F Bruinsma, Mother & Child Health Research, La Trobe University, Level 3, 215 Franklin Street, Melbourne 3000, Vic, Australia. Email firstname.lastname@example.org
Please cite this paper as: Bruinsma F, Quinn M. The risk of preterm birth following treatment for precancerous changes in the cervix: a systematic review and meta-analysis. BJOG 2011;118:1031–1041.
Background Studies investigating the association between treatment for precancerous changes in the cervix and risk of preterm birth have used a variety of comparison groups.
Objectives To investigate whether treatment for precancerous changes in the cervix is associated with preterm birth (<37 weeks) and to examine the impact of the type of comparison group on estimates of risk.
Search strategy PubMed, Embase and CENTRAL were searched for studies pubished between 1950 and 2009.
Selection criteria Eligible studies were those that reported preterm birth outcomes for excisional and ablative treatments separately and included a comparison group.
Data collection and analysis Pooled relative risks (RR) and 95% confidence intervals were computed using a random effects model.
Main results Thirty eligible studies were located. Excisional treatment was associated with an increased odds of preterm birth, when compared with an external (RR 2.19, 95% CI 1.93–2.49) or internal (RR 1.96, 95% CI 1.46–2.64) comparison group. In comparison with women who were assessed but not treated, the risk estimate was smaller (RR 1.25, 95% CI 0.98–1.58). Ablative treatment was associated with an increased risk of preterm birth when an external comparison group (RR 1.47, 95% CI 1.24–1.74) but not an internal comparison group (RR 1.24, 95% CI 0.73–2.10) or untreated comparison group (RR 1.03, 95% CI 0.90–1.18) was used.
Authors’ conclusions Excisional treatment was associated with a significantly increased risk of preterm birth. It provides new evidence that some types of ablative treatment may also be associated with a small increased risk. The type of comparison group used is an important consideration when comparing the outcomes of studies.
Cervical cancer screening programmes aim to detect precancerous changes in the cervix that can be treated before the development of invasive disease. Although screening has resulted in reduced mortality from cervical cancer, the diagnosis of a precancerous lesion may have unintended adverse consequences for young women. The question of whether treatment for precancerous changes in the cervix is associated with subsequent adverse pregnancy outcomes, and in particular, preterm birth, has generated considerable interest from both a clinical and research perspective. Treatments are classified as excisional or ablative. Excisional treatments include cold-knife conisation (CKC), laser conisation and loop electrosurgical excision procedure (LEEP) [also known as large loop excision of the transformation zone (LLETZ)]. Ablative treatments include cryotherapy, laser ablation (vaporisation), cold coagulation and radical electrocoagulation diathermy.
Two small meta-analyses of preterm birth following excisional treatment for precancerous changes were published in 19931 and 2003.2 A comprehensive review was published in 20063 and subsequently extended in 2008 to examine severe outcomes such as perinatal mortality.4 In addition, a systematic review (without meta-analysis) was published in 2007.5 A number of large studies6–14 examining this question have been published since the most recent review,3 making it timely to update the analysis and examine methodological issues that may influence the findings.
To date, the meta-analyses have consistently shown that excisional,1–4 but not ablative,3 treatment is associated with an increased risk of preterm birth.
