Design and Conduct of Search

A review protocol⁴⁶ was developed and posted by one SR.^4,39 The other made no reference to a protocol or a priori objectives so it is unclear whether all design elements and analyses were pre-planned.⁴⁰ One SR included duplicate study selection.^4,39 The number of authors involved in data extraction was unclear as one report indicated two reviewers³⁹ and one indicated that data was extracted by one reviewer and reviewed by another.⁴ The other SR⁴⁰ did not disclose the number of reviewers involved in study selection and extraction. Lack of duplicate study selection increases the risk of overlooking potentially relevant literature, and lack of duplicate data extraction increases the risk of errors. Both SRs^4,39,40 conducted a comprehensive literature search using multiple databases. However, with the exception of searching of reference lists, no other grey literature search methods were disclosed. The searches were limited to English and Japanese language publications.^4,39,40 No date restrictions were noted with the exception of one database used in one SR,⁴⁰ which was searched from 1983 onward. Further, this SR failed to disclose search dates for two databases used. Neither SR^4,39,40 mentioned publication status as an exclusion criteria for the search, and one SR^4,39 included unpublished work.

Reporting

Both SRs provided information about included studies, though one SR⁴⁰ failed to present a list of included studies and presented study characteristics by clinical outcome. One SR^4,39 provided a list of excluded studies with reasons for exclusion. The other SR did not disclose a list of excluded studies.⁴⁰.

Quality Assessment

Both SRs^4,39,40 assessed study quality. One SR^4,39 used several tools and conducted quality appraisal in duplicate with input from content experts. The Cochrane Risk of Bias tool was used for RCTs, NRSs were assessed using tools based on the Research Triangle Institute item bank and McMaster Quality Assessment Scale of Harms. This quality assessment was used to inform a designation of strength of evidence as per AHRQ’s Effective Health Care Program’s Methods Guide for Effectiveness and Comparative Effectiveness Reviews.⁴⁷ One SR⁴⁰ used Cochrane Risk of Bias and Grading of Recommendations, Assessment, Development and Evaluations (GRADE) criteria^48,49 to appraise the quality of studies and certainty of evidence and strength of recommendation. Both SRs discussed quality extensively in the presentation of results and conclusions. Overall, one SR⁴⁰ rated the evidence as high to moderate quality for breastfeeding efficacy, and weak for non-primary outcomes including milk supply, breastfeeding continuation, weight gain, and adverse events. The other SR reported mixed study quality, with RCTs generally being higher quality and NRSs and case series or reports typically being of lower quality, and the strength of evidence being low to insufficient for the majority of outcomes.^4,39

Method of Pooling and Publication Bias

One SR⁴⁰ conducted meta-analyses for several outcomes; however, the appropriateness of this approach is debatable. The other SR^4,39 which reviewed all of the same primary clinical studies, concluded that “the small number of studies, the study designs and the heterogeneity of interventions and outcomes made a meta-analysis inappropriate” (page 43).⁴ Consistent with this, the meta-analytic results of Ito et al.,⁴⁰ included a maximum of three studies per outcome, and all analyses reported high heterogeneity (I² = 81% to 90%). Also two pooled outcomes included NRSs of varying design. Thus, results from meta-analyses should be interpreted with caution. Publication bias was assessed informally by both SRs.^4,39,40 Publication bias was considered for each primary study as part of the evidence profile of one SR,⁴⁰ but the method used for assessment was unclear. One SR discussed publication bias in the limitations, noting that there was a possibility of publication bias, and that “most controlled studies reported only positive outcomes, and we identified no negative trials” (page 49).⁴ Grey literature search methods were limited for both SRs^4,39,40 and no other methods of reducing publication bias (e.g., contacting authors and manufacturers, searching trial databases) were used, so publication bias in this literature base cannot be ruled out.

Conflict of Interest

Both SRs declared that there was no funding sources or affiliations that would result in conflict of interest.

