MbmJARE

Ov

Stc

Rwt�

Cm

Tswpbr

F(CTPWTHHBL

P

R

RB

TC

0

SMFM Papers www.AJOG.org

ulticenter randomized trial of cerclage for pretermirth prevention in high-risk women with shortenedidtrimester cervical length

ohn Owen, MD; Gary Hankins, MD; Jay D. Iams, MD; Vincenzo Berghella, MD; Jeanne S. Sheffield, MD;nnette Perez-Delboy, MD; Robert S. Egerman, MD; Deborah A. Wing, MD; Mark Tomlinson, MD;ichard Silver, MD; Susan M. Ramin, MD; Edwin R. Guzman, MD; Michael Gordon, MD; Helen Y. How, MD;ric J. Knudtson, MD; Jeff M. Szychowski, PhD; Suzanne Cliver, MSPH; John C. Hauth, MD

gcm

Ctp3

K

BJECTIVE: The objective of the study was to assess cerclage to pre-ent recurrent preterm birth in women with short cervix.

TUDY DESIGN: Women with prior spontaneous preterm birth lesshan 34 weeks were screened for short cervix and randomly assigned toerclage if cervical length was less than 25 mm.

ESULTS: Of 1014 women screened, 302 were randomized; 42% ofomen not assigned and 32% of those assigned to cerclage delivered less

han 35 weeks (P � .09). In planned analyses, birth less than 24 weeks (P

idtrimester cervical length. Am J Obstet Gynecol 2009;201:375.e1-8.

smpas

roup. There was a significant interaction between cervical length and cer-lage. Birth less than 35 weeks (P � .006) was reduced in the less than 15m stratum with a null effect in the 15-24 mm stratum.

ONCLUSION: In women with a prior spontaneous preterm birth lesshan 34 weeks and cervical length less than 25 mm, cerclage reducedreviable birth and perinatal mortality but did not prevent birth less than5 weeks, unless cervical length was less than 15 mm.

ey words: cerclage, cervical length, prior preterm birth, vaginal

.03) and perinatal mortality (P� .046) were less frequent in the cerclage sonography

ite this article as: Owen J, Hankins G, Iams JD, et al. Multicenter randomized trial of cerclage for preterm birth prevention in high-risk women with shortened

aimln

b

he role of cervical cerclage to pre-vent preterm birth is controver-

ial.1,2 Originally proposed for use inomen with recurrent midtrimesterregnancy loss that was unaccompaniedy bleeding, contractions, infection, oruptured membranes,3 cerclage has sub-

rom the Departments of Obstetrics and GynDrs Owen and Hauth and Ms Cliver); Univerenter, Columbus, OH (Dr Iams); Jefferson Mexas Southwestern Medical Center, Dallas, Terez-Delboy); University of Tennessee Healting); Northwestern Perinatal Center, Portla

exas Medical School at Houston, Houston, Tall, USAF Medical Center, San Antonio, TXealth Sciences Center, Oklahoma City, OK (irmingham, AL (Dr Szychowski). Dr Gordoouisville, KY.

resented in abstract form at the Annual Meetin

eceived July 10, 2009; revised July 29, 2009; a

eprints: John Owen, MD, Department of Obsteldg., Birmingham, AL 35249-7333.

his study was supported in part by Grants 5K2hild Health and Development.

002-9378/free • © 2009 Published by Mosby

For Editors’ Commentary, see Table

equently been more broadly recom-ended for women with a history of

reterm birth, especially if the gestationalge at birth was less than 26 weeks.4 Ultra-ound studies5,6 showing that the cervix

logy at the following institutions: University oof Texas Medical Branch, Galveston, TX (Dr

ical College of Thomas Jefferson University, PDr Sheffield); College of Physicians and Surgcience Center, Memphis, TN (Dr Egerman); U, OR (Dr Tomlinson); Northshore Medical GDr Ramin); Saint Peter’s University Hospital

r Gordon); University of Cincinnati, CincinnKnudtson); and the Department of Biostatistcurrently practicing in San Antonio, TX. Dr H

the Society for Maternal-Fetal Medicine, San D

pted Aug. 14, 2009.

s and Gynecology, University of Alabama at Birm

D43314-5 (to J.O.) and U01 HD039939-05 from

. • doi: 10.1016/j.ajog.2009.08.015

ontents

OCTOBER 2009 Americ

ppeared to shorten without contractionsn women destined for preterm birth led

any to consider cerclage as prophy-axis,7-9 but several randomized trials haveot supported this practice.10-12

Althuisius et al13 observed a significantenefit in a small clinical trial of women

labama at Birmingham, Birmingham, ALnkins); The Ohio State University Medical

ladelphia, PA (Dr Berghella); University ofs, Columbia University, New York, NY (Drversity of California, Irvine, Irvine, CA (Drp, Evanston, IL (Dr Silver); University ofw Brunswick, NJ (Dr Guzman); Wilford

OH (Dr How); University of OklahomaUniversity of Alabama at Birmingham,is currently with Norton Health Care,

