The search strategy used in the previous review for English language articles 11 was replicated for this review update using the search terms: "(GIC sealant* OR Glass ionomer cement sealant) AND (caries OR tooth decay)". Only the start and cut-off dates were changed. The English databases, Biomed Central, Cochrane Oral Health Reviews, Cochrane Library, Directory Of Open Access Journals, Expanded Academic ASAP PLUS, Meta Register Of Controlled Trials, PubMed and Science-Direct, were searched for relevant papers published between 15 January 2008 (the search cut-off date of the original systematic review) and 26 August 2010. Criteria for trial inclusion were:

- 2-or multiple arm clinical prospective study designs;

- Comparison of GIC versus Resin fissure sealants;

- Publication in English.

Included trials were excluded after further review if:

- No outcome measure related to caries was reported;

- No computable data, dichotomous or continuous, per treatment group was reported;

- Resin-modified GIC (RMGIC) was used instead of conventional, chemically cured GIC.

Included trials that passed the exclusion criteria test were accepted for further quality assessment and data extraction. Reviewing, data extraction and quality assessment of the accepted trials was undertaken independently by two reviewers (SM and VY). Differences were resolved through discussion and consensus.

Unlike in the original published systematic review 11 , quality assessment of accepted trials was undertaken on the basis of availability of evidence indicating successful prevention of selection- and detection/performance bias from the start to end of each trial. The new criteria (Table 1) differed from those previously used in the first review 11 . It has been argued that the inclusion of bias-preventing measures (e.g. randomisation, blinding/masking) into the trial methodology only demonstrates an attempt to reduce bias risk but does not carry proof in itself that such attempt was indeed successful and that it is far more important to judge trial quality according to evidence that indicates to what extent such attempt has succeeded 13 . Against this background the quality criteria were adjusted accordingly. For example; if a trial merely reported that randomisation was conducted; reported only the name of the randomisation method used or included a detailed description of the randomisation process without providing any evidence that randomisation was indeed effective throughout the trial; then this was regarded as inadequate.

Potential attrition- and publication bias was not investigated in the original meta-analysis 11 . In this update, sensitivity analysis was done, using the RevMan Version 4.2 statistical software of The Nordic Cochrane Centre, The Cochrane Collaboration (Copenhagen; 2003), in order to investigate potential attrition bias risk in trials. To investigate publication bias a funnel plot was generated, using the datasets from the included clinical trials. The standard error (SE) of the mean differences was plotted on the Y-axis, and the log of the Relative Risk (RR) on the X-axis, using MIX Version 1.7 meta-analysis software 17 . In addition, Egger's linear regression method 18 was used to calculate an intercept with a 95% Confidence Interval (CI), with statistical significance set at α = 0.05.

We anticipated that many of the identified trials would be of split-mouth design. The split-mouth study design is commonly used in dentistry to test interventions and has the advantage of having an individual serve as both the experiment and control. In this study design, one or more pairs of teeth (e.g. primary molars) form the unit of randomisation. Strictly, these pairs are not independent and should be analysed as "paired data" on a patient basis. However, similar to other systematic reviews where split-mouth trials are included 2 we analysed the pairs of tooth surfaces independently for differences whether they are carious or not. The fact that this would result in slightly narrower confidence intervals was considered in the discussion of the overall results.

All data concerning primary and secondary outcomes of accepted trials were extracted either as single dichotomous datasets (containing the number of observed effects (n) and the total number of evaluations (N) for both the control and the test groups) or as single continuous datasets (containing the mean value, standard deviation and total number of evaluations for both the control and the test groups). For dichotomous datasets the Relative Risk (RR, 95% CI), and for continuous datasets the Mean Difference (MD, 95% CI), was computed, using the Cochrane RevMan, Version 4.2 software package. Statistical significance was set at α = 0.05.

Meta-analysis, using RevMan Version 4.2 statistical software by The Nordic Cochrane Centre, The Cochrane Collaboration (Copenhagen; 2003), was considered for datasets only if they complied with criteria for clinical homogeneity. Datasets were considered to be clinically homogeneous if they covered the same type of dentition; type of teeth; study length; type of evaluation method; type of GIC (high- or low viscosity) and stated whether the resin-based fissure sealant material contained fluoride. The percentage of total variations across datasets (I²) was used in assessing statistical heterogeneity 19 . Statistical significance for assessing statistical heterogeneity was set at α = 0.10. A fixed effects model was used for meta-analysis under condition of statistical homogeneity of datasets (p > 0.10) and a random-effects model was used for the others. Pooled datasets were assigned a Mantel-Haenszel weight directly proportionate to the sample size.

