含氟漱口水预防儿童和青少年龋齿

Types of studies

We included randomised or quasi‐randomised controlled trials where 'blind outcome assessment' was stated or indicated (e.g. caries examinations performed independently of previous results, radiographic examinations registered separately from clinical examinations/added later, examiners clearly not involved in giving treatment, use of placebo described), and in which the length of follow‐up was at least one year/school year. We included cluster‐randomised trials, except when only one cluster was assigned to each study group .

We excluded randomised or quasi‐randomised controlled trials with open outcome assessment or no indication of blind assessment of outcome (blind assessment was considered unlikely if the following were not described: a caries examination performed independently of previous results, X‐rays registered independently of clinical examination, examiners clearly not involved in giving treatment and use of placebo), or lasting less than one year/school year, or where random or quasi‐random allocation was not used or indicated. We also excluded split‐mouth studies as they are unsuitable for fluoride mouthrinse owing to unavoidable contamination.

Types of participants

Children or adolescents aged 16 or younger at the start of the study (irrespective of initial level of dental caries, background exposure to fluorides, dental treatment level, nationality, setting where intervention was received or time when it started).

We excluded studies where participants were selected on the basis of special (general or oral) health conditions.

Types of interventions

Intervention: topical fluoride in the form of a mouthrinse that is swished and expectorated, not swallowed. We included fluoride mouthrinses irrespective of formulation, concentration (ppm F), volume, duration or frequency of application, or application technique of application.

Comparison: placebo or no treatment.

Therefore, the following comparison is of interest: fluoride mouth rinse versus placebo or no treatment.

We excluded studies where the intervention consisted of use of any other caries‐preventive agent or procedure (e.g. other fluoride‐based measures, chlorhexidine, sealants, oral hygiene interventions, xylitol chewing gums), in addition to fluoride rinse.

Types of outcome measures

The primary outcome measure in this review is caries increment, as measured by change from baseline in the number of decayed (missing) and filled permanent tooth surfaces (D(M)FS), or in the number of decayed (extracted/missing) and filled primary tooth surfaces (d(e/m)fs) or both (and change in the number of permanent or primary teeth (D(M)FT/d(e/m)ft). Dental caries is defined here as clinically and radiographically recorded at the dentin level of diagnosis. If caries data were reported only with dentin and enamel lesions combined, this was used in the analysis. (See Data collection and analysis for different ways of recording caries and reporting D(M)FT/S and d(m)ft/s scores in permanent and primary dentitions in clinical trials of caries preventive interventions, and for ways in which data were selected for analysis.)

We excluded studies reporting no dental caries data, reporting only on plaque/gingivitis/gingival bleeding, calculus, dentin hypersensitivity or fluoride physiological outcome measures (fluoride uptake by enamel or dentin, salivary secretion levels, etc).

Electronic searches

We searched the following electronic databases:

• Cochrane Oral Health’s Trials Register (searched 22 April 2016) (see Appendix 2 );

• Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 3) in the Cochrane Library (searched 22 April 2016) (see Appendix 3 );

• MEDLINE Ovid (1946 to 22 April 2016) (see Appendix 1 );

• Embase Ovid (1980 to 22 April 2016) (see Appendix 4 );

• CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 22 April 2016)(see Appendix 5 );

• LILACS BIREME (Latin American Caribbean Health Sciences Literature; 1980 to 22 April 2016) (see Appendix 6 );

• BBO BIREME (Brazilian Bibliography of Odontology; 1980 to 22 April 2016) (see Appendix 6 );

• Proquest Dissertations and Theses (1861 to 22 April 2016) (see Appendix 7 ); and

• Web of Science Conference Proceedings (1990 to 22 April 2016) (see Appendix 8 ).

We placed no restrictions on the language or date of publication when searching electronic databases.

For ongoing trials, we searched the following trial registries (see Appendix 9 for details of search terms):

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov ( http://clinicaltrials.gov/ ; searched 22 April 2016);

• World Health Organization International Clinical Trials Registry Platform ( apps.who.int/trialsearch ; searched 22 April 2016).