Reporting

All studies^41–45 described a hypothesis, aim, or objectives within their methods or introduction section, with the exception of the RCT⁴¹ where it was only presented in the abstract. Two studies^43,45 described outcomes measures clearly. The RCT⁴¹ did not explicitly list outcomes, and two studies only described outcomes in brief^42,44 until the results section. It was unclear whether study outcomes were pre-defined. All studies presented inclusion criteria with varying degrees of detail.^41–45 The RCT⁴¹ described both the inclusion and exclusion criteria clearly. The NRSs^42–45 did not provide details about patient exclusion criteria. This is of concern as including patients with certain genetic disorders, concomitant malformations, and other clinical conditions, who were excluded from other reports, may affect the generalizability of the findings. In addition, two studies^42,43 provided very limited information about the inclusion criteria. The intervention of interest was clearly described by four of the studies.^41,43–45 One study⁴² only gave the name of the intervention (i.e., frenotomy) and did not provide detail regarding the specific method of conduct. The distribution of potential confounders was underreported by all studies.^41–45 Studies presented limited patient information, often in aggregate, rather than by group. Important confounders including method of diagnosis, degree of impairment or immobility, gender, ethnicity, maternal anatomy, milk supply, adjunct therapies, and relevant comorbidities were often unclear. As these factors have the potential to influence the relationship between frenectomy and breastfeeding or other clinical outcomes, the lack of information makes it difficult to assess their impact. Four studies^42–45 presented their main findings clearly; however, there was incomplete information about analytical approach in one case⁴² and unclear statistical significance for some outcomes in another.⁴³ These issues made it difficult to interpret the results with certainty. The RCT⁴¹ failed to present change from baseline statistics, despite making conclusions reflecting this comparison. In addition, there were some inconsistencies in the group means presented for between group differences and within group differences in some cases, but it is unclear whether this would have affected the observed outcomes.⁴¹ Estimates of random variability were presented for the main outcomes for four studies.^41–43,45 One study reported simple rate estimates and did not provided any estimates of random variability.⁴⁴ Adverse events were generally underreported by all studies. The RCT⁴¹ reported on intraoperative complications, bleeding, and adhesion, and one study⁴² reported on major bleeding, infection and ulceration, but methods of assessing these outcomes was unclear. Adverse events were indicated as self-reported in one study and none were identified,⁴⁴ and two studies^43,45 did not comment on adverse events. Long term function issues, pain, and discomfort were not assessed by any study. Three studies^41,43,45 reported no dropouts or losses to follow-up. One⁴⁵ was a database study, so it is unclear whether there were eligible patients without follow-up data who weren’t included in the study. One study conducted assessments immediately after surgery and no follow-up measures were made.⁴⁴ One study reported significant losses to follow-up due to missing questionnaire responses, and it is unclear how the patients who didn’t respond differed from those who were included in analysis.⁴² One study reported significant numbers of patients who refused the procedure.⁴³ It is unclear how these patients differed from the patients who received the surgery. The four studies that conducted statistical analysis all reported actual P-values.^41–43,45 One study didn’t conduct any statistical tests.⁴⁴

External Validity

None of the studies reported explicitly on sampling methods. In most cases the study sample was representative of the entire eligible population of children at the facility where the trial was being conducted. For example, the RCT included all eligible children attending pediatric surgery clinics over a two year period.⁴¹ Two NRSs included all patients who attended an out-patient ankyloglossia clinic,⁴⁴ or all patients with documented tongue tie from a single hospital population.⁴⁵ All parents of children who had ankyloglossia and underwent frenectomy were contacted in one study.⁴² Only patients whose parents agreed to frenectomy were included in another study, after screening by a speech language pathologist.⁴³ Regarding willing versus unwilling participants, two studies were retrospective chart reviews so all eligible patients were assessed.^44,45 Two studies mentioned that some patients either did not receive the procedure⁴³ or did not attend the clinic.⁴² One study⁴² reported that non-participants either sought alternative care or the reasons for not participating were unknown. Differences between willing and unwilling participants were unclear for two studies.^41,43 The settings and staff were generally highly specialized with a higher standard of care.^41–44 For instance, studies were conducted at pediatric surgery clinics at a teaching hospital,⁴¹ a dedicated tongue-tie clinic,⁴² specialized ankyloglossia clinic,⁴⁴ hospital with highly specialized assessment of patients.⁴³ One study did not provide details about the study setting.⁴⁵ Overall, the patients included in these studies appeared to be more visible (i.e., had chosen to attend a specialty clinic or undergo assessment or procedure) and may have received a higher standard of care than all patients in the general ankyloglossia population who might be eligible for frenectomy.

