Precocious Cervical Ripening as a Screening Target to Predict Spontaneous Preterm Delivery among Asymptomatic Singleton Pregnancies: A Systematic Review

Abstract

Routine second-trimester transvaginal ultrasonographic (TVU) screening for short cervical length (CL) predicts spontaneous preterm delivery (SPTD), albeit with limited sensitivity (35–40%) and moderate positive likelihood ratio (LR+: 4–6). However, CL describes one of multidimensional changes associated with precocious cervical ripening (PCCR), which also include cervical softening, cervical funneling (CF), and dilation. PCCR, a precursor and a strong predictor for SPTD, was proposed as a potential screening target. We hypothesized that screening for composite measures of PCCR (e.g. CL, CF, cervical consistency, and dilation) using either digital exam (DE) or TVU would improve prediction of SPTD compared to screening for short CL alone. We searched PubMed and EMBASE electronic databases for observational cohort studies to evaluate cervical screening in asymptomatic obstetric populations. Multidimensional composite cervical measures were assessed in 10 datasets (n=22,050 pregnancies) and 12 publications. Appreciable heterogeneity in cervical measurements, data quality, and outcomes across studies prevented quantitative meta-analysis. Only one study reported intra- and inter-observer reliability of cervical measurements. The prevalence of CF ranged from 0.7% to 9.1%. Five studies compared composite measures of PCCR (i.e., CL and CF) to short CL alone, and consistently reported improved screening performance. Among three TVU studies, gains in sensitivity ranged from 5% to 27%, and increases in LR+ ranged from 3 to 16. Our findings suggest composite measures of PCCR might serve as valuable screening targets. High-quality interdisciplinary studies integrating epidemiologic approaches are needed to test this hypothesis and accelerate the translation of advances in cervical pathophysioloy into effective preventive interventions.

INTRODUCTION

Spontaneous preterm delivery (SPTD) is an unsolved public health problem of global proportions,¹ requiring more effective prevention strategies.^{2, 3} Timely prevention of SPTD commences with early identification of a modifiable target by means of effective screening programs.^{4, 5} In general, short cervical length (CL) is the screening target in routine second-trimester transvaginal ultrasonography (TVU, Figure 1, left and middle), which represents a simple, safe, and reproducible technical advance,^{6, 7} compared to digital examination (DE). While TVU has distinct advantages, DE doesn’t require sonographic training or equipment, and may therefore be more suitable for resource-limited settings. Whereas the false-positive rates of TVU and DE screening are similar,⁸ the limited sensitivity (35–40%) of TVU screening for shortened CL is marginally better than that of DE (25–30%).^{2, 8} Considering short CL, defined at less than 25 mm at <20 weeks, a UK review reported a moderate LR+ of 6.29 to predict SPTD before 34 weeks,⁹ and a Canadian review reported a LR+ of 4.31 to predict SPTD before 35 weeks.¹⁰ Moderate LR+s of short cervices assessed between 20–24 weeks were consistently reported,^{11, 12} including a value of 2.86 for asymptomatic high-risk women with a SPTD history.¹⁰

Figure 1.

Transvaginal ultrasonography (left) to assess a cervix (middle), [illustration by James A. Cooper, M.D., San Diego, CA in Callen (2008)³⁶], showing cervical effacement and funneling (right)³⁷ [reproduced with permissions from Elsevier and the American Institute of Ultrasound in Medicine].

If the obstetric community had an effective and efficient means of screening for SPTD, this approach could be expanded to routine use in all pregnant women. Previously, there was a lack of evidence for the value of early intervention. However, a recent meta-analysis demonstrated that vaginal progesterone administration to asymptomatic women with a sonographic short cervix not only reduced the risk of SPTD but also led to a 43% reduction in neonatal morbidity and mortality.¹³ Although universal screening for short cervices followed by progesterone treatment is cost-effective,^14–16 large (400 or 588) numbers of mothers must be screened to prevent one SPTD.¹⁷ Clearly, more efficient screening strategies are needed. According to The American Congress of Obstetricians and Gynecologists Practice Bulletins (#101 in 2009 and #130 in 2012), the maternal cervix “should be examined as clinically appropriate when technically feasible;”^18(p452) and universal cervical screening of pregnant women without a prior preterm birth may be considered despite “limited or inconsistent scientific evidence (Level B recommendation).”^7(p970) Furthermore, evidence-based research is required for greater quality assurance.⁷ Although the U.S. Preventive Services Task Force acknowledges the importance of predicting preterm delivery through screening, it has not recommended any screening targets.^19–21