, CA, Jan. 26-31, 2009.

gham, 619 19th St. South, 458 Old Hillman

e Eunice Kennedy Shriver National Institute of

eco f Asity Haed hi

X ( eonh S nind rouX ( , Ne(D ati,Dr ics,n is ow

g of iego

cce

tric in

4 H Th

, Inc

of C

an Journal of Obstetrics & Gynecology 375.e1

wcfiocahtpt3sfggmlsltgtt

atpvpotw(

rttpipaw

MTpcartsptc

mtctiprw

pccniptcfWncfdwt

cagtmtamarcprtftsp

fvstattdee

cmpaslosnatclsws

tmwatewwtltuis

wecinvslwtchqspscbpc

c

SMFM Papers www.AJOG.org

3

hose history or symptoms suggestedervical insufficiency; preterm birth be-ore 35 weeks of gestation was observedn 44% of the no-cerclage group vs nonef the women who were assigned to re-eive cerclage (P � .002). Larger tri-ls10,11 included women with variousistoric risk factors for spontaneous pre-erm birth: Rust et al10 observed rates ofreterm birth before 34 weeks of gesta-ion in 35% of cerclage-group women vs6% of controls. Berghella et al11 also ob-erved similar rates of preterm birth be-ore 35 weeks of gestation, regardless ofroup assignment: 45% in the cerclageroup vs 47% in controls. Finally, a largeultinational trial12 enrolled unse-

ected, but mostly low-risk women withhortened cervical length of 15 mm oress and also found no significant reduc-ion in preterm birth before 33 weeks ofestation in women randomly assignedo treatment with cerclage (22%) vs con-rols (26%).

More recently a patient-level meta-nalysis14 of these 4 randomized cerclagerials uncovered a relationship betweenregnancy history and cerclage: inter-ention was effective only in singletonregnancies (there was significant harmbserved in women with a multiple ges-ation), and it was especially beneficial inomen who had a prior preterm birth

adjusted odds ratio, 0.6).Thus, significant controversy remains

egarding appropriate candidate selec-ion for cerclage.15,16 We hypothesizedhat cerclage would reduce the rate ofreterm birth before 35 weeks’ gestation

n women with a prior early spontaneousreterm birth before 34 weeks’ gestationnd whose midtrimester cervical lengthas less than 25 mm.

ATERIALS AND METHODShis randomized controlled trial waserformed by a consortium of 15 USlinical centers between January 2003nd November 2007. Healthy multipa-ous women carrying a singleton gesta-ion who enrolled for prenatal care werecreened to identify those with at least 1rior spontaneous preterm birth be-ween 170/7 and 336/7 weeks’ gestation,

onfirmed by a review of the patient’s d

75.e2 American Journal of Obstetrics & Gynecolo

edical records. When efforts to retrievehe records of the prior birth were unsuc-essful, we accepted women as eligible ifhe events surrounding the prior birthncluded spontaneous causes such asreterm labor or preterm membraneupture, and the reported birth weightas less than 2 kg.Exclusion criteria were fetal anomaly,

lanned history–indicated cerclage for alinical diagnosis of cervical insuffi-iency, and clinically significant mater-al-fetal complications (eg, fetal red cell

soimmunization, treated chronic hy-ertension, insulin-dependent diabetes)hat would increase the risk of an indi-ated preterm birth and potentially con-ound the primary study outcome.

omen with cerclage in a prior preg-ancy were not excluded if review indi-ated that the cerclage had been placedor an indication other than classicallyefined cervical insufficiency. Qualifyingomen were invited to enroll in the ul-

rasound screening phase of the study.Gestational age was established by a

ertain last menstrual period (if avail-ble), confirmed by standard sono-raphic biometric measurements at lesshan 20 weeks’ gestation. If a certain last

enstrual period was not reported, ges-ational age was defined using the earliestvailable sonographic biometric infor-ation, and a second-trimester fetal an-

tomic assessment was performed toule out structural anomalies. The con-eption date was used for women whoseregnancies were conceived by assistedeproductive techniques. As part of rou-ine obstetric care, women were screenedor Neisseria gonorrhoeae and Chlamydiarachomatis, and treatment was pre-cribed for those who were cultureositive.Fifty-six sonologists underwent a uni-

orm certification process by a single in-estigator (J.O.) to ensure uniformity inonographic equipment, measurementechnique, completion of study forms,nd adherence to protocol. Specifics ofhis sonographic evaluation based on theechnique of Iams have been previouslyescribed.17 Briefly, the cervical length atach visit was measured along a closedndocervical canal, in which minimal

egrees of apparent dilation (ie, echolu- s

gy OCTOBER 2009

ency along the entire canal) less than 5m were considered closed. Fundal

ressure was also applied for 30 secondss a provocative maneuver, and eachcan included an evaluation period of ateast 5 minutes to detect spontaneouslyccurring cervical shortening. Thehortest cervical length for each exami-ation that clearly displayed the internalnd external cervical os with equivalenthickness of the anterior and posteriorervix was recorded as the cervicalength, regardless of whether the mea-urement was obtained with pressure oras the result of spontaneous dynamic

hortening.Eligible women consented to serial

ransvaginal ultrasound examinations toeasure their cervical length, the first ofhich was scheduled in the gestational

ge window of 160/7 to 216/7 weeks’ ges-ation. Follow-up scans were scheduledvery 2 weeks unless the cervical lengthas observed to be 25-29 mm, afterhich the scan frequency was increased

o every week. Women with a cervicalength that remained at least 25 mm byhe final sonographic evaluation, sched-led to be no later than 226/7 weeks, were

neligible for randomization and re-umed their obstetric care.

