noninvasive uterine electromyography for prediction of preterm … · 2017-05-19 · evaluate...

��ncl

tlwp

Research www.AJOG.org

OBSTETRICS

Noninvasive uterine electromyographyfor prediction of preterm deliveryMiha Lucovnik, MD; William L. Maner, BSc, BA; Linda R. Chambliss, MD, MPH; Richard Blumrick, MD;James Balducci, MD, MBA; Ziva Novak-Antolic, MD, PhD; Robert E. Garfield, PhD

OBJECTIVE: Power spectrum (PS) of uterine electromyography (EMG)can identify true labor. EMG propagation velocity (PV) to diagnose laborhas not been reported. The objective was to compare uterine EMGagainst current methods to predict preterm delivery.

STUDY DESIGN: EMG was recorded in 116 patients (preterm labor, n20; preterm nonlabor, n � 68; term labor, n � 22; term nonlabor, n6). A Student t test was used to compare EMG values for labor vs

onlabor (P � .05, significant). Predictive values of EMG, Bishop score,ontractions on tocogram, and transvaginal cervical length were calcu-
ated using receiver-operator characteristics analysis.
2011;204:228.e1-10.

dotmns

ecle

bamEthe prognostic capabilit

doi: 10.1016/j.ajog.2010.09.024

228.e1 American Journal of Obstetrics & Gynecology MARCH 2011

RESULTS: PV was higher in preterm and term labor compared withnonlabor (P � .001). Combined PV and PS peak frequency predictedpreterm delivery within 7 days with area under the curve (AUC) of 0.96.Bishop score, contractions, and cervical length had an AUC of 0.72,0.67, and 0.54.

CONCLUSION: Uterine EMG PV and PS peak frequency more accuratelyidentify true preterm labor than clinical methods.

Key words: prediction, preterm labor, propagation velocity, uterine
electromyography
Cite this article as: Lucovnik M, Maner WL, Chambliss LR, et al. Noninvasive uterine electromyography for prediction of preterm delivery. Am J Obstet Gynecol

ie

iunavtp

od

O ther than childbirth, threatenedpreterm labor is the most common

diagnosis that leads to hospitalizationduring pregnancy.1 Up to 50% of pa-ients admitted for threatened pretermabor are, however, not in true labor andill eventually deliver at term.2 Twentyercent of symptomatic patients who are

From the Department of Obstetrics andGynecology (Drs Lucovnik, Chambliss,Blumrick, Balducci, and Garfield and MrManer), St Joseph’s Hospital and MedicalCenter, Phoenix, AZ; and the Division ofPerinatology, Department of Obstetrics andGynecology (Drs Lucovnik and Novak-Antolic), University Medical Center,Ljubljana, Slovenia.

Presented orally at the 57th Annual ScientificMeeting of the Society for GynecologicInvestigation, Orlando, FL, March 24-27, 2010.

Received June 2, 2010; revised Aug. 12, 2010;accepted Sept. 22, 2010.

Reprints: Robert E. Garfield, PhD, St Joseph’sHospital and Medical Center, DowntownCampus at TGen, 445 N 5th St., Phoenix, AZ85004. [email protected].

This study was supported in part by NationalInstitutes of Health Grant R01HD037480 andthe St Joseph’s Foundation, Phoenix, AZ.

0002-9378/$36.00© 2011 Mosby, Inc. All rights reserved.

iagnosed as not being in preterm labor,n the other hand, will deliver prema-urely.3 This leads to unnecessary treat-

ents, missed opportunities to improveeonatal outcome, and largely biased re-earch of treatments.

Myometrial activation, required forffective contractions and true labor, isharacterized by molecular changeseading to an increase in coupling andxcitability of cells.4-7 Electrical activity

of the myometrium, which can be mon-itored noninvasively by measuring theuterine electromyography (EMG),changes at delivery as a result of theseevents.8-14 Bursts of electrical signals re-sponsible for contractions have been re-ported to be more frequent and their du-ration more constant in labor.15,16

An increase in peak amplitude and fre-quency of EMG signals, assessed by power-spectrum (PS) analysis, has also been ob-served prior to labor.15,17,18 Propagationvelocity (PV) of electrical signals in themyometrium has been shown in vitro toincrease before delivery when gap junc-tions are increased.19,20 As a result, it has

een suggested that EMG could be used tossess the PV in vivo. Previous studiesainly focused on methods for assessing

MG signal propagation.21-28 However,
y of PV for predict-
ng labor (term or preterm) has not beenvaluated yet.

This study investigated whether uter-ne EMG can be used to evaluate PV ofterine electrical signals in labor andonlabor patients at term and pretermnd compared diagnostic accuracy ofarious EMG parameters, including PV,o methods currently used in the clinic toredict preterm delivery.