Theory and Reasoning for Prediction

For years, multidimensional cervical features were used to predict the early onset of labor or SPTD. In 1964, the Bishop scoring system (cervical dilation, effacement, consistency, and position as assessed by DE), was correlated with the time to the onset of labor.²² In 1965, Wood et al. first reported that an internal cervical os dilated to one finger breadth and an effaced cervix predicted SPTD.²³ Papiernik and colleagues reported a decline in SPTD prior to 32 weeks (1.7% in 1971–1974 vs 0.8% in 1979–1982) in the French city of Haguenau after implementation of uncontrolled and multilevel interventions.^{24, 25}Prominent among targets of this population-based risk assessment and screening system²⁵ were both shortened cervices and patency of the internal os.²⁶ These precocious signs of cervical ripening can be recognized during a vaginal examination several weeks before the onset of SPTD and may be useful in predicting it.²⁷ Despite the presentation of this French screening experience at conferences^{28, 29} and in a March of Dimes monograph²⁴ aimed to the U.S. medical audience,³⁰ this paper published in 1986²⁷ hasn’t been cited widely in three past decades (78 citations on Web of Knowledge and 102 on Google Scholar in March 2014), and deserves a new look. Furthermore, we acknowledge again recent progress in available effective treatments, such as vaginal progesterone,⁷ which is an essential criterion required to support screening.^{4, 5, 31}

Identifying effective screening targets for SPTD relies on an understanding of its natural history and pathophysiology; in the latter circumstance, our understanding is lacking. Because precocious cervical ripening (PCCR) is an important precursor state in the SPTD pathway and a strong predictor for it, PCCR is a potential target for screening. Precursors are pathologic states that have a high probability of progressing to disease after a latent stage.³² Accordingly, ascertainment of properly defined precursors can increase the effectiveness of screening and prevention.³² As a recognizable stage in parturition,^33–35 the term PCCR was initially coined by Papiernik and colleagues in 1986.²⁷ PCCR describes multidimensional cervical changes including softening, shortening, funneling, and dilation of the internal os. These changes, visible using ultrasound,³⁶ progress from T-, to Y-, V- or U-shape funnels (Figure 1, right) prior to the onset of SPTD.^{27, 37} Cervical pathophysiology has been further investigated through molecular and cellular approaches.^38–40 Romero and colleagues described cervical ripening as a general feature of the “premature parturition syndrome.”^{41, 42} In 2011, routine recording of cervical ripening was recommended by the Global Alliance to Prevent Prematurity and Stillbirth.⁴³ In 2012, Caritis and Simhan proposed that the term PCCR was more appropriate and less confusing than either “cervical incompetence”^{6, 44, 45} or “cervical insufficiency,”⁴⁶ both being ill-defined biologically.⁴⁷ In this review, we use the term PCCR and operationalize it as at least two measurable cervical dimensions.

It is logical to ask how well the performance of PCCR has been evaluated to date in predicting SPTD. The effectiveness of a screening program depends on the interrelations between: 1) the performance, timing and frequency of screening procedures; 2) the efficacy of timely interventions; and 3) the risk profile of target populations.³⁵ We chose to investigate both reviews as well as individual studies; but we confine our comments regarding reviews to the introduction. Reviews by Owen and colleagues⁴⁴ and Honest and colleagues^{9, 12} grouped only observational studies; other reviews mixed clinical trials and observational studies together.^48–50 Despite providing useful insights concerning diverse populations, designs and analytical methods, prior reviews^{9–12, 17, 44, 48–52} failed to consider PCCR with most investigators focusing entirely on CL as measured by TVU.^{17, 44, 49–51} Reports from five investigative teams^{9–12, 48, 52, 53} over the past 15 years did not cite Papiernik et al.’s screening paper from 1986²⁷ but considered some of the hypotheses which form the basis for the present analysis. The first was Leitich and colleagues from Austria who concluded dilatation of the internal cervical os to be among the most effective markers for preterm delivery.⁵² The second was Honest and colleagues from the UK who published three reviews^{9, 11, 12} and reported that 1) the larger the funnel (e.g., dilatation of internal os >5 mm), the more accurately the prediction of SPTD; and 2) CL and cervical funneling (CF), used alone or in combination, appeared useful for SPTD prediction, but no data were highlighted. The third team was Crane and Hutchens from Canada, who included CF in their tables but did not summarize its predictive performance.¹⁰ The fourth team was Reiter and colleagues from Denmark, who published the only review that chose to target “premature cervical ripening” and reported unclear methods for the estimation and the insufficient evidence for routine screening;⁴⁸ however, they neglected to justify this target and to include CF from studies, such as the one from Iams and colleagues.⁸ Finally, Barros-Silva and colleagues from Portugal reported inconsistent findings in comparing combined screening targets to short CL alone in three studies, and recommended combining CL “with other markers (sonographic, biochemical and/or clinical) that reflect the multiplicity of mechanisms involved in the pathogenesis of SPTD”.⁵³⁽⁵⁾

