9
Proteomic biomarkers that predict the clinical success of rescue cerclage Carl P. Weiner, MD, a, * Keun-Young Lee, MD, b Catalin S. Buhimschi, MD, c Rob Christner, PhD, d Irina A. Buhimschi, MD c Department of Obstetrics, Gynecology, and Reproductive Sciences, University of Maryland School of Medicine, Baltimore, Md, a Department of Obstetrics and Gynecology, Kangnam Sacred Heart Hospital, Hallym University, Seoul, Korea, b Department of Obstetrics, Gynecology, and Reproductive Sciences, Yale University, New Haven, Conn, c Ciphergen Biosystems, Fremont, Calif d Received for publication June 1, 2004; revised September 20, 2004; accepted October 4, 2004 KEY WORDS Incompetent cervix Cervical cerclage Surface-enhanced laser-desorption ionization Proteomic biomarker Objective: The origin of incompetent cervix is multifactorial, and the success of rescue cerclage is unpredictable. We tested amniotic fluid from women who were preparing to undergo rescue cerclage for proteomic biomarkers and correlated their presence with clinical outcome. Study design: Amniocentesis was performed to facilitate rescue cerclage in 37 consecutive women with painless dilation (>2 cm) and no detectable uterine activity for 4 hours (range, 1-24 hours) before cerclage. Thirty-nine consecutive women with a sonographically normal pregnancy and cervix who underwent amniocentesis for chromosomal testing during the same study interval at the same clinical site provided the control samples. A proteomic fingerprint was generated with the discarded sample and the Mass-Restricted score (MR score) for inflammation calculated. Peaks corresponding to free hemoglobin chains were sought as evidence of decidual hemorrhage or intra-amniotic bleeding. Results: Amniocentesis was performed at 23.5 weeks in cerclage (mean dilation, 4 cm) versus 19.5 weeks in control subjects. Cerclage subjects were delivered at 28.8 weeks; control subjects were delivered at 39.2 weeks. Thirty-two of 37 of cerclage subjects (86%) were delivered prematurely. Ten of 37 of cerclage subjects (27%), but no control subject, had a MR score that was indicative of inflammation (P ! .001). Hemoglobin peaks were present in 12 of 37 of cerclage subject (32%), but no control subjects. Among cerclage subjects, those with a MR score of 3 to 4 were delivered earlier than those with a MR score of 0 to 2 (P ! .001). Women with a MR score of 3 to 4 had a shorter latency period (days from amniocentesis to delivery; 3 days) and a shorter percentage of prolongation (1.8%) than women with a MR score of 0 to 2 (35 days; P ! .05; 17.9%; P ! .05). Women with hemoglobin had a shorter latency period (6 days) and a shorter percentage of prolongation (3.8%) than women without hemoglobin (38 days; P ! .05; 21.8%; P ! .05). Hemoglobin was present in 7 of 10 of the cerclage subjects (70%) with a MR score of 3 to 4. Women with both a MR score of 3 to 4 and hemoglobin had the shortest intervals to delivery. Oral presentation at the 25th Annual Meeting of the Society for Maternal-Fetal Medicine, February 7-12, 2005, Reno, Nev. * Reprint requests: Carl P. Weiner, MD, MBA, Department of Physiology and of Obstetrics and Gynecology and Reproductive Science, University of Maryland School of Medicine, 655 W Baltimore St, BRB-11-033, Baltimore, MD 21201-1559. E-mail: [email protected] 0002-9378/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.ajog.2004.10.588 American Journal of Obstetrics and Gynecology (2005) 192, 710–8 www.ajog.org

Proteomic biomarkers that predict the clinical success of rescue cerclage

Embed Size (px)

Citation preview

Page 1: Proteomic biomarkers that predict the clinical success of rescue cerclage

American Journal of Obstetrics and Gynecology (2005) 192, 710–8

www.ajog.org

Proteomic biomarkers that predict the clinical success ofrescue cerclage

Carl P. Weiner, MD,a,* Keun-Young Lee, MD,b Catalin S. Buhimschi, MD,c

Rob Christner, PhD,d Irina A. Buhimschi, MDc

Department of Obstetrics, Gynecology, and Reproductive Sciences, University of Maryland School of Medicine,Baltimore, Md,a Department of Obstetrics and Gynecology, Kangnam Sacred Heart Hospital, Hallym University,Seoul, Korea,b Department of Obstetrics, Gynecology, and Reproductive Sciences, Yale University, New Haven,Conn,c Ciphergen Biosystems, Fremont, Calif d

Received for publication June 1, 2004; revised September 20, 2004; accepted October 4, 2004

KEY WORDSIncompetent cervixCervical cerclageSurface-enhanced

laser-desorptionionization

Proteomic biomarker

Objective: The origin of incompetent cervix is multifactorial, and the success of rescue cerclage isunpredictable. We tested amniotic fluid from women who were preparing to undergo rescue

cerclage for proteomic biomarkers and correlated their presence with clinical outcome.Study design: Amniocentesis was performed to facilitate rescue cerclage in 37 consecutive womenwith painless dilation (>2 cm) and no detectable uterine activity for 4 hours (range, 1-24 hours)

before cerclage. Thirty-nine consecutive women with a sonographically normal pregnancy andcervix who underwent amniocentesis for chromosomal testing during the same study interval atthe same clinical site provided the control samples. A proteomic fingerprint was generated with

the discarded sample and the Mass-Restricted score (MR score) for inflammation calculated.Peaks corresponding to free hemoglobin chains were sought as evidence of decidual hemorrhageor intra-amniotic bleeding.

