Spontaneous preterm birth is a heterogeneous phenotype. A multitude of pathophysiologic pathways culminate in the final common denominator of cervical softening, shortening, and dilation that leads to preterm birth. A precise description of specific microstructural changes to the cervix is imperative if we are to identify the causative upstream molecular processes and resultant biomechanical events that are associated with each unique pathway. Currently, however, we have no reliable clinical tools for quantitative and objective evaluation, which likely contributes to the reason the singleton spontaneous preterm birth rate has not changed appreciably in >100 years. Fortunately, promising techniques to evaluate tissue hydration, collagen structure, and/or tissue elasticity are emerging. These will add to the body of knowledge about the cervix and facilitate the coordination of molecular studies and ultimately lead to novel approaches to preterm birth prediction and, finally, prevention.
Spontaneous preterm birth (sPTB), the leading global cause of neonatal death, affects 13 million babies annually. Premature babies who survive are at risk for serious complications that include cerebral palsy, respiratory morbidity, mental retardation, blindness, deafness, cardiovascular disease, and cancer. This problem has been the object of considerable research efforts for decades, yet the singleton sPTB rate remains unacceptably high, both in the United States and throughout the world. A prevailing theory is that the approach to research and treatment has been simplistic, although the pathways to sPTB are extremely complex.
SPTB is the final common denominator of the interaction of a multitude of factors such as social stress, infection/inflammation, poor nutrition, genetics, and others. Studies of biomarkers, genetics, and “-omics,” although adding essential information to the body of knowledge about sPTB, have been disappointing thus far with regard to its prevention. Of >100 biomarkers that were evaluated in a recent metaanalysis, none emerged as reliably predictive, and the assessment of biomarkers for early intervention has not decreased the incidence of sPTB. Associations between sPTB risks and polymorphisms in candidate genes have been modest, and “-omics” studies (genomics, proteomics, transcriptomics, metabolomics) have provided no concrete associations with sPTB. This is not unexpected, however, given the multiple and varied pathways to sPTB.
Fortunately for study, the pathways to sPTB dovetail into final unifying processes such as cervical remodeling. As recently stated by Global Alliance to Prevent Prematurity and Stillbirth : “The finding that heterogeneous origins result in common downstream biological pathways and outcomes … provides the opportunity to develop rational treatment strategies that target the upstream initiators … .” In other words, identification of cervical microstructural changes during specific stages of remodeling should facilitate the study of isolated molecular events and interactions at critical time points.
The role of the cervix in preterm birth
Currently, a short cervix in the second trimester is our best predictor of sPTB. But treatment for the short cervix remains controversial, despite >600 publications in the past 2 decades on the relationship of sPTB, the short cervix, and proposed interventions. Cerclage for a cervix of ≤25 mm in an unselected population demonstrated no benefit in a metaanalysis; however, cerclage does reduce the risk of sPTB by 30% if there is also a history of sPTB. Two recent randomized controlled trials of progesterone for a short cervix in an unselected population had opposite results: intramuscular progesterone for a cervix of <30 mm demonstrated no benefit, but vaginal progesterone gel for a cervix of 10-20 mm was associated with a 45% reduction in sPTB. In February 2012, however, the Food and Drug Administration found the latter study insufficient to support approval of this gel.
This is unsurprising, given the conflicting data about the effect of progestins on cervical length. Some studies suggest that progestins attenuate shortening ; other studies suggest that progestins do not. Further, the variety of doses, administration routes, and formulations makes it difficult to reach definitive conclusions. Therefore, although the measurement of cervical length has become a fundamental part of clinical practice, the cervix remains mysterious. Most women with a short cervix (but no history of sPTB) deliver at term without intervention. The risk reduction for those who have any intervention is usually modest, and the vast majority of preterm births in low-risk women occur in those with a normal midtrimester cervical length.
Perhaps our confusion can be blamed on the complexity of the cervix itself. This remarkable structure, with its diametrically opposed functions, is comprised of interwoven layers of collagen that remodel independently and progressively throughout gestation via, presumably, different molecular processes ( Figure 1 ). Shortening at term is associated with differential expression of 687 genes, and preterm shortening appears to be even more complicated. Further, by the time shortening is recognizable, marked microstructural change has already occurred. Much less is understood about cervical softening than shortening, although the latter is arguably more critical. Softening starts soon after conception and occurs progressively throughout pregnancy. Testimony to its importance is that uterine contractions do not effect delivery if the cervix is firm ; a soft cervix is associated with preterm delivery, even without contractions. Yet, clinical assessment is entirely subjective; the cervix is labeled “soft, medium, or firm” based solely on digital examination.