Studies investigating the risk of preterm birth following precancerous changes in the cervix have used three main sources of comparison group. The first, an external comparison group, comprises women who gave birth at the same time but who did not have a diagnosis of precancerous changes in the cervix. Women are selected either from women giving birth at the same hospital or from a population-based registry. A variation of this method compares the observed number with the outcome of interest (in this case, preterm birth) with the number expected (based on a standardised population). The second is an internal comparison group, where the outcomes of births before treatment are compared with those of births after treatment (also called a pre–post design). The third compares the outcome of births to women who had treatment for precancerous changes in the cervix with the outcome for women who had a diagnosis of precancerous changes but who were not treated (treated/untreated comparison). The use of each of these comparison groups has strengths and weaknesses. A strength of using an external comparison group is that the group can be matched on factors that may be related to the outcome of interest. In practice it is often difficult to match on all these factors or they may not be available (in medical records or registries) at the time when matching takes place. A strength of using an internal comparison group is that unknown as well as known confounders are adjusted for. A weakness is that the woman was younger (and perhaps more fertile), and of lower parity. Also her life situation (marriage, education, career, desire to have children) may change over time.15 A strength of using women diagnosed but who do not have treatment as the comparison group is that many risk factors for the acquisition of precancerous changes of the cervix will be present in both groups.4,5 A weakness is that women who do not have treatment tend to have lower grade lesions and to be younger.4,5,8
This systematic review and meta-analysis of cohort studies aims to investigate the impact of type of comparison group on any association between treatment for precancerous changes in the cervix and subsequent preterm birth (<37 weeks of gestation) and whether the risk of preterm birth varies across the type of treatment.
Inclusion criteria and analytic methods were specified in advance.
Types of studies
Randomised controlled trials and observational studies (cohort, case–control) published between 1950 and 2009 of treatment for precancerous changes in the cervix (also known as cervical intraepithelial neoplasia [CIN] or cervical dysplasia) were included. Eligible studies were those that included a comparison group, which could be: no treatment, women not diagnosed with precancerous changes in the cervix, births before diagnosis, or women selected from the general population.
Types of participants
Women diagnosed with precancerous changes of the cervix who subsequently delivered a singleton pregnancy after 20 weeks of gestation. Studies that included outcomes for twins or higher-order multiple births were excluded because of the association between plurality and preterm birth.
Types of interventions
Eligible treatments for pre-cancerous changes in the cervix were (i) CKC biopsy, (ii) laser conisation, (iii) LEEP (also known LLETZ), (iv) laser ablation (also known as vaporisation), (v) electrocoagulation diathermy, (vi) cold coagulation, or (vii) cryotherapy. If the study included both excisional and ablative treatments the outcomes for each must have been reported separately. The treatment must not have occurred during pregnancy. Studies reporting outcomes associated with treatment using amputation or combined amputation and conisation of the uterine cervix were excluded.
Types of outcome measures
The primary outcome is preterm birth defined as birth before 37 completed weeks of gestation.
A systematic search of Pubmed, EMBASE and CENTRAL (Cochrane Central Register of Trials) was undertaken by FB. The following keyword searches were used: precancerous changes in the cervix, cervical intraepithelial neoplasia, carcinoma in situ, cervix neoplasm, pregnancy outcome, cervical dysplasia, preterm birth, and low birthweight, LEEP, LLETZ, laser vaporisation, laser ablation, cryosurgery, cryotherapy, CKC, treatment. Eligible papers were those published between 1950 and 2009. In addition, bibliographies of published papers were searched to locate papers missed using database searching. No limits were applied for language. The last search was run on 21 April 2010 and 118 papers were found.
Study selection and data collection process
All studies were reviewed on eligibility criteria by FB. Data were collected using a standardised data extraction sheet. In a small number of cases (where presentation of published data was unclear) additional information or clarification was sought from the original researchers. Data abstraction was conducted by FB. If there was ambiguity regarding eligibility or presentation of results MQ provided a second opinion.
Information was extracted from each included study on: (i) study design, including data collection method; (ii) characteristics of study participants (including age and severity of disease), and any inclusion or exclusion criteria; (iii) type of treatment; (iv) type and source of comparison group; (v) definition of preterm birth; (vi) length of follow up; (vii) separation of singleton and multiple births; and (viii) separation of index and subsequent births.
The meta-analyses were performed by computing pooled relative risks (RRs) and 95% confidence intervals using a random effects model (DerSimonian and Laird method). Where population-based standardised measures were used, a standardised prevalence ratio (SPR) was produced using a random effects model.
