Cervical length as a predictor for spontaneous preterm birth in high-risk singleton pregnancy: current knowledge
Epidemiology and definitions
Preterm birth is a major contributor to the global burden of disease1: it is the leading cause of neonatal mortality worldwide and has effects on survivors that may be lifelong. These include increased susceptibility to early severe infection, neurological impairment, psychiatric or behavioral issues and elevated risk of non-communicable diseases in adult life2. The scale of the problem and the severity of the consequences of prematurity have spurred extensive research into potential causes and predictors of preterm birth and methods of prevention; however, many questions remain unanswered.
Premature or preterm birth, by definition, occurs prior to 37 completed weeks of gestation, and may be further categorized into extremely premature (< 28 weeks' gestation), severely premature (28–31 weeks), moderately premature (32–33 weeks) and near term (34–36 weeks)3. While a significant proportion of preterm births are medically indicated due to fetal or maternal requirement (such as intrauterine growth restriction or pre-eclampsia), the majority are spontaneous3. Spontaneous preterm birth (SPTB) may be considered according to the status of the amniotic membranes at the onset of labor, i.e. with membranes intact, or with preterm prelabor rupture of the membranes (PPROM). The incidence of preterm birth is increasing for two reasons: the rising number of medically indicated deliveries in singleton pregnancies and a higher rate of multiple pregnancy due to more widespread use of assisted reproductive technologies3. The ability to identify women who will deliver prematurely is important, as it allows targeted administration of effective treatments, such as vaginal progesterone4-6, steroids7, and magnesium sulfate for neuroprotection8. Both progesterone pessaries and cervical cerclage have been shown to reduce the incidence of preterm birth and improve neonatal outcome9; however, cerclage in particular carries significant risks, including those of PPROM, preterm labor and, rarely, maternal septicemia10.
Pathophysiology
Comprehending the pathophysiology of SPTB will pave the way to precise methods of prediction and identify appropriate therapeutic targets for prevention. SPTB was once thought to be due to the inability of the cervix to remain closed and maintain the pregnancy, from which arose the term ‘cervical incompetence’. Now more commonly known as ‘cervical insufficiency’, this is a diagnosis applied in retrospect, in which progressive cervical shortening and painless dilation have resulted in recurrent mid-trimester pregnancy losses or preterm births, possibly preceded by prolapsing membranes or PPROM9, 11. The advent of transvaginal sonography (TVS) allowed measurement of the shortened cervix, which was believed initially to be evidence of the proposed inherent functional or structural inadequacy12; however, over time this paradigm has shifted quite dramatically.
Increasingly, ongoing research suggests a limited role for the concept of true cervical insufficiency as a cause of SPTB, except perhaps in cases of inherited defects in collagen synthesis (such as Marfan syndrome), which are associated with an increased risk of SPTB13. Indeed, if SPTB were caused uniformly by a mechanical failure of the cervix, then structural reinforcement via cervical cerclage ought to be a panacea14. Instead, it would seem that a shortening cervix is evidence that parturition may commence long before signs and symptoms of labor appear. This is certainly the case for term labor, in which the cervix has softened and ripened over the preceding weeks, then decidual activation occurs, and myometrial contractions follow (although with considerable variation in the timing and overlap of these processes)15.
Romero et al.16 described recently the preterm parturition syndrome, a heterogeneous condition with preterm labor as the common endpoint. They propose that, rather than preterm labor being a process that is abnormal only in its timing, it is a result of pathological activation of one or more of the signals that subsequently initiate labor. It appears that this may be due to various etiologies, which could trigger a range of different components that converge ultimately on a shared pathway. (In contrast, generally, term labor is initiated by physiological activation of the pathway.) For example, complement activation is implicated in the cervical ripening process in preterm but not in term birth17. The pathological processes involved in the preterm parturition syndrome fall into the broad categories of infection/inflammation, cervical disease, vascular disorders, decidual senescence, uterine overdistension, decline in progesterone action, stress and breakdown of maternal–fetal tolerance18. This multifactorial model implies that, in many instances, the shortening cervix may be simply a sign that accompanies some of the abnormal processes leading to SPTB, rather than a cause in itself.
