Overall completeness and applicability of evidence

The small number of studies that have examined each intervention type limits our ability to draw firm conclusions about effectiveness, although interventions that integrate settings-based oral health education with toothbrushing programmes or professional preventive oral care appear to be most effective. When interventions of this nature are considered, issues of access, cost and appropriateness of care must be explored. However, investigators have rarely reported evidence related to approaches to intervention development and delivery or cost-effectiveness and equity of impact and adverse outcomes, leaving gaps in the evidence base. The stage of childhood when delivery of interventions is most effective for children remains unclear, although interventions that coupled oral health education with toothbrushing programmes including fluoridated toothpaste have been found effective in reducing caries in deciduous teeth, and interventions that comprise oral health education and professional dental care have been effective in reducing caries in children’s permanent teeth. Investigators have not often examined the long-term sustainability of impact, but this is clearly an important consideration, as oral disease is a chronic condition.

Our ability to draw firm conclusions from the evidence reviewed is limited because of the nature of the included studies. In particular, we highlight the following limitations of these studies.

• Less than one in five of these studies (18%) were conducted in lower-middle-income or low-income countries.

• Included studies were conducted in various regions, but most were undertaken in Asia.

• Most studies lasted less than one year.

• Close to two-thirds of studies assessed post-intervention follow-up, and the period of follow-up ranged from less than one year to four years. Reported sustainability of impacts varied across studies, and only a few studies looked at long-term sustainability.

• Most studies did not include strategies to address diversity and disadvantage, although about one-quarter of studies were implemented solely in highly disadvantaged groups. Although included studies that provided access to education and oral health services addressed disadvantage in some way, evidence related to implementation and acceptability of the intervention, long-term impact and sustainability was generally lacking, as was evidence on whether characteristics of the study population (e.g. socio-economic status (SES), ethnicity, education level) were related to the effectiveness of interventions. These aspects limit our ability to draw conclusions on the applicability and transferability of intervention strategies to other populations and contexts.

• Researchers did not widely report stakeholder engagement and involvement in intervention development and implementation, and only a few studies appear to have taken such an approach.

• Most studies targeted early (38%) and middle (41%) childhood. Only a few targeted adolescence.

• Most studies reported on at least one of the PROGRESS categories at baseline, and the categories most frequently reported were SES, education, gender, race and residence.

• Only one-third of studies reported analysis of results by any of the PROGRESS items, and items reported against varied across studies.

• Many published reports did not provide a clear description of implementation factors.

• Investigators have only rarely collected or reported economic or cost-effectiveness data.

• Study findings generally show inconsistency in intervention components tested, in intensity and duration of tested interventions and in reporting of outcomes.

The 38 included studies show substantial variation in intervention strategies and components, age groups of participants, study settings, intervention doses provided, training for implementation and personnel who delivered the interventions, as well as considerable differences in impact and outcome measures and in personnel who collected data. Various moderators such as social environment, parent and teacher involvement and use of dental health services have been identified to influence oral health behaviour change and clinical outcomes. Our review identified that more than half of the interventions tested were developed on the basis of sound theory for health promotion, population health or behaviour change. Although intervention development was not always based on theory, this represents improvement in findings from those of earlier evidence reviews and is a promising development. Despite this fact, only a few studies have investigated how these moderators may have affected intervention outcomes. Further, only a few studies that reported behaviour change outcomes also examined associations of these with clinical outcomes. Researchers must undertake these important analyses to further our understanding of the pathways involved, ranging from oral health promotion to clinical impact. Although several studies have reported on the cost of implementing interventions, we found few economic evaluations of the interventions included in this review, resulting in lack of cost-effectiveness data upon which decision makers can rely.

There does not appear to be any particular bias associated with the studies that were commercially funded when compared with the others of the same intervention type, however the number of studies is small and definitive conclusions cannot be made (see Table 11).

Quality of the evidence

We graded the evidence as having moderate to very low quality. All studies had some form of limitation regarding risk of bias. Several studies lacked thorough reporting methods and provided insufficient information to permit informed judgement about risk of bias. This review included both randomised and non-randomised controlled trials. Selection bias could present concern in non-randomised studies. A few such studies included in this review attempted to minimise the impact of potential selection bias. Whilst selection bias can be controlled in randomised controlled trials, a few such trials lacked information on randomisation methods and the process of allocation concealment. When sufficient details were lacking, we reported unclear risk for these domains. Similarly, only a few trials reported blinding of participants and personnel. A few reported blinding but failed to explain the blinding process. In studies in which only personnel were blinded, results suggest that lack of participant blinding probably had minimal consequences for outcome assessment. In only a few studies was outcome assessment reportedly conducted by examiners blinded to treatment allocation. Contamination was another important factor in this review. Many studies in this review were cluster-randomised trials, and although a few acknowledged the risk of contamination and its impact on study findings, it is often not possible to avoid contamination when community-based health promotion interventions are implemented. We further downgraded the quality of the evidence as a result of serious imprecision arising from small sample sizes, or from wide confidence intervals around estimated effects.

Potential biases in the review process

We must note certain limitations of this review. Although we conducted a comprehensive search for studies, we were able to include only studies published in some form. It is likely that important studies have been undertaken but remain unpublished in the public domain. Furthermore, studies with positive results favouring treatment were more likely to be published, and this fact could introduce bias into the results. As the result of resource constraints, we did not contact the authors of included studies to request missing data, and this could have resulted in exclusion of specific studies. To be included in this review, studies had to report on one or more of the primary outcomes while presenting baseline and post-intervention measurements, or change scores. Inclusion of studies based only on primary outcomes could have resulted in exclusion of certain types of studies.