with saliva collection (eg, bleeding gums), the presence of a maternal medical complication that would require specialized obstetric care (eg, chronic hypertension, diabetes mellitus) or of any other serious medical disease, and a history of drug or ethanol abuse.

Subjects were withdrawn from the study and their results were not analyzed if they did not supply saliva samples for 3 consecutive weeks or did not attend 2 consecutive clinic visits (n = 189), unless the subject was hospitalized and was supplying weekly saliva samples. Because of previously recognized suppression of salivary estriol with use of corticosteroids and possibly with tocolysis,18 women who received betamethasone or dexamethasone to induce fetal lung maturation or for tocolysis were also not included in the analysis (n = 35). Women with premature rupture of membranes and preterm birth (n = 13) and those whose preterm births were deemed to be caused by obstetric indications (n = 10) were also excluded from the analysis because it was considered that preterm premature rupture of membranes was associated with causal mechanisms not linked with estriol. Women who had preterm labor and term birth (>37 weeks gestation) were also excluded. Most of these subjects did not have delivery within 72 hours of the onset of preterm labor (n = 55). These exclusions were created to ensure focused evaluation of the primary outcome, preterm birth occurring within 72 hours of preterm labor. Subjects were instructed to collect unstimulated saliva between 9 AM and 8 PM, because no significant diurnal variations in estriol levels have been found between these hours.19 Subjects were asked to rinse their mouths with water, wait 10 minutes, and then drool 2 mL saliva into the plastic tube that was provided. Subjects were asked to avoid eating, drinking, smoking, tooth brushing, mouth washing, or gum chewing 1 hour before sample collection. Previous evaluations (unpublished data) showed no differences between mailed samples and those collected in the clinic. Samples collected at home were therefore sent to the clinic in mailers, where they were stored at 20C and then sent frozen to Biex for batch assay. Samples were tested for unconjugated (free) estriol by means of a previously described competitive microplate enzyme immunoassay (SalEst; Biex).14 Salivary estriol values 2.1 ng/mL were designated as single positive. A finding of levels 2.1 ng/mL on 2 consecutive tests was deemed a rescreen positive finding. Statistical methods. The McNemar test was used to compare the predictive accuracy of salivary estriol monitoring with that of Creasy scoring. Statistical significance was set at an alpha of .05. Results Patient demographics. A total of 956 pregnant women were enrolled in the study, 302 (31.6%) of whom were designated as at high risk (Creasy score 10) and 654 (68.4%) of whom were designated as at low risk. The demographic characteristics of the enrolled and evaluable women at baseline were similar among test centers with respect to age distribution, prepregnancy weight, obesity,

S216 Heine, McGregor, and Dullien

January 1999 Am J Obstet Gynecol

Table IA. Numbers of enrolled and analyzed subjects for primary and secondary end points
Enrolled Subjects High-risk group Low-risk group Combined groups Preterm births (<37 wk) Spontaneous preterm births PPROM Medically indicated No. 302 654 956 121 71 21 29 % 31.5 68.5 12.6 7.4 2.2 3.0 Analyzed No. 152 449 601 23 23 0 0 % 25.3 74.7 3.8 3.8 0 0

Table IIA. Contingency table for the McNemar test for the high-risk group (n = 152): single test positive
Creasy score correct Estriol assay correct Estriol assay incorrect
TOTALS

Creasy score incorrect 96 45 141

Totals 103 49 152

7 4 11

McNemar test P .001, odds ratio 24.00 (95% binomial confidence limit 8.83-65.25).

PPROM, Preterm premature rupture of membranes.

Table IIB. Contingency table for the McNemar test for the total study group (n = 601) with 2 consecutive positive salivary estriol assay results
Creasy score correct Creasy score incorrect 132 21 153 Totals 544 57 601

Table IB. Subjects excluded from analysis (n = 355)

No. Treated with betamethasone/dexamethasone and/or tocolytics PPROM Medically indicated preterm births Term with preterm labor Nonevaluable subjects PPROM, Preterm premature rupture of membranes. 35 13 10 55 242 Estriol assay correct Estriol assay incorrect
TOTALS

412 36 448

McNemar test P .001, odds ratio 3.67 (95% binomial confidence limit 2.54-5.30).

Table IIC. Contingency table for the McNemar test for the high-risk group with 2 consecutive positive salivary estriol assay results (n = 152)
Creasy score correct Creasy score incorrect 127 14 141 Totals 132 20 152

Table IC. Subjects included in analysis

Total Low-risk group High-risk group Combined groups 449 152 601 Preterm births 12 11 23 Term births 437 141 578 Estriol assay correct Estriol assay incorrect
TOTALS

5 6 11

McNemar test P .001, odds ratio 21.17 (95% binomial confidence limit 9.33-48.00).