Risk factors for SPTB
Large epidemiological studies have identified a number of risk factors for SPTB, which help define a population who may benefit from closer monitoring. Some of the well-described risk factors include a history of SPTB19, previous surgery for cervical intraepithelial neoplasia (CIN)20, multiple previous cervical dilatations (e.g. via termination)21 and Mullerian anomalies22. Some factors may predispose to infection or inflammation, for example, smoking during pregnancy, bacterial vaginosis and, perhaps, a short interpregnancy interval23; however, causal links are generally unclear.
Ultrasound imaging of the cervix
TVS is the current standard for imaging of the cervix (Figure 1), having been shown to be superior to digital examination and transabdominal sonography, techniques which tend to overestimate cervical length (CL) (Figure 2), and have poor interobserver reliability24. The TVS technique has been well described and provides a standardized, reproducible measurement, with interobserver reliability of 5–10%25. A measurement is obtained over approximately 5 min to allow assessment of any dynamic shortening. The empty maternal bladder is visualized, along with the internal os, external os and cervical canal, with application of only minimal pressure required to gain a clear image (see images of normal CL in Figure 3 and short CL in Figure 4). Fundal pressure may be applied by the sonographer to assess the presence of dynamic changes in CL26. The presence of funneling and/or amniotic fluid ‘sludge’ may also be noted (Figure 5), as should any dilation of the internal or external ora (Figure 6).
Cervical-length screening at a single time-point
Iams et al.27 published a seminal work that established norms for CL at 24 weeks' gestation, identifying CL < 25 mm (the 10th centile) as a clinically important threshold in predicting SPTB, and CL at 24 weeks as the most predictive value. This multicenter, prospective study (the Preterm Prediction Study) assessed a sample of both low- and high-risk women, selected to represent the population presenting for antenatal care. CL < 25 mm features prominently in the literature as a threshold of significance for SPTB prediction, and has informed clinical practice. The positive predictive value (PPV) of a short CL, however, is poor when assessed in a normal antenatal population including low- and high-risk women: only 18% of women in their study with CL < 25 mm (at 22–25 weeks) delivered prior to 35 weeks' gestation27. Routine TVS screening, with use of vaginal progesterone in women with a shortened cervix to prevent SPTB, is reportedly cost-effective, but it is low in yield, requiring up to 588 women to be screened and treated to prevent one SPTB28. Although it has achieved wide acceptance as a predictor of SPTB, routine TVS screening in the general antenatal population is therefore not recommended29.
Because CL assessment by TVS cannot be utilized as a universal screening tool, a slightly different body of research has emerged, examining its value in prediction of SPTB in women considered at high risk of preterm birth. Durnwald et al.30 and de Carvalho et al.31 assessed the value of short CL as a single measure in combination with obstetric history during the second trimester of high-risk pregnancies. Durnwald et al.30 found similar rates of CL < 25 mm at 22–25 weeks' gestation in women with a history of one previous SPTB (at > 18 weeks but < 37 weeks) vs women with two or more SPTBs. With their small sample size they could not determine whether an increasing number of prior SPTBs significantly modified the value of a short cervix in predicting future SPTB before 32 or 35 weeks' gestation. They did not analyze the predictive value of CL < 25 mm in the cohort. de Carvalho et al.31 found, amongst a larger sample, a significantly shorter CL between 21 and 24 weeks' gestation in women with previous SPTB (at < 37 weeks) who went on to deliver prior to 34 weeks' gestation compared to women without a history of SPTB. CL < 25 mm had a PPV of 12.5% in women without a history of SPTB, but 33.5% in women with a prior SPTB, suggesting an additive effect of past obstetric history. A potentially important flaw in both these studies is that women who deliver prior to 20 weeks' gestation may do so for different reasons from those with a near-term birth, and what predicts recurrence in one subgroup may not be relevant to the other. Thus, the literature suggests that a single evaluation of CL will not identify accurately women who will proceed to a SPTB.
Cervical-length surveillance
The study of Iams et al.27 collected CL measurements at both 24 and 28 weeks' gestation. The results indicated that cervical shortening between these time points was significant in predicting SPTB, even independently of the initial CL measurement. This provided a basis for CL surveillance in a high-risk population, in whom cervical shortening may be expected.
