Risk-based, 6-monthly and 24-monthly dental check-ups for adults: the INTERVAL three-arm RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 06/35/99. The contractual start date was in September 2011. The draft report began editorial review in August 2019 and was accepted for publication in June 2020. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Clarkson et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2020 Queen’s Printer and Controller of HMSO

Introduction

The subsequent chapters of this monograph describe the INTERVAL (Investigation of NICE Technologies for Enabling Risk-Variable-Adjusted-Length) Dental Recalls Trial, a National Institute for Health Research (NIHR) Health Technology Assessment (HTA) programme-funded trial testing of the effectiveness and cost–benefit of dental check-ups at different recall intervals. The trial protocol has been published. 1

The reason for the trial

Epidemiology and pathogenesis

Periodontal disease is an inflammatory disease of the soft and hard tooth-supporting tissues. Periodontal diseases comprise gingivitis and periodontitis. Gingivitis is a reversible condition characterised by gingival redness and oedema, and absence of periodontal attachment loss. 21 The 2017 world workshop and classification system gives clear definition of a gingivitis site and a gingivitis case (patient). 21 A patient can be defined as a ‘gingivitis case’ when bleeding on probing at > 30% of sites is evident at a minimum of 10% of sites. 21 Bleeding at between 10% and 30% of sites is defined as localised gingivitis and bleeding on probing at > 30% of sites is generalised gingivitis. Gingival health is defined as bleeding on probing at < 10% of sites. Gingivitis is a pre requisite for periodontitis and is also a risk indicator for dental caries progression.

Periodontitis is the irreversible destruction of the tooth-supporting periodontal structures (periodontal ligament, cementum and alveolar bone) due to inflammation. 22 Periodontitis is characterised by periodontal pocket formation and gingival recession. In addition, tooth mobility and migration may occur as a result, as well as dentine hypersensitivity of the exposed root surface, root caries and, ultimately, tooth loss.

Gingivitis and periodontitis are a continuum of the same inflammatory disease process,23 with evidence that gingivitis is a risk factor for periodontitis,24 and that absence of gingival bleeding is a reliable predictor for the maintenance of periodontal health. 25 However, it is not currently possible to predict progression from gingivitis to periodontitis at either the individual or the site-specific level. Accumulation of microbial dental plaque is the primary aetiological factor for gingivitis and periodontitis, as well as dental caries. 26–28 Disease progression is also known to be affected by genetic factors (host defence mechanism), calculus, smoking and systemic comorbidities, including type 2 diabetes. 29–32

Despite the largely preventable nature of periodontal disease, it is considered the most common disease of mankind,33 remains the major cause of poor oral health globally and is the primary cause of tooth loss in older adults. 22,34 Global estimates of gingivitis range from 50% to 90% of populations. 35–37 Severe periodontitis is the sixth most prevalent human disease globally, with a prevalence of 11.2%,38 which appears to be rapidly increasing. 33

Dental caries is a multifactorial chronic oral disease that affects most populations globally and is considered the most important global oral health burden. 39 Dental caries results from production of organic acids by acidogenic bacteria within dental plaque (a biofilm formed on the tooth surface soon after tooth cleaning). Dental caries is considered a consequence of an ecological shift in the balance of the normally beneficial oral microbiota driven by a change in lifestyle and in the oral environment. 40 Organic acids can cause mineral loss from the tooth surface by removing calcium and phosphate ions from surface apatite crystals (demineralisation). In favourable conditions, a reversal of this process is possible (remineralisation). The development of a carious lesion is a dynamic process that may progress, halt or reverse. Progression of the carious lesion occurs where the demineralisation process prevails over remineralisation. Carious lesions can range from early non-detectable mineral loss restricted to enamel, through lesions that extend into dentine without any surface cavitations, to cavitated lesions visible as holes in the teeth. Progression rates of carious lesions appear to be more rapid in dentine than in enamel, with variable rates between individuals as well as between lesions within an individual. 41,42 Dental caries and its consequences are considered the most important burden of oral health, affecting up to 100% of adults in most countries. 43 It is not just a disease of children, but appears to occur at a relatively constant rate throughout the life course. 44

