Seal or Varnish? A randomised controlled trial to determine the relative cost and effectiveness of pit and fissure sealant and fluoride varnish in preventing dental decay

Children at risk of dental caries

Mean population values of dental caries prevalence mask an important fact. Similar to many chronic, lifestyle-associated diseases, caries prevalence is markedly linked to social and economic deprivation,5,6 and the prevalence is markedly skewed. At the age of 5 years, the prevalence of dental decay in children resident in the most deprived localities is more than twice that of children living in the least deprived communities. As oral health has improved, dental caries has become concentrated in children from poorer backgrounds.

Teeth at risk of dental caries

Individual teeth differ in their susceptibility to dental decay. This largely reflects the fact that dental plaque (oral biofilm) is more likely to form in specific areas. One such area is the occlusal (or biting) surface of molar teeth. This fissured surface predisposes teeth to oral bacteria multiplying in the depths of the fissures and forming an environment favourable to the fermentation of sugars into acid.

The first permanent molars (FPMs) erupt at the back of the mouth around 6 years of age. The occlusal surface of these teeth is particularly prone to dental caries, often within a short period of eruption into the mouth. Management of occlusal caries on permanent molars has proven to be a great challenge to the dental profession. 7 Occlusal caries accounts for the majority of affected tooth surfaces in adolescents and adults. 8–11

It is, therefore, apparent that preventing dental caries on the occlusal surfaces of FPMs is a key objective in preventative dental care. These surfaces represent the surfaces most susceptible to decay and are present in susceptible children. Actions to prevent and arrest dental caries, targeted at these surfaces, have the potential to significantly improve oral health and to address inequalities.

Preventative dental technologies

There are two preventative dental technologies that have the potential to be targeted specifically at occlusal surfaces of FPMs: pit and fissure sealant and fluoride varnish (FV). The terms ‘pit and fissure sealant’ and ‘fissure sealant’ (FS) refer to the same material. Unless referring to the term used specifically by a referenced source, the term FS is used throughout this report.

The importance of establishing the relative clinical effectiveness, cost-effectiveness and patient preferences in relation to fissure sealant or fluoride varnish

Although the relative clinical effectiveness of FS and FV is important, a number of other factors should also be considered when choosing between these technologies.

Trial aim

The overall aim of the study was to identify and compare the relative clinical effectiveness and cost-effectiveness of two established technologies, FS and FV, for the prevention of dental caries in FPM teeth in children aged 6 or 7 years.

Trial objectives

Participant selection and eligibility criteria

Children were considered eligible to join the trial if they met all of the inclusion criteria and none of the exclusion criteria. 20 It was possible to assess whether or not children met some of these criteria prior to a clinical dental examination; however, for other eligibility criteria, examination by a dentist was required.

Number of participants

It was planned to recruit 920 participants, who would be randomised equally (n = 460 per arm) to the two technologies (FS and FV) investigated in this trial. The statistical justification for this number is discussed below.

Recruitment of schools

Eligible schools involved in the Cardiff and Vale University Health Board (UHB) CDS Designed to Smile programme were invited to participate in the trial. The trial management team visited those schools that expressed an interest in participating to discuss the practical and logistical aspects of the trial. When applicable, agreement to participate in the trial was documented.

Recruitment of participants

In line with the procedures for informing parents of children of the existing Designed to Smile programme, teachers distributed invitation packs to children for them to take home to their parents.

The following documents were included in the invitation packs:

• participant information sheet (PIS) for parents (see Appendix 4)

• MHF (see Appendix 5)

• informed consent form (ICF) (see Appendix 6).

Study setting

The study interventions took place in the MDC, which visited schools in areas of high social and economic deprivation in the Cardiff and Cwm Taf Health Board areas in South Wales, UK. The MDCs were operated by the CDS, Cardiff and Vale UHB, under the Designed to Smile programme,24 which was a national oral health improvement programme.

Technologies evaluated

The two technologies evaluated in this trial (FS and FV) are well established and have been used routinely for several decades to prevent dental caries, as discussed in Chapter 1. Eligible participants were randomised to receive either FS or FV and remained on the intervention to which they were randomised for the duration of the study.

Participant eligibility

The demographic and medical history information collected from the parents was reviewed to determine if the child met the inclusion/exclusion criteria prior to a clinical examination. Children considered eligible underwent a baseline clinical examination to determine their eligibility, based on the clinical inclusion/exclusion criteria detailed in Inclusion criteria and Exclusion criteria.

Clinical dental examination

Study participants were examined in the MDC while supine under a standard overhead dental clinical light, using a plane dental mirror and a ball-ended probe. The probe was used only to remove debris and to determine surface texture. It was not used to probe for cavitation. Teeth were not dried prior to clinical dental examination. Gross debris was removed using a toothbrush.

Caries assessment using International Caries Detection and Assessment System

Caries status was assessed and recorded at baseline, for all children considered eligible, by trained and calibrated dentists using conventional diagnostic caries criteria at the d1/D1–d6/D6 level in accordance with nationally recognised diagnostic criteria. 25 Data were recorded using charts specifically designed for the collection of International Caries Detection and Assessment System (ICDAS) dental codes26 (see Appendix 11). An assessment of molar incisor hypomineralisation (MIH) was also carried out.

Further clinical examination, including ICDAS caries assessment, was performed at 12-month intervals for 36 months. Experienced community dental officers undertook the clinical dental examinations. A total of six officers were used during the study, with one examiner involved in all years of the project.

Training and calibration

An intensive training and calibration exercise was undertaken in advance of each round of annual clinical examinations. Prior to each calibration exercise, the examiners were asked to complete the online e-learning package developed by the ICDAS team. 26 At each calibration event, the examiner team received an interactive lecture to review the ICDAS caries criteria. The calibration exercise was based on projected images of 74 extracted primary teeth. The gold standard was determined by Professor Chris Deery (a trained and calibrated ICDAS examiner), who provided the images. At each calibration exercise, the images were projected under standardised conditions and compared with the gold standard examiner. The caries code (the second digit of the ICDAS code) was recoded as 0, ‘free of caries into dentine’ (caries code 0–3), and 4, ‘caries at ICDAS level 4 or higher‘, that is, ‘caries into dentine’ (caries code 4–6). Calibration was based on these categorisations at the ‘caries into dentine’ level.

As part of the annual caries assessment, approximately 5% of study participants were re-examined to determine intra-examiner reproducibility. The details of the analysis of ICDAS caries data are provided in Statistical methods.

Nomenclature for recording dental caries in the Seal or Varnish? trial

In the SoV trial dental caries was recorded using ICDAS. 26 As there is potential for confusion between the different codes and thresholds used to describe dental caries when using ICDAS and previous caries indices/scoring systems, Table 1 describes the terms used to define thresholds and levels of dental caries experience.

In this study, dental caries status is described using the following principal terms: caries free, free of caries into dentine, enamel caries and dentine caries.

Outcome measures for dental caries

The primary outcome measure was the development of dental caries on FPMs at 36 months. This was recorded as:

• the proportion of children developing new caries in dentine on any one of up to four treated FPMs at 36 months

• the number of treated FPM teeth that were free of caries into dentine at 36 months.

Outcome measures for health economic analysis

Indicators of clinical effectiveness were used alongside costs to estimate the relative cost-effectiveness of FS and FV.

Two additional outcome measures were used for the cost-effectiveness analysis. These were used to estimate incremental costs per quality-adjusted life-year (QALY) estimates.

Treatment time

The treatment times at each time point were examined and found to be non-normally distributed. A natural log-transformation of the treatment time data was performed and a linear regression analysis was used to compare arms. Treatment effects are, therefore, interpreted as percentage difference between arms. The regression model was adjusted for the randomisation balancing variables sex and baseline caries.

Sealant retention

The sealant status at each treatment visit and the retention at each clinical examination were tabulated.

Molar incisor hypomineralisation

The presence and severity of MIH in the FPMs was tabulated at each dental examination, and the prevalence was reported by treatment arm. Severe MIH presenting with PEB at baseline excluded a FPM from the study.

Secondary analysis of primary outcomes

Additional important covariates were collected via the DHQ. The variables relating to tooth brushing frequency, toothpaste, mouthwash, fluoride tablets and fluoride drops were utilised in the primary model if there were ≥ 10 percentage points of difference between the binary categories. The variables relating to tooth brushing, amount of toothpaste and FV/gel applied were utilised in the primary model if there were ≥ 20 percentage points of difference between the binary categories. The following categorisations were used for covariates added to the primary analysis:

• frequency of tooth brushing – ‘twice or more a day’ versus ‘once or less a day’

• toothpaste type – ‘family toothpaste’ versus ‘children’s toothpaste or other’

• amount of toothpaste – ‘smear, pea or never’ versus ‘cover bristles’

• additional fluoride intake – ‘any mouthwash, drops or tablets or gel’ versus ‘none’.

Sensitivity analyses specific to primary outcomes

• Adjustment of the primary analysis for all treatments within the specified time window, 4 weeks either side of the scheduled treatment due date.

• The efficacy of the number of FV treatments on the primary outcome was estimated in a way that preserves randomisation using complier average causal effect (CACE) modelling. As-treated or per-protocol analyses have been widely used as secondary analyses if the effect of treatments when actually received is of interest. However, these secondary analysis methods may yield seriously biased estimates of treatment effects. Given the questionable validity of these analyses, statistical methods such as CACE estimation have been developed to better estimate treatment efficacy, taking into account non-compliance. The efficacy of the number of treatments on the primary outcome is estimated in a way that preserves randomisation. The structural mean model (SMM) approach for CACE was implemented here by fitting an instrumental variables regression. Applications of SMMs were developed to adjust for adherence in trials with active treatments and an untreated control arm. Although SoV has two active treatments, we have assumed, for the purposes of the CACE analysis, that the sealant is a one-off standard treatment comparable with a control arm, and so treatment visits were set to a constant zero. SoV also has a binary outcome, whereas CACE is more usually applied to linear outcomes. Hence, the standard instrumental variables regression in Stata^® (version 13, StataCorp LP, College Station, TX, USA) was applied with the addition of robust standard errors. Adherence in the FV arm was defined as the number of FV treatment visits that a child attended (the instrumental variable) and the outputted estimate is interpreted as the efficacy per visit. A small, non-significant efficacy estimate indicates little effect of additional treatment visits and negligible adherence effects in the trial. Future work beyond the scope of this report will take these methods further and adjust for compliance, as well as for predictors of compliance in both arms.

• The SMM approach was implemented by fitting an instrumental variables regression. 38 Adherence was defined as the number of FV treatment visits that a child attended.

Subgroup analyses specific to primary outcome

Appropriate interaction terms were entered into the primary regression analysis in order to conduct prespecified subgroup analyses: cohort, baseline caries group. The trial was not powered to detect significant interactions and, as such, these analyses are exploratory only. ORs and 95% CIs were reported.

Imputation

Imputation was to be considered after an examination of the number of missing primary outcome data and the likely mechanism for missing data. If the primary outcome was deemed to be missing completely at random and the proportion of data missing for reasons other than moving out of the SoV area was less than 10%, then imputation would not be performed and a compete case analysis would be carried out to estimate an unbiased treatment effect. If differential dropout was observed between arms, this would indicate a missing at random or missing not at random mechanism. Mixed modelling is appropriate if the data are missing at random, whereas multiple imputation and associated sensitivity analyses may be more appropriate if the missing data were determined to be missing not at random.

Reliability

Calibration of the dental examiners was carried out each year of the study, and kappa scores were calculated for reliability. At each clinical examination, a 5% sample was re-examined and caries charts were completed a second time.

Statistical software employed

All analyses were carried out in IBM SPSS Statistics version 20 (IBM Corporation, Armonk, NY, USA), except the zero-inflated Poisson regression, zero-inflated negative binomial regression and ivregress (CACE) models, which were carried out in Stata.

Participant information sheet

As described in section 4.8 of the ICH Guideline for Good Clinical Practice,23 in order that research participants (or their legal guardians) are able to make an informed decision on whether or not to take part in a clinical trial, all relevant information must be provided to the participants in such a way that it is as non-technical and understandable as possible.

With this in mind, it was essential that all participant materials (including the PIS) were developed taking into consideration the diverse literacy levels of the target population (i.e. parents of children in Communities First areas).

To achieve this, the PIS was developed in consultation with a parents’ group in Newport. This group is in a Communities First area of South Wales but one that is not served by the Cardiff and Vale CDS; therefore, it was not within the catchment area of the trial.

After the initial draft by the SoV project team, the PIS and consent form were reviewed by the parents’ group, and the following points that they made were taken into account for subsequent revisions.

• All of the parents agreed that the overall amount of information/text presented was too extensive and would be very likely to discourage them from thoroughly reading the information sheet (if they read it at all). Furthermore, they felt that the tone of the wording needed to be made more ‘positive’. Once the parents had had time to digest the information and understand the study, they agreed that it was a ‘positive’ thing and would want their children to participant if a similar study was run in their area.

• Parents felt that having separate information sheets for parents and children was unnecessary, and that it would be easier for parents to be given one simplified information sheet, which they could use to discuss the study with their child. Parents felt that the wording used in the information sheet for children was much more accessible and that the one simplified information sheet should contain wording more similar to this.

