Types of studies

We included all randomised controlled trials (RCTs), including quasi‐randomised and cluster‐randomised trials. We excluded cross‐over trials as they are an inappropriate study design (as the intervention may have a lasting effect that compromises entry to subsequent periods of the trial). We excluded observational studies (including cohort studies, case‐control studies, etc.). We included both clinical and community‐based trials.

Types of participants

We included pregnant women, regardless of their dental caries, exposure to fluorides, level of dental treatment, nationality or level of education. The women may or may not have had access to fluoridated water (naturally or artificially).

Types of interventions

We included studies of fluoride supplementation (tablets, drops, lozenges or chewing gum) of any dosage, frequency, duration and timing of delivery, which may or may have not included the use of topical fluorides such as fluoride dentifrice, fluoride rinse and topical fluoride application, compared with no fluoride supplementation.

Control group: no treatment or placebo.

Types of outcome measures

Electronic searches

Cochrane Oral Health's Information Specialist conducted systematic searches in the following databases for randomised controlled trials and controlled clinical trials. There were no language, publication year or publication status restrictions:

• Cochrane Oral Health's Trials Register (searched 25 January 2017) (Appendix 1);

• Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 11) in the Cochrane Library (searched 25 January 2017) (Appendix 2);

• MEDLINE Ovid (1946 to 25 January 2017) (Appendix 3);

• Embase Ovid (1980 to 25 January 2017) (Appendix 4);

• LILACS BIREME Virtual Health Library (Latin American and Caribbean Health Science Information database; 1982 to 25 January 2017) (Appendix 5);

• CINAHL EBSCO (Cumulative Index to Nursing and Allied Health Literature; 1937 to 25 January 2017) (Appendix 6).

Subject strategies were modelled on the search strategy designed for MEDLINE Ovid.

Searching other resources

We searched the following trial registries for ongoing studies:

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (clinicaltrials.gov; searched 25 January 2017) (Appendix 7);

• World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch; searched 25 January 2017) (Appendix 8).

We searched the reference lists of included studies and relevant systematic reviews for further studies.

We did not perform a separate search for adverse effects of interventions used, and we considered adverse effects described in included studies only.

Selection of studies

Two review authors (Rena Takahashi (RT) and Keika Hoshi (KH)) independently screened the titles and abstracts (when available) of all reports identified through the electronic searches and handsearching that were entered into EndNote X7 software. We obtained the full text of potentially relevant studies or studies where it was difficult to make a clear decision from only the title and abstract. The full‐text articles were assessed independently by two review authors to determine if they met the inclusion criteria. We resolved any disagreements by discussion. If this was not possible, we consulted a third review author. At the time of exclusion, we recorded reasons for exclusion in the 'Characteristics of excluded studies' tables.

Data extraction and management

Two review authors (RT and Erika Ota (EO)) independently extracted data using data extraction forms. We extracted data related to settings, participants (e.g. inclusion and exclusion criteria for pregnant women), interventions (e.g. type of intervention, comparison), outcomes (e.g. outcomes reported in the paper, duration and rates of follow‐up, adverse effects), methods (e.g. study designs, randomisation methods) and other information (e.g. pharmaceutical sponsorship data). We recorded if clinical trials reported the presence of calcium in the fluoride supplement. We noted the topical fluoride exposure of children (up to 6 years of age) in the follow‐up period of the studies. We included unpublished research data from recognised research groups and experts in the field, obtained via personal contact if the study was in the clinical register.

We resolved any disagreements by discussion and consultation with a third review author. In this review, we did not contact trial authors. In future updates, we will contact trial authors (if possible) asking for assistance with data transformation or raw data, if the data are not reported in a format suitable for analysis.

Assessment of risk of bias in included studies

Two review authors (KH and Yoshihiro Toyoshima (YT)) assessed the risk of bias independently for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved any disagreements by discussion or by involving a third assessor (EO). We assessed the following domains of risk of bias: (1) random sequence generation (selection bias), (2) allocation concealment (selection bias), (3) blinding of participants and personnel (performance bias), (4) blinding of outcome assessors (detection bias), (5) incomplete outcome data (attrition bias), (6) selective reporting (reporting bias) and (7) other bias (recruitment bias, bias influenced by funding source, etc.).

Each domain was assessed as at either low, high or unclear risk of bias. We categorised the overall risk of bias of individual studies as follows:

• low risk of bias (plausible bias unlikely to seriously alter the results) if all domains were at low risk of bias;

• unclear risk of bias (plausible bias that raises some doubt about the results) if one or more domains had an unclear risk of bias, but none at high risk of bias;

• high risk of bias (plausible bias that seriously weakens confidence in the results) if one or more domains were at high risk of bias.