Objective

We hypothesized that comprehensive assessment for multidimensional PCCR (e.g., CL, CF, cervical consistency, and cervical dilation in combination) is more effective (e.g., improved sensitivity and LR+) than screening for short CL alone using either TVU or DE.⁵⁴ DE is suitable for resource-limited settings, serves as a historical comparison, and is included. The primary outcome measure was SPTD at specified gestational weeks. In this systematic review, we aimed to identify, appraise, select and synthesize all high-quality research evidence. We assessed elements of study methodology and variations in cervical assessment, risk profiles of participants, and healthcare contexts. Further, we identified research gaps and suggest future research to improve the performance of cervical screening and its cross-cultural applicability.

METHODS FOR REVIEW

Selection criteria and data sources

A comprehensive literature search was conducted to identify articles published in English language journals from 1980 to March 2014. We included high-quality studies, which evaluated multidimensional aspects of PCCR to predict singleton SPTDs in observational cohort studies of unselected obstetric populations. We excluded studies assessing only one cervical dimension, i.e., CF⁵⁵ or CL.^56–58 This systematic review included only published manuscripts and was therefore exempt from Institutional Review Board (IRB) review.

Using the key words Cervi*, Preterm, Prematurity, and Predict*, singleton*, we searched PubMed and EMBASE electronic databases in March, 2014. We identified 538 reports and two from other sources as depicted in a flow diagram⁵⁹ (Figure 2) from the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA);⁶⁰ 397 citations and abstracts yielded 160 potentially relevant articles for full-text review. We excluded 119 papers considering only one cervical dimension and 29 papers including multiple cervical measures. Reference lists were manually searched but failed to reveal other studies. Twelve reports describing ten datasets comprising 22,050 pregnancies met the inclusion criteria.^{27, 61–65}

Figure 2.

Flow diagram of information through the different phases of the systematic review on comprehensive cervical screening for precocious cervical ripening to predict singleton preterm delivery in large observational cohort studies, 1980–2014.

Screening is the identification of asymptomatic disease or risk factors.⁶⁶ Assessing high-risk (e.g., mothers with a history of SPTD) or symptomatic mothers is not an appropriate design for comparative screening evaluation and will limit the generalizability of findings to other populations. We have been prudent to the inclusion and exclusion criteria so that we can identify all high-quality evidence of screening effectiveness. Twenty-five of the 29 excluded studies (citations are available upon request) assessed cervices of high-risk (n=4) or symptomatic (n=21) mothers; one excluded women with the history of SPTD;⁶⁷ Among three studies with asymptomatic participants, two did not evaluate predictive performances of cervical assessment,^{68, 69} and one evaluated cervical measures other than cervical length.⁷⁰

Measurements and evaluation process

The two essential characteristics of a screening test are its reliability and validity.⁶⁶ Therefore, when available we abstracted data on reliability, sensitivity, specificity, positive and negative predictive values (PPV, NPV), positive and negative likelihood ratios (LR+, LR-), and receiver operating characteristic (ROC) curves.⁷¹ However, due to the variation in reporting across studies, we only focused on reporting sensitivity and LRs. We calculated LRs based on the following formulas.⁶⁶

$$LR+=\frac{\text{sensitivity}}{1-\text{specificity}}\mathrm{L}\mathrm{R}-=\frac{1-\text{sensitivity}}{\text{specificity}}$$

To assess study quality, we used the recommendations from the Standards for Reporting of Diagnostic Accuracy (STARD) Steering Group⁷² and a review of screening tests⁷³ to document the following study criteria: sample characteristics (consecutive sample), study design (prospective or retrospective cohort), cervical assessment (evaluator background, sonographers, obstetricians, etc.), blinding, reliability, outcomes (definition of SPTD), and quantitative analysis (screening performance, and statistical association between cervical measures and SPTD).

