Regarding the prediction of preterm delivery (PTD) inasymptomatic low-risk population, numerous investiga-tors have reported the usefulness of transvaginal sono-graphic (TVS) assessment of the uterine cervix in thesecond trimester (Hasegawa et al. 1996; Hassan et al.2000; Heath et al. 1998; Iams et al. 1996; Taipale andHiilesmaa 1998). Among the different sonographic pa-rameters, majority of investigators have focused on cer-vical length (CL) measurement. It has become widelyaccepted as a well standardized method of cervical as-sessment, and requires a relatively short period of train-ing (Hasegawa et al. 1996; Hassan et al. 2000; Heath etal. 1998; Iams et al. 1996; Taipale and Hiilesmaa 1998).In addition, CL measurement was shown to be a betterscreening test regarding the prediction of PTD in low-risk population compared to other sonographic parame-ters (e.g., funneling) and to classic bimanual examination(Berghella et al. 1997; Gomez et al. 1994; Hoesli et al.
Address correspondence to: Dr Ozren Grgic, Sveti Duh Hospital,
University Department of Obstetrics and Gynecology, Sveti Duh 64,10000 Zagreb, Croatia. E-mail: [email protected]
333
2003; Iams et al. 1996; Matijevic and Grgic 2004; Owenet al. 2001; To et al. 2001). Disadvantages of this screen-ing test are low sensitivity, low positive predictive value(PPV) and the low prevalence of PTD in a low-riskpopulation; therefore, to get acceptable specificity thecutoff values have to be set below the fifth percentile(Hasegawa et al. 1996; Hassan et al. 2000; Heath et al.1998; Hoesli et al. 2003; Iams et al. 1996; Taipale andHiilesmaa 1998).
Recently, cervical gland area (CGA) was describedas a new sonomorphological parameter (Fukami et al.2003; Sekiya et al. 1998; Yoshimatsu et al. 2002). Thesonographic disappearance of CGA in the second trimes-ter implies accelerated cervical maturation, and could beused as an additional sonographic screening test for theprediction of PTD in low-risk population (Fukami et al.2003). Presence or absence of cervical glands was clas-sified as an “all or nothing” phenomenon, but the inves-tigators pointed out that its qualitative assessment mighthave some use in prediction of PTD (Fukami et al. 2003;Sekiya et al. 1998).
In this prospective study, we compared the diagnos-
tic accuracy of CL measurement and qualitative glandu-
334 Ultrasound in Medicine and Biology Volume 32, Number 3, 2006
lar cervical score (QGCS) assessed by TVS in the secondtrimester, regarding the prediction of PTD �34 com-pleted wk and between 34 to 37 wk in the asymptomaticlow-risk population.
MATERIALS AND METHODS
Study design and subjectsThis prospective cohort study was part of a research
project designed to assess the efficacy of different diag-nostic methods as potential screening tests for late spon-taneous miscarriage and preterm delivery (Croatian Min-istry of Science, project no. 0129111). The target popu-lation was: nulliparous women with uncomplicatedsingleton pregnancy who were sonographically exam-ined between 16 to 23 completed wk at University De-partment of Obstetric and Gynecology, “Sveti Duh” Hos-pital, Zagreb, Croatia. At that gestational age, all womenwere routinely booked for anomaly scan, and those sat-isfying our inclusion criteria were approached. The studyprotocol was approved by local and national ethicalcommittees, and all women included in the study gavetheir informed written consent. The exclusion criteria forenrollment were: (1) Clinically and laboratory suspectedinfection (white cell count �14 � 1012/L, C reactiveprotein � 10 mg/L) at enrollment. (2) Evidence of cer-vical and vaginal infection on swabs, if taken. (3) Historyof surgical procedure on the cervix. (4) Developmentalmalformations of the Müllerian ducts. (5) Cervical cer-clage before enrollment. (6) Clinically palpable contrac-tions. (7) Other medical conditions that are risk factorsfor preterm delivery. (8) Major congenital fetal anoma-lies. (9) Intrauterine fetal death.
Gestational age was determined by comparing a lastcertain menstrual period (if available) with the sono-graphic evaluation at the first trimester scan. Concor-dance between the biometric parameters and the men-strual date of 7 d or less confirmed the last menstrualperiod; otherwise, the biometric data were used. Theresults obtained in the study were known to the principalinvestigators only and the course of the pregnancy wasunaltered by the participation in the study.