MATERIALS AND METHODSPatientsOne hundred sixteen pregnant womenwere included in the study at a single in-stitution (St Joseph’s Hospital and Med-ical Center, Phoenix, AZ). From previ-ous EMG studies, there has beenreported difference in means of EMG PSpeak frequency in labor vs nonlabor pa-tients of (0.4708 – 0.3982 � 0.0726 Hz),and an average SD of ([0.0459 �0.0231]/2 � 0.0345 Hz).16 Using a power

f 0.80 and an alpha of – 0.05, with a Stu-ent t test, gives a desired sample size of 5

per group minimum.Eighty-eight consecutive preterm pa-

tients were included. They were admit-ted with the diagnosis of preterm labor atless than 34 weeks of gestational age. Thecutoff of 34 weeks was chosen because
the risk of death and handicap is mainly
mailto:[email protected]

cm(aod

odscfilsbdi

pt(chdghi

lbbtofd

www.AJOG.org Obstetrics Research

increased if delivery occurs prior to thistime point, and attempts to stop pretermlabor are very rarely done at later gesta-tions.29 Preterm labor was diagnosedlinically as at least 6 contractions in 60inutes assessed by tocodynamometer

TOCO) and/or maternal perceptionnd a cervical dilatation of at least 2 cmr effacement of at least 80% assessed byigital cervical examination.Calculation of gestational age was based

n the last menstrual period or, when itiffered by 7 days or longer from the ultra-onographic estimation (calculated byrown-rump length measured within therst trimester) on ultrasound. Women de-

ivering within 7 days from the EMG mea-urement were classified in the preterm la-or group, and those delivering outside 7ays from the measurement were classified

FIGURE 1Flow sheet diagram for better visu

EMG, electromyography.

Lucovnik. Electromyography for prediction of preterm deliver

n the preterm nonlabor group.

Twenty-eight consecutive patientsresenting with regular uterine contrac-ions with intact membranes at term�37 weeks of gestation) were also in-luded. Women delivering within 24ours from the EMG measurement wereefined as being in labor (term laborroup) and those delivering outside 24ours from the measurement as not be-

ng in labor (term nonlabor group).Different cutoff measurement-to-de-

ivery intervals for term and preterm la-or vs nonlabor groups were chosenased on previous studies, which showedhat an increase in uterine EMG activityccurs within approximately 24 hoursrom delivery at term and within severalays from delivery preterm.17

All women included provided writteninformed consent for study participation.

ation of study groups

m J Obstet Gynecol 2011.

Data from patients who ultimately under-

MARCH 2011 Americ

went cesarean section were not used foranalysis (Figure 1). We chose to excludethe cesarean section patients because thedecision on when exactly the surgery willbe performed is based on several consider-ations, including those on fetal well-beingand the subjective assessment of laborprogress. This decision is therefore toosubjective and arbitrary to include in aproper receiver-operating characteristics(ROC) analysis in which one wants to alsoaccurately determine the mean measure-ment-to-delivery interval.

We chose to include patients with pre-term premature rupture of membranes(PPROM) because our objective was toevaluate whether uterine EMG measure-ments can differentiate between truepreterm labor (ie, patients who are goingto deliver spontaneously within a short

aliz

y. A

period of time) and those who are not in

an Journal of Obstetrics & Gynecology 228.e2

tfp

PP

Research Obstetrics www.AJOG.org

true preterm labor, during which theclinical evaluation does not allow us tomake this differentiation.

The St Joseph’s Hospital and MedicalCenter Institutional Review Board ap-proved the study.

Uterine EMG signal recordingsUterine EMG measurements were per-formed by 5 different researchers within 24hours from the patient’s admission to thehospital. We standardized the electrode ar-rangement to the following factors: 2.5 cmelectrode-electrode vertical and horizontal

FIGURE 2Propagation velocity of EMG signal

The propagation velocity of the EMG signals wasample recording from 2 electrode pairs (channel(measured by TOCO). Bottom trace shows the eEMG, electromyography; TOCO, tocodynamometer.


separation distances (measured from cen-

228.e3 American Journal of Obstetrics & Gynecolo

ter to center) in a square-shaped patternabout the navel and with each electrodepositioned in the vertex of each of the 4corners of the square. Uterine EMG wasmeasured for 30 minutes from each pa-tient using a custom-built uterine EMGpatient-monitoring system. Patients wereasked to remain still while supine withoutdisturbing any of the probes and wires forthe recordings.

Signal analysisAnalog EMG signals were digitally filteredto yield a final band-pass of 0.34-1.00 Hz to

lculated from the time difference of the signal aand 2). Note the excellent temporal correspondended EMG burst, with the individual voltage peak


exclude most components of motion, res- t

gy MARCH 2011

piration, and cardiac signals from the anal-ysis. Data were sampled at 100 Hz (thishigh sampling rate was chosen to increasethe resolution of PS analysis later).

Previously described EMG parametersEMG parameters were chosen from pre-vious publications.17,30 The definition ofhese parameters, as well as the rationaleor their use, can also be found in thoseublications.

ropagation velocity analysisV can be calculated by dividing the dis-

al at adjacent electrodes (t). Top trace shows abetween EMG and mechanical contractile eventslearly distinguishable.

s

s ca rrivs 1 cexpan s c

y. A

ance that the propagating wave travels

ivery


by the amount of time required for thepropagating wave to traverse this dis-tance. All the time differences in corre-sponding action potential peaks for eachburst of action potentials were calcu-lated, and the average of absolute valuesof all time differences for bursts in a pa-tient’s uterine EMG recording was usedto calculate the PV.

