Comparative effectiveness and risk of preterm birth of local treatments for cervical intraepithelial neoplasia and stage IA1 cervical cancer: a systematic review and network meta-analysis

Probably not a substantial claim: may not pass S3 (load-bearing).

Within-study risk of bias was assessed separately for each outcome using RoB 2 for randomised controlled trials and ROBINS-I for non-randomised studies.

“Two reviewers (AA and IK) assessed within-study risk of bias separately for each outcome using RoB 224 for randomised controlled trials and ROBINS-I25 for non-randomised studies (appendix pp 12–14).”

• cites [24] Sterne, 2019, RoB 2: a revised tool for assessing risk of bias in randomised trials, BMJ sourceSupported
• cites [25] Sterne, 2016, ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions, BMJ sourceSupported

Classified as probably not a substantial claim.

Local treatment techniques for cervical preinvasive and early invasive disease were previously thought to be interchangeable with regards to treatment failure rates, largely based on a 2013 Cochrane review of randomised controlled trials.

“Local treatment techniques for cervical preinvasive and early invasive disease were previously thought to be interchangeable with regards to treatment failure rates, largely based on a 2013 Cochrane review of randomised controlled trials.”

• cites [14] Martin-Hirsch, 2013, Surgery for cervical intraepithelial neoplasia, Cochrane Database Syst Rev sourcePartly supported

  The Background (Main results) reports that 29 randomised trials found no significant differences in persistent disease after treatment for cervical intraepithelial neoplasia (CIN). But the Discussion (Overall completeness and applicability of evidence) says the trials were largely underpowered and that no firm conclusion about differences in treatment effect could be drawn; it does not establish that techniques are interchangeable. The review concerns CIN, not treatment of early invasive disease. (LLM, conf=0.88)

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim.

The risk of prematurity in subsequent pregnancies increases with increasing treatment radicality and length of cone excised.

“The risk of prematurity in subsequent pregnancies has been known to increase with increasing treatment radicality and length of cone excised.”

• cites [8] Kyrgiou, 2006, Obstetric outcomes after conservative treatment for intraepithelial or early invasive cervical lesions: systematic review and meta-analysis, Lancet sourceNot checked — no full text of the source
• cites [9] Arbyn, 2008, Perinatal mortality and other severe adverse pregnancy outcomes associated with treatment of cervical intraepithelial neoplasia: meta-analysis, BMJ sourcePartly supported

  The Discussion (Biological mechanisms; page unavailable) reports a positive association between depth of excision and adverse obstetric events, but presents the link between the amount of tissue removed and obstetric morbidity as a hypothesis. The Results (Severe and extreme preterm delivery; page unavailable) show higher risks after some procedures, but do not establish a consistent increase in prematurity with each increase in treatment radicality or cone length. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [10] Kyrgiou, 2014, Fertility and early pregnancy outcomes after treatment for cervical intraepithelial neoplasia: systematic review and meta-analysis, BMJ sourceSupported
• cites [11] Kyrgiou, 2016, Adverse obstetric outcomes after local treatment for cervical preinvasive and early invasive disease according to cone depth: systematic review and meta-analysis, BMJ sourceSupported
• cites [12] Kyrgiou, 2015, Proportion of cervical excision for cervical intraepithelial neoplasia as a predictor of pregnancy outcomes, Int J Gynaecol Obstet sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The credibility of the evidence included in the network meta-analysis was rated using CINeMA.

“We rated the credibility of the evidence included in the network meta-analysis using CINeMA33 (appendix pp 16–17).”

• cites [33] Salanti, 2014, Evaluating the quality of evidence from a network meta-analysis, PLoS One sourcePartly supported

  The Introduction and the section on confidence in each effect size propose a GRADE-based approach to rating confidence in network meta-analysis evidence. Our copy does not identify that approach as CINeMA or describe a tool by that name (searched: CINeMA). The citing authors’ statement that they used it is their own report, not something this source could establish. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

May combine two claims.

