Best-practice prevention alone or with conventional or biological caries management for 3- to 7-year-olds: the FiCTION three-arm RCT

Dental caries (decay) experiences of children (globally and in the UK)

Dental caries remains an important public health problem globally,8 recognised by the World Health Organization as a disease that affects 60–90% of school children and the vast majority of adults, contributing to extensive loss of natural teeth in older people. 9,10

The social gradient in health inequity means that the poorest have the worst health,11 and there is strong evidence relating oral health (especially dental caries) and socioeconomic status; children in lower socioeconomic groups are disproportionately affected by the disease, which has a linear relationship with poverty. 12 Disease levels vary starkly in the UK, with 56% of 5-year-old children in Blackburn and Darwen but only 4% of children in less deprived South Gloucester showing visible carious lesions. 13 In England, dental caries remains the primary reason for children’s admissions to hospital. 1 In 2013/14, a total of 62,747 children and young people were admitted to hospital in England, Scotland and Wales with a diagnosis of dental caries; the most common age group was 5–9 years. 1–3

Dental caries and its consequences are not limited to young children or the primary dentition. There is evidence surrounding the trajectory of disease, showing that children who experience carious lesions in their primary dentition carry a much greater burden of disease into adolescence and beyond. 14 Dental caries also affects one in three UK 12-year-olds, is positively associated with deprivation15 and is experienced by almost half of those aged 12–15 years in deprived areas. In 2013 in England, 32% of 12-year-olds experienced dental caries and required treatment, ranging from 46% of those eligible for free school meals to 30% of those ineligible. 16

Changing strategies for management of carious lesions in primary teeth

Until recently, dental caries was considered an infectious disease. As a result, treatment required removal of all bacteria, biofilm and affected tooth tissue before restoration. However, the disease is now understood to result from an ecological shift in the varying proportions of different commensal bacteria, resulting in a dysbiotic microflora in the dental biofilm. 17 Acidogenic, aciduric and cariogenic bacteria are highly competitive within the biofilm and, where there is a frequent supply of fermentable (dietary) carbohydrates, conditions develop that favour them, supporting their domination of the biofilm ecology. This ecological shift in biofilm composition leads to a concurrent shift in bacterial activity, with long periods of acidity and net mineral loss from the dental enamel and dentine resulting in the carious lesion. 18

This more recent view of the disease has allowed a rationale for the development of alternative treatments, rather than the traditional complete surgical excision (drill-and-fill) model. These alternative treatments involve controlling precipitating factors for the disease, through carbohydrate restriction (diet change to reduce sugars intake), biofilm removal (through tooth-brushing), promoting tooth tissue remineralisation (through professional- and home-applied fluoride) and biofilm sealing from substrate (using restorative techniques). 19 A resultant shift away from cariogenic- to non-cariogenic biofilm promotes remineralisation, inactivates the disease process and preserves tooth tissue. This minimal intervention dentistry approach20 follows the contemporary tenet of quaternary prevention. In addition, it avoids, or slows down, the cycle of re-restoration21 and is less invasive for patients.

The evidence for effective management of carious lesions in primary teeth

Teaching in UK dental schools has been based on guidance from the British Society of Paediatric Dentistry (BSPD). Until recently, the BSPD recommended that the optimum treatment of carious lesions in primary teeth is complete removal of the carious tissue, followed by the placement of a conventional restoration (filling) to repair lost tooth tissue. 22,23 However, these recommendations were largely based on evidence from studies conducted in either a secondary care or a specialist paediatric dental practice setting,24 rather than the primary dental care environment, where the vast majority of children are treated. More recently, the Scottish Dental Clinical Effectiveness Programme (SDCEP) has recognised the growing evidence for minimally invasive, biologically based approaches to carious lesion management and has developed national guidance for the management of caries in children,25 which is based on this approach.

Despite long-standing teaching in paediatric dentistry and high-quality guidance for primary dental care in the UK, the proportion of primary teeth with visible carious lesions that are restored (the Care Index) remains low. In Scotland it is 13%,26 whereas in England it is 14%. 27 The perceived ineffectiveness of the traditional ‘drill-and-fill’ methods of managing decayed primary teeth has been hypothesised as one reason why this approach is unpopular with GDPs. 28

Pain and dental infection (sepsis)

Although a highly prevalent condition, the impact of childhood dental caries is often under-appreciated, as the disease itself is rarely life-threatening or overtly limiting on daily activities. However, the consequences of the disease, including pain62, interference with sleep and reduced school attendance,63 can have significant effects on children’s daily lives. In 2013, 6% of 12-year-olds and 3% of 15-year-olds reported difficulty with schoolwork because of the condition of their teeth and mouth over the previous 3 months. 16 Dental caries can also affect the general health and quality of life (QoL) of children, impairing growth and cognitive development,64 as well as interfering with nutritional status,63 and may even have an effect on attainment. 65

Estimates of pain and infection from dental caries in children are difficult to establish. However, the 2016 National Dental Inspection Programme26 epidemiological study in Scotland, in which 86% of Primary 1 children were inspected (at approximately 5 years of age), reported that 7.5% of the children examined had a dental status that required them to be issued with a letter indicating that they ‘should seek immediate dental care on account of severe decay or abscess’.

One of the difficulties with measuring pain and infection in young children is their limited ability to communicate. Young children do not find it easy to describe and report pain. 66 Cognitively, the ability to understand pain and differentiate a chronic pain from the absence of pain is a complex phenomenon. Young children who grow up with pain from an early age may not realise that this is not normal. When they do realise they have pain, it can be difficult to have them describe this precisely to help with reaching an accurate diagnosis.

Dental caries incidence

Since the turn of the millennium, it has been widely acknowledged that the commonly used threshold for the measurement of dental caries in clinical trials is no longer appropriate, because this threshold represents disease that is already well advanced (into dentine). 67,68 The National Institutes of Health (US) (NIH) Consensus Development Conference in Bethesda, MD, USA, in 200167 identified the benefits of including the recording of early dental caries in clinical trials, stating ‘There was a paradigm shift in the management of dental caries toward improved diagnosis of early non-cavitated lesions and treatment for prevention and arrest of such lesions’. In addition, the panel recommended that ‘clinical trials of established and new treatment methods . . . should conform to contemporary standards of design implementation, analysis and reporting’. Many technical adjuncts to carious lesion detection and diagnosis have become available over the years,69 but their use is not widespread in clinical dentistry and they remain, primarily, a clinical research tool.

The most widely used adjunct for carious lesion detection, which is in general use in clinical dentistry, is radiographs. The mandatory use of radiographs as part of a clinical dental research trial is universally agreed as unethical. However, they are recommended at risk-based intervals in UK guidelines. 70 Nevertheless, it is clear from the literature,71 and was highlighted in the FiCTION pilot trial and feasibility study,72,73 that radiograph use in children does not follow the expected frequency.

Child- and parent-reported outcomes: oral health-related quality of life

The value of patient-reported observations on dental treatment experiences is becoming widely accepted and patients’ perspectives are seen as valuable. Patient-reported outcome measures (PROMs) are now included in the vast majority of trials to evaluate changes in the impact of the condition following treatment from the patient’s perspective.

Many PROMs have been produced to measure oral health-related quality of life (OHRQoL), which was defined by Locker and Allen74 as ‘the impact of oral diseases and disorders on aspects of everyday life that a patient or person values, that are of sufficient magnitude, in terms of frequency, severity or duration to affect their experience and perception of their life overall’. Most existing measures of OHRQoL in children, including the Child Perceptions Questionnaire (CPQ), the Child – Oral Impacts on Daily Performances (C-OIDP) index and the Child – Oral Health Impact Profile (C-OHIP), are designed to cover a variety of oral conditions, such as dental caries, malocclusion and craniofacial anomalies. Although these measures have been developed for children to self-report, for young children (< 8 years of age) most studies rely on parental reports of the impact of their child’s health on their daily lives. 75

Child- and parent-reported outcomes: dental anxiety and worry

The prevalence of child dental anxiety, across Europe, ranges from 3% to 21%,76–78 with 14% of children experiencing extreme dental anxiety. 78 These findings suggest a continuum of child dental fear across populations of children, including those who attend for dental treatment. 79

The literature suggests that younger preschool children may find any form of dental intervention frightening; consequently, they may be disruptive during treatment. 80 For the younger, preschool child, it is not the actual dental intervention or seriousness of the treatment that is important, but the imaginings or fantasies stirred up by it. 81 These imaginings include feeling helpless, having to submit passively to treatment, fear of pain and being separated from their parent. Similar observations are made for children who experience dental pain and for those who suffer from high dental anxiety. 79

With psychological development, the once fearful and disruptive preschool child usually becomes able to manage dental treatment. 80 Therefore, it is important when investigating different dental treatment modalities, for preschool and primary school-aged children, to assess child dental anxiety and to examine how dental anxiety status changes with chronological age. Different dental treatment modalities may also influence children’s dental fear, and some may be more appropriate for children with dental anxiety. Therefore, it is important to assess not only dental trait anxiety (anxiety that is more stable) but also dental state anxiety (anxiety that is associated with specific situations) to assess the effect of the treatment on the dental anxiety experienced by the child patient during and after treatment.

In summary, to assess child dental anxiety in the FiCTION trial, to demonstrate how this affect changed with the experience of pain and chronological age and to identify the most appropriate treatment modality to treat carious lesions in the dentally anxious child, state measures of dental fear were chosen. The justification for choice of measures to assess child dental anxiety is described in Chapter 2, Child- and parent-reported outcomes: participant dental anxiety, worry and discomfort during dental treatment.

Economic analysis to determine the cost-effectiveness of different treatment approaches

(See Chapter 2, Cost-effectiveness of managing dental decay in primary teeth, and Chapter 4). Dental care for dental caries has high direct costs for the NHS;82 therefore, we need to ensure that health benefits are being maximised within the budget. An economic evaluation was conducted alongside the FiCTION trial to determine the cost-effectiveness of the different treatment approaches. These results can be used to facilitate efficient resource allocation for managing carious lesions in primary teeth.

Acceptability and associated experiences of the FiCTION trial arms for children, parents/guardians and dental professionals

Dental treatments vary in their degree of acceptability to patients. The more acceptable a treatment is to the patient, parent or care-deliverer, the more likely it is to be delivered and received. The value of children’s opinions on their dental treatment experiences has become more widely accepted and their perspectives are seen as credible. Although a vast body of research has compared different types of treatment to manage carious lesions, most of the research has focused on outcomes related to choice and characteristics of dental materials;33 very little research has included the preferences or perspectives of participants as outcomes.

Objectives

The objectives were to assess:

• the clinical effectiveness and cost-effectiveness of three treatment strategies for managing dental caries in primary teeth

• the three treatment strategies with respect to –

  • child- and parent-reported outcomes, including child OHRQoL, dental anxiety, worry and discomfort during treatment

  • acceptability and associated experiences for children, parents and members of the dental team

  • caries development and/or progression in primary and permanent teeth.

Practice recruitment

The initial target was to recruit 50 practices, 10 from each of the five CCs (see Sample size calculation). The selection of practices was designed to reflect the sociodemographic mix of the catchment communities. A range of strategies were used to identify and invite practices to participate. Practices from which an expression of interest in trial participation was received were visited by the research team to assess their eligibility before their inclusion was confirmed.

Practice retention

The trial manager, clinical leads and their secretaries, and the clinical researcher actively maintained regular contact with all participating practices throughout the trial. They identified practice retention or associated problems early, using the formally established communication strategy and informally through their ongoing engagement with the practice, working closely with the practitioners to troubleshoot any problems. Regular e-mail and telephone updates from clinical leads’ secretaries and the trial manager, plus quarterly newsletters from the trial core team, were issued during the trial. Active support was sought from the PCRNs, research networks and local research champions for recruitment and retention of practices.

Continuing professional development (CPD) credit for all members of each FiCTION trial practice dental team was made available for attendance at any trial training events.

The practices were remunerated with service support costs for participant screening and recruitment activities, as well as with research costs based on the additional time spent on trial-related data collection. Dental treatment was remunerated in the normal way, depending on the nation where the practice was located.

Child participant inclusion criteria

Children (aged 3–7 years) who:

• were willing to be dentally examined

• had at least one primary molar tooth with decay into dentine (i.e. carious lesion) on clinical examination

• were known regular attendees or, if new to the practice, considered likely to return for follow-up.

Child participant exclusion criteria

• Children aged < 3 or > 7 years.

• Children aged 3–7 years who:

  • at the recruitment appointment, were accompanied by an adult who lacked the legal or mental capacity to give informed consent

  • at the recruitment appointment presented with either dental pain and/or dental sepsis due to caries (as diagnosed by the GDP from patient history, examination, radiographs). These participants were not enrolled into the trial at that point, but, after treatment, could be reassessed for eligibility. Discomfort or pain associated with erupting or exfoliating teeth or an incident of trauma or oral ulceration was not an exclusion criterion

  • had a medical condition requiring special considerations with their dental management, for example cardiac defects or blood dyscrasias

  • were currently involved in any other research that might have affected this trial

  • were part of a family that knew they would be moving out of the area during the 3 years following recruitment.

Conventional management of carious lesions, with best-practice prevention

[This intervention will hereafter be referred to as conventional with prevention (C+P).] (See Chapter 1, Evidence from conventional/traditional approaches to managing carious lesions.) Conventional management is commonly known as the ‘drill-and-fill’ method and is the traditional approach to managing dental caries that has been taught and practised for many years. It is based on active management of carious lesions by complete removal of the carious tissue. The teeth are numbed with local anaesthesia (a dental injection), carious tissue is mechanically removed using rotary instruments (drill) or by hand excavation (using hand tools), and a restoration (filling) is placed in the tooth to fill the cavity. If the dental pulp (nerve) is exposed during carious tissue removal or there are symptoms of pulpitis, a pulpotomy (removal of some of the nerve) may be carried out, followed by provision of a metal crown (usually) or a filling. Retained roots, and teeth for which the crowns are unrestorable or the pulp chamber is open, are managed by extraction (removal) of the tooth/root following local anaesthesia.