Only those pairs of peaks on differentchannels (electrode pairs) with congru-ent shapes and within 2 seconds tempo-ral separation of one another were in-cluded in the analysis, ensuring that anypeak observed on 1 channel during any 2second window matched the associatedpeak on the other channel (Figure 2).

For our EMG instrument, we use dif-ferential, bipolar electrode pairs. The ad-vantage of a differential bipolar setupover a monopolar setup is signal quality,allowing us to more accurately identifyindividual uterine voltage peaks. Onlythose bursts for which the mean voltagepeak value was greater than 2 times themean baseline voltage peak value were

TABLE 1Clinical background variables, pret

Variable

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

Nulliparous...................................................................................................................

Number of previous gestations...................................................................................................................

Previous preterm delivery or late abortion...................................................................................................................

Twin gestations...................................................................................................................

Gestational age at measurement...................................................................................................................

PPROM...................................................................................................................

Smoking...................................................................................................................

BMI, kg/m2

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

Illicit drug abuse...................................................................................................................

Tocolytic treatment...................................................................................................................

Antenatal corticosteroids...................................................................................................................

Contractions on TOCO...................................................................................................................

Bishop score...................................................................................................................

Transvaginal cervical length, cm (n � 59)...................................................................................................................

Data are median (range) and n. P value was calculated by MaBMI, body mass index; EMG, electromyography; PPROM, preta Statistical significance (P � .05).

Lucovnik. Electromyography for prediction of preterm del

used in these calculations to clearly see

and compare uterine voltage peaks at ad-jacent electrodes. Within each of theseelectrical contractile bursts, there werefound anywhere from approximately 30to 60 voltage peaks (associated withpropagating voltage waves), which wereanalyzed (Figure 2). More than 25,000voltage peaks pairs were analyzed, withan average of approximately 215 peaksper patient.

Common obstetric measuresPresence or absence of contractions onTOCO at the time of EMG measurementwas documented by the researcher re-cording the EMG. Transvaginal cervicallength and Bishop score were also docu-mented but only when they were as-sessed no more than 24 hours before orafter the EMG measurement. Bishopscore has not been developed as a predic-tor of preterm birth. Nevertheless, itgives a metric evaluation of digital cervi-cal examination, which is commonlyused to diagnose preterm labor and pre-dict delivery. That is why we chose to

patients

Women deliveringwithin 7 days (n � 20)

24 (18–40).........................................................................................................................

5.........................................................................................................................

1 (0–8).........................................................................................................................

2.........................................................................................................................

1.........................................................................................................................

27 5/7 (22 6/7 to 33 4/7).........................................................................................................................

3.........................................................................................................................

1.........................................................................................................................

28 (24–47).........................................................................................................................

1.........................................................................................................................

16.........................................................................................................................

11.........................................................................................................................

7.........................................................................................................................

7 (2–13).........................................................................................................................

2.0 (0.5–3.5) (n � 7).........................................................................................................................

hitney U-test and Student’s t test.premature rupture of membranes; TOCO, tocodynamometer.

. Am J Obstet Gynecol 2011.

compare predictive values of Bishop

MARCH 2011 Americ

score with those of uterine EMG. Trans-vaginal cervical length was measured by1 of 5 board-certified perinatologists or 1of 4 certified ultrasound technicians.Digital examination was performed by 1of 26 resident physicians involved in thecare of the patients included in the study.

StatisticsA Student t test and a Mann Whitney Utest (when appropriate because of non-normal distribution of variables) wereused to compare delivery within, vs out-side, 24 hours from the measurement interm patients, and 7 days from the mea-surement in preterm patients. Statisticalcomparison between preterm and termpatients was performed to determinewhether gestational age has an impact onEMG PV. Data were analyzed by analysisof variance, and Dunn’s test was used forpair-wise comparisons among groups.Pearson’s correlation analysis was usedto determine whether demographicand/or clinical parameters influence the

omen deliveringter 7 days (n � 68) P value

(18–43) .59..................................................................................................................

.99..................................................................................................................

(0–11) .64..................................................................................................................

.54..................................................................................................................

.65..................................................................................................................

6/7 (21 5/7 to 33 6/7) .51..................................................................................................................

.42..................................................................................................................

.58..................................................................................................................

(20–45) .15..................................................................................................................

.72..................................................................................................................

.89..................................................................................................................

.09..................................................................................................................

.64..................................................................................................................

(1–10) .01a

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

.8 (0.3–4.8) (n � 52) .16..................................................................................................................

erm

Waf

27......... .........

16......... .........

1......... .........

13......... .........

8......... .........

28......... .........

2......... .........

9......... .........

27......... .........

7......... .........

53......... .........

54......... .........

19......... .........

5......... .........

2......... .........

nn-Werm

PV. A P � .05 was considered significant.