Inappropriately superficial excisions in an attempt to minimise adverse pregnancy sequelae might compromise long-term oncological safety and could be partly accountable for the increased post-treatment incidence of cervical cancer.

“Conversely, inappropriately superficial excisions in an attempt to minimise adverse pregnancy sequelae might compromise long-term oncological safety45 and could be partly accountable for the increased post-treatment incidence of cervical cancer.15, 18”

• cites [15] Strander, 2014, Effect of ageing on cervical or vaginal cancer in Swedish women previously treated for cervical intraepithelial neoplasia grade 3: population based cohort study of long term incidence and mortality, BMJ sourceSupported
• cites [18] Arbyn, 2014, Long term outcomes for women treated for cervical precancer, BMJ sourceNot checked — no full text of the source
• cites [45] Sasieni, 2016, Risk of preterm birth following surgical treatment for cervical disease: executive summary of a recent symposium, BJOG sourceSupported

May combine two claims.

Untreated women attending for colposcopy have a higher risk of preterm birth than women without CIN, probably explained by inherent characteristics and genetic, epigenetic, or microbiome factors.

“Untreated women attending for colposcopy have a higher risk of preterm birth than do women without CIN, probably explained by inherent characteristics and genetic, epigenetic, or microbiome factors.40, 41, 42”

• cites [40] Castanon, 2012, Risk of preterm birth after treatment for cervical intraepithelial neoplasia among women attending colposcopy in England: retrospective-prospective cohort study, BMJ sourcePartly supported

  The Results (Internal comparison; page not provided) report a preterm-birth rate of 8.3% after punch biopsy only, while Table 2 reports 6.7% for births in England. The paper does not directly compare untreated colposcopy attendees with women confirmed not to have CIN. The Discussion (Strengths and limitations; page not provided) considers pre-existing risk factors and disease-related confounding, but does not identify genetic, epigenetic, or microbiome factors as the probable explanation. (LLM, conf=0.84)

  Experimental — accuracy not yet measured
• cites [41] Kindinger, 2017, The interaction between vaginal microbiota, cervical length, and vaginal progesterone treatment for preterm birth risk, Microbiome sourcePartly supported

  The Results section reports an association between vaginal microbiota and preterm birth in women already at high risk, and the Discussion suggests possible microbiome-related mechanisms. The paper does not compare untreated women attending colposcopy with women without CIN, or attribute such a difference to inherent, genetic, or epigenetic factors. Page numbers are unavailable in our copy. (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites [42] Mitra, 2016, The vaginal microbiota, human papillomavirus infection and cervical intraepithelial neoplasia: what do we know and where are we going next?, Microbiome sourcePartly supported

  The ‘Limitations of current literature and areas for future study’ section says dysbiosis increases preterm-birth risk, and the ‘Vaginal microbiota and HPV/CIN/cervical cancer’ section reports an association between CIN severity and vaginal microbiota. The paper does not establish that untreated women attending colposcopy have a higher preterm-birth risk than women without CIN, or attribute such a difference to genetic or epigenetic factors. Page numbers are unavailable in our copy. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

May combine two claims.

No difference in risk of treatment failure was found between cold coagulation and LLETZ (OR 1.09, 95% CI 0.68-1.74), but direct data were based on two small studies only.

“No differences were found for the comparison of cold coagulation versus LLETZ (1·09 [0·68–1·74]) but direct data were based on two small studies only.”

• Also stated in: § Results · plain

  “with the exception of cold coagulation, which had a similar risk (1·09 [0·68–1·74]), but only two small studies compared cold coagulation directly with LLETZ”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

Cone lengths up to and including 10 mm were associated with a small increase in the risk of preterm birth when compared with the untreated colposcopy group, whereas lengths over 10 mm with a greater increase in risk (10 mm [10·4% absolute risk] vs colposcopy [7·9%]: OR 1·31 [95% CI 1·11-1·55]; 15 mm [17·2%] vs colposcopy: OR 2·17 [1·69-2·80]; 20 mm [32·4%] vs colposcopy: OR 4·09 [2·68-6·23]).