Current data suggest that alterations to collagen processing, assembly, and structure are primarily responsible for cervical remodeling. There are 4 stages, each associated with specific collagen alterations: (1) softening, (2) shortening and marked softening (“ripening”), (3) active dilation during labor, and (4) recovery after delivery. The inability to describe microstructural change objectively within each of these stages limits our ability to target the associated molecular processes. Ultimately, this makes it nearly impossible to conceive of novel approaches to prediction and treatment. This review discusses emerging methods for the objective assessment of the pregnant cervix.
Approaches to assesssment of cervical microstructure
Tissue hydration, collagen structure, and tissue elasticity all change progressively with cervical remodeling: hydration increases, collagen disorganizes, and elasticity (softness) increases. Current approaches to microstructural assessment attempt to assess hydration, to describe various aspects of collagen structure, and/or to measure directly some aspect of tissue elasticity. Methods are grouped by approach.
Tissue hydration
Cervical surface area
Cervical tissue hydration increases as pregnancy progresses. Measurement of cervical surface area indirectly evaluates that property. In a rat model, digital images of the cervix were captured from which cervical surface area was calculated ( Figure 2 ). The surface area of the cervix increased throughout gestation, which led the authors to hypothesize that this may be useful for the evaluation of human cervical remodeling.
The advantage of this technique is its ease of use. The main disadvantage is noted by the authors: all the “folds and furrows” of the rat cervix cannot be measured, which makes the calculation imprecise. This also makes it unclear how well it would translate to the human cervix.
Electrical impedance
Flow of electrical current is affected by tissue hydration. Impedance to flow can be measured by the placement of electrodes on the cervix. Good correlation between impedance and cervical consistency was demonstrated in hysterectomy specimens, and a significant difference was found between pregnant and nonpregnant tissue. However, a recent study demonstrated low predictive values, which led the authors to conclude that the method “offers no immediate clinical utility.”
Acoustic attenuation
Attenuation is the loss of ultrasound signal amplitude with depth as a function of ultrasound frequency. Attenuation should decrease with hydration. Forty-one women (10-41 weeks’ gestation) underwent transvaginal ultrasound imaging. Images were converted to raw radiofrequency echo signals. Attenuation estimates were calculated within regions of interest that were chosen manually in areas that “appeared to be homogeneous” because tissue inhomogeneity violates the assumptions of the attenuation algorithm ( Figure 3 ).
Attenuation was a predictor of interval from ultrasound examination to delivery, but not of gestational age or cervical length. The study was not powered to look at attenuation as a predictor of sPTB; however, its potential was suggested by 2 cases of women with a short cervix but moderate-to-high attenuation values at 18 or 29 weeks’ gestation, respectively, both of whom delivered at term. The authors suggested that attenuation may be a better predictor of sPTB than cervical length.
Advantages of this method are that it is quantitative and relatively easy. Disadvantages noted by the authors are considerable between-subject variability and inability to standardize comparisons between subjects (because regions of interest are selected manually). A seemingly more significant issue is discussed by Parra-Saavreda et al ; tissue inhomogeneity violates the assumptions of their attenuation algorithm, but the cervix is definitely not homogeneous.
Cervical stromal differentiation
On T 2 -weighted magnetic resonance images, the signal intensity difference between the inner (near the canal) and outer cervix diminishes with remodeling as tissue hydration increases. A prospective cohort study of 100 women who were admitted for preterm labor at 18-34 weeks’ gestation determined “stromal differentiation” as high, intermediate, or low on the basis of visual inspection of signal intensity difference between the inner and outer cervix. Low differentiation was associated with shorter cervical length, and low or intermediate differentiation was associated with a higher risk of preterm delivery. This method is relatively easy; however, its subjectivity, low sensitivity, and negative predictive values and the fact that cervical length via ultrasound imaging was as predictive led the authors to conclude the “clinical usefulness of this test is limited by practical and cost issues.”
Collagen structure
Cervical gland area
Absence of normal mucosal glands (cervical gland area), which is demonstrated by a hyperechoic or hypoechoic segment around the cervical mucosa has been proposed as a predictor of sPTB. Six hundred pregnant women were scanned at 16-19 weeks of gestation. Cervical gland area was detected in 77% of the women who delivered at term vs 55% of those who did not ( Figure 4 ).