Table IID. Accuracy (%) of salivary estriol assay and the modified Creasy score
Group Single test Rescreen positive High risk Combined High risk Salivary estriol assay* 67.7 90.5 86.8 Creasy score* 7.2 74.5 7.2 Significance P < .001 P < .001 P < .001 OR 24 3.67 21.1 95% CI 8.8-65.2 2.5-5.3 9.3-48.0

OR, Odds ratio; CI, confidence interval. *Percentage of test results correctly predicting outcome.

weight at enrollment, height, number of children, gravidity, number of abortions, and number of previous preterm and term pregnancies. Most enrolled women were between 20 and 35 years old (78%), weighed >100 lbs (99%), had completed 12 years of education (79%), were married or living with a mate (70%), had not had a previous preterm pregnancy (82%), had not had an abortion (53%), and had 1 child (57%). Most (83.0%) of the women who were designated as being at high risk had 1 risk factor worth 10 points (eg, previous premature

delivery, uterine anomaly, second-trimester abortion, exposure to diethylstilbestrol). A total of 13,347 samples were submitted. Only 31 of these 13,347 samples (0.23%) could not be analyzed because of tube leakage or insufficient sample. The mean interval between samples was 1.1 weeks (SD = .1 week) for evaluable subjects. Numbers of enrolled and analyzed subjects are shown in Table IA. Subjects excluded from analysis of primary end points are described in Table IB. Noncompliance

Volume 180, Number 1, Part 3 Am J Obstet Gynecol

Heine, McGregor, and Dullien S217

was the most common reason for discontinuation from the study (189/242), followed by development of exclusion criteria after enrollment (49/242) and fetal death (4/242). The numbers of subjects analyzed for accuracy assessment are shown in Table IC. A total of 601 patients were included in this analysis, 152 (25.3%) in the highrisk group and 449 (74.7%) in the low-risk group. The accuracy analysis was done on 2 groups (n = 601), women with preterm labor followed by preterm delivery and women with term births. There were 23 spontaneous preterm births (11 in the high-risk group and 12 in the low-risk group) and 578 term deliveries (141 in the highrisk group and 437 in the low-risk group; Table IC). Contingency tables and results for McNemar tests for selected comparison groups are shown in Tables IIA, IIB, and IIC. Table IIA shows results among women judged to be at high risk (Creasy score 10). A single positive salivary estriol test was more predictive of the primary outcome than was Creasy scoring (McNemar test P < .001, odds ratio 24.00). Tables IIB and IIC illustrate the contingency tables for the total study and the high-risk group for subjects with rescreen positive results, that is, 2 consecutive weeks of positive estriol assay results (2.1 ng/mL). For the entire group the odds ratio for estriol testing versus Creasy score being correct was 3.67 (95% confidence interval 2.54-5.30). Among high-risk subjects a positive rescreen test status had an even greater likelihood of correctly predicting the end point (odds ratio 21.17, 95% confidence interval 9.33-48). Table IID shows the accuracy (percentages of tests correctly predicting outcome) of single positive and rescreen positive salivary estriol test results compared with Creasy scoring performed at study entry. For the entire (combined) group, salivary estriol assay rescreen positive status correctly predicted the end point 91% of the time, whereas the traditional Creasy scoring predicted the outcome 75% of the time (P < .001). Among the high-risk subjects, a rescreen positive result correctly predicted the outcome 87% of the time, versus only 7.2% of the time for Creasy scoring (McNemar test P < .001). Comment We conducted a large (956 enrolled and 601 evaluable subjects) triple-blinded multicenter assessment of serial testing of salivary estriol level versus the traditionally used Creasy risk scoring system for predicting preterm birth. A single positive test (2.1 ng/mL) and 2 consecutive positive weekly tests (rescreen positive) were both superior to Creasy scoring for correctly predicting preterm birth. The increased accuracy was greatest for the high-risk group, which was rescreen positive. Factors in this study that support the internal and external validations of these results include the studys (1) relatively large size (601 evaluable subjects), (2) tripleblinded prospective design, (3) diverse study popula-

tions, and (4) strict definition of the study end point. Factors that limit applicability of our results to other populations include exclusions at enrollment and during the study, including subjects who were noncompliant or had premature rupture of membranes or use of corticosteroids or tocolytics. Conclusion. This prospective study analyzed the accuracy of serial estriol testing in comparison with Creasy score in the prediction of study-defined preterm birth. This analysis is part of a larger assessment of salivary estriol test performance presented elsewhere.20 The data present a promising role for salivary estriol testing in traditional high-risk populations to determine which women are truly not at risk for early delivery, creating opportunities for reducing unnecessary interventions in high-risk populations. Further research is underway, focusing on (1) understanding possible roles for salivary estriol testing in detecting and evaluating putative placental insufficiency and premature rupture of membranes and prolonged pregnancy (postdates) and (2) formulating specific clinical responses for woman-fetus dyads at risk for preterm parturition and birth as a result of presumed estriol-linked endocrine or paracrine mechanisms.