• Parents showed a strong aversion to the word ‘research’ being used (the image of children being used as ‘guinea pigs’ was expressed by several parents). They felt that ‘evaluation’ or ‘assessment’ conveyed the same concept in a more acceptable and understandable way.

• The use of cartoons/graphics/photographs was encouraged.

• Parents felt that the duplication of wording in the information sheet and the consent form was unnecessary, and that cross-referencing should be used to help the parents understand better which items on the consent form referred to which parts of the PIS.

Despite considering the recommendations of the parents’ group, we were mindful of the guidance provided in section 4.8.10 of ICH topic E 6 (Guideline for Good Clinical Practice23), and tried to convey to those in the group that a minimum level of information was required when developing an information sheet for a clinical trial of an investigational medicinal product (CTIMP). The revised PIS was therefore intended to represent a risk-based approach to the information provided to parents for a CTIMP.

In addition to ensuring that the content of the PIS was suitable, the visual appeal of the document was further developed by a graphic designer, incorporating the child-friendly logos and colour schemes of the CDS.

Follow-up questionnaires: development and return rate

Information relating to caries risk-related habits and parental resource utilisation was obtained from parents via a single combined DHQ (see Appendix 13) at baseline and at 12, 24 and 36 months. This was accompanied (at the 12-, 24- and 36-month time points only) by the CHU-9D (see Appendix 14), both of which were distributed to participants’ homes at the required time points.

The initial return rates of these questionnaires were lower than expected. The initial follow-up strategy allowed for one follow-up telephone call 2 weeks after the questionnaires were handed out. However, unanswered calls and inactive numbers made this strategy ineffective.

Initially, a £5 shopping voucher was included with the questionnaires in an effort to boost return rates, but responses remained poor, with an overall return rate of 53%. A consultation exercise followed with parents of children at a primary school in a Communities First area not involved in the trial. Between this consultation and ideas developed by the trial management group, the following changes to the strategy for follow-up of questionnaires were made.

• The questionnaire pack was to be given to the child at treatment/examination.

• A trial-branded giveaway (pen, eraser or fridge magnet) was included.

• A text message reminder was sent if the questionnaire was not returned 1 week after distribution.

• A second copy of the questionnaire was sent to participants’ homes if the original was not received 2 weeks after distribution.

• A telephone call was made directly to parents to remind them/aid with questionnaire completion (DHQ only) and return.

• Participants were informed that all returned questionnaires gained entry to a prize draw for an iPad^® (Apple Inc., Cupertino, CA, USA) at each time point.

• The questionnaires were reformatted to increase their visual appeal, and were designed to be consistent with the PIS and ICF, using the colours and graphics of the Designed to Smile programme.

Fluoride exposure

From Table 8, it is apparent that the arms were reasonably well matched in terms of reported exposure to fluoride outside the trial interventions. Reported tooth brushing frequency was higher in the FV group, with 74.6% claiming twice-daily brushing, compared with 68.5% in the sealant group.

A greater proportion (55.3%) of parents in the FS arm than in the FV arm (48.9%) reported that their child used family toothpaste to clean their teeth. The amount of toothpaste used on toothbrushes was broadly similar between the trial arms. In both trial arms, 91% of the participants were lifetime residents of non-fluoridated South Wales.

Global oral hygiene regimen

The combined global rating scale of fluoride use and tooth brushing regimen, calculated as described in Chapter 2, is shown in Table 9. This shows that the global oral hygiene score increased in both trial arms from baseline to 12 months and was marginally higher in the FV arm than in the FS arm at 12, 24 and 36 months.

Diet

Reported dietary intake at baseline is shown in Table 10 and summarised in Table 11.

Dental attendance

On the topic of dental attendance, 92.6% of children in the FS arm and 90.9% of children in the FV arm were reported as being regular dental attenders, with the remainder described by their parents as attending only when they were in pain.

The proportion of children developing caries into dentine, having a restoration or extraction on at least one first permanent molar at 36 months

The primary outcome measure of the trial was the proportion of children developing new caries into dentine on any one of up to four treated FPMs (Table 13 and Figure 5). The proportion of children with dentine caries (D_4–6MFT) on any one of their FPMs was compared between those who had sealant and those who had FV.

FIGURE 5.

The proportion of children with dentine caries (D_4–6MFT) on any FPM in the trial at 36-month follow-up by trial arm.

The proportion of children who developed dentine caries (D_4–6MFT) on at least one FPM at 36 months was broadly similar in the FS (19.6%) and the FV (17.5%) arms.

The OR for developing caries in the FV arm (when compared with the FS arm) was 0.87 (95% CI 0.61 to 1.23) in the unadjusted model, that is, children who received FV were slightly less likely to develop caries than those who received FS, but the difference was not statistically significant (Table 14).

As sex and baseline caries prevalence were used to balance the randomisation, an adjusted model was also performed, and this was taken as the primary analysis. The OR for developing caries in the FV arm was 0.84 (95% CI 0.59 to 1.21) in the adjusted model.

In summary, the final model (see Table 14) shows that there was no significant difference in the proportion of children with dentine caries (D_4–6MFT) on any FPM in the trial at 36 months regardless of whether the children received FS or FV. Children who had more than three carious primary teeth at baseline were significantly more likely to develop caries into dentine on a FPM at 36 months. There was no difference between the proportion of boys and girls developing caries into dentine on at least one FPM.

Secondary analysis for the primary outcome: proportion of children developing dentine caries (D_4–6MFT) on any first permanent molar at 36 months

To determine the impact of potential confounding factors, a covariate analysis of the primary outcome was undertaken from two perspectives: child level (Table 19) and school level (Table 20).

The numbers for the child-level covariate analyses were substantially lower than those for the primary outcome because of questionnaire non-responses. The only covariates that were significantly associated with outcome were frequency of tooth brushing (OR 0.36, 95% CI 0.21 to 0.60) and toothpaste type (OR 0.45, 95% CI 0.27 to 0.75). Those brushing twice or more a day were less likely to develop caries on their FPMs than those brushing once or less a day. Those using children’s or other toothpaste were less likely to develop caries on their FPMs at 36 months. None of the covariates altered the main effect for arm.

Two further potential influencing factors on the primary outcome were examined: the size of the school, in terms of the number of children enrolled, and the socioeconomic level of the locality in which the school was situated, as determined by the WIMD. 36

When modelled using a two-level logistic model of child within school, neither school size nor the WIMD of the locality in which the school was located was significantly associated with the caries outcome at 36 months, and including them did not alter the main effect of intervention arm.

The proportion of first permanent molar teeth caries-free at 36 months

The analyses reported above relate to the proportion of children developing new caries on any one of up to four treated FPMs, that is, a child-level outcome. The second main objective in this clinical trial was:

• the number of treated FPM teeth that were caries free at 36 months.

In both the FS and the FV arms of the trial, 120 FPM teeth (7.5% of all teeth) developed caries into dentine, required a restoration or were extracted (Table 21). A multilevel model (Table 22), adjusted for the number of decayed primary teeth at baseline and sex, confirmed that the difference between children receiving FS and children receiving FV was not statistically significant (OR 0.97, 95% CI 0.73 to 1.28; p = 0.83). A preliminary analysis showed that there was no significant differences between quadrants (upper right, upper left, lower right and lower left), but quadrant was retained in the model as a cluster level (ICC = 0.12).

This means that there was no difference in the proportion of FPM teeth developing decay into dentine between the arms of the trial.

A further analysis examined the severity of caries attack as determined by the number of teeth becoming carious per child. The proportion of children with dentine caries (D_4–6MFT) at 36 months by the number of FPMs by intervention arm is shown in Table 23. The proportion of FPMs per child developing dentine caries differed marginally between the intervention arms.

The significance of the difference between the intervention arms was tested using ordinal regression modelling (Table 24). Those children with three or four FPMs with caries into dentine were grouped together. A marginally smaller proportion of children who received FV developed caries into dentine in their FPMs at 36 months, although the difference from the FS arm was not statistically significant (OR 0.87, 95% CI 0.64 to 1.20).

‘Threshold’ represents the response variable in the ordered logistic regression.

The threshold estimates are the cut-off points that differentiate between ordered categories of the outcome. The OR for trial arm is not statistically significant (and no different from the logistic model), indicating no difference between the arms for the ordinal outcome of number of FPMs with caries per child.

Caries experience on occlusal and non-occlusal surfaces

Although there was no difference between arms in the proportion or the total number of tooth surfaces affected by dentine caries, or in the severity of attack, it was necessary to check if there was a difference in treatment effect by type of surface. Table 29 shows the distribution of caries-affected occlusal and non-occlusal surfaces by trial arm. Although, overall, a significantly greater proportion of occlusal surfaces than smooth surfaces had developed caries into dentine at 36 months (6.4% vs. 1.1%), it is clear that the proportion of occlusal surfaces affected by dentine caries was remarkably similar in the FS and FV arms of the trial, at 6.5% and 6.3%, respectively. Minimal differences between the intervention arms in the proportion of smooth surfaces of FPMs developing caries were observed at 36 months (1.0% vs. 1.3%). A multilevel model that accounted for quadrant and surface-level clustering, adjusted to account for caries prevalence in the primary dentition at baseline, sex, surface type and surface by arm interactions, is reported in Table 30. As previously, children with three or more affected primary teeth at baseline were significantly more likely to develop dentine caries on a FPM by 36 months and, overall, there was a significant difference between the proportion of occlusal surfaces developing dentine caries and the proportion of non-occlusal surfaces developing dentine caries (OR 7.72, 95% CI 5.57 to 10.69). However, no significant difference was observed between the trial arms for the main effect or for the interaction of arm with surface type, indicating no differential effect of treatment on occlusal and non-occlusal surfaces.

Complier average causal effect analysis

In a trial such as that reported here, compliance with the intended treatment regimen is an important consideration. Children participating in the trial were scheduled to be seen by the dental hygienist every 6 months, meaning a maximum of six times (at 0, 6, 12, 18, 24 and 30 months) during the course of the trial. The number of treatment visits attended by each child who also underwent a final clinical examination at 36 months is shown in Table 31. It is apparent that fidelity in this trial was excellent, with in excess of 70% of children attending all six treatment visits and 95% of children in the FS and FV arms of the trial attending at least five treatment visits. The association of treatment visits and final outcome by trial arm is shown in Table 31.

From Table 31, it is clear that there was little association between the number of treatment visits and the development of caries in either trial arm.

To determine the effects of adherence to the scheduled number of treatment visits, a CACE analysis was carried out (Table 32). As FS was the standard treatment in the Designed to Smile programme prior to the implementation of this trial, a CACE analysis was carried out, assuming FS as the ‘control’ with treatment visits set to a constant (zero) and FV as the active treatment. In this way it can be established if the number of FV treatment visits had any significant impact on the main treatment effect.

The coefficient for the number of treatment visits for FV was non-significant, indicating that increasing the number of visits does not have an impact on the proportion of children experiencing caries on their FPMs. The efficacy per visit was –0.005, indicating only a small decrease in the proportion of children with caries on any FPM for additional FV visits. This small effect is attributable to the high proportion of children attending at least five visits out of a maximum of six.

From this, it is concluded that adherence to the treatment protocol was not an influencing factor in the results obtained at 36 months.

Adherence to the prespecified treatment window

The study protocol dictated that treatment visits should be scheduled within a 4-week window either side of the due 6-month treatment schedule. Adherence to this schedule is shown in Table 33. Overall, 71.6% of treatments were within the treatment window throughout the trial, and a further 26.1% were outside the schedule on only one occasion, a further indicator of the fidelity of the trial.

To test for the effect of adherence to the treatment window, a binary indicator was added to the primary model that categorised the participants as having received all of their treatment visits within the time window stated in the study protocol or having had at least one visit outside the treatment window. The likelihood of developing D_4–6MFT on any FPM was lower for those who received at least one treatment outside the study treatment window (OR 0.66, 0.43 to 1.00; p = 0.052). Although this effect was of borderline significance, the main effect of FS versus FV remained unchanged compared with the primary analysis (Table 34).

This demonstrates that any deviation from the scheduled treatment window did not impact on the overall results of the trial.

Assessment of costs

Costs were considered from the perspective of the NHS (i.e. direct health-care costs associated with the interventions) and from a partial societal perspective that considered the costs to the family and school (education sector).

The intervention costs were obtained from discussion with the trial team and MDC staff and from data collected as part of the trial case report forms. Research costs (such as the costs of the calibration training or costs of storing the technologies to be compliant with MHRA guidelines) were not included. With the technologies for both interventions being delivered in an established MDC, the costs of the clinic in the base case (e.g. costs of the van, running costs or equipment costs) were not included, as these are equivalent. However, to illustrate the costs associated with the delivery of the technologies within a MDC setting, we reported a separate cost with the MDC included.