All analysis was conducted using the REVMAN software (review manager 4.2, 2003; The Nordic Cochrane Centre, Copenhagen, Denmark). To examine the influence of adjustment for possible confounders, subgroup analyses were undertaken to show outcomes according to whether: no adjustment for confounders was undertaken; participants were matched; or possible confounding variables were included as co-variables in multivariable analyses (e.g. multivariable logistic regression). Adjusted results generated using multivariable analysis techniques were entered using the generic inverse variance approach. This approach requires the computation of standard errors. The confidence interval was manually converted into a standard error using the method described in the Cochrane Handbook.16 Studies that presented their results as adjusted odds ratios were manually converted to relative risks, also using the method described in the Cochrane Handbook.16
Statistical heterogeneitywas assessed in two ways; the chi-square statistic and I2.16 The chi-square statistic has low power when studies have small sample sizes or there are only a small number of available studies, for this reason a P value of 0.10 (rather than 0.05) was used to determine statistical significance. The I2 statistic assesses the impact of heterogeneity on the meta-analysis. It was calculated in the following way:
where Q is the chi-square statistic and df is its degrees of freedom. It describes the percentage of the variability in the effect estimate that is the result of heterogeneity rather than sampling error (chance).16 The Cochrane Reviewers’ handbook provides a guide to its interpretation and suggests that over 50% may represent substantial heterogeneity.16
The analysis examined excisional and ablative treatment modalities separately by type of comparison group.
To assess the impact of methodological quality on the findings of the meta-analyses, two sensitivity analyses were performed. The first excluded studies if there was ambiguity about any of the selection criteria (particularly about the definition of preterm birth), and if there was no separation of index and subsequent births. Methodologically this was important because the inclusion of more than one birth to a particular woman introduces clustering into the data. The second excluded studies that were rated of low methodological quality (based on criteria developed by Khan et al.17). Studies were rated of high quality if most (7–10) of the criteria were met, of medium quality if some (4–6) were met and of low quality if few (1–3) of the criteria were adequately addressed. An additional consideration was the power of the study to detect the outcome of interest. Results from these analyses are not presented here as the findings were not substantially changed; they can be obtained from the authors on request.
A total of 118 studies were identified and 30 (Table 1)7–15,18–38 were considered eligible for inclusion in the meta-analysis (further detail is contained in Table S1). Eighty-eight studies were excluded because there was no comparison group, insufficient detail or no separation of singleton and multiple births or because only an abstract was available (Figure 1). Studies used a range of different comparison groups and some studies used multiple types of comparison groups (Table S1). Only three studies12,23,36 examined pregnancy outcomes prospectively, all the rest were conducted retrospectively.
Table 1. Summary of included studies
Study time period
Assessment of study quality
*Denotes that multivariable modeling was used.
**Data for for CKC excluded from meta-analysis due to problem of multiple use of same comparison group.
***n of comparison group in forest plot is half of actual number (1 056 855) because package does not accept seven-digit numbers.
****Comparison selected from prelaser interval of the same woman.
Association of excisional methods of treatment and preterm birth
Overall, there was a significantly increased risk of giving birth preterm associated with excisional treatments when compared with an external comparison group (RR 2.19, 95% CI 1.93–2.49) (Table 2). The point estimate was larger when the comparison was women giving birth at the same hospital rather than women selected from population-based registers; however, the confidence intervals overlap (Table 2; see Figure S1).
Table 2. Risk of preterm birth associated with treatment type, reported by type of comparison group
Type of comparison group
No of studies
Number of women giving birth preterm among treated group
Number of women giving birth preterm among comparison group
The risk of subsequent preterm birth associated with CKC (RR 3.41, 95% CI 2.38–4.88) and laser conisation (RR 3.58, 95% CI 1.93–6.61) was more pronounced than LEEP (OR 1.85, 95% CI 1.59–2.15) (see Figure S2). These pooled relative risk estimates were generated using raw numbers. To further investigate the effect of the type of study design and analysis, studies using an external comparison group were analysed according to whether the study carried out no adjustment, the study was matched or multivariable modelling was used, for the most commonly used excisional treatment type, namely LEEP. The risk of preterm birth associated with LEEP was similar when only studies adjusting for a range of possible confounders (RR 1.84, 95% CI 1.63–2.08)2,11,12,38 were included compared with when raw data were used (RR 1.85, 95% CI 1.59–2.15).