The use of multidimensional cervical changes in the inclusion criteria allowed us to evaluate the performance of individual PCCR components. First, within a given population, we calculated the difference in sensitivity and/or LRs of composite measures of PCCR compared to short CL alone. Second, we assessed the consistency of within-study comparisons across studies.⁷⁴ However, a quantitative meta-analysis could not be performed due to 1) the lack of sensitivity and LRs data or data required to reproduce the 2×2 contingency tables and 2) the appreciable heterogeneity in design, screening, definitions of CF, and variable cutoff values.

RESULTS

General time and geographic trends

The 12 studies reviewed were published at a rate of approximately 3–4 per decade in the past thirty years (Table 1). They were conducted in Europe (5 publications, 41%), North America (3 publications, 25%), Asia (2 publications, 17%), and South America (2 publications, 17%). Most were from high-income countries (i.e., France, Sweden, Japan, Finland, the U.S., and the United Kingdom) or high-income regions (i.e., Hong Kong), but two originated in middle-income countries (i.e., Brazil and Colombia).

Table 1.

Characteristics of comprehensive cervical screening for precocious cervical ripening to predict singleton preterm delivery in large observational cohort studies, 1980–2014

Author, Published Year	Screen period	Location, venue	Population	Gestational weeks	Exam
Papiernik et al., 198627	1971–1976	*Haguenau, France	4430	18, 19–24, 25–28, 29–31, 32–34, and 35–36	DE
Bouyer et al., 198661	Ibid	Ibid	4390		DE
Mortensen et al,198765	1982	*Skaraborg, Sweden	581	24, 28 & 32	DE
Hartmann et al,199975	1995–2000	4 sites in North Carolina, USA	871	24–29	DE
Newman et al., 200877 Iams et al., 19968	1992–1994 Ibid	10-site MFMU network, USA Ibid	2916, 2538 2915, 2531	22–24, 26–29 Ibid	DE & TVU DE & TVU
Hasegawa et al,199681	1994	10 centers, Japan	729	15–34	TVU
Taipale & Hiilesmaa, 199863	1995–1996	*Helsinki, Finland	3694	18–22	TVU
To et al., 200164	1997–2000	*London, the UK	6334	22–24	TVU
de Carvalho et al, 200576	1998–2001	São Paulo, Brazil	1958	21–24	TVU
Leung et al, 200562	2000–2002	*Hong Kong, China	2880	18–22	TVU
Parra-Saavedra et al., 201180	2009–2010	Barranquilla, Colombia	1115	5–36	TVU

Note: DE: digital examination; TVU: transvaginal sonography.

^*for routine care. Ibid, the same as above; MFMU, NIH Maternal Fetal Medicine Unit Network in the U.S.

Study quality

All studies used prospective cohorts with consecutive cases (Table 2). The patients and providers in two studies and providers in five studies were blinded to cervical measures.^{8, 62, 63, 75–78} Intra/inter observer reliability in studies of ultrasound measures was poorly described; despite three studies claiming to have used rigorous quality control processes,^{77, 79, 80} only one reported the intra- and inter-observer reliability of the cervical consistency index.⁸⁰ Older studies reported associations (e.g., adjusted relative risk) between cervical measurements and the risk of SPTD.^{27, 61, 64} However, two of these did not include measures of sensitivity, specificity, or LRs,^{27, 64} and three others reported only a single criterion (i.e., sensitivity or PPV).^{76, 80, 81}

Table 2.

Assessment of study quality

Author, Published Year	Consecutive sample	Prospective cohort design	Exam	Evaluators	Blinded	Reliability quantified	Outcome Defined	Predictive performance Assessed	Statistical association reported
Papiernik et al., 198627	Yes	Yes	DE	OBs	Unclear	No	Yes	No	Yes
Bouyer et al., 198661	Yes	Yes	DE	Ibid	Unclear	No	Yes	Yes	Yes
Mortensen et al,198765	Yes	Yes	DE	Midwives	Unclear	No	Yes	Yes	Yes
Hartmann et al, 199975	Yes	Yes	DE	OBs & nurses	Provider	No	Yes	Yes	No
Newman et al., 200877	Yes	Yes	DE	Nurses & examiners	Provider	No	Yes	Yes	No
Iams et al., 19968	Yes	Yes	DE & TVU	Nurses & examiners	Provider	No	Yes	Yes	Yes
Hasegawa et al, 199681	Yes	Yes	TVU	?	Unclear	No	Yes	Yes	Yes
Taipale & Hiilesmaa, 199863	Yes	Yes	TVU	OBs and midwives	Provider	No	Yes	Yes	Yes
To et al., 200164	Yes	Yes	TVU	OBs & sonographers	Provider	No	Yes	No	Yes
de Carvalho et al, 200576	Yes	Yes	TVU	Sonographers	Patient & Provider	No	Yes	Yes	No
Leung et al, 200562	Yes	Yes	TVU	Sonographers	Patient & Provider	No	Yes	Yes	No
Parra-Saavedra et al., 201180	Yes	Yes	TVU	OBs	Unclear	Yes	Yes	Yes	No