Cumulative meta-analysis, using MIX Version 1.7 meta-analysis software 17 , was performed on datasets of consecutive follow-up periods, in order to investigate whether a trend of the available evidence might be observed in line with increasing time after sealant application. Care was taken to select only clinically homogenous datasets. A random-effects model was used under condition of statistical heterogeneity of datasets (p < 0.10).

In addition to the 12 trials included during the original systematic review 20 21 22 23 24 25 26 27 28 29 30 31 , five trials 32 33 34 35 36 were identified during the new database search. A further two trials 37 38 from the hand search and reference check were included (Figure 1). Of these 19 trials, 3 were excluded 34 35 36 , owing to lack of reported caries outcome 35 and reporting on RMGIC as a test material 34 36 . Sixteen trials passed the exclusion criteria and were accepted for data extraction and quality assessment 20 21 22 23 24 25 26 27 28 29 30 31 32 33 37 38 .

Twenty-six dichotomous (DS 01-12,15-28,30) and 4 continuous datasets (DS 13,14,23,29) were extracted from the 16 accepted trials. Characteristics of these trials and their datasets are shown in Table 2. The articles by Forss et al. 37 and Forss and Halme 28 reported each of different datasets from the same trial. Three of the accepted 16 trials followed a parallel group design 25 26 33 , while all other trials were split-mouth studies.

Sixteen of the 30 datasets showed no difference between the two materials after periods lasting from 6 months to 7 years (Table 2 and 3). Twelve dichotomous (DS 05, 10,11,17-21,27,30) and two continuous datasets (DS 13,14) showed statistically significant (p < 0.05) different results. Of these, seven dichotomous datasets (DS 06,10-12,20,21,27) extracted from five trials 22 23 24 27 31 were in favour of resin-based fissure sealants after 2 to 3 years. Five dichotomous datasets (DS 05,17-19,30) extracted from three trials 21 26 38 were in favour of GIC-based fissure sealants after 3 to 5 years. Two continuous datasets (DS 13,14) from one trial 25 reporting on secondary outcomes, such as the DFS and DMFS increment, were also in favour of resin-based fissure sealants after 2 years.

Three of the datasets with statistically significant results (DS 17-19) were derived from one trial including high-viscosity GIC 26 .

Of the 30 datasets, two groups of datasets, DS 04,21,27 after 3 years and DS 05,07 after 4 years, were considered to have each met the criteria for clinical homogeneity and were combined in two meta-analyses. The results were generated in the form of two forest plots (Figure 2 and 3). These datasets included first permanent molars sealed either with low-viscosity GIC or with resin material lacking fluoride and were evaluated by visual, clinical examination.

Figure 2 shows a pooled relative risk of 2.62 (95% CI: 1.71 - 4.00; p < 0.00001), suggesting a 2-3 times higher chance of caries for teeth sealed with low-viscosity GIC than for those filled with resin, after 3 years. Additional analysis established a low risk of statistical heterogeneity (I² = 56.7%, p = 0.13). For that reason a fixed-effects model was used for this meta-analysis.

Figure 3 shows a pooled relative risk of 0.53 (95% CI: 0.07 - 3.73; p = 0.52), suggesting no difference between the chance of caries development in teeth sealed with low-viscosity GIC or resin after 4 years. Additional analysis established a high risk of statistical heterogeneity (I² = 92.8%, p = 0.0002). For that reason a random-effects model was used for this meta-analysis. The significant statistical heterogeneity may be related to the inconsistency in the size of the treatment effects as the trials from which both datasets were extracted were similar in type of materials, sample size, outcome measure, evaluation criteria and method, as well as type of dentition, teeth, age of patients and study period (Table 2). Further investigation, by using the available trial information, did not result in any other possible explanation.