The main intervention costs were (1) acquisition costs for FS and FV and (2) costs associated with the delivery of the technologies (including the clinical assessments). The costs of the FS and FV were taken from the trial protocol, that is, it was assumed that a standard dosage was applied to all trial participants and, when required, the costs of any reapplication were also calculated when they were recorded. The costs of delivery were calculated from ‘time in chair’ information contained within the case report form for each participant visit. The skill mix and grade of staff involved in the visit were obtained from the clinical staff providing the interventions. In addition, any specific materials or equipment used in the application of the FS or FV were obtained from the clinical staff. These costs were aggregated and an average intervention cost per participant in each trial arm was calculated (Table 48).

Information about the other health-care visits associated with dental-related problems for each trial participant was obtained from a parent resource utilisation questionnaire (the DHQ; see Appendix 13). This captured information on dental health-care visits (e.g. to the family dentist), other health-care contacts (e.g. general practitioner or hospital) as a result of dental health problems and any prescriptions (e.g. painkillers or antibiotics) for dental health problems. As it would have been implausible to ask the parent/carer precise information about which tooth was affected, information was collected from the parent/carer on all dental health visits and events related to his or her child. The DHQ was completed by the main parent/carer at baseline and at 12, 24 and 36 months.

Additional costs incurred by the family were also captured in the DHQ. This included any travel time and time off work/other activities associated with a health-care contact. The cost of caring for the child was also considered as a result of a dental health problem (e.g. day off work to care for the child). The main parent/carer was asked to estimate the time the child had off school as a result of a dental health problem.

Cost associated with travel time (using published government mileage rates) and time off work to attend health-care appointments with the child or the time taken to care for a child off school was calculated using a human capital approach. The value of work as well as leisure time was estimated by applying hourly rates, according to the parents’ professions, to the time expended because of dental health problems during the trial period. Hourly rates of household/domestic help were used to cost the time of unemployed parents.

To capture the costs to the school, a series of questions were given to the head teacher as part of the process evaluation. These questions asked the head teacher to estimate the time the child had away from school activities in order to receive the interventions, any time required by the teacher or other support staff (e.g. to settle a child back into the classroom) and any other costs incurred by the school to support (e.g. administration time).

Prior to commencement of the trial, the DHQ, alongside the CHU-9D, was piloted for comprehensiveness, ease of completion, recall and accuracy in a primary school in South Wales, which was not included in the trial, with a similar sociodemographic to the trial population. A protocol was developed (available from the chapter author on request), with ethics approval given by Swansea University School of Human and Health Sciences prior to the pilot evaluation. Of 30 children who were given a pack to take home, six responses were obtained (20% response rate). All parents completed a postal questionnaire, with all responses indicating that their child had no difficulties understanding or completing the CHU-9D. In addition, the PPI representative on the TSC reviewed both questionnaires to check for readability and comprehension.

When possible, unit costs for the UK were applied [e.g. Personal and Social Services Research Unit,44 British National Formulary (BNF),45 NHS reference costs46] to increase generalisability. However, to ensure that the skill mix of the MDC was fully considered in the analysis, local salary scales from NHS Wales47 were obtained, and these were fully agreed with the trial team prior to analysis. These are reported in Table 48.

A summary of the other resource items, unit costs and sources is given in Table 49.

The intervention costs and health-care utilisation costs were summated to give the total cost per participant for each arm (NHS costs), with the wider costs to the family and school reported separately (see Table 54).

Valuation of costs

All data were valued in monetary terms (£), and the price year of 2015 was applied to costs (representing the end point of the trial). As follow-up was beyond 12 months, discounting of 3.5% was applied, in line with the National Institute for Health and Care Excellence (NICE) reference case for health technologies57 for the base-case costs. All costs were summarised into the different cost categories, with means and standard deviations (SDs) produced. A total cost (means and 95% CIs) was calculated for each trial arm, with differences reported using a Levene’s test for equality t-test and with statistical significance accepted at the 5% level. When missing data existed, a simple weighted average cost based on known cost histories based on accepted methods was used. 49

To take into account the possible impact of skewness in cost data, the distribution of costs for normality were examined, as was the impact of removing extreme outliers from the analysis. When required, the sensitivity of the results was checked by considering appropriate transformation. 50

Caries avoided

The data from the main trial were used to calculate the percentage of dentine caries (D_4–6MFT) avoided at 36 months.

Quality-adjusted life-years using the Child Health Utility Index 9D

The CHU-9D is a child-specific generic preference-based measure of HRQoL. It consists of a descriptive system of nine questions and a set of preference weights based on five levels with a recall period of today/last night. The CHU-9D allows the generation of utility values for each health state to enable the calculation of QALYs. 51,52

The CHU-9D was originally intended for use in a population aged 7–11 years. Prior to commencement of the trial, the original developer (Dr Katherine Stevens) was contacted to discuss whether or not it could be used in a younger age group, and there were no major concerns with regard to the trial’s slightly younger age group. The trial team were mindful of the trial population and the potential variance in ability to self-complete the CHU-9D. The CHU-9D was piloted in a non-participating school in South Wales, with a debriefing interview completed by the parent/carer to assess any problems in its use. No issues were identified. Since the commencement of the trial, further studies have reported the use of the CHU-9D in a comparable age group to the trial population at study entry. 53,54

The CHU-9D was delivered to the child’s home address with instructions given to complete at 12, 24 and 36 months. The instruction given was that either the child should self-complete or that the parent/carer would read the questions aloud to the child who would give his or her response. Owing to the practicalities of the trial, it was deemed unfeasible to collect the CHU-9D from the child, for example in school. Thus, the administration of the CHU-9D in this context may have resulted in under-/over-estimation of QoL, as it was not possible to verify whether or not this had been reported by the child himself/herself. However, this is expected to be similar across both study arms. The CHU-9D scores were computed for each time point and the QALY was calculated to obtain the difference in QALYs between the two trial arms at 36 months. In accordance with good practice, appropriate methods were explored to take into account missing data on the CHU-9D and subsequent QALY calculation.

Quality-adjusted tooth-year

Published utilities and clinical data used to derive the tooth-year were derived in order to generate a QATY. A QATY will estimate the production of additional years of life (tooth-year) of each tooth adjusted for the quality of that tooth. An unrestored tooth has a QATY of 1 in the year it was restoration free, although a restored, crowned or root canal-treated tooth has a QATY of less than perfect (i.e. less than 1) in the year that it was restored and in subsequent years. The QATY for an extracted tooth is equal to 0 in that year and subsequent years.

Although the QATY is seen as potentially useful in providing a more dental-specific analogy to the QALY, it has had limited use within the literature because of the limitations of being compared across different interventions and the lack of robust methodological work in deriving suitable utilities in order to populate the ‘Q’ element of the QATY. Owing to the lack of resources to undertake necessary methodological work to derive precise estimates of dental-specific utilities in children, this was intended to be an exploratory analysis only. Thus, only a one-way sensitivity analysis was undertaken.

The trial data were used to provide a series of health states associated with the tooth state of the four index teeth in the trial, based on the caries status. Published utilities were assigned from the published utilities from the Fyffe and Kay55 study and from an economic analysis on the cost-effectiveness of oral health programmes56 and were agreed with the trial team (Table 50).

A combined ‘FPM utility’ from the scores using the mean of the four tooth state utilities is as follows:

Within-trial analysis

A series of incremental cost-effectiveness analyses were planned: (1) an incremental cost per caries avoided at 36 months and (2) an incremental cost per QALY (cost–utility). As part of a secondary analysis, costs and QATYs were examined. If one technology was both more effective and less costly, it would be considered to dominate the other technology and, as such, an ICER would not be reported. When there was no such dominance, an ICER would be calculated.

Univariate sensitivity analyses for all ICERs were recalculated after changing the value of a range of parameters individually to examine the uncertainty around the ICER (Table 51).

A probabilistic sensitivity analysis undertaken for the cost–utility analysis (CUA), with changes to the values of all chosen parameters (usually within the 95% CI or a reasonable, defined range), calculated the probability that the intervention was cost-effective in accordance with commonly accepted norms (£20,000–30,000 per incremental QALY gained). 57 Uncertainty associated with the individual parameters was considered.

Bootstrapping was used to construct data sets of the same trial population size by sampling at random with replacement over 1000 times. Cost-effectiveness acceptability curves (CEACs) were produced.

Budget impact analysis

A budget impact analysis to compare the costs to the NHS of FS versus FV was undertaken using the estimations of costs of the two interventions derived from the trial. The total budget impact for all patients within the trial population was calculated and examined the potential cost saving on a per-patient basis based on being treated with FS versus FV within the current intervention mix (based on the MDC). The impact to NHS budgets was estimated at 3 years. No discounting was applied to the budget impact analysis.

Model-based analysis

Deterministic sensitivity analyses were conducted to estimate the worst- and best-case scenarios of the budget impact of FS versus FV by using the lower and upper costs derived from the trial. The different scenarios that could impact on NHS budgets were (1) the cost of the intervention using a best-case/worst-case scenario; (2) the inclusion of the costs of the MDC; (3) the delivery of FV by a community dental nurse in a ‘classroom’ setting (i.e. not requiring a MDC but reflecting the potential for this technology to be delivered in a single application with minimal equipment/resources); and (4) the application of FV by a dental hygienist alone within a classroom/school medical room setting).

To assess the longer-term cost-effectiveness of FS versus FV in preventing dental caries in children, post-trial modelling was undertaken. As no suitable model could be identified from the literature search for adaptation, a de novo decision-analytic model was constructed. A UK NHS and Personal Social Services perspective was adopted in the analysis, in line with NICE methodological recommendations. 57 In the base case, costs and benefits were discounted at a rate of 3.5%.

Model structure

A Markov cohort simulation model, written in Microsoft Excel^® 2013, version 15 (Microsoft Corporation, Redmond, WA, USA), was developed, with a 3-month time cycle employed in order to reflect an appropriate time point for caries progression to occur. Half-cycle corrections were applied. The time horizon of 5 and 10 years was chosen as sufficient to reflect potential changes in the change of progression of dental caries as a result of FS and FV in children up until the onset of adulthood. Although a lifetime horizon was considered, the lack of suitable evidence to make plausible assumptions about the longer-term effects of FS and FV resulted in an amendment to the original protocol plan. This was agreed with the trial team.

For FS and FV utilisation, Markov modelling allows for the construction of an arbitrary complex chain of events that represents the natural history of sealant retention, caries formation and their associated health states. 58 A sealed tooth may lose its seal and remain sound, may be resealed or may transition to caries. Caries may be restored. A restored tooth may develop further caries and this may finally lead to the tooth being lost. Varnished teeth represented the second intervention arm. A varnished tooth may develop caries, after which the pathway is the same as that for a tooth that develops caries after losing its seal. A lost/removed tooth was the absorbing state in the model.

A model structure was developed outlining the disease pathways and validated with the trial team before the model was built in Microsoft Excel. An outline of the model structure is given in Figure 6.

FIGURE 6.

Markov state diagram: outline of modelled intervention pathways for FS vs. FV. The states are represented by ovals and possible transitions between states are represented by arrows.

Key model assumptions

The model assumed that the population was the same as the trial population in terms of baseline characteristics at entry into the model. Children would be eligible to receive FS and FV as specified in the trial. After trial completion, children would no longer be in the Designed to Smile programme (i.e. they would have entered secondary school). It was assumed that all children would be entitled to NHS dental care until the age of 16 or 17 years, which would represent the ages of children up until the 10-year time horizon. 59

We assumed that children would continue to receive their original intervention (i.e. they would not receive both interventions) and that the same products would be applied in a similar way to that in the trial. The standard of application and the efficacy of the staff administering the intervention would be commensurate with the trial, as would the accuracy in diagnosing and appropriately treating the tooth. For the base case, we assumed that children would visit a NHS community dentist and receive the intervention in accordance with standard practice, including, when required, the reapplication of FV or resealing of FS.

The trial permitted the child’s tooth to be assessed to ensure that the FPMs were sufficiently erupted to be treated, as part of the early detection of caries risk provided by the MDC. The same dental recall periods were used within the first 5 years of the model, that is, periods of 6 months, as the child would still be in the Designed to Smile programme. Between 5 and 10 years, it would be assumed that the recall period would be based on an individual risk-based assessment for each child and thus it would be problematic to provide a precise estimate of the time interval. NICE guidance60 recommended that the shortest interval for oral health reviews for all patients should be 3 months, and, for patients < 18 years old, the longest interval should be 12 months. The health economic analysis undertaken for this guideline used Markov-based modelling to assess the cost-effectiveness of different dental recall periods, emphasising that, with considerable uncertainty around model parameters (particularly clinical efficacy), conclusions about the relative cost-effectiveness of different recall times could not be drawn.

On the basis of the trial outcomes, and as agreed with the trial team, it was assumed that all children attended a dental visit every 6 months. This was assumed to be the same for both trial arms.

Within this model, no assumptions were made regarding the accuracy of clinical assessments or the impact of different clinical reversal rates and it was assumed that a similar reversal rate could be achieved as that in the trial results on caries progression. The possible impact of dental health problems (e.g. caries progression in other teeth) or other oral health interventions on the subsequent health states of the FPMs was not considered.

Base-case inputs

As this was a model developed from the SoV trial, when possible and appropriate, the clinical evidence from the trial, alongside the costs and outcomes used in the within-trial health economic analysis, were used to provide data inputs. The FS was the base case, with incremental cost-effectiveness reported for FV compared with FS.