When studies examining preterm birth subtypes; preterm prelabour rupture of membranes (pPROM) and spontaneous preterm birth were analysed, the pattern for spontaneous preterm birth, with a pooled relative risk of 1.99 (95% CI 1.68–2.36),11,13,26,36,38 was similar to that of all preterm birth using an external comparison group. The estimate for the risk of preterm prelabour rupture of membranes was higher, with a wider confidence interval (pooled RR 3.40, 95% CI 1.63–8.11).10,13,14,18,20,38
Three studies used standardisation with population-based data7,8,27 giving an SPR of preterm birth associated with excisional treatment 2.46 (95% CI 1.96–3.08).
Births following treatment had an increased risk of preterm birth in comparison to those before treatment (RR 1.96, 95% CI 1.46–2.64) (Table 2; see Figure S3).8,14,24,28,30,39 When the adjusted data were used this changed to a pooled relative risk of 2.25 (95% CI 1.73–2.93) because an additional study,11 which did not present raw data, was able to be included.
Table 2 shows the pooled risk of giving birth preterm for excisional treatment compared with women who had a diagnosis of precancerous changes in the cervix but who did not receive treatment (RR 1.25, 95% CI 0.98–1.58; see Figure S4); once again this changed very little if adjusted estimates are used (RR 1.27, 95% CI 1.01–1.59).7,11,13,15,22,31,37
Ablative treatment and preterm birth
There was a significant association with ablative treatment and risk of preterm birth (RR 1.47, 95% CI 1.24–1.74) (Table 2; see Figure S5),9,11,13,19,22,23,32,34,36 when an external comparison group was used; including adjusted estimates produced very little change (RR 1.42, 95% CI 1.29–1.56). When studies were grouped according to type of ablative treatment, no significantly increased risk of preterm birth was associated with laser ablation (RR 1.27, 95% CI 0.67–2.40) but when all types were combined there was an increased risk (RR 1.53, 95% CI 1.32–1.78).
Only one study, using standardisation techniques, reported pregnancy outcomes following laser ablative treatment (SPR 1.56, 95% CI 1.28–1.92).39 No significant association was found between ablative treatment and preterm birth (RR 1.24, 95% CI 0.73–2.10) when an internal comparison group was used.25,39
Table 2 shows the risk of giving birth preterm for ablative treatment compared with women who had a diagnosis of precancerous changes in the cervix but who did not receive treatment (RR 1.03, 95% CI 0.90–1.18; see Figure S6)7,11,13,15,22,31 and once again the results changed very little if available adjusted data were used (RR 1.03, 95% CI 0.89–1.18).7,31 One of the included studies,7 reported outcomes associated with both laser ablative and radical electrocoagulation diathermy treatment. Only outcomes associated with laser ablative treatment were included in this analysis. Radical electrocoagulation diathermy is no longer widely used as the destruction of cervical tissue was deemed excessive. However, if the data on radical electrocoagulation diathermy were included then the pooled relative risk changed very little (RR 1.09, 95% CI 0.91–1.31).
The findings of this review are based on data obtained from 30 cohort studies, 27 of which were retrospective. It examined the risk of preterm birth associated with excisional and ablative treatment separately and the impact on the findings of the use of different types of comparison groups.
The results of the current meta-analysis indicate that women treated using excisional treatments were at a higher risk of preterm birth than women who did not have a diagnosis of precancerous changes in the cervix. This was consistent whether the comparison group was women giving birth at the same hospital, women selected from population-based registries, standardised comparison with population-based registries or births to women before treatment. When women who were assessed, but not treated were used as the comparison group, the combined risk associated with excisional treatment, although still increased, was smaller.