Note: OB, obstetrician; DE: digital examination; TVU: transvaginal sonography; ?: not clear.

Study populations and definitions of SPTD

The low incidence rates of SPTD (e.g., <9% before 37 weeks, <5% before 34 or 35 weeks, and <1% before 33 weeks; Table 3) reflected generally low-risk populations. Mothers were recruited from multi-center studies^{8, 75, 77, 81} or hospitals.^{76, 80} Only hospitals from Finland, France, Hong Kong, Sweden, and the UK integrated cervical assessment into routine prenatal service an institutionalized preventive intervention to predict SPTD.^{8, 27, 61–65, 75–77, 80, 81}

Table 3.

Measurements and performance of screening for precocious cervical ripening to predict singleton preterm delivery in large cohort studies, 1980–2014

Author, Published Year	Gestational weeks at initial assessment	Measurements (cm for CL)	Funneling Dilation %	Preterm (<wks)	Incidence of PTD %	Sensitivity %	Specificity %	PPV %	NPV %	ROC	LR+	LR−	Association
Papiernik et al., 198627	18	Dilation, Station, Length, Uterine contraction, Expanded lower segment	0.8–12.4 By ≤18, 24, 28, 31, 34, 36 wks	37	?								2.5–3.4 (vary by 0.8–4.3 weeks)
													0.9–2.9
													1.2–2.9
													0.6–1.9
Bouyer et al., 198661	18	Score: short open cervix, contraction, parity, age	1.1–14	37	5.9*	44–57*	71–78*				1.5–2.4*	0.59–0.78*	0.8–3.1/1.8–6.7*
					5.5	56–64	73–78				2.1–2.8	0.47–0.58	0.9–2.7/1.6–3.5
Mortensen et al,198765	24	Modified Bishop score	4.0	37	1.5	33	88	4	97		2.8	0.76
		Dilation				11	97	6	99		3.6	0.92	8.3
		Effacement											3.1
Hartmann et al,199975	24–29	CL<2.0	6.0	37 or pPROM	8.3	13	93	15	92		1.9	0.94
		Dilation ≥1.0				8	99	38	92		8	0.93
		Cervical Score^<2				20	93	21	92		2.9	0.86
Newman et al., 200877	22–24	T1: Bishop score ≥4	?	35	4.4	28	90.9	12		0.66	3.0	0.80
		Cervical score^<1.5				13	97.9	21		0.61	6.4	0.88
		T2: Bishop score ≥5	?			32	93.0	14		0.68	4.6	0.73
		TVU CL<2.0				32	95	17	98	0.68	6.4	0.72
		TVU CF present				32	91	11	98		3.6	0.75
		Cervical score^<1.5				36	95	20	98		7.4	0.67
Iams et al., 19968	15–34	T1: Bishop≥4	6.4	35	4.3	28	91	12	97		3.0	0.80
		CL ≤2.5				37	92	18	97		4.8	0.68	6.19
		CF present				25	95	17	97		4.6	0.79
		T2: Bishop≥4	9.1			43	83	10	97		2.4	0.70
		CL ≤2.5				49	87	11	98		3.7	0.58	9.57
		CF present				33	92	17	98		3.9	0.34
Hasegawa et al,199661	15–34	CL ≤2.7	7.8	36	3.3			10*/2					4.86 (1.85–12.72)*
		Open internal os. Funneling index						7*/11					6.00 (1.65–21.71)
Taipale & Hiilesmaa, 199863	18–22	CL ≤2.9	0.7	35	0.8	19	97	6			6.3	0.84	8
		Dilation ≥0.5		37	2.4	16	99	20			16	0.85	28
		Either				29	97	7			9.7	0.73	11
To et al., 200164	22–24	CL	4	33	0.9								24.9
		Internal os. ≥0.5											1.8
de Carvalho et al, 200576	21–24	CL ≤2	1.5	34	3.4	7
		Add CF present				34
Leung et al, 200562	18–22	CL ≤2.7	6.3 width 6.4 depth 4.3	34	0.7	37	96	6	100		9.8	0.66
		CF				32	94	3	100		5.2	0.73
		Both				26	99	15	100		26	0.74
		Either				42	91	3	100		4.7	0.64
Parra-Saavedra et al., 201180	5–36	Consistency index	Excluded	34	2.1	64	98	47	99	0.94	39.7
		CL				9	98	9	98		4.3