Although the trial provided estimates of the incidence of caries at baseline and the rate of caries progression as a result of the two technologies up until 36 months’ follow-up, it was problematic to estimate precise estimates of progression beyond the trial duration.

The strongest clinical evidence would be derived from the updated Cochrane review that evaluated the relative clinical effectiveness of FS versus FV (and FS + FV vs. FV alone) for preventing dental caries in the occlusal surfaces of permanent teeth of children and adolescents. 16 From this review, only one study provided longer-term outcome data,61 which was assessed to have both an unclear risk of bias because no information was provided at 48 months, and a high risk of bias as only 33% of trial participants were included at 9 years’ follow-up. This trial was also assessed as having a high risk of bias for four key domains (outcome assessment, incomplete outcome data, selective reporting and baseline comparability). The model was therefore constrained by using evidence from the trial, and discussions with the clinical trial team were used to make assumptions, when necessary. All data inputs underwent full validation by the trial team, and uncertainty was considered within the sensitivity analysis.

Clinical data

The trial data were used to inform the probability of the different progression into caries through the model.

Costs

It was assumed that modelled patients accrued costs associated with any assessment and intervention and that they continued to receive interventions to prevent/manage caries progression in the FPMs. The costs reflected a NHS and Personal and Social Services perspective, that is, drug costs, intervention costs and any other resource use that may be required (e.g. adverse events).

The same costing approach was used from the within-trial analysis of costs, valued in monetary (£) terms at 2015 prices from unit costs from published sources, for example BNF, Personal and Social Services Research Unit and NHS reference costs and, when applicable, local financial records from Cardiff and Vale UHB. Tables 48 and 49 reported the costs utilised in the model.

Health utilities

The following estimates of QoL were used (Table 52), using the trial data to estimate utility decrements.

One-way sensitivity analysis

Table 51 presented the range of parameter estimates applied to the comparison of FS with FV in the one-way sensitivity analysis of the CUA, and these were applied to the model.

Probabilistic sensitivity analysis

A probabilistic sensitivity analysis was performed to test the robustness of the modelling conclusions in the face of uncertainty surrounding the choice of modelling inputs. Parameter values were varied within a reasonable range in each of 1000 runs and the results averaged across runs. Costs were sampled from gamma distributions, utilities from beta distributions, and rates and probabilities from log-normal or beta distributions. A CEAC was generated to depict the probability of the intervention being cost-effective at different willingness-to-pay (WTP) thresholds.

Intervention costs

The interventions costs for FS versus FV are shown in Table 53. The overall intervention costs were aggregated across all the treatment and clinical assessment time points, as per the trial protocol. For the base case, the overall mean costs were lower for FV than for FS (£32,146.50 vs. £38,098.71, respectively). These were mainly attributed to the difference in time receiving treatment (in chair) between the two treatments. The overall time in chair across the trial period was 24 minutes for FS and 16 minutes for FV; even taking into account the reapplication of FV over the trial period, this cost of treatment remained lower (£74.12 per FS patient vs. £64.16 per FV patient). When the costs of the MDC were added, this increased the mean per-patient cost in each arm, but the incremental difference was small (£95.73 vs. £86.33 for FS vs. FV, respectively).

Dental visits

The DHQ was used to estimate the resource use over the 3 years based on all available data for the two interventions. Children in the FV arm had, on average, fewer visits to the dentist (outside the MDC) than children in the FS arm (3.5 vs. 3.9, respectively). However, there were small differences in the reasons documented for these visits, with more children in the FS arm than in the FV arm attending for ‘check-ups’ (3 vs. 2.7, respectively), which equates to a difference of £28 per patient across the arms. The estimated mean cost per patient associated with additional dental visits in the FS was £426, compared with £368 in the FV (£58.17, 95% CI –£4.33 to £120.67). The results are shown in Table 54.

An important cost driver to assess differences between the two technologies would be the treatment required as a result of caries within the FPMs, for example the cost of filling or removing the index tooth. One of the limitations of the DHQ is that it would not be precise enough to estimate direct costs attributed to the FPMs. Thus, the clinical trial data were examined to give a more accurate estimation of the costs of treating caries, assigning a cost to each event based on the unit cost of treatment. These were used as the primary source of data for these key cost drivers, with the DHQ providing information of other dental health-related events.

With regard to hospital visits (e.g. hospital-based dental outpatients) and hospital admissions associated with dental health problems, there were no recorded events across both treatment groups. No other health-care attendances were recorded (e.g. accident and emergency department attendance or out-of-hours service).

Prescription costs

Parents were asked to estimate the prescriptions given by a dentist or other health-care professional given to treat dental health problems during the trial. On examining the data, there were some anomalies noted, insofar as a prescription was recorded but not the prescribing health-care professional (e.g. dentist or general practitioner). This was discussed by the trial team and it was assumed that a community dentist would have prescribed. Painkillers and antibiotics were the only medications prescribed. It was assumed this was done as part of a dental visit already counted.

The estimated mean per-patient cost for medications prescribed in the FS arm was £0.28, compared with £0.44 in the FV arm (–£0.17, 95% CI –£0.36 to £0.03). This was seen as a non-statistically significant difference (see Table 53; p = 0.091).

The base-case analysis shows that the children randomised to the FS arm accumulated more costs, overall, than children in the FV arm. Table 54 also shows the comparison of total mean health-care costs between the two technologies. These results were used in the subsequent cost-effectiveness analyses and budget impact analysis.

Costs to the family

The costs to the family were estimated from the DHQ. With regard to over-the-counter medications, the total costs were £12.54 for FS versus £13.49 for FV. In terms of carer time as a result of taking time off to attend dental health appointment with the participating child or time to look after the child, the mean per-carer cost was estimated to be £29 versus £25 for FS versus FV, respectively. Table 54 shows the costs to the family regarding the costs resulting from time off.

Cost of hosting mobile dental clinic to the school

The schools questionnaire was returned by 30% of the participating schools. In their responses, all schools stated that there were negligible costs associated with the school’s participation related to the MDC, with little disruption to classroom time for the teacher or administrative tasks. Thus, the costs to the school of hosting the service were not formally analysed.

Overall NHS costs

Table 54 shows the health-care contacts across the trial arms from the NHS perspective. Non-statistically significant differences were seen in the contacts, that is, £425 versus £367 (FS vs. FV, respectively), with a difference of £58.17 (95% CI –£4.33 to £120.67; p = 0.068).

Total combined NHS and family costs

Table 54 shows the health-care contacts across the trial arms from the perspective of the NHS and a partial societal (family) perspective. Statistically significant differences were seen in the ‘total cost of time off work’ contacts, that is, £28.79 versus £24.97 (FS vs. FV, respectively), a difference of £3.83 (95% CI £0.55 to £7.10; p = 0.022).

The costs of the two technologies showed small but statistically significant differences; the mean cost to the NHS (including intervention costs) per child was £500 for FS, compared with £432 for FV, a difference of £68.13 (95% CI £5.63 to £130.63; p = 0.033) in favour of FV.

When costs to the family were included (with NHS costs), the costs were £529 for FS and £457 for FV, with a difference of £71.96 (95% CI £7.21 to £136.71; p = 0.029) in favour of FV.

Dentine caries avoided at 36 months

The data presented in Table 13 were used to estimate the cost of each 1% reduction in the proportion of children who at 36 months had developed dental cares.

Health utilities and quality-adjusted life-years

The mean utility values derived from the CHU-9D at each of the follow-up points were similar (p-values greater than 0.05) for both technologies (Table 55). In the FS arm, the mean utility value (and resultant QALY) was 0.928 at 12 months and this increased to 0.933 at 36 months. In the FV arm, the mean utility value (and resultant QALY) was 0.931 at 12 months and this increased to 0.933 at 36 months.

The effects of these very small changes in the CHU-9D (both within and across the two arms) are reflected in the negligible utilities lost/gained and subsequent QALYs from 12 to 36 months. Overall, there was a small and non-statistically significant difference in QALYs at 36 months in the FV arm (0.931) compared with the FS arm (0.929) (difference 0.002, 95% CI –0.010 to 0.005). However, the very small difference and the 95% CIs indicate that these data should be treated with some caution (Table 56). In effect, there is a non-statistically significant difference and negligible clinically significant difference in QALYs, which was a similar result to the clinical outcomes reported in the trial.

Tooth utilities and quality-adjusted tooth-years

Table 57 reports the tooth utilities associated with the two technologies and QATY differences. The mean tooth utility at 36 months was 0.932 in the FS arm and 0.925 in the FV arm. The overall QATY difference was 0.007 in favour of the FS arm, a small and non-significant numerical difference. The clinical significance of this result is negligible and thus should be regarded with caution.

Incremental cost per child with dentine caries avoided

The base case of the analysis for the incremental cost per child with dentine caries avoided is presented in Table 58.

With very small numerical differences in dental caries between the two trial arms, the ICER computed that FV was dominant over FS. However, appropriate caution should be exercised in using and interpreting the term ‘dominant’ in the context of these results.

Sensitivity analysis

As shown in Table 58, the results remained consistent when adjusted around the upper and lower range (95% CI) for costs and outcome. The CEAC for the incremental cost per child with dentine caries avoided is shown in Figure 7. Although there is no formal threshold, this illustrates the probability of FV being cost-effective at different monetary WTP thresholds.

FIGURE 7.

The CEAC for incremental cost per child with dentine caries avoided.

Incremental cost per quality-adjusted life-year

Table 59 shows the incremental cost per QALY.

Figure 8 presents the CEAC.

FIGURE 8.

The CEAC (CUA).

For the cost per QALY, the probability of FV being cost-effective is 70% at a NICE societal WTP threshold of £20,000 per additional QALY.

Incremental cost per quality-adjusted tooth-year

The base case of the analysis of the incremental cost per QATY is presented in Table 60. It can be seen that, in comparison with FV, FS generated slightly more QATYs. However, when the ICER calculation is performed, FV was dominant.

This is a result consistent with the previous ICERs, and the same caution is applied in the interpretation of these results. This remained consistent when a one-way sensitivity analysis was performed. As this was an exploratory analysis, no CEAC is presented.

Sensitivity analysis of the budget impact analysis

Deterministic sensitivity analyses were conducted to estimate the worst-case and best-case scenarios of the budget impact of FS versus FV by using the lower and upper costs derived from the trial. To examine different scenarios that could impact on NHS budgets, three additional scenarios were explored: (1) the budget impact when the costs of the MDC are included, (2) the budget impact of a community dental nurse delivering FV in a ‘classroom’ setting (i.e. not requiring a MDC but reflecting potential for this technology to be delivered in a single application with minimal equipment/resources), and (3) the budget impact when a dental hygienist alone applies FV within a classroom/school medical room setting.

Table 62 shows the different cost scenarios for the intervention over a 3-year period as seen in the trial. It can be seen that the deterministic sensitivity analysis is unchanged in most of the modelled scenarios, that is, FV is dominant and is seen to be the optimal strategy. However, when the best-case cost scenario of £51.88 for FS is compared with the worst-case cost scenario of £83.41 for FV, FS technology dominates. That FS could cost as little as this in a real-life setting is questionable, and so caution should be exercised with regard to this. Further work would be required to explore the plausibility of this scenario.

Table 63 shows the different cost scenarios for the intervention to be administered if the costs of the MDC are included. It can be seen that the deterministic sensitivity analysis is unchanged in most of the modelled scenarios, that is, FV is dominant and is seen to be the optimal strategy. However, when the best-case cost scenario of £67.01 for FS is compared with the worst-case cost scenario of £112.23 for FV, FS dominates. The plausibility of this scenario reflecting a real-life cost needs further discussion.

Table 64 represents the analysis of one FV application by a dental nurse in the school. It can be seen that, by reducing the costs of the FV technology, it becomes dominant and is seen to be the optimal strategy across all the scenarios. The plausibility of this scenario reflecting a real-life cost would need further investigation.

Finally, Table 65 represents the analysis of one FV application by a dental hygienist in the school. Again, it can be seen that, by reducing the costs of the FV technology, it becomes dominant and is seen to be the optimal strategy across all of the scenarios. Again, the plausibility of this scenario reflecting a real-life cost would need further investigation.

Additional scenario analysis

To provide further illustration of the practical implications of the findings, an additional scenario analysis was undertaken to provide a hypothetical illustration of the time savings that could be achieved in delivering FS versus FV. As shown in the results, one of the resources attributed to costs is the ‘time in chair’ and another is the skill mix of staff that would be needed to deliver the technologies and how these findings can be used to optimise capacity and the number of children treated within the service. The ideal service configuration of preventative dental health for a population of children at high risk of caries is informed by translating the time savings gained from using FV with a dental care team that, compared with FS, needs to be less highly trained and does not include a qualified dentist. Table 66 presents an illustrative scenario (using time to treat each child gathered from the trial) in which, compared with FS, using FV allows 93 more children to be treated within the same time (assuming a day’s session, based on the school working day of 9 a.m. to 3 p.m., at a school that has this number of children).