Where gingival recession has migrated apical to the amelocemental junction, the exposed root surface of the tooth may be susceptible to root caries. Like coronal caries, the main aetiological factor for the initiation and progression of root caries is the presence of a cariogenic biofilm and fermentable carbohydrates. Owing to the lower level of mineralisation of dentine, a smaller decrease in pH will induce demineralisation of the root surface. 45 As with coronal caries, the formation of root caries is a dynamic process of demineralisation and remineralisation, with progression of caries occurring where the balance of factors favours demineralisation. 46 Importantly, and unlike caries in enamel, coronal dentine and root caries both involve not only demineralisation but also collagen degradation,47 resulting in a demineralisation process that is approximately twice as rapid on enamel. 48

Root caries, like other forms of the disease, can be associated with pain, discomfort and tooth loss,49,50 which has the most significant impact on the oral health-related quality of life (OHRQoL) of the elderly. 51,52 Although a well-recognised disease, its prevalence is increasing as populations age and retain more of their natural teeth into older life. 47,53,54 There is a wide range in the reported global prevalence of root caries for diverse populations ranging from 29% to 89%. 55

Individuals and dental care professionals have different roles to play in the prevention and control of periodontal diseases and dental caries. Consistent removal of the intraoral plaque biofilm by means of personal tooth brushing and interdental cleaning is considered the foundation of successful primary prevention of periodontal disease and dental caries. 56,57 The dental care professionals’ role in primary prevention involves assessment of an individual’s risk of developing oral disease and tailoring preventative advice, including oral hygiene and dietary advice based on this risk assessment, although the evidence relating to the beneficial effects of chairside provision of dental health education advice is conflicting. 58,59

In 2004, the National Institute for Health and Care Excellence (NICE) published a guideline entitled Dental Recall: Recall Interval Between Routine Dental Examinations,60 following a remit received from the Department of Health and Social Care and the Welsh Assembly Government. Within this guideline, the role of the oral health review, or dental check-up, in providing primary prevention and secondary prevention is highlighted. Subsequently, in 2011, the Scottish Dental Clinical Effectiveness Programme (SDCEP) published their guidance on oral health assessment and review. The SDCEP guidance was based on and was a tool to assist implementation of the NICE recommendations. 61

Given that the global burden of oral disease is not shared evenly, the risk of developing oral disease between patients is clearly variable. It has, therefore, been suggested that the preventative needs of patients are also variable and that intervals between oral health reviews should be appropriate for the needs of individual patients.

Dental check-ups at 6-month intervals have been customary in the general dental service (GDS) in the UK since the inception of the NHS. Although a recall interval of 6 months is not explicitly recommended by the NHS, this practice is implicitly recognised by NHS regulations that have remunerated dental practitioners for providing a dental check-up at 6-month intervals for decades, and, since 2006, the dental check-up has been free in Scotland. It has been argued that a 6-month dental recall policy is too rigid and that recall intervals should match the individual needs of patients more closely – needs that may change over time. Analysis of dental attendance patterns in NHS primary dental care using the Dental Practice Board’s longitudinal data demonstrates an attendance pattern that is variable, with many patients attending less frequently than every 6 months. 59 Data from the Information Services Division (ISD) in Scotland show that the vast majority of Scottish adults did not attend NHS primary care dental services on an annual basis, with only 23% attending at least once per year in each of the previous 6 years, and 21% not attending at all in the past 6 years. 62

In addition, it has been consistently observed that caries experience is generally more extensive in lower socioeconomic status groups,63 reinforcing the case for patient-specific recall intervals, based on an assessment of the patient’s risk of oral disease. 3,60,64 Evidence from the Dutch health system suggests that there is an increase in general dentists moving away from recalling all patients at the same interval in favour of applying an individualised recall interval, resulting in more frequent screening for periodontal disease than with those dentists using a fixed-period recall protocol. 65

