Survival of smart restoratives in pediatric dentistry: a PRISMA-guided meta-analysis of randomized trials

2.1 Selection criteria

Eligibility was defined using a pediatric restorative Patient/Problem, Intervention, Comparison, Outcome (PICO) framework (Table 1). Randomized controlled trials (parallel, split-mouth, or cluster/child-level) involving children or adolescents receiving direct posterior restorations were included. Interventions included ion-releasing restorative systems or GIC-related technique modifications, with resin composites as the primary comparator. Trials reporting extractable arm-level failure outcomes based on ART or USPHS/Ryge criteria at any follow-up were eligible. Non-randomized studies, adult-only cohorts, prevention-only trials, and protocols without results were excluded. In addition to the primary comparison of ion-releasing materials versus composite, trials comparing different ion-releasing formulations (e.g., GIC vs. GIC) were included to address the prespecified secondary objective of evaluating within-class differences.

2.2 Data sources and search strategy

Systematic searches of PubMed (MEDLINE), EMBASE (Ovid), and Cochrane CENTRAL were performed from inception to 18 September 2025 with no language or year restrictions. Strategies combined controlled vocabulary and free-text terms for pediatric dentition, ion-releasing restorative materials, and randomized trial designs. Full database-specific search strings are provided once in Supplementary Table 1 to avoid repeating similar terms within the text. Reference lists of all included studies and relevant reviews were manually screened for additional records. All citations were managed in Zotero for de-duplication before screening.

2.3 Selection of studies

Two reviewers independently screened titles/abstracts and full texts using predefined eligibility criteria. Disagreements were resolved by discussion or third-reviewer adjudication. Prevention-only trials were catalogued separately and not pooled with restoration-failure analyses. The PRISMA flow diagram (Fig. 1) summarizes study selection. Prevention-only trials identified during screening were excluded from all restoration-survival analyses but are briefly summarized separately to maintain PRISMA-compliant reporting transparency.

Fig. 1.PRISMA 2020 flow diagram of study selection.

2.4 Data extraction

A standardized, piloted extraction form captured study identifiers, design characteristics, participant age range, tooth type, intervention/comparator details, and outcome assessment. Data extraction also included allocation concealment and specification of dentition type (primary or permanent) for each trial. Arm-level failures and totals at the longest follow-up were recorded for risk ratio calculation, and hazard ratios were extracted for time-to-event studies. Multi-arm, split-mouth, and cluster trials were handled with appropriate combination or design-effect adjustments. All extracted data were double-checked by two reviewers.

2.5 Risk-of-bias assessments

Risk of bias was assessed using the Cochrane RoB 2 tool [11] for failure outcomes at the longest follow-up. Two reviewers conducted assessments independently, resolving differences by consensus or third-reviewer input. Domain-level and overall judgments are summarized in Supplementary Table 2 (Ref. [9, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]).

2.6 Meta-analytic methods

The primary effect measure was the risk ratio (RR) for restoration failure comparing ion-releasing materials with composites in primary teeth. Random-effects models (DerSimonian-Laird τ², Hartung-Knapp confidence interval (CI)) were used when ≥2 studies contributed data. For survival outcomes, log-hazard ratios and standard errors were synthesized separately. Prespecified subgroup analyses examined tooth type, material class, evaluation criteria, and technique/adjunct comparisons. All analyses were conducted in R using the meta and metafor packages.

3.1 Study selection

Database searches identified 380 records (PubMed 153; Cochrane 173; EMBASE 54). After removal of 227 duplicates, 153 records remained for title/abstract screening. Of these, 115 were excluded, and 38 full-text articles were assessed for eligibility. Sixteen randomized trials [9, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26] met the inclusion criteria and were included in the review; the remaining 22 full-texts were excluded with documented reasons (e.g., non-randomized design, prevention-only outcomes, no extractable arm-level data, or protocol-only). The study selection process is depicted in the PRISMA 2020 flow diagram in Fig. 1.

3.2 Characteristics of included studies

The 16 included Randomized controlled trials (RCTs) [9, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26] were conducted in school, community, and university settings across Asia, South America, and Europe. Most evaluated posterior restorations in primary molars using ART or USPHS/Ryge criteria at 12–24 months. Two trials were conducted on permanent molars in children, including teeth affected by molar incisor hypomineralization (MIH). Interventions comprised high-viscosity or glass-hybrid GICs, resin-modified/bioactive GICs, giomer/S-PRG systems, alkasites, and GIC-related technique modifications. Key characteristics are detailed in Supplementary Table 3 (Ref. [9, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]).