“Cone lengths up to and including 10 mm were associated with a small increase in the risk of preterm birth when compared with the untreated colposcopy group, whereas lengths over 10 mm with a greater increase in risk (10 mm [10·4% absolute risk] vs colposcopy [7·9%]: OR 1·31 [95% CI 1·11–1·55]; 15 mm [17·2%] vs colposcopy: OR 2·17 [1·69–2·80]; 20 mm [32·4%] vs colposcopy: 4·09 [2·68–6·23]; appendix”

May combine two claims.

On the basis of P-scores, laser conisation and CKC were the highest-ranking treatments, whereas cryotherapy was the lowest-ranking treatment.

“On the basis of P-scores, laser conisation and CKC were the highest-ranking treatments, whereas cryotherapy was the lowest-ranking treatment (appendix p 70).”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

Most ablative techniques had a risk of preterm birth similar to that of the untreated colposcopy group (7·9% [95% CI 6·3-9·9]), except for cold coagulation, which had a lower risk but much uncertainty (5·5% [0·2-71·5]) because the estimate was based on a single study of low quality.

“Most ablative techniques had a risk of preterm birth similar to that of the untreated colposcopy group (7·9% [95% CI 6·3–9·9]), except for cold coagulation, which had a lower risk but had much uncertainty (5·5% [0·2–71·5]) because the estimate was based on a single study of low quality.”

May combine two claims.

Secondary analyses of treatment failure using different cutoffs for the definitions of treatment failure (cytological ASC-H or worse, or histological CIN2 or worse: 30 studies; histological CIN1 or worse: 22 studies; histological CIN2 or worse: 18 studies; high-risk HPV positivity rates at 6 months: eight studies) were consistent with the main analysis, although there was more uncertainty in the估计

“We did secondary analyses of the outcome of treatment failure using different cutoffs for the definitions of treatment failure (cytological ASC-H or worse, or histological CIN2 or worse: 30 studies; histological CIN1 or worse: 22 studies; histological CIN2 or worse: 18 studies; high-risk HPV positivity rates at 6 months: eight studies). Results were consistent with the main analysis, although”

May combine two claims.

The absolute risk of any treatment failure was lowest for laser conisation and CKC, and highest for cryotherapy and laser ablation.

“The absolute risk of any treatment failure was lowest for laser conisation and CKC, and highest for cryotherapy and laser ablation (figure 5).”

May combine two claims.

In post-hoc subgroup analyses by grade of the lesion, the odds ratios of treatment failure after laser ablation and cryotherapy versus LLETZ increased with increasing CIN grade and were highest for CIN3 (laser ablation OR 2·38 [95% CI 1·03-5·48]; cryotherapy OR 3·11 [1·29-7·49]).

“In the post-hoc subgroup analyses for grade of the lesion, we found that odds ratios of treatment failure after laser ablation and cryotherapy (vs LLETZ) increased with increasing CIN grade, and were the highest for CIN3 (laser ablation: OR 2·38 [95% CI 1·03–5·48]; cryotherapy: 3·11 [1·29–7·49]; appendix pp 143–45).”

May combine two claims.

P-scores identified laser ablation, followed by cryotherapy and cold coagulation, as being the best treatments, and CKC as being the worst treatment for preterm birth.

“P-scores identified laser ablation, followed by cryotherapy and cold coagulation, as being the best treatments, and CKC as being the worst treatment (appendix p 77).”

May combine two claims.

Although the distribution of CIN2 or worse was similar, the proportion of women treated for CIN3 or worse was lower in ablative than excisional techniques.

“Although the distribution of CIN2 or worse was similar, the proportion of women treated for CIN3 or worse was lower in ablative than excisional techniques.”

May combine two claims.

When compared with LLETZ, ablative techniques had a reduced risk of preterm birth but with uncertainty (laser ablation: OR 0·77 [95% CI 0·56-1·04]; cold coagulation and cryotherapy also had a reduced risk, but estimates were much more uncertain especially for cold coagulation).