Collection of data and ultrasonographyTransvaginal sonography was performed with a
7-MHz transducer with an angle of 160° (model 6117,Aloka 5500, ALOKA Co. Ltd., Tokyo, Japan) that op-erated at a constant output power of 0 dB and a total gaincontrol of 55 to 65 dB to avoid false positive or negativedetection of cervical mucus or glandular invasion. EachTVS was performed according to the protocol describedby Owen et al. (2001). This method has a mean interob-server variation of 4% to 10% and a mean intraobserver
variation of 5% (Burger et al. 1997; Zorzoli et al. 1994).
An adequate image was defined as presence of externalcervical axis, endocervical canal and internal cervicalaxis, which is the cranial end of the cervical canal adja-cent to the intrauterine cavity.
Cervical length was measured with electronic cali-pers as the linear distance between the external axis andthe functional internal axis along a closed endocervicalcanal, and the length below the fifth percentile for ourpopulation was defined as shortened.
Qualitative glandular cervical score (score range 0to 4) was defined as a summary of the cervical mucusarea and glandular invasion score. It is presented in Table1. The score below the fifth percentile for our populationwas defined as low. The hypoechogenic translucency ofcervical mucus area in the middle of the endocervicalcanal was measured by area trace (Fig. 1a). Hypo- orhyperechogenic translucency surrounding the canal thatprobably corresponds to the histologic cervical glandarea was measured under a 90° angle from the endocer-vical canal, as the linear distance from the outer bound-ary of the deepest invasion of cervical glands (Fig. 1b).During the study period, each woman was scanned onlyonce and cervical assessment was carried out over aperiod of 3 min to detect spontaneous dynamic changes.At least three measurements were made and the lowestvalues were used for further calculations.
The outcomes of the 327 pregnancies were followedto determine whether PTD � 34 completed wk andbetween 34 to 37 wk had occurred. In the 93% (304/327)of the cases, this information was obtained from theclinical records at our department. In the remainingcases, the information was received by personal or tele-phone contact with the participants.
Data analysisAs a part of the study design, a sample size calcu-
lation was performed (sample size calculator, MaCorrInc., Toronto, Ontario, Canada). The calculation wasdesigned to detect at least 10% difference between thesensitivity of CL measurement and QGCS score, regard-
Table 1. Scoring system for qualitative glandular cervicalscore (QGCS)
Item Score Item Score
No mucus present 0 Glandular invasion absent 0Mucus area � 20 mm2† 1 Invasion � 5 mm‡ 1Mucus area � 20 mm2 2 Invasion � 5 mm 2
† Mucus area of 20 mm2 represents the 50th percentile for ourpopulation.
‡ Glandular invasion of 5 mm represents the 50th percentile for ourpopulation.
ing the prediction of PTD in low risk population. Con-
New sonographic test for preterm delivery ● O. GRGIC et al. 335
fidence interval (CI) was determined using the confi-dence level of 95%. The program has calculated that CIis 5.9%. Using confidence level of 95% and CI of 5.9%,the program has calculated that for our population thesample size needs to be 276 participants.
Data were analyzed using SPSS version 11.0 (SPSSInc., Chicago, IL, USA). Cervical length (continuousvariable) and QGCS (composite measure of cervicalmucus area and deepest invasion of the cervical glands)were the variables used to predict outcomes. The Kol-mogorov-Smirnov test was used to determine whetherthe cervical measurements were normally distributed.
The outcomes were PTD � 34 completed wk andbetween 34 to 37 wk. Percentiles for the CL and QGCSwere analyzed with the use of chi-square tests. Thesensitivity, specificity, PPV and negative predictive
Fig. 1. (a) Measurement of cervical mucus area (six arrows). (b)Measurement of cervical glandular invasion (two arrows).
value (NPV) with 95% CI were calculated for CL of �
24 mm (fifth percentile) and QGCS of � 1 (fifth percen-tile). We calculated likelihood ratios (LR) for positiveresults and 95% CI comparing subjects at or below thefifth percentile for CL, and QGCS with those above thefifth percentile.
RESULTS
During the study period (October 2002 to April2005), 387 women were approached. The flow diagramof enrollment and exclusion is presented in Fig. 2. Finalresults were based on 327 women.