Best-practice prevention was carried out in line with current guidelines and as per the prevention alone (PA) arm.

Biological management of carious lesions, with best-practice prevention

[This intervention will hereafter be referred to as biological with prevention (B+P).] (See Chapter 1, Evidence from non-destructive/minimally invasive approaches to managing carious lesions.) This minimally invasive approach to managing carious lesions involves sealing decay into the tooth and separating it from the oral cavity; this is achieved by application of an adhesive filling material over the caries or by covering with a metal crown. It may be clinically necessary, on occasion, to partially remove superficial carious tissue prior to the tooth being sealed. Local anaesthetic injections are rarely needed. Retained roots and teeth for which the crowns are unrestorable, or dental pulp is exposed, are managed on a tooth-by-tooth basis. In situations when a tooth has active carious lesions (decay still progressing) or when the clinician decides that the tooth is likely to give the patient pain or sepsis before it exfoliates (falls out), the caries is managed by extraction following local anaesthesia.

Best-practice prevention was carried out in line with current guidelines and as per the PA arm.

Best-practice prevention alone

[This intervention will hereafter be referred to as prevention alone (PA).] (See Chapter 1, Evidence from Good Practice Prevention approaches to managing caries.) For the PA arm, no drilling, filling or sealing of primary teeth occurred. Treatment plans for participants were based on best-practice preventative care for teeth and oral health. This involved the four component pillars of prevention, carried out according to current guidelines:25,87

• dietary investigation, analysis and intervention to reduce fermentable carbohydrates in the diet

• tooth-brushing twice daily with a fluoridated toothpaste, plus fluoride mouth-rinsing in children > 7 years of age

• topical fluoride varnish applied to primary and permanent teeth by a dental professional

• fissure sealants for permanent teeth.

Training of dentists and practice staff

Between July and October 2012, each CC hosted a practice training day to deliver clinical and trial process training (which included GCP and informed consent training) to all enrolled dentists. Whenever possible, dental therapists/hygienists/nurses and practice receptionists/managers were also trained at the practice training day. For dental team staff who could not attend a practice training day, training was delivered as part of a site initiation visit by the trial manager and clinical researcher.

Training was provided for the individual clinical procedures with which dentists were unfamiliar and was tailored as far as possible to each group of dentists. Topics included, but were not limited to, recording dental caries using ICDAS, taking radiographs in children, the Hall Technique and conventional crown provision. Additional training materials were developed for taking radiographs and for the Hall Technique;35 these materials were made available to participating dentists in their practice. Although the detection of dental sepsis (infection) is a standard part of a dental clinical examination, given its importance as one of the primary outcomes, specifically directed training was included.

Training was given during appropriate treatment planning and delivery to children according to the randomised trial arm. This comprised a didactic teaching session from members of the core team followed by practical treatment planning with cases and discussion with the local clinical lead and co-chief investigators.

Primary outcomes

When the trial was originally designed it was powered only for a single primary outcome of incidence of dental pain and/or dental sepsis. However, as the trial progressed, it became clear that the number of episodes of dental pain and/or dental sepsis experienced by a child is a more clinically relevant outcome and, statistically, a more sensitive measure (compared with dichotomising the number of episodes into zero episodes vs. one or more episodes). Therefore, it was decided that the number of episodes of dental pain and/or dental sepsis should be a co-primary outcome.

Originally, it was planned that participants would be followed up in the study for a fixed 3-year period. However, the extension to the trial recruitment period resulted in the maximum potential follow-up ranging from 23 to 36 months; hence, the co-primary outcomes were assessed across a variable follow-up period.

Secondary outcomes

See the project web page [www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)] for the SAP and study documentation, patient information sheet and questionnaires, for additional detail on the secondary outcome measures used.

Safety (harms)

The three interventions tested in this trial are all in common use in general dental practice and were considered to be of relatively low risk. Expected adverse events (AEs) are summarised in Appendix 3.

Because of the type of trial, non-serious AEs were not captured. However, the practice lead was required to report all serious adverse events (SAEs) to NCTU via a secure fax line within 24 hours of the practice learning of an occurrence.

A SAE is defined as any untoward and unexpected medical occurrence or effect that:

• results in death

• is life-threatening (refers to an event in which the subject was at risk of death at the time of the event; it does not refer to an event that hypothetically might have caused death if it were more severe)

• requires hospitalisation (for > 24 hours), or prolongation of the participant’s existing hospitalisation

• results in persistent or significant disability or incapacity

• is a congenital anomaly or birth defect.

Dental practice staff were reminded that clinical judgement should be exercised in deciding whether or not an AE was serious in other situations. Important AEs that were not immediately life-threatening or did not result in death or hospitalisation but might jeopardise the subject or require intervention to prevent one of the other outcomes listed in the definition above were also considered to be serious.

Hospitalisations for elective treatment of a pre-existing condition did not require reporting as SAEs. Unrelated hospitalisations were to be elicited at the post-treatment follow-up appointment, scheduled subsequent appointments and all unscheduled/emergency appointments.

All SAEs that, in the opinion of a co-chief investigator, were ‘related’ (i.e. resulted from the administration of any of the research procedures) and ‘unexpected’ (see Appendix 3) were reported to the REC.

Identification and screening of participants

Potential trial participants were children aged 3–7 years who were identified from participating dental practices, which invited the potential participants to participate through two routes:

1. Simple searches of practice databases to identify potentially eligible children, using a date of birth query. Potentially eligible children due for a recall visit were invited to participate by letter of invitation from the child’s GDP. This letter, together with information sheets for parents and the child (the child’s version was pictorial and age-appropriate), was sent with their dental appointment card at least 1 week ahead of the scheduled recall visit.

2. Parents of children presenting opportunistically, and identified as being potentially eligible for participation, were invited to participate at the time of presentation. Unless they declined immediately these parents were given the invitation letter and the parent and child information sheets, and time (a minimum of 24 hours) was allowed to consider participation in the trial before consent was sought.

Potential participants identified through both routes had a routine screening examination (‘check-up’) to confirm eligibility. This screening consisted of standard dental recall clinical investigations (including medical and dental history taking, questioning regarding oral pain since last visit and current oral pain, clinical examination of the soft tissues and teeth, and radiographs in line with national guidance). This allowed the identification of children with dental caries and the exclusion of children with current dental pain or dental sepsis due to caries, as well as medically compromised children.

For those children without evidence of caries into dentine or where dental pain due to caries and/or dental sepsis were present at the screening visit, the GDP explained why it was not possible to take part in the FiCTION trial at that time and, if relevant, treated the carious lesion(s) and related pain and/or sepsis. The family was informed that if the child re-presented with dental caries but without pain or sepsis, he or she could then be considered for the FiCTION trial. If a child was caries-free at screening but subsequently presented to the practice with caries during the recruitment phase of the trial, he or she could be invited to join the trial.

Recruitment and consent

Post screening, if there was evidence of dental caries and absence of both dental pain due to caries and dental sepsis, a FiCTION trial-trained dentist in the practice discussed the trial with the parent and child, supplementing the written trial information, and answered any questions they may have had. If the parent and child were willing to participate, and once eligibility had been confirmed, parental written informed consent, and oral or written assent from the child, were obtained by the FiCTION trial-trained dentist prior to any trial-specific procedures being carried out.

The child was then given a subsequent treatment appointment. It was intended that randomisation to a treatment arm should be carried out via the NCTU secure web-based randomisation service before the child returned for the subsequent visit, at which time the child and parent would be informed of the treatment arm to which the child had been allocated. A detailed ICDAS dental chart was completed at the initial treatment visit. Treatment was commenced as per the child’s randomised arm and clinical protocol. Families were also presented with a letter to give to their general medical practitioner to inform them of the child’s involvement in the trial.

For those children for whom consent was not given for participation in the trial, the dentist carried out the child’s normal dental care.

Participant management and visit schedule

The timing and frequency of subsequent visits was at the discretion of the GDP and the family. The original trial design assumed a fixed follow-up period of 3 years, but, as indicated in Sample size calculation, an extension to the trial recruitment period resulted in the maximum potential follow-up, post randomisation, ranging from 23 to 36 months, with those recruited after May 2014 having a truncated follow-up.

Any recurrence of caries in the teeth that were originally treated and any subsequent incidence of caries in other primary teeth were expected to be managed according to the randomly allocated arm and its associated clinical protocol. At any visit during which there was deviation from the allocated clinical protocol, either within or between arms, treatment deviation forms (TDFs) were completed by the GDP.

Data on all treatment provided over the trial period were collected at each FiCTION trial visit from the dental team (usually the GDP), and parents and children completed outcome questionnaires at all visits [see Table 1 and the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)].

A practice used its usual contact procedures to reschedule any missed visits until the child had not been seen for > 540 days. After this point had been reached the practice sent a did-not-attend (DNA) questionnaire, with a covering letter and pre-paid envelope, to the parent for them to complete and return to the NCTU.

All participants had a final trial visit date, with an associated defined final visit window, for collection of ‘final-visit’ data (see Table 1). A similar approach to that described above was used for children failing to attend their final visit: in this case, a DNA final visit questionnaire was posted by the practice to the parent for them to complete and return to the NCTU.

Participant retention

The concept of ‘marketing’ of clinical trials informed participant recruitment and retention strategies. 110 A distinctive FiCTION trial logo and branding was developed and used on all trial-related materials. On enrolment, parents and children were given FiCTION trial membership cards; this card carried details of their dentists and whom to contact should they change dentist/practice or require out-of-hours or emergency dental care. The card also included FiCTION trial website details, which gave the children access to colouring-in and other activities as well as providing trial updates and contact details of the trial team. Leaflets and posters were distributed for display in practice waiting rooms to convey the research team’s thanks to the FiCTION trial families for their participation, while enhancing the trial’s prominence in practices. Children were sent FiCTION trial birthday cards via the practice and practice feedback was used to guide the development of additional suitable FiCTION trial-branded merchandise promoting the trial.

Participant withdrawal

Participants had the right to withdraw from the trial at any time without having to give a reason. Practices were asked to try to ascertain the reason for withdrawal and document this on a Withdrawal Form [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)]. Participants who failed to return for follow-up visits were not considered to have withdrawn, unless an explicit request to withdraw was received or the practice confirmed that the child had moved to a practice that was not participating in the FiCTION trial. When practices withdrew from the trial, all of their FiCTION trial participants were considered to have been withdrawn at the time of the practice leaving.

Trial dates

The trial opened to recruitment on 1 October 2012. The first child was recruited on 12 October 2012 and the last child was recruited on 18 June 2015. The end of the trial was defined as last patient, last visit; follow-up of the last child was completed on 29 June 2017.

Sequence generation

The allocation sequence was generated by a statistician not otherwise involved in the trial, using variable length random-permuted blocks of size 6 and 9 (based on the ‘ralloc’ function in Stata).

Allocation concealment

Allocation concealment was achieved by use of a centralised web-based randomisation facility hosted by NCTU. Randomisation was intended to be carried out once the consent process was completed and the child had left the dental practice. The variable length random-permuted blocking ensured concealment of allocation.

Implementation

The intention was that participants were managed throughout their time in the trial according to the arm to which they had been randomly allocated, that is any subsequent dental caries would be managed in the same way (as per random allocation) as the initial occurrence. Any deviation from the allocated treatment strategy (i.e. when treatment of carious lesions involved elements of a treatment strategy other than that to which the child had been randomised, because of decisions or behaviours of the dental practitioner, parent or child) was recorded on a TDF at each follow-up visit at which it occurred.

Primary analysis of the co-primary outcomes

At the start of the trial, the single proposed primary outcome was the proportion of children with at least one episode of dental pain and/or dental sepsis during the planned 3-year follow-up period. However, the extension to the trial recruitment period resulted in the maximum potential follow-up ranging from 23 to 36 months; hence, the analysis strategy was adjusted to incorporate variable follow-up. As indicated in Primary outcomes, it was agreed that the incidence of dental pain and/or dental sepsis during the follow-up period and the total number of episodes of dental pain and/or dental sepsis for each child during the follow-up period would become co-primary outcomes. The original power calculation for the trial was based on a comparison of proportions and, as a result, was the only powered analysis; therefore, an exploratory hypothesis test for the unpowered comparison of the mean number of episodes is reported.

The analyses of the co-primary outcomes were conducted on the basis of intention to treat (ITT), defined as all randomised children with at least one CRF completed and in MACRO (Elsevier, Amsterdam, the Netherlands), including protocol violators and ineligible randomised participants, retaining participants in their randomised treatment groups. The results for the co-primary outcomes are reported as unadjusted and adjusted models. Models were adjusted for age in years at randomisation and time in the trial in years (defined as the time from randomisation to the date of last CRF in MACRO). Differences between dental practices were included as a random intercept; therefore, all adjusted models were mixed-effects models. Randomised treatment arm was included in the models as a factor with three levels, with the C+P arm as the reference group. The adjusted analyses were defined as the primary analyses and, therefore, the models to use for reporting purposes.

The primary analyses of the co-primary outcomes were:

1. The analysis of the incidence of dental pain and/or dental sepsis during the follow-up period, using logistic regression.