Owen et al.32 performed a blinded observational study of CL surveillance in 183 women with history of one or more SPTBs as their only risk factor. This study examined the predictive value of a short cervix in the early second trimester (16–19 weeks) and whether serial measurements up to 24 weeks improved this. The rate of change in CL (cervical slope) throughout the surveillance period was shown to be a risk factor for a further SPTB, with women who delivered prematurely showing a more rapid rate of shortening. Even after controlling for a short baseline CL measurement, a steeper cervical slope predicted SPTB.
Guzman et al.33 conducted a similar study with the aim of establishing norms for rates of cervical shortening amongst women with a ‘competent’ or ‘incompetent’ cervix between 15 and 24 weeks' gestation. They too reported a more rapid rate of shortening (−0.49 to −0.80 cm/week) amongst women in their incompetent-cervix group (the competent-cervix group, in contrast, had a plateau period initially, with stable CL, followed by a later decline in CL), along with more mid-trimester losses. A limitation of this study was that the incompetent-cervix group received ultrasound-indicated cerclage, which may affect the natural history of cervical shortening34 and therefore alter the predictive value of shortened CL.
In another confirmatory study, Iams et al.15 performed secondary analysis on data collected for the Preterm Prediction Study, and also found that the rate of cervical shortening was higher in women who went on to have SPTB. CL measurements were taken at a later gestational age, at three visits between 22 and 29 weeks; however, women who delivered prior to 28 weeks' gestation were excluded from the analysis. The authors used their findings to contend that premature parturition is a prolonged process that begins well before 24 weeks' gestation, and is evidenced by a faster rate of change of CL compared with the rate in women who deliver at term. This may be true for SPTB in this subgroup of women, and is a useful contribution to our knowledge of the role of cervical shortening in a certain phenotype of SPTB. However, eliminating women who delivered prior to 28 weeks' gestation potentially excludes a different phenotype of SPTB, in which cervical shortening occurs very acutely; for example, there have been reports of women undergoing surveillance who had normal CL only 6 days prior to spontaneous delivery at 27 weeks' gestation35.
Overall, it would seem that serial CL measurements are more likely than is a single assessment to identify high-risk women who will go on to deliver prematurely. This is, however, with the caveat that phenotypes of SPTB vary, and cervical shortening and preterm labor may occur acutely, perhaps even between frequently scheduled visits.
Due to its utility in predicting SPTB, CL surveillance has been used in a number of studies as an alternative management strategy to universal prophylactic cerclage in high-risk pregnancy, and as a result is an accepted alternative to history-indicated cerclage29. Cerclage rates varied between 32%36 and 59.6%37 in the surveillance groups reported in the literature, which represents a substantial reduction compared with previous practice. CL surveillance allows more judicious application of a treatment that bears considerable risks – albeit at low frequency – for the woman and fetus10.
Cervical-length surveillance in women with a history of SPTB
A key risk factor for SPTB is a previous preterm birth, with reported recurrence rates of up to 50%38, although not necessarily in the immediately subsequent pregnancy. Earlier gestational age at previous delivery elevates the risk further, and some studies report that increasing numbers of prior SPTBs do likewise23. Yost et al.39 found that an increased number of previous SPTBs did not add to the predictive value of a short mid-trimester CL, which is consistent with the idea that obstetric history alone is insufficient to predict recurrent SPTB. In a secondary analysis of the same cohort of women with previous SPTB, Owen et al.40 noted that those with an early short cervix (≤ 30 mm at 16–18 weeks' gestation) or who experienced early shortening (before 22 weeks) were more likely to deliver before 26 weeks' gestation. A small proportion of women with a long cervix, or shortening at 22–24 weeks, however, still went on to deliver extremely prematurely, which could be the result of a different phenotype of SPTB; a lack of delivery information makes this difficult to ascertain. These studies highlight that, even amongst women with a previous SPTB, CL surveillance is important, but does not capture perfectly all of those who will benefit from treatment to prevent subsequent SPTB.