Internal Validity – Bias

Patient blinding was not possible as the patient is awake during the procedure, but due to the patient age this is likely low risk. Similarly, the surgeons or health practitioners performing the surgery could not be blinded. None of the NRSs^42–45 blinded subjects, parents, or outcome assessors. The RCT⁴¹ blinded parents, but it was unclear whether outcome assessors were blinded. No studies adjusted for different lengths of follow up. In four cases the duration of follow-up was reported to be the same between comparison groups.^41,43–45 One study⁴² reported different lengths of follow-up with no adjustment. In this case it is unclear how this may have affected the outcomes observed. It is possible that shorter follow-up times may have been associated with better outcomes as they were closer to the intervention. The RCT⁴¹ reported using paired t-tests for assessing differences between groups, which is appropriate for the change from baseline values, but not for between group comparisons as it increases the chance for false positives. Another study reported using analysis of variance tests to detect differences between groups, but there were only two comparison groups.⁴³ Again, this is not the appropriate statistical test, but in this case it is unlikely that it affected the outcome in a meaningful way.⁴³ One study did not describe the statistical tests used to assess the main outcomes, and did not use statistical tests to assess differences before and after the intervention at all time points. The statistical significance of differences presented for long-term follow-up is unclear.⁴² One study did not conduct statistical analyses and only presented summary statistics.⁴⁴ One study used parametric statistics for continuous outcomes but due to the small sample sizes, non-parametric tests may have been more appropriate.⁴⁵ All studies had reliable compliance as it was a surgical intervention, but it is unclear whether there was compliance with other aspects of the treatment (e.g., following advice from lactation consultant, care of surgical site). The latter may have varied between and within study populations and may have to some extent contributed to variation in outcomes. For the majority of outcomes assessed by all primary studies, assessments were made by the mother and were subjective. In some cases nominal or ordinal scales were used but they appeared to be arbitrary.^41–44 These outcomes are subject to recall bias (e.g., when conducted retrospectively) and performance bias. Further, harms were self-reported in one study,⁴⁴ and appeared to be potentially self-reported in others.

Internal Validity – Confounding

Patients in all studies were recruited from the same populations including single hospitals,^41,42 an ankyloglossia clinic,⁴⁴ a hospital population screened for ankyloglossia,⁴³ and a group of patients diagnosed with tongue tie.⁴⁵ Patients in all studies were recruited over the same time period which ranged from five months to two years. The RCT⁴¹ randomized patients using permuted block randomization with a block size of four. It is unclear whether random allocation was concealed. Regarding adjustment for confounders, all studies failed to present a comprehensive overview of baseline characteristics of study groups. In the case of the RCT⁴¹ it was unclear whether there were significant baseline differences to account for. Two studies^42,44 stated that potential confounders were not assessed and could not be adjusted for. The remaining two studies,^43,45 did not adjust for any potential confounders, and did not describe reasons for this omission. Confidence in the results is limited due to lack of information regarding the potential risk of confounding, and lack of justification for not performing analyses (e.g., multivariate regression) to explore the influence of potential confounders on the effectiveness of frenectomy. Four studies reported either no follow-up time, were database studies so there were no losses, or reported zero dropouts, and did not report planned methods of accounting for losses.^41,43–45 One study reported significant losses to follow-up and did not take them into account. As it is unclear how these patients differed from those who completed the study, the potential for bias is unclear.⁴².