3.3 Risk-of-bias assessment

Sixteen randomized trials were assessed. Overall, most were judged to have some concerns of risk of bias, reflecting frequent uncertainties in sequence concealment and potential unblinding of outcome assessment. One trial was judged low risk overall, one was judged high risk, and the remaining fourteen had some concerns The summary of the characteristics of the included RCTs is given as Supplementary Table 3. For example, unclear allocation concealment was noted in several trials such as Maru et al. [13], Deepika et al. [14], and Phonghanyudh et al. [15]. Potential lack of assessor masking was evident in studies where material appearance could reveal allocation, including Akman et al. [12] and Pesaressi et al. [9]. Missing outcome data at the longest follow-up contributed to “some concerns” ratings in trials such as Garbim et al. [19] and Moura et al. [25]. These study-specific methodological issues are detailed in Supplementary Table 2 and are referenced here to improve transparency.

By domain, concerns most commonly arose in the randomization process (unclear allocation concealment in several parallel and split-mouth designs) and in measurement of the outcome (assessor masking sometimes uncertain or infeasible due to recognizable material appearance, despite use of standardized criteria such as ART or USPHS/Ryge). Missing outcome data was generally low to moderate—attrition was typically balanced and below thresholds likely to bias results, though a few studies reported notable losses by the longest follow-up. Deviations from intended interventions were largely low risk given standardized clinical protocols, and selection of the reported result was usually low risk where protocols or prespecified outcomes were available.

For the primary meta-analysis set (smart/ion-releasing vs. composite in primary teeth), both contributing trials were rated with some concerns overall (unclear concealment and/or potential assessor unblinding). Technique/adjunct comparisons within GICs also raised some concerns, while the single low-risk judgment arose in a permanent-tooth trial (MIH). These patterns should be considered when interpreting the pooled effects.

3.4 Qualitative synthesis

Sixteen randomized trials with results were included in the narrative synthesis. The studies were conducted in school or university settings across Asia, South America, and Europe, and enrolled children from preschool age to adolescence. Designs were predominantly parallel-group randomized trials, with several split-mouth studies. Outcomes were most often assessed with ART or USPHS/Ryge criteria at 12–24 months, and one trial reported survival using hazard ratios. Risk of bias was commonly judged as “some concerns”, typically owing to unclear allocation concealment and the practical difficulty of masking examiners when materials were visibly distinct in Supplementary Table 2. These technique-related trials are presented as an exploratory subgroup because they evaluate procedural modifications within ART/GIC protocols; however, the interventions themselves are mechanistically different, and the results should therefore be interpreted cautiously.

For the primary question—smart/ion-releasing restorative materials versus conventional composites in primary teeth—two trials contributed data at the longest follow-up. Akman et al. [12] (2020) compared a high-viscosity glass-ionomer (Equia Fil) against multiple composite comparators under USPHS at 12 months and observed very few failures overall, with a numerical advantage for composites. Maru et al. [13] (2022) contrasted Equia Forte with a silorane composite using Roeleveld criteria at 24 months and reported more failures in the GIC arm than in the composite arm. Considered together, these trials provide no clear evidence that smart/ion-releasing materials outperform conventional composites over 12–24 months; the findings trend toward fewer failures with composites but remain imprecise because of limited study numbers and differing outcome frameworks.

Head-to-head comparisons within ion-releasing systems did not reveal consistent superiority of any single class. Deepika et al. [14] (2022) compared a Resin-Modified Glass Ionomer (RMGI)/bioactive material (ACTIVA) with a giomer in a split-mouth design and found comparable retention at 12 months. Garbim et al. [19] (2024) compared two GIC formulations (Equia Forte versus Riva Self Cure) using survival analysis over 24 months and reported no significant difference in hazard of failure. Moura et al. [25] (2020) reported larger arm totals for Ketac versus Vitro at 12 months but did not permit the extraction of explicit per-arm failure counts from the published tables; adjusted analyses favoured Ketac. Faustino-Silva et al. [27] (2019) followed toddlers for 1–4 years and found high overall ART success without clear between-material differences. Collectively, ion-releasing materials appeared to perform similarly to one another over the observed time frames, with signals dependent on outcome definitions and follow-up length.