228.e5 American Journal of Obstetrics & Gynecology MARCH 2011

The ROC curves were used to estimatethe predictive values of the EMG param-eters that were significantly higher inpreterm patients delivering within 7 daysand to assess the diagnostic accuracy ofBishop score, contractions on TOCO,and transvaginal cervical length for pre-dicting preterm delivery within 7 days.Diagnostic accuracy of EMG, Bishopscore, TOCO, and the transvaginal cer-vical length were then compared.

The ROC analysis was also performedon the cohort of patients for whom theresults of all the examined methods wereavailable. The ROC curves constructedon this cohort were compared by Stu-dent t test.

Repeatability and reproducibilitymeasuresThe PV analysis was reperformed on asubset of the data to determine intraob-server and interobserver agreement. Ap-proximately 175 voltage spikes from pa-tients in term labor, term nonlabor,preterm labor, and preterm nonlaborgroups were reanalyzed. The intraobserverand interobserver agreements were calcu-lated according to the statistical methodsproposed previously.31

RESULTSOne hundred sixteen patients were eval-uated in the study. The data regardingthe preterm and term groups will be pre-sented in separate sections.

Preterm patientsGeneralThe study population consisted of 88pregnant women admitted to our insti-tution between September 2009 andFebruary 2010 with the diagnosis of pre-term labor at less than 34 weeks’ gesta-tion. Uterine EMG was initially recordedin 98 preterm patients, but 10 patientswho underwent cesarean section weresubsequently excluded from the analysis.Patients were included in the study at amedian of 28 5/7 weeks of gestational age(range, 21 5/7 to 33 6/7 weeks). Nine pa-tients (10%) were less than 24 weeks; 31(35%), 24-28 weeks; 33 (38%), 28-32weeks; and 15 (17%) longer than 32weeks. Delivery within 7 days from the

FIGURE 3Uterine electromyography propagation velocity values

Comparison of uterine electromyography propagation velocity values for preterm patients deliveringwithin 7 days of measurement vs those delivering more than 7 days from measurement. Means andSDs shown. The asterisk represents statistical significance (P � .05).Lucovnik. Electromyography for prediction of preterm delivery. Am J Obstet Gynecol 2011.

FIGURE 4Uterine electromyography PS peak frequency values

Comparison of uterine electromyography PS peak frequency values for preterm patients deliveringwithin 7 days of measurement vs those delivering more than 7 days from measurement. Means andSDs shown. Asterisk represents statistical significance (P � .05).PS, power-spectrum.

Lucovnik. Electromyography for prediction of preterm delivery. Am J Obstet Gynecol 2011.

EMG measurement occurred in 23%


(20/88) of the patients. Of 68 patientswho did not deliver within 7 days fromadmission, 23 delivered at term (after 37weeks), and 45 delivered before 37 weeksof gestation.

The clinical background variables aresummarized in Table 1. It should benoted that at the time of EMG measure-ment (no more than 24 hours after ad-mission to the hospital), the contrac-tions were detected by TOCO in only30% of patients (26/88), although theywere all initially admitted for preterm la-bor. In spite of this factor, uterine EMGsignals were being produced and easilydetected in these patients. This fact con-

TABLE 2Diagnostic EMG parameters, preter

EMG parameterWomen delivewithin 7 d

PS peak frequencya 0.56 � 0.15...................................................................................................................

PVa 52.56 � 33.9...................................................................................................................

PS median frequency 0.64 � 0.12...................................................................................................................

PS peak amplitude 50.98 � 84.7...................................................................................................................

PS median amplitude 16.27 � 44.1...................................................................................................................

Mean burst duration 35.53 � 9.00...................................................................................................................

SD of burst duration 7.44 � 5.85...................................................................................................................

Mean interburst interval 307.5 � 178.3...................................................................................................................

SD of interburst interval 184.14 � 136....................................................................................................................

P value was calculated by Mann-Whitney U test and Student tEMG, electromyography; PS, power-spectrum; PV, propagatioa Statistical significance (P � .05).

Lucovnik. Electromyography for prediction of preterm del

TABLE 3Uterine EMG predictive measures

Variable

1 d to delivery 2 d to

PV PFr

PVplusPFr PV

AUC 0.91 0.61 0.90 0.92...................................................................................................................

Best cutoff 22.13 cm/s 0.87 Hz 191.96 28.00...................................................................................................................

Sensitivity, % 100 14 14 77...................................................................................................................

Specificity, % 80 100 100 92...................................................................................................................

PPV, % 30 100 100 63...................................................................................................................

NPV, % 100 92 92 96...................................................................................................................

Best cutoff values are presented as centimeters per second aAUC, area under the curve; EMG, electromyography; NPV, neg

Lucovnik. Electromyography for prediction of preterm delivery

firms that EMG is a much more sensitivetechnology for assessing the contractileactivity than is the TOCO.

Transvaginal cervical length was mea-sured in 67% of patients (59/88). Cervi-cal length was not significantly shorter inwomen who delivered within 7 days(median, 2.0 cm) compared with that inwomen who did not (median, 2.8 cm)(P � .16).