Results: Amniocentesis was performed at 23.5 weeks in cerclage (mean dilation, 4 cm) versus 19.5weeks in control subjects. Cerclage subjects were delivered at 28.8 weeks; control subjects weredelivered at 39.2 weeks. Thirty-two of 37 of cerclage subjects (86%) were delivered prematurely.Ten of 37 of cerclage subjects (27%), but no control subject, had a MR score that was indicative

of inflammation (P ! .001). Hemoglobin peaks were present in 12 of 37 of cerclage subject(32%), but no control subjects. Among cerclage subjects, those with a MR score of 3 to 4 weredelivered earlier than those with a MR score of 0 to 2 (P! .001). Women with a MR score of 3 to

4 had a shorter latency period (days from amniocentesis to delivery; 3 days) and a shorterpercentage of prolongation (1.8%) than women with a MR score of 0 to 2 (35 days; P ! .05;17.9%; P ! .05). Women with hemoglobin had a shorter latency period (6 days) and a shorter

percentage of prolongation (3.8%) than women without hemoglobin (38 days; P ! .05; 21.8%;P ! .05). Hemoglobin was present in 7 of 10 of the cerclage subjects (70%) with a MR score of 3to 4. Women with both a MR score of 3 to 4 and hemoglobin had the shortest intervals todelivery.

Oral presentation at the 25th Annual Meeting of the Society for Maternal-Fetal Medicine, February 7-12, 2005, Reno, Nev.

* Reprint requests: Carl P. Weiner, MD, MBA, Department of Physiology and of Obstetrics and Gynecology and Reproductive Science,

University of Maryland School of Medicine, 655 W Baltimore St, BRB-11-033, Baltimore, MD 21201-1559.

E-mail: [email protected]

0002-9378/$ - see front matter � 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.ajog.2004.10.588

Page 2: Proteomic biomarkers that predict the clinical success of rescue cerclage

Weiner et al 711

Conclusion: These findings illustrate 2 pathologic mechanisms that are associated with pretermdelivery are also associated with incompetent cervix. Either an intrauterine inflammatoryresponse or decidual hemorrhage predates surgery in one half the women whose conditionrequires rescue cerclage. The activation of either mechanism predicts cerclage failure.

� 2005 Elsevier Inc. All rights reserved.

Early mid trimester pregnancy loss that is associatedwith painless cervical dilation of undetermined causeremains a frustrating challenge. The traditional treat-ment of incompetent cervix is the placement of a pursestring suture close to the internal cervical os. Althoughtypically placed in the early second trimester electivelyafter a pregnancy loss that is consistent with incompe-tent cervix, some pregnant women are already dilatedwith either no relevant or confusing clinical histories. Inthis setting, a cerclage can still be performed in anattempt to ‘‘rescue’’ the pregnancy. Exposure of theamniotic membranes to the vagina increases the risk offailure,1 and the failure rate of rescue cerclage is higherthan after elective placement.2,3

Amniotic fluid microbial invasion occurs in O50% ofwomen with painless cervical dilation of O2 cm between14 and 24 weeks of gestation.4 In one retrospective study,Mays et al5 noted 7 of 18 women (39%) who wereconsidered candidates for rescue cerclage already hadindirect markers of inflammation in their amniotic fluid.They declined to operate and attributed their worseoutcome to inflammation rather than the lack of surgicalprocedure. However, a small, randomized trial con-ducted by Althusius et al6 concluded rescue cerclageplus indomethacin and bed rest were superior to cerclageplus bed rest alone. Thus, candidate selection for rescuecerclage and the involvement of inflammation remains anissue.

Cerclage failure manifests as either preterm labor withintact membranes or labor preceded by preterm pre-mature rupture of membranes (PPROM). Although it isnot unusual for preterm labor to be associated withPPROM, they are usually considered separate entities.7

At least 4 distinct pathophysiologic pathways are impli-cated in the triggering of preterm birth8: stress, excessivemyometrial stretch, decidual hemorrhage, and infectionor inflammation.9

Decidual hemorrhage and the resulting placentalabruption is recognized histologically in up to 60% ofspontaneous preterm deliveries, yet is associated withrecurrent vaginal bleeding in just 20%.8 Hemosiderindeposition is common, which demonstrates that thebleeding had occurred at least 24 to 48 hours beforedelivery. Hemosiderin deposition in the placental basalplate and extraplacental decidua is associated with lowermean birth weight percentiles in deliveries at !32 weeksof gestation.10,11

We have noted that there are multiple causes of in-flammation, only one of which is infection. Directanalysis of amniotic fluid is the most accurate techniquefor the detection of inflammation, even if it is not theresult of infection.12,13 Tests that are used includedamniotic fluid white blood cell and neutrophil counts,glucose concentration, Gram stain, and microbial cul-tures.14 But even in combination, these tests have poorclinical performance,15 and attempts to introduce theminto practice have failed because their sensitivities andpredictive values do not justify an invasive test. Asa result, diagnostic amniocentesis usually is not per-formed before rescue cerclage. It is performed, however,to decompress the bulging amniotic sac and facilitatecerclage placement.