Technique and adjunct modifications within GIC showed clearer signals. Pesaressi et al. [9] (2024) demonstrated that adding retentive grooves to Class II ART preparations reduced failures relative to no grooves at 12 months, suggesting a tangible benefit of additional mechanical retention in proximal lesions. Phonghanyudh et al. [15] (2011/2012) observed numerically fewer failures when GIC restorations followed rotary burr caries removal rather than hand excavation at 12 months, although the difference was modest and derived from a single study. Ferreira et al. [16] (2013) reported substantially fewer overall failures when a medicated antibiotic liner was used beneath a GIC compared with a conventional liner at 12 months, but the composite endpoint and small sample size warrant cautious interpretation. These trials together indicate that how GICs are applied—particularly the use of retentive grooves—can materially influence clinical performance.

Two trials in permanent teeth of children (adolescents and MIH-affected molars) compared GIC variants and reported broadly similar failure rates. D’Costa et al. [17] (2020) contrasted metal-reinforced with high-strength GIC at 18 months, and Mahfouz et al. [18] (2025) compared zirconia-reinforced with glass-hybrid GIC at 12 months in MIH. Pooled estimates for this subgroup did not identify a clear difference and showed moderate heterogeneity attributable to differences in material class, clinical context, and follow-up.

Four randomized trials addressed prevention rather than restoration survival and were summarized separately. Lam et al. [20] (2024) found no difference between Sodium fluoride (NaF) varnish and GIC sealant for 12-month progression to dentinal caries in child-randomized cohorts, alongside low sealant retention. Guler et al. [21] (2013) reported lower caries presence and better retention with Fuji VII relative to an ormocer sealant at 24 months in a split-mouth design. Schraverus et al. [22] (2021) showed that GIC sealants reduced new dentinal caries versus no sealant over 12 months in MIH without affecting posteruptive breakdown. Rodrigues et al. [23] (2021) demonstrated substantially higher lesion arrest with RMGIC sealants compared with fluoride varnish at 12 months in erupting molars. These studies support sealant-based prevention in appropriate pediatric contexts but are not commensurate with restoration-failure outcomes.

Adverse events and safety outcomes were rarely detailed; where reported, no serious adverse events were attributed to the restorative materials or techniques, and postoperative sensitivity or pulp complications were uncommon within the follow-up horizons.

Overall, the randomized clinical evidence in children indicates no demonstrable advantage of smart/ion-releasing materials over conventional composites for restoration survival in primary teeth at 12–24 months. Differences among ion-releasing systems are small and inconsistent, whereas certain technique choices—most notably the incorporation of retentive grooves for Class II ART—appear to improve performance. These patterns, together with the prevalent “some concerns” risk-of-bias judgments and variability in outcome criteria and follow-up, inform the quantitative results and the certainty assessments presented subsequently.

3.5 Quantitative synthesis/meta-analytic findings

Two randomized trials by Akman et al. [12] and Maru et al. [13] contributed data at the longest follow-up (12 and 24 months). Using a random-effects model (DerSimonian-Laird τ² with Hartung-Knapp adjustment), the pooled risk ratio (RR) for restoration failure favored composites but was not statistically significant: RR = 1.74 (95% CI: 0.67–4.57). Between-study heterogeneity was low (I² ≈ 12%, τ² ≈ 0.18). A continuity correction (0.5) was applied to the Akman et al. [12] study due to a zero cell in the composite arm. These findings indicate no clear survival advantage for smart/ion-releasing materials over conventional composites at 12–24 months. Fig. 2 (Ref. [12, 13]) shows the study-wise effects and pooled estimate.

Fig. 2.Forest plot—primary teeth (smart/ion-releasing vs. composite). Study-specific risk ratios (RRs) for restoration failure at the longest follow-up (Akman et al. [12], 12 months; Maru et al. [13], 24 months). Points show RRs; horizontal bars show 95% CIs; the vertical dashed line indicates no effect (RR = 1). Effect measure: RR of failure comparing smart/ion-releasing materials with conventional composites (RR <1 favors smart/ion-releasing; RR >1 favors composite). Random-effects model (DerSimonian-Laird τ²; Hartung-Knapp CI) used for pooling in the text: pooled RR = 1.74 (95% CI: 0.67–4.57), I² ≈ 12%. A 0.5 continuity correction was applied to one study with a zero cell. Outcomes were assessed using USPHS or ART-style criteria.