Maternal age (mean � standard deviation) at enroll-ment was 28 � 5 y (range 18 to 43 y), and gestationalage was 20 � 2 wk (range 16 to 24 wk). Cervical lengthwas 39 � 8 mm (range 10 to 65 mm), and the median ofQGCS was 2 (interquartile range 2 to 3). The distribu-tions of CL measurement and QGCS are presented inFig. 3. The incidence of PTD � 34 completed wk was2.1% (7/327) and between 34 to 37 wk it was 3%(10/327). The results of CL and QGCS of all women whodelivered �34 completed wk and between 34 to 37 wkare presented in Table 2.
The efficacy of shortened CL (�24 mm) and lowQGCS (�1) regarding the prediction of PTD � 34completed wk and between 34 to 37 wk is presented inTable 3. These values represent the fifth percentile forour population. Regarding the prediction of PTD � 34completed wk, the shortened CL and low QGCS hadsensitivity 57% versus 86% and PPV 20% versus 33%.The specificity and NPV were similar. Low QGCS incomparison with shortened CL had a twofold higher LR
Satisfying our inclusion criterian = 387
1. Length of gestation was beyond the pre-specified limits; n = 18; (4.7%) 2. Inadequate cervical image; n = 14; (3.6%) 3. Declined to participate; n = 25; (6.4 %)
Original enrollement n = 330
Lost to follow-up; n = 3; (0.7%)
Final results based on n = 327
Fig. 2. Flow diagram of enrollment and exclusions.
336 Ultrasound in Medicine and Biology Volume 32, Number 3, 2006
(23; 95%CI [12 to 43] versus 11; 95%CI [5 to 25]). Thecombination of these two tests had the highest PPV(50%) and LR (46; 95%CI [11 to 188]) for prediction of
Fig. 3. (a) Normal distribution (solid line) and frequency dis-tribution (bars) of cervical length (CL). (b) Normal distribution(solid line) and frequency distribution (bars) of qualitative
glandular cervical score (QGCS).
PTD � 34 completed wk.
Regarding the prediction of PTD between 34 to 37wk, the shortened CL and low QGCS score had sensi-tivity 20% versus 50% and PPV 10% versus 28%. LowQGCS had fourfold higher LR (12; 95%CI [5 to 28]versus 3; 95%CI [1 to 13]) for prediction of PTD be-tween 34 to 37 wk. The combination of these two testsdid not improve diagnostic accuracy.
DISCUSSION AND SUMMARY
Spontaneous preterm delivery and premature pre-term rupture of membranes are responsible for the ma-jority of preterm births (Tucker et al. 1991). Despite thefact that perinatal mortality is relatively low after 32completed wk, preterm delivery before 35 wk remainsresponsible for a high neonatal morbidity (Seubert et al.1999). Our target population was low risk primigravideawith singleton pregnancy because presently availablescreening tests for PTD in this population have lowsensitivity and low PPV. The prevalence of PTD in thispopulation is low, which makes the screening protocolsdifficult (Hasegawa et al. 1996; Hassan et al. 2000;Heath et al. 1998; Hoesli et al. 2003; Iams et al. 1996;Taipale and Hiilesmaa 1998). Primigravidae account foralmost half of all pregnancies, and we believe that by ourselection and exclusion criteria we identified the womenwhose major risk factor for PTD was primary cervicalinsufficiency (Mercer et al. 1999). As well as that, thecervical maturation process in primigravidae is differentthan in the multiparous population (Danforth 1983).
Table 2. Cervical length (CL) and qualitative glandularcervical score (QGCS) in women who delivered before 34
Abbreviations: n, number; GAE, gestational age at examination; CL,
cervical length; QGCS, qualitative glandular cervical score; GAD,gestational age at delivery.
cteristipositiv
New sonographic test for preterm delivery ● O. GRGIC et al. 337
Cervical length measurement is superior comparedto other ultrasound based screening tests (e.g., funneling)for prediction of PTD in low-risk population (Hoesli etal. 2003; Owen et al. 2001; To et al. 2001). The distri-bution of CL measurements in our study was similar tothe results published in the literature (Hasegawa et al.1996; Hassan et al. 2000; Heath et al. 1998; Iams et al.1996; Taipale and Hiilesmaa 1998). Comparing the ac-curacy for prediction of PTD � 34 completed wk, thefifth percentile (�24 mm) in our study and the fifthpercentile (�22 mm) in the study of Iams et al. (1996)showed comparable PPV (20% versus 26%). The sensi-tivity was better in our study (57% versus 30%), mainlybecause of relatively low number of women who wereincluded. The accuracy of our results regarding the pre-diction of PTD between 34 to 37 wk is comparable withthe results of Taipale and Hiilesmaa (1998).