   The dependent variable was a binary indicator of whether or not there was a reported incidence of dental pain and/or dental sepsis during the follow-up period. We generated 97.5% CIs for the difference between treatment arms (PA vs. C+P and B+P vs. C+P), expressed as risk differences [by integrating over the covariates in the model (time in study and age) and the random effects]; p-values were reported.

2. The numbers of episodes of dental pain and/or dental sepsis were analysed using negative binomial regression.

   The dependent variable was the total number of episodes reported by a child during the follow-up period. Both negative binomial and zero-inflated negative binomial models were considered; the Vuong test was used to determine the best-fitting model. We generated 97.5% CIs for the difference between treatment arms (PA vs. C+P and B+P vs. C+P), expressed as incidence rate ratios (IRRs); exploratory p-values were reported.

Sensitivity analysis specific to the primary analysis of co-primary outcomes

As a planned sensitivity analysis, the models were also re-fitted including only participants with at least 23 months’ follow-up, to assess the influence of shorter lengths of follow-up on the treatment effect estimates (the rationale being that all participants had the opportunity to have at least 23 months of follow-up).

Secondary analysis of the co-primary outcomes

The time to the first episode of dental pain and/or dental sepsis has been identified as particularly important in relation to the age of the child, as the impact of dental pain and/or dental sepsis is greater on younger children and they have a reduced capacity to tolerate dental treatment.

Therefore, an analysis of time to first episode of dental pain and/or dental sepsis was undertaken using a Cox proportional hazards model, including age in years at randomisation as a continuous covariate and dental practices as a random effect. We generated 97.5% CIs for the difference between treatment arms (PA vs. C+P and B+P vs. C+P), expressed as hazard ratios (HRs); exploratory p-values were reported.

Analysis of the secondary outcomes

As described in Secondary outcomes, a number of secondary outcomes were measured during the period of follow-up. The timing of measurements is described earlier (see Table 1), with further description of the outcome measure index or scale given in the SAP [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)].

Quantitative secondary outcome data (P-CPQ-16 and MCDASf) were analysed using the same approach as for the co-primary outcomes. Models were adjusted for age in years at randomisation and time in the trial in years. When a baseline measure was taken into consideration, the baseline measurement was also included as a covariate. Differences between dental practices were included as a random effect. If outcomes were measured at each visit (MCDASf), an additional random effect was added to account for repeated measures nested within child. Randomised treatment arm was included in the models as a factor with three levels, with the C+P arm as the reference group. Within this framework, the mean difference between randomised treatment arms at final visit or across the whole of the follow-up period was estimated.

Ordinal categorical outcomes (anticipatory and treatment-related anxiety and worry) measured at each visit were modelled using multilevel mixed-effects ordinal logistic regression. The assumption of proportional odds was assessed graphically and using the approximate likelihood ratio test. When this assumption was not satisfied, multilevel mixed-effects multinomial regression was considered. The frequencies of observations in each outcome category at each visit were also taken into consideration. When necessary, outcome categories were collapsed to create a binary variable and multilevel mixed-effects logistic regression was used.

Scoring of questionnaires and handling of missing data

For the parent questionnaire (P-CPQ-16) measuring perceived child OHRQoL, missing and ‘do not know’ responses were treated as ‘missing data’ and the same method of imputation applied to both, without distinction or prioritisation. The method adopted was the imputation of the ‘subject–subscale mean’ [i.e. the respondent-specific mean across a minimum of two valid (i.e. 0–4) responses in the item’s own P-CPQ-16 subscale] for missing and ‘do not know’ responses in the subscale [only if two or more of the constituent subscale items had originally valid (0–4) responses]. There was no imputation for participants with missing or ‘do not know’ responses for all 16 items of the P-CPQ-16.

A similar approach was used for MCDASf, although, as there are no subscales in this instrument, this was effectively a ‘subject–overall mean’ imputation, if three or more of the six items had originally valid (1 to 5) responses.

These methods assume that the data are missing completely at random.

Adherence to randomised treatment arm’s clinical treatment protocol

Deviations from the randomised treatment arm’s clinical treatment protocol were recorded on the TDF. These forms were entered into an electronic clinical data management system, MACRO, and then verified by the clinical researcher and classified as a ‘major’ deviation from the randomised treatment arm clinical treatment protocol (if there had been a from-arm tooth treatment change) or otherwise. For the purposes of assessing adherence to the clinical treatment protocol, only ‘major’ deviations were considered.

Adherence to the clinical treatment protocol for each randomised treatment arm could not be evaluated at an individual tooth level because TDFs did not collect tooth numbers. Because multiple teeth could have been treated at any particular visit, a TDF could not be linked to a specific tooth or number of teeth. In addition, tooth numbers were not collected for prevention activities. Therefore, adherence to the clinical treatment protocol was evaluated using child-level summaries.

The prespecified per-protocol analysis set

The prespecified per-protocol (PP) analysis set excluded participants from the ITT analysis set who:

• were deemed likely to have had dental pain and/or dental sepsis at consent and/or

• were ‘non-compliant’ with the clinical treatment protocol corresponding to their randomised treatment arm (i.e. defined as having a TDF involving a ‘major’ deviation from the randomised treatment arm’s clinical treatment protocol at > 20% of their visits). A ‘major’ deviation was defined as a cross-arm tooth treatment change.

Planned exploratory multivariable analyses

The variables listed below were included in prespecified exploratory multivariable regression models for both the co-primary outcomes of incidence and the number of episodes of dental pain and/or dental sepsis, regardless of their univariate association with the outcome:

• Age of participant, in years, at randomisation as a continuous covariate.

• Number of decayed teeth (level 5/6 cavitation) at baseline from ICDAS charting for each participant.

• Participant ethnicity. Six ethnicities were listed in the parent baseline questionnaire: (1) white, (2) black, (3) Indian, Pakistani or Bangladeshi, (4) Chinese, (5) mixed race and (6) other. Ethnicity was explored as far as possible within the limitations imposed as a result of the distribution of participants across the categories and incomplete data.

• Fluoride concentration in tap drinking water. Tap-water fluoride concentration [in parts per million (fluoride ion) (p.p.m.F^–)] at the dental practice was used as a proxy for a child’s fluoridation status, in terms of the tap-water supply they received at home.

• Index of deprivation. The dental practice index of deprivation, based on the dental practice postcode, was used as a proxy for a child’s index of deprivation. The index was extracted from census data collected prior to the recruitment phase of the trial. Index of deprivation can change significantly both spatially and temporally (dependent on the neighbourhood), and so was expected to be a very approximate measure of the individual children’s level of household deprivation. In addition, each country of the UK uses different domains, indicators and weighting of domains to calculate the index of deprivation [England: see www.gov.uk/government/statistics/english-indices-of-deprivation-2015 (accessed 22 February 2019); Scotland: see www.gov.scot/Topics/Statistics/SIMD (accessed 22 February 2019); and Wales: see http://wimd.wales.gov.uk/ (accessed 22 February 2019)]. Methods have been proposed111,112 for cross-country comparisons, but were not employed in this trial.

Dental practice recruitment

The original practice (site) recruitment target was 50 dental practices, comprising approximately 10 practices from each of the five CCs (see Chapter 2, Practice recruitment): Scotland, Newcastle, Leeds/Sheffield, Wales and London. However, patient recruitment rate was lower than predicted within practices and some practices that originally initiated later withdrew. Following consultation with the TSC, the IDMC and the NIHR HTA programme, the target number of practices was increased to 70. The CCs were also expanded geographically, as described in Chapter 2, Practice recruitment.

Of the 93 practices that agreed to participate and received a site initiation visit, 21 did not randomise any participants, leaving 72 sites across the five CCs that recruited at least one participant. Ten of these 72 practices subsequently withdrew.

Dental practice characteristics

Practice characteristics collected were size (number of registered patients), deprivation index in quintiles (see Chapter 2, Planned exploratory multivariable analyses) and tap-water fluoridation status (p.p.m.F^–). These data are presented descriptively (Table 2) for the 72 practices that recruited at least one participant.

Screening at the sites

Patient/participant journey through invitation, screening, eligibility assessment, randomisation and subsequent progress through the trial is shown in the Consolidated Standards of Reporting Trials (CONSORT) flow diagram in Figure 2.

FIGURE 2.

The CONSORT flow diagram of participant journey through the trial. a, Prior to the start of the study, it was estimated that 18,717 children would be invited; b, prior to the start of the study, it was estimated that 65% of children invited would attend a screening appointment; 64% attended; c, prior to the start of the study, it was estimated that 85% of children screened would be ineligible – 76% were ineligible and 1% declined screening; d, prior to the start of the study, it was estimated that 20% of children screened and found eligible would decline to take part in the trial – 19% of those eligible declined; e, prior to the start of the study, it was estimated that 12% of children screened would be randomised – 15% of those screened were randomised. Reproduced from Innes NP, Clarkson JE, Douglas GVA, Ryan V, Wilson N, Homer T, et al. , Journal of Dental Research, 99(1), pp. 36–43, copyright © 2020 by International & American Associations for Dental Research 2019. Reprinted by Permission of SAGE Publications, Inc. 88

Reproduced from Innes NP, Clarkson JE, Douglas GVA, Ryan V, Wilson N, Homer T, et al. , Journal of Dental Research, 99(1), pp. 36–43, copyright © 2020 by International & American Associations for Dental Research 2019. Reprinted by Permission of SAGE Publications, Inc. 88

Recruitment rate

Recruitment opened on 1 October 2012; the first child was randomised on 12 October 2012. The trial closed to recruitment on 30 June 2015; the last child to enter the trial was randomised on 18 June 2015.

Figure 3 shows the recruitment rates originally anticipated, the revised target (October 2013), the further revised target (August 2014) and actual recruitment.

FIGURE 3.

Cumulative number of child participants randomised by month (all randomised analysis set, n = 1144).

Recruitment by arm

Seventy-two sites had randomised at least one child by the trial recruitment closure date. There was variation across centres in the numbers of participants recruited and randomised: Scotland, 307; Newcastle, 315; Leeds/Sheffield, 171; Wales, 125; and London, 226 participants. The number of randomised participants per practice ranged from 1 to 55, with a median of 14 [interquartile range (IQR) 5–23.5] (see Appendix 4, Table 22).

Numbers analysed: intention-to-treat analysis (1058 participants)

The ITT analysis set was defined as all randomised children with at least one CRF completed and in MACRO, including protocol violators and ineligible randomised participants, retaining participants in their randomised treatment groups. Of the 1144 randomised individuals, 86 (7.5%) did not return for any treatment and therefore had no CRF data collected: C+P, 34 out of 386 (8.8%); B+P, 29 out of 381 (7.6%); and PA, 23 out of 377 (6.1%). This left 1058 participants (from 68 of the 72 randomising practices) who constituted the ITT analysis data set on which subsequent analyses were based.

Ineligible participants

Ineligible participants in the ITT analysis set were classed as those randomised but subsequently found to not meet the eligibility criteria of the trial (see Chapter 2, Trial participants). Of the 22 participants who were considered thus ineligible, five were outside the age range of 3–7 years at consent date and 17 had dental pain and/or dental sepsis associated with carious teeth at the consent date. The 22 ineligible participants were distributed across the three arms (C+P, n = 6; B+P, n = 5; and PA, n = 11). As the proportion of ineligible participants was small [22 of 1058 in the ITT analysis set (2.1%)], a sensitivity analysis without them was not conducted.

First trial visits for intention-to-treat analysis set (1058 participants)

For the ITT analysis set, time from randomisation to first trial visit was similar across arms, with most participants (n = 736, 69.6%) having their first trial visit within 1 month of randomisation. For the remaining participants, the timing of their first trial visit post randomisation was as follows:

• more than 1 and up to and including 4 months, n = 241 (22.8%)

• more than 4 and up to and including 12 months, n = 63 (6.0%)

• more than 12 months, n = 18 (1.7%).

The distribution of time to first trial visit was balanced across arms.

Follow-up and final trial visits

Summary data for the time in trial and number of visits, by arm, are shown in Table 3. Participants spent a median of just less than 34 months in the trial, with a median of 2.5 visits per year. These parameters were well balanced across arms. A total of 711 participants attended a final trial visit (67% of the ITT participants); this was well balanced across the three arms.

Withdrawals

There were 118 documented withdrawals from the ITT analysis set. All withdrawals were complete; that is, there was consent to use participants’ data up to the point of withdrawal but with no further follow-up or data collection after that point. Reasons for withdrawal were ascertained for 108 recruited participants.

Seven randomising practices withdrew (n = 38 participants). The median time from first randomisation to date of practice withdrawal was 561 days (IQR 482–822 days).

Other reasons for withdrawal were the participant moved away (n = 34); trial fatigue (n = 4); dental reason, for example a traumatic event (n = 8); and personal or another reason (n = 24). The numbers of participants who withdrew were evenly distributed across the three trial arms with no differences in the length of time in trial (time from randomisation to time of last known trial visit) or in the number of trial visits (see Appendix 4, Table 23).

Baseline comparability

Demographic and clinical baseline characteristics were compared across treatment arms descriptively (Table 4). No significance testing was carried out because of the randomised nature of the trial. Characteristics were well balanced across arms.

As Table 4 and Appendix 4, Table 24, show, children’s caries experience at entry to the trial (baseline) was balanced between the three trial arms for the number of primary teeth affected and the amount and severity of the lesions. The mean number of primary teeth affected by caries (d₃mft) was 2.72 [standard deviation (SD) 2.66] [the national mean d₃mft values for 5-year-olds who are not free of obvious carious lesions (i.e. d₃mft of > 0) are England (2015) 3.4, Scotland (2016) 3.93 and Wales (2015–16) 3.58]. 13,26,115 It should be noted, however, that the national surveys are of randomly selected children recruited in the school setting and not from those attending general dental practices, whose oral health may be better than non-attenders.