Technique influenced outcomes in several trials. In the study by Pesaressi et al. [9], adding retentive grooves to Class II ART preparations reduced failures compared with no grooves at 12 months (10/122 vs. 29/127; RR ≈ 0.36). Phonghanyudh et al. [15] observed numerically fewer failures after rotary burr excavation than hand (ART) excavation at 12 months (10/88 vs. 15/89; RR ≈ 1.49; imprecise, single trial). Ferreira et al. [16] reported fewer overall failures with a medicated antibiotic liner under GIC than a conventional liner at 12 months (4/20 vs. 14/16; RR ≈ 0.23), noting the composite failure definition (clinical + radiographic) and small sample. These point estimates are summarized in Fig. 3 (Ref. [9, 15, 16]).

Fig. 3.Technique/adjunct within GIC—single-study effects (log scale). Risk ratios for failure at 12 months for: (i) retentive grooves vs. no grooves (Pesaressi et al. [9], Class II ART), (ii) hand excavation (ART) vs. rotary bur (Phonghanyudh et al. [15]), and (iii) medicated antibiotic liner vs. conventional liner under GIC (Ferreira et al. [16]). Points show trial RRs with 95% CIs on a logarithmic x-axis; the dashed line indicates no effect. RR <1 favors the first-listed technique/adjunct. Estimates are single-study and not pooled. ART: atraumatic restorative treatment.

Across ion-releasing systems, no consistent superiority was identified. Deepika et al. [14] (2022) found similar retention at 12 months for ACTIVA RMGI vs. giomer (0/33 vs. 1/33; continuity-corrected RR ≈ 0.33). Garbim et al. [19] (2024) compared Equia Forte vs. Riva Self Cure using survival analysis and reported HR = 1.31 (95% CI: 0.85–2.01), indicating no significant difference in hazard of failure. Moura et al. [25] (2020) favored Ketac over Vitro in adjusted analyses, but per-arm failure counts were not extractable for pooling. Because D’Costa et al. [17] and Mahfouz et al. [18] differed in both patient populations (healthy vs. MIH-affected molars) and material types (metal-reinforced vs. zirconia-reinforced GIC), this subgroup is clinically heterogeneous, and the pooled estimate should be regarded as exploratory. Although the statistical heterogeneity was low, this does not imply that the studies are truly comparable; the substantial clinical and methodological differences—particularly in substrate condition, material type, and outcome criteria—limit the validity of pooling, and the combined estimate should therefore be regarded as exploratory.

Two trials contributed to a separate pooled analysis in permanent molars (adolescents/MIH). The random-effects estimate suggested no clear difference between GIC variants: RR = 0.52 (95% CI: 0.21–1.31), with moderate heterogeneity (I² ≈ 45%, τ² ≈ 0.20). The corresponding forest appears in Fig. 4 (Ref. [17, 18]).

Fig. 4.Forest plot—permanent teeth in children (smart vs. smart). Study-specific RRs for restoration failure at the longest follow-up D’Costa et al. [17], 18 months; Mahfouz et al. [18], 12 months in MIH. Points show RRs; horizontal bars show 95% CIs; the vertical dashed line indicates no effect (RR = 1). Effect measure: RR comparing the first-listed smart/ion-releasing material with the comparator (RR <1 favors the first-listed material). With only two studies, the random-effects pooled estimate may not lie midway between individual study effects. Random-effects pooled estimate: RR = 0.52 (95% CI: 0.21–1.31), I² ≈ 45%.

Observed heterogeneity was low for the primary comparison and moderate for the permanent-tooth subgroup; likely sources include follow-up duration (12 vs. 24 months), lesion class, outcome criteria (ART vs. USPHS), and material formulation. Sensitivity checks prespecified in the protocol (restriction to 12-month outcomes; exclusion of studies at higher overall risk of bias; removal of non-standard failure definitions) either reduced the evidence to single-study estimates or did not alter the direction of effect. Funnel plots and Egger’s tests were not conducted because each quantitative set included fewer than 10 studies.

Across pediatric RCTs, smart/ion-releasing materials did not demonstrate superior survival to composites for primary molar restorations over 12–24 months. Technique choices—notably retentive grooves for Class II ART—appear to improve outcomes, while head-to-head differences among ion-releasing systems are small or inconsistent under current evidence.

Given the limited number of eligible trials and variation in outcome criteria, the meta-analytic estimate should be interpreted as exploratory, with the primary interpretive value derived from the narrative synthesis summarizing trends and research gaps.