Cervical-length surveillance after surgery for cervical intraepithelial neoplasia (CIN)
Treatment for cervical dysplasia is also a risk factor for SPTB, with an apparent association between SPTB and cone biopsy, and possibly large loop excision of the transformation zone (LLETZ, also known as loop electrosurgical excision procedure, or LEEP). An early view was that the physical removal of tissue from the cervix compromised its structural and functional integrity, causing a shorter CL during pregnancy. A small prospective study questioned this theory41: the CL in non-gravid women prior to and ≥ 3 months after LLETZ demonstrated a negligible difference (mean difference, 0.0 ± 0.4 cm). Nevertheless, a number of studies have reported an increased risk of SPTB associated with the LLETZ procedure42-45. An alternate theory is that excisional treatment may remove some of the cervical glands responsible for producing the mucous plug, which, in its intact state, has a role in preventing infection46. Importantly, though, Bruinsma and Quinn47 found that there is an increased risk of SPTB in women with cervical dysplasia, whether treated or untreated, suggesting an underlying pathological process in this group of women, perhaps irrespective of CL.
Only two studies have utilized CL surveillance in women who have undergone surgery for CIN. Pils et al.48 measured CL at fortnightly intervals from 16 until 22 weeks' gestation in women with a previous history of one or more LLETZ procedures (although other risk factors for SPTB were present in some), comparing their findings to those of a control group with a history of SPTB. The rate of change of CL was found to be highly significant in the cervical-surgery group, but adding the serial TVS measures to multivariate analysis showed that the subsequent measurements did not add to the predictive value of a short cervix at 16 completed weeks' gestation. The authors argued in favor of the increased rate of SPTB being caused by a short cervix as a consequence of surgery in this population; however, the cut-off that emerged as significant was CL < 30 mm, compared with < 25 mm in most other studies. This study suggests that women who have undergone one or more LLETZ procedures may only require a single TVS CL assessment at 16 weeks rather than undergoing many weeks of surveillance.
Berghella et al.49 performed surveillance between 16 and 24 weeks' gestation in 109 women after surgical treatment for CIN, and analyzed their findings according to procedure type. They defined a short cervix as < 25 mm, in keeping with studies in other high-risk groups. Cone biopsy emerged as a significant risk factor for SPTB, but LLETZ did not. Unfortunately, the authors did not report on the significance of cervical slope, or on the optimal gestational age for prediction of SPTB, which may have strengthened the potential clinical implications of the Pils et al.48 study. They did note a higher PPV (64%) of CL < 25 mm in women with prior cone biopsy compared with other populations.
Other studies in this group of women used only a single measure of CL, and many reported CL < 30 mm, rather than < 25 mm48, 50-52, as being significant for SPTB prediction. If the cause of SPTB in women with previous excisional cervical surgery was indeed due to loss of cervical volume, then the CL predicting SPTB ought to be shorter than in women with different risk factors for SPTB. A retrospective case–control study by Nam et al.53 (again, using only one TVS CL assessment), that also collected data on the volume of cervix excised, found no significant link between the size of the biopsy and the likelihood of SPTB. As in the majority of studies in other groups of high-risk women, the authors found that CL was the only factor associated significantly with premature birth53.
Using a heterogeneous study group in which one of the exposures is of arguable significance makes results difficult to interpret, as does the choice of control group in some cases. Indeed, the one study that used women who had undergone only LLETZ to treat their cervical dysplasia54 found that LLETZ was not a significant independent predictor of SPTB once adjustment for CL was carried out. Pils et al.48 utilized women with a history of previous SPTB as controls for their cohort, but it is possible that the women in the control group who delivered prematurely did so via a different pathological process from those in the surgery group, and to compare them is of dubious value.
The literature does not clearly support LLETZ as an independent risk factor for SPTB in women with CIN. Data on CL surveillance in this group may support a longer CL cut-off (< 30 mm) compared with that in other high-risk groups, and may even support a one-off CL assessment at 16 weeks' gestation, instead of serial measurements. Further study is required to confirm these findings.