If on any evaluation the cervical lengthas less than 25 mm, the woman became

ligible for randomization. Informedonsent was then obtained for central-zed random assignment to cerclage oro cerclage. Because the cerclage inter-ention was not masked, managing phy-icians might infer that the cervicalength was less than 25 mm, but theyere otherwise masked to the results of

he sonographic evaluations except inases of complete placenta previa, oligo-ydramnios, or fetal death.17 At theualifying cervical length evaluation, aterile speculum examination was alsoerformed to rule out acute cervical in-ufficiency, which we defined as a cervi-al dilation of at least 2 cm with mem-ranes visible. In these cases, managinghysicians were notified, and women be-ame ineligible for randomization.

Consenting women assigned to theerclage intervention group were to be

cheduled for their surgery within 96

hDscltapmccrstd

ssptoostts

tasdd

tOcipfnwtcpscoetsg

ca

AaTlamcbsdtAt3t

dmwpidbdact

fodtlhtlptm.tlw

bqm21brdbg

wptsa

ftowwWfKtacow

fbmpFoa

ROmbusptscb(statmaalyctcl

www.AJOG.org SMFM Papers

ours of the qualifying scan, and a Mc-onald procedure3 with nonabsorbable

uture was the cerclage technique ofhoice. The use of perioperative prophy-actic antibiotics and tocolytic medica-ions was not specified in the protocolnd left to the discretion of the managinghysicians. Postrandomization patientanagement was similar in both cer-

lage and no-cerclage groups and in-luded the recommendation for pelvicest, described as abstinence from anyexual activity involving penetration ofhe vagina; no use of tampons; and noouching.Recommended physical activity re-

trictions consisted of no prolongedtanding longer than 4 hours; no heavyhysical work involving lifting morehan 20 pounds or straining; exercisenly in moderation with no-impact aer-bics or other activity that involvestraining or Valsalva, such as weightraining; and avoidance of any activityhat brings on symptoms of pelvic pres-ure or discomfort.

Women were also educated regardinghe signs and symptoms of preterm labornd preterm membrane rupture and in-tructed to report any changes in vaginalischarge, vaginal fluid, bleeding, or ab-ominal pain to their care providers.Research nurses at each center main-

ained weekly contact with participants.therwise, management was directed by

linical practice at each center. Womenn the no-cerclage group could receive ahysical examination-indicated cerclage

or acute cervical insufficiency diag-osed on clinical examination, whereasomen who had undergone cerclage as

heir trial intervention could undergoerclage revision if clinically indicated;ostrandomization transvaginal ultra-ound information was not utilized forlinical decision making. In the absencef pregnancy complications requiringarlier removal (eg, chorioamnion rup-ure, labor, hemorrhage), the cerclageuture was removed at 37 weeks’estation.The protocol and consent forms re-

eived local institutional review board

pproval at all centers. b

ssessment of outcomend statistical analysishe primary study outcome was birth at

ess than 35 weeks’ project gestationalge. From a previous report,17 we esti-ated that 57% of women in the no-cer-

lage group would experience a pretermirth before 35 weeks’ gestation. Thetudy was designed to have 80% power toetect a 30% reduction in the rate of pre-erm birth or to an absolute rate of 40%.llowing also for a maximum 10% lost-

o-follow-up rate, we planned to enroll00 women in the randomized interven-ion trial.

Because of previous observationsemonstrating a preponderance ofidtrimester births in these high-riskomen with shortened cervical length,18

lanned secondary outcomes of interestncluded the rates of birth less than 7ays from randomization; previableirth (�24 weeks); and perinatal death,efined as either a stillbirth orpostnatal death prior to hospital dis-

harge. We also planned to evaluate pre-erm birth less than 37 weeks.

Because cervical length as a surrogateor cervical competence is believed toperate on a continuum19 with a well-ocumented inverse relationship be-ween shortest midtrimester cervicalength and the risk of preterm birth,17 wead also hypothesized an interaction be-ween cerclage efficacy and cervicalength at randomization. Thus, welanned an analysis to assess the interac-ion between cervical length and treat-

ent, and if found significant at the P �10 level, associated analyses similar tohe primary aims and within cervicalength strata (�15 mm vs 15-24 mm)ould be performed.Randomization in predetermined

locks was stratified by each center andualifying cervical length less than 20m vs 20-24 mm. Early in the trial (May

003), the results of a randomized trial of7-alpha-hydroxyprogesterone caproateecame available.20 In response to thiseport, the steering committee and an in-ependent data and safety monitoringoard recommended that the use of pro-esterone for preterm birth prevention

e an option for study participants. This s

OCTOBER 2009 Americ

as included in the informed consentrocess, and an additional randomiza-ion stratum, reflecting the woman’stated intent to use progesterone, wasdded.