Note: The bolded composite measures had better predictive performance than cervical length alone. T1 and T2, assessments at two time points.

^*for nulliparous women, other values in the same cell for parous women;

^{^}Cervical Score= Cervical length (cm) – Cervical dilation (cm). CL: cervical length; CF: cervical funneling; DE: digital examination; TVU: transvaginal ultrasonography; PTD, preterm delivery; wks: weeks; Ibid: the same as above; ROC: receiver operating characteristic curves; PPV: positive predictive value; NPV: negative predictive value; LR: likelihood ratio, calculated results based on original values in papers before being rounded.

Cervical assessment specifics

The timing, mode, and training of evaluators varied across studies (Table 2). Four studies screened only once during pregnancy,^{62–64, 76} two screened twice,^{8, 77} and one screened three times.⁶⁵ In studies conducted in the 1980s, obstetricians or midwives performed DE only. Three American studies evaluated DE by obstetricians and nurses⁷⁵ or by nurses and standard examiners^{8, 77} who had at least 5 years of experience in cervical examinations and were designated as the “standard” to which all cervical examiners were compared.⁷⁷ However, since the 1990s TVU, as performed by sonographers, obstetricians and midwives, has been common. No harmful effect of either method was reported.

Measurements and screening performance

The evaluation of multidimensional cervical screening (Table 3) began with the French study on PCCR,²⁷ which was followed by three others using the 5-component Bishop score or modifications thereof.^{8, 65, 77} Overall, insufficient standardization exists in terms of methods and definitions to compare performance criteria across studies. For example, the prevalence of CF ranged from 0.7% to 9.1%. Using data from the Maternal Fetal Medicine Unit (MFMU) Network, sensitivity was improved (e.g., 49.4% vs 37.3% for CL≤ 25 mm) at the second cervical assessment (26–28 weeks) compared to the first (22–24 weeks).^{8, 77} The change of performance over three assessments⁶⁵ was not reported.

The focus in cervical evaluation within a single screening section changed from their associations with SPTD to their predictive performance of SPTD. Most studies reported significant associations between cervical dilation and later CF and SPTD (Table 3). Specifically, Papiernik and colleagues²⁷ and Mortensen and colleagues⁶⁵ reported cervical dilation as being the strongest predictor for SPTD compared to other assessable dimensions. Bouyer and colleagues⁶¹ reported that cervical dilation had a persistently higher relative risk among parous women compared to nulliparous women. In contrast, To and colleagues noted that the significant association between CF and SPTD in univariate analysis became null after adjustment for CL.⁶⁴ As shown in Table 3, three studies reported larger LR+s for cervical dilation or CF compared to that of short CL alone (range of difference: 0.2–9.7),^{8, 63, 75} and two studies reported smaller LR+s (0.2–4.6).^{8, 62} Despite null findings from studies performed in the MFMU Network,^{8, 77} one reported better predictive performance of screening in parous mothers,⁶¹ whereas the other suggested improved performance in nulliparous mothers.⁸¹

Using both CL and CF (i.e., either one being abnormal, both abnormal, or a combined score), only five of 12 studies compared composite measures of PCCR to that of a short CL alone within a single screening section, and reported gains in sensitivity (from 4% to 27%) and LR+ (1 for DE and comparing DE to TVU, and from 3.4 to 16 for TVU) (Table 4). Hartmann and colleagues⁷⁵ and Newman and colleagues⁷⁷ used a cervical score (CL minus dilation in centimeters) based on DE (i.e., sensitivity: 20% vs 13%, 36% vs 32%; LR+: 2.9 vs 1.9, 7.4 vs 6.4). Using TVU, Taipale and Hiilesmaa⁶³ used a measure of either a short CL or cervical dilation (sensitivity: 29% vs 19%; LR+: 9.7 vs 6.3) to predict SPTD. Leung and colleagues,⁶² combining CL and CF, improved LR+ (26 vs 9.8) but reduced sensitivity (26% vs 37%), whereas using either indicator improved sensitivity (42% vs 37%), but reduced LR+ (4.7 vs 9.8). De Carvalho and colleagues⁷⁶ combined CF with short CL and improved sensitivity (34% vs 7%).