One of the persistent arguments in favour of maintaining 6-month dental check-ups is that dentists may miss the opportunity to diagnose oral cancer lesions at an early stage in patients who attend at longer recall intervals. The incidence of oral cavity cancer in the UK is highest in Scotland, at 10.0 per 100,000 males,66 and has been relatively stable in Scotland since 2000 but rising in England and Wales. However, it has been reported that 53.7% of patients diagnosed with oral cancer had not attended an NHS primary care dentist in the 2 years preceding diagnosis,67 thus radically decreasing the opportunity for early detection. From these data it is estimated that a dentist potentially encounters one case of oral cancer every 10 years. 67 Brocklehurst and Speight68 reviewed the pros and cons of a national screening programme for mouth cancer, concluding that studies into mouth cancer screening have provided evidence to satisfy only 5 of the 20 criteria required by the UK Screening Committee, with no evidence of a test effective in the detection of oral lesions in the context of a screening programme,69 and that more research is needed to develop diagnostic tests more specific than conventional oral examination. 68 The authors also report that screening programmes have not resulted in a demonstrable reduction in mortality, apart from in high-risk groups, in which there is some evidence that screening may be effective and cost-effective. 68 Instead of asserting the need for shorter intervals between dental recall appointments, these papers67–69 highlight the need for oral health services to develop strategies to reach out to populations that do not attend primary care dental services regularly, instruct patients about high-risk habits including alcohol and tobacco use, and better network with other primary care services.

Evidence base

The recommendations in the NICE guideline on dental recall60 are designed to aid dentists in assigning individualised recall intervals to patients based on their risk of developing oral disease. The Guideline Development Group produced a checklist for use by dentists when assessing risk, including specific risk factors for consideration. 60 The guideline recommends an adjustable recall interval for adults, ranging from a minimum of 3 months to a maximum interval of 24 months between recall appointments for patients who have repeatedly demonstrated an ability to maintain oral health. The guideline panel recommended that the recall interval be regularly assessed, discussed and agreed based on each individual’s oral health risk profile and amended accordingly. The Guideline Development Group considered a balance of benefits and harms related to caries, periodontal disease and also oral mucosal lesions in making its recommendations. The recommendations are, however, based on low-quality evidence and the clinical experience of the Guideline Development Group. This lack of evidence has complicated the implementation of this guideline for dentists and health service commissioners.

Although the concept of assigning risk-based recall intervals has gained increasing standing internationally, the clinical effectiveness of this recall protocol is not supported by scientific evidence from clinical trials. Furthermore, there remains significant variation in professional recommendations within and between countries regarding the maximum time interval between dental check-ups that can reasonably be assigned for patients at low risk of oral disease. This can be considered inevitable given that many guideline recommendations regarding this issue have been informed primarily by professional consensus and are subject to variation in interpretation. 70

There is also a paucity of reliable scientific evidence to support the effectiveness of routine dental checks of differing recall frequencies in adults. The scientific basis for the 6-month dental examination was questioned more than 30 years ago. 2 Since then, systematic reviews investigating this key question have reported limited evidence of poor overall quality, which is insufficient to reach any conclusions regarding the potential beneficial and harmful effects of varying recall intervals between dental check-ups, and concluding that there is no evidence to support or refute the practice of encouraging patients to attend for dental check-ups at 6-month intervals. 71,72 The NICE guideline on dental recall reiterated the need for research in this area to examine the effects of varying dental recall intervals on oral health. 59 The NICE guideline also concluded that the research base was severely lacking in terms of determination of the optimal dental recall intervals on the basis of cost-effectiveness. 60

There is, therefore, an urgent need to assess the relative effectiveness and cost-effectiveness of different dental recall intervals in a robust, sufficiently powered randomised control trial (RCT) in primary dental care.

The questions addressed by the INTERVAL trial

Study design

The trial was a UK-wide [England (London, Manchester, Birmingham, North East), Wales (Cardiff), Northern Ireland (Belfast, County Down) and Scotland] multicentre, parallel-group, RCT with blinded outcome assessment at 4-year follow-up.

The trial interventions were three recall intervals – a fixed-period 24-month recall interval, a risk-variable adjusted-length recall interval (risk-based recall) based on the NICE guideline60 and a fixed-period 6-month recall interval.

A two-stratum trial design was proposed to overcome potential ethical considerations and dental clinician and/or participant concerns. Participants were randomised to the fixed-period 24-month recall interval only if the recruiting dentist considered them clinically suitable.

Randomisation was organised within the two strata (Figure 1):

1. For those participants considered suitable for a fixed-period 24-month recall (stratum 1), randomisation was to one of three groups:

   1. Fixed-period 24-month recall versus risk-variable-adjusted-length recall (risk-based recall) versus fixed-period 6-month recall.

2. For those participants not considered suitable for a fixed-period 24-month recall (stratum 2), randomisation was to one of two groups:

   1. Risk-variable-adjusted-length recall (risk-based recall) versus fixed-period 6-month recall.

FIGURE 1.

Study design.