Intergroup comparisons were per-ormed using the principle of intent toreat. The primary study outcome andther categorical variables were analyzedith �2, whereas continuous variablesere analyzed using a Student t test orilcoxon rank-sum test. Treatment dif-

erences in time to birth were assessed byaplan-Meier curves and the log-rank

est. Multivariable logistic regressionnd Cox proportional hazard modelsonsidered possible confounders for theutcomes of preterm birth less than 35eeks and time to birth respectively.A single interim analysis was per-

ormed after half the planned sample hadeen randomized (yielding approxi-ately one third of the planned 300 with

regnancy outcomes) using O’Brien-leming boundaries with critical valuesf P � .0064 at the interim assessmentnd P � .0498 for the final assessment.

ESULTS

f the 1044 women who were deter-ined to have a qualifying prior preterm

irth, 1014 (99%) were consented andnderwent their initial sonographic as-essment of cervical length. Review ofrior pregnancy information indicatedhat of these 1014, 831 (82%) enteredcreening after medical record reviewonfirmed a qualifying prior pretermirth. From this cohort, we observed 31831%) who experienced cervical lengthhortening less than 25 mm. Sixteen pa-ients did not consent to randomization,nd 302 (95%) were randomly assignedo no-cerclage or cerclage groups. Pri-

ary outcome information was avail-ble for all 153 in the no-cerclage groupnd for 148 of 149 in the cerclage group,eaving a total of 301 women in the anal-sis (Figure 1). Only 1 patient was ex-luded from randomization because ofhe diagnosis of acute cervical insuffi-iency at the randomization visit. Se-ected baseline characteristics of the

tudy population are shown in Table 1,

an Journal of Obstetrics & Gynecology 375.e3

O

SMFM Papers www.AJOG.org

3

FIGURE 1CONSORT flow diagram

16 exclusions 13 declined randomization 2 ineligible at randomization visit 1 withdrew from trial

302 randomized

318 observed cervical length shortening <25 mm

7035 not eligible 691 declined participation

8770 initially screened for eligibility

30 exclusions 16 ineligible on further review 14 withdrew from trial

1044 met initial criteria and consented

1014 women began ultrasound screening

663 cervical length ≥25 mm 33 exclusions 21 lost or unable to contact 9 withdrew from trial 3 became ineligible

149 assigned to cerclage group 138 received assigned treatment 3 cerclage contraindication 8 declined to undergo surgery 1 emergent cerclage revision

1 unknown outcome

153 assigned to no-cerclage group 139 received no cerclage 10 received emergent cerclage 4 received off-protocol cerclage

148 analyzed 153 analyzed

wen. Cerclage for preterm birth prevention in shortened midtrimester cervical length. Am J Obstet Gynecol 2009.

75.e4 American Journal of Obstetrics & Gynecology OCTOBER 2009

sb

gnpttpimtcpdcpdt

lsg(ipvbg(

dltp0flTc�mame9fi0

M(btpsm

slpa.tl

ma0

w3i

www.AJOG.org SMFM Papers

howing the 2 groups to be wellalanced.Compliance with the intervention was

ood; a total of 14 women assigned to theo-cerclage group (9.1%) underwent therocedure, 4 solely at the discretion ofheir managing physicians (off-protocolreatment crossover), whereas 10 werelaced for a diagnosis of acute cervical

nsufficiency (protocol-sanctioned treat-ent crossover), confirmed by review of

he maternal records. Eleven women in theerclage group (7.4%) did not receive thelanned intervention: 8 declined to un-ergo surgery, whereas 3 procedures wereontraindicated because of obstetric com-lications (intraamniotic infection, fetaleath, and cervicitis) and were canceled byhe managing physicians.

The primary outcome of preterm birthess than 35 weeks’ gestation was ob-erved in 32% of women in the cerclageroup vs 42% in the no-cerclage groupodds ratio [OR], 0.67, 95% confidencenterval [CI], 0.42–1.07; P � .09). As de-icted in Figure 2, the Kaplan-Meier sur-ival analysis, considering the time toirth (ie, duration of gestation), sug-ested an overall benefit from cerclagelog-rank test, P � .053).