Power and Conflict of Interest

No power or sample size calculations were disclosed by any included study, so it is unclear whether there was sufficient power to detect the outcomes of interest. This is of particular concern where small sample^43,45 and imbalanced group^41,45 sizes were observed. None of the study authors noted any conflict of interest, financial or otherwise. Although one study did not explicitly disclose a conflict of interest statement.⁴⁵

Breastfeeding Effectiveness

Regarding immediate or short-term maternally reported breastfeeding effectiveness, one SR^4,39 reported that maternally reported improvements in breastfeeding were observed by three studies, one that compared frenectomy to lactation consultant support at 48 hours,⁵⁰ one to sham surgery immediately after intervention,²⁷ and one to no intervention immediately post-procedure.³⁵ Based on the Breastfeeding Self-Efficacy Scale Short Form (BSES-SF) and HATLFF score, mothers reported significant improvements at five days in the group who received the procedure versus no intervention.³⁰ However, at 8 weeks the groups were similar, which may have been confounded by significant crossover of patients from the control to intervention group. One SR⁴⁰ pooled results from two RCTs and reported an increased likelihood of improved breastfeeding as evaluated by the mother in the group that received frenotomy compared to placebo. Of note, the incremental improvements were greater in the non-blinded²⁷ versus the blinded study.⁵⁰ Of the primary studies identified, one RCT reported that subjective breastfeeding scores improved following both simple release and Z-plasty procedures, with no differences observed between groups.⁴¹ Further, jaw locking was improved following both procedures with no differences between groups, suggesting better latching.⁴¹ One NRS⁴⁴ reported that the majority of mothers communicated that they experienced either a significant or moderate improvement in breastfeeding after correction of anterior or posterior ankyloglossia with or without concomitant lip-tie. The statistical significance of these findings is unclear as no analysis was conducted.⁴⁴ One NRS⁴³ reported a significant improvement in the average number of sucks in the three first groups of sucking after surgery in the infants who underwent frenotomy. Before surgery, quantity was lower than infants without tongue tie, and there were no differences after surgery. The same was observed for pause length.⁴³ One NRS⁴⁵ reported that maternal ratings of improvement in breastfeeding were higher in patients who received frenotomy, versus support from an infant feeding coordinator, and that this effect did not differ in children older or younger than 30 days of age.⁴⁵ However, group sample sizes were small and imbalanced and results should be interpreted with caution.

Regarding long-term maternally reported outcomes, one SR^4,39 reported that based on longer-term follow-up data, the majority of mothers reported improved feeding, approximately half reported complete resolution of feeding issues, and the majority of infants were still breastfed at 3 and 4 to 5 months in one study.⁵¹ No differences in breastfeeding effectiveness at 2 weeks,²⁸ or latch score at 8 weeks,³⁰ were reported by two other studies. One study³⁵ reported that a greater number of patients who received the procedure were still breastfeeding and for a longer duration than those who received no intervention. Similarly, a numerically lower proportion of patients reported discontinuation of breastfeeding due to pain.³⁵ None of the primary studies reported on long-term breastfeeding efficacy outcomes.

Regarding objective scales or measures of breastfeeding efficacy, one SR^4,39 reported on two studies that assessed the Infant Breastfeeding Assessment Tool (IBFAT) scores. One study reported significantly improved scores in the intervention group compared to sham surgery immediately after surgery but not at 2 weeks follow-up.²⁸ The other study reported no significant differences in IBFAT score at 5 days post-procedure.³⁰ One study assessed a score adapted from the LATCH tool (latch, audible swallowing, nipple type, mother’s level of comfort, and help the mother needs to hold her infant to the breast) and IBFAT and reported 50% improvement in the intervention and 40% improvement in the control group, which was not significantly different.²⁷ One study assessed LATCH score alone and reported that there were no significant differences between the intervention and control group at 5 days post-intervention.³⁰ One SR⁴⁰ pooled results from two NRSs^52,53 and reported significant improvement in LATCH scores after frenotomy. This finding was supported by observational studies presented narratively in this review⁴⁰ which reported improvements in sucking or latch ranging from 57% to 92% after frenotomy.^22,54–57 Of the primary studies assessed, one NRS⁴⁵ reported significant improvements in IBFAT score in the frenotomy group post-intervention, while the scores in the control group stayed the same. Due to small and imbalanced group sample sizes these results may not be reliable.