As the mean age of participants was 6 years, they had few FPMs present on entry to the trial; the mean number of FPMs recorded as at least partially erupted was 1.74 (SD 1.88). The decay experience in these teeth was very low, again as might be expected for relatively newly erupting permanent teeth, with a mean D₃MFT for FPMs of 0.05 (SD 0.35), and this was balanced between the trial arms (see Appendix 4, Table 25).

Baseline questionnaires were returned for almost all children [n = 1045 (99%)] and had a high level of completeness.

Child OHRQoL at baseline also showed balance between arms, with an overall mean score of 8.4 (SD 6.4), reflecting a group with an OHRQoL score similar to that found in other child populations. 95 The results of the MCDASf, with an overall mean score of 14.1 (SD 5.1), were also balanced across arms at baseline and similar to scores reported in other UK children. 103

Treatment received (intention-to-treat analysis set, n = 1058)

A detailed breakdown of the treatments received is provided as part of the economic evaluation (see Chapter 4). Appendix 5, Table 74, summarises the average total treatment provided at the first visit and at follow-up visits by arm. This table provides insight into how the children were initially treated, how the treatment progressed over the follow-up period and the differences in treatment by arm.

Adherence to protocol

‘Major’ deviations (i.e. clinical treatment provided not according to the randomised arm) reflected non-adherence to the clinical treatment protocol. At 6% (429/7713) of visits, we observed a ‘major’ deviation from the clinical protocol for a participant’s randomised treatment arm, that is a cross-arm deviation. These major deviations occurred in 263 participants, with 46% among participants randomised to the C+P arm, 15% in the B+P arm and 39% in the PA arm. The most frequently reported direction of treatment deviation for the C+P arm was to B+P (69%); for the B+P arm it was to C+P (80%); and for the PA arm it was to C+P (53%). The main reason for major deviations were parent factors (28%) and dentists’ clinical judgements (29%). The participants with major deviations were not equally distributed across trial arms: C+P, 99 out of 352 (28%); B+P, 51 out of 352 (14%) and PA, 113 out of 354 (32%). Further details on the reason for and direction of all major deviations are shown in Appendix 4, Tables 26 and 27.

For 104 participants (10%), > 20% of their treatment visits involved a major treatment deviation from their randomised treatment arm, and they accounted for 182 of the 429 major treatment deviations recorded on TDFs. Across the trial arms, these participants were distributed as follows:

• C+P arm, 37 out of 352 participants (11%)

• B+P arm, 19 out of 352 (5%)

• PA arm, 48 out of 354 (14%).

These 104 participants remained in the ITT analyses but were removed from the PP analyses.

Use of radiographs

For the ITT analysis set, radiographs were taken at 11% (886/7713) of visits over the duration of the trial. This was less than originally anticipated [see Chapter 2, Schedule of assessment of primary outcomes, and the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)]. The 886 radiographs were taken in 48% (511/1058) of children and distributed evenly across all three treatment arms: C+P, 49% (174/352); B+P, 47% (165/352); and PA, 49% (172/354). Sixty per cent (308/511) of children who had a radiograph taken had at least one taken during their first year in the trial; this accounted for 39% (349/886) of all radiographs taken.

The reasons for not using radiographs were collected via the CRF at 94% of visits. The justifications given were similar across treatment arms:

• not due to be taken, according to FGDP guidelines (42%)

• radiograph(s) not attempted (child compliance issues) (18%)

• already taken during this course of treatment (18%)

• other (12%)

• attempted but failed (5%)

• parental wish (4%).

When radiographs were taken, the presence or absence of periradicular infection was recorded in 73% of cases. The rate of periradicular infection was 11% and was balanced across arms; however, 17% of radiographs assessed for periradicular infection were considered to be of limited diagnostic use.

Given that the percentage of children with radiographs taken was less than our pre-defined threshold of 80% [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)], this source of data was excluded from contributing to the analysis of outcomes related to dental sepsis.

Proportion of children with at least one episode of dental pain and/or dental sepsis during the follow-up period (incidence)

Total number of episodes of dental pain and/or dental sepsis during the follow-up period

Descriptives: number of episodes of dental pain and/or dental sepsis

For the ITT analysis set, a total of 676 episodes of dental pain and/or dental sepsis were recorded over the follow-up period, which is an average of 0.64 episodes per participant. The number of episodes ranged from 0 to 7, but the majority of participants (86%) had 0 or 1 episodes (Table 7). The distribution of the number of episodes was similar across the arms, although the unadjusted rate of episodes in the PA arm was slightly higher, at 0.72 episodes per participant, than in the C+P arm (0.62 episodes per participant) or the B+P arm (0.58 episodes per participant). The PA arm had more participants experiencing ≥ 2 episodes of dental pain and/or dental sepsis than the C+P or B+P arm (17.5% vs. 12% vs. 12.5%, respectively) (Figure 4).

TABLE 7 - Number of episodes of dental pain and/or dental sepsis (ITT analysis set)

Reproduced from Innes NP, Clarkson JE, Douglas GVA, Ryan V, Wilson N, Homer T, et al. , Journal of Dental Research, 99(1), pp. 36–43, copyright © 2020 by International & American Associations for Dental Research 2019. Reprinted by Permission of SAGE Publications, Inc. 88

Episodes of dental pain and/or dental sepsis	C+P (N = 352)	B+P (N = 352)	PA (N = 354)	Total (N = 1058)
Minimum	0	0	0	0
Median (IQR)	0 (0–1)	0 (0–1)	0 (0–1)	0 (0–1)
Mean (SD)	0.62 (0.95)	0.58 (0.87)	0.72 (0.98)	0.64 (0.94)
Maximum	7	6	5	7
Number, n (%)
0	204 (58.0)	211 (59.9)	193 (54.5)	608 (57.5)
1	106 (30.1)	97 (27.6)	99 (28.0)	302 (28.5)
2	23 (6.5)	29 (8.2)	40 (11.3)	92 (8.7)
3	15 (4.3)	13 (3.7)	15 (4.2)	43 (4.1)
4	2 (0.6)	1 (0.3)	5 (1.4)	8 (0.76)
5	0 (0.0)	0 (0.0)	2 (0.6)	2 (0.2)
6	1 (0.3)	1 (0.3)	0 (0.0)	2 (0.2)
7	1 (0.3)	0 (0.0)	0 (0.0)	1 (0.1)

FIGURE 4.

Number of episodes of pain and/or sepsis by randomised treatment arm (ITT analysis set). (a) C+P (n = 352); (b) B+P (n = 352); (c) PA (n = 354); and (d) total (n = 1058).

When the descriptive analysis of episodes was restricted to participants with at least 23 months’ follow-up (n = 797), the overall mean number of episodes per participant increased from 0.64 to 0.72 (see Appendix 4, Table 31). For the PP analysis set (n = 940), the overall rate of episodes reduced to 0.59 (see Appendix 4, Table 32).

Secondary analysis of the primary outcome: time to first episode of dental pain and/or dental sepsis

Descriptives: time to first episode of dental pain and/or dental sepsis

The Kaplan–Meier survival curves for the time to first episode of dental pain and/or dental sepsis for the ITT analysis set are shown in Figure 5. From these curves, the estimated probability of having no dental pain and/or dental sepsis at 2 years post randomisation in the C+P arm is 0.64 (97.5% CI 0.58 to 0.69), in the B+P arm it is 0.65 (97.5% CI 0.59 to 0.70) and in the PA arm it is 0.56 (97.5% CI 0.50 to 0.61). Further Kaplan–Meier probabilities by length of time in the trial in years and treatment arm are given in Appendix 4, Table 34. The median time to first episode and associated quartiles by treatment arm are given in Appendix 4, Table 35.

FIGURE 5.

Kaplan–Meier estimates of survivor functions by randomised treatment arm, generated using Stata version 14 (StataCorp LP, College Station, TX, USA) (ITT analysis set, n = 1058). (a) Total; (b) C+P arm; (c) B+P arm; and (d) PA arm.

Pre-planned exploratory analyses: dental pain and/or dental sepsis ever (ITT analysis set)

The relationship between the incidence of dental pain and/or dental sepsis during the follow-up period and the explanatory variables of age, ethnicity, practice tap water fluoride level, index of deprivation and number of decayed teeth at baseline from ICDAS charting (level 5/6 cavitation), which are described in Chapter 2, Planned exploratory multivariable analyses, were explored descriptively (Table 10) and in univariate and multivariable logistic regression models (see Appendix 4, Tables 36 and 37). From the summary statistics presented in Table 10, those participants who had dental pain and/or dental sepsis had, on average, more teeth with level 5/6 cavitation at baseline than those who did not. All other variables had very similar distributions across the two groups. This is also shown in the exploratory univariate analysis in which the only variable significantly associated with the incidence of dental pain and/or dental sepsis was the number of teeth with level 5/6 cavitation (see Appendix 4, Table 36). The exploratory multivariable model for incidence of dental pain and/or dental sepsis that was adjusted for all exploratory variables did not provide evidence at the 2.5% level of a difference between the arms (see Appendix 4, Table 37).

New dental carious lesions

As children entered the trial with caries experience in some of their primary teeth, the assessment of carious lesion incidence in the primary dentition over the trial period considered only teeth that were neither filled nor had dentinal lesions at baseline. We have used the term ‘sound/reversible’ for these teeth, that is those that were caries free or had non-cavitated enamel lesions only. Tables 24 and 25 in Appendix 4 show that the numbers of such teeth were balanced at baseline between the three trial arms. Teeth that were ‘sound/reversible’ at baseline but that experienced dentinal decay during the trial were taken as having caries progression and included in the analysis of caries incidence.

As children exhibited little decay experience in their FPM teeth on entry to the trial (mean age 6 years, around the time when the FPMs are erupting), and both the number of FPMs and their limited caries experience was balanced between the groups, caries incidence in the permanent teeth is based on only the assessment of the FPMs at the end of the trial.

Carious lesion incidence was analysed in this way at both the child (progression in any primary tooth or FPM) and the tooth level (number of primary teeth or FPMs that exhibited progression).

For all participants in the ITT data set with at least one ‘sound/reversible’ primary tooth or FPM at baseline and ICDAS data at follow-up (n = 653), the baseline characteristics were balanced across the three arms (see Appendix 4, Table 40).

Overall, 399 out of 653 (61.1%) participants had caries development/progression in one or more ‘sound/reversible’ primary teeth over the follow-up period [C+P, 57.9%; B+P, 61.6%; and PA, 64.1% (see Appendix 4, Table 41)]. The adjusted risk difference in the incidence of caries development/progression in primary teeth over the follow-up period (Table 11) in the B+P compared with the C+P arm was 0.03 (97.5% CI –0.06 to 0.11), indicating, on average, a 3% increased risk of caries development/progression in the B+P arm compared with the C+P arm. For the PA arm compared with the C+P arm, the adjusted risk difference was 0.05 (97.5% CI –0.03 to 0.14), indicating, on average, a 5% increased risk of lesion development/progression in the PA arm compared with the C+P arm. However, it must be noted that this analysis is on a subset (n = 653) of the ITT analysis set (61.7% of participants are included), and so must be treated with caution (see Table 11).

On average, this subset (n = 653) had 1.3 progressed primary teeth that had been ‘sound/reversible’ (i.e. were sound or had non-cavitated enamel lesions at baseline) (see Appendix 4, Table 41). The range of progressed primary teeth per participant was 0–6, but the majority of participants were at the lower end of this range; 81.3% of the 653 participants had between 0 and 2 teeth that progressed (0 teeth progressed, 38.9%; 1 tooth progressed, 25.6%; 2 teeth progressed, 16.8%) and the distribution between arms was similar (Figure 6). The unadjusted mean number of progressed primary teeth was 1.2 in the C+P arm, 1.2 in the B+P arm and 1.4 in the PA arm.

FIGURE 6.

Number of progressed teeth (primary teeth and FPMs) by randomised treatment arm (n = 653). (a) C+P arm, primary teeth; (b) C+P arm, FPMs; (c) B+P arm, primary teeth; (d) B+P arm, FPMs; (e) PA arm, primary teeth; and (f) PA arm, FPMs.

With regard to the number of primary teeth in which dental caries developed/progressed from ‘sound/reversible’ at baseline, as Table 12 shows, the IRR for the B+P arm relative to the C+P arm, adjusted for time in trial, age and including a random effect for practice, was 1.01 (97.5% CI 0.83 to 1.24), and for the PA arm to relative to the C+P arm it was 1.15 (97.5% CI 0.95 to 1.41). There was no evidence of a statistically significant difference at the 2.5% level between the arms when comparing the B+P with the C+P arm or the PA with the C+P arm.

There was no evidence of a difference in carious lesion development/progression in FPMs between the three arms (see Tables 11 and 12, Figure 6 and Appendix 4, Table 41).

Children’s oral health-related quality of life

The P-CPQ-16 was administered at baseline and at the final visit. At baseline, 93% (984/1058) of parents returned at least partially completed questionnaires. Baseline and final visit parent questionnaires were received from 59% (627/1058) of parents. The P-CPQ-16 was well completed at both baseline and the final visit, with P-CPQ-16 scores being calculable, following imputation, for 94% and 95% of respondents at baseline and the final visit, respectively. Just over half of the children in the trial [560/1058 (53%)] had a P-CPQ-16 score following imputation at both baseline and the final visit (see Appendix 4, Tables 42–46). Box plots of the P-CPQ-16 scores at baseline and the final visit, and the change in score from baseline to the final visit, are shown in Figure 7.