Cervical-length surveillance in women with uterine anomalies
During embryonic development, failure of Müllerian duct differentiation can lead to a unicornuate uterus, while abnormal unification of the paramesonephric ducts may result in a bicornuate or septate uterus or even two separate uteri (uterus didelphys)55. Suggested potential causes for SPTB amongst women with uterine anomalies include decreased muscle mass, abnormal uterine perfusion and cervical insufficiency56. Congenitally abnormal uteri have thinner walls, which may produce relative uterine overdistension as gestation progresses, leading to early activation of myometrial stretch receptors. The concept of a mechanical deficiency of the cervix may be more plausible in the context of an organ that is anomalous in development of connective tissues or in even distribution of progesterone receptors, as well as in gross anatomy.
The likelihood of SPTB appears to vary according to the type of uterine anomaly, with a small cohort study (n = 64) by Airoldi et al.57 reporting that a septate uterus gave the best odds of a term delivery, while a unicornuate uterus, with a preterm birth rate of 44%, gave the poorest. Notable in this cohort were two women with uterus didelphys who delivered prematurely in the absence of a short cervix. No further delivery data were provided, so it is unknown whether other factors may have contributed to these SPTBs. Some of the study population also had other risk factors, including previous cervical surgery and use of assisted reproductive technologies. If any of these factors lead to SPTB via a different mechanism from that caused by uterine anomaly, then these are important potential confounders. Although CL < 25 mm in the presence of any uterine malformation had a 50% PPV for delivery < 35 weeks in this particular study, this may not reflect the true association.
Crane et al.58 performed the only other study examining specifically CL in women with uterine anomalies. This study found a CL < 30 mm between 16 and 30 weeks' gestation to be a significant cut-off value; below this threshold the PPV was clinically unhelpful (37.5%), but the 100% negative predictive value (NPV) would provide useful reassurance for women undergoing surveillance, and help to avoid unnecessary prophylactic treatments. They reported that women with a bicornuate or unicornuate uterus tended to have a shorter CL, and that shortening occurred earlier in women with uterus didelphys or uterine septum. Their sample included only two women with a septum and two with unicornuate uterus, however, which makes generalizing from these findings inappropriate and highlights a need for more research in this area. The group with uterine anomalies also had historical risk factors for SPTB, whereas the control group did not, and pregnancies complicated by antepartum hemorrhage remained in the study. Antepartum hemorrhage is a separate risk factor for SPTB that appears to operate independently of cervical shortening59, so inclusion of these cases may have reduced artificially the PPV of short CL in this population.
While CL surveillance appears to have a role in this high-risk group of women, there is a paucity of literature in the area. Agreement is lacking regarding a significant cut-off for short CL, and it is unclear whether a short cervix carries a higher PPV for SPTB in women with uterine malformations than it does in other high-risk groups.
Cervical-length surveillance after multiple cervical dilatations
Numerous studies have observed a relationship between a previous history of termination and subsequent SPTB60-62; however, we could identify only one study that examined the predictive value of serial TVS in this population. Visintine et al.63 analyzed retrospectively CL surveillance data for 65 women with a history of one or more prior terminations and found the PPV for a cervix < 25 mm to be 47% for delivery before 35 weeks' gestation. The NPV of a cervix ≥ 25 mm was 86% in this population, leaving a small but significant group of women who will deliver prematurely in the absence of a short cervix. This suggests that CL is moderately predictive of SPTB in women who have had one or more terminations; however, replication of these findings is needed to guide practice in this high-risk group.
Cervical funneling
All the studies discussed so far found a significant association between short CL (usually < 25 mm) and SPTB, yet all showed a low PPV, which limits the utility of CL as a single predictor of SPTB. In an effort to find a more accurate predictor of SPTB, some authors have therefore looked to combine CL measurement with other markers, including biological indicators as well as cervical funneling, another TVS finding.
Cervical funneling is defined as dilation of the internal os of the cervix, with protrusion of the amniotic membrane 5 mm or more into the endocervical canal64. It provides sonographic evidence of a precocious cervical ripening process, whereby the internal os and some of the cervical canal opens, with progressive opening towards the external os. This has been described in terms of the shapes observed on ultrasound: first a ‘T’-shape before any dilation occurs; then a ‘Y’-shape whilst some closed length of cervix remains; as effacement continues, the canal opens almost to the external os, and the cervix appears as a ‘V’; finally it appears as a ‘U’ as dilation progresses65 (Figures 5 and 6).