EMG parametersEMG PV was significantly higher in pa-tients delivering within 7 days from themeasurement (52.56 � 33.94 cm/sec-ond) compared with those who deliv-

patients

Women deliveringafter 7 d P value

0.44 � 0.07 Hz .002..................................................................................................................

/s 11.11 � 5.13 cm/s � .001..................................................................................................................

0.68 � 0.05 Hz .11..................................................................................................................

V2 70.56 � 134.64 �V2 .63..................................................................................................................

V2 10.16 � 16.29 �V2 .58..................................................................................................................

39.32 � 12.26 s .21..................................................................................................................

10.16 � 7.0 s .07..................................................................................................................

348.96 � 227.27 s .65..................................................................................................................

s 149.76 � 157.92 s .26..................................................................................................................

.ocity; SD, standard deviation.


ery 4 d to delivery 7 d t

PFr

PVplusPFr PV PFr

PVplusPFr PV

0.66 0.90 0.96 0.74 0.95 0.9.........................................................................................................................

0.87 Hz 191.96 24.88 cm/s 0.64Hz 95.33 22.8.........................................................................................................................

8 8 82 18 53 85.........................................................................................................................

100 100 93 100 100 94.........................................................................................................................

100 100 74 100 100 81.........................................................................................................................

83 84 69 80 88 96.........................................................................................................................

rtz. Their rescaled sum has no units.predictive value; PFr, PS peak frequency; PPV, positive predictiv


MARCH 2011 Americ

ered after 7 days (11.11 � 5.13 cm/sec-ond) (P � .001; Figure 3). No statisticallysignificant correlation was found be-tween PV and the following demograph-ic/clinical variables: maternal age, bodymass index, number of previous gesta-tions, number of previous pretermbirths, gestational age, twin gestations,PPROM, smoking, drug abuse, tocolytictreatment, and antenatal corticosteroids.

PS peak frequencies were also signifi-cantly higher in women who deliveredwithin 7 days (0.56 � 0.15 Hz) comparedwith those who did not (0.44 � 0.07 Hz)(P � .002; Figure 4). All other EMG pa-rameters analyzed did not differ signifi-cantly among groups (Table 2).

Predictive values of EMG PV, PSpeak frequency, and the combination(rescaled sum) of these parameters forpredicting preterm delivery at varioustime points were calculated (Table 3).The PV and PS peak frequencies werecombined by looking at the sum of theirrescaled values. Specifically, PS peak fre-quency was multiplied by 100 and addedto the corresponding PV value. Thecombination of these 2 parametersyielded the best predictive values at 7days to delivery. These values werehigher than for any parameter alone atany time point. A similar combination(product) using PV and PS peak fre-quency yielded no better results.

The sum of the PV and PS peak fre-quency increased as the measurement-to-delivery interval decreased, as shownin Figure 5. Based on this figure, it is pos-

livery 14 d to delivery

PFr

PVplusPFr PV PFr

PVplusPFr

0.78 0.96 0.89 0.71 0.89..................................................................................................................

/s 0.64 Hz 84.48 26.6 cm/s 0.64 Hz 84.48..................................................................................................................

15 70 70 13 61..................................................................................................................

100 100 99 100 100..................................................................................................................

100 100 94 100 100..................................................................................................................

78 90 90 71 85..................................................................................................................

lue; PS, power-spectrum; PV, propagation velocity.

m

ring

Hz.........

4 cm.........

Hz.........

0 �.........

4 �.........

s.........

s.........

8 s.........

68.........

testn vel

ivery

deliv o de

5......... .........

cm/s 8 cm......... .........

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

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

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

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

nd heative e va


auf((woww

(

m


sible to speculate that the EMG activityincreases even before 7 days from deliv-ery, and it may be useful to identify pa-tients in preterm labor even prior to thiscutoff.

Figure 6 and Table 4 present ROCnalysis results. The ROC curve forterine EMG differed significantly

rom the ROC curve for Bishop scoreP � .02), transvaginal cervical lengthP � .03), and TOCO (P � .001), evenhen the ROC analysis was performedn the same group of patients forhich the results of all the methodsere available.

Term patientsUterine EMG was recorded in 36 termpatients, and 8 patients who under-went cesarean section were subse-quently excluded from the analysis.Gestational age at inclusion did notdiffer significantly between the 2groups (P � .216). The median gesta-

FIGURE 5Uterine electromyography PV plus

Uterine electromyography PV plus PS peak freinterval decreases. Open triangle indicates prete

ent; closed triangle indicates delivery more thaPS, power-spectrum; PV, propagation velocity.


tional age for labor patients was 39 2/7


weeks (range, 38 0/7 to 40 6/7 weeks)and for nonlabor patients, 38 5/7 weeks(range, 37 1/7 to 41 1/7 weeks). Themedian measurement to delivery inter-val for nonlabor patients was 8 days(range, 3–14 days) and in the laborgroup, 4 hours (range, 2–14 hours). PVwas significantly higher (P � .001) inlabor (31.25 � 14.91 cm/second) com-pared with nonlabor patients (11.31 �2.89 cm/second) (Figure 7).