In order to investigate whether a possible trend may be assumed in the comparison of GIC and resin, the chronological results from 5 datasets, concerning sealed teeth after 2 years (DS 06), 2-3 years (DS 21), 3 years (DS 04,27) and 3.64 years (DS 05) 20 21 22 27 31 , were included in a cumulative meta-analysis and its results were generated in the form of a forest plot (Figure 4). As with both meta-analyses (Figure 2 and 3), all datasets included first permanent molars sealed either with low-viscosity GIC or resin material without fluoride and were evaluated by visual, clinical examination. They were thus considered clinically homogenous in all aspects except their follow-up periods. However, a statistical heterogeneity was established (I² = 89.2%; p < 0.00001) that may be attributed to the differences in the length of follow-up periods, so a random-effects model was used. The cumulative Relative Risk indicates no statistical significant difference between the two materials after 5 years (RR 1.33; 95% CI: 0.39 - 4.45; p = 0.65). A shift of the cumulative relative risks towards the value of 1.00, as well as a shift of the cumulative 95% confidence intervals below a relative risk of 1.00 from the period after 2 to 3.64 years can be observed (Figure 4).

Of the accepted 16 trials, only 3 trials followed a parallel group design 25 26 33 and all other were split-mouth studies. Mejáre et al. have cautioned against the split-mouth design as "randomised", as the common practice of including subjects with at least one pair of caries-free molars results in exclusion of caries-active subjects 39 . An obvious selection bias is thus created, as not all subjects will have the same chance to participate. Mejáre et al. have rightfully suggested that the split-mouth trial design should therefore be regarded as "quasi-randomised". Thus, reviews where inclusion criteria include only randomised-control trials should, in theory, exclude trials that use the split-mouth study design. Following the example of other systematic reviews 2 , the data from split-mouth trials was analysed as independent data. This will have caused narrower confidence intervals and thus may have favoured the reported outcomes of one type of treatment above the other. However, the so caused differences in confidence intervals may be considered to only be slight 2 and its correction would not have affected the general impact of selection bias on the relative risk (RR, 95% CI) per dataset (Table 3) due to lack of adequate randomisation 39 40 41 . In addition, wider confidence intervals would have provided only stronger indication of no difference between both types of treatment and would not have changed the overall conclusion from the current analysis results.

All of the accepted trials appear to be limited by risk of selection- and detection-/performance bias. Bias or systematic error may affect studies, causing either an over- or under-estimation of the treatment effect of an investigated clinical procedure. Overestimation has been observed to be the most common 40 . Kjaergard et al. reported a treatment effect overestimation of 48% caused by lack of random sequence allocation 41 and Egger et al. reported a treatment effect overestimation of 54% and 53% due to lack of allocation concealment and lack of evaluator blinding, respectively 42 .

It has been emphasized that selection bias can only be successfully prevented if the allocation sequence remains truly random and free from potential interference throughout the trial 13 14 . Thus, it is important that trials should include an effective process for concealing the random allocation sequence and that by the end of each trial this process has indeed prevented direct observation and prediction of the random sequence allocation 13 14 . Quality assessment in terms of the internal validity of trials should therefore be a measure of the result of random sequence allocation and allocation concealment, and not only of its reported attempt. All trials accepted in this systematic review failed to report not only on evidence of successful sequence allocation and allocation concealment results, but also on necessary details about how sequence allocation and allocation concealment were attempted (Table 4). None of the trials, therefore, provide any guarantee that each patient had an equal chance of being allocated to either treatment group and thus, their allocation may have favoured the outcome of one type of treatment above the other. One measure for testing whether random sequence allocation has not been successful is testing whether covariates differ between treatment groups at baseline 13 . Only three articles had included such a test and reported on its outcome 25 26 33 . The statistically non-significant results (p > 0.05) suggest a successful random allocation (Table 4). However, doubt remains regarding potential bias risk, as other non-balanced covariates may exist, that were not tested for and/or not reported.

From the onset, in all trials successful blinding or masking appeared not to have been possible, owing to the obvious differences in clinical appearance between GIC and resin sealants. For that reason the allocation to either treatment group was visible to patients, operators and evaluators. However, the difficulties of successful blinding still carry the danger of detection-/performance bias, which may thus have affected the trials' results. Potential knowledge of superiority claims prior to the trial may have led patients to change their oral hygiene habits, operators to place restorations more carefully or evaluators to apply evaluation criteria more subjectively. This in turn may have favoured the outcome of one type of treatment over the other.