Proteomics provides a snapshot of all proteins thatwere present in a biologic sample at a given moment.Surface-enhanced laser-desorption ionization (SELDI)time-of-flight (TOF) technology combines chromatog-raphy with mass spectrometry.16 It may have futureapplications for the diagnosis of ovarian17 and breastcancers18 and Alzheimer’s disease.19 In an earlier study,we described a proteomic signature profile (termed themass restricted [MR] score) using the Ciphergen Pro-teinChip (Ciphergen Biosystems, Fremont, Calif) thatidentified women with intrauterine inflammation thatwas the result of chorioamnionitis without error.20,21

The MR score is the arithmetic sum of Booleanindicators for absence (0) or presence (1) of 4 specificprotein peaks (calgranulin A and C and neutrophildefensins 1 and 2) that were identified on a hydrophobicProteinChip array. The presence of R3 indicators in theamniotic fluid of women in preterm labor stronglypredicts intra-amniotic inflammation that is identifiedwith traditional criteria and imminent preterm deliv-ery.21 The MR score is historically the first proteomicdescriptor with demonstrated diagnostic value. Theavailability of such a highly accurate diagnostic toolcould influence clinical treatment.

We hypothesize that preterm labor and incompetentcervix share pathophysiologic elements and that eitherintra-amniotic inflammation (as reflected in the MRscore) or decidual hemorrhage (as identified by thepresence of hemoglobin chains), occur in at least somewomen with incompetent cervix. We further hypothesizethat the success of rescue cerclage reflects their absence.We tested these hypotheses in a consecutive series of

Page 3: Proteomic biomarkers that predict the clinical success of rescue cerclage

712 Weiner et al

women with cervical dilation in the second trimesterresulting from incompetent cervix and no clinicalevidence of chorioamnionitis who received care ata single institution.

Material and methods

Cervical incompetence was diagnosed in 37 consecutivewomen with painless dilation (>2 cm) and no detectableuterine activity for a mean of 4 hours (range, 1-24 hours)before cerclage. In each instance, bulging but intactmembranes were visualized at the level of the externalcervical os. Each woman was afebrile and free ofsymptoms of chorioamnionitis at the time surgery. Anemergency cerclage (McDonald type purse string with 3mm Mersilene) was performed under general anesthesiaafter amnioreduction to reduce the intra-amniotic fluidpressure.22 A balloon device was used to hold themembranes above the cervix during suture placement.A cohort of 39 women with a sonographically normalpregnancy and cervix who underwent amniocentesis forchromosomal testing during the same time interval atthe same clinical site provided the control amniotic fluidsamples. All samples were centrifuged to remove cellulardebris and stored at �80(C. The samples were collectedunder a research protocol that was approved by theInstitutional Review Board at the Hallym UniversitySchool of Medicine and Kangnam Sacred Heart Hos-pital, Seoul, Korea; all women provided written in-formed consent. Each amniocentesis and cerclage wasperformed by a single investigator (K.L.).

After delivery, the samples were coded by one of theinvestigators (K.L.) and transported to the remaininginvestigators without the key code. The samples wereimported on dry ice with Centers for Disease Controlapproval and in accordance to all federal regulations.The proteomic fingerprint was generated; the MR score(0-4) was calculated, and hemoglobin peaks weresought. The results were transmitted back to the in-vestigator; the blinding was broken, and the relationshipbetween the MR score, hemoglobin chains, and clinicaloutcome of the rescue cerclage was assessed.

Protein profiling protocol for SELDI–TOF massspectrometry

The technique was described previously.20,21 Briefly, 2mL of 10-fold diluted amniotic fluid in phosphate-buffered saline solution (PBS) was placed on spots ofduplicate H4 arrays (8-spot H4 array; Ciphergen Bio-systems). Some spots were covered with PBS alone.After a 1-hour incubation in a humidified chamber, thesample was aspirated, and the spots were washedindividually with 25% aqueous acetonitrile solution,air-dried, and covered with energy absorbing molecule(EAM) matrix solution of either 1 mL of 20% saturated

a-cyano-4-hydroxycinnamic acid (CHCA) in 25% tri-fluoroacetic acid/50% acetonitrile on one array, or 2sequential applications of 0.5 mL saturated solution (in0.5% trifluoroacetic acid/50% acetonitrile) of sinapinicacid (SPA) on the other.

The arrays were read in the ProteinChip Reader(model PBS II; Ciphergen Biosystems) using the Pro-teinChip software (version 3.0; Ciphergen Biosystems).SELDI tracings were examined in 3 mass ranges: 3000 to4000 Da (CHCA), 10 to 12.5 kDa (SPA), 14 to 17 kDa(SPA). The data for the MR score peaks (P1-P4; Figure1) were extracted from the 3000 to 4000 Da (CHCA: P1and P2) and 10 to 12.5 kDa (SPA: P3 and P4) ranges.The data for the hemoglobin peaks were analyzed in the14 to 17 kDa region (SPA: P5 and P6; Figure 1).