In an ROC analysis to distinguish be-tween term patients in true vs false labor,PV had an area under the curve (AUC) of0.98. For predicting delivery within 24hours, a PV greater than 13.19 cm/sec-ond had 100% sensitivity, 83% specific-ity, 96% positive predictive value (PPV),and 100% negative predictive value(NPV).

Comparison between pretermand term patientsPV was significantly higher in labor at

peak frequency

ncy increases as the measurement-to-deliverylabor: delivery 7 days or less from the measure-days from the measurement.


term and preterm compared with nonla-

gy MARCH 2011

bor patients at term and preterm. Thedifferences between the term and pre-term labor and term and preterm nonla-bor groups were not significant (P � .05)Figure 7).

Repeatability and reproducibilitymeasuresIntraobserver and interobserver agree-ments were 99.92 � 0.04% and 99.67 �0.20%, respectively. The technique to as-sess the PV of uterine EMG signals, de-scribed in Materials and Methods, istherefore highly repeatable andreproducible.

COMMENTNoninvasive measurement of uterineEMG propagation and frequency canidentify true preterm labor more accu-rately than the currently used methods.It can therefore identify those patientswho will really benefit from early institu-tion of tocolytic therapy, transport to ahospital with facilities for neonatal in-tensive care, and administration of ste-roids. At the same time, uterine EMGalso identifies patients in false pretermlabor who are not going to deliver withinthe next 7 days. This can help to avoidsubstantial economic costs associatedwith unnecessary hospitalization, thematernal risks associated with tocolytics,and the potential fetal risks associatedwith steroids.

In the case of low PV and peak fre-quency values, it therefore stands to rea-son that it would be safe not to admit,treat, or transfer the patient, regardlessof the presence of contractions onTOCO and regardless of digital cervicalexam and transvaginal cervical length re-sults because all of the changes in themyometrium required for labor are notyet fully established.

Methods currently available to clini-cians to diagnose preterm labor have sev-eral major drawbacks. Digital cervicalexamination is subjective and does notprovide accurate diagnosis of true pre-term labor.32,33 In the present study, thepredictive measures of Bishop score werehigh only at scores of greater than 10,which is not useful clinically because atthat point imminent delivery is already

PS

quermn 7

y. A

obvious.

ceibcttPbblE


An interesting finding of this studywas that the uterine EMG is better thantransvaginal cervical length in predictingpreterm birth within 7 days of evalua-tion. This raises an important aspect re-lated to the clinical utility of the cervicallength in predicting preterm deliverywithin a short time interval. There arecompelling data to suggest that cervicallength is a good predictive marker forpreterm birth between 24 and 32 weeksof gestation but on the long and not onthe short time interval.34

It can be argued that the cervicallength was measured only in 67% of the

FIGURE 6Comparison of ROC curves for uterand currently used clinical method

The comparison of ROC curves for uterine EMfrequency) and currently used clinical methods tEMG, electromyography; PS, power-spectrum; PV, propagation ve


TABLE 4Uterine EMG predictive measures c

Method

EMG (PV plus PS peak frequency) (n � 88)...................................................................................................................

Bishop score (n � 88)...................................................................................................................

Transvaginal cervical length (n � 59)...................................................................................................................

Contractions on TOCO (n � 88)...................................................................................................................

AUC, area under the curve; EMG, electromyography; n, numpropagation velocity; TOCO, tocodynamometer.

Lucovnik. Electromyography for prediction of preterm delivery

patients included and in only 7 patientswho delivered within 7 days. In many ofthe patients who presented with ad-vanced cervical dilatation, cervicallength was not obtained, and those pa-tients were more likely to deliver within 7days. The predictive values would there-fore most likely be better if the transvag-inal cervical length of all patients wereknown. However, several patients withshort cervices in this study did not de-liver within 1 week, and some did notdeliver preterm at all.

Further studies are needed to deter-mine whether the addition of the fetal

EMG parameters

arameters (rescaled sum of PV and PS peakedict preterm delivery within 7 days.y; ROC, receiver-operating characteristics.


bined

AUC Best cutoff Sensitivity, %

0.96 84.48 70.........................................................................................................................

0.72 10 18.........................................................................................................................

0.67 0.7 cm 14.........................................................................................................................

0.54 N/A 35.........................................................................................................................

of patients included in the analysis; NPV, negative predictive va


MARCH 2011 Americ

fibronectin test can improve the predic-tive value of transvaginal cervical lengthalone. However, given that the cervicallength has much lower predictive valuescompared with uterine EMG, we doubtthat the fetal fibronectin test will signifi-cantly change these results.

Our results also confirm that monitor-ing uterine activity with TOCO is nothelpful in identifying patients in pretermlabor.35,36 Only 23% of patients withcontractions on TOCO during the 30minutes of EMG recording deliveredwithin 7 days, and the absence of con-tractions apparently does not rule outpreterm labor reliably because the NPVwas only 79%. Approximately 1 in 5 pa-tients who exhibited no contractionsregistering on TOCO did neverthelessdeliver preterm within 1 week.

We are highly encouraged that mea-suring uterine electrical activity is supe-rior to TOCO. It not only detects uterinecontractions as does TOCO but can alsoidentify myometrial properties that dis-tinguish physiological preterm contrac-tions from true preterm labor.