FIGURE 7.

Box plots of P-CPQ-16 scores for the participants with both baseline and final data (n = 560). (a) Baseline scores; (b) final visit scores; and (c) change in score from baseline to final visit.

By the end of the trial scores had changed, on average, very little. In the analysis adjusted for baseline P-CPQ-16 score, age at randomisation and time in trial, the estimated mean P-CPQ-16 at the final visit was, on average, 0.27 points higher (97.5% CI –1.08 to 1.62) in the B+P arm than in the C+P arm, and 0.17 points higher, on average, (97.5% CI –1.20 to 1.53) in the PA arm than in the C+P arm. These estimated positive mean differences represent a very slightly poorer OHRQoL for children in the B+P and PA arms than for children in the C+P arm, but the differences were very small, not statistically significant and would not be considered clinically meaningful (see Appendix 4, Tables 42–46).

Children’s general anxiety and worry about dentistry (see Appendix 4, Tables 47–50)

The MCDASf was used to assess whether or not the way in which children’s teeth were managed influenced how they felt about dentistry (trait dental anxiety); lower scores represent lower anxiety. Child questionnaires were completed at every visit and were returned for 88% (6793/7713) of visits.

After adjusting for baseline MCDASf, age at randomisation and time in the trial, there was no evidence of any statistically or clinically significant difference between arms in levels of trait dental anxiety, averaged over all follow-up visits. The estimated post-baseline mean MCDASf score was 0.07 points lower for children in the B+P arm than for those in the C+P arm (97.5% CI –0.74 to 0.59), and was 0.22 points higher in the PA arm than in the C+P (97.5% CI –0.44 to 0.89 (see Appendix 4, Table 50).

Anticipatory and treatment-related anxiety and worry (state anxiety) (see Appendix 4, Tables 51–59)

Using the child and adult questionnaires, anticipatory anxiety was recorded at baseline; at every subsequent treatment visit, both anticipatory and treatment-related anxiety were recorded in the same way, with one question posed before treatment and one after treatment. A small proportion of questionnaires had missing data for these pre- and post-treatment questions (child questionnaires: < 5% missing; parent questionnaires: 6% missing). See Appendix 4, Table 51. The percentages of children and parents answering ‘not worried’ at baseline and over all subsequent visits were compared between trial groups.

Discomfort during treatment

Discomfort during treatment was recorded independently from the child’s, parent’s and dentist’s perspectives at each treatment visit. Out of the 7713 visits, the question on treatment discomfort was answered 6468 (84%) times by children, 6498 (84%) times by parents and 7301 (95%) times by dentists. The data are summarised in Appendix 4, Tables 60–63.

Children reported having experienced no discomfort at the majority of their appointments; 71% (4571/6468) of the time they reported that their treatment did ‘not hurt at all’. The proportion of visits eliciting this response was higher in the PA arm, 76% (1528/2018), than in either the C+P arm [68% (1551/2278)] or B+P arm [69% (1492/2172)].

At 77% (4713/6093) of visits, parents thought that their child’s treatment was ‘not at all painful’. This was reported more often in the PA arm, 83% (1596/1915), than in the C+P arm [73% (1536/2116)] or B+P arm [77% (1581/2062)].

Dentists perceived ‘no apparent discomfort’ on behalf of the children in 68% (4990/7301) of treatment visits. They reported this more often for the PA arm, 76% (1736/2292), than for the C+P arm [63% (1614/2578)] or B+P arm [68% (1640/2431)].

Child dental discomfort (caused by toothache rather than treatment discomfort)

The Dental Discomfort Questionnaire (DDQ-8) was well completed, with few missing items (< 8%) (see Appendix 4, Table 68). This descriptive additional analysis of parent-reported child dental discomfort suggested that the scores between treatment arms were very similar and at the low end of the scale, indicating that parents’ perception was of little child discomfort experienced from toothache (see Appendix 4, Table 69).

Serious adverse events

No SAEs were reported.

Cost data collection

Costing methods

As previously mentioned, two costing techniques were undertaken as part of the FiCTION trial economic evaluation: microcosting and the current charges to the NHS. A sensitivity analysis explored FFS and UDAs individually and the impact the two different fee structures had on the overall cost-effectiveness of the treatment strategies being explored.

Charges to the NHS

Fee for service

In Scotland, GDPs are reimbursed for every unit of activity they perform. As it has been suggested that this fee-per-item system may create an incentive to overtreat participants,122 this was explored in a sensitivity analysis. Unit costs were collected from the Information Services Division Scotland. 123 The cost of treating each child for every visit was used to estimate the average total cost in Scotland for treating dental decay in primary teeth over the time in the trial of at least 23 months. The unit costs used in the FFS analysis are detailed in Appendix 5, Table 71.

Units of dental activity

In England and Wales, a reimbursement similar to the current Scottish FFS-based system was in play until 2006, when UDAs were introduced. 124 UDAs are a fixed reimbursement that GDPs receive for dental work and vary between practices (between £16 and £40 per UDA). 119 Under the UDA system, dental procedures can be classified into three bands:124

1. Band 1 (1 UDA) – diagnosis, treatment planning and maintenance. It includes examination, radiography, scale and polish and preventative care.

2. Band 2 (3 UDAs) – simple treatments such as fillings, extractions and periodontal treatment.

3. Band 3 (12 UDAs) – complex treatment that includes a laboratory element. Examples include bridges, crowns and dentures.

In addition, GDPs can receive 1.2 UDAs for a band 1 urgent treatment. To incorporate urgent treatment into the UDA analysis, we defined urgent treatment as an unscheduled visit that occurred at the start of a course of treatment. We assumed that needing urgent treatment would begin a new course of treatment, as it was unlikely that a child would require urgent care during an existing course of treatment. This assumption underestimated UDA reimbursement by 0.2 UDAs if an unscheduled visit occurred during a course of treatment. However, the number of visits affected by this assumption was minimal [n = 15 (< 1%)] and, in these cases, the number of days between the unscheduled visit and the previous visit was < 28 (mean 18.8); hence, it was assumed that the unscheduled visit was related to the previous visit.

It is important to note that UDA(s) are reimbursed for only one course of treatment. A course of treatment definition is provided in the health economic analysis plan [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)] and Figure 8 is an illustrative presentation of the course of treatment pathway.

FIGURE 8.

Course of treatment pathway [see HEAP at project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/074403/#/ (accessed 27 November 2019)].

We collected UDA values for FiCTION trial practices in England (n = 43) from the NHS Business Services Authority website,119 which has information on dental processing and payments. The UDA value was estimated to be the value of the total contractual payments divided by the number of contracted UDAs based on the value of GDS contracts only. The average health board UDA value was adopted for the Welsh practices (n = 4). This information was obtained through personal communication (Research and Development Division, Welsh Government, October 2016, personal communication). Practice-level UDA values used in the UDA analysis are summarised in Appendix 5, Table 72. UDAs were discounted based on when a course of treatment started.

The UDAs incurred for each course of treatment were combined into a total number of UDAs received for each child over their time in the trial (see Appendix 5, Table 75). These UDAs were combined with practice UDA values to produce a total cost per child and, hence, the average total costs in England and Wales for treating dental decay in primary teeth.

Effectiveness measure

Health outcomes were measured in natural units; incidence of dental pain and/or dental sepsis and episodes of dental pain and/or dental sepsis. The methods for deriving these outcomes measured are described in detail in Chapter 2. It was assumed that those who did not return for regular appointments during their follow-up did not experience dental pain and/or dental sepsis.

Cost-effectiveness analysis

Adjusted analyses were performed to estimate cost-effectiveness. All results were presented as point estimates of the mean incremental differences in costs and effects, the incremental cost per incidence of dental pain and/or sepsis avoided, and the incremental cost per episode of dental pain and/or dental sepsis avoided.

Sensitivity analyses

Sensitivity analyses were conducted to assess the robustness of the results to realistic variations in the levels of underlying data. Deterministic sensitivity analyses were used to address any uncertainty in the assumptions used in our analysis. A stochastic sensitivity analysis using the bootstrapping technique130 explored the impact of the statistical imprecision surrounding estimates of costs, effects and cost-effectiveness. The bootstrapped results from the three-arm comparison were presented as a cost-effectiveness frontier. The cost-effectiveness frontier131 allowed us to determine the treatment strategy that maximised net benefits at each willingness-to-pay (WTP) value for an additional unit of health effect (i.e. dental pain and/or dental sepsis avoided).

The base-case analysis was replicated to include children who were treated in accordance with the clinical protocol for at least 80% of visits (see Chapter 2, The prespecified per-protocol analysis set, and Chapter 3, Adherence to protocol) to determine what effect, if any, this had on the overall results. The results of the PP analysis are presented in Appendix 5, Table 79.

All costs and effects were discounted at the recommended rate of 3.5%. 132 Discounting for incidence was based on when the child incurred the dental pain and/or dental sepsis during their follow-up period. Discounting for episodes was based on when the episode began. Similar to the co-primary outcome analysis, 97.5% CIs were generated for point estimates of costs and effects and for the differences between the management strategies.

Complete-case analysis

A complete-case analysis was performed on all children who attended their final visit, regardless of the number of visits they had over their period of follow-up. The inclusion of only these children allowed us to estimate the cost-effectiveness of the three treatment strategies with at least 23 months’ follow-up data and overcome some of the issues relating to unscheduled visits. It is important to note that this analysis was underpowered (as the sample size was based on the primary outcomes) and so was less likely to detect a difference should one exist. The results of the complete-case analysis are presented in Appendix 5, Table 80.

Data validity and completeness

This section reports on the completeness of the data used to estimate costs and effects.

Resource use and costs

The average total resource use based on staff time spent providing treatment and type of treatment was estimated for all three treatment strategies and is presented in Appendix 5, Table 74. A summary of types of treatment provided according to arm is also included [see Chapter 3, Treatment received (intention-to-treat analysis set, n = 1058)]. On average, all three treatment strategies were similar in terms of average total frequency and duration of visits. The three strategies differed in terms of operative treatment, with < 20% of all PA visits receiving operative treatment, compared with > 40% of B+P and C+P visits. The type of operative treatment also differed across the arms, which is reflective of the different treatment strategies (e.g. complete carious tissue removal was more routinely performed during C+P visits, whereas B+P visits with operative treatment were expected to involve partial or no carious tissue removal). GDPs provided the majority of prevention and operative treatment.

Costs were estimated for the first visit and for all follow-up visits during the time in the trial. The average total cost per randomised arm was estimated based on the three costing strategies.

Table 13 illustrates the difference in the total cost per child of treating dental decay in primary teeth, based on costs estimated using microcosting over a minimum of 23 months (maximum 36 months) in the trial. As would be expected, the total cost is lower for PA than for the other two treatment strategies, as children being managed by the PA arm had less operative treatment and shorter appointment times.

If we assumed a cost for the missing medications, mentioned in Data validity and completeness, it would have a trivial effect on the overall average total cost estimates for each treatment strategy and, given that B+P had the lowest average prescription costs, it would not affect the order of the three treatment strategies in terms of mean (or median) total cost.

The mean costs have a direct relationship to the total cost of adopting a given strategy. However, the median costs presented in Table 13 are arguably more representative of the typical total cost associated with managing dental decay in primary teeth over a minimum of 23 months and up to a maximum of 36 months. Typically, PA had more referrals and more referrals involving a general anaesthetic, which has inflated the average total cost of this treatment strategy. The assumptions we made previously to exclude costs when information was missing do not affect our results and would only strengthen the difference in costs between PA and the other treatment strategies if they were included.

Effectiveness

Summaries of incidence of dental pain and/or dental sepsis and episodes of dental pain and/or dental sepsis are provided in Table 15. The effectiveness outcomes were discounted at 3.5%, as per the National Institute for Health and Care Excellence guidelines. 132 The effectiveness point estimates used in the economic analysis are presented in Table 16, alongside the comparative results, which are described in the following section.

Cost-effectiveness results

Incremental analysis

Arguably, an incremental analysis is more appropriate from an economics perspective as it allows us to compare the three treatment strategies simultaneously. In this analysis, we first ordered the treatment strategies in terms of increasing average total cost and then compared a more costly strategy with a less costly strategy in terms of incremental cost-effectiveness. As PA is the least costly option,6 we compared this to B+P, the next least costly option, and estimated an ICER as B+P is, on average, more costly but more effective than PA. We then incorporated the most costly treatment strategy, C+P, into the analysis and compared this with B+P. As B+P was, on average, less costly and more effective than C+P, it dominated the comparison; hence, we would not consider C+P to be cost-effective compared with the other two treatment strategies. The ICER for B+P is approximately £300 to avoid one case of incidence and £130 to avoid one episode of dental pain and/or dental sepsis, compared with PA.

Again, the bootstrapped results present the uncertainty surrounding our results; Table 17 shows the probabilities of the three treatment strategies, compared with each other, being considered cost-effective at the different WTP thresholds. Figures 9 and 10 illustrate which treatment strategy has the highest probability of being considered cost-effective at each WTP threshold when the measure of effectiveness is, respectively, incidence of dental pain and/or sepsis avoided or episode of dental pain and/or sepsis avoided.

TABLE 17 - Cost-effectiveness analysisa for the comparison of PA vs. B+P vs. C+P based on microcosting (n = 1058)

Costs and effects are discounted at 3.5%.

Estimated based on adjusted analysis (n = 1057).