Recently, a new sonomorphological cervical param-eter CGA was reported as a screening test for PTD(Fukami et al. 2003; Sekiya et al. 1998; Yoshimatsu et al.2002). It was classified as an “all or nothing” phenome-non (Sekiya et al. 1998). With the use of high-frequencytransducers, the CGA, if present, can be clearly visual-ized in the second trimester (Sekiya et al. 1998). Bio-chemical changes in the connective tissue associatedwith cervical maturation and increase in water contentare likely to be responsible for its disappearance, espe-cially after 32 wk (Sekiya et al. 1998). The sonographicabsence of the CGA in the second trimester indirectlyreflects accelerated cervical maturation, and could beused as a screening test for PTD (Fukami et al. 2003).Fukami et al. (2003) assessed CL and presence or ab-sence of CGA (quantitative assessment) regarding theprediction of PTD in low-risk population. The absence of
Table 3. Shortened cervical length (�24 mm) and lpreterm delivery before 34 comp
PC
PTD �34 Completed Weeks
SENS% (n)
[95% CI]
SPEC% (n)
[95% CI]
PPV% (n)
[95% CI]
NPV% (n)
[95% CI]
CL �24 mm 57(4/7)[20–88]
95(304/320)[92–97]
20(4/20)[7–44]
99(304/307)[97–100]
QGCS �1 86(6/7)[42–99]
96(308/320)[93–98]
33(6/18)[14–59]
100(308/309)[98–100]
CL � QGCS�24mm � �1
43(3/7)[12–80]
99(317/320)[97–100]
50(3/6)[14–86]
99(317/321)[97–100]
Abbreviations: PTD, preterm delivery; CL, cervical length; QGCS, qlength and low qualitative glandular cervical score; PC, predictive charaNPV, negative predictive value; n, number, LR�, likelihood ratio for
CGA had better sensitivity (75% versus 50%) and PPV
(55% versus 8%) then CL measurements regarding theprediction of PTD before 32 wk (Fukami et al. 2003).Comparing the accuracy of absence of CGA with CLmeasurements regarding the prediction of PTD between32 to 37 wk, the differences in sensitivity and PPV werenot statistically significant (Fukami et al. 2003). How-ever, a qualitative assessment of CGA was not carriedout.
The detection rate of CGA in our study was morethan 99% (325/327), and these results are in accordancewith published reports (Fukami et al. 2003; Sekiya et al.1998).
As the proportion of women with absent CGA in thesecond trimester is low, and there is no data about theCGA in earlier gestation (before 16 wk), we introducedqualitative assessment (Sekiya et al. 1998). The param-eters used in our assessment were: invasion of cervicalglands, represented by the infiltration of cervical mucosainto the cervical stroma and cervical mucus area, whichcould be a direct sign of glandular activity. We believethat the disappearance of CGA is a gradual process thatcan be assessed by QGCS. To support this, low QGCS(�1) was detected in six women who delivered �34completed wk, and in five women who delivered be-tween 34 to 37 wk. These findings probably indicate thatthe process of cervical maturation started earlier thanexpected.
The principal goal of different sonographic screen-ing tests for PTD in the low-risk population is improve-ment of sensitivity and PPV. We found that the combi-nation of gradual disappearance of CGA assessed byQGCS and CL measurement might improve screeningperformance for prediction of PTD � 34 completed wk.The women who tested positive for both of these screen-
alitative glandular cervical score (�1) in predictingeeks and between 34–37 weeks
ing tests had the highest LR for PTD � 34 completed
338 Ultrasound in Medicine and Biology Volume 32, Number 3, 2006
wk, twofold higher than QGCS and fourfold higher thanCL measurement taken individually. For prediction ofPTD between 34 to 37 wk, LR of combined CL mea-surement and QGCS was worse than QCGS assessmentalone. This might be explained by the fact that thecervical shortening and gradual disappearance of CGAcould be two independent physiological processes ofcervical maturation.
We are aware that our sample size is relatively lowcompared to the other studies assessing shortened CLregarding the prediction of PTD. Our goal was not todegrade the CL measurement; it remains the best sono-graphic screening test regarding the prediction of PTD inthe low-risk population. However, the qualitative sono-graphic analysis of cervical glands may have some valuein addition to CL measurement for the prediction of PTDin the low-risk population. This observation could havesome implications for better understanding of cervicalphysiology in pregnancy, and may be of some value indesigning study protocols for further investigations.