Calculation of the MR and hemoglobin scores

The presence or absence of a peak was determinedobjectively from numeric data with an algorithm thatdefines a peak that was based on the relative difference insignal-to-noise ratio (S/N) compared with the tracingsthat were obtained from the PBS-covered spots at thecorresponding mass value. A peak was present and as-signed a value of 1 if its S/N exceeded the average S/NG 2SD from PBS spots (n = 5) at the corresponding mass.

The MR score was derived originally by non–hierarchical analyses of SELDI tracings of amnioticfluid from women with and without intra-amnioticinflammation because of bacterial infection anda data-mining algorithm that was based on the sequen-tial application of 5 Boolean logic criteria.20,21 In thepresent study, the peaks that composed the MR scorewere identified by their conspicuous aspect at or inproximity to their known respective masses: 3377.0 Daand 3448.1 Da on the CHCA tracing (corresponding toneutrophil defensin-2 and –1, respectively) and at10443.8 Da and 10834.5 Da (corresponding to calgra-nulin C and A, respectively).

The reported SwissProt masses of the alpha and betahemoglobin chains are 15126.4 and 15867.2 Da, re-spectively. We recognize that there also may be a com-ponent in the later peak of the delta (15924.29 dalton) orgamma chains (16,009 dalton) that is not distinguishedfrom the beta chain. Both alpha and beta chains werepresent in all instances where hemoglobin was detected.Analogous to the MR score, the presence or absence ofhemoglobin peaks was based objectively on the S/Nratio of the PBS tracings, and a score (hemoglobinscore) of 0 (hemoglobin chains absent) or 1 (hemoglobinchains present) was assigned for their absence orpresence, respectively.

Statistical analysis

Because surgical procedures were conducted acrossa wide range of gestational ages, we sought to remove

Page 4: Proteomic biomarkers that predict the clinical success of rescue cerclage

Weiner et al 713

Figure 1 Representative protein profiles of amniotic fluid from 5 patients who underwent rescue cerclage (cases 1-5) and onecontrol patient (Crl-6). A, CHCA; B, SPA. The arrowheads indicate the peaks that compose the MR score (P1, neutrophil defensin-2; P2, neutrophil defensin-1; P3, calgranulin C; P4, calgranulin A) or representative hemoglobin (P5 and P6, alpha and beta

hemoglobin chains). The x-axis of the tracings represents the molecular weight in daltons (Da) and kilodaltons (kDa); the y-axisrepresents normalized peak intensity. R notes a reference protein peak that was present in all (R1) or most (R2) fluid samples. PBSwas the sample diluent. The S/N ratio from the PBS tracing was used to estimate objectively the presence or absence of a peak in the

samples of amniotic fluid.

Page 5: Proteomic biomarkers that predict the clinical success of rescue cerclage

714 Weiner et al

any positive bias by normalizing the data to reflect thedegree a pregnancy was prolonged after operation. First,we calculated the ‘‘latency period’’ interval, defined as thenumber of days gestation at delivery minus the number ofdays of gestation at amniocentesis. Second, we calculatedthe ‘‘percentage prolongation of pregnancy’’ by dividingthe latency period by the number of days of gestation atamniocentesis times 100. We recognize that this calcula-tion actually underestimates the benefit of cerclage be-cause the standard gestational age that is based on the lastmenstrual period includes 2 weeks at the beginning whenthe pregnancy does not yet exist. Statistical analysesincluded testing for normality with the Kolmogorov-Smirnov test, the Student t test, the Mann Whitney test,the KruskalWallis analysis of variance on ranks followedby Dunn’s tests and survival analysis, as appropriate.Comparisons between proportions were performed withthe Fisher’s exact test. A probability value of %.05 wasassumed to indicate a significant difference amongmeans,medians, or proportions.

Results

The clinical characteristics of the study groups aresummarized in Table I. The average gestation at amnio-centesis in the rescue cerclage group was 23.5 weeks (95%CI, 22.2-24.8 weeks) compared with 19.5 weeks (95% CI,18.5-20.5 weeks) in the control group (P ! .001, t-test).The median cervical dilation at cerclage was 4 cm (range,2-9 cm). Thirty-two women (86%) in the rescue cerclagegroup were delivered preterm: 10 women (27%) were at!24 weeks of gestation, and 22 women (59%) werebetween 24 and 37 weeks of gestation. Cervical dilation atcerclage was correlated inversely to both outcome meas-ures (latency period: r = �.398; P ! .02; percent pro-longation of pregnancy: r = �.378; P = .02).

Women in the rescue cerclage group were delivered atan average of 28.8 weeks of gestation (95% CI, 26.7-30.9weeks), with a median latency period of 21 days (range,0-148 days). Identifiable postoperative complicationsthat were related to preterm delivery were intraoperativeiatrogenic rupture of membranes (2 cases), late abortion(!20 weeks of gestation, but at O12 hours aftercerclage; 3 cases), preterm contractions with intactmembranes (11 cases), fever (>38(C; 2 cases), sponta-neous PPROM (with or without preterm contractions, 9cases) or abruptio placenta (1 case).