Investigation strategy	Cost (£) (97.5% CI)	Incremental cost (£) (97.5% CI)b	Effects (97.5% CI)	Incremental effects (97.5% CI)b	ICER (£)	Probability of each treatment strategy being considered cost-effective at different threshold values for society’s WTP to avoid dental pain and/or dental sepsis
						£0	£50	£100	£250	£500
Incremental cost per incidence of dental pain and/or dental sepsis avoided
PA (n = 354)	206.42 (176 to 237)		0.445 (0.39 to 0.50)			0.87	0.81	0.75	0.54	0.31
B+P (n = 352)	226.76 (201 to 252)	18.77 (–18 to 55)	0.390 (0.33 to 0.45)	–0.058 (–0.14 to 0.02)	323.62	0.12	0.17	0.22	0.36	0.58
C+P (n = 352)	244.65 (219 to 271)		0.410 (0.35 to 0.47)		Dominated by B+P	0.01	0.02	0.03	0.05	0.11
Incremental cost per episode of dental pain and/or dental sepsis avoided
PA (n = 354)	206.42 (176 to 237)		0.701 (0.58 to 0.82)			0.87	0.72	0.54	0.23	0.19
B+P (n = 352)	226.76 (201 to 252)	18.77 (–18 to 55)	0.565 (0.46 to 0.67)	–0.143 (–0.29 to 0.01)	131.26	0.12	0.25	0.41	0.64	0.72
C+P (n = 352)	244.65 (219 to 271)		0.603 (0.49 to 0.71)		Dominated by B+P	0.01	0.03	0.05	0.13	0.09

FIGURE 9.

Probability of being cost-effective at different WTP thresholds to avoid an incidence of dental pain and/or dental sepsis: PA vs. B+P vs. C+P. C+P does not appear in the figure because C+P did not have the highest probability of being considered cost-effective at the WTP values presented.

FIGURE 10.

Probability of being cost-effective at different WTP thresholds to avoid an episode of dental pain and/or dental sepsis: PA vs. B+P vs. C+P. C+P does not appear in the figure because C+P did not have the highest probability of being considered cost-effective at the WTP values presented.

The stochastic analysis shows that PA would be the most cost-effective treatment if a decision were to be based on cost alone. As society’s WTP for avoidance of dental pain and/or dental sepsis increases, it becomes more probable that B+P would be considered cost-effective (see Figures 9 and 10, respectively). C+P would not be considered cost-effective compared with PA and B+P; hence, it is not presented in Figures 9 and 10.

The complete-case analysis, as described in Complete-case analysis, has similar results, in that B+P would be considered cost-effective at reducing the number of episodes of dental pain and/or dental sepsis at a WTP threshold of £130 per episode avoided (see Appendix 5, Table 80).

Sensitivity analysis

Fee for service analysis

Data on 287 children treated in Scotland were used in the FFS analyses (C+P = 90, B+P = 100 and PA = 97). The average total cost per treatment arm was higher in the FFS analysis than in the analysis in which costs were based on microcosting (see Table 16). Estimates of mean effect remain similar to the microcosting-based estimates, except that now PA has fewer incidences of dental pain and/or dental sepsis than C+P (see Appendix 5, Table 76). This change in direction of effects can be explained by the smaller number of children included in the FFS analysis than in the base-case analysis (257 vs. 1058) and suggests that children in the PA arm based in Scotland are less likely to experience dental pain and/or dental sepsis.

When comparing the full incremental analysis based on FFS costs (Table 18) with that based on the microcosting (see Table 17), the conclusions drawn are similar. C+P is unlikely to be considered cost-effective regardless of the basis of costs. Regarding incremental cost per episode of dental pain and/or dental sepsis avoided, C+P was extendedly dominated by B+P. Extended dominance occurs when the ICER for the management strategy is higher than that of the next, more effective, alternative. 134 B+P would probably not be considered cost-effective at reducing incidence of dental pain and/or dental sepsis compared with PA with an ICER of > £1000. However, the incremental cost to avoid an episode of dental pain and/or dental sepsis was £235.

TABLE 18 - Cost-effectiveness analysisa for the comparison of PA vs. B+P vs. C+P based on FSS costs (n = 287)

Costs and effects are discounted at 3.5%.

Estimated based on adjusted analysis (n = 287).

ICER is estimated as B+P vs. PA.

Investigation strategy	Cost (£) (97.5% CI)	Incremental cost (£) (97.5% CI)b	Effects (97.5% CI)	Incremental effects (97.5% CI)b	ICER (£)	Probability of each treatment strategy being considered cost-effective at different threshold values for society’s WTP to avoid dental pain and/or dental sepsis
						£0	£50	£100	£250	£500
Incremental cost per incidence of dental pain and/or dental sepsis avoided
PA (n = 97)	321.46 (282 to 360)		0.423 (0.31 to 0.54)			0.98	0.98	0.97	0.91	0.69
C+P (n = 90)	359.28 (310 to 408)	34.89 (–6 to 76)	0.442 (0.33 to 0.56)	0.016 (–0.14 to 0.17)	Dominated by PA	0.02	0.02	0.03	0.04	0.22
B+P (n = 100)	374.54 (324 to 425)	52.84 (12.79 to 93)	0.37 (0.26 to 0.48)	–0.052 (–0.20 to 0.10)	1016.15c	0.00	0.00	0.00	0.05	0.09
Incremental cost per episode of dental pain and/or dental sepsis avoided
PA (n = 97)	321.46 (282 to 360)		0.706 (0.47 to 0.94)			0.98	0.94	0.81	0.38	0.13
C+P (n = 90)	359.28 (310 to 408)	34.89 (–6 to 76)	0.586 (0.40 to 0.77)	–0.126 (–0.39 to 0.14)	276.90	0.02	0.05	0.13	0.26	0.25
B+P (n = 100)	374.54 (324 to 425)	17.95 (–59 to 23)	0.478 (0.31 to 0.64)	–0.097 (–0.36 to 0.17)	185.05	0.00	0.01	0.06	0.36	0.62
B+P vs. PA		52.84 (12.79 to 93)		–0.223 (–0.49 to 0.04)	236.95	0.00	0.01	0.08	0.50	0.82

Aims and objectives

To explore the acceptability of the three treatment strategies to manage carious lesions from the perspective of the child participants and their parents/guardians.

The objectives were to:

• explore child participants’ and parents’/guardians’ experiences of the three treatment strategies and compare and contrast the impact of the treatment strategies on their daily lives

• identify factors of importance when considering the acceptability of the three treatment strategies.

Methods

Results

Data saturation was reached after 13 interviews had been conducted across Scotland and Yorkshire between January 2017 and November 2017. The 13 child participants were aged from 5 to 11 years and comprised eight girls and five boys; four of the participants were being managed in the C+P arm, five in the B+P arm and four in the PA arm. These children had been treated by six dental teams and treatment deviation forms had been completed for four of them. Three of the 13 parents were male (see Appendix 6, Table 81).

Aim and objectives

The aim was to explore the three treatment strategies with respect to the experiences of dentists and dental practice team members (hereafter collectively referred to as ‘dental professionals’).

The objectives were to explore:

• the experiences of dental professionals providing the three treatment strategies and whether or not their past use had an impact

• dental professionals’ perspectives on children’s and parent/guardians’ preferences between the three treatment strategies

• how dental professionals’ experiences of the trial might shape future management of children with carious lesions in primary teeth and identify any associated training needs.

Methods

Results

The FiCTION trial dental practice staff from the Scotland, Newcastle, Leeds/Sheffield and London CCs were interviewed, either face to face or by telephone. Overall, 31 dental professionals were interviewed between August 2016 and March 2017: 22 dentists (four community-based and 18 GDPs), two dental therapists, four nurses and three practice managers (see Appendix 6, Table 82).

Trial findings

The FiCTION RCT compared three treatment approaches to the management of carious lesions involving dentine in the primary teeth of children aged 3–7 years and found that over a median follow-up period of 33.8 (IQR 23.8–36.7) months, there was no evidence of a difference between the arms for the co-primary outcomes of (1) the proportion of children with at least one episode (incidence) and (2) the number of episodes of dental pain or dental sepsis or both.

Overall, 43% of the 1058 children in the ITT analysis set experienced at least one episode of dental pain and/or dental sepsis over a 3-year period (dental pain ever, 36%; dental sepsis ever, 25%).

For incidence, there was no evidence of a difference between arms. Comparing B+P arm children with those in the C+P arm, the risk difference was –0.02 (97.5% CI –0.10 to 0.06; 2% less pain/sepsis for B+P arm); comparing PA arm children with those in the C+P arm, the risk difference was 0.04 (97.5% CI –0.04 to 0.12; 4% less pain/sepsis for C+P arm).

For number of episodes, there was no evidence of a difference between arms. Comparing the B+P and C+P arms, the IRR was 0.95 (97.5% CI 0.75 to 1.21) (slightly fewer for the B+P arm); comparing the PA and C+P arms, the IRR was 1.18 (97.5% CI 0.94 to 1.48) (slightly more for the PA arm).

A similar pattern, but with less marked observed differences, was noted in the pre-planned PP analysis, which excluded from the analysis children with treatment deviations on > 20% of visits.

Child oral health-related quality of life remained high (median P-CPQ-16 score was 5–7) throughout the trial and MCDASf had a median score of 14–15, representing low to moderate dental anxiety. There was no evidence of a difference in the child OHRQoL or dental anxiety measures between treatment arms at the end of the trial and no evidence that they changed over the period of the trial. However, parent-reported child anticipatory anxiety was, on average, 6% lower in the PA arm than in the C+P arm (risk difference –0.06, 97.5% CI –0.11 to –0.003).

Caries development or progression throughout the trial was considered for teeth that were entirely sound or had non-cavitated carious lesions restricted to enamel at baseline, hereafter referred to as ‘sound/reversible’. This analysis was on a subset of the ITT analysis set (61.7% of the 1058 ITT participants included) and so must be treated with caution. Overall, 399 out of 653 (61.1%) of the participants with complete ICDAS data exhibited development or progression of a carious lesion in one or more primary teeth (C+P 57.9%, B+P 61.6% and PA 64.1%). There was no statistical evidence of a difference in carious lesion development/progression in primary teeth. Comparing the B+P arm with the C+P arm, the risk difference was 0.03 (97.5% CI –0.06 to 0.11). Comparing the PA arm with the C+P arm, the risk difference was, on average, 5% higher in the PA arm: risk difference 0.05 (97.5% CI –0.03 to 0.14). Of the 399 participants with development/progression of caries, 69% had one (42%) or two (28%) teeth that were ‘sound/reversible’ at baseline, but in which dental caries developed or progressed [mean 1.3 (SD 1.4) teeth per child], and the distribution between arms was similar. There was no statistical evidence of a difference in carious lesion development/progression in FPMs between the three arms.

Findings from the economic evaluation

In terms of cost-effectiveness, the PA arm was, on average, the least costly treatment but the least effective for both of the co-primary outcomes. This means that if society is not willing to pay to avoid dental pain and/or dental sepsis, then PA would have the highest probability of being considered cost-effective compared with B+P and C+P for both of the co-primary outcomes. The other two treatment strategies would, on average, provide more benefits, albeit at higher cost. A judgement is required as to what value the NHS places on avoiding dental pain and/or dental sepsis, as B+P is, on average, more costly and more effective than PA and dominates C+P. Should the WTP to avoid an episode of dental pain and/or dental sepsis be ≥ £130, then B+P would have the highest probability (49%) of being considered cost-effective compared with PA (45%) and C+P (6%). As this WTP threshold increases, so does the probability of B+P being considered cost-effective. A WTP threshold of approximately £330 per incidence of dental pain and/or dental sepsis avoided would be needed for B+P to be considered the most cost-effective treatment strategy to avoid an incidence of dental pain and/or dental sepsis.

Fee-for-service and UDA analyses represent the current methods of reimbursement for NHS treatments in Scotland, England and Wales. Three variants of the analysis were completed: one using FFS data only, one using UDA data only and a combined analysis in which FFS and UDA data were both used (an individual participant would have either a FFS cost or a UDA cost but not both).

For these analyses, the average total cost per arm was higher than the corresponding average total costs that were estimated in the microcosting analysis. This was to be expected, as the FFS and UDA analyses account for additional overheads and capital costs that were excluded from the microcosting analysis. The impact of this is that the difference in average total costs between B+P and C+P was smaller in the FFS and UDA analyses than in the base-case analysis (which was based on the microcosting). This was expected for the UDA analysis, given that the operative treatments provided in both the B+P and the C+P arms are reimbursed at a rate of 3 UDAs per treatment episode. In the FFS analyses C+P was, on average, less costly than B+P because the reimbursement of a filling (used in C+P) is less than the reimbursement of a crown (used in B+P). Overall, analyses based on FFS and UDAs result in similar conclusions being drawn: PA has the highest probability of being considered cost-effective if we are not willing to pay to avoid an episode or an incidence of dental pain and/or dental sepsis. However, unlike the base-case analyses (which used the microcosting data), PA continues to have the highest probability of being considered cost-effective compared with B+P and C+P at higher WTP thresholds to avoid an episode or incidence of dental pain and/or dental sepsis. For example, at a WTP threshold of £130 to avoid an episode of dental pain and/or dental sepsis, PA has the highest probability of being considered cost-effective compared with B+P and C+P in the FFS (72% vs. 17% vs. 11%, respectively), UDA (99% vs. < 1% vs. < 1%, respectively) and FFS and UDA (99% vs. < 1% vs. < 1%, respectively) analyses. At a WTP threshold of £330 to avoid an incidence of dental pain and/or dental sepsis, PA has the highest probability of being considered cost-effective compared with B+P and C+P in the FFS (84% vs. 6% vs. 10%, respectively), UDA (97% vs. 2% vs. 1%, respectively) and FFS and UDA (99% vs. < 1% vs. < 1%, respectively) analyses. This could lead to incentive incompatibility in that practices could be driven to provide a treatment (i.e. PA) that, when costs are considered rather than fees, might be viewed as inefficient. Further work is needed to understand the implications of this and how dental practices might change practice as reimbursement policies change.