The addition of cervical length at ran-omization as a continuous variable to a

ogistic regression model strengthenedhe association between cerclage andreterm birth less than 35 weeks (OR,.60; 95% CI, 0.37– 0.98; P � .04). Weurther evaluated the effect of cervicalength at randomization in the 2 strata.he interaction between randomizationervical length strata less than 15 mm (n

64) vs 16-24 mm (n � 237) and treat-ent was significant (P � .03). Stratified

nalyses indicated that in the less than 15m stratum, there was a significant ben-

fit from cerclage assignment (OR, 0.23;5% CI, 0.08 – 0.66; P � .006) vs a nullnding in the 15-24 mm stratum (OR,.84; 95% CI, 0.49 –1.4; P � .52).As depicted in Figure 3, the Kaplan-eier graph and associated log-rank test

P � .024) demonstrated a significanteneficial effect of cerclage in the lesshan 15 mm stratum. Similarly, a Coxroportional hazards model demon-trated that the women in the less than 15

m cervical length stratum who were as- a

igned to cerclage had a significantlyower hazard for an earlier birth as com-ared with the no-cerclage group (haz-rd ratio, 0.57; 95% CI, 0.34 – 0.95; P �03). As observed in the above-men-ioned logistic regression analysis, the re-ationship between cerclage assignment

TABLE 1Baseline characteristics for 301 woassigned to cerclage or to no-cercl

Characteristic

Race/ethnicity, n (%)a..........................................................................................................

Black (non-Hispanic)..........................................................................................................

White (non-Hispanic)..........................................................................................................

Hispanic..........................................................................................................

Asian..........................................................................................................

Other...................................................................................................................

Marital status, n (%)..........................................................................................................

Single/never married..........................................................................................................

Married..........................................................................................................

Divorced..........................................................................................................

Widowed..........................................................................................................

Other...................................................................................................................

Cigarette use, n (%)...................................................................................................................

Any drug abuse, n (%)...................................................................................................................

Cervicovaginal microbiology, n (%)..........................................................................................................

Chlamydia..........................................................................................................

Gonorrhea...................................................................................................................

�1 prior induced abortions, n (%)...................................................................................................................

Prior cerclage, n (%)...................................................................................................................

Maternal age, y...................................................................................................................

Body mass index, kg/m2

...................................................................................................................

Number of prior births, n...................................................................................................................

Years of education, n...................................................................................................................

Gestational age of qualifying birth, wk...................................................................................................................

Weeks of gestation at first vaginal sonogram,...................................................................................................................

Cervical length at first vaginal sonogram, mm...................................................................................................................

Weeks of gestation at randomization, wk...................................................................................................................

Cervical length at randomization, mm...................................................................................................................

Total number of vaginal sonograms, n...................................................................................................................

Plus-minus values are means and 1 SD.a Race and ethnic group are self-reported; b Median and inter

Owen. Cerclage for preterm birth prevention in shortened

nd pregnancy duration in the 15-24 r

OCTOBER 2009 Americ

m stratum in the survival analysis waslso null (hazard ratio, 0.84; 95% CI,.65–1.09; P � .20).When the progesterone-use stratumas introduced, only 10 of the eventual02 women (3.3%) had been random-zed. Of the subsequent 292, 117 were

en randomlye groups

No cerclage(n � 153)

Cerclage(n � 148)

..................................................................................................................

93 (61) 80 (54)..................................................................................................................

28 (18) 25 (16.9)..................................................................................................................

17 (11) 27 (18.2)..................................................................................................................

0 (0) 1 (0.7)..................................................................................................................

15 (9.8) 15 (0.1)..................................................................................................................

..................................................................................................................

99 (65) 85 (57)..................................................................................................................

42 (27) 49 (33)..................................................................................................................

10 (6.5) 13 (8.8)..................................................................................................................

1 (0.7) 0 (0.0)..................................................................................................................

1 (0.7) 1 (0.7)..................................................................................................................

30 (20) 24 (16)..................................................................................................................

10 (6.5) 5 (3.4)..................................................................................................................

..................................................................................................................

8 (5.2) 6 (4.0)..................................................................................................................

2 (1.3) 1 (0.7)..................................................................................................................

25 (16) 25 (17)..................................................................................................................

12 (7.8) 8 (5.4)..................................................................................................................

26.6 � 5.1 26.4 � 5.5..................................................................................................................

29.9 � 7.5 29.2 � 7.8..................................................................................................................

2 (1, 4)b 2 (1, 4)b..................................................................................................................

11.9 � 2.4 12.0 � 2.8..................................................................................................................

24.9 � 4.7 24.4 � 4.9..................................................................................................................

17.4 � 1.4 17.4 � 1.2..................................................................................................................

29.5 � 12.9 28.5 � 12.7..................................................................................................................

19.5 � 2.0 19.4 � 1.9..................................................................................................................

19.5 � 5.3 18.6 � 6.3..................................................................................................................

2 (1, 4)b 2 (1, 4)b..................................................................................................................

e range.

trimester cervical length. Am J Obstet Gynecol 2009.

mag

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

.........

wk.........

.........

.........

.........