Searching other resources

Selection of studies

At least two review authors performed screening for eligibility independently for all reports identified from all searches performed. We considered it essential to identify all reports related to the same study. When a trial report thought to be potentially relevant was written in a language not known to the review authors, it was translated and the inclusion criteria form completed by a review author with reference to the translator. We attempted to contact authors of trials that could not be classified to ascertain whether inclusion criteria were met. We noted trials not fulfilling the inclusion criteria and our reasons for excluding them in the Characteristics of excluded studies table.

Data extraction and management

At least two review authors extracted data from all included studies in duplicate using a predesigned pilot‐tested data extraction form. We extracted numerical data presented only in graphs and figures whenever possible. We attempted to contact study authors by using an open‐ended request to obtain missing information or for clarification when necessary.

We extracted information related to study methods, including study design, study duration (overall length of follow‐up in years) and objectivity/reliability of primary outcome measurement (diagnostic methods and thresholds/definitions used and included, and monitoring of diagnostic errors).

We recorded information on sponsoring/funding institutions and manufacturers involved.

We extracted characteristics related to participants, including age (mean or range or both) at start, caries severity at start (average DMFS/dmfs, DFS/dfs or other caries increment measure, for sample analysed), background exposure to other fluoride sources (toothpaste, water, etc), year study began, location where study was conducted (country), setting where participants were recruited (and setting of treatment) and total sample randomised (at baseline) and analysed (at relevant final examination).

We extracted characteristics of the interventions, including mode of application (how the intervention was delivered/supervision), methods (technique/device) of application, before and after application, fluoride active agents and concentrations used (in ppm F), frequency and duration of application and amount applied. We recorded information on what the fluoride mouthrinse was compared with (no treatment or placebo), together with numbers for each group. We have described these data in the Characteristics of included studies table.

We recorded different ways of reporting caries increment (change from baseline as measured by the DMF index) separately and/or combined according to the components of the index chosen and units measured (DMFT/S, or DFT/S, or DT/S, or FT/S), types of tooth/surface considered (primary/permanent teeth/surfaces, first molar teeth approximal surfaces, etc), diagnostic thresholds used (cavitated/dentin lesions, non‐cavitated incipient enamel lesions or both), methods of examination adopted (clinical or radiographical, both or other), state of tooth eruption considered (teeth erupted at baseline and/or erupting teeth (or surface) during the trial) and approaches to account or not for reversals in caries increment adopted (in a net caries increment or observed/crude increment, respectively). In addition, we recorded caries increment data at all reported time periods (at various follow‐ups).

As we were aware that caries increment would be recorded differently in different trials, we developed a set of a priori rules to choose the main outcome data (D(M)FS) for analysis from each study: DFS data would be chosen over DMFS data, and these would be chosen over DS or FS; data for 'all surface types combined' would be chosen over data for 'specific types' only; data for 'all erupted and erupting teeth combined' would be chosen over data for 'erupted' only, and these over data for 'erupting' only; data from 'clinical and radiological examinations combined' would be chosen over data from 'clinical' only, and these over 'radiological' data only; data from 'clinical and FOTI examinations combined' would be chosen over data from 'clinical' examination only; data for dentinal/cavitated caries lesions would be chosen over combined data for dentinal/cavitated and for enamel/non‐cavitated lesions, and these over enamel caries data only; net caries increment data would be chosen over crude (observed) increment data; and follow‐up nearest to three years (often the one at the end of the treatment period) would be chosen over all other lengths of follow‐up, unless otherwise stated. When no specification was provided with regard to the methods of examination adopted, diagnostic thresholds used, groups of teeth and types of tooth eruption recorded and approaches for reversals adopted, we assumed the primary choices described above.