REFERENCES
Berghella V, Tolosa JE, Kuhlman K et al. Cervical ultrasonographycompared with manual examination as a predictor of preterm de-livery. Am J Obstet Gynecol 1997;177:723–730.
Burger M, Weber-Rossler T, Willmann M. Measurement of the preg-nant cervix by transvaginal sonography: an interobserver study andnew standards to improve the interobserver variability. UltrasoundObstet Gynecol 1997;9:188–193.
Danforth DN. The morphology of the human cervix. Clin ObstetGynecol 1983;26:7–13.
Fukami T, Ishihara K, Sekiya T, Araki T. Is transvaginal ultrasonog-raphy at mid-trimester useful for predicting early spontaneouspreterm birth? J Nippon Med Sch 2003;70(3):135–140.
Gomez R, Galasso M, Romero R et al. Ultrasonographic examinationof the uterine cervix is better than cervical digital examination as apredictor of the likelihood of premature delivery in patients withpreterm labor and intact membranes. Am J Obstet Gynecol 1994;171(4):956–964.
Hasegawa I, Tanaka K, Takahashi K et al. A prospective clinical studyfor the prediction of preterm delivery in a low risk population. J
Matern Fetal Invest 1996;6:148–151.
Hassan SS, Romero R, Berry SM et al. Patients with an ultrasono-graphic cervical length � 15 mm have nearly a 50% risk of earlyspontaneous preterm delivery. Am J Obstet Gynecol 2000;182:1458–1467.
Heath VC, Southall TR, Souka AP, Elisseou A, Nicolaides KH. Cer-vical length at 23 wk of gestation: prediction of spontaneouspreterm delivery. Ultrasound Obstet Gynecol 1998;12:312–317.
Hoesli I, Tercanli S, Holzgreve W. Cervical length assessment byultrasound as a predictor of preterm labor—is there a role forroutine screening? Br J obstetrics and gynecology (BJOG): aninternational journal of obstetrics and gynecology 2003;110 Suppl20:61–65.
Iams JD, Goldenberg RL, Meis PJ et al. The length of the cervix andthe risk of spontaneous premature delivery. N Engl J Med 1996;334:567–572.
Matijevic R, Grgic O. Clinical examination and transvaginal sonograpyin mid trimester as potential screening tests for preterm labor.Preliminary results on low risk population. J Matern Fetal NeonatalMed 2004;16 Suppl 1:48.
Mercer BM, Goldenberg RL, Moawad AH et al. The Preterm Predic-tion Study: effect of gestational age and cause of preterm birth onsubsequent obstetric outcome. Am J Obstet Gynecol 1999;181:1216–1221.
Owen J, Yost N, Berghella V et al. Mid-trimester endovaginal sonog-raphy in women at high risk for spontaneous preterm birth. J theAmerican Medical Association (JAMA) 2001;286:1340–1348.
Sekiya T, Ishihara K, Yoshimatsu K et al. Detection rate of the cervicalgland area during pregnancy by transvaginal sonography in theassessment of cervical maturation. Ultrasound Obstet Gynecol1998;12:328–333.
Seubert DE, Stetzer BP, Wolfe HM, Treadwell MC. Delivery of themarginally preterm infant: what are the minor morbidities? Am JObstet Gynecol 1999;181:1087–1091.
Taipale P, Hiilesmaa V. Sonographic measurement of uterine cervix at18–22 weeks gestation and the risk of preterm delivery. ObstetGynecol 1998;92:902–907.
To MS, Skentou C, Liao AW, Cacho A, Nicolaides KH. Cervicallength and funneling at 23 weeks of gestation in the prediction ofspontaneous early preterm delivery. Ultrasound Obstet Gynecol2001;18:200–203.
Tucker JM, Goldenberg RL, Davis RO et al. Etiologies of preterm birthin an indigent population: is prevention a logical expectation?Obstet Gynecol 1991;77:343–347.
Zorzoli A, Soliani A, Perra M et al. Cervical changes throughoutpregnancy as assessed by transvaginal sonography. Obstet Gynecol1994;84:960–964.
Yoshimatsu K, Sekiya T, Ishihara K et al. Detection of the cervicalgland area in threatened preterm labor using transvaginal sonogra-phy in the assessment of cervical maturation and the outcome of