In contrast to the cerclage group, all women in thecontrol group were delivered at term, with a meangestation of 39.2 weeks (95% CI, 38.9-39.6 weeks) anda median latency period of 143 days (range, 56-170 days).

SELDI tracing analysis

Representative SELDI tracings from 5 patients whowere treated by rescue cerclage and 1 control patient

who underwent genetic amniocentesis are given inFigure 1. The baseline noise level on the respectiveCHCA and SPA SELDI readings is illustrated by thetracing that included PBS instead of amniotic fluid. Thepeaks of the MR score (P1-P4) and those correspondingto the hemoglobin chains (P5 and P6) are shown at theirexpected masses. Two additional, conspicuous peakswere present in the mass regions of interest in most ofthe tracings (reference peaks, R1 and R2). The R1 peakwas present in all fluids that were studied. We previouslyidentified the R1 peak as a fragment of beta-2 micro-globulin.23 Its characteristic SELDI peak was used forrapid orientation to the mass axis of the SPA spectra.The R2 peak was less conspicuous than the R1 peakunder these experimental conditions and was present ononly 34 of the tracings (67.5% of cases and 23% ofcontrol subjects). Although these proportions differ(P ! .01, Fisher’s exact test), the R2 peak does notfulfill our criteria for a disease biomarker (all biomarkerpeaks must be absent in control subjects).20,21 Table IIshows the average experimental masses of the biomarkerpeaks, with Boolean indicators of 1 observed in the 76samples of amniotic fluid that were included in thisstudy.

Ten of 37 rescue cerclage patients (27%), but nocontrol subjects, had an MR score of 3 to 4 (P ! .001,Fisher’s exact). Twelve women in cerclage group hadhemoglobin peaks (hemoglobin score, 1). The womenwith hemoglobin present each showed both the alphaand beta peaks (hemoglobin score, 1). No controlpatient sample had hemoglobin peaks. Seven womenhad both an MR score of 3 to 4 and hemoglobin score of1. Thus, slightly O40% (15/37) of the women had anabnormal proteomic profile.

Excluding the 2 women with accidental intraoper-ative rupture of membranes, women with cerclage withan MR score of 3 to 4 were delivered earlier than womenwith an MR score of 0 to 2 (P ! .05, Mann Whitney).Women with an MR score of 3 to 4 had a shorterlatency period (median, 3 days; range, 0-105 days) anda shorter percentage of prolongation (median, 1.8%;range, 0%-62.8%) compared with women with scores of0 to 2 (latency period, 35 days; range, 0-148 days; P !.04 Mann Whitney; median percentage prolongation,17.9%; range, 0-116.5%; P ! .03, Mann Whitney).

Women with hemoglobin peaks were delivered earlierthat women without (latency period: hemoglobin score0; median, 38 days; range, 0-148 days; vs hemoglobinscore 1; median, 6 days; range; 0-100 days; P ! .04;percentage of prolongation: hemoglobin score 0; me-dian, 21.8%; range, 0.7%-116.5%; vs hemoglobin score1; median, 3.8%; range, 0-56.8; P = .03, Mann Whit-ney). Three of 10 women (30%) with an MR score of 3to 4 had no hemoglobin score; 5 of 12 women (42%)with a hemoglobin score had an MR !3 (P = nosignificance). There was no relationship between the

Page 6: Proteomic biomarkers that predict the clinical success of rescue cerclage

Weiner et al 715

Table I Patient, amniotic fluid and outcome characteristics

Variable Cerclage cases n = 37 Control subjects n = 39 P value

Clinical characteristics at cerclageAge (y)* 31.5 (30.5-32.6) 32.9 (31.5-34.4) .137z

Gestational age (wk)* 23.5 (22.2-24.8) 19.5 (18.5-20.5) ! .001z

Cervical dilatation (cm)y 4 (2-9) 0 (0-0) ! .001x

Amniotic fluid SELDI analysisMR score 3 or 4 (n) 10 (27%) 0 (0%) ! .001||

Hemoglobin score 1 (n) 12 (32%) 0 (0%) ! .001||

MR score 3 or 4 and hemoglobin score 1 (n) 7 (19%) 0 (0%) ! .001||

Delivery and outcome measuresGestational age at delivery (wk)* 28.8 (26.7-30.9) 39.2 (38.9-39.6) ! .001z

Birth weight (g)y 1190.0 (140.0-3700.0) 3400.0 (2106.0-4100.0) ! .001x

PPROM (n) 9 (25%) 0 (0%) ! .001||

Latency (d)y 21.0 (0.0-148.0) 143.0 (56.0-170.0) ! .001x

% Prolongation of pregnancyy 14.3 (0.0-116.5) 111.5 (27.6-149.5) ! .001x

* Data presented as mean (95% CI).y Data presented as median (range).z Student t test.x Mann-Whitney U test.|| Fisher’s exact test. Two cases with intra-operative PPROM were excluded from the cerclage group when delivery and outcome measures were

reported.