The Characteristics of included studies table provides a description of all main outcome data reported from each study, with the chosen primary outcome measure featured at the top. When assessments of caries increments were made during a postintervention follow‐up period, we noted the length of time over which outcomes were measured after the intervention ended. We also listed in this table all other relevant outcomes identified as assessed in the trials.

Assessment of risk of bias in included studies

At least two review authors independently undertook assessment of risk of bias in all included trials. We resolved disagreements by discussion or by involvement of another review author. This was carried out using the tool of The Cochrane Collaboration for assessing risk of bias, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions ( Higgins 2011 ), but according to predefined criteria that were adapted and refined for the Cochrane topical fluoride review updates. We assessed eight domains according to the tool, namely, sequence generation, allocation concealment, blinding of participants/personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, balance of baseline characteristics and freedom from contamination or co‐intervention. Each domain included one or more specific entries in a 'Risk of bias' table. Within each entry, we described information reported in the study and assigned a judgement related to risk of bias for that entry. When the study clearly reported the methods used, we made a judgement of 'low risk of bias' or ' high risk of bias' as appropriate. Where trial methods were unclear, we judged a domain as at 'unclear risk of bias' until further information becomes available.

After taking into account additional information provided by trial authors, we assessed the overall risk of bias in included trials over all eight domains. We categorised studies as being at overall:

• low risk of bias (plausible bias unlikely to seriously alter the results: all eight domains assessed as at low risk of bias);

• unclear risk of bias (plausible bias that raises some doubt about the results: at least one domain assessed as at unclear risk of bias, but none at high risk of bias); or

• high risk of bias (plausible bias that seriously weakens confidence in the results: at least one domain assessed as at high risk of bias).

Measures of treatment effect

The chosen measure of treatment effect for the primary outcome, caries increment, was the prevented fraction (PF), that is, mean increment in control group minus mean increment in treated group, divided by mean increment in controls. For an outcome such as caries increment, where discrete counts are considered to approximate to a continuous scale and are treated as continuous data, we considered this measure more appropriate than the mean difference or the standardised mean difference because it allows the combination of different ways of measuring caries increment and a meaningful investigation of heterogeneity between trials. It is also simple to interpret.

For outcomes other than caries increment, we planned that we would summarise continuous data as average mean differences (MDs) in treatment effects along with their 95% confidence intervals (95% CIs), or, if different scales were used to measure the same outcome in different trials, standardised mean differences (SMDs) and their 95% CIs. We analysed dichotomous outcome data by calculating risk ratios (RRs) or, for adverse effects of fluoride treatment, risk differences (RDs).

Unit of analysis issues

Dealing with missing data

We decided that when missing standard deviations for caries increments could not be obtained by contacting the original researchers, we would impute these values through linear regression of log standard deviations on log mean caries increments. This is a suitable approach for caries prevention trials because, as they follow an approximate Poisson distribution, caries increments are closely related (similar) to their standard deviations ( Van Rijkom 1998 ).

Assessment of heterogeneity

We assessed heterogeneity by inspecting a graphical display of estimated treatment effects from trials along with their 95% CIs and by conducting formal tests of homogeneity undertaken before each meta‐analysis ( Thompson 1999 ). We quantified this by using the I ² statistic and classified it according to the Cochrane Handbook for Systematic Reviews of Interventions ( Higgins 2011 ). A rough guide to interpretation follows: 0% to 40% might not be important, 30% to 60% may represent moderate heterogeneity, 50% to 90% may represent substantial heterogeneity and 75% to 100% may indicate very substantial ("considerable") heterogeneity.

Assessment of reporting biases

Reporting bias can be assessed as within‐study outcomes reporting bias or as between‐study publication bias.

Data synthesis

We conducted meta‐analyses for the PFs as inverse variance weighted averages in Review Manager 5.3 ( RevMan 2014 ), where the prevented fraction and standard error data [PF (SE)] were entered by using the generic inverse variance (GIV) method. We estimated variances using the formula presented in Dubey 1965 , which was more suitable for use in a weighted average, and for large sample sizes the approximation should be reasonable. Two previous reviews ( Marinho 2013 ; Marinho 2015 ) noted that this formula was inappropriate for studies with small increments, and that the data from such studies were to be excluded from the analysis in this review. We used random‐effects meta‐analyses throughout and analysed primary and permanent teeth separately throughout.