Findings from the qualitative study

The qualitative interviews with child–parent dyads indicated that each treatment arm was felt to be generally acceptable to children and parents, but trust in the dental professional played a significant role. Certain procedures, including local anaesthetic and extractions, were more likely to be viewed negatively. Other factors identified by children and parents were anticipatory anxiety, perceptions of effectiveness and the impact on home or work life from attending dental appointments. Children and parents had similar perspectives on most aspects of the management of carious tooth tissues, except with respect to PMCs: some parents were concerned about the aesthetics of PMCs, but children did not share this view.

The qualitative interviews with dental professionals illustrated how managing carious lesions was a recognised activity, involving the competence of selecting the most appropriate treatment option to act in the best interests of the child patient. Being a dental professional involves drawing on experiential and interpersonal knowledge, as well as research-based evidence. Participants indicated that parents/guardians shared this understanding, trusting dental professionals to treat each child in the ‘best’ way possible. This understanding of dentistry and the activity of managing carious lesions meant that some dental professionals deviated from the allocated trial arm to treat a particular patient in a way other than that to which the child had been randomised. Such treatment deviations indicate the importance to dental professionals of being able to select the most appropriate treatment option; when they continued with a treatment that they felt was less than ideal, they described this as a ‘difficult’ and an ‘uncomfortable’ experience. The interviews with dental professionals also revealed that treatment options were assessed in terms of perceived effectiveness. Although delivering local anaesthetic, removing carious tissue and filling tooth cavities were understood as elements constituting the ‘conventional’ activity of treating carious lesions, dental professionals familiar with minimally invasive biological methods of sealing-in decay spoke about this as more effective, particularly in terms of the negative connotations of local anaesthetic with the traditional ‘drill-and-fill’ approach. Nonetheless, it was recognised that, as an unfamiliar activity, the materials involved in the biological method could be ‘strange’ and dental professionals needed to develop the competencies of selecting the right crown and judging when to use a crown and when to use adhesive restorative material. Providing preventative treatment was viewed positively (and was also ‘routine’), but this was often seen as an activity that needed to be combined with other treatments, otherwise it could be ‘insufficient’.

Dental professionals reflected on how parents and children viewed different treatment options: fillings were ‘expected’, but local anaesthetic could be problematic for many children; crowns were initially viewed with suspicion but understood as beneficial when the process was fully explained; and providing only preventative treatment was often questioned and seen as ‘doing nothing’. Nevertheless, dental professionals reported that parents had other concerns, particularly around the aesthetic aspects of different treatment options. Children were also seen to prefer no intervention if possible; therefore, providing preventative treatment could have positive meanings if this involved avoiding unpleasant experiences. Participants also commented on how they planned to engage in different activities of treating carious lesions in the future. The data revealed some movement towards minimally invasive biological treatments and away from conventional methods; this was also reflected in the numbers and directions of the reported treatment deviations. Nevertheless, it was recognised that it might be difficult to change one’s habitual way of treating carious lesions. It was also clear that dental professionals expected to continue to use a range of different treatment options to manage carious lesions. Although some dental professionals spoke about accepting the results of the trial (should one management strategy be proven to be the best), it was also suggested that dental professionals should continue to have a choice in how they treat patients. The qualitative data indicate that, although the results of the FiCTION trial will form part of their knowledge base, dental professionals will also continue to draw on what they know about an individual patient and their own clinical experience to manage carious lesions.

Overview

The size of the FiCTION trial is a huge strength, with 1144 children recruited across 72 general dental practices in three countries; therefore, the trial is thoroughly embedded in the environment where this research question is most pertinent. The trial has collected a vast number of important data from different perspectives (dental professionals and their teams, children and parents) and using different methods (quantitative and qualitative). To our knowledge, no other RCT of this size has been undertaken in this area of dental care and no RCTs have taken the pragmatic approach adopted in the FiCTION trial, with a child (as opposed to a tooth) forming the basis of the assessment level. Efforts were made to recruit a random and representative sample of practices and to also recruit dentists in clinical equipoise in each of our chosen areas. 86

The pragmatic approach taken, observing what dental professionals did for patients in each of the arms when requested to follow carious lesion management protocols, is akin to establishing what might happen if guidance or policy were put in place to direct clinical practice towards managing carious lesions in primary teeth using one particular approach.

Strengths

The sample of children who participated in the RCT was balanced across groups. At baseline, there was balance between arms in terms of age [mean age 5.96 (SD 1.3) years], sex (51% female), ethnicity (76% white) and d₃mft [2.72 (SD 2.66)], including the amount of untreated decay (d₃) [2.04 (SD 2.15)]. There was also balance across the arms in respect of fluoridation status and index of multiple deprivation, although these variables were captured only at dental practice level. Overall, 67% of the ITT participants attended a final visit and the median follow-up period was 33.8 (IQR 23.8–36.7) months, also balanced across the three arms, with no evidence of differential attrition across the arms. This balance confirms our confidence in the randomisation process, and the baseline participant characteristics demonstrate the representativeness of the participants as typical high-risk child dental patients attending dental practices with untreated decay. This aspect of the trial is discussed further in Generalisability.

Although the rate of recruitment was relatively slow and a time-only extension was required, the FiCTION trial exceeded its revised target sample size of 1113 participants, and 1058 out of 1144 (92%) children randomised attended at least one trial visit and were included in the ITT analysis set. Allowing for 25% attrition, the revised effective sample size was to retain 834 children who were followed up to their maximum potential of between 23 and 36 months. The FiCTION trial was able to follow up 797 out of 1144 (70%) participants for over 23 months and maximised the utility of the data collected by including in the analyses all data available for each of the 1058 children in the ITT analysis set, regardless of time in the study. The trial was adequately powered to detect the clinically meaningful differences between the arms in the incidence of dental pain and/or dental sepsis, hypothesised at the design stage. However, the estimated level of dental pain and/or dental sepsis was higher than had been anticipated at the design stage; the consequence of this was that the associated CIs were also wider.

A further strength of the trial was the simultaneous evaluation of clinical effectiveness and cost-effectiveness of the three treatment approaches. The economic evaluation was based on extensive data collection undertaken in the RCT. In particular, the different costing approaches adopted allowed us to estimate the cost-effectiveness using both the ‘true’ cost and the NHS reimbursement rates for each of the treatments. The economic results allow us to draw out the implications of the trial findings on policy and practice.

The inclusion of the qualitative study component also added much value to the trial, in addition to providing data for two secondary objectives: first, an understanding of the acceptability and associated experiences of the three treatment strategies for child participants and parents/guardians; and, second, an understanding of dentists’ and dental practice team members’ preferences between the three treatment strategies. The latter, combined with information from the CRFs and TDFs, provided a detailed insight into GDPs’ ability and willingness to deliver the three treatment strategies. By explaining patients’ and health-care professionals’ perceptions of the interventions, the qualitative study also helped the interpretation of findings,151 while also providing some insight into how the three treatment strategies were delivered in practice.

In the qualitative component, we aimed to follow best practice. In research with children, efforts are required to minimise the power imbalance that exists between the researcher and the child. During the interviews with children, efforts were made to reduce the power imbalance by ensuring that interviews took place where children were comfortable, using participatory activities and emphasising that children could stop the interview at any time. 109 The study was designed with input from the multidisciplinary Trial Management Group; the interviews were conducted by experienced qualitative interviewers, including two research associates (non-dental) and one clinical academic. The interviewers were experienced at assessing signs of distress or boredom in participants, but none was observed. Most of the interviews were conducted in children’s own homes, with steps taken to ensure interviewer safety. The analysis was conducted by several members of the team from different disciplines, with two researchers (one dentally qualified) independently identifying initial themes before discussing these with the wider team. The qualitative work with dental professionals was designed by the Trial Management Group with further support from a sociologist experienced in the use of practice theory. The decision to use practice theory to guide the analysis was driven by the data themselves. The interviews were conducted by three experienced qualitative researchers, two of whom were clinical dental academics, to enable the more technical aspects of the dental treatment to be explored while also ensuring that the broader sociocultural and economic aspects were not ignored. The analysis was initially conducted independently by two members of the team (one dentally qualified), with ongoing discussions with the Trial Management Group and undertaken before the results of the primary and secondary outcomes were revealed to the lead for the qualitative component.

Limitations

The trial experienced challenges throughout its course, some of which were anticipated and others that could not have been predicted. These resulted in some limitations, as discussed below and in Methodological rigour.

Recruitment and retention of participants in RCTs is recognised as a significant challenge,152,153 especially among young children and in a relatively research-naive environment such as NHS primary dental care services. The slower than predicted recruitment rate affected the length of time a practice was involved in the trial; some practices were involved for up to 5 years in the RCT if they were practices that had recruited participants at the beginning of the recruitment period and had continued to recruit participants throughout the whole duration of the trial. The trial length also had an impact on the FiCTION trial practice training, which sometimes required repetition or supplementation, in view of staff turnover, particularly in long-serving and vocational training practices. In addition, some secondary outcomes were measured only at baseline and the scheduled final visit and, as 33% of ITT participants did not attend their scheduled final visit, the opportunity to capture these data for those individuals was missed. From some of the qualitative interviews, as well as anecdotally, it was also clear that increasing numbers of practices and participants experienced study fatigue as the trial approached the final visits period, requiring greater levels of motivational input from the research support staff and clinical leads.

As noted above, fluoridation status and deprivation were based on practice address, not patient address, and may well have masked some intrapractice variation (especially in deprivation). Given that the study was patient randomised, it would have been preferable, although more resource intensive, to have been able to calculate these variables at the participant level.

There were several limitations in the economic evaluation. First, there were a number of challenges associated with implementing the different costing methods. In particular, the FFS is a more complex system with reimbursements based on different criteria; for example, an extraction rate was based on the number of teeth extracted in one course of treatment and a per-visit extraction rate was also available. However, the detailed costing undertaken allowed us to cope with the complexity of both reimbursement systems and provided us with robust estimates of the current charges to the NHS to manage dental decay in primary teeth.

Second, an inflated UDA value was incorporated for urgent band 1 visits, which may have led to an overestimation of UDAs for that course of treatment. In the UDA analysis, 3% of courses of treatment (n = 133) were classified as urgent visits, although such visits were evenly distributed across the three treatment strategies. In addition, for UDAs, discounting was applied to costs that occurred after the first year of follow-up based on when a course of treatment started, not when it ended. This assumption may not be reflective of current practice, given that practices are reimbursed retrospectively; however, given that resources used are incurred at the start of the course of treatment, we deemed this to be a sensible assumption. Both of these assumptions potentially give an overinflation of costs that may not be true. However, the effect is small overall and is balanced across the three treatment strategies.

Third, the SUR model used in the adjusted analysis may not be an appropriate fit for the co-primary outcomes. In the primary adjusted analyses, a logit model was adopted for incidence and a negative binomial model was fitted for the number of episodes. Although these models are arguably more appropriate for estimating these outcomes, there is a trade-off between fitting the most appropriate model and applying a model that allows for the correlation of costs and outcomes, which is arguably more appropriate for the economic analysis.

Finally, capital costs were excluded from the analysis as all three management strategies are being provided as part of current care; therefore, these costs would have been incurred regardless of which strategy was considered. Excluding capital costs reduces the total cost of each arm equally; hence the incremental costs and ICER would remain unchanged.

There were several limitations of the qualitative study. First, it was not possible to embed the qualitative component throughout the main trial, with interviews conducted only once recruitment had been completed. Second, only the perspectives of those parents and children who were retained in the trial were gained; it was not possible to interview children or parents who had withdrawn from the trial. In addition, during the interviews with children and parents, it was difficult to distinguish between procedures children had received during the trial from those before or since, especially if the child had subsequent treatment to their permanent teeth, although this is unlikely to alter the findings about acceptability. Finally, although a range of dental professionals were interviewed, it was not possible to interview those who had withdrawn from the study.

Generalisability

To our knowledge, the FiCTION trial is the first multicentre three-arm, parallel-group participant-RCT to evaluate three clinical management approaches for carious primary teeth at high risk of decay, in young children, in NHS primary dental care services, the environment where the vast majority of dental care for children takes place. As a pragmatic trial in NHS primary dental care it reflects, as closely as possible, the real clinical practice and issues encountered in the provision of dental care services for children in this environment, including the fact that the inclusion criteria were broad, thereby minimising exclusion, to try to ensure full representation of this group of children. However, the restrictions that were imposed for inclusion meant that children did not join the trial from a very early stage before they had developed a carious lesion. The age criterion for the trial was 3–7 years but the median age of the recruited children was 6 years, at the older end of the planned age range, and, therefore, arguably more compliant with treatment. However, the requirement for children to have at least one primary molar with dental caries into dentine and knowing that most children were screened for carious lesions on a visual basis alone meant that these FiCTION trial participants had advanced caries, which meant that they may or may not have been amenable to avoiding dental pain and/or sepsis. In terms of generalisability of the trial to the reality of general dental practice, management of carious lesions in a trial encourages continuity of care, which is not always observed in primary dental care.

The demographic and clinical characteristics of the practices and participants demonstrate a wide representation of the characteristics of UK primary care dentistry, including the different contractual and funding systems currently operating. Hence, we believe that this study provides generalisable findings for regularly attending child patients at high risk of decay and attending with decay in their primary teeth.