Table II Presence and experimental masses of the biomarker peaks in the 76 patients studied

Biomarker EAM Observed mass (95% CI) Peak present (n) Protein identity Calculated mass

MR scoreP1 CHCA 3374.9 (3373.9-3375.9) 14 Neutrophil defensin-2 3377.0P2 CHCA 3446.0 (3444.9-3447.1) 16 Neutrophil defensin-1 3448.1P3 SPA 10441.4 (10437.2-10445.6) 15 Calgranulin C (S100 A12) 10443.8P4 SPA 10843.64 (10829.8-10857.4) 7 Calgranulin A (S100 A8) 10834.5

Hemoglobin chainsP5 SPA 15119.81 (15113.7-15125.9) 12 Hemoglobin a-chain 15126.4P6 SPA 16134.75 (16125.8-16143.6) 12 Hemoglobin b or

Hemoglobin d-chain orHemoglobin g-chain

15867.2 or15924.3 or16009.0

EAM, Energy absorbing molecule (matrix: CHCA - a-cyano-4-hydroxycinnamic acid; SPA - sinapinic acid).

The calculated mass is the mass reported by Swiss-Prot database (www.expasy.ch). All masses are expressed in daltons.

degree of cervical dilation and the presence of hemo-globin.

The combination of the MR score and hemoglobinproved most valuable in the prediction of cerclagesuccess. Women with an MR score !3 and no hemo-globin had a median latency period of 40.5 days (range,1-148 days) compared with women with either an MRscore of 3 to 4 or hemoglobin of 1 (but not both) whosemedian latency period was 14 days (range, 0-105 days)compared with women with both an MR score of 3 to 4and hemoglobin of 1 whose median latency period was 3days (range, 0-43 days). Likewise, the percentage ofprolongation among these 3 groups declined stepwisefrom 23.3% to 9.6% to 1.9% (survival analysis using logrank chi-squared test, 7.69; P = .005; relative risk, 10.0;95% CI, 9.6-10.4; Figure 2).

Comment

The causes of preterm delivery are multiple, and theclinical distinction between preterm labor and incompe-tent cervix often is difficult. Rescue cerclage is per-formed typically after the unexpected discovery ofcervical dilation in the absence of detectable uterineactivity before viability. The success of rescue cerclage isrelatively low, certainly compared with elective cerc-lage.2,3 It is less clear how valid this comparison is andunknown whether losses occur despite the cerclage orbecause of it. Likely, the two are not mutually exclusive.We have demonstrated the existence of two pathophys-iologic mechanisms that are associated with pretermdelivery that are also associated with incompetentcervix, absent detectable contractions. Almost one half

Page 7: Proteomic biomarkers that predict the clinical success of rescue cerclage

716 Weiner et al

the women had evidence predating the rescue cerclage ofeither an intrauterine inflammation20,21 or presumablydecidual hemorrhage.10 There is a stepwise worsening ofoutcome when one or both of these pathways are active.In contrast, the absence of activity along either pathwaysuggests that the likelihood of success is high, absent anoperative complication. If these biomarkers reflect path-ways of causation, their presence suggests that cervicalincompetence is not necessarily recurrent in subsequentpregnancies.

Neutrophil defensins (alpha-defensins) belong toa family of cationic antimicrobial peptides. These keycomponents of the host-defense mechanism24 exert their

Figure 2 Scattergram illustrates the relationship between (A)latency period (days amniocentesis-to-delivery), (B) percentageof prolongation of pregnancy (latency period divided by the

gestational age (days) at amniocentesis ! 100), and MR scoreand hemoglobin (Hb) scores of the amniotic fluid samples thatwere obtained from patients before they underwent emergent

cerclage (n = 35). The 2 patients with documented intra-operative rupture of the membranes were excluded.

bactericidal activity by punching pores into bacterialmembranes.25 Elevated defensin levels are found inpatients with sepsis,26 meningitis,27 and cystic fibrosis.28

Calgranulins are members of the S100 calcium-bindingproteins that are implicated in Alzheimer’s disease,cancer, cardiomyopathy, psoriasis, rheumatoid arthritis,and other inflammatory disorders.29 There are manycauses of inflammation, and one of the limitations of thepresent study is that we do not know whether the causeof inflammation in cerclage subjects was bacterial in-fection. Such a determination might be clinically rele-vant in the future.

Similar to inflammation, evidence of decidual hem-orrhage is present in many women who are deliveredafter spontaneous preterm labor.9,10 It is suggested thatthe release of thrombin leads to myometrial activationand uterine contractions.30,31 In the present series, onequarter of the women with incompetent cervix hadhemoglobin in their amniotic fluid before the cerclage.Its uniform absence from the samples of control womenstrongly suggests that the hemoglobin that was detectedwas not the result of the amniocentesis, but ratherpredated it. Further, we have found that the simpleaddition of blood to amniotic fluid in vitro does notresult in hemoglobin being detectable on SELDI, nodoubt because of the centrifugation step. Consideringthe association between thrombin and uterine contrac-tion, it is interesting to note that these women were feltrelatively free of uterine activity. The observation that30% of women with a high MR score had no hemoglo-bin and that 42% of women with hemoglobin hada normal MR score suggests that the two pathologicpathways have separate origins but can intersect as theyact additively. How might bleeding be associated withcervical ripening? Neutrophils play an important role incervical ripening. Thrombin reportedly stimulates ma-trix metalloproteinase expression in cultured endome-trial stromal and decidual cells.32 Further, intracerebralhemorrhage is known to cause activation of bothmacrophages and matrix metalloproteinases in theabsence of bacterial infection.33

We cannot exclude the possibility that some un-known bias was introduced by the use of a control groupthat was younger in gestational age when sampled thanthe study group. However, the gestational age rangeoverlapped, and both groups that were studied sharea similar stage of myometrial quiescence (supported byabsent uterine activity) and no clinical evidence ofinfection. We do not believe it would have been ethicalto have enrolled a control group of women who werematched by gestational age for an unindicated invasiveprocedure.