Maternal Nipple Pain

One SR⁴⁰ conducted a meta-analysis of three NRSs^52,57,58 and reported that there was a significant reduction in nipple pain after the procedure. Two SRs^4,39,40 reported narratively on the findings of four RCTs.^27,28,30,55 Two studies reported no difference in pain scores versus sham operation immediately post-procedure²⁷ or after 5-days,³⁰ while the other two reported significant improvement in pain versus sham operation²⁸ and compared to pre-procedure values,^28,29 immediately,^28,29 and after four weeks.²⁸ One SR⁴⁰ presented further findings from five NRSs,^{52,53,55,57,58} four of which suggested an improvement in pain scores following the procedure. Of the primary studies identified, one RCT⁴¹ reported reduced breast pain in patients who received simple release or Z-plasty with no significant differences between groups. One NRS⁴² reported a reduced incidence of breastfeeding related problems (e.g., breast pain and cracked nipples) at 48 hours and a mean follow-up of four months, although the long-term follow-up findings may be unreliable due to potential confounding and unclear statistical significance. One NRS reported that based on qualitative assessment of breastfeeding, the quantity of reported problems related to breastfeeding technique or effectiveness were reduced after the procedure.⁴³

Feeding Sequelae and Growth

One SR^4,39 presented limited and primarily non-comparative evidence on feeding sequelae. Three studies suggested numerically greater improvements in feeding problems (i.e., dribbling and excess gas),⁵⁰ eating difficulty (i.e., ability to clean teeth with tongue, lick outside of lips and eat ice cream) at 3 years,⁵⁹ and a greater reduction in the number of breastfeeding and supplementary bottle feeding sessions at 2 weeks post-procedure.⁵⁵ One SR⁴⁰ reported on milk supply and production, noting a significant increase in milk transfer and 24 hour milk production in a sample of six mothers in one study.⁵² Breastfeeding continuation rates were reported on by eight uncontrolled studies,^{22,23,27,32,50,51,54,56} which suggested rates between 43% and 78% at 3 months, which was noted to be higher than the United Kingdom’s national average of 29% by the authors.⁴⁰ A single study addressed weight gain and reported that neonates gained significant weight at 2 weeks post-procedure.⁵⁵ Of the primary studies identified, one NRS⁴² reported an increase in the rate of exclusive breastfeeding at 48 hours, which trended back to baseline at a mean follow-up of four months. Formula milk use was reduced at 48 hours, but increased above baseline at follow-up.⁴² No statistical analysis was conducted at long-term follow-up so the significance of these findings is unclear. This study⁴² reported reduced odds of women reporting frequent or prolonged feed, shallow latch, and infant fussiness and restlessness at 48 hours.

Parent Satisfaction

Of the primary studies identified, one RCT⁴¹ reported that Z-plasty resulted in greater parent satisfaction with the procedure versus simple release.

Harms

Adverse events of the procedure were mainly addressed by non-comparative evidence including case series and case reports, and were often self-reported. Both SRs^4,39,40 presented data on harms. The majority of studies reported no significant harms or minimal harms. The most common harm reported was minor and/or limited bleeding. Other potential harms were re-operation and scarring,³⁵ development of a healing slough requiring a week long recovery,³⁰ pain,⁶⁰ and case reports noted surgical site infection, swelling, post-surgical mucous cyst, and hemorrhagic shock after administration of procedure by an untrained worker.⁶¹

Of the primary studies identified, one RCT reported a single case of minor hemorrhage not requiring surgical intervention and a single case of re-operation.⁴¹ One NRS⁴² reported five cases of re-operation, in some case requiring multiple procedures. One NRS⁴⁴ reported no occurrences of self-reported complications.