.........

decil

mid

andomized within the progesterone

an Journal of Obstetrics & Gynecology 375.e5

sgwwcwitttt(ctgt.tatl9

pdc

(wHwc(wcod1

totnfblt

ctr1

(eepepth

CWibtWtccmegcpce

ptlfvmamoddcmmshtscltimbtct

KgO

SMFM Papers www.AJOG.org

3

tratum: 56 were assigned to the cerclageroup and 61 to no cerclage. Of the 175ho did not plan to use progesterone, 89ere assigned to cerclage and 86 to no

erclage (P � .62). The single womanho was lost to follow-up was random-

zed both to the cerclage group and withhe intent to use progesterone. In a logis-ic regression model, the effect of the pa-ient’s plan to use progesterone on pre-erm birth less than 35 weeks was nullOR, 0.97; 95% CI, 0.6 –1.6). We also in-luded the progesterone strata in a mul-ivariable model with the interventionroup and an interaction term. The in-eraction term was not significant (P �94). The inclusion of the patient’s plano use progesterone in the model had noppreciable effect on the relationship be-ween cerclage intervention and birthess than 35 weeks (adjusted OR, 0.67;5% CI, 0.42–1.1; P � .09).Secondary perinatal outcomes are de-

icted in Table 2. Delivery less than 7ays from randomization was very un-

FIGURE 2Survival curves of women who rem

aplan-Meier survival curves indicating the proproups who remained undelivered across gestatwen. Cerclage for preterm birth prevention in shortened mi

ommon, affecting only 7 women g

75.e6 American Journal of Obstetrics & Gynecolo

2.3%), and the intergroup distributionas not significantly different (P � .72).owever, previable birth less than 24eeks occurred in 14% of the no-cer-

lage group vs 6.1% of the cerclage groupP � .03), and preterm birth less than 37eeks was also less common in the cer-

lage group (P � .01). Intergroup ratesf perinatal death were also significantlyifferent: 8.8% in the cerclage group vs6% in the no-cerclage group (P � .046).We also examined the homogeneity of

he effect of cerclage on preterm birthutcomes across the participating cen-ers with the Breslow-Day test. There waso significant heterogeneity across sites

or birth less than 35 weeks (P � .06),irth less than 37 weeks (P � .33), birth

ess than 24 weeks (P � .067), or perina-al death (P � .24).

Surgical adverse events associated witherclage placement were uncommon. Ofhe women who underwent protocol-di-ected cerclage (cerclage group, n �38), emergent cerclage (no-cerclage

ed undelivered across gestation

ons of women in the cerclage and no-cerclage

ester cervical length. Am J Obstet Gynecol 2009.

roup, n � 14), or a cerclage revision s

gy OCTOBER 2009

cerclage group, n � 1), only 2 experi-nced a reported complication: 1 experi-nced chorioamnion rupture during therocedure, and 1 experienced a postop-rative hemorrhage. There were 2 re-orted surgical anesthetic complica-ions: 1 failed spinal and 1 postspinaleadache.

OMMENTe did not observe a statistically signif-

cant benefit from cerclage in preventingirth before a gestational age of 35 weeks,he primary outcome for the trial.

hereas somewhat arbitrary, this gesta-ional age endpoint was chosen to avoidases of near-term birth, which are asso-iated with much lower rates of neonatalorbidity and only rare mortality. Nev-

rtheless, the weight of our findings sug-ests that cerclage, utilized for shortenedervical length in selected women with arior early spontaneous preterm birth,an improve pregnancy outcomes withssentially no demonstrable harm.

We believe that the most clinically im-ortant finding from this randomizedrial is the interaction between cervicalength at randomization and cerclage ef-ectiveness. The risk of prematurity is in-ersely proportional to cervical lengtheasured with endovaginal sonography

t various times in gestation, and theidtrimester has been the focus of most

f the research in this area.1,5,17 We haveemonstrated a biologically predictable,ifferential benefit of cerclage when theervical length is very short, less than 15m. We chose a priori to examine 15m as an alternate cutoff to define

hortened cervical length because thisas been utilized by other investigatorso assess both the predictive value ofonographic cervical length6 and cer-lage effectiveness for shortened cervicalength.7,13 Still unclear are the factorshat incite pathologic cervical shorteningn these women.15 Similarly, the precise

echanism by which cerclage confers aenefit is unknown, but it may supporthe immunological barrier between thehorioamnion-extraovular space andhe vaginal microbiologic flora.21

Because of the well-known relation-

ain

ortiion.dtrim

hip between preterm birth history and

sagotwbttclmegewbsldi

ciscdbcpmwdiec

dcalysaaattpirmotss

wcpcffiup

c

ppWsrpaNsp

Kg1O

www.AJOG.org SMFM Papers

ubsequent pregnancy outcome,2 we hadlso considered the possible effect of theestational age of the prior preterm birthn cerclage efficacy with regard to therial’s primary outcome; the effect hereas null (data not shown). However,ased on prerandomization data fromhis trial, we recently reported the rela-ionship between birth history and cervi-al length.22 Women with a prior birthess than 24 weeks were significantly

ore likely to experience cervical short-ning (�25 mm) and did so at an earlierestational age than women whose earli-st prior birth occurred at 24-336/7

eeks. Thus, we conclude that, whereasirth history affects cervical length in aubsequent pregnancy, once shorteningess than 25 mm is observed, this historyoes not significantly affect the cerclage

ntervention.Possible limitations to our trial in-

lude the open treatment because blind-ng may have been possible only withham surgery. Even then, evidence of theerclage suture would be readily visibleuring a pelvic examination. However,ecause the primary and secondary out-omes were objective, the potential im-act from lack of blinding may be mini-al. The possibility of missing womenho underwent rapid shortening andelivery during the sonographic screen-

ng was a concern, but only 1 woman wasxcluded from the randomized trial be-ause of acute cervical insufficiency.