The ability to identify women at risk women accu-rately and rapidly before cerclage creates an opportunityto intervene and improve outcome pharmacologically.There are several therapeutic possibilities. Antibiotics

Page 8: Proteomic biomarkers that predict the clinical success of rescue cerclage

Weiner et al 717

alone have failed to either improve cerclage results orprolong pregnancy in women with preterm labor andintact membranes. They have also failed to preventdelivery and maintain fetal health in animal models ofpreterm labor that results from inflammation. Anti-inflammatory or antioxidant therapies may providea pathophysiologically directed therapy. For example,antioxidant treatment with N-acetyl cysteine preventspreterm delivery and rescues the fetuses from inflamma-tion-induced wastage in a model of preterm delivery thatresulted from Escherichia coli lipopolysaccharide.34

Treatment with an anti-inflammatory cytokine, inter-leukin-10, has a similar effect.35 It may be possible totreat women with evidence of inflammation beforeoperation with a combination of N-acetyl cysteine orother anti-inflammatory agents perhaps coupled witha broad-spectrum antibiotic agent that achieves highconcentrations in the conceptus. Such treatment willrequire the rapid response time provided by SELDI(!1 hour).

In conclusion, the mid trimester loss of an otherwisehealthy fetus after cervical dilation, absent detectableuterine activity, shares with preterm labor an associationwith intrauterine inflammation and presumed decidualhemorrhage. Although the phrase ‘‘cervical incompe-tence’’ suggests a structural weakness or deficiency, thisinvestigation demonstrates that there is more to cervicalincompetence than just physical strength. The simulta-neous assessment of a number of distinct biomarkerswith a convenient platform such as SELDI allows forthe rapid and accurate identification of high-riskwomen, creating the possibility for new therapeuticinterventions to reduce the failure rate of rescuecerclage.

References

1. Yip SK, Fung HY, Fung TY. Emergency cervical cerclage: a study

between duration of cerclage in situ with gestation at cerclage,

herniation of forewater, and cervical dilation at presentation. Eur J

Obstet Reprod Biol 1998;78:63-7.

2. Charles D, Edwards WR. Infectious complications of cervical

cerclage. Am J Obstet Gynecol 1981;141:1065-71.

3. Wu MY, Yang YS, Huang SC, Lee TY, Ho HN. Emergent and

elective cerclage for cervical complications. Int J Gynaecol Obstet

1996;54:23-9.

4. Romero R, Gonzalez R, Sepulveda W, Brandt F, Ramirez M,

Sorokin Y, et al. Infection and labor: VIII, microbial invasion of

the amniotic fluid cavity in patients with suspected cervical

incompetence: prevalence and clinical significance. Am J Obstet

Gynecol 1992;167:1086-91.

5. Mays JK, Figueroa R, Shah J, Khakoo H, Kaminsky S, Tejani N.

Amniocentesis for selection before rescue cerclage. Obstet Gynecol

2000;95:652-5.

6. Althuisius SM, Dekker GA, Hummel P, van Geijn HP. Cervical

incompetence prevention randomized cerclage trial: emergency

cerclage with bed rest versus bed rest alone. Am J Obstet Gynecol

2003;189:907-10.

7. Blanco JD. Clinical problem of preterm labor. Clin Obstet

Gynecol 2000;43:713-6.

8. Lockwood CJ, Kuczynski E. Markers of risk for preterm delivery.

J Perinat Med 1999;27:5-20.

9. Lockwood CJ, Kuczynski E. Risk stratification and pathological

mechanisms in preterm delivery. Paediatr Perinat Epidemiol

2001;15(suppl):78-89.

10. Sherer DM, Salafia CM. Chronic intrauterine bleeding and fetal

growth at less than 32 weeks of gestation. Gynecol Obstet Invest

2000;50:92-5.

11. Salafia CM, Lopez-Zeno JA, Sherer DM, Whittington SS, Minior

VK, Vintzileos AM. Histologic evidence of old intrauterine

bleeding is more frequent in prematurity. Am J Obstet Gynecol

1995;173:1065-70.

12. Yoon BH, Romero R, Kim CJ, Jun JK, Gomez R, Choi JH, et al.

Amniotic fluid interleukin-6: a sensitive test for antenatal diagnosis

of acute inflammatory lesions of preterm placenta and prediction

of perinatal morbidity. Am J Obstet Gynecol 1995;172:960-70.

13. Angus SR, Segel SY, Hsu CD, Locksmith GJ, Clark P, Sammel

MD, et al. Amniotic fluid matrix metalloproteinase-8 indicates

intra-amniotic infection. Am J Obstet Gynecol 2001;185:1232-8.