The advantage of this method is that it uses standard imaging techniques. The disadvantages are the relative subjectivity of ascertaining the presence of mucus glands (limits of detection depend on the system, transducer, and scanning technique). The other disadvantage is low performance; cervical gland area was detected in most of the women in their study, irrespective of delivery timing.
Light-induced fluorescence
Light-induced fluorescence measures the natural fluorescence of nonsoluble (cross-linked) collagen in the cervix. As the cervix remodels, the amount of cross-linked collagen decreases; a collascope can measure this change. The instrument is a steel probe that is attached by a fiber-optic cable to a large main unit, which is attached to a computer. The probe, when placed against the anterior lip of the cervix, delivers excitation light onto the cervix and carries the fluorescent light back to the instrument; the computer displays the spectrum of fluorescence.
The collascope was used weekly in 21 pregnant patients beginning at 24-40 weeks’ gestation and in 29 patients in labor. Light-induced fluorescence was correlated (weakly) inversely with gestational age and correlated(weakly) with time to delivery. Another study enrolled 41 patients who were admitted for induction of labor. Light-induced fluorescence and Bishop scores before and 4 hours after prostaglandin application were recorded. There was no correlation between initial light-induced fluorescence score and Bishop score, but there was an inverse relationship between final light-induced fluorescence and Bishop scores. Eleven of 22 patients with initial low light-induced fluorescence score showed no change after ripening, in contrast to 16 of 19 patients with a high preripening light-induced fluorescence score. The authors suggested that the technique may help distinguish women who would benefit from preinduction ripening.
The advantage of this technique is its ability to assess cervical crosslinking objectively, which is a measure of collagen organization. A disadvantage is that it requires a specialized, fairly cumbersome apparatus.
Second harmonic generation (SHG)
SHG is a nonlinear microscopic method for imaging collagen. In ex vivo mouse cervices, SHG identified changes in signal intensity and collagen fiber size that accompanied normal remodeling changes ( Figure 5 ). The advantage of this method is that it directly visualizes collagen microstructure. Its disadvantage is that it would require the development of an endoscopic microscope for in vivo use.
Raman spectroscopy
This optical technique measures subtle microstructural component changes that are based on the Raman effect, which is a description of energy exchange between photons and scattering molecules. In a windowless room (no ambient light), energy (light) is sent to the tissue through a fiber-optic probe. When a photon collides with a molecule, the energy causes a change in the molecular vibration; energy differences between the incident and the scattered photons are indicative of particular molecules. Therefore, the spectrum of scattered photon energies provides information about the composition of tissues and allows observation over time (Raman shifts) of specific components such as collagen.
Investigators obtained measurements from the distal mouse cervix throughout gestation. Trends early in pregnancy followed a logical process, but those late in pregnancy were difficult to recognize and were without a clear pattern. Much greater variation was noted in a pilot study in humans.
The advantage of this method is the precision with which it may be able to identify collagen changes. A disadvantage is its need of highly specialized equipment and stringent conditions. Also, the authors note that the significant variability that is noted in women could hinder the ability to find clear Raman shifts of statistical significance.
Backscattered power loss (BSPL)
When an ultrasound wave encounters a scatterer (eg, collagen), the modes of vibration excited in it causes backscatter (“echo”); these modes can vary as the incident angle of the acoustic beam changes. Measurement of BSPL as a function of the beam angle thus provides information about the shape and organization of scatterers, such as collagen. Raw radiofrequency echo signals are recorded while the ultrasound beam is steered electronically; the power spectrum of the echo signal is computed for each angle to estimate the BSPL. We have used this technique to assess cervical microstructure in (ripened vs unripened) human hysterectomy specimens, corroborating findings with micron-scale maps of collagen distribution and alignment from SHG microscopy images. We find that BSPL is sensitive to collagen microstructure organization; when the ultrasound B-mode images are spatially registered and superimposed with the SHG images and quantitative ultrasound measurements, the BSPL correlates well with collagen fiber organization and orientation ( Figure 6 ).
The advantage of this quantitative method is that it seems able to detect small changes in tissue properties within heterogeneous areas. The disadvantage is that the current implementation uses a prototype transducer and specialized software. Also, progress toward a viable clinical tool has been slow because of the need to map out the entire cervix to determine the most appropriate and reproducible location(s) for measurement.