Another possible limitation was ourecision to cap the upper gestational ageutoff for screening and randomizationt 226/7 weeks of gestation, potentiallyimiting the generalizability of results be-ond this gestational age. Althoughomewhat arbitrary, we were concernedbout the possibility of cerclage-associ-ted complications at the threshold of vi-bility and the possibility of an interac-ion with other common postviabilityreatments for women with threatenedreterm birth. We recognize that other

nvestigators have extended this tempo-al window to include more of theidtrimester.10 To the extent that some

f our high-risk patients may have con-inued to experience pathologic cervicalhortening after completion of ultra-

ound screening (as evidenced by the 10

omen who later presented with acuteervical insufficiency and underwenthysical examination-indicated cer-lage) and who may also have benefittedrom earlier cerclage placement, ourndings may have underestimated thetility of the intervention in thisopulation.The finding of no interaction between

erclage and progesterone and the com-

FIGURE 3Survival curves of women whose clength at randomization was <15 m

aplan-Meier survival curves indicating the proproups, limiting the analysis to the 64 women wh5 mm.wen. Cerclage for preterm birth prevention in shortened mi

TABLE 2Secondary perinatal outcomes forassigned to cerclage or no-cerclag

Outcome

Birth �7 d from randomization, n (%)...................................................................................................................

Previable birth �24 wks, n (%)...................................................................................................................

Preterm birth �37 wks, n (%)...................................................................................................................

Perinatal death, n (%)a...................................................................................................................a One neonate in the cerclage group was lost to follow-up.

Owen. Cerclage for preterm birth prevention in shortened mid

OCTOBER 2009 Americ

lete lack of effect of progesterone onreterm birth in this trial was surprising.e purposefully added the progesterone

tratum after a large randomized trialeported a reduced rate of recurrentreterm birth in women treated with 17-lpha-hydroxy progesterone caproate.20

evertheless, in spite of that demon-trated benefit, only 39% of our partici-ants stated their intent to use progester-

ical

ons of women in the cerclage and no-cerclagecervical length at randomization was less than

ester cervical length. Am J Obstet Gynecol 2009.

women randomlyroupscerclage

� 153)Cerclage(n � 148) P value

(2.0) 4 (2.7) .72..................................................................................................................

(14) 9 (6.1) .03..................................................................................................................

(60) 66 (45) .01..................................................................................................................

(16) 13 (8.8) .046..................................................................................................................

ervm

ortiose

dtrim

301e gNo(n

3.........

21.........

91.........

25.........

trimester cervical length. Am J Obstet Gynecol 2009.

an Journal of Obstetrics & Gynecology 375.e7

oHppacpotbptm

thwo3firwptgmcsa

wwrlrcswb

dcat2tn

oiswl

ADcMSGMt

R1rG2cn3b14Saw15lp36Agd37Ngs18tsrp19Di

1BbtG1fsr21cwt1Bccws1Atd21Ttd21i21th21svO1GacO2vh22tc2Ti

SMFM Papers www.AJOG.org

3

ne for preterm birth prevention.owever, because 17-alpha-hydroxy-

rogesterone caproate has to be extem-oraneously compounded and is vari-bly covered by third-party payers, weould not control the precise form ofrogestin locally available to participantsr the gestational age at the initiation ofreatment. Moreover, the stratum wasased on only a subject’s intended use ofrogesterone at the time of randomiza-ion, not an intention to treat by the

anaging physicians.We emphasize that this screening and

reatment regimen was limited to aighly selected population of womenith a prior spontaneous preterm birthf a nonanomalous singleton at less than4 weeks of gestation, primarily con-rmed by history and review of maternalecords. We have demonstrated thatomen with a prior early spontaneousreterm birth represent a populationhat can benefit from endovaginal sono-raphic cervical assessment. We recom-end that women with this history be

onsidered for serial cervical length mea-urement at 2 week intervals, beginnings early as 16 weeks of gestation.

