Patients

One hundred sixteen pregnant women were included in the study at a single institution (St Joseph’s Hospital and Medical Center, Phoenix, AZ). From previous EMG studies, there has been reported difference in means of EMG PS peak frequency in labor vs nonlabor patients of (0.4708 – 0.3982 = 0.0726 Hz), and an average SD of ([0.0459 + 0.0231]/2 = 0.0345 Hz). Using a power of 0.80 and an alpha of –0.05, with a Student t test, gives a desired sample size of 5 per group minimum.

Eighty-eight consecutive preterm patients were included. They were admitted with the diagnosis of preterm labor at less than 34 weeks of gestational age. The cutoff of 34 weeks was chosen because the risk of death and handicap is mainly increased if delivery occurs prior to this time point, and attempts to stop preterm labor are very rarely done at later gestations. Preterm labor was diagnosed clinically as at least 6 contractions in 60 minutes assessed by tocodynamometer (TOCO) and/or maternal perception and a cervical dilatation of at least 2 cm or effacement of at least 80% assessed by digital cervical examination.

Calculation of gestational age was based on the last menstrual period or, when it differed by 7 days or longer from the ultrasonographic estimation (calculated by crown-rump length measured within the first trimester) on ultrasound. Women delivering within 7 days from the EMG measurement were classified in the preterm labor group, and those delivering outside 7 days from the measurement were classified in the preterm nonlabor group.

Twenty-eight consecutive patients presenting with regular uterine contractions with intact membranes at term (>37 weeks of gestation) were also included. Women delivering within 24 hours from the EMG measurement were defined as being in labor (term labor group) and those delivering outside 24 hours from the measurement as not being in labor (term nonlabor group).

Different cutoff measurement-to-delivery intervals for term and preterm labor vs nonlabor groups were chosen based on previous studies, which showed that an increase in uterine EMG activity occurs within approximately 24 hours from delivery at term and within several days from delivery preterm.

All women included provided written informed consent for study participation. Data from patients who ultimately underwent cesarean section were not used for analysis ( Figure 1 ). We chose to exclude the cesarean section patients because the decision on when exactly the surgery will be performed is based on several considerations, including those on fetal well-being and the subjective assessment of labor progress. This decision is therefore too subjective and arbitrary to include in a proper receiver-operating characteristics (ROC) analysis in which one wants to also accurately determine the mean measurement-to-delivery interval.

Flow sheet diagram for better visualization of study groups

EMG , electromyography.

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

We chose to include patients with preterm premature rupture of membranes (PPROM) because our objective was to evaluate whether uterine EMG measurements can differentiate between true preterm labor (ie, patients who are going to deliver spontaneously within a short period of time) and those who are not in true preterm labor, during which the clinical evaluation does not allow us to make this differentiation.

The St Joseph’s Hospital and Medical Center Institutional Review Board approved the study.

Uterine EMG signal recordings

Uterine EMG measurements were performed by 5 different researchers within 24 hours from the patient’s admission to the hospital. We standardized the electrode arrangement to the following factors: 2.5 cm electrode-electrode vertical and horizontal separation distances (measured from center to center) in a square-shaped pattern about the navel and with each electrode positioned in the vertex of each of the 4 corners of the square. Uterine EMG was measured for 30 minutes from each patient using a custom-built uterine EMG patient-monitoring system. Patients were asked to remain still while supine without disturbing any of the probes and wires for the recordings.

Signal analysis

Analog EMG signals were digitally filtered to yield a final band-pass of 0.34-1.00 Hz to exclude most components of motion, respiration, and cardiac signals from the analysis. Data were sampled at 100 Hz (this high sampling rate was chosen to increase the resolution of PS analysis later).

Propagation velocity analysis

PV can be calculated by dividing the distance that the propagating wave travels by the amount of time required for the propagating wave to traverse this distance. All the time differences in corresponding action potential peaks for each burst of action potentials were calculated, and the average of absolute values of all time differences for bursts in a patient’s uterine EMG recording was used to calculate the PV.

Only those pairs of peaks on different channels (electrode pairs) with congruent shapes and within 2 seconds temporal separation of one another were included in the analysis, ensuring that any peak observed on 1 channel during any 2 second window matched the associated peak on the other channel ( Figure 2 ).

Propagation velocity of EMG signals

The propagation velocity of the EMG signals was calculated from the time difference of the signal arrival at adjacent electrodes (t). Top trace shows a sample recording from 2 electrode pairs (channels 1 and 2). Note the excellent temporal correspondence between EMG and mechanical contractile events (measured by TOCO). Bottom trace shows the expanded EMG burst, with the individual voltage peaks clearly distinguishable.

EMG , electromyography; TOCO , tocodynamometer.

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

For our EMG instrument, we use differential, bipolar electrode pairs. The advantage of a differential bipolar setup over a monopolar setup is signal quality, allowing us to more accurately identify individual uterine voltage peaks. Only those bursts for which the mean voltage peak value was greater than 2 times the mean baseline voltage peak value were used in these calculations to clearly see and compare uterine voltage peaks at adjacent electrodes. Within each of these electrical contractile bursts, there were found anywhere from approximately 30 to 60 voltage peaks (associated with propagating voltage waves), which were analyzed ( Figure 2 ). More than 25,000 voltage peaks pairs were analyzed, with an average of approximately 215 peaks per patient.

Common obstetric measures

Presence or absence of contractions on TOCO at the time of EMG measurement was documented by the researcher recording the EMG. Transvaginal cervical length and Bishop score were also documented but only when they were assessed no more than 24 hours before or after the EMG measurement. Bishop score has not been developed as a predictor of preterm birth. Nevertheless, it gives a metric evaluation of digital cervical examination, which is commonly used to diagnose preterm labor and predict delivery. That is why we chose to compare predictive values of Bishop score with those of uterine EMG. Transvaginal cervical length was measured by 1 of 5 board-certified perinatologists or 1 of 4 certified ultrasound technicians. Digital examination was performed by 1 of 26 resident physicians involved in the care of the patients included in the study.

Statistics

A Student t test and a Mann Whitney U test (when appropriate because of nonnormal distribution of variables) were used to compare delivery within, vs outside, 24 hours from the measurement in term patients, and 7 days from the measurement in preterm patients. Statistical comparison between preterm and term patients was performed to determine whether gestational age has an impact on EMG PV. Data were analyzed by analysis of variance, and Dunn’s test was used for pair-wise comparisons among groups. Pearson’s correlation analysis was used to determine whether demographic and/or clinical parameters influence the PV. A P < .05 was considered significant.

The ROC curves were used to estimate the predictive values of the EMG parameters that were significantly higher in preterm patients delivering within 7 days and to assess the diagnostic accuracy of Bishop score, contractions on TOCO, and transvaginal cervical length for predicting preterm delivery within 7 days. Diagnostic accuracy of EMG, Bishop score, TOCO, and the transvaginal cervical length were then compared.

The ROC analysis was also performed on the cohort of patients for whom the results of all the examined methods were available. The ROC curves constructed on this cohort were compared by Student t test.