REFERENCES

1. American College of Obstetricians and Gynecologists. ACOG technical bulletin 133: preterm labor. Washington: American College of Obstetricians and Gynecologists; Oct 1989. 2. Stubblefield P. Causes and prevention of premature birth: an overview. In: Fuchs R, Fuchs F, Stubblefield P. Preterm birth: causes, prevention, and management. 2nd ed. New York: McGraw Hill; 1993, P. 3-40. 3. Carroll SG, Sebire NJ, Nicolaides KH. Preterm prelabour amniorrhexis. Pearl River (NY): Parthenon Publishing; 1996. 4. McGregor JA, French JI, Parker R, et al. Prevention of premature birth by screening and treatment for common genital tract infections: results of a prospective controlled evaluation. Am J Obstet Gynecol 1995;173:157-87. 5. Romero R, Mazor M. Infection and preterm labor. Clin Obstet Gynecol 1988;31:553-84. 6. American College of Obstetricians and Gynecologists. ACOG technical bulletin 206: preterm labor. Washington: Jun 1995. 7. Papiernik E. Coefficient de risque daccouchement premature. Presse Med 1969;77:793-4. 8. Creasy RK, Gummer BA, Liggins GC. System for predicting spontaneous preterm birth. Am J Obstet Gynecol 1980;55:692-5. 9. Speroff L, Glass RH, Kase NG. The endocrinology of pregnancy. In: Brown C. Clinical gynecologic endocrinology and infertility. 4th ed. New York: Williams & Wilkins; 1994, pp.317-50. 10. Fuchs A, Fuchs F. Physiology and endocrinology of parturition. In: Gabbe SG, Niebyl JR, Simpson JL, editors. Obstetrics. 3rd ed. New York: Churchill Livingstone; 1996, pp.111-36. 11. Darne J, McGarrigle HH, Lachelin GC. Saliva oestriol, oestradiol, oestrone and progesterone levels in pregnancy: spontaneous labour at term is preceded by a rise in the saliva oestriol:progesterone ratio. Br J Obstet Gynaecol 1987;94:22735. 12. Anderson JN, Peck EJ, Clark JH. Estrogen-induced uterine responses and growth: relationship to receptor estrogen binding by uterine nuclei. Endocrinology 1975;96:160-7. 13. Goodwin TM, Jackson GM, McGregor JA, et al. Increased incidence of preterm labor and preterm delivery associated with in-

S218 Heine, McGregor, and Dullien

January 1999 Am J Obstet Gynecol

creased salivary estriol level [abstract 59]. Am J Obstet Gynecol 1996;174:326. 14. McGregor JA, Jackson GM, Lachelin GC, et al. Salivary estriol as risk assessment for preterm labor: a prospective trial. Am J Obstet Gynecol 1995:173:1337-42. 15. Quissel D. Steroid hormone analysis in human saliva. In: Malamud D, Tabak L, conference co-chairs. Saliva as a diagnostic fluid. Conference proceedings. Panama City, Florida, October 22-25, 1992. Ann NY Acad Sci 1993, 694: 1-348. 16. Lachelin GC, McGarrigle HH. A comparison of saliva, plasma unconjugated, and plasma total oestriol throughout normal pregnancy. Br J Obstet Gynaecol 1984;91:1203-9.

17. Mandel ID. The diagnostic used of saliva. J Oral Pathol Med 1990;19:119-25. 18. Hendershott CM, Dullien V, Goodwin TM. Serial betamethasone administration and effect on maternal salivary estriol levels [abstract 437]. J Soc Gynecol Invest 1996;3:276A. 19. McGregor J, Barrett J, Hastings C, Diurnal variation of salivary estriol in pregnancy [abstract 167]. Am J Obstet Gynecol 1997;176:S56. 20. McGregor JA, Heine P, Artal R, et al. Results of a prospective, blinded, multicenter trial of salivary estriol for risk of preterm labor and delivery [abstract 39]. Am J Obstet Gynecol 1998;178:S18.