Tissue elasticity
Cervical consistency: cervical consistency index or mean gray-level histogram
These techniques address cervical softening. The cervical consistency index evaluates compressibility (softer tissue deforms more easily than stiffer). An image of the cervix is obtained in the standard manner, then a second image is obtained after pressure is applied with the transducer until no more shortening of the anterior-posterior diameter is observed (eg, the cervix will not compress further). The ratio of the anterior-posterior diameter on the before and after deformation image is the cervical consistency index and a lower cervical consistency index indicates softer tissue ( Figure 7 ). In a cross-section of 1031 women who were measured at 5-36 weeks’ gestation, a lower cervical consistency index was seen with increasing gestational age; the sensitivity for prediction of preterm birth at <32, <34, and <37 weeks’ gestation was 67%, 64%, and 45%, respectively, which was superior to the sensitivity of prediction by cervical length. They proposed that this technique may be useful for preterm birth prediction.
The mean gray-level histogram addresses differences between the anterior and posterior cervical wall. From a B-mode image of the cervix that is obtained in the usual manner, a grayscale histogram is developed from a midsection of the anterior and posterior cervix, and the 2 histograms are compared ( Figure 8 ). According to the authors, a greater difference in echogenicity between the anterior and posterior region of interest indicates a “harder” cervix. Measurements were performed in 214 women at 27-30 weeks’ gestation. A greater anterior-posterior difference was associated with a lower Bishop score, and the method had a sensitivity of 71% to identify a hard cervix. The authors suggested that this measurement may allow evaluation of cervical consistency.
An advantage of both methods is their simplicity; they use standard transvaginal cervical imaging. The primary disadvantage is their relatively low performance. For mean gray-level histograms, this may be related to a common drawback of grayscale histograms: adjustments to most front panel controls on an ultrasound system can significantly change the values. Also, there is little physical relationship between the shear modulus (stiffness) of a material and its acoustic scattering properties (that cause echogenicity in grayscale images).
Elastography
This method is based on the determination of motion in areas of the cervix relative to other areas, which is described by a color map. The cervix is scanned in the usual manner; slight pressure is applied with the transducer to deform the tissue, and specialized software is used to produce a color map that describes deformation of the tissue relative to neighboring areas ( Figure 9 ). In one protocol, points are assigned to the map to obtain an elastography index: purple (0, hardest) → blue (1) → green (2) → yellow (3) → red (4, softest). Twenty-nine women who underwent term labor induction were evaluated. The mean elastography index of the internal os demonstrated a statistically significantly different elastography index of 1.32 vs 0.39 for patients with successful vs failed induction, respectively. The elastography index from tissue near the external os and the middle of the canal was not predictive. The authors suggested that elastography may guide decisions about need for cervical ripening before labor induction but note that this would be most useful if measurements could be standardized between subjects.
A reproducibility study was performed in a cross-section of 112 pregnant women at all points in gestation. There were no statistically significant differences between cervices, except in the area that received direct force from the transducer; the authors concluded that measurements may be “a mere reflection of the force being applied by the transducer.” They cautioned that it is premature to conclude that “measurements of rate-of-change in tissue displacement reflect histological changes that could provide a measure of cervical ripening.”
The major advantage of elastography is its ease; although specialized system software is required, the technique is compatible with standard equipment. The disadvantages are that measurements are dependent on transducer pressure by the operator; thus, there is no control for variability or adequate means for standardization.
Aspiration
Tissue compliance is related to tissue softness; thus, tissue displacement provides a measure of softening. One group uses a 10-mm diameter tube pushed against the distal end of the cervix while a vacuum sucks tissue into it and an optical fiber illuminates the inner part of the tube. Images of the side view of the tissue before and after suction are reflected by a mirror, and displacement is calculated from the images.
Initial studies noted marked variability among subjects, which suggests that the method may not be useful. However, a more recent study demonstrated successful measurement in 12 of 20 subjects (technical difficulties inhibited the others), and a 20% decrease in stiffness was shown 1 month later in all 3 of the women who had subsequent evaluation.
The advantage of this quantitative method is its potential for monitoring incremental tissue changes. A disadvantage that was noted by the authors is that there is currently no way to standardize the force that is applied by the operator. Further, they indicate concern that the measurement cannot address cervical heterogeneity adequately.
Shear wave speed (SWS)
Measurement of SWS provides an objective description of tissue softness and microstructure because waves travel more slowly in softer tissue and shear modulus (stiffness) is dependent on fiber (eg, collagen) organization. Our methods are a hybrid of techniques for on-line data acquisition and off-line motion tracking, image formation, and parameter estimation. SWS are slower in ripened vs unripened cervical tissue from hysterectomy specimens; when SWS measurements are registered spatially and superimposed with ultrasound B-mode images and BSPL measurements, the result provides an objective description of tissue softness ( Figure 10 ).