Measures of treatment effect

For dichotomous outcomes, we calculated risk ratios for differences between the intervention and comparison groups, along with 95% confidence intervals (CI). For continuous outcomes, we calculated the mean difference (MD) and 95% CIs where means and standard deviations (SD) were presented or were calculable. We did not calculate the standardised mean difference (SMD) in this review. In future updates, where a continuous outcome is measured using different scales, we will calculate the SMD and SDs.

Unit of analysis issues

We did not include any cluster‐randomised trials in the analyses. In future updates, if we include cluster‐randomised trials, we will adjust their sample sizes using the methods described in Section 16 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) using an estimate of the intracluster correlation coefficient (ICC) with either data from the trials (if possible), data from a similar trial or data from a study of a similar population. We will conduct sensitivity analyses to see the effect of variation in the ICC when we use ICCs from other sources. We plan to combine the relevant outcomes from both cluster‐randomised and individually randomised trials if there is no substantial heterogeneity between the two study designs.

Dealing with missing data

In this review we did not contact trial authors. In future updates, where possible, we will contact trial authors to provide missing data. We will note levels of attrition for included studies. We will conduct sensitivity analysis to check the impact of including studies with substantial levels (more than 20%) of missing data in the overall assessment of the intervention effect.

If possible, we will conduct analysis on an intention‐to‐treat basis. If there are missing data, the denominator for each outcome in each trial will be the number randomised minus any participants whose outcomes are known to be missing in the analysis. If there are missing standard deviations of the continuous data, we will use the methods described in the Cochrane Handbook of Systematic Reviews of Interventions Section 7.7.3 (Higgins 2011) to estimate them.

Assessment of heterogeneity

We could not assess heterogeneity due to the inclusion of only one study. In future updates, we will assess heterogeneity by inspection of forest plots of the estimates and confidence intervals of treatment effects. We will assess statistical heterogeneity in each meta‐analysis using the I² and Chi² statistics. We defined substantial heterogeneity as I² greater than 50%, or if there was a low P value (less than 0.10) in the Chi² test for heterogeneity.

Assessment of reporting biases

In future updates, if more than 10 trials are identified for any meta‐analysis, we will assess publication bias using visual assessment of funnel plots according to recommendations in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Should asymmetry be identified in the contour‐enhanced funnel plots, we will investigate possible causes by performing exploratory analyses.

Data synthesis

We could not conduct data synthesis as only one study was included in the review. In future updates, we will conduct meta‐analyses for studies with comparable analyses that report similar outcome measures using forest plots in Review Manager software (RevMan) (RevMan 2014). We will combine risk ratios for dichotomous data and mean differences for continuous data using random‐effects models, assuming that the identified studies allow this procedure. If random‐effects analyses are conducted, we will present the results as the average treatment effect with 95% CIs, and estimates of I². If there are few studies or small sample sizes, it may be impossible to estimate between‐study variance with any precision. In that case, a random‐effects analysis would provide poor estimates of the distribution of intervention effects, therefore we would use a fixed‐effect model (Higgins 2011).

In future updates, we will analyse cluster‐RCTs at the individual level. We will meta‐analyse results from appropriately analysed cluster‐RCTs using the generic inverse variance method in RevMan. If original analyses do not account for clustering, we will conduct an adjusted analysis, provided that the necessary information (e.g. mean cluster size, proportion of individuals with events, ICCs) can be extracted (Higgins 2011). We will calculate the prevented fraction.

Subgroup analysis and investigation of heterogeneity

We could not assess heterogeneity by inspection of forest plots of the estimates and confidence intervals of treatment effects as only one study was included in the review. In future updates, if we identify substantial heterogeneity in the primary outcomes, we will conduct subgroup analyses for these relevant and clinically meaningful subgroups where sufficient data are available. In studies with more than one intervention group, such as those comparing different frequencies of application or different types of supplements, we will consider the results from all relevant experimental groups separately in the meta‐analyses.

Moreover, we could not conduct two different subgroup analyses:

• low‐ and middle‐income countries versus high‐income countries (defined by World Bank criteria);

• region of high‐level fluoride concentration in tap water or well water (1.5 mg/L or more, 0.3 mg/L or more) versus region of low‐ to medium‐level fluoride concentration.

Sensitivity analysis

In future updates, we will undertake sensitivity analysis for primary outcomes based on the risk of bias assessment for allocation concealment and attrition rates. We will redo analyses and remove studies that are at high risk of bias for these domains in order to assess whether this makes any difference to the overall result.