PS peak frequency has been the mostpredictive of true labor in both humanand animal studies.16,17 Our resultsonfirm that shifts to higher uterinelectrical signal frequencies occur dur-ng transition from a nonlabor state tooth term and preterm labor states andan be reliably assessed by noninvasiveransabdominal uterine EMG. In addi-ion, we demonstrate that not only canV of EMG signals in the myometriume determined by EMG measurementut also that PV predicts preterm de-

ivery more accurately than any otherMG parameter investigated so far.

Specificity, % PPV, % NPV, %

100 100 90..................................................................................................................

100 100 81..................................................................................................................

98 50 90..................................................................................................................

72 27 79..................................................................................................................

PPV, positive predictive value; PS, power-spectrum; PV,

ines

G po prlocit

y. A

om

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

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

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

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

ber lue;


.tps


Similar data were obtained with pa-tients at term.

Possible reasons for the increased PVduring preterm and term labor are thepresence of gap junctions, number ofmuscle cells recruited, ion channels (sizeand number), cellular resting potential,cellular threshold potential, and conduc-tion pathways: all of these play a part inwhat types of EMG signals are ultimatelyreceived at the electrodes.

We evaluated many EMG parametersand their predictive values for pretermdelivery among patients admitted forthreatened preterm labor. EMG PV, PSpeak frequency, and their combination(rescaled sum) identified which patientswill deliver soon and which will not withhighest PPV and NPV. It would be usefulto do a prospective clinical trial to con-firm our results. They may be extremely

FIGURE 7PV comparison of labor vs nonlabo

PV was significantly higher (P � .001) in labor gences between term vs preterm labor and term vs05). Data are presented as box plots and not vererm labor group. Term labor is delivery within 2reterm labor is delivery within 7 days; preterm ntatistical significance.

PV, propagation velocity.


important clinically but could also lead


to more reliable analyses of treatmentsand eventually to more effective inter-ventions to prevent preterm birth. f

REFERENCES1. Bennett TA, Kotelchuck M, Cox CE, TuckerMJ, Nadeau DA. Pregnancy-associated hospi-talizations in the United States in 1991 and1992: a comprehensive view of maternal mor-bidity. Am J Obstet Gynecol 1998;178:346-54.2. McPheeters ML, Miller WC, Hartmann KE, etal. The epidemiology of threatened preterm la-bor: a prospective cohort study. Am J ObstetGynecol 2005;192:1325-30.3. Iams JD, Romero R. Preterm Birth. In: GabbeSG, Niebyl JR, Simpson JL, eds. Obstetrics:normal and problem pregnancies, 5th ed. Phil-adelphia, PA: Churchill Livingstone Elsevier;2007:668-712.4. Tezuka N, Ali M, Chwalisz K, Garfield RE.Changes in transcripts encoding calcium chan-nel subunits of rat myometrium during preg-

roups

ps compared with nonlabor groups. The differ-eterm nonlabor groups were not significant (P �l bar charts because of nonnormal distribution inours; term nonlabor is delivery after 24 hours;

abor is delivery after 7 days. Asterisk represents


nancy. Am J Physiol 1995;269:1008-17.

gy MARCH 2011

5. Balducci J, Risek B, Gilula NB, Hand A, EganJF, Vintzileos AM. Gap junction formation in hu-man myometrium: a key to preterm labor. Am JObstet Gynecol 1993;168:1609-15.6. Garfield RE, Blennerhassett MG, Miller SM.Control of myometrial contractility: role and reg-ulation of gap junctions. Oxf Rev Reprod Biol1988;10:436-90.7. Young RC. Tissue-level signaling and controlof uterine contractility: the action potential-cal-cium wave hypothesis. J Soc Gynecol Investig2000;7:146-52.8. Buhimschi C, Boyle MB, Saade GR, GarfieldRE. Uterine activity during pregnancy and laborassessed by simultaneous recordings from themyometrium and abdominal surface in the rat.Am J Obstet Gynecol 1998;178:811-22.9. Buhimschi C, Garfield RE. Uterine contractil-ity as assessed by abdominal surface recordingof electromyographic activity in rats duringpregnancy. Am J Obstet Gynecol 1996;174:744-53.10. Garfield RE, Chwalisz K, Shi L, Olson G,Saade GR. Instrumentation for the diagnosis ofterm and preterm labour. J Perinat Med1998;26:413-36.11. Garfield RE, Saade G, Buhimschi C, et al.Control and assessment of the uterus and cer-vix during pregnancy and labour. Hum ReprodUpdate 1998;4:673-95.12. Devedeux D, Marque C, Mansour S, Ger-main G, Duchene J. Uterine electromyography:a critical review. Am J Obstet Gynecol 1993;169:1636-53.13. Marque C, Terrien J, Rihana S, Germain G.Preterm labour detection by use of a biophysi-cal marker: the uterine electrical activity. BMCPregnancy Childbirth 2007;7(Suppl 1):S5.14. Vinken MPGC, Rabotti C, Mischi M, Oei SG.Accuracy of frequency related parameters ofthe electrohysterogram for predicting pretermdelivery. A review of the literature. Obstet Gyne-col Surv 2009;64:529-34.15. Buhimschi C, Boyle MB, Garfield RE. Elec-trical activity of the human uterus during preg-nancy as recorded from the abdominal surface.Obstet Gynecol 1997;90:102-11.16. Maner WL, Garfield RE. Identification of hu-man term and preterm labor using artificial neu-ral networks on uterine electromyography data.Ann Biomed Eng 2007;35:465-73.17. Maner WL. Garfield RE, Maul H, et al. Pre-dicting term and preterm delivery with transab-dominal uterine electromyography. Obstet Gy-necol 2003;101:1254-60.18. Marque C, Duchene J, Leclercq S, PanczerG, Chaumont J. Uterine EHG processing forobstetrical monitoring. IEEE Trans Biomed Eng1986;33:1182-7.19. Miller SM, Garfield RE, Daniel EE. Improvedpropagation in myometrium associated withgap junctions during parturition. Am J Physiol1989;256:C130-41.20. Miyoshi H, Boyle MB, MacKay LB, GarfieldGE. Gap junction currents in cultured musclecells from human myometrium. Am J Obstet