A total of 72 practices participated in the study, which were linked to five centres in Scotland, England and Wales. The large number of practices and diversity of location, including urban and rural areas, strengthened the trial not only in terms of being able to achieve the required sample size but also in terms of its generalisability to UK dental practice.

The selection of the area of north-east England around Newcastle upon Tyne as one of the five study centres was made as the city sits in a water fluoridation area. Based on the fluoridation status of the practice, the proportion of randomised participants [114/1144 (10%)] associated with a practice receiving optimally fluoridated water across the trial was similar to the proportion of the UK population receiving a fluoridated water supply (12%). 154

The RCT was also designed to include regions of the UK where children from ethnic minority populations could be recruited, as it was considered important to understand any cross-cultural differences between treatment approaches. The potential for such differences was anticipated to relate to cultural, behavioural and acceptability aspects of dental treatment, rather than to the actual clinical management of dental decay, which would be expected to be similar for all children in a given practice. The non-white population of the UK is 8.17 million (12.9% of the overall UK population),155 and 24% of the FiCTION trial children were non-white. Major efforts were made to engage with and recruit from these populations as there is evidence to suggest that families from ethnic minority groups in populations researched internationally are less likely to express an interest in or be recruited to RCTs. This is an ongoing problem in terms of research evidence, in view of the disparity in caries experience seen between the larger ethnic minorities and majorities in the UK;13,156 further research is needed to ascertain the main reasons for the difficulties seen in recruiting ethnic minority participants, including children, to RCTs. Achieving 24% of children in the FiCTION trial who were identified as non-white was an additional strength of the trial.

The overall mean caries experience in the primary dentition for all FiCTION trial participants at baseline was 2.72 (SD 2.66) d₃mft. This is lower than the caries experience found in 5-year-old children with experience of dental decay (i.e. d₃mft > 0) in Scotland, for whom the mean d₃mft in 2016 was 3.93;156 in England, for whom the mean d₃mft in 2015 was 3.4;13 and in Wales, for whom the mean d₃mft in 2015/16 was 3.58. 115 It should be noted, however, that the national surveys are of randomly selected children recruited in the school setting and not from those attending general dental practices and are therefore likely to include those with the poorest oral health, who are not engaged with regular dental care.

In summary, we believe that the practices recruited to the FiCTION trial represented practices delivering primary dental care services in the NHS in Scotland, England and Wales in terms of their size, location (in terms of social deprivation), skill mix and patient mix and were therefore generalisable to the UK. With regard to participants recruited to the FiCTION trial, we believe that they were representative of UK 3- to 7-year-old regular attenders who are at high risk of dental caries, in terms of their age, caries experience, water fluoridation status and ethnicity.

Choice of primary and secondary outcomes

Failure to prevent or treat dental decay ultimately results in dental pain and/or dental sepsis with a substantial impact on individuals, families and NHS resources. Pain and sepsis are therefore appropriate outcomes to measure when comparing the impact of decay management approaches, and were the two primary outcome measures requested in the HTA commissioning brief. The use of co-primary outcomes of the incidence of dental pain and/or dental sepsis in the child (yes/no) and the number of episodes of dental pain and/or dental sepsis in the child (an interval measure) was the optimal way of measuring these most important outcomes, although challenging in terms of the data collection and its analysis.

At the trial design stage, these child- or mouth-based measures were deliberately chosen to focus on the primary mouth-based consequences arising from failure to prevent, arrest or manage decay. The addition of the co-primary outcome of number of episodes of dental pain and/or dental sepsis to the originally planned measure of incidence alone provided a refined combination of measures that strengthened the analysis and also provided further insight into the child participant experiences. Although 43% of children experienced dental pain and or dental sepsis, 67% of these children experienced it only once during the FiCTION trial. We believe that the wording of the pain question may have contributed to the high levels of ‘dental pain ever’ recorded (an overall incidence of 36%, with no evidence of differences between arms), as it asked about ‘a history of pain at this visit’ rather than pain since the last visit, leading to a probable inflation of the instances of positive recordings.

The use of dental sepsis as an outcome measure provided the necessary objectivity, which was less evident with the measure of dental pain. It was the prevalence of dental sepsis seen in other studies that provided the basis for the sample size and power calculation for the trial. However, although pain was a less objective outcome measure, being participant/parent observed (and reported to the dentist), it was the measure that is important to patients.

The secondary outcomes included in the trial were diverse, but all were important contributors to the overall assessment of the clinical, economic, behavioural and psychological impact of dental management approaches to managing decay in children’s primary teeth. At baseline and at the end of follow-up, the use of the ICDAS caries index – an objective measure suitable for recording carious lesion prevalence and incidence – allowed the derivation of a binary indicator variable at a child level (progression of disease in one or more teeth: yes/no) and a measure of disease severity (number of teeth with caries development/progression). From the behavioural and psychological perspective, the child- and parent-reported outcomes measuring OHRQoL, anxiety/worry and dental discomfort were important to capture, particularly as the three treatment approaches differ substantially from one another in that one (C+P) requires the routine use of local anaesthetic when removing decay involving dentine. Parents tend not to recognise dental anxiety in their children very well: they usually under-report it, which is why child self-reported measures are preferred if possible;106 however, it was difficult to identify instruments appropriate and/or valid for the target age group. For example, the use of anticipatory and treatment-related anxiety and worry questions (MCDASf scales) at every visit aimed to distinguish between state and trait dental anxiety. However, in 2007, Howard and Freeman103 concluded that, although the MCDASf scale was a reliable measure of dental anxiety, demonstrating good reliability and validity, this had been demonstrated only with children aged 8–12 years. At baseline scoring, FiCTION trial participants were aged 3–7 years. Notwithstanding this, the pilot study gave some reassurance that the tools used were valid.

With regard to parental and caregiver perceptions, the P-CPQ-16 scale was used. In 2014, Thomson et al. 96 compared the responsiveness and functionality of the Early Childhood Oral Health Impact Scale (ECOHIS) and P-CPQ scales and found them similar and of good quality for internal consistency, reliability, validity and responsiveness. However, when used for health services research, when the child participants had significant amounts of decay as in the FiCTION trial, the P-CPQ did not have the shortcomings that ECOHIS157 demonstrated (it focuses minimally on any ‘oral symptoms’ domain), instead centring on general function and social and emotional well-being.

Financial costs for managing decay in the primary teeth of children in the UK are a substantial call on NHS resources. 5 It was important, therefore, that the FiCTION trial included appropriate clinical effectiveness and cost-effectiveness comparisons using valid, reliable measures. The costs of the interventions were assessed in terms of incremental cost, using data collected at an individual participant level using microcosting, as well as being based on charges to the NHS on an aggregate level using UDA and FFS values. These robust estimates of cost-effectiveness can be used in conjunction with the clinical effectiveness results to help inform policy decisions regarding the management of dental decay in primary teeth.

Blinding and other methodological issues

For ethics reasons, the trial design did not allow for an arm receiving no treatment; therefore, all three arms included the same active prevention element.

Dental professionals, parents and children could not be blinded to the intervention, as major features of the C+P and B+P arms were the use of local anaesthetic and the Hall Technique, respectively. However, they were all responsible for recording a number of the outcome variables, leaving the study open to an element of (detection) bias. The measurement of caries would have been less open to bias if undertaken by independent examiners. However, this would have been another very large resource implication in terms of time and cost given the number of participants, their locations and the 3-month final visit time period.

Although blinding of treatment arm was not possible, we tried to minimise other forms of bias. Of the 12,078 people sent a screening invitation, 4379 (36%) never attended, and, of the 1756 screened and known to be eligible, 612 (35%) did not participate. It is almost certain that biases were introduced at both of these stages, especially in respect of the 331 children who were eligible but declined the trial, but the impact of this cannot be assessed. Selection bias was minimised by using a central randomisation process operated by the NCTU, while detection bias was limited by including a simple observable reporting outcome (dental sepsis) as part of the composite primary outcome alongside the more subjective dentist- and participant-/parent-reported pain measures. We also tried to validate the reported pain by recording the tooth involved.

Dentists were asked to recruit eligible children when they attended their routine dental check-up appointments and, in view of the pragmatic nature of the trial, subsequent treatments and recall intervals varied as they do in real life. As a result, the number of data collection points varied not only with length of follow-up but also between participants and between arms. Overall, the median rate of visits was 2.5 per year in the trial, ranging from 2.1 for the PA arm to 2.5 for the other two arms. This is a lower rate of attendance than would be anticipated as, on the basis of the inclusion criteria for the trial, all children had at least one primary molar with dental caries involving dentine and, therefore, all were at high caries risk. This further indicates that, in the trial, GDPs tended to stick to a 6-monthly ‘check-up’ regime, rather than providing more frequent preventative interventions (e.g. application of fluoride varnish three or four times per year). The reason for this is unclear, but may relate to administrative or financial factors tending to dictate the frequency of recall intervals for children, or may be parent determined (e.g. not attending or arranging visits).

In addition, primarily for ethics reasons, the radiographs taken were only those that would be taken during normal clinical practice, that is those taken in line with FGDP guidelines;70 therefore, they were not a mandatory requirement for data collection. We had originally planned to use any available radiographs to assess the primary outcome of dental pain and/or dental sepsis as well as the caries incidence and progression outcomes. However, the level of adherence to guidelines recommending clinical radiography in children is known to be low. 158 The children entering the trial fell into the ‘high-risk’ category for dental caries because having a carious primary tooth was an inclusion criterion. Therefore, all children in the trial should, in principle, have had a radiograph taken within 1 year of entry to the trial (as they may have had one taken prior to entry) and then annually or more frequently until they moved out of the ‘high-risk’ category. However, in practice, we found that only 308 (29%) of the children had radiographs taken on entry to the trial (or within 1 year of entry). We decided that this rate of radiography was too low (and not representative enough of the children across the trial) to be able to use the radiograph data to supplement the clinical data. We therefore relied on assessment of the clinical data for the primary outcome measure of sepsis and/or pain due to caries. Approximately half of the carious lesions that exist in a child’s mouth, especially affecting proximal surfaces, are not detected by clinical visual examination alone;159 instead, detection relies on the use of further investigation through the use of bitewing radiographs.

Adherence to protocol

As reported in Chapter 3, Treatment provision and adherence to protocol, and Appendix 5, Table 74, we are confident that dentists and dental teams provided treatment for the children in line with the clinical protocols. Prevention was common to all arms and at least one item of prevention was delivered, on average, at 81% of all visits, balanced across arms. According to the trial protocol, recording of treatment deviations was required at every visit at which they occurred. Of the 7713 CRFs completed, a major treatment deviation recorded on a TDF occurred at 429 (6%) visits affecting 263 children, which indicated good fidelity with the protocol. The majority (46%) of deviations were away from the C+P arm and most of these were because of issues associated with local anaesthetic. The reported major deviations from clinical protocol resulted in 10% of the children from the ITT analysis set having > 20% of their treatment visits deviating from the clinical protocol. There was some evidence from the data collected via the CRF that additional protocol deviations had occurred around the failure to use local anaesthetic in the C+P arm, but, as the CRF had not been designed to specifically collect this information, the data around additional deviations lacked sufficient clarity for any meaningful analysis.

As discussed earlier, the clinical protocol for the FiCTION trial was developed in 2007 and was based on evidence-based guidelines available at that time. 25 These provided a degree of flexibility between adhesive intracoronal restorations and Hall Technique use and it is clear that some dentists chose to avoid the Hall Technique use in many situations where the protocol provided a legitimate alternative (ie. intracoronal adhesive restoration); in the B+P arm, the majority of restorations used GIC (11%), followed by PMCs placed using the Hall Technique (9%), and RMGIs (7%).

Implications for health care

• The results of the FiCTION trial emphasise the importance of preventing disease before it occurs to avoid dental pain and/or dental sepsis. Prevention must start early to avoid the need for operative management of disease.

• The FiCTION trial results indicate that, for successful management of carious lesions in young children, the health-care system may require adjustment to ensure effective detection and diagnosis of dental caries and the use of effective materials and techniques.

• The FiCTION trial’s findings open the debate around (1) where dental care for children at high risk of dental caries is best provided, (2) the most appropriate way of identifying children at high risk of dental caries and (3) the optimal clinical and funding environment for children at high risk of dental caries.

Implications for parents and practice

• Parental and dental professional roles are key to ensure that prevention starts early at home, and early attendance at a dental practice is an opportunity for reinforcement of prevention, monitoring of oral health, provision of effective advice and early intervention, if necessary.

• The FiCTION trial findings indicate that treatment of carious lesions in young children can be tailored to meet their behavioural and clinical needs and, therefore, the role of shared decision-making is important. Although there is no evidence of a difference in clinical outcome between arms, the biological approach may offer an advantage when taking into account the disease process, the child’s needs and the technical requirements, and is the strategy most likely to provide value for money if preventing dental pain/sepsis is valued.

Implications for dental education/teaching and training

• There appears to be a need for knowledge tools, key evidence-based guidance and additional clinical training and support of dental professionals to address their clinical skills with regard to use of radiographs and choice of restorative materials and techniques.

• The FiCTION trial has evidence of facilitators of and barriers to the investigation, detection, diagnosis and management of carious lesions, learnt through the qualitative study, which could inform implementation strategies for Public Health England and SDCEP guidance. Messages and techniques need to be clear, evidence based and tailored for delivery and uptake in the primary care environment.

The undergraduate and postgraduate curricula in all dental schools may need to be reviewed to ensure that all caries management strategies evidenced as suitable for use in young children are included, along with competencies in managing the dental care of young children and shared decision-making.