Optimal tooth brushing initiation age and frequency for preventing early childhood caries: a systematic review and meta-analysis

Abstract

Background

Early childhood caries (ECC) is a prevalent global public health concern. Although daily tooth brushing is a key preventive measure, uncertainty remains regarding the ideal age to initiate tooth brushing and the optimal brushing frequency to minimise ECC risk in preschool children. A systematic review and meta-analysis were conducted to explore the association between tooth brushing practices (initiation age and frequency) and the early childhood caries among children under 6 years of age.

Methods

We searched three databases (Proquest, Web of Science, and Scopus) for observational studies (2000–2024). Quality was assessed with the Newcastle–Ottawa Scale, and random-effects meta-analyses (DerSimonian–Laird) were conducted. Heterogeneity (I2), publication bias (Egger’s test), and sensitivity analyses (outlier exclusion) were performed.

Results

Thirty-two studies (27 cross-sectional, 5 longitudinal) out of 2833 initially identified records met inclusion criteria. 10 contributed to the brushing initiation meta-analysis and 8 to the frequency analysis. Including all 10 studies, late initiation showed a borderline nonsignificant pooled odds ratio (OR = 1.40, 95% CI: 0.91–2.16, p = 0.13) with I2 > 90%. Excluding one outlier yielded a significant OR = 1.75 (1.50–2.03, p < 0.0001). Brushing < 2 × /day was associated with 2.11‐fold higher ECC odds (p = 0.0052), also with heterogeneity but no evidence of publication bias (Egger’s p = 0.1161).

Conclusions

Despite considerable between-study variability, the overall evidence suggests that commencing tooth brushing by 12 months of age and brushing at least twice daily maybe protective against ECC, highlighting the importance of integrating these practices into early childhood health programs.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12903-025-07179-5.

Background

Early childhood caries (ECC) is defined as the presence of one or more decayed, missing (due to caries), or filled tooth surfaces in any primary tooth in a child under six years of age [1]. It ranks among the most prevalent childhood diseases worldwide, prompting the World Health Organization (WHO) to identify ECC as a significant global public health challenge [2, 3]. Recent estimates indicate that approximately 1.76 billion children suffer from ECC in their primary teeth [4]. A global analysis of data from 193 United Nations member states (2007–2017) revealed mean ECC prevalence rates of 23.8% among children younger than three years and 57.3% among those aged three to six [5]. If left untreated, ECC can lead to pain, infection, nutritional deficits, and compromised quality of life during these formative years [6]. Furthermore, early caries strongly predicts subsequent dental problems in the permanent dentition, underscoring the long-term health implications of inadequate prevention [7]. Despite various preventive strategies, ECC remains widespread, emphasising the need to clarify how to implement oral hygiene practices effectively in early childhood [8].

Daily tooth brushing is a key preventive measure for ECC as it mechanically disrupts dental plaque biofilms, removes dietary carbohydrates, and delivers fluoride to tooth surfaces when used with fluoridated toothpaste [41]. Uncertainties exist regarding the optimal age at which brushing should begin and the frequency necessary to significantly reduce ECC risk in preschool children [42–44]. Guidelines provide conflicting information, with some advocate initiating brushing with eruption of the first tooth (6 and 12 months of age), whereas others suggest delaying until two years of age may not significantly alter outcomes [16, 44]. Similarly, brushing at least twice per day is commonly promoted to remove plaque biofilms and allow for adequate fluoride application. However, the evidence on the benefits of twice-daily brushing compared to once-daily brushing is inconclusive [15, 27, 32, 45]. Discrepancies in study methodologies, cultural practices, and confounding factors like fluoride exposure and sugar intake further complicate consensus on best practices for parents and caregivers [46].

In this study, we conducted a systematic review and meta-analysis of published observational studies to quantify the association between tooth brushing practices (initiation age and frequency) and early childhood caries in children under 6 years.

Methods

Protocol and registration

This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [47]. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) (ID: CRD42024570017).

Eligibility criteria

This study employed a modified Population, Intervention (Exposure), Comparison, and Outcomes (PICO) framework. The Population comprised healthy children aged 6 months to 6 years with no craniofacial anomalies, immunocompromising conditions, or syndromes known to affect dental development. The Exposure involved tooth brushing initiation age (before, after 12 months of age) or brushing frequency (e.g., ≥ 2 times/day, < 2 times/day). Comparisons were therefore “early vs. late” in age-based studies and “high vs. low” in frequency-based studies. The primary Outcome was the incidence, prevalence, or severity of ECC, such as having at least one decayed, missing, or filled primary tooth, or reporting a decayed, missing, and filled teeth/surfaces (dmft/dmfs) index.

Observational studies (cross-sectional, longitudinal and cohort designs) of children under age six reporting ECC incidence or prevalence were deemed eligible. We focused on observational designs to capture real-world associations of brushing and caries in preschool children, as randomised trials on brushing start/frequency remain rare due to ethical considerations of randomising children to potentially harmful delayed brushing practices. Only articles published in English in peer-reviewed journals between 2000 and 2024 were included to ensure contemporary evidence. Studies that did not clearly report ECC outcomes or brushing exposures, as well as editorials, opinions, and case reports lacking primary data, were excluded.

Information sources and search strategy

In consultation with an experienced librarian, we conducted a comprehensive search of ProQuest, Web of Science, and Scopus for studies published from January 2000 to August 2024. This combination was selected to ensure comprehensive coverage of the biomedical literature. ProQuest provides access to multiple health and medical databases including MEDLINE-indexed content, Web of Science covers high-impact journals across disciplines, and Scopus offers extensive coverage of peer-reviewed literature including nearly all MEDLINE and PubMed indexed journals. We used a combination of Medical Subject Headings (MeSH) and relevant keywords relating to early childhood caries, tooth brushing practices, ensuring alignment with our PICO framework (Table a).

Table a.

Search Strategy

("Preschool Child" OR "Infant" OR "Toddler" OR "young child" OR "early childhood" OR "DMF index") AND ("Toothbrushing" OR "Tooth Brushing" OR "Oral Hygiene" OR "mouth hygiene") AND (Initiation OR "Start Age" OR habits OR commence OR frequency OR start)

Study selection

All retrieved studies were imported into Covidence (https://www.covidence.org/) for systematic screening and selection. Two reviewers (HMA, NA) independently assessed titles and abstracts against the predefined eligibility criteria. Discrepancies were resolved through discussion or arbitration by a third reviewer (TZ). HMA and NA also performed full-text evaluations of articles deemed potentially eligible, documenting the reasons for any exclusions.

Data extraction

Two reviewers (HMA, NA) conducted data extraction in Covidence. Extracted variables included study design, geographic location, sample size, participants’ age range, brushing initiation threshold, brushing frequency category, caries definition (dmft/dmfs or presence/absence of ECC), and adjusted effect sizes (OR or RR) with their 95% confidence intervals (CIs). Where multiple thresholds were reported, those aligning best with “ < 12 vs. ≥ 12 months” or “ ≥ 2 vs. < 2 times/day” were prioritised. Any discrepancies in data extraction were resolved by discussion.

Quality assessment

We used the Newcastle–Ottawa Scale (NOS), adapted for observational studies, to assess methodological quality across three domains (Selection, Comparability, and Outcome) [48]. Each study could earn up to nine stars. Those scoring ≥ 7 stars were considered low risk of bias, 4–6 stars moderate, and < 4 stars high.

Two reviewers (HMA, NA) performed independent scoring, achieving high interrater agreement (Cohen’s kappa ≥ 0.87). When discrepancies arose, reviewers conducted joint review sessions to resolve differences. Disagreements most frequently occurred regarding the "non-respondents" criterion and "Ascertainment of Exposure" domain, particularly whether parental reports of brushing habits constituted validated measurement. In these cases, reviewers re-examined the original papers together, consulting the NOS guidelines and study methods sections. Through discussion and joint review of the evidence, consensus was reached for the studies. The third reviewer (TZ) was available for arbitration if needed. Documentation of all discrepancies was maintained in Covidence.

Statistical analysis

Quantitative data was analysed for 1) early versus late brushing initiation and 2) high versus low brushing frequency. We combined studies that contrasted "early" brushing initiation (< 12 months or < 2 years) with "late" initiation (≥ 12 months or ≥ 2 years) to evaluate effects on ECC risk. This categorisation was selected based on common clinical recommendations and previous literature suggesting 12 months as a meaningful threshold for early intervention. Where multiple thresholds were reported, the category most closely matching " < 12 months vs. ≥ 12 months" was selected to maintain consistency across studies and enable meaningful pooling of results. Similarly, we pooled studies that reported effect sizes comparing brushing ≥ 2 times/day (high frequency) vs. < 2 times/day (low frequency) or once vs. twice daily, as twice-daily brushing is widely recommended in clinical guidelines. If multiple frequency categories existed, we selected the contrast best aligning with " ≥ 2/day vs. < 2/day" to ensure comparability.

Random-effects models were fitted using the DerSimonian–Laird method in the metafor package (R version 4.2.2). We converted each OR or RR and its 95% CI to the log scale to calculate the overall pooled estimate (log(OR) or log(RR)) and variance. Heterogeneity was assessed with the Q-statistic (p < 0.10 considered significant) and I2, and τ2 (tau-squared) was estimated to quantify between-study variance. Where applicable, we performed leave-one-out sensitivity analyses to explore whether single studies dominated the results. Funnel plots and Egger’s regression test (p < 0.05 indicating potential small-study effects) assessed publication bias.

Results

Study selection and characteristics

An initial search across three databases yielded 2826 records, supplemented by 7 additional citations identified through citation searching (Fig. 1). After removing 442 duplicates identified by Covidence, 2391 unique titles and abstracts were screened; 2173 were excluded, leaving 218 articles for full-text review. All 218 articles were successfully retrieved for assessment. A final total of 32 studies met the inclusion criteria, representing 34,050 preschool-aged children from diverse regions, drawn from multiple regions in Europe, Asia, Africa, the Americas, and Australia. Most studies (n = 27) were cross-sectional, while 5 employed prospective or longitudinal designs.

Fig. 1.

PRISMA Flow Chart

Quality assessment revealed distinct methodological patterns across risk categories. High risk studies (n = 2) exhibited major defects, including a lack of representative sampling, no sample size justification, failure to assess non-respondents, poor exposure ascertainment through unvalidated parental recall, and, critically, no control for confounding variables beyond basic demographics. Moderate risk studies (n = 16) commonly demonstrated convenience sampling from single sites, absence of non-respondent analysis, incomplete confounder adjustment, and reliance on self-reported brushing practices without validation. Low risk studies (n = 14) showed only minor limitations, primarily the absence of non-respondent analysis typical in cross-sectional designs, but employed representative sampling through multi-stage recruitment, justified sample sizes, validated exposure assessment methods, comprehensive confounder adjustment, and standardized clinical examinations by calibrated examiners (Table 1).

Table 1.

Quality Assessment of Included Studies

Study ID	Selection (Max 4 stars)	Comparability (Max 2 stars)	Outcome (Max 3 stars)	Total Stars (Max 9)	Risk of Bias
Subramaniam et al. (2012) [9]	⋆⋆	⋆	⋆⋆	5	Moderate
Özen et al. (2016) [10]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Retnakumari et al. (2012) [11]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Wyne et al. (2003) [39]	-	⋆	⋆⋆	3	High
Dini et al. (2000) [12]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Kraljevic et al. (2017) [13]	⋆	⋆	⋆⋆	4	Moderate
Aida et al. (2008) [14]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Pereira et al. (2021) [15]	⋆	⋆	⋆⋆⋆	5	Moderate
Al-Haj et al. (2021) [16]	⋆	⋆⋆	⋆⋆	5	Moderate
Alhabdan et al. (2018) [17]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Bulut & Kilinç (2023) [18]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Ogunlade et al. (2024) [19]	-	⋆	⋆⋆	3	High
Jain et al. (2015) [20]	⋆⋆	⋆	⋆⋆	5	Moderate
Selen et al. (2024) [21]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Li et al. (2011) [22]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Gao et al. (2018) [23]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Cariño et al. (2003) [24]	⋆⋆	⋆	⋆⋆	5	Moderate
Peres et al. (2005) [25]	⋆⋆⋆	⋆⋆	⋆⋆⋆	8	Low
Wagner et al. (2014) [26]	⋆⋆⋆	⋆⋆	⋆⋆⋆	8	Low
Wagner et al. (2017) [27]	⋆⋆⋆⋆	⋆⋆	⋆⋆⋆	8	Low
Skeie et al. (2006) [28]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Prakash et al. (2012) [29]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Hallett et al. (2006) [30]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Elidrissi et al. (2016) [31]	⋆⋆⋆	⋆	⋆⋆	6	Moderate
Wigen et al. (2018) [6]	⋆⋆⋆	⋆⋆	⋆⋆⋆	8	Low
Buckeridge et al. (2021) [32]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Liu et al. (2024) [33]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Sun X et al. (2017) [34]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Sun HB et al. (2017) [35]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Xu et al., (2024) [36]	⋆⋆	⋆⋆	⋆⋆	6	Moderate
Zhang et al. (2019) [37]	⋆⋆⋆	⋆⋆	⋆⋆	7	Low
Chouchene et al. (2022) [38]	⋆⋆	⋆⋆	⋆⋆	6	Moderate

Qualitative synthesis of tooth brushing behaviours

Among the 32 included studies, (27 cross-sectional and 5 longitudinal studies, comprising 34,050 children from diverse regions across Asia, Europe, North America, South America, Africa, and Australia). There was considerable variation in how tooth brushing initiation age and frequency were defined across the studies. Several studies reported that beginning to brush by or before 12 months of age was associated with a significantly lower incidence or severity of ECC incidence and prevalence [9, 10, 15, 34]. For example, Subramaniam et al. [9] found a significant reduction in caries prevalence (27.5% overall prevalence), with children who began brushing before 2 years of age, showing 42.7% lower odds of developing ECC (adjusted OR = 2.43, 95% CI: 1.33–4.46, p < 0.01) compared to those who started later. Pereira et al. [15] observed that establishing brushing habits before 1 year of age was strongly associated with 37% lower ECC incidence at age 5 years (p < 0.05), with baseline prevalence rates of 40.1% for ECC and 11.3% for S-ECC. Few studies [14, 32] did not detect a significant relationship between brushing initiation and ECC, finding instead that parental education and sugar consumption were stronger predictors that masked the effects of early brushing, with Aida et al. [14] reporting that socioeconomic factors explained up to 32% of the variance in caries prevalence while brushing timing contributed only 7%.

Regarding brushing frequency, majority of studies indicated a consistent benefit of brushing ≥ 2 times/day for reducing ECC prevalence [9, 11, 13, 17, 19, 25, 26, 28, 30, 33, 39, 49, 50]. Eight studies suggested no meaningful frequency effect. Four studies [14, 16, 27, 32] specifically attributed null results to recall bias in parental reporting of brushing habits. Two studies [24, 35] cited inadequate sample sizes, while two others [28, 32] reported insufficient statistical power to detect smaller effect sizes as the primary limitations. Among the studies supporting frequency effects, Subramaniam et al. [9] found children brushing less than twice daily had 1.82 times higher ECC odds (95% CI: 0.99–3.33), Wigen et al. [6] reported 2.10 times greater caries increment (95% CI: 1.50–2.95), and Alhabdan et al. [17] observed a strong association with an odds ratio of 30.1 (95% CI: 15.00–60.00) for infrequent brushing (Table 2).

Table 2.

Characteristics of Included Studies

Study (Author, Year), Country	Study Design	Sample Size & Age	Brushing Initiation categories	Brushing Frequency categories	Key ECC Outcome & Measure	Main Findings
Subramaniam et al. (2012), India [9]	Cross-sectional	n = 1500, ages ~ 8–48 months	Yes (≥ 2 vs. ≤ 2 yrs)	Yes (≥ 2 vs. < 2/day)	ECC presence (dmft); logistic regression	Earlier brushing (< 2 yrs) correlated with lower ECC; less frequent(< 2/day) brushing correlated with higher ECC risk
Özen et al. (2016), Turkey [10]	Cross-sectional	n = 408, ages 24–71 months	Yes (< 18 vs. ≥ 18 mo)	Yes (regular vs. none)	ECC incidence; dmft	Early(< 18 mo) brushing significantly reduced ECC risk; unsupervised brushing increased caries risk, though not always significant
Retnakumari et al. (2012), India [11]	Cross-sectional (descriptive)	n = 350, ages 12–36 months	Yes (< 1 vs. > 1 yr)	Yes (once vs. twice/day)	ECC severity; cross-tabs & p-values	Brushing after 1 yr correlated with higher severity of ECC; twice‐daily brushing linked to significantly less severe caries
Wyne et al. (2003), Saudi Arabia [39]	Cross-sectional	n = 74, mean age ~ 4.58 yrs	Yes (delayed: ~ 37 mo)	Yes (≥ 2/day vs. < 2/day)	All had ECC (nursing caries); descriptive only	No control group (all ECC). Delayed brushing start (~ 37 mo) presumably contributed to higher risk; no adjusted OR reported
Dini et al. (2000) Brazil [12]	Cross-sectional	N = 245 children, ages 3–4 yrs	Yes: < 1 vs. ≥ 1 yr	Yes: occasional vs. once or more daily	ECC prevalence by dmft index	Started brushing after 1 yr + brushing < 2/day significantly correlated with higher caries
Kraljevic et al. (2017), Switzerland [13]	Observational	n = 82, ages 11–71 months	Yes (from first tooth vs. later)	Possibly frequency	ECC severity (dmft); sedation needed as outcome factor	Earlier initiation (from first erupted tooth) associated with fewer lesions; no separate final OR for frequency
Aida et al. (2008), Japan [14]	Cross-sectional (multilevel)	n = 3086, age = 3 yrs	Yes (< 18 vs. ≥ 18 mo)	Yes (< 1/day vs. ≥ 1/day)	Caries (dmft); multilevel analysis	Found no significant link between initiation age and ECC; brushing < 1/day → higher dmft
Pereira et al. (2021), Portugal [15]	Observational longitudinal	n = 146, age ~ 3 → 5 yrs	Yes (< 1 vs. > 1 yr)	Yes (twice/day vs. once)	ECC, S-ECC, caries incidence; logistic model	Establishing brushing before 1 yr strongly associated with lower ECC; brushing 2/day also protective
Al-Haj et al. (2021), Saudi Arabia [16]	Cross-sectional	n = 241, up to ~ 71 months	Yes (0–12 vs. > 12 mo)	Yes (none, once, ≥ 2)	ECC presence; found no link with initiation, but freq was sig	No significant relationship with initiation; frequent brushing was protective. Did not provide final adjusted OR for freq
Alhabdan et al. (2018), Saudi Arabia [17]	Cross-sectional with logistic	n = 578, age 6–8 yrs	Yes (> 2 vs. ≤ 2 yrs)	Yes (< 1/day vs. ≥ 2/day)	ECC prevalence; adjusted OR for initiation/freq	Delayed brushing (> 2 yrs) had higher ECC risk; brushing < 1/day → OR = 30.1. Strong outlier in freq meta-analysis
Bulut & Kilinç (2023) Turkey [18]	Cross-sectional (clinic-based)	n = 255, 0–5 yrs	Yes (when erupted, 1–2, 2–3, 3–5 yrs, or No)	Yes (≥ 2 vs. < 2/day)	ECC per WHO criteria	Delayed brushing & infrequent brushing → higher ECC
OGUNLADE et al. (2024), Nigeria [19]	Cross-sectional	n = 200, mean ~ 21.8 months (6–48 mo)	Yes < 12 vs. ≥ 12	Yes (< 2/day vs. ≥ 2/day)	No direct association data for ECC	Preliminary. Provided no effect size on ECC. Excluded from meta
Jain et al. (2015) India [20]	Cross-sectional	n = 1400 children, 0–71 mo	Yes < 12 vs. ≥ 12	Yes (< 2/day vs. ≥ 2/day)	Caries using DEFT index	Late teeth cleaning = > higher caries rates
Selen et al. (2024), Turkey [21]	Cross-sectional	n = 475, ages 0–72 months	Yes (< 1 vs. > 1 yr)	Yes (twice vs. once, etc.)	ECC presence (dmft, ECC)	Late brushing and infrequent brushing correlated with higher ECC, but no final logistic OR
Li et al. (2011), China [22]	Cross-sectional	n = 1523, ages 3–6 yrs	Possibly (initiation)	Yes (< 2 vs. ≥ 2/day)	ECC incidence; adjusted OR but freq not significant	Found freq not significantly associated with ECC in adjusted model. Provided an OR for freq but non-significant
Gao et al. (2018), Hong Kong [23]	Cross-sectional	n = 5167 examined; final ~ 945 in some analysis	Yes (> 24 vs. < 24 mo)	Yes (< 1/day, ≥ 1/day)	Caries experience among 3 yr olds; logistic regression	Late(> 24 mo) start linked to higher ECC; freq also measured. Provided an adjusted effect for initiation
Cariño et al. (2003), Philippines [24]	Cross-sectional	n = 993, ages 2–6	Yes (later start → more caries)	Yes (< 2/day vs. ≥ 2/day)	ECC prevalence, mean dmft	Later start → higher caries rates. No final logistic measure for freq
Peres et al. (2005), Brazil [25]	Nested cross-sectional	~ 359 children, age = 6	Possibly (initiation)	Yes (< 2/day vs. ≥ 2/day)	High caries (dmft ≥ 4) vs. low; logistic model	Less frequent brushing (< 2/day) → higher OR for caries. Strong association
Wagner et al. (2014), Austria [26]	Observational longitudinal	n = 471, age = 5	Yes (≥ 2 vs. < 1 yr)	No final measure for freq	Caries experience (dmft); stepwise logistic	Late brushing after 1 st yr associated with higher dmft. Adjusted OR = 1.51 in final model
Wagner et al. (2017), Germany [27]	Prospective birth cohort	n = 289, aged 5 yrs	Yes (< 12 vs. ≥ 12 mo)	Yes (< 2 vs. ≥ 2/day)	ECC (dmft ≥ 1), logistic OR	Early(< 12 mo) start strongly protective (OR = 0.06). Higher freq also protective
Skeie et al. (2006), Norway [28]	Cross-sectional	n = 735, ages 3–5	Yes (< 1 vs. ≥ 1)	Possibly freq not main	Caries (dmfs); adjusted OR	Starting < 1 yr lowered ECC, OR = 1.80 for ≥ 1 vs. < 1. Freq didn’t emerge strongly
Prakash et al. (2012) India [29]	Cross-sectional	n = 1500 preschoolers, 3–6 yrs old	Yes, 6 −12 M, 13–18 M, 19–24 M, and > 24 M)	Yes (1/day, 2/day and < 2/day)	ECC (caries severity)	≥ 2/day correlated with lower ECC
Hallett et al. (2006), Australia [30]	Cross-sectional	n = 123, mean age ~ 3.3 (1.9–4.3 yrs)	Yes (< 6 vs. 6–12 vs. > 12 mo)	Yes (≥ 2 vs. ≤ 1/day)	Caries severity (dmfs); referred children	Early brushing (< 6 mo) → lower dmfs; more freq brushing linked to lower severity but no final logistic measure
Elidrissi et al. (2016), Sudan [31]	Cross-sectional	n = 553, ages 3–5	Yes (earlier vs. older)	Yes (once, twice, thrice)	dmft, p-values; no significant association freq	Found earlier brushing = lower dmft, but not sig. No final OR. Frequencies not sig
Wigen et al. (2018), Norway [40]	Observational	n = 392 baseline (211 follow-up)	Yes (< 7 vs. ≥ 7 mo)	Yes (≥ 2 vs. < 2/day)	Caries increment from 2 → 5 yrs; logistic	Later initiation → higher caries increment. < 2/day → more caries increment. Both strongly associated
Buckeridge et al. (2021), Australia [32]	Cross-sectional	n = 155, ages 18 mo–5 yrs	Yes, but no significant	Yes, no significance	Caries experience; no final effect size	Found no link between brushing variables & ECC. No adjusted OR for meta
Liu et al. (2024), China [33]	Cross-sectional	n = 1281, age ~ 5.5 yrs	Yes (≥ 3 vs. < 3 yrs)	No (not significant)	ECC prevalence, logistic model	Later(≥ 3 yrs) brushing → higher ECC. Frequency not significant. Provided adjusted OR for initiation
Sun X et al. (2017), China [34]	Cross-sectional	n = 9722, age = 5	Yes (≥ 3 vs. ≤ 1, RR)	Yes (< 2 vs. ≥ 2/day, RR)	dmft or ECC presence; negative binomial model	Each year delay → increased caries. Lower freq → higher ECC. Provides RR
Sun HB et al. (2017), China [35]	Cross-sectional	n = 392, ages 2–71 mo (mean = 9.9 ± 7.5)	Not in final model	Not in final model	Focus on diet, parent's knowledge; excludes brushing factor	No effect measure for ECC brushing exposure
Xu et al. (2024), China [36]	Cross-sectional	n = 570, ages 3–6	Yes (≥ 3 vs. < 3 yrs)	Yes (≥ 2 vs. < 2/day)	ECC presence, logistic model	Later start → higher ECC (OR = 1.94). < 2/day → higher ECC (OR = 0.57). Both used for meta
Zhang et al. (2019) China [37]	Cross-sectional (Kindergarten-based)	n = 404, all Lisu children aged 5 yrs	Yes (≥ 24 vs. < 24 mo) but no direct final OR	Yes (≥ 1 vs. < 1 time/day), but not ≥ 2 vs. < 2	Caries by dmft + zero-inflated negative binomial model (ZINB)	Paradoxical result: kids brushing ≥ 1/day had higher dmft; large proportion untreated
Chouchene et al. (2022), Tunisia [38]	Cross-sectional	n = 381, ages 3–5 yrs	Yes (< 3 vs. ≥ 3 years)	Yes (≥ 2/day vs. < 2/day)	ECC using WHO criteria, mean dmft = 0.89	No significance correlation between tooth brushing and ECC

Meta-analysis

Early vs. Late brushing initiation

Ten studies meeting our methodological criteria were included in the brushing initiation meta-analysis. These studies [9, 14, 17, 23, 26–28, 33, 34, 44] represented approximately 23,200 children from eight countries across four continents. Seven of these studies employed cross-sectional designs, while three used longitudinal approaches. All studies provided adjusted odds or rate ratios contrasting "early" brushing (commonly < 12 months or < 2 years) with "late" brushing initiation (≥ 12 months or ≥ 2 years) (Table 3).

Table 3.

Meta-Analyses of Tooth Brushing Initiation Age

A. Early vs. Late Brushing Initiation (Meta-analysis A)
Study (Author, Year)	Comparison	Effect Type	Effect	95% CI	Log(Effect)	SE (log)
Subramaniam et al. (2012) [9]	> 2 years vs. ≤ 2 years	OR	2.43	(1.33, 4.46)	0.888	0.323
Alhabdan et al. (2018) [17]	> 2 years vs. ≤ 2 years	OR	2.20	(1.30, 3.40)	0.788	0.230
Gao et al. (2018) [23]	> 24 months vs. < 12 months	OR	2.30	(1.18, 4.50)	0.835	0.319
Wagner et al. (2014) [26]	After 1 st yr vs. within 1 st yr	OR	1.51	(1.03, 2.21)	0.415	0.195
Wagner et al. (2017) [27]	< 12 months vs. ≥ 12 months	OR	0.06	(0.03, 0.12)	–2.813	0.290
Skeie et al. (2006) [28]	≥ 1 year vs. < 1 year	OR	1.80	(1.00, 3.30)	0.588	0.305
Liu et al. (2024) [33]	≥ 3 years vs. < 3 years	OR	1.85	(1.10, 3.11)	0.615	0.264
Xu et al. (2024) [36]	≥ 3 years vs. < 3 years	OR	1.94	(1.19, 3.16)	0.664	0.253
Sun X et al. (2017) [34]	≥ 3 years vs. ≤ 1 year	RR	1.42	(1.22, 1.66)	0.352	0.078
Wigen et al. (2018) [6]	≥ 7 months vs. < 7 months	OR	2.10	(1.40, 3.10)	0.742	0.196
Pooled < sup > b </sup >	Late vs. Early	-	1.40	(0.91, 2.16)	0.336	0.221
(Sensitivity) Excluding Wagner2017 < sup > 2 </sup >	Late vs. Early	-	1.75	(1.50, 2.03)	0.558	0.077

Heterogeneity (All 10): Q(df = 9) = 95.19, p < 0.0001; I² = 90.5%

Heterogeneity (Excluding Wagner2017): Q(df = 8) = 10.09, p = 0.259; I² = 20.7%

Egger’s Test (All 10): p = 0.85 (no strong sign of funnel-plot asymmetry)

A random-effects model (DerSimonian-Laird) produced a pooled log(OR) = 0.336 (SE = 0.221, z = 1.52, p = 0.128), corresponding to an OR of 1.40 (95% CI: 0.91–2.16). Although this estimate implies a 40% higher odds of ECC among children starting to brush "late," the 95% confidence interval crossed 1.0, yielding a non-significant overall result. Heterogeneity was significantly high (Q = 95.19; df = 9; p < 0.0001; I2 = 90.5%). This substantial variability stems from differences in how studies defined "late initiation." Consequently, caution is needed when generalising these results across different populations and contexts.

A leave-one-out approach identified Wagner et al. [27] as a potential outlier, given it reported an unusually small OR (~ 0.06) for "late vs. early" brushing. Excluding this study reduced the pooled heterogeneity considerably (I2 = 20.7%) and produced a pooled OR of 1.75 (95% CI: 1.50–2.03), with the p-value changing from non-significant (p = 0.13) to highly significant (p < 0.0001). This indicates a strong, statistically significant association between delayed brushing initiation and higher ECC risk after removing the outlier study.

Figure 2a displays the forest plot for the 10‐study analysis, while Figure 2b shows the outlier-excluded model (k = 9). The latter highlights how a single study may disproportionately affect the pooled effect size and heterogeneity. The full 10‐study estimate (OR = 1.40, p = 0.13) remains the primary analysis for completeness, while the 9-study model provides important insights into the potential effect after addressing heterogeneity concerns.

Fig. 2.

a Forest plot, brushing initiation (Meta-analysis A). b Sensitivity Analysis (Excluding Wagner 2017)

High vs. Low brushing frequency

Eight studies [6, 9, 17, 22, 25, 34, 36, 38] (6 cross-sectional and 2 longitudinal designs, encompassing approximately 14,981 children from diverse regions in Asia, Europe, North America, and South America) compared “high frequency” brushing (≥ 2 times/day) with “low” or “less frequent” brushing (< 2 times/day). One study reported a rate ratio (RR), whereas the rest provided odds ratios. The random-effects model yielded a pooled log(OR) = 0.745 (SE = 0.266, z = 2.80, p = 0.0052), which corresponds to an odds ratio (OR) of about 2.11 (95% CI: 1.25–3.55). This finding implies that children brushing less than twice daily face 2.1‐fold higher odds of ECC. Figure 3 presents the forest plot for this brushing frequency meta-analysis, detailing individual study effect sizes and the overall estimate. Heterogeneity was significantly high (Q = 105.4, df = 7, p < 0.0001; I2 = 93.36%). Egger’s regression test (p = 0.1161) did not indicate strong funnel-plot asymmetry. A leave-one-out sensitivity analysis identified Alhabdan et al. [17], which reported an extreme OR (~ 30.1) for brushing < 1/day vs. ≥ 2/day—as an outlier inflating the pooled estimate. After excluding this study, the pooled OR decreased from 2.39 to approximately 2.0, with a concurrent reduction in the I2 heterogeneity value from 93.36% to 87.5%. The statistical significance of the association between brushing frequency and ECC remained (p = 0.003) (Table 4).

Fig. 3.

Funnel plot, brushing initiation (Meta-analysis A)

Table 4.

Meta-Analyses of Tooth Brushing Frequency

B. High vs. Low Brushing Frequency (Meta-analysis B)
Study (Author, Year)	Comparison	Effect Type	Effect	95% CI	Log(Effect)	SE (log)
Subramaniam et al. (2012) [9]	< 2/day vs. ≥ 2/day	OR	1.82	(0.99, 3.33)	0.601	0.317
Alhabdan et al. (2018) [17]	< 1/day vs. ≥ 2/day	OR	30.10	(15.00, 60.00)	3.401	0.343
Li et al. (2011) [22]	< 2/day vs. ≥ 2/day	OR	1.70	(1.20, 2.40)	0.531	0.176
Peres et al. (2005) [25]	< 2/day vs. ≥ 2/day	OR	2.30	(1.20, 4.70)	0.834	0.302
Xu et al. (2024) [36]	< 2/day vs. ≥ 2/day	OR	1.43	(1.04, 1.96)	0.357	0.162
Wigen et al. (2018) [6]	< 2/day vs. ≥ 2/day	OR	2.10	(1.50, 2.95)	0.742	0.181
Sun X et al. (2017) [34]	< 2/day vs. ≥ 2/day	RR	1.05	(0.90, 1.20)	0.049	0.071
Chouchene et al. (2022) [38]	< 2/day vs. ≥ 2/day	OR	0.503	(0.069, 1.684)	− 0.688	0.619
Pooled < sup > b </sup >	Low vs. High Frequency	-	2.11	(1.25, 3.55)	0.745	0.266

Heterogeneity Q(df = 7) = 105.4, p < 0.0001; I² = 93.36%

Egger’s Test p = 0.1161 (no strong evidence of funnel‐plot asymmetry)

Publication bias and sensitivity checks

Funnel plots were constructed for quantitative analysis (Figs. 4 and 5), illustrating considerable dispersion consistent with the high heterogeneity estimates. High I2 was observed and that this is expected given the diverse settings, definitions (e.g., different age cutoffs for “early” vs. “late”), and sample characteristics across the included studies.

Fig. 4.

Forest plot, brushing frequency (Meta-analysis B)

Fig. 5.

Funnel plot, brushing frequency (Meta-analysis B)

In the frequency meta-analysis, Egger's test (p = 0.1161) indicated potential funnel-plot asymmetry, likely driven by smaller studies with large effect (sample sizes < 600) with disproportionately large effect sizes. In contrast, the brushing initiation meta-analysis showed no significant asymmetry (p = 0.85), though the outlier study [27] overshadowed the overall effect. Beyond these formal tests, the leave-one-out examinations underscore how single outlier studies can drastically alter pooled estimates, emphasising the caution needed in interpreting the combined results.

Discussion

The findings of this systematic review and meta-analysis found both the timing of tooth brushing initiation and the frequency of brushing may play important roles in modulating ECC risk in preschool-aged children. Although the qualitative synthesis and pooled results generally support earlier brushing (before age 1 year) and higher frequency (≥ 2 times/day) as protective, the meta-analyses revealed considerable heterogeneity. This high heterogeneity indicates substantial real differences in effect sizes across studies that cannot be attributed to chance alone, suggesting that the relationship between brushing practices and ECC varies across different populations, settings, and methodological approaches. These findings underscore both the benefits of early, frequent brushing and the methodological complexities affecting effect size estimation.

Most included studies reported lower ECC incidence or severity when brushing began before 12 months or when brushing was maintained at least twice daily [9, 23, 36]. However, a minority found no significant link, frequently attributing null results to confounding factors such as dietary sugar intake, parental oversight, or limited statistical power [14, 32].

Our findings support the importance of both early brushing initiation and adequate frequency for ECC prevention, though with notable methodological challenges. After accounting for heterogeneity through sensitivity analyses, we found strong evidence that delayed brushing initiation significantly increases ECC risk, supporting current clinical recommendations to begin brushing with the eruption of the first tooth [9, 10, 34]. Similarly, our frequency analysis reinforces the value of twice-daily brushing as a protective measure against ECC [9, 23, 36]. These findings have important implications for paediatric healthcare providers, who should emphasise these specific brushing practices during routine child health visits [26, 28], and for public health campaigns targeting parents of infants and toddlers [14, 15].

Several factors may explain the wide variation in effect sizes across the included studies, consistent with challenges identified in previous oral health research [41, 42, 46]. First, definitional variations complicate direct pooling: "early initiation" ranged from < 6 months to < 2 years, and "high brushing frequency" could mean ≥ 2/day or ≥ 3/day in different contexts, reflecting inconsistencies in clinical recommendations [44]. Second, population differences—including diverse continents, cultures, and socioeconomic settings—shape dietary habits (especially sugar intake), fluoride exposure, and parental involvement, thereby magnifying variability as reported by Aida et al. [14] and Li et al. [22]. Methodological and reporting discrepancies arise when some studies adjust only for age or sex while others control for diet, fluoride toothpaste usage, or supervised brushing, an issue noted by Berzinski et al. [46] as common in paediatric oral health research. Outlier effects, which we identified through sensitivity analyses, remain a common challenge in oral health systematic reviews. Although Egger’s test did not indicate significant publication bias in our brushing frequency meta-analysis (p = 0.1161), smaller sample sizes can reduce the power to detect funnel-plot asymmetry. These methodological issues align with findings from other dental meta-analyses and underscore the importance of sensitivity analyses when interpreting pooled estimates [8, 51].

A notable strength of this review is its comprehensive search strategy, capturing studies from multiple databases and diverse geographical contexts, thereby covering > 30,000 preschoolers. We employed a standard quality-assessment tool (Newcastle–Ottawa Scale) and consistently extracted and converted effect sizes for two separate brushing behaviours (initiation age vs. frequency) [48]. Nonetheless, limitations remain. All included studies were observational, leaving open the possibility of residual confounding and the inability to establish causality [10]. This is a recognised constraint in dental epidemiology where randomised trials may be ethically challenging. Second, the heterogeneity—particularly when definitions of "late" or "high" brushing vary—hampers the interpretability of a single pooled estimate, a common issue in systematic reviews of observational data [52]. Third, our inclusion of only English-language publications may have excluded relevant studies published in other languages, potentially introducing language bias and limiting the generalizability of our findings to non-English speaking populations. Fourth, most studies relied on parent-reported brushing behaviours, which are subject to recall bias and social desirability bias. Finally, we cannot exclude publication bias, as studies finding no association may be less likely to be published.

To enhance future meta-analytic clarity, large-scale, well-designed studies should adopt standardised definitions—for example, defining “early” brushing as < 12 months and “high frequency” as ≥ 2 times per day. These studies should consistently measure key confounders such as dietary sugar intake, fluoride use, and the level of parental supervision. Additionally, employing prospective designs or interventional frameworks, where feasible, would help reduce recall bias and allow for stronger causal inferences. These methodological refinements would significantly strengthen the evidence base, ultimately leading to more precise clinical recommendations.

Despite these challenges, the broader evidence suggests that encouraging parents to begin brushing around or before the eruption of the first primary tooth (roughly 6–12 months) and maintaining at least two brushing sessions per day can significantly lower the risk of ECC. Public health interventions aimed at parental education, especially in socioeconomically disadvantaged populations, may be critical in achieving these behaviours consistently.

Conclusions

This systematic review and meta-analysis found that earlier tooth brushing (before age 1) and brushing at least twice daily generally reduced ECC risk in preschool children. However, significant heterogeneity in effect sizes, definitions, and potential outlier impacts underscore the need for caution when interpreting pooled estimates. Sensitivity analyses reveal that a single outlier study can transform the significance and magnitude of results, reaffirming the importance of standardised methodologies and thorough confounder adjustment. Nevertheless, the collective direction of evidence aligns with current guidelines advocating early and frequent brushing to prevent early childhood caries.

Abbreviations

ECC

Early Childhood Caries

dmft

Decayed, missing, and filled teeth

dmfs

Decayed, missing, and filled surfaces

OR

Odds Ratio

RR

Rate Ratio

CI

Confidence Interval

NOS

Newcastle-Ottawa Scale

WHO

World Health Organization

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PROSPERO

International Prospective Register of Systematic Reviews

PICO

Population, Intervention (Exposure), Comparison, and Outcomes

MeSH

Medical Subject Headings

S-ECC

Severe Early Childhood Caries

SE

Standard Error

df

Degrees of freedom

Authors’ contributions

HMA conceptualized the study, developed the search strategy with librarian consultation, performed database searches, conducted title/abstract and full-text screening, extracted data, performed quality assessments, conducted statistical analyses, interpreted results, and drafted the initial manuscript. NS provided methodological expertise, supervised the systematic review process, and critically revised the manuscript for important intellectual content. NA independently performed title/abstract and full-text screening, data extraction, and quality assessments, and reviewed the manuscript. YZ provided statistical expertise for meta-analyses, assisted with data interpretation, and critically reviewed the manuscript. TZ served as third reviewer for resolving screening and extraction discrepancies, provided methodological input, and critically revised the manuscript. All authors read and approved the final manuscript.

Funding

No funding was received for this study.

Data availability

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study. All data extracted from included studies are available within the manuscript and its supplementary files.

Declarations

Ethics approval and consent to participate

Not applicable. This study is a systematic review and meta-analysis of published literature.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.