Because of the small number of eligibile studies and the range of treatment types, the risk associated with particular forms of excisional and ablative treatment was only able to be examined among studies that used an external comparison group. An increased risk of preterm birth was associated with all forms of excisional treatment—CKC, laser excision and LEEP. The estimate produced for LEEP was significantly lower than for CKC, which may reflect the different criteria for selection of treatment modality or the amount of tissue excised. The current meta-analysis contains the largest number of studies, ten (using an external comparison group), and found a pooled relative risk of 1.88 (95% CI 1.61–2.19), which was similar to the estimates produced by previous meta-analyses based on five2 and eight3 studies respectively.
Some investigators have suggested that the outcome of interest should be spontaneous preterm birth rather than all preterm births combined.31,38 Only a small number of studies, using an external comparison group, examined the risk of preterm birth associated with spontaneous preterm birth.11,13,26,36,38 The finding was consistent with that of all preterm birth. The risk associated with more severe outcomes such as extreme prematurity and perinatal morbidity and mortality was not included in this review because of the small number of studies using comparison groups other than an external comparison group; this question has been addressed by another recent review.4
There is increasing evidence that the presence of a lesion (even without treatment) is associated with an increased risk of preterm birth, perhaps through shared risk factors for preterm birth. A number of studies investigating the risk of preterm birth associated with treatment for precancerous changes in the cervix compared with women who were diagnosed but did not receive treatment found high rates of preterm birth among untreated women.13,22,31 Excisional treatment then confers an additional small, but clinically significant risk.
A novel finding of this meta-analysis, and one that warrants further investigation, is that ablative treatment was associated with a small but significantly increased odds of preterm birth when compared with women giving birth either at the same hospital or in the population (external comparison group). Previous systematic reviews failed to find an association;3,5 however, they were based on a small number of studies. When studies were analysed according to type of treatment, no increase was seen among studies that investigated only laser ablative treatment; however, an increase was seen among studies that grouped findings for laser ablative treatment with other ablative treatments such as cold coagulation and diathermy. There are very limited data on outcomes associated with cryothery.9,22,25,36 Further studies are needed to assess whether the risk is consistent across ablative treatment modalities. Given that ablative treatment has been considered a less destructive treatment, it is possible that ablative treatment may not confer an additional risk above that associated with the presence of a lesion.
The choice of comparison group is complex and there is no ‘correct’ answer. The use of each comparison group addresses a slightly different question. Researchers and clinicians alike should consider the choice of comparison group when examining the findings of studies. Across both excisional and ablative treatment modalities the estimates produced using an external comparison group were higher than those produced using an untreated comparison group. One interpretation of the higher risk estimates of preterm birth associated with an external comparison group is that the risk is over-estimated, perhaps because of inadequate control of confounding. Use of a population comparison investigates whether women treated for precancerous changes in the cervix are at an increased risk of preterm birth compared with women in the general population, usually matched or adjusted for a number of possible confounding factors such as age and smoking. However, many of the predictors of preterm birth, e.g. low socioeconomic status, early sexual activity and smoking, are also risk factors for diagnosis of precancerous changes in the cervix and these may not be adequately controlled for. An alternative explanation is that the two groups of women come from different populations with differing distributions of risk factors. In clinical practice, this is still important information because it indicates that women treated for precancerous changes in the cervix may be at increased risk of preterm birth compared with other women giving birth in the hospital. Interestingly, using an internal control group (where some, but not all possible confounders are controlled: e.g. age at first sexual activity will remain constant but age, parity, number of sexual partners may change) gave estimates similar to those generated using an external comparison group, although once again there may have been inadequate adjustment of confounders. In one study, the rate of preterm birth among women after diagnosis (compared with prediagnosis) was similar among women who were untreated whereas the rate among women treated using excisional modalities was significantly increased39 with a point estimate similar to that seen in the current analysis. A strength of using women who were assessed but not treated as the comparison group is that many possible confounding factors (e.g. smoking status, age at first intercourse) may be similar across the two groups.5 A weakness is that women who do not undergo treatment may be younger and may have lower grade lesions.
The findings of the current analysis regarding excisional treatment confirm the findings of previously published systematic reviews and meta-analyses1–3,5 and a recent large, population-based registry study excluded from the current meta-analysis as it did not appear to report the findings for singleton and multiple births separately.6 Kyrgiou et al.3 located 27 studies that met their inclusion criteria, 22 of which examined preterm birth as an adverse outcome. Nine of these were excluded from the current meta-analysis because outcomes for multiple births were not reported separately,40–44 there was inadequate information to determine eligibility,45 preterm birth was inadequately defined46,47 and microinvasive cancer was the indication for treatment.48 Kyrgiou et al.3 reported the findings by type of treatment. The estimates produced by Kyrgiou et al.3 were lower than our estimates associated with treatment and use of an external comparison group. This was not surprising given that all types of comparison group were included. Our analysis showed that the treated/untreated comparison was associated with a smaller increase in the risk of preterm birth. Our analysis supports the importance of reporting outcomes by the type of comparison group, particularly reporting outcomes from treated and untreated women separately.
Studies included in the meta-analyses ranged in methodological quality. Ascertainment bias may have occurred as some studies did not specify the number of women lost to follow up and hospital-based studies may have missed births occurring at other centres. Misclassification bias may have occurred where women classified as untreated or undiagnosed were diagnosed or treated previously at another centre. This would act to dilute the effect of treatment. The majority of studies were conducted retrospectively. While in some instances this can be a disadvantage, in terms of minimising bias it was an advantage. Both the exposure and the outcome of interest were clearly defined and had occurred before data collection, making data collection less subject to bias.
The current meta-analysis was constrained by the limitations of the included studies; many of which did not report the size of the excision or the ablated area and as such the review was unable to adjust for these factors. An additional limitation was the design of available studies. All the studies were observational; conducting a meta-analysis does not remove the biases of individual observational studies. Unfortunately, risk of preterm birth according to individual treatment type was only able to be examined among studies using an external comparison group, because of the small number of eligible studies in each of the other comparison groups. However, in all analyses studies were grouped according to whether the treatment was excisional or ablative. Clinician preference, training, availability and cost of equipment as well as the grade of lesion all influence the choice of treatment modality. Further information on outcomes associated with different types of ablative treatment is needed, as is information on fertility and early pregnancy loss associated with all treatment types. Ideally, a randomised trial examining fertility and pregnancy outcomes, as well as rates of recurrence, of excisional versus ablative treatment should be conducted.
In conclusion, this meta-analysis confirms the findings of previous systematic reviews and meta-analyses that excisional treatment is associated with a significantly increased risk of preterm birth. Studies that used women who were assessed but not treated as a comparison group, consistently reported high rates of preterm birth among such women, suggesting shared risk factors for preterm birth.
The meta-analysis provides new evidence that ablative treatment (or particular types of ablative treatment) may be associated with a small increased risk, although this risk is smaller than that associated with excisional treatment.
The analysis has clearly demonstrated that the type of comparison group used is an important consideration when comparing the outcomes of studies. Studies that use women who were assessed but not treated as their comparison group should be considered separately from those that use an external comparison group.
Disclosure of interests
No conflict of interest to disclose.
Contribution to authorship
FB was responsible for all aspects of the the review. She designed the protocol, conducted the search and data extraction, entered the data and undertook the analysis. MQ participated in the design of the study protocol, contributed to decisions about inclusion and exclusion of studies, and to the intrepetation of findings. FB and MQ prepared the paper for publication.
Details of ethics approval
No ethics approval was required for this review.
FB was supported by a National Health and Medical Research Council Capacity Building Grant in Population Health Research. NHMRC ID: 433043.
The contribution of Professor Judith Lumley in the development of the protocol and discussion of conceptual issues is gratefully acknowledged and we would like to thank Professor Rhonda Small for commenting on drafts of the paper.