Sensitivity analysis may be used in establishing whether missing data could have affected trial outcomes by assuming that the numbers of restoration lost to evaluation were either failures or successes 43 . Comparison of the analysis results with reported trial outcomes indicates whether different conclusions should be drawn. Sensitivity analysis was conducted for all datasets. The analysis results differed from reported outcomes of four datasets (DS 06,12,24,25) extracted from three trials 22 24 29 . According to the analysis results, more datasets (DS 24,25) would have been in favour of GIC and fewer in favour of resin if all sealed teeth that were lost to follow-up were assumed to have developed caries in their pits and fissures. How high the caries rate in the teeth lost to evaluation really was remains unknown. Owing to uncertainty regarding the real caries prevalence within those lost to follow-up, the results of the sensitivity analysis cannot serve as evidence that GIC would perform better than resin. However, the validity of the four datasets (DS 06,12,24,25) can be questioned on grounds of attrition bias and thus, their results need to be regarded with caution.

None of these datasets were included into the meta-analyses (Figure 2 and 3) and thus do not affect their results. However, one dataset (DS 06) was also included in the cumulative meta-analysis (Figure 4). A re-computation of the data including all loss-to-follow-up (under assumption of caries) did not change the initially observed trend. Rather, the trend was reinforced, as the change from a cumulative RR after 3.65 years of 1.33 to 1.14 would suggest.

A run-in phase is considered to be a stage during a trial where all patients receive, for example, the test treatment and only those patients that respond well to the treatment are later used for random allocation in either the control or the test group 14 . Such practice would effectively exclude patients from the randomisation process and potentially favour one of the groups. No such run-in phase was indicated in any of the accepted trials (Table 4).

Publication bias was investigated by generating a funnel plot (Figure 5). Publication bias is present when the results of published research differ from those of all the studies that have been done 44 . Funnel plots are scatter graphs showing the size of studies on the Y-axis (large studies on top; small studies at the bottom) and the effect size, observed in these studies, on the X-axis. The effect sizes of larger studies have the tendency to cluster near the mean. Small studies have effect sizes that are dispersed across a wider range. Results of both types of studies, plotted on a scatter graph, give the shape of an inverted, in absence of publication bias, symmetrical funnel 45 . Publication bias affects a funnel plot in the form of a concentration of studies to one side only (asymmetry). Such asymmetry is created when particular smaller studies are published only when they show a larger than average effect. However, if the number of studies (n) is less than ten, any asymmetry may be due to chance and not to publication bias 46 . For that reason the decision was made to plot results of the 26 extracted dichotomous datasets as units of investigation. These are not all independent from the published trials and this forms a departure from the common application of funnel plots in investigating for publication bias. Despite this departure, the use of datasets (instead of published trials) will also indicate potential publication bias when only datasets that show a larger than average effect are published and other datasets are not. In our update, the funnel plot concerning dichotomous data on caries progression showed a symmetrical spread of dataset results (Figure 5). As the visual judgement of funnel plots is subjective, we calculated intercepts (95% CI), using Eggers regression 18 . The calculated non-significant intercept confirmed the observations from the funnel plot. Both suggest that any potential impact of publication bias may be low in regard to this topic.

The extended scope of this update did not change the overall results of the original published systematic review 11 . However, it has to be noted that these results are limited by risk of selection- and detection-/performance bias. As the true extent of such bias impact remains unknown within the reviewed trials, the results need to be regarded with caution.

Five of the accepted studies reported on fluoride exposure of subjects from drinking water 22 23 25 27 30 , a fluoride-rinsing programme 25 and toothpaste 27 . Two trials compared GIC fissure sealants with resin materials that reportedly contained fluoride 23 32 . In addition, most of the trials followed a split-mouth design, providing the possibility of fluoride release from the GIC fissure sealants into the oral cavity. This may have provided a caries-preventive effect for the resin-sealed teeth. It can be assumed that this may have increased caries resistance and thus confounded a potential caries-preventive effect of GIC, as suggested by Hara et al. 47 . Notwithstanding any possible confounder impact, the overall results suggest that resin-based materials may be more caries-preventive after 2-3 years of sealant placement (Figure 2). Such benefit does not translate beyond the duration of 3 years as Figure 3, as well as the majority of datasets without statistically significant outcomes (Table 3), suggests. Other statistically significant outcomes in favour of resin from datasets (DS 06,12), not included in the meta-analysis, may be invalid because of attrition bias and datasets of non-significant outcomes (DS 24,25) could very well have shown statistically significant outcomes in favour of GIC if all studied teeth could have been evaluated at the end of the trials. All these observations suggest that weakness exists in any claims of resin superiority over GIC and further, rather suggest that the caries-preventive effects of both types of material are equal.