14. Romero R, Yoon BH, Mazor M, Gomez R, Diamond MP,

Kenney JS, et al. The diagnostic and prognostic value of amniotic

fluid white blood cell count, glucose, interleukin-6, and gram stain

in patients with preterm labor and intact membranes. Am J Obstet

Gynecol 1993;169:805-16.

15. Gibbs RS, Cassell GH, Davis JK, St Clair PJ. Further studies on

genital mycoplasmas in intra-amniotic infection: blood cultures

and serologic response. Am J Obstet Gynecol 1986;154:717-26.

16. Weinberger SR, Morris TS, Pawlak M. Recent trends in protein

biochip technology. Pharmacogenomics 2000;1:395-416.

17. Zhang Z, Bast RC Jr, Yu Y, Li J, Sokoll LJ, Rai AJ, et al. Three

biomarkers identified from serum proteomic analysis for the de-

tection of early stage ovarian cancer. Cancer Res 2004;64:5882-90.

18. Paweletz CP, Trock B, Pennanen M, Tsangaris T, Magnant C,

Liotta LA, et al. Proteomic patterns of nipple aspirate fluids

obtained by SELDI-TOF: potential for new biomarkers to aid in

the diagnosis of breast cancer. Dis Markers 2001;17:301-7.

19. Davidsson P, Westman-Brinkmalm A, Nilsson CL, Lindbjer M,

Paulson L, Andreasen N, et al. Proteome analysis of cerebrospi-

nal fluid proteins in Alzheimer patients. Neuroreport 2002;13:

611-5.

20. Buhimschi IA, Christner R, Buhimschi CS. Proteomic biomarker

analysis of amniotic fluid for identification of intra-amniotic

inflammation. BJOG 2005;112:173-81.

21. Buhimschi I, Christner R. WO Pat 2004/043238 A2, Pat Appl

60/426,096; filed 14 November 2002.

22. Locatelli A, Vergani P, Bellini P, Strobelt N, Arreghini A,

Ghidini A. Amnioreduction in emergency cerclage with prolapsed

membranes: comparison of two methods for reducing the mem-

branes. Am J Perinatol 1999;16:73-7.

23. Buhimschi IA, Buhimschi CS, Christner R, Weiner CP. Non-

invasive assessment of intra-amniotic environment. WO Pat 2004/

072638 A1, PatAppl 60/444,976; filed 5 February 2003.

24. Ganz T, Lehrer RI. Antimicrobial peptides of vertebrates. Curr

Opin Immunol 1998;10:41-4.

25. Kagan BL, Selsted ME, Ganz T, Lehrer RI. Antimicrobial

defensin peptides form voltage-dependent ion-permeable channels

in planar lipid bilayer membranes. Proc Natl Acad Sci USA

1990;87:210-4.

26. Ihi T, Nakazato M, Mukae H, Matsukura S. Elevated concen-

trations of human neutrophil peptides in plasma, blood, and body

fluids from patients with infections. Clin Infect Dis 1997;25:

1134-40.

27. Panyutich AV, Panyutich EA, Krapivin VA, Baturevich EA,

Ganz T. Plasma defensin concentrations are elevated in patients

Page 9: Proteomic biomarkers that predict the clinical success of rescue cerclage

718 Weiner et al

with septicemia or bacterial meningitis. J Lab Clin Med

1993;122:202-7.

28. Soong LB, Ganz T, Ellison A, Caughey GH. Purification and

characterization of defensins from cystic fibrosis sputum. Inflamm

Res 1997;46:98-102.

29. Kretsinger RH, Tolbert D, Nakayama S, Pearson W. The EF-

hand, homologs and analogs. In: Heizmann CW, editor. Novel

calcium binding proteins: fundamentals and clinical applications.

New York: Springer-Verlag; 1991. p. 17-37.

30. Rosen T, Kuczynski E, O’Neill LM, Funai EF, Lockwood CJ.

Plasma levels of thrombin-antithrombin complexes predict pre-

term premature rupture of the fetal membranes. J Matern Fetal

Med 2001;10:297-300.

31. Elovitz MA, Baron J, Phillippe M. The role of thrombin in

preterm parturition. Am J Obstet Gynecol 2001;185:1059-63.

32. Rosen T, Schatz F, Kuczynski E, Lam H, Koo AB, Lockwood CJ.

Thrombin-enhanced matrix metalloproteinase-1 expression:

a mechanism linking placental abruption with premature rupture

of the membranes. J Matern Fetal Neonatal Med 2002;11:11-7.

33. Power C, Henry S, Del Bigio MR, Larsen PH, Corbett D, Imai Y,

et al. Intracerebral hemorrhage induces macrophage activation

and matrix metalloproteinases. Ann Neurol 2003;53:731-42.

34. Buhimschi IA, Buhimschi CS, Weiner CP. Protective effect of N-

acetylcysteine against fetal death and preterm labor induced by

maternal inflammation. Am J Obstet Gynecol 2003;188:203-8.

35. Terrone DA, Rinehart BK, Granger JP, Barrilleaux PS, Martin JN

Jr, Bennett WA. Interleukin-10 administration and bacterial

endotoxin-induced preterm birth in a rat model. Obstet Gynecol

2001;98:476-80.