Our screening schedule includedeekly assessment if the cervical lengthas within 5 mm of our action point for

andomization (25-29 mm). Neverthe-ess, our findings may not be prescriptiveegarding the optimal cervical lengthutoff for cerclage for the indication ofhortened cervical length in theseomen at risk for recurrent pretermirth.In planned secondary analyses we

emonstrated improved obstetric out-omes in the form of lower rates of previ-ble birth and perinatal mortality usinghe trial’s entry cervical length cutoff of5 mm. However, we also recognize thathe beneficial effect of cerclage for preg-

ancy prolongation varies, depending G

75.e8 American Journal of Obstetrics & Gynecolo

n the degree of cervical length shorten-ng prior to 23 weeks of gestation and isignificantly more pronounced inomen with very shortened cervical

ength less than 15 mm. f

CKNOWLEDGMENTSata safety and monitoring committee in-luded: Dr Andrew Satin (Chair), Dr CoracPherson, Dr Alessandro Ghidini, Dr Rogeroll, and Heidi Maloni. We also acknowledgeeorge Howard, DrPH, Rachel Copper, RN,SN, and Robin Steele for their many contribu-

ions to the trial.

EFERENCES. Owen J, Iams JD, Hauth JC. Vaginal sonog-aphy and cervical incompetence. Am J Obstetynecol 2003;188:586-96.. Spong CY. Prediction and prevention of re-urrent spontaneous preterm birth. Obstet Gy-ecol 2007;110:405-15.. McDonald IA. Suture of the cervix for inevita-le miscarriage. J Obstet Gynecol Br Empire957;64:346-53.. Cromblehome WR, Minkooff HL, Delke I,chwarz RH. Cervical cerclage: an aggressivepproach to threatened or recurrent pregnancyastage. Am J Obstet Gynecol 1983;146:68-74.. Iams JD, Goldenberg RL, Meis PJ, et al. The

ength of the cervix and the risk of spontaneousremature delivery. N Engl J Med 1996;34:567-72.. Heath VCF, Southall TR, Souka AP, Elliseou, Nicolaides KH. Cervical length at 23 weeks ofestation: prediction of spontaneous pretermelivery. Ultrasound Obstet Gynecol 1998;12:12-7.. Heath VCF, Souka AP, Erasmus I, Gibb DMF,icolaides KH. Cervical length at 23 weeks ofestation: the value of Shirodkar suture for thehort cervix. Ultrasound Obstet Gynecol998;12:318-22.. Berghella V, Daly SF, Tolosa JE, et al. Predic-ion of preterm delivery with transvaginal ultra-onography of the cervix in patients with high-isk pregnancies: does cerclage preventrematurity? Am J Obstet Gynecol 1999;81:809-15.. Hassan SS, Romero R, Maymon E, et al.oes cervical cerclage prevent preterm delivery

n patients with a short cervix? Am J Obstet

ynecol 2001;184:1325-31. n

gy OCTOBER 2009

0. Rust OA, Atlas RO, Reed J, van Gaalen J,alducci J. Revisiting the short cervix detectedy transvaginal ultrasound in the second trimes-er: why cerclage may not help. Am J Obstetynecol 2001;185:1098-105.1. Berghella V, Odibo AO, Tolosa JE. Cerclage

or prevention of preterm birth in women with ahort cervix found on transvaginal ultrasound: aandomized trial. Am J Obstet Gynecol004;191:1311-7.2. To MS, Alfirevic Z, Heath VCF, et al. Cervicalerclage for prevention of preterm delivery inomen with short cervix: randomised con-

rolled trial. Lancet 2004;363:1849-53.3. Althuisius SM, Dekker GA, Hummel P,ekedam DJ, van Geijn HP. Final results of theervical incompetence prevention randomizederclage trial (CIPRACT): therapeutic cerclageith bed rest versus bed rest alone. Am J Ob-tet Gynecol 2001;185:1106-12.4. Berghella V, Odibo AO, To MS, Rust OA,lthuisius SM. Cerclage for short cervix on ul-

rasonography, meta-analysis of trials using in-ividual patient-level data. Obstet Gynecol005;106:181-9.5. Romero R, Espinoza J, Erez O, Hassan S.he role of cerclage in obstetric practice: canhe patient who could benefit from this proce-ure be identified? Am J Obstet Gynecol006;194:1-9.6. Alfirevic Z. Cerclage: we all know how to do

t but can’t agree when to do it. Obstet Gynecol006;107:219-20.7. Owen J, Yost N, Berghella V, et al. Mid-rimester endovaginal sonography in women atigh risk for spontaneous preterm birth. JAMA001;286:1340-8.8. Owen J, Yost N, Berghella V, et al. Canhortened mid-trimester cervical length predictery early spontaneous preterm birth? Am Jbstet Gynecol 2004;191:298-303.9. Iams JD, Johnson FF, Sonek J, Sachs L,ebauer C, Samuels P. Cervical competences a continuum: a study of ultrasonographicervical length and obstetric performance. Am Jbstet Gynecol 1995;172:1097-106.0. Meis PJ, Klebanoff M, Thom E, et al. Pre-ention of recurrent preterm birth by 17 alpha-ydroxyprogesterone caproate. N Engl J Med003;348:2379-85.1. Jones G. The weak cervix: failing to keephe baby in or infection out? Br J Obstet Gynae-ol 1998;105:1214-5.2. Szychowski JM, Owen J, Hankins G, et al.iming of mid-trimester cervical length shorten-

ng in high-risk women. Ultrasound Obstet Gy-

ecol 2009; 33:70-5.