Although several studies have included cervical funneling as an outcome21, 31, 32, 64, 66, 67, in general it was not found to be an independent predictor of SPTB, nor did its presence add to the ability of CL to predict SPTB. Andrews et al.68 found that funneling before 20 weeks' gestation was associated with delivery within 2–4 weeks and before 35 weeks' gestation, although only 53 women were examined. Cook and Ellwood69 studied prospectively 120 high-risk women and developed a descriptor of an ‘abnormal cervix’ based on a composite of CL < 30 mm and internal os dilation > 5 mm (effectively denoting a short and funneled cervix). They found that the addition of funneling did not add to the predictive value of a short cervix, still only providing a PPV of 56% for delivery before 34 weeks' gestation.
Berghella et al.64 observed that cervical funneling was associated consistently with a short cervix, and appeared to be associated with an earlier preterm birth; however, they noted technical difficulties in obtaining reliable measures of funneling. A recent survey of Australian specialists and sonographers70 confirmed that, while CL measurement is a well-described procedure with good interrater reliability, agreement between raters is poor when funneling is present. This presents an obvious impediment to assessing the use of cervical funneling as a predictor of SPTB.
Amniotic fluid ‘sludge’
Kusanovic et al.71 used CL measurement as a comparator for the predictive value of amniotic fluid ‘sludge’ noted on TVS. Sludge, usually lying near the internal os, is intra-amniotic debris: a collection of particulate matter composed of biofilm and inflammatory cells72, sourced either from ascending infection or the placental microbiome73, and indicating chronic intra-amniotic infection. Infection has been recognized increasingly as a cause of SPTB, although most of the studies reviewed here did not collect data on this. The presence of sludge on TVS has been found to be associated with a shorter CL, PPROM and earlier gestational age at delivery74. This is an important finding in light of the study of Odibo et al.75, which looked for factors predicting repeat SPTB and derived different significant cut-off values for CL in PPROM (25 mm) vs preterm labor with intact membranes (15 mm). These studies support the concept of different pathogenesis of the two subtypes of SPTB, and also underscore the need for research that separates SPTBs according to type, as different tests may be more useful and predictive in one or other clinical scenario, while the effect may be obscured by combining the two groups.
Fetal fibronectin (fFN)
Another marker that has been incorporated widely as a test for preterm labor is fFN, a glycoprotein released from the interface of the decidua and chorion during preclinical labor. Two recent studies have compared and combined it with CL as a predictor of SPTB in asymptomatic women, and both found fFN to be a superior test. Zhou et al.76 found the presence of fFN to improve the predictive value of short CL, while Bolt et al.77 concluded that fFN was a superior test to which CL added no benefit. Zhou et al. used a single measure of CL, whereas the cohort of Bolt et al. attended a specialist clinic for surveillance, but only the shortest CL was used in the analysis. Given that cervical slope has been shown in other studies to be a better predictor of SPTB than is a single CL value, the former may have been a more useful measure for comparison with fFN.
Cervical-length surveillance following physical intervention
Physical interventions employed in women undergoing CL surveillance who demonstrate cervical shortening may include cervical cerclage or placement of a cervical pessary, although the role of the latter is likely to be limited78-80, with a recent large randomized controlled trial identifying no reduction in early SPTB rates following pessary insertion81. Total CL has been shown to predict SPTB < 34 weeks following both elective and emergency cerclage82. Perioperative cervical TVS has demonstrated that emergency cerclage results in an immediate improvement in the ultrasonographic status of the cervix83, with positive change in CL on serial assessment thereafter associated with a later gestational age at delivery84. Figure 7 shows the appearance on TVS of a cerclage in situ. In contrast, the presence of funneling is only predictive of preterm birth following elective cerclage placement82, 84. fFN testing remains valid in the presence of a cerclage85, although has a higher false-positive rate than does testing without cerclage86. Sonographic assessment of the cervix with a pessary in situ can be technically challenging87, and may require placement of the probe in the center of the pessary to achieve adequate images88. The role of sonographic CL surveillance following pessary placement has not been elucidated, nor has fFN assessment been validated in patients with a pessary in situ.