“When compared with LLETZ, ablative techniques had a reduced risk of preterm birth but with uncertainty (laser ablation: 0·77 [0·56–1·04]; cold coagulation and cryotherapy also had a reduced risk, but estimates were much more uncertain especially for cold coagulation; figure 3).”

May combine two claims.

In another post-hoc subgroup analysis of treatment failure, the odds ratios of cumulative treatment failure rates for laser ablation and cryotherapy compared with LLETZ increased with increasing follow-up duration and were highest in studies with median follow-up duration of at least 60 months (laser ablation OR 3·02 [95% CI 1·23-7·39]; cryotherapy OR 3·17 [1·36-7·35]).

“In another post-hoc subgroup analysis of treatment failure, the ORs of cumulative treatment failure rates for laser ablation and cryotherapy compared with LLETZ increased with increasing follow-up duration and were the highest in studies with median follow-up duration of at least 60 months (laser ablation: OR 3·02 [95% CI 1·23–7·39]; cryotherapy: 3·17 [1·36–7·35]); for cold coagulation, the”

May combine two claims.

In all these analyses, no differences were found between cold coagulation and LLETZ, but data on cold coagulation were scarce.

“In all these analyses, no differences were found between cold coagulation and LLETZ, but data on cold coagulation were scarce.”

May combine two claims.

Results for laser ablation and cryotherapy were quite conclusive even when non-randomised studies were down-weighted more aggressively and high-risk studies completely removed; for the rest of the treatments, uncertainty precluded definitive conclusions.

“Results for laser ablation and cryotherapy were quite conclusive even when we down-weighted non-randomised studies more aggressively and completely removed high-risk studies. For the rest of the treatments, there was uncertainty precluding definitive conclusions (appendix p 93).”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

CKC was more effective than LLETZ for treatment of endocervical lesions (post hoc) and when done in older (median age of ≥33 years) women (prespecified).

“In subgroup analyses, we found that CKC was more effective than LLETZ for treatment of endocervical lesions (post hoc) and when done in older (median age of ≥33 years) women (prespecified).”

May combine two claims.

The two sensitivity analyses that excluded non-randomised studies and that excluded non-randomised studies at high risk of bias showed consistent results with the main analyses, although there was more uncertainty due to the lower number of included studies than in the main analyses.

“The two sensitivity analyses that excluded non-randomised studies and that excluded non-randomised studies at high risk of bias showed consistent results with the main analyses, although there was more uncertainty due to the lower number of included studies than in the main analyses (appendix pp 143–5).”

Probably not a substantial claim: may not pass S3 (load-bearing).

Results from pairwise meta-analyses were in line with the network meta-analysis.

“Results from pairwise meta-analyses were in line with the network meta-analysis (appendix pp 66–67).”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

Cold coagulation had a similar risk of treatment failure as LLETZ, but data on cold coagulation were limited.

“Cold coagulation had a similar risk of failure as LLETZ, but data on cold coagulation were limited.”

Classified as probably not a substantial claim.

A 2017 meta-analysis reported lower rates of margin positivity and recurrence for knife and laser cones than for LLETZ.

“Another meta-analysis published in 2017 reported lower rates of margin positivity and recurrence for knife and laser cones than for LLETZ.”

• cites [16] Arbyn, 2017, Incomplete excision of cervical precancer as a predictor of treatment failure: a systematic review and meta-analysis, Lancet Oncol sourceNot checked — no full text of the source

May combine two claims.

The evidence base for the network meta-analysis was mostly observational studies with inherent risks of bias, and the credibility of many comparisons was low.

“The evidence was based mostly on observational studies with their inherent risks of bias, and the credibility of many comparisons was low.”

• Also stated in: § Introduction · plain

  “Results should be interpreted with caution because the evidence base consisted of mostly observational studies with inherent risks of bias.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Protocols were registered with PROSPERO (CRD42018115508 and CRD42018115495).

“Protocols have been registered with PROSPERO, CRD42018115508 and CRD42018115495, and published previously.19, 20”

• Also stated in: § Methods · plain

  “This systematic review is registered with PROSPERO, CRD42018115495 and CRD42018115508.”