We used random‐effects models to calculate a pooled estimate of effect for outcomes other than caries increment data.

Subgroup analysis and investigation of heterogeneity

We specified three potential sources of heterogeneity a priori, as these formed part of the primary objectives of this review. We hypothesised that the effect of fluoride mouthrinses on caries differs according to: (1) baseline levels of caries severity; (2) exposure to other fluoride sources (in water, in toothpastes, etc); and (3) frequency of application and fluoride concentration. We examined the association of these factors with estimated effects (D(M)FS PFs) by performing random‐effects metaregression analyses in Stata version 12.0 (Stata Corporation, College Station, Texas, USA) using the 'Metareg' command ( Sharp 1998 ).

To allow such investigation, we dealt with relevant data as follows. We calculated data on 'baseline levels of caries' from the study sample analysed (final sample) unless otherwise stated, and we averaged values among all relevant study groups. Data on 'background exposure to other fluoride sources' represented combined data on use of fluoride toothpaste and consumption of fluoridated water (or salt) and were grouped into two categories: one for studies that were based on samples provided with non‐fluoride toothpaste and that were obtained from non‐fluoridated areas (non‐exposed), and another for studies based on samples using fluoride toothpaste or studies in fluoridated communities or both. We considered exposure to water fluoridation when fluoride levels in water were stated to be above 0.3 ppm F. Use of fluoride toothpaste reported for 30% or more of the study sample would indicate exposure to fluoridated toothpaste. When use or non‐use of fluoride toothpaste was not clearly indicated in studies carried out in high‐income countries, we assumed that fluoride toothpaste was widely used from the middle of the 1970s ( Ripa 1989 ); we sought this information from study authors (or obtained it from other sources) when missing from studies carried out in other locations. When data on the year a study had begun were not provided, we calculated a 'probable date' by subtracting the duration of the study (in years) plus one extra year, from the publication date of the study. We have not categorised data on 'frequency of application' and 'concentration applied'. We averaged concentrations in multiple‐arm studies over fluoride mouthrinse groups. We dealt with incomplete data for frequency of mouthrinsing as follows: In studies of supervised daily rinse at school where participants were provided with mouthrinse for home use, we assumed rinsing frequency of 365 times a year if not precisely reported. We assumed rinsing frequency of 320 times a year in studies of 'unsupervised' daily rinse at home (even if instructions to rinse more than once a day were given); we assumed frequency of 160 times (days) a year when it was not precisely reported in studies of supervised daily rinse at school where children were not provided with any rinse for home use; frequency of 30 times a year for weekly rinse at school and frequency of 17 times a year for fortnightly rinse at school.

We investigated further potential sources of heterogeneity by metaregression ‐ for different types of control groups (placebo (PL) or no treatment (NT), length of follow‐up (years) and dropout rate (%). These 'post hoc' analyses were reported as such and findings should be treated with caution.

Sensitivity analysis

For the main meta‐analysis of D(M)FS prevented fraction, we planned to undertake a sensitivity analysis including trials with an overall assessment of low risk of bias, but we found no trials satisfying this criterion. We undertook a sensitivity analysis excluding trials where we imputed missing standard deviations. We performed a sensitivity analysis to take account of additional uncertainty related to the cluster‐randomised trial by Ruiken 1987 , and another excluding one trial ( Spets‐Happonen 1991 ) in which a non‐fluoride active agent was present in both fluoride and control groups (the trial was different in this way from all others). We also undertook a sensitivity analysis excluding trials at high risk of bias for allocation concealment, and another excluding trials at high and unclear risk of bias for blinding of outcome assessment. We performed these meta‐analyses using a random‐effects model.