Shortcomings in existing research
Studies investigating risk factors for SPTB, summarized in Table S1, are heterogeneous with respect to the sample population, treatment provided and outcomes of significance. The provision of antenatal care specific to the needs of women at high risk of SPTB provides convenient samples to study; the majority of the studies reviewed were performed in this setting. However, an accompanying difficulty is that many of these women undergo interventions aiming to reduce the risk of SPTB, such as cervical cerclage or tocolysis30, which can affect the usual prognosis of a short cervix. While some studies excluded women who received cerclage under any circumstances, others enrolled women with cerclage in situ, or continued to collect data from those who underwent the procedure during the course of the study. These retrospective cohort studies were typically not blinded, which adds to the risk of bias.
The populations of the reviewed studies were all described as being at high risk for SPTB, yet they were a heterogeneous group overall, as inclusion criteria varied from one study to another. Conditions known to increase the risk of preterm delivery, such as in-vitro fertilization89, multiple cervical dilatations63, previous excisional treatment for cervical dysplasia43 and uterine malformations22, were excluded from some, but not all, studies. Authors did not always report on the inclusion of women with certain risk factors, making comparison between studies difficult. Conversely, in some cases, researchers would aim to focus on CL surveillance in the presence of a specific risk factor for SPTB, but include women with multiple risk factors.
A known predictor of SPTB is racial origin: the rate of preterm birth amongst African-American or Afro-Caribbean women is 16–18%, compared with 5–9% in Caucasian women23. The underlying reason for this is unknown; it may be an effect that is completely independent of lower socioeconomic status, which is itself a risk factor for SPTB23. Race is likely to represent a confounding variable in many studies; for example, Macdonald et al.67 did not report demographic data, whilst other studies used different proportions of black women in their sample, from 86%66 through to none90. Yost et al.39 observed a higher proportion of African-American compared with Caucasian and Hispanic women in their early preterm birth group, but did not report whether this difference reached statistical significance. Odibo et al.75 noted black race to be a strong predictor of both PPROM and preterm labor < 35 weeks, but failed to show a tendency towards one pathway or another that was unique to African-American women.
A further issue is that the definition of preterm birth differs amongst studies, making comparisons more difficult. Some include only moderately and extremely preterm births as their primary endpoints, whilst others take a broader approach, using 37 weeks' gestation as a cut-off value. The largest burden of morbidity and mortality due to prematurity occurs prior to 34 completed weeks' gestation36, which makes this a clinically important threshold for use in research. It is possible that different pathophysiology underpins extreme vs late preterm birth, and that a particular assessment or treatment modality may not be relevant to all of the different mechanisms, which would favor a stratified approach to reporting and analyzing according to gestational age at delivery. Some recent studies (for example, Manuck et al.91) have attempted to describe SPTB according to different phenotypes, which may provide a useful structure for reporting further research into predictors of SPTB.
Conclusion
Prematurity is a major cause of perinatal mortality worldwide, but its pathogenesis remains elusive. Both clinical and laboratory-based research in the last two decades point to multiple pathways converging on a common endpoint of preterm labor. The difficulty with studying a single endpoint is that the distinct pathways may well be contingent on different initiators and risk factors. Adding to this issue, the existing body of research includes a heterogeneous population with a variety of risk factors for SPTB, many of which have not been examined individually. This may explain why CL consistently emerges as a significant, but frustratingly imprecise, predictor of SPTB in high-risk women. Although it could be argued that this reflects real-world use of a test, it seems unreasonable to expect that a single parameter can serve as a universal predictor when multiple mechanisms are involved. Whilst CL surveillance remains an important aspect of assessment in high-risk women, and cervical slope seems to perform better than does a single CL measure, it is not adequate as a stand-alone predictor of subsequent SPTB. Ongoing basic science research is required to discover the pathways that lead to premature birth, and further clinical research needs to examine associations between single risk factors (including short CL) and different phenotypes of SPTB.