• cites [19] Athanasiou, 2019, Comparative efficacy and complication rates after local treatment for cervical intraepithelial neoplasia and stage 1a1 cervical cancer: protocol for a systematic review and network meta-analysis from the CIRCLE Group, BMJ OpenNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [20] Athanasiou, 2019, Comparative fertility and pregnancy outcomes after local treatment for cervical intraepithelial neoplasia and stage 1a1 cervical cancer: protocol for a systematic review and network meta-analysis from the CIRCLE group, BMJ OpenNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

For observational studies, relative treatment effect estimates adjusted for potential confounders were extracted when available.

“For observational studies, we extracted relative treatment effect estimates adjusted for potential confounders, when available.”

• Also stated in: § Methods · plain

  “for observational studies, we extracted relative treatment effect estimates adjusted for potential confounders, when available”

Classified as probably not a substantial claim.

The trade-off between treatment failure rates and preterm birth among different treatment methods remains unclear.

“The trade-off between treatment failure rates and preterm birth among different treatment methods remains unclear to date.”

• Also stated in: § Background · plain

  “The trade-off between comparative effectiveness and reproductive morbidity of different treatment methods for cervical intraepithelial neoplasia (CIN) remains unclear.”

May combine two claims.

Ablative treatments probably do not increase the risk of preterm birth compared with untreated CIN, but they are associated with higher failure rates than excisional techniques.

“Although there is uncertainty, ablative treatments probably do not increase risk of preterm birth, but are associated with higher failure rates than excisional techniques.”

• Also stated in: § Introduction · hedged

  “Conversely, ablative techniques might not increase the risk of preterm birth compared with women with untreated CIN, but there is uncertainty and they have a higher risk of treatment failure than excisional procedures, particularly in the long term and when used for treatment for CIN3.”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

More radical excisional techniques (cold knife conisation and laser conisation) reduce the risk of treatment failure but increase the risk of subsequent preterm birth compared with more conservative treatments (LLETZ and ablation).

“More radical excisional techniques reduce the risk of treatment failure but increase the risk of subsequent preterm birth.”

• Also stated in: § Introduction · plain

  “We found that more aggressive treatments (ie, cold knife conisation and laser conisation) are associated with lower failure rates but a higher risk of subsequent preterm birth than more conservative treatments (ie, large loop excision of the transformation zone [LLETZ] and ablation).”

• AssumptionThe authors state this as an assumption: Heterogeneity (τ2) is the same across all treatment comparisons of each network.

May combine two claims.

Prespecified design-adjusted analyses were done to combine randomised and non-randomised evidence, down-weighting studies with higher risk of bias.

“We did prespecified design-adjusted analyses to combine randomised and non-randomised evidence,31 in which studies with higher risk of bias were assigned less weight and studies of lower risk of bias were assigned more weight.”

• cites [31] Efthimiou, 2017, Combining randomized and non-randomized evidence in network meta-analysis, Stat Med sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The systematic review and network meta-analysis was done according to the Cochrane Handbook and reported using PRISMA guidelines for network meta-analysis.

“This systematic review and network meta-analysis was done according to the Cochrane Handbook22 and was reported using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for network meta-analysis (appendix pp 3–5).23”

• cites [22] HigginsNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [23] Hutton, 2015, The PRISMA extension statement for reporting of systematic reviews incorporating network meta-analyses of health care interventions: checklist and explanations, Ann Intern Med sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Screening and data extraction were done independently by two reviewers, and discrepancies were resolved through discussion with a third reviewer.

“Screening and data extraction were done independently by two reviewers (AA and IK) and discrepancies were resolved through discussion with a third reviewer (MK; more details on data extraction are in the appendix [p 12]).”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Possible inclusion of women with clinically insignificant lesions, preferential use of excision or ablation on the basis of the severity and location of the lesion, and differences in the distribution of publication year, age, adenocarcinoma in situ, cervical cancer, and CIN3 or worse across treatment comparisons might have led to further bias in calculations for the risk of treatment failure.