Table 4.

Indicators and performance of screening for composite measures of precocious cervical ripening in selected large observational cohort studies, 1980–2014

Mode of assessment	Study, Year	Time (wks)	Indicators # (cm for CL)	PTD (wks)	Prevalence	Sensitivity %	Specificity %	+PV %	−PV %	+LR
I: DE only	Hartmann et al,199975	24–29	CL<2.0	37	8.3	13	93	15	92	1.9
			Dilatation ≥1.0			8	99	38	92	8
			Score*>2			20	93	21	92	2.9
			Difference			+7				+1
II: DE and TVU	Newman et al., 200877	26–29	CL ≤2.0	35	4.3	32	95	17	98	6.4
		T2	Funneling			32	91	11	98	3.5
			Score<1.5			36	95	20	98	7.4
			Difference			+4				+1
	Taipale and Hiilesmaa, 199863	18–22	CL ≤2.9	35	0.8	19	97	6		6.3
			Dilatation≥0.5	37	2.4	16	99	20		16
			Either			29	97	7		9.7
			Difference			+10				+3.4
III: TVU	de Carvalho et al,200576	21–24	CL ≤2	34	3.4	7
			CL+CF			34
			Difference			+27
	Leung et al, 200562	18–22	CL ≤2.7	34	0.7	37	96	6	100	9.8
			Funneling			32	94	3	100	5.2
			Both			26	99	15	100	26
			Either			42	91	3	100	4.7
			Difference			+5				+16.2

Note. CL, cervical length; DE: digital examination; Difference: performance difference between the composite cervical assessment and cervicallength only; PTD, preterm delivery; TVU transvaginal ultrasonography;

^*Score=Cervical length-Dilatation; PV, predictive value; LR, likelihood ratio; wks: weeks.

We found other targets of cervical measures in addition to CL and CF and the consideration of parity. One study investigated cervical consistency, the ratio of the anteroposterior cervical diameter measured before and after application of pressure on the cervix using the transvaginal probe, multiplied by 100.⁸⁰ Another study compared the performance of summary scores through stratifying by parity.⁶¹ Two studies^{75, 77} applied the formula of a cervical score validated among twin gestations^82–84 to singleton gestations.

COMMENT

This review systematically reviewed the available evidence from three decades of comprehensive cervical screening for PCCR in large observational cohorts of asymptomatic singleton gestations to predict SPTD. Only studies from Finland, France, Hong Kong, Sweden, and the UK integrated cervical assessment into routine prenatal care as a standard. Most of the studies reviewed had insufficient standardization and varied by methodological quality, outcomes (i.e. early or all SPTD), and cervical assessment (i.e. dimensional measurements, and timing and frequency of examination). Shorter CL and CF had high specificity, but low sensitivity to predict SPTD. All five studies that used composite measures consistently showed improved predictive performance compared to those which used CL alone. Taken together, our findings indicate that composite measures of PCCR could represent valuable targets of future research to optimize the prediction of SPTD.

Existing studies are limited in their scope, and do not include reports from low-income countries.⁸⁵ Variable screening performance could be explained by measurement issues including methods, evaluators, facilities, and global populations. For example, regional- and racial-specific cervical assessments (e.g., gestational end points for racial groups across continents) were reported.^{62, 86–88} Blinded cervical evaluation was not widely used. Standardization of cervical assessment⁸⁹ is necessary to improve the validity of cervical screening. Despite the difficulty in obtaining reliable CF measures,^{58, 90} such details need further investigation and quantification to improve their reliability. Only one study reported reliability.⁸⁰ Future high-quality studies should assess intra-rater and inter-rater reliability of cervical measurements and evaluate reasonable cervical predictors to generate empirical evidence.

We also determined that the lack of the underlying theory and the logic behind prediction might explain the varied performances across screening tests. As stated in our introduction, the features of an ideal screening test start with a detectable target based on pathophysiology. PCCR is an important but under-researched concept; only five groups^{9–12, 46, 50, 51} considered some of the hypotheses which form the basis for the present analysis. Using an operational definition of PCCR and available empirical measurements, we found evidence that composite measures of PCCR improved predictive performance compared to CL alone.^{62, 63, 75–77} and may thus serve as valuable potential screening targets.