These considerations should be regarded as hypothetical, owing to the high risk of selection/detection/performance bias in all accepted trials. Plausibility for this conclusion is provided by the results of the included cumulative meta-analysis (Figure 4). Cumulative meta-analysis is used to provide insights into how much the efficacy of treatments, often reported as mean results with 95% confidence intervals, change over time as evidence accumulates 48 . It is the result of conducting a new meta-analysis each time a new set of evidence emerges 49 . A cumulative meta-analysis not only allows evaluation of any additional contributions made by individual studies to the cumulatively combined results of preceding studies 50 but also allows observation of a trend in the direction of evidence over time. The pooling of datasets of different study periods may be questioned on the grounds that their heterogeneity (due to differences in the length of observation periods) could render the result of a meta-analysis meaningless. On the other hand, it can be assumed that datasets covering longer study periods insert a larger weight into the cumulative results and thus generate a potential trend concerning treatment effects observed after increasing periods of time. The length of an observation period is related to the trial outcome as with its length also the length of, e.g. exposure to cariogenic factors increases. Thus the weight of trial outcomes after a longer observation period would be larger. Such period effect may be investigated through cumulative meta-analysis because it allows for the pooling of outcomes from (otherwise clinically homogeneous) trials after continuously increasing time periods. Should the period effect be stronger for one of the compared treatment types than the other, a trend in the direction of the cumulated point estimates and confidence intervals can be shown by use of a forest plot. Such cumulative results may not provide proof in support of a particular hypothesis due to prevailing statistical heterogeneity. However, the results may assist in providing plausibility for hypothesis development. The forest plot in Figure 4 shows a shift of the cumulative results towards the value of RR 1.00, thus indicating that 'no difference' between low-viscosity GIC and resin may be a more plausible hypothesis with increasing time after sealant application than the hypothesis that 'resin is superior to GIC'. Such hypothesis requires thorough testing through further high quality randomised control trials.

The results of two datasets (DS 17,19) from trials in favour of high-viscosity GIC after 3 and 5 years 26 could not be repeated in later trials 32 33 . However, the results of the latter two trials may not be comparable with the former, as Beiruti et al. did not report on annual results that were independent from those of each previous year in this trial 26 .

Systematic reviews have been reported to provide the highest form of clinical evidence 51 . However, the internal validity of such evidence can only be as good as the internal validity of the trials reviewed. Although the trials accepted in this update may be considered to be less affected by attrition- and publication bias, their risk of selection- and detection-/performance bias is high. Overestimation due to bias may have affected particularly those datasets that indicated a statistically significant difference between both material groups (Table 3) by increasing (or decreasing) the value of point estimates and by artificially narrowing their confidence intervals. However, the precise effect of bias on all results remains unknown. Thus the results need to be regarded with caution and require verification. For that reason, further high quality randomised control trials (RCT) are needed. Such RCTs should adopt a parallel group design and include randomisation and allocation concealment methods that can effectively prevent direct observation and prediction of the allocation sequence. For this purpose, the maximum randomisation method has been suggested 13 . Covariates of both treatment groups should be tested as to whether they differ at baseline (after randomisation). Recently, inclusion of the Berger-Exner test has been suggested, to enable authors of trials to investigate whether selection bias has been introduced into their studies 13 14 . Where bias risk has been found, it may be adjusted statistically 13 . Both outcomes should be included in the final trial report. In order to ensure that the lack of blinding may not have led to favouring one treatment over the other, trials should use and report on procedures and tests employed that may limit or at least monitor potential bias risk. In addition to these recommendations, future trials should base their reporting on the CONSORT statement 52 .