r g

roupr

tica4 honl

y. A

Gynecol 1997;178:588-93.


21. Duchene J, Marque C, Planque S. UterineEMG signal: propagation analysis. Conf ProcIEEE Eng Med Biol Soc 1990;1990:831-2.22. Rabotti C, Mischi M, van Laar JOEH, OeiSG, Bergmans JWM. On the propagation anal-ysis of electrohysterographic signals. Conf ProcIEEE Eng Med Biol Soc 2008;2008:3868-71.23. Rabotti C, Mischi M, van Laar JOEH, OeiSG, Bergmans JWM. Inter-electrode delay es-timators for electrohysterographic propagationanalysis. Physiol Meas 2009;30:745-61.24. Mischi M, Rabotti C, Vosters LPJ, Oei SG,Bergmans JWM. Electrohysterographic con-duction velocity estimation. Conf Proc IEEE EngMed Biol Soc 2009;2009:6934-7.25. Terrien J, Marque C, Germain G, KarlssonB. Sources of bias in synchronization measuresand how to minimize their effects on the estima-tion of synchronicity: application to the uterineelectromyogram. In: Naik GR, ed. Recent ad-vances in biomedical engineering. Vienna (Aus-tria): I-Tech Education and Publishing; 2009:73-99.26. Euliano TY, Marossero D, Nguyen MT, Eu-liano NR, Principe J, Edwards RK. Spatiotem-
poral electrohysterography patterns in normal
and arrested labor. Am J Obstet Gynecol2009;200:54.e1-7.27. Hassan M, Terrien J, Karlsson B, Marque C.Spatial analysis of uterine EMG signals: evi-dence of increased in synchronization withterm. Conf Proc IEEE Eng Med Biol Soc2009;2009:6296-9.28. Hassan M, Terrien J, Karlsson B, Marque C.Synchronization between EMG at different uter-ine locations investigated using time-frequencyridge reconstruction: comparison of pregnancyand labor contractions. EURASIP J Adv SignalProcess 2010;20:1-10. Article ID 242493.29. Marlow N, Wolke D, Bracewell MA, SamaraM. Neurologic and developmental disability atsix years of age after extremely preterm birth.N Engl J Med 2005;352:9-19.30. Garfield RE, Maner WL, MacKay LB,Schlembach D, Saade GR. Comparing uterineelectromyography activity of antepartum pa-tients versus term labor patients. Am J ObstetGynecol 2005;193:23-9.31. Bland JM, Altman DG. Statistical methodsof assessing agreement between methods of
clinical measurements. Lancet 1986;1:307-10.
MARCH 2011 American

32. Gomez R, Galasso M, Romero R, et al. Ul-trasonographic examination of the uterine cer-vix is better than cervical digital examination asa predictor of the likelihood of premature deliv-ery in patients with preterm labor and intactmembranes. Am J Obstet Gynecol 1994;171:956-64.33. Jackson GM, Ludmir J, Bader TJ. The ac-curacy of digital examination and ultrasound inthe evaluation of cervical length. Obstet Gyne-col 1992;79:214-8.34. Iams JD, Goldenberg RL, Meis PJ, MercerBM, Moawad A, Das A. The length of the cervixand the risk of spontaneous premature delivery.N Engl J Med 1996;334:567.35. Iams JD, Newman RB, Thom EA, et al. Fre-quency of uterine contractions and the risk ofspontaneous preterm delivery. N Engl J Med2002;346:250-5.36. Maul H, Maner WL, Olson G, Saade GR,Garfield RE. Non-invasive transabdominal uter-ine electromyography correlates with thestrength of intrauterine pressure and is predic-tive of labor and delivery. J Matern Fetal Neo-
natal Med 2004;15:297-301.
Journal of Obstetrics & Gynecology 228.e10

noninvasive uterine electromyography for prediction of preterm … · 2017-05-19 · evaluate...

Documents