Our review also discovered analytical issues that might explain observed variances in screening performance. Few studies used LR±⁶² and ROC curves.⁷⁷ To optimize analytical processes, we propose a 2-step multivariable prediction model.^{91, 92} First, identify and model measurements to generate a formula. Second, quantify scores and cut-off points based on the best performance,⁶⁶ evaluating the effect of adding additional characteristics to a short cervix.⁹³ It is necessary to consider profiles of target populations. Originally, To and colleagues⁶⁴ reported an insignificant association between CF and SPTD, but their data suggested an additive value of CF depending on CL (Figure 3). The correlation between CF and CL and the potential multicollinearity⁹⁴ are worthy of consideration.

Figure 3.

Association of cervical length and percent of funneling based on To et al.⁶⁴

Despite strengths, including “within-study comparisons” and aligning the pathophysiology, definition and function of PCCR as a precursor to optimize the prediction of SPTD, this review is subject to several limitations. First, related reports may not have been identified if they did not include key words in abstracts or titles. Second, we could not access original data to test whether observed differences in screening performance were significantly different. In this regard, we could neither perform individual patient data meta-analyses, nor synthesize cut-off points of measures across studies (e.g., definitions of SPTD, features of CF, percentile values or Z-scores for CL).⁹⁵ Third, the generalizability of the findings from studies using convenience samples of women is limited. Finally, repeat cervical screening,^96–98 screening coupled with interventions,⁵⁴ the concurrent use of other predictors (e.g., collagen structure, elasticity,⁹⁹ consistency,⁸⁰ fetal fibronectin,¹⁰⁰ amniotic fluid sludge,^{101, 102} maternal position,^{103, 104} and multiple gestations¹⁰³) as well as cost effectiveness deserve future attention.

CONCLUSIONS

Further research on comprehensive cervical assessment for multidimensional PCCR to predict singleton SPTD is justified for several reasons:

First, the recent increase in global attention to SPTD invites the development of clinical innovations with potential for primary and secondary prevention. Although the concept of PCCR is grounded in clinical tradition, its application and empiric measurement in screening tests require innovative and epidemiologic approaches to generate contemporary evidence.

Second, important gaps must be addressed in the preparation of an optimal, evidence-based protocols and high-quality comparative studies on screening for PCCR. Key unknowns in this daunting task include the lack of data on analytical approaches to incorporate cervical measurements, timing, frequency, and reliability of screening. In this regard, secondary data analyses also can be valuable. Benefits and harms should be assessed (e.g., under- or over-prediction).⁵ Standardized research is required to improve the conceptualization, measurement, and validation of comprehensive cervical screening for PCCR.

Third, the U.S. has much to contribute to global initiatives to predict and prevent SPTD, as it contributes half of all SPTD cases among high-income countries.⁸⁵ In 1986, Papiernik noticed the lack of routine cervical assessment in both the U.S. and Great Britain compared to continental Europe;²⁷ whereas the United Kingdom adopted it later in 1997,⁶⁴ the U.S. has not and high-quality studies^{8, 75, 77} included in this review were designed 20 years ago. Today’s pressing needs include the national approaches to advance research and practice using high-quality design and new data. In this regard, cervical data are being collected from 10,000 nulliparous American women from eight sites in a research network.¹⁰⁵

Finally, high-quality studies using an interdisciplinary approach including epidemiology are needed to test the hypothesis of PCCR as a target and accelerate the translation of advances in pathophysioloy into effective preventive interventions. This journal previously has called for accelerating efforts in collaborative and translational research.¹⁰⁶ By synthesizing knowledge across disciplines (e.g., cervical pathophysiology, clinical epidemiology, and maternal fetal medicine), epidemiology can play a central role and provide methods and tools to enhance translational research¹⁰⁷ and facilitate evidence-based practice.^{106, 108} Precursors,³² predisease,¹⁰⁹ and “predictor of poor health”⁵ can advance preventive interventions, a successful example being cervical intraepithelial neoplasia grade 3 detection for cervical cancer prevention.³²

Abbreviations

CF

cervical funneling

CL

cervical length

DE

digital examination

MFMU

Maternal Fetal Medicine Unit

PCCR

precocious cervical ripening

PPV

positive predictive values

LR+

positive likelihood ratio

SPTD

spontaneous preterm delivery

ROC curves

receiver operating characteristic curves

TVU

transvaginal ultrasonography