“Possible inclusion of women with clinically insignificant lesions (ie, non-persistent CIN1) in some studies, preferential use of excision or ablation on the basis of the severity and location of the lesion, and differences in the distribution of publication year, age, adenocarcinoma in situ, cervical cancer, and CIN3 or worse across treatment comparisons might have led to further bias in our”

May combine two claims.

Because odds ratios of treatment failure and preterm birth after ablation (vs LLETZ and colposcopy, respectively) were higher in the subgroup analysis restricted to patients with CIN3 or worse than in the main network meta-analyses, the magnitude of risks from ablation for severe disease might be underestimated in the main analyses, but our conclusions would not change.

“Because odds ratios of treatment failure and preterm birth after ablation (vs LLETZ and colposcopy, respectively) were higher in the subgroup analysis restricted to patients with CIN3 or worse than in the main network meta-analyses, the magnitude of risks from ablation for severe disease might be underestimated in the main analyses, but our conclusions would not change.”

May combine two claims.

More aggressive local CIN treatments are associated with a reduced risk of treatment failure but an increased risk of preterm birth in subsequent pregnancies.

“Our findings suggest that more aggressive local CIN treatments are associated with a reduced risk of treatment failure but an increased risk of preterm birth in subsequent pregnancies.”

May combine two claims.

Data were dominated by observational data, thus bias might be present in the estimated treatment effects because of unmeasured confounding.

“Data were dominated by observational data, thus bias might be present in the estimated treatment effects because of unmeasured confounding.”

May combine two claims.

Heterogeneity in the network for preterm birth (τ2=0·02) was low compared with the empirical distribution, no evidence of inconsistency was found between direct and indirect data, and no small-study effects or publication bias were detected.

“Heterogeneity in the network (τ2=0·02) was low compared with the empirical distribution (appendix p 15). We found no evidence of inconsistency for risk of preterm birth between direct and indirect data (appendix pp 85–86).”

• Also stated in: § Results · plain

  “No small-study effects or publication bias were detected (appendix pp 227–30).”

May combine two claims.

No evidence of small-study effects or publication bias was found.

“We did not find any evidence of small-study effects or publication bias (appendix pp 217–21).”

May combine two claims.

Most non-randomised studies and some randomised controlled trials were at high risk of bias.

“Most non-randomised studies (34 [74%] of 46 studies for treatment failure and 14 [52%] of 27 for preterm birth) and some randomised controlled trials (seven [28%] of 25 for treatment failure and none for preterm birth) were at high risk of bias.”

May combine two claims.

Most non-randomised studies were downgraded because of absence of adjustment for confounders, whereas most randomised controlled trials were downgraded because of absence of a published protocol.

“Most non-randomised studies were downgraded because of the absence of adjustment for confounders, whereas most randomised controlled trials were downgraded because of the absence of a published protocol (appendix pp 40–65).”

May combine two claims.

The design-adjusted network meta-analyses gave almost identical results to the unadjusted network meta-analysis and did not affect conclusions drawn.

“The design-adjusted network meta-analyses gave almost identical results to the unadjusted network meta-analysis and did not affect conclusions drawn (appendix p 99).”

Classified as probably not a substantial claim.

The plausibility of the transitivity assumption was assessed by examining the distribution of suspected effect modifiers across treatment comparisons.

“To assess the plausibility of the transitivity assumption, we examined the distribution of suspected effect modifiers across treatment comparisons (appendix p 15).”

May combine two claims.

The study excluded studies that preferentially used excision for high-grade disease and ablation for low-grade disease, and did subgroup analyses of younger women, ectocervical lesions only, and studies excluding adenocarcinoma in situ or cervical cancer, which did not change the results.

“We excluded studies that preferentially used excision for high-grade disease and ablation for low-grade disease, and we did subgroup analyses of younger women, ectocervical lesions only, and studies excluding adenocarcinoma in situ or cervical cancer, which did not change the results.”