Post-trial modelling

Table 67 reports the estimations of the incremental cost per QALY at 5 and 10 years. These base-case results estimate that FV continues to be dominant over FS because of lower costs and slightly more QALY gains.

The results of the univariate sensitivity analysis estimated that FV remained dominant.

Figures 9 and 10 show the CEACs generated for 5 and 10 years, respectively. They show that, at a NICE societal WTP threshold of < £20,000 per QALY gained, FV is 99% (at 5 years) and 96% (at 10 years) likely to be cost-effective.

FIGURE 9.

The CEACs for a 5-year time horizon.

FIGURE 10.

The CEACs for a 10-year time horizon.

Data collection

Interviews with children

Interviews with participating parents

The interviews with parents were conducted by telephone 4–6 weeks after their child’s first and last treatments, in years 1 and 3 of the study. Parents were interviewed after their children had been interviewed.

The interview schedule was piloted with parents at a local community centre in South Wales in an area that was not part of the trial. The following comments were received.

• The term ‘interview’ was perceived as intimidating. Parents preferred the terms ‘courtesy call’ or ‘chat’. However, because there was a need to be clear to study participants that data were going to be collected and analysed, the term ‘interview’ was retained, but simpler language was used when parents were contacted. For example, parents were told that the purpose of the interview was to find out what they thought about SoV.

• Similarly, parents thought that ‘recorded’ sounded less ominous than ‘tape-recorded’, because ‘recorded’ sounded similar to the phrase ‘your call may be recorded for training purposes’, which people were used to hearing.

• Parents also suggested clarifying that interviews would take place by telephone; they said that this would encourage more parents to consent, as it was perceived as easier to participate in a telephone interview than in a face-to-face interview.

• Parents also made recommendations about recruiting other parents, suggesting that schools could be more involved in this. They suggested using posters with pictures to remind parents about the study and to increase trust because recruitment would be through the school. They also suggested enabling parents to collect and return forms directly to the school rather than by a postbox, especially to class teachers, as they are particularly trusted. Using plain English on posters and advertising, and clarifying that assistance from staff was available for parents with low literacy/English language skills, were also suggested as ways to increase accessibility.

  Schools were not asked to assist further with recruitment as there were concerns about the research burden. Nevertheless, some schools did assist with recruiting. For example, one school inserted a reminder about study recruitment into the school newsletter, and one head teacher spoke to parents as a group about the trial, as he thought that the parents would not understand the paperwork.

Forty-nine parents were interviewed in year 1 of the study and 30 parents were interviewed in year 3. The aim of the interviews in year 3 was to follow up all interviewees from year 1; attempts were made to follow up as many children and parents from year 1 as possible, even if the child and parent could not be interviewed. The most common reason for not following up a child or parent was that the parent had changed their telephone number or did not answer the telephone despite several calls (the initial telephone call was partly to reconfirm consent, so if the parent was not contactable then their child was not interviewed). Another problem was the poor coverage of mobile phone reception in some areas of the South Wales valleys (many parents did not have landlines). Furthermore, telephone calls from the university were routed through the university switchboard; this resulted in the recipients of calls not being able to identify the caller number, which probably resulted in some parents not answering. A mobile phone was not a viable alternative because it could not be used in conjunction with a voice recorder for the interviews, and the reception/quality of the line tended to be poor. Three parents were followed up after the school was contacted for an up-to-date telephone number (permission to do this had been obtained in the study consent forms). Secondary reasons for not following up were that a few children were absent on the day of the interview at the school or had been withdrawn from the trial. As a result of the absent children and the various problems recontacting parents, follow-up in year 3 of the study with some parents and children was not possible. The intention was to interview the same pairs of parents and children in years 1 and 3; this was achieved for 17 child–parent pairs. To increase the overall number of interviews, additional schools were sampled and children and parents from these schools were also interviewed.

The interviews with parents were relatively short, lasting approximately 10–15 minutes, owing to parents’ brief answers and often busy schedules; on some occasions, during the interviews, young children could be heard in the background. The interview topics included the acceptability of the treatment and the treatment setting, social influences on attitudes towards treatment and treatment settings (such as prior experience of dental treatment), changes in the child’s oral hygiene behaviour, the perceptions of treatment impact and the acceptability and impact of the trial processes. See Appendix 18 for the parent interview schedule.

Questionnaire and interviews with school staff

In year 1, a questionnaire was sent by e-mail to all 65 schools participating in the study. A second e-mail was sent to non-responding schools and, if required, further contact was made by sending postal questionnaires to any schools that had still not responded. The questionnaire included both closed and open questions about schools’ experiences of the implementation of SoV (both the treatment programme and the trial) and acceptability of the intervention. Twenty-four questionnaires (37%) were returned.

In the third year of the trial, at the end of the school year, the 24 schools that had returned a questionnaire in year 1 were contacted by telephone to request their participation in an interview; those that agreed were sent consent forms to complete. Of the 24 schools contacted, five agreed to take part in an interview, and four interviews were actually conducted (one was cancelled because of sickness). Three interviewees were head teachers and one was a class teacher. The interviewees were asked about the impact of the trial and treatment delivery in the school, contextual factors that affected the feasibility and acceptability of delivering preventative treatment in a school setting, the acceptability of treatments and the recruitment of hard-to-reach parents for interventions. The interview schedule is in Appendix 19. The original intention had been to purposively sample schools based on their questionnaire responses; however, as only 24 questionnaires were returned in year 1, and as the sample of teachers or head teachers willing or able to spare time for an interview was likely to be only a fraction of this figure, all 24 schools were contacted so that the number of staff interviewed would not be too small.

Interviews with community dental unit staff

The dental team delivering the treatment was interviewed annually; this took place at the end of the school year. These interviews were conducted face to face at the CDS offices in Cardiff. Nine staff were interviewed in the first year and 12 staff were interviewed in both the second and the third years. This group included hygienists, dental nurses and the staff members who collected children from the classrooms. The staff makeup of the groups was largely the same each time, but there were occasional omissions or additions because of staff leaving or joining, or because a staff member was unavailable (e.g. through sickness) on the days of the interviews. The interviews explored the feasibility and acceptability of delivery in a school setting, perceptions of acceptability of the treatment for children, child co-operation and perceptions of parent recruitment issues. The interview schedule is in Appendix 20.

Questionnaire and interviews with non-participating parents

For the first and second cohorts, the interviews with parents who did not participate in the trial were conducted after the trial recruitment periods had ended; this was to avoid any confusion or perceptions of harassment about recruitment to the trial itself rather than to a research interview. Two recruitment methods were used; the intention was to recruit up to 20 parents by each method. In the first method, if parents had not responded to the initial SoV recruitment attempt, they were sent information about the trial a second time. This information included a consent form for a telephone interview to discuss their reasons for non-participation, and a one-page questionnaire about their reasons, for those parents who did not wish to participate in the SoV study but were willing to take part in research about non-participation. The parents were invited to return the questionnaire, the consent form or both if they did not wish to participate in the trial. Those who returned the consent form were followed up with a telephone interview.

The second recruitment method was speaking to parents who had not returned any paperwork for the SoV trial itself and had also not returned the questionnaire or consent form for the interview about non-participation. This was done to collect data from parents who may have been willing to take part in the research but were unlikely to respond via a written invitation. Direct contact, including face to face, is likely to be effective for recruitment for interviews for hard-to-reach groups that have difficulty with literacy or paperwork and has the added benefit that a researcher visiting schools in person can familiarise people with a study and reduce anxiety about particular elements. It was also expected that teacher and school staff support for an intervention would have a positive influence on recruitment rates. 64 Seven schools were visited to recruit this group of parents and these were sampled by identifying schools that had lower than average recruitment rates for SoV, although very small schools were excluded as there were too few potential interviewees. The initial plan was to recruit parents from a parents’ evening at a sample of schools, at which teachers and school staff would introduce the researcher to groups of parents and request participation in a research interview, but the research team was advised by several schools that it would be difficult because of how busy these events are and that parents would have limited time for interviews. Therefore, a second approach was devised.

In this second approach, a researcher waited with the year 2 class teacher by the exit door at the end of the school day, a time when a direct handover of the child to the parent or carer is required. The researcher and teacher/teaching assistant handed out information sheets directly to parents. This helped to associate the researcher with the school and with the class teacher, which it was hoped would build trust, and meant that the researcher was available for parents to ask any questions about the research. The researcher did not approach parents, except to give them the information sheet, to avoid parents feeling harassed to take part. This recruitment was conducted after the first child dental examination to establish trial eligibility had been carried out, so there was no possibility that this approach could be identified with trial recruitment, which had finished by this point. The information sheet made it clear that the trial recruitment had finished and that the current recruitment was for a research interview only, to find out why some parents decided not to take part in dental studies and programmes to inform future interventions. Information sheets were given only to parents who had not returned any paperwork to the trial. Parents were told that the researcher would be in the school (accessible via the school reception desk) one day during the following week, when parents could drop in to do an interview if they wanted. The interviews were set up on a ‘drop-in’ basis to make it easy and flexible for parents, rather than arranging an appointment time, and to further avoid parents feeling harassed to take part in the research. Year 2 class teachers were asked to remind parents about the interviews the day before. The information sheet also included the researcher’s telephone number in case parents had any queries about the interviews. This recruitment was carried out in years 1 and 2 of the study (for cohorts 1 and 2), and for cohort 2 an incentive of a £10 shopping voucher was offered for a completed interview.

Parents who withdrew their children from the trial were also sent a letter inviting them for a telephone interview about their reasons for deciding to no longer take part. The target for the study was to recruit up to 20 withdrawing parents and to compare data by trial arm.

The results of these recruitment efforts and the data collected on reasons for non-participation are discussed in detail in Findings.

Data analysis

Signs of distress

Immediately after treatment in the MDC, the dental hygienist and the dental nurse recorded signs of adverse behaviour during FS or FV placement. Observations during the baseline examination are reported in Tables 69 and 70. The equivalent data at the 30-month treatment visit are shown in Tables 71 and 72.

Overall, the proportion of children displaying adverse behaviours or showing signs of distress during treatment at baseline, as observed by the dental hygienist, was small: 97 of the 468 children in the sealant treatment arm and 40 out of 464 children in the FV arm of the study. Children in the FS arm were more likely to gag and to cry. There was good agreement between the observations made by the dental hygienist and the accompanying dental nurse, although the nurses were more likely to observe and report excess movement of arms and legs during treatment.

To facilitate analysis, the adverse outcomes over the course of the trial as observed by the dental hygienist and the dental nurse were dichotomised (any adverse outcome vs. no adverse outcomes recorded) for analysis (Tables 73 and 74). The behaviours observed differed across the intervention arms of the study. At baseline, 21% of participants displayed some form of adverse behaviour, but this fell steadily at subsequent treatment visits up to the final treatment visit at 30 months, at which only 2.6% of the participants were observed to have some form of adverse behaviour during treatment. In contrast, the proportion of children showing signs of adverse behaviour during the application of FV remained more constant in the course of the trial, ranging from 8.7% at baseline to 3.8% and 4.1% at the 24- and 30-month visits, respectively. During the trial, increased familiarity with the treatment process and increased ability to cope as the children got older are probable influences on these findings.

The difference between trial arms at baseline as observed by the dental hygienists at the baseline examination was subject to two-level and single-level logistic regression (Tables 75 and 76).

Model 1 is a two-level logistic model of child within school, whereas model 2 is a single child-level model for the dichotomised adverse outcome scale. It can be seen that the addition of school as a level does not alter the main conclusion that a significantly higher proportion of children in the FS arm experienced adverse outcomes at baseline. Both models adjust for the balancing variables from the randomisation (sex and baseline caries).

A similar analysis was of the behaviours observed by the treating dental hygienist conducted at the 30-month treatment visit (Table 77). Slightly more adverse events were recorded in the FV arm at 30 months, although the number of events was low in both intervention arms.

Again, model 1 is a two-level logistic model of child within school, whereas model 2 is a single child-level model for the dichotomised adverse outcome scale. It can be seen that the addition of school as a level did not alter the main conclusion that the proportion of children experiencing adverse outcomes at 30 months was slightly higher in the FV arm than in the FS arm, but the difference was not significant (Table 78). Both models adjust for the balancing variables from the randomisation (sex and baseline caries).

Participants’ perception of treatment acceptability

Treatment acceptability was assessed from the child’s perspective through a DTF scale (see Appendix 12), completed by all children in the MDC immediately following the initial application of FS/FV and at each follow-up visit. The scores recorded at the baseline treatment visit are recorded in Table 79. It is apparent that time and taste were the dominant factors influencing acceptability. To facilitate analysis of the trends observed over the course of the trial, the data were dichotomised to most happy/happy versus neither happy nor sad/sad/most sad.

The proportion of children who opted for the neither happy nor sad face, the sad face or the very sad face at baseline to describe their feelings on completion of treatment was much higher in the FS group than in the FV group (49.2% vs. 27.1%). This proportion reduced during the course of the study so that by the 30-month examination, only 16.8% of the FS-treated group reported that they were neutral or sad immediately following treatment. This contrasts with the findings in the FV arm, in which there was a gradual increase in the proportion of children opting for the neutral or sad options on the DTF scale between baseline and 30 months (Table 80).

Two-level and single-level logistic regression models were constructed to determine the difference between intervention arms at baseline and at 30 months (Tables 81–84).

Model 1 is a two-level logistic model of child within school, whereas model 2 is a single child-level model for the DTF scale. It can be seen that the addition of school as a level does not alter the main conclusion that the proportion of children reporting that they were happy or very happy at baseline was significantly higher in the FV arm than in the FS arm, and that the proportion of children reporting that they were happy or very happy at 30 months was significantly higher in the FS arm than in the FV arm. Both models adjust for the balancing variables from the randomisation (sex and baseline caries).

From the above, it is obvious that the acceptability of the technologies under test as judged by the children immediately post treatment was different at different time points in the trial. There are two main issues:

1. At baseline, the children in the FV arm were significantly happier than those in the FS arm, this situation being reversed at the final treatment at 30 months.

2. Those receiving FS became happier as the trial progressed; however, in contrast, in the FV arm there was a modest decrease in the number of children who chose a happy face over the course of the trial.

These findings are in line with what might be expected from a clinical perspective. As the FSs were applied at the beginning of the study as soon as the FPMs erupted, the treatment was more invasive and intense and, as described in Chapter 3, took longer. The children receiving FV, however, had to experience treatment of approximately the same intensity throughout the study.

Parental reports of treatment acceptability

Parent accounts of child acceptability corroborated child reports. Parents said that children found it novel and exciting to visit the MDC (known colloquially as ‘the van’) and that they liked various aspects of treatment such as being with their friends and being given stickers:

Has he ever said anything about the taste or about how long it takes or anything like that?

He said it’s banana taste.

Does he like that or does he not like that?

He likes that.

He likes the taste, OK. And does he say anything else about the van or going with his friends or anything like that?

He just says he likes going on the van and having a sticker.

And he likes the sticker, do you know what he likes about going on the van?

I think it’s the chair, because they let him play with the chair.

Parent interview, FS arm, year 3

Many parents said that their children did not say much about the treatment but would have complained if they had found the treatment very objectionable. A small number of children had reported to their parents that they did not like the taste or another aspect of treatment (e.g. that it was ‘gluey’) but, again, acceptability was often based on more than one factor:

She said that it tasted horrible.

OK.

That was it really just it didn’t taste very nice but she was delighted because she had stickers [laughs] you know I was like ‘What was it like, what did they do?’ and she said ‘They put something in my mouth and it didn’t taste very nice’.

Parent interview, FS arm, year 1

In year 1, children either did not say anything about the treatment to their parents or said that it tasted funny or that they did not like the taste. In year 3, children still either said little or reported that they did not like the treatment, but children at this stage also mentioned to parents aspects of the treatment that they liked. There were no differences between trial arms or by deprivation level.

Parents in the high-deprivation schools in the first year of the study were most likely to report discoloration of teeth from the treatment, in both trial arms. In the final year of the study, a very small number (n = 3) of parents in the FV arm reported discolouration. If parents did notice discoloration, they were not always sure that this was because of the treatment; other explanations were that it might be because of their child’s age or inadequate brushing. No parent was concerned about discoloration, and perceptions did not differ by trial arm.

A few parents thought that their child’s teeth were whiter or shinier; some were not sure about changes but expressed a hope that the child’s teeth were stronger or more protected because of the treatment received.

School staff reports of treatment acceptability

Staff reported that the MDC and CDS service was well liked in schools by children; one head teacher commented that usually children were disappointed if they were not called to attend. One school staff member commented that if children did not like an aspect of treatment they forgot about it soon after returning to the classroom.

Community Dental Service staff reports of treatment acceptability

Throughout the study, CDS staff had mixed opinions about which treatment was more acceptable to children. They observed varied reactions to the treatment and different aspects of treatment that were more or less popular:

The children react differently to be honest. Some of them love the varnish or some of the hate the varnish, some of them play up because their friends are there or whatever, they have a different reaction and then some of them are good with the sealants as well.

CDS staff interview, year 2

Community Dental Service staff commented that children were generally tolerant and well behaved, regardless of whether they liked the treatment.

In year 1, more staff thought that FV was better liked by children because it was quicker and did not involve cotton wool rolls, and because some children did not like the taste of the FS fixer. In year 2, the staff commented in more detail about the pros and cons of both treatments: FV was quicker and easier to apply, but it was also sticky, could cause nausea and had a lingering taste, whereas children could rinse after FS treatment; if children gagged easily it meant that applying FS could be difficult, and children sometimes found it difficult to keep their mouth open for a long time, but FS usually required only check-ups rather than reapplication. They also mentioned that FV was easier than FS for a difficult child as it was the quicker treatment. Staff used various tactics to encourage children, such as bringing nervous children back later in the day to retry treatment, treating a nervous child first (so that the child did not have to wait) or last (so that the child could observe other children receiving treatment), and sitting a child further upright to reduce gagging.

Staff in years 2 and 3 also said that children were very accepting of treatment and that refusals were rare, but they did note that acceptability still varied and that some children liked the treatment or taste and some did not. By this point, some children remembered not liking the treatment and could react negatively when attending again, which could then affect the other children they were with. One staff member thought that children from more deprived areas reacted more negatively when coming back for treatment. However, some FS children were disappointed that they did not receive any treatment when other children did.

Various CDS staff comments mirrored what children reported. One staff member commented that children receiving FS disliked the cotton wool rolls more than they disliked the length of treatment. A number of staff members reported that whether or not children found the treatment acceptable was largely driven by taste, and also that some children thought the treatment was too quick and would have liked more time out of the classroom. Staff also commented that children liked different aspects of treatment, particularly laughing at each other wearing the protective glasses; preferring to be with their friends and being less nervous if they were accompanied; and enjoying the novelty and excitement of being on the MDC:

Oh they love coming on the van. They love it. They always ask who drives it and where do people, some, some think that they sleep there overnight and it turns into a caravan but they love it, they love coming on. What’s this drawer for, you know, yeah, they love coming on.

CDS staff interview, year 2

The CDS staff members’ comments that children were chattier and asked more questions in year 3 of the study also corroborated some children’s reports that they felt more confident now that they were older.

Global intervention acceptability for children

Turning to the information analysed from the interviews with the participating children, the primary finding on the overall acceptability of FS/FV treatment for children was that there were many factors that contributed to whether or not children found the treatment acceptable. Most of these factors were related not to the FS/FV aspects of the MDC visit but rather to more general aspects of treatment or to the context in which the treatment was delivered.

Most children commented that ‘going to the van’ was ‘fun’, ‘brilliant’ and ‘good’ and that it was exciting to go there (the ‘van’ is the term children used to describe the MDC). Receiving stickers was a very popular aspect of MDC visit, and three children mentioned that they liked the CDS staff. When comparing a visit to the MDC with a visit to a family/community dentist, most children also expressed their views in relation to factors beyond the actual treatment received, such as being given stickers, the speed of the visit, which visit was warmer and the van being fun. Some children also mentioned that they liked having their teeth looked after or protected:

I like the dentist because they look after your teeth and keep them clean.

Child interview, FV arm, year 1

Fourteen children in year 1 of the study mentioned feeling scared or nervous about the treatment; of these, 12 said that being with friends (children normally attended the MDC in groups of three) made them less nervous or feel safer. In year 3 of the study, fewer children mentioned feeling nervous, and in the low-deprivation schools four children commented that they felt more confident about being at the dentist now that they were older and more used to the treatment visits:

We did both say we were nervous when we went, when we were.

Yeah, when we were younger, ‘cos it was our first time.

Yeah ‘cos it was less.

We were like . . . not confident.

Yeah.

‘Cos we were just too young to go on our own, not like on our own like.

You were quite young when you started though.

Yeah, I was talking about before.

Because you’re like quite scared, because you haven’t been to the dentist oftenly[sic].

Then you might need . . .

So when you get older and older and . . .

You’ll be more confident.

Yeah.

Paired child interview, one FS arm and one FV arm, year 3

Although the older children generally felt more relaxed and confident about a visit to the MDC, they were also more aware of what was going on around them. In year 3 of the study, a small number of children reported that the examination made them feel anxious because they did not understand the tooth-numbering terminology the CDS hygienist used and were worried that there might be something wrong with their teeth:

They say stuff like numbers and stuff and we don’t really know what that means.

You don’t know what that means, OK.

So like, they could be saying that our teeth are really bad or really good but we don’t know.

Like A2 and A3 and that.

Yeah.

Yeah.

AC.

Sounds funny doesn’t it?

Yeah.

Yeah.

So would you like to know more, you’d like to understand what they’re saying?

Yeah.

Paired child interview, one FS arm and one FV arm, year 3

Children also frequently mentioned generic aspects of the dental treatment they liked, particularly the moving dental chair and protective (and often coloured) glasses. Feelings about the three-in-one syringe were mixed, as some children found it unpleasant but some liked it because it was ‘ticklish’. The cotton wool rolls were the least popular aspect of treatment and a few children expressed a dislike of having their teeth ‘scraped’/’poked’. There were no differences in preferences between the FS and FV trial arms. Children’s responses to what they thought about the overall treatment often centred on these wider factors rather than on the FS/FV itself:

So can you tell me, what did you think about going to the dentist earlier this week?

It was good, I liked it.

It was good . . .

It was fun.

And you thought it was fun [to child 2], OK. And what did you like about it? [Pause.] Or are you not sure?

Because you got to sit on the special chair.

The special chair, a magic chair?

Yeah.

What was special about it? Why was it a good chair?

Because it can move.

Is that fun?

Yeah, and I like the glasses.

What do you like about the glasses?

Because they protect our eyes and face.

They protect your eyes. Are they a special colour?

White.

White?

Yeah.

OK. Is there anything else that you liked about going?

Yeah, I was excited.

You were excited . . .

And because you got to go in the van.

Oh right, what do you like about the van?

Because um [pause] can’t remember.

You’re not sure, that’s OK. But it’s fun going in the van?

Yeah.

Paired child interview, FV arm, year 1

Taste was the main aspect of the actual treatment acceptability mentioned by children, but it was one of several factors:

If somebody said you’ve got to go and have that same treatment done again tomorrow, how would you feel about that?

I wouldn’t mind.

You wouldn’t mind [to child 1] and how about you [to child 2]?

It would be OK.

It would be OK. So it wouldn’t be fantastic but it would be OK?

It would be fantastic.

Why would it be fantastic?

Because I love the banana gel.

Is that the best thing about going?

No.

What’s the best thing about going?

Getting the sticker.

Paired child interview, FV arm, year 3

A small number of children were so enthusiastic about the treatment visits that they responded (unprompted) that they would be happy to have the treatment done all day/every day:

If the van came back tomorrow and you had to have it done all over again, how would you feel about that?

It don’t matter, I’ll have it every day.

OK. Why would you have it every day?

Because the chair is fun.

Is that your favourite thing?

Yeah.

Paired child interview, FV arm, year 1

Children in both trial arms mentioned similar acceptability factors in year 1 and year 3 of the study, such as liking being with their friends, treatment being good for their teeth, and liking the stickers and the moving chair, and mixed attitudes towards different aspects of the treatment, such as the three-in-one syringe:

And what do you think about going there for the treatment?

I think it’s really good idea.

Mm hm.

And what’s good about it?

It’s like, we can get to wear some glasses, we can get a sticker, stuff like that.

Mm hm.

We don’t get bad teeth when we’re older.

Paired child interview, FV arm, year 3

In year 3 of the study, children reported that some people in their class liked treatment and some did not. However, most children in year 3 were more likely to report that they were ‘fine’ or happy/excited to go for treatment, even if they did not like all aspects of it.

Acceptability of treatment setting for children

To elicit information about the acceptability of the FS/FV treatment, children were asked if they preferred being in class or leaving class to visit the dentist at the MDC. In year 1 of the study, most children did not express a clear preference or were indifferent. Children who did express a preference gave various reasons: they thought that it was important to get dental care, they did not like doing schoolwork/they liked missing class, or they liked going to the MDC. One child said that they preferred being in class because they liked their teacher. There were no differences between the FS and FV trial arms.

In the final year of the study, children were more aware of having to catch up on schoolwork if they missed part of a class to visit the MDC, although this was not a major problem, as children mentioned receiving help from their friends or from the teacher to do this. The main reason that children preferred staying in the classroom was avoiding having to catch up on work later. Two children were not sure about their preference; three children said that, if the lesson was mathematics, they would prefer to go to the dentist.

Acceptability of treatment setting for parents

Several parents thought that their children behaved better, made less fuss or were less nervous about seeing a dental clinician in a school environment:

He will do it in school.

So he’s a bit better in school?

Yeah, as the teachers are there.

The teachers are around?

Yeah yeah.

And you’re not?

That’s right he’s a good boy in school see [laughs] it’s an awful thing to say but it’s true.

Right, so is it just mainly because of the teachers and because it’s a school environment that he behaves better?

He’s never naughty in school.

Parent interview, year 1

Parents did not mind their children missing some class time to attend the MDC. Some pointed out that, as the children were young, missing a small amount of time was not significant. A few also pointed out that the children missed less class time going to the MDC than they would if they went to a community dentist. Parents also appreciated not having to escort children to the appointments and commented that this was very convenient for them. A small number commented that the intervention was good for children who would not otherwise see a dental professional.

Acceptability of treatment setting for school staff

In response to the questionnaire in year 1, the majority of schools indicated that hosting SoV in their school had been straightforward, the programme ran smoothly with minimal disruption to the school, the information received was easy to understand and SoV was similar to existing dental programmes running in schools. CDS communication and co-ordination was good and both CDS and schools were flexible (e.g. in response to each other’s timetables). Six also said that it was interesting to be part of a study and two said that they were pleased to offer it to their children. One head teacher mentioned not knowing details of the study, which would have been helpful in explaining the intervention in response to parent queries. One support teacher had had problems with not being aware that the vans were coming and this causing disruption. One class teacher said that some children did not like the taste of the treatment and that staff had to comfort them when the vans arrived the next time. The only other problems were taking time to explain the study to parents (four respondents mentioned this), explaining the study to teachers in the school (two respondents) and finding the time to respond to requests for data (four respondents). Some schools distributed letters for the study, although some also actively reminded parents to return them, by telephoning or texting parents, reminding them in person and by displaying a poster (in one case). Eleven schools agreed that SoV fitted in with other health programmes or initiatives they were involved in; the most common were tooth-brushing programme and the Healthy Schools scheme. Some schools were unsure whether or not SoV had changed dental behaviour, several thought it made children more aware of the importance of looking after their teeth, although another respondent commented that they did various dental programmes so it was hard to know the effect of SoV on behaviour.

Responses in telephone interviews with school staff (mainly school heads) in year 3 were similar to the questionnaire findings. Two school staff mentioned minor disruption to classes when children left to visit the MDC and a small amount of extra paperwork but said that this was not a problem. Overall, visits to the MDC were not perceived as a problem because they were quick and efficient, and schools were very used to their visits:

Very, very minimal impact really. I mean they come in they do their stuff and off they go again.

School staff interview, year 3

The preventative programme was also perceived as meaning that children would miss less school than they would visiting a community dentist for dental problems later on:

Well I’m hoping, from the information that was provided when I took over the class room is that it was going to protect the children’s teeth, that there was going to be less cavities, less absences with them disappearing for full days when they have teeth removed and teeth repaired because they’re going to be better protected.

School staff interview, year 3

One school head also commented that the service was very child-friendly, and, because of this, the children were not stressed and their visit to the MDC did not disrupt the school day too much.

School staff were positive about having the CDS service in schools and regarded it as fitting in with their promoting health agenda and also helping them engage with families, which was seen as part of their remit:

What do you think about having the dental clinic on the school grounds?

Oh absolutely fine.

Mm hm.

Absolutely fine. I think it was nice for the parents to see.

Mm hm. Why is that?

Well, because I think a lot of them find dentists, doctors, even crossing the threshold of the school quite intimidating.

Mm hm.

It’s made it more accessible for the community.

School head interview, year 3

School staff thought that SoV, along with other dental programmes, had improved children’s awareness of dental health and oral hygiene.

Feasibility and acceptability of treatment setting for Community Dental Unit staff

Community Dental Service staff found the treatment setting unproblematic on the whole, largely because SoV was integrated with the Designed to Smile dental programme for children in South Wales, which already used MDCs to deliver dental treatment to school children and typically ran very smoothly. CDS staff reported that being part of the SoV trial had created extra paperwork and that this took some extra time, especially early in the trial when staff were still getting used to it but that they were generally positive about being part of the trial and reported no significant impact on how treatment was delivered compared with other CDS services. A small number of staff mentioned that they had little time with each child, which could be a disadvantage as chatting to a child could reduce anxiety sometimes.

Minor difficulties noted included parking a MDC in tight spaces or driving one in bad weather, and occasionally school administrators (the main liaison persons for the CDS in schools) complaining about paperwork or not being as helpful as they might be. One SoV staff member commented that because of the various dental programmes running in schools, each with their own paperwork, schools could get confused between them sometimes. However, most school administrators were helpful; for example, one would assist SoV with recruitment if numbers were low and another would text reminders to parents. Overall, CDS staff found schools very helpful and found that they valued CDSs. CDS staff reciprocated, being careful to organise themselves around the school timetable and events to minimise disruption. Two CDS staff members thought that schools in the more deprived areas were particularly welcoming to the MDC.

Community Dental Service staff also mentioned several aspects of the school setting that made treating children easier, including children being more likely to turn up for appointments:

In the schools we’ve got a captive audience.

Mm.

In the clinic we have horrendous, horrendous DNA [did not attend] rates.

CDS staff member, year 3 of study

The CDS staff were also able to treat larger numbers of children more quickly in a school setting.

Reasons for participation

Most parents were not fully aware of the details of the treatment or study but were participating to generally benefit their children’s teeth and to feel that they had done as much as possible to promote their child’s dental health:

I’m thinking myself there’s every chance now that I’ve given her the best chance to look after her teeth then and she may not have decay.

Parent interview, year 3

Most made the decision about participation themselves, although some also consulted their spouse or the child as well. A small number checked with their regular dentist first. Some parents were able to recall that it was a protective treatment that was put on teeth, although not all remembered that it was treating only molars or that there were two different treatments being used. Some thought that SoV included dental education. Randomisation to trial arm was not always understood: one parent thought that their child might receive no treatment, and another thought that their child was ‘lucky’ to be picked for treatment. Few knew how long the study lasted, and many asked about this during research interviews. A few parents had very little understanding or had completely forgotten the details of the study:

Do you understand what the study is about?

Not really no.

OK, is there anything, if you had to tell a friend about what Seal or Varnish? was what would you say?

Well to be honest I don’t really know a lot about it.

Parent interview, year 1

However, parents trusted the school and several said that they participated partly because of this:

It was with the school I was quite happy because it was recognised through the school.

Parent interview, year 1

I trust the school with everything.

Parent interview, year 1

Several parents also gave accounts of family members’ dental histories and experience of FS/FV, either their own or that of the participating child or another of their children, both positive and negative, and gave this as a reason for participation. Some parents liked that the programme would mean that their children would have dental check-ups more regularly (even if they were already having 6-monthly check-ups at a community dentist) because problems would get picked up more quickly.

A small number of parents (both participating and non-participating) also mentioned the value of doing research and were also willing to participate for this reason (although it was not a primary reason).

Reasons for non-participation

Nineteen parents returned a consent form (including in the second trial recruitment package sent home with children) to be interviewed about non-participation. When contacted by telephone, four said that they were taking part in the trial or had signed up, so no interview was attempted. Five did not answer the telephone or the number was wrong, two had a level of English that was too limited for an interview, two said that they were not participating because they were moving house, one was too busy to do an interview because of family illness and one said that they did not want to do an interview and also commented that they did not participate because they already had a dentist in town. On subsequent cross-checking of the 19 consent forms with trial records, it was found that only 8 of the 19 ‘non-consenter’ parents had not given consent. Six had given consent but were ineligible and five were actually in the trial. It is likely that a majority of parents had completed all the forms their child had brought home from school without checking the content of the form thoroughly and did not realise that the consent forms for non-participation interviews were different from the trial forms and completed and returned all the forms they received. CDS staff reported that this was common in parents receiving paperwork about other dental programmes. Hence, the contacts made with parents who returned these forms did not yield any data on reasons for non-participation, except that parents had not always understood the paperwork they received or completed it correctly.

Twelve parents returned the questionnaire about reasons for non-participation (parents could give multiple responses, and so the number of responses is > 12).

The responses indicate positive and negative reasons for non-participation. Some parents did not participate because they just wanted to continue with existing dental care, some were unsure about SoV’s benefit or acceptability to the child, and some were unwilling to take part because it was also a research study (Table 85).

The questionnaire also asked about dental care and tooth-brushing habits, to determine whether children with poor dental habits were more or less likely to take part in this type of intervention. According to their parents, three children brushed their teeth at least twice a day, four children brushed their teeth once a day and one child did so less frequently than once a day. All parents said that their child visited a dental surgery regularly.

Face-to-face recruitment at the school handover at the end of the day, through distributing an invitation to participate in a research interview, did not work well for the hardest to reach non-participating parents (those who had never responded to any of the study recruitment materials they had received). This recruitment effort was a second measure to try to collect data on why some parents did not participate in the SoV intervention or study. During recruitment for non-participants at handover time at the end of the school day, no parents queried the researcher, although one or two made more general queries about dental services available in relation to specific problems that their child was having. These parents were given the telephone number for the CDS for information about appropriate dental services.

Five parents across three schools gave consent to be interviewed. One decided not to participate in SoV because they were moving house but would have liked to otherwise and had attended the interview primarily to find out about their own dental treatment. One was participating in the trial and two thought that they had returned consent forms. One of these parents was familiar with FS because their child’s older sibling had had this treatment, and they thought that they had consented and so did not know why they were not taking part. Three did not remember receiving any information about SoV but were generally positive about school dental programmes. Two commented that they thought that their child had probably lost the letter sent out from the school and that this had happened with school letters before: ‘sometimes she makes paper aeroplanes out of it’. None expressed any negative feelings about taking part in a trial. Therefore, this set of interviews yielded no data on reasons for non-participation, as parents were generally positive about the dental programme, except that, again, that there was confusion about or difficulties with completing the paperwork. This recruitment effort was eventually abandoned, as these five interviews yielded few useful data and had required 14 school visits (two visits each to seven schools) to achieve. The addition of a shopping voucher as an incentive in year 2 increased the participation rate only very marginally.

The overall finding from the recruitment efforts and results from speaking to these ‘non-participating’ parents was that there was a lack of understanding of and engagement with the paperwork in various respects. In the initial piloting of the materials, the paperwork such as the information sheet and consent form was found to be very long, even though it had been shortened as much as possible and a graphic design company had produced a very friendly format. Despite this, none of this group of parents could articulate what the study was about in full or mentioned the comparative aspect of SoV, indicating that the paperwork had not communicated the study to parents clearly or that the information had not been retained. One had understood that it was to ‘seal their teeth’ but thought that it was to address existing holes in children’s teeth and one thought that the SoV programme also handed out toothbrushes. Another thought that it addressed tooth grinding and that it was a student research project. No negative attitudes towards the SoV intervention or trial were expressed.

In interviews, both school and CDS staff discussed similar problems with contacting and recruiting parents for programmes, and commented that they did not always get paperwork back from parents whatever the programme. School and CDS staff thought that parents with little English would also have problems, especially with long forms; of the five non-participating parents recruited for interview through school visits, one had no English (a friend attended the interview to translate) and another spoke English as a second language. CDS staff also mentioned that some parents had problems completing paperwork because of literacy problems. One parent commented they would have preferred to receive information about SoV face to face rather than in a ‘big booklet’. School staff thought that some parents needed face-to-face support in order to understand the study and to complete recruitment forms correctly, and so would sometimes assist parents with filling in paperwork. One school staff member commented that it would have been helpful for teachers to know more about SoV so that they could have answered parents’ queries about it. Some of the difficulty was also a result of the volume of paperwork and participation in other dental programmes running in the school, leading parents to think that they were already taking part. CDS staff reported that schools often found it a challenge themselves to keep track of the paperwork from different programmes operating in the school, including several dental programmes. School and CDS staff also commented that parents who were already engaged with health and dental services were another group that was less likely to participate because they might not see a need beyond their existing dental visits. One parent commented to the researcher in the playground that their child already saw a dentist and they did not want to use up additional resources unnecessarily. Other families who tended not to engage, according to school staff, were those who were nervous of school environments or who had chaotic lives or work/other commitments that were difficult to juggle. Some school staff thought that parents needed reminders in person from school staff to sign up for programmes. School staff sometimes made an effort to recruit those families, in person, that they thought might be particularly in need of services, either to explain letters or to discuss any potential barriers.

What is the effect of the trial on recruitment or acceptability to parents, dental staff and schools?

The reasons offered by parents for consenting their children to the trial included the perceived benefits. However, it is clear that, on occasion, parents were not entirely clear about or could not fully recall the purpose of the trial, again resorting to trust in the school and familiarity with the CDS.

The main trial findings and conclusions are discussed in Chapter 6.

Clinical results

This trial found that the proportion of children who developed dentine caries or had a restoration or tooth extracted because of caries (D_4–6MFT) on any FPM after 36 months did not differ significantly between those treated with FS and those treated with FV (OR 0.84, 95% CI 0.59 to 1.21; p = 0.351). Similarly, when the proportion of teeth, the number of teeth or the number of tooth surfaces was studied as the outcome measure, no significant difference was observed between treatments. However, across all of these measures, marginally fewer teeth were affected in the FV-treated children.

This study was commissioned because the relative effectiveness of FS and FV was unknown. Cochrane systematic reviews published just before17 this study started and immediately before it finished16 came to the conclusion that, when comparing resin-based FS and FV application, there is some evidence to suggest that FS is more effective. However, these studies reported that it is not possible to draw definitive conclusions based on the four previous studies to have compared resin-based FS and FV. The authors attributed this to scarce and clinically diverse data, and to the low quality of previously reported studies.

Other clinical findings

Beyond the primary and main clinical outcomes, some additional clinical parameters were explored.

Strengths and limitations of the health economic analysis

This health economic analysis has been undertaken as part of a rigorous randomised trial to assess the clinical effectiveness and cost-effectiveness of FS and FV delivered within a MDC. Data were collected prospectively alongside the trial with the economic evaluation, which was an integral component of the trial design from the outset. It also benefits from reporting a budget impact analysis. Any changes to funding arrangements in health care over the next 5 years and the impact of any decision on local and national NHS dental budgets can now be informed by our findings. However, some limitations are acknowledged, particularly in relation to undertaking health economic evaluations in this context, and should be fully considered.

Consideration of the longer-term impact of fissure sealant versus fluoride varnish in preventative dental health programmes

Exploration of the longer-term costs and outcomes of the two technologies was an important consideration in developing the health economic analysis, given that this was done as a preventative intervention to improve the longer-term dental health of children who are at high risk of caries. A Markov model was built and validated with the trial clinicians as commensurate with the expected caries pathway as a result of having preventative treatment with FS or FV. The results show that FV remains dominant over FS, with this result remaining reasonably robust in the sensitivity analysis. However, caution must be exercised when interpreting these results.

The model-based analysis beyond the trial period was significantly hampered by the lack of evidence of the longer-term costs and outcomes in order to populate the model. The Cochrane review16 highlighted the lack of data available to compare FS and FV applications; only one study provided any data beyond our trial period,61 but this was limited by several biases, particularly the fact that only a limited number of children were followed up in the longer term. Only informed assumptions could be made about the longer-term costs and outcomes of FS versus FV based on clinical opinions from the trial team, informed by the trial results. Although it could be argued that, in the absence of good-quality evidence, the feasibility of modelling outside the trial period is questionable, this has been done to ascertain the framework for a model-based economic analysis and, importantly, the evidence that would be needed to populate the model. In the absence of data from previous trials, undertaking further follow-up of the trial cohort could provide an excellent opportunity to formally address the question of the longer-term clinical effectiveness and cost-effectiveness of FS versus FV.

Choosing and using appropriate health outcomes to inform the health economic analysis

An important issue that has arisen from this study is the need to give thorough attention to the outcomes of interest for the health economic analysis. A plethora of literature has examined oral health outcomes in the adult population but, in comparison, the evidence base for children is sparse. The strength of our evaluation is that we choose three outcomes that we expected to provide a comprehensive picture of costs and outcomes of the two technologies to support decision-making, both from a UK NHS decision-making perspective and in the clinical dental community.

Providing an estimation of the incremental cost per QALY is seen as the cornerstone of demonstrating cost-effectiveness within a UK NHS context. For example, the NICE reference case for health technology assessment57 highlights the inclusion of a QALY as its health outcome of choice. When the trial was originally conceptualised, a pragmatic decision had to be made on which measure to use, given the paucity of measures available for children and, in particular, the lack of evidence in the utility of generic HRQoL measures. We decided, after discussion with the developer and following recommendations made by the Health Technology Assessment funding panel, to use the CHU-9D, even though it had only just been published and its utility in a dental health context was still unknown. CHU-9D score was not measured at baseline; however, as this was a randomised trial, it is reasonable to assume that any differences between the two groups at that point were due to chance. The inability to consider utility curves over outcome time points only, rather than including baseline, could be seen to result in a reduction in power, but, given the preventative nature of this study (i.e. the study was not missing assessing utility during an episode of illness) and the fact that there were assessments of utility on multiple occasions, this is unlikely to be a major issue.

A study published since the start of the trial suggests that CHU-9D can be used in children aged 6 or 7 years. 53 That work, which considered the psychometric performance of the CHU-9D versus the EuroQol-5 Dimensions Youth version, suggested that children at this age can understand and complete the measure. This supported our own informal and limited evaluation of the CHU-9D that was undertaken prior to the start of the trial. The CHU-9D was assessed in the context of a child dental community clinic in New Zealand54 and was found to have the potential to capture HRQoL across different caries experience, showing that those with no apparent caries had a higher mean score than those with caries (0.88 vs. 0.87, SD 0.10), although the difference was small. Interestingly, this cross-sectional sample had lower HRQoL scores than our trial sample, although it is difficult to assess whether or not the patient populations are comparable. In terms of QALYs generated, this study reported that the difference in QALYs between those with and those without caries was 0.12. This is a fairly moderate difference, but it could suggest the potential of the CHU-9D to identify small but important changes in QALYs as a result of caries. Because of the lack of clinical differences in caries between our trial arms, we were unable to detect differences in utilities and QALYs. Subsequently, Foster Page et al. 69 undertook a randomised controlled trial to further assess the CHU-9D in terms of its responsiveness to changing components of the decayed, missing, filled tooth surfaces in primary dentition (dmfs) and DMFS scores in children aged 6–10 years over a 1-year period. They found that the mean CHU-9D improved from 0.88 to 0.9 from baseline to follow-up, with no significant difference found between caries status and CHU-9D. Again, CHU-9D scores from their study are lower than those in the present trial and lower than those in other studies, such as the 0.86 reported in the Canaway and Frew study,53 indicating that, as in this study, the impact of children receiving dental care under trial conditions cannot be ruled out. Similar to the later Foster Page et al. study,69 this trial suggests that the CHU-9D is insufficiently responsive to changes in caries experience.

An important question is ‘what could be done to improve measures aimed at capturing HRQoL in relation to this patient population?’. Whether the focus should be on generic HRQoL measures or on condition-specific measures is the subject of much debate, and it is important also to consider the context of this trial within a preventative/public health context. For example, the NICE public health guidance70 recognises the possible challenges of undertaking a CUA, and other approaches such as a cost–consequence analysis could be considered, although CUA is advocated (along with the need for HRQoL measures that can derive appropriate utilities) to ensure comparability across other parts of NICE.

The authors suggest that patient-reported outcome measures (including HRQoL) for use in a dental public health context should have a stronger conceptual and methodological basis. Understanding children’s preferences for treatment, the properties of measures and the association that these have with (1) dental-related health and (2) HRQoL warrants careful attention. Moreover, understanding and capturing the utility/disutility associated with caries experience over time will assist in ensuring longer-term examination of the cost-effectiveness of different strategies.

The cost-per-QATY analysis undertaken in this trial allows the consideration of a dental health state utility, which has been suggested as a more meaningful measure in the context of dental health. Although the QATY has been reported in studies,31,71 there has been little ‘methodological movement’ since the original work. 55,72 In this trial, published utilities were used alongside clinical opinion from the trial team to attribute a utility to (1) each individual tooth and (2) a composite QATY drawing on other studies31 and preliminary work undertaken as part of a NICE public health guideline. 56 The present analysis suggested that QATY gain was greater with FS than with FV, suggesting that FS could potentially be cost-effective. The results should be viewed with appropriate caution, given the approach taken to derive the number of QATYs. QATY gain appears to be a potentially useful dental health outcome, and further work is recommended to improve the underpinning work to derive suitable utilities for calculation of QATYs.

The importance of collaboration between health economists and the dental health research and clinical community will be instrumental to improving the quality of economic evaluations within preventative dental health.

Number of treatments received

In the course of the trial, participants were scheduled to receive treatment on six occasions. Clearly, the number of treatments received, whether application of FV at each visit or check-up or, if necessary, ‘top-up’ of the FS, is important. More than 70% of children attended all six treatment visits, and 95% of children in the FS and FV arms of the trial attended at least five treatment visits, which is excellent adherence to the scheduled treatment protocol. The CACE analysis demonstrated that such a high level of compliance means that adherence to the treatment protocol was not an influencing factor in the results obtained at 36 months. To obtain such a level of compliance reflects the relationship between the CDS team, the schools and their parents.

The CACE analysis was performed as a sensitivity analysis in order to explore the effect of attendance for treatment on treatment efficacy. A CACE analysis is preferable to a per protocol analysis, as it respects the trial randomisation, but is usually more suited to trials with two active arms rather than trials with one active trial arm and one control arm. Hence in SoV, the FS arm was selected as the ‘control’ treatment for the CACE analysis, as ideally it is a one-off treatment whereas FV needs to be applied regularly for full effectiveness. It was also the current standard treatment for children in these schools. The results indicated that, in fact, there was very little variation in treatment adherence, with the majority of children attending either five or all six FV treatments. As such, the CACE analysis indicated only slight gains in treatment efficacy with additional visits. In a wider context, this means that, for FV to be effective in the community as a caries prevention tool, treatment adherence needs to be ensured.

Treatment window

The six treatment appointments were scheduled at 6-month intervals over the course of the trial. Owing to the practicalities of scheduling visits of the MDC to schools to accommodate holidays and acknowledging that there were days when it was not convenient for the schools to host the MDC and so forth, it was agreed at the outset that a treatment window of 4 weeks either side of the treatment due date would not compromise the integrity of treatment delivery and would reflect the practical nature of a pragmatic clinical trial such as SoV. Over the course of the trial, adherence to this schedule was good. Overall, in 71.6% of schools, all treatment visits were within the treatment window, and in a further 26.1% of schools, only one of the six treatment visits was outside the schedule. This further indicates the fidelity of the trial.

Fissure sealant retention

The level of FS retention in this study was very high. This is because the study participants were seen at 6-monthly intervals and FS was ‘topped up’ if deficient. Furthermore, in the case of the maxillary FPMs, the teeth were sealed if part of the mesial–occlusal pit and fissure was erupted and the distal aspect was then sealed at a subsequent 6-month visit. At the 30-month treatment visit, only a small proportion of previously sealed teeth were reported as being ‘lost’, with 74% of upper teeth and 88–91% of lower teeth reported as ‘partial’, that is, sealant did not cover the whole of the occlusal surface.

Crossover

One deviation from protocol was observed. At the 12-month treatment visit, one child who had been randomised to the FV arm received FS. This was logged and the child continued in the trial in the arm to which they were randomised.

Examiner reproducibility

Throughout the trial, efforts were made to ensure that the dentists undertaking the clinical examinations were appropriately trained and calibrated. A professor of paediatric dentistry who was involved in the development of the ICDAS index trained the examiners at baseline and a second professor of paediatric dentistry undertook the annual training and calibration exercise before each round of examination began. The kappa scores achieved indicated that all examiners demonstrated a high level of intra-examiner reproducibility, and that inter-examiner reproducibility was also high.

Study considerations

From the issues previously discussed, the authors are of the view that this was a very robust and well-conducted study. The trial protocol was rigidly adhered to and compliance was excellent, as were follow-up and completion rates. Fidelity was as high as could possibly be expected. The trial team are content that there was nothing that they would have done differently in answering the question proposed at outset. Some aspects of the trial are considered further here.

Study numbers

The caries risk status of the participants is an important consideration. The SoV trial was set in areas of social and economic deprivation and more than half of the participants had decay into dentine in their primary dentition at the start of the study. The finding that approaching one in five of the participants experienced dental caries in at least one FPM at 36 months, across trial arms, makes clear that the study population was appropriate and typical of those whom the technologies tested are designed to benefit. The consistent finding from the regression analysis of caries experience in the primary dentition at baseline being a significant predictor of caries in the permanent dentition is as would be expected. 73 This of course did not affect the conclusions about differences in the relative caries-protective effect of FS and FV.

It is clear that the randomisation and stratification procedure employed was very successful and that the trial arms were well balanced at baseline.

The trial was more than adequately powered. This arose for the following reason. Calculations based on a previous study in South Wales and estimates of treatment effectiveness suggested that 313 participants were required per arm. The original intention was to recruit all participants in the same school year. However, this effectively meant recruiting all children in the period September–December in the first school term in order that the 6-monthly treatment visits could be accommodated. By December just over 500 children had been recruited, too few to satisfy the power calculations. With the funder’s support, recruitment was extended to a second school year. This resulted in two study cohorts, each of which was followed for 36 months. There was no observed cohort effect on the clinical outcome.

In planning the study, allowance was made for an 8% withdrawal per annum. In the event, many fewer children were lost from the study, with the most common reason for failure to complete the study being moving away from the area or to a different school that was not participating in the trial. When children moved to a participating school they continued in the trial. Recruitment to the trial in the second cohort proceeded at a much quicker pace than in the first year and so the study over-recruited by around 100 participants. The view was taken that, having invited the parents of the children to participate in the study, it would have been problematic to refuse the opportunity to participate to already consented children.

This, together with the much lower than expected dropout rate, meant that in both trial arms the number of children completing was in excess of the numbers required to satisfy the power calculation.

Questionnaire return rate

Information to inform oral health-related habits and the health economics utility scores was collected via a self-complete questionnaire. The efforts taken to maximise the return rate are described in Chapter 2. The revised follow-up strategy was found to boost return of the follow-up questionnaires, with the overall return rate of the DHQ rising to 60% for the whole trial.

Clearly, this limited the number of participants who were available for the secondary analyses, which were dependent on information on covariates gained via the questionnaire. However, there was no evidence of a difference in return rates between treatment arms. It should be remembered that this trial was conducted in Community First areas, which are areas designated by the Welsh Government as being of high social and economic deprivation. A response rate of 60% is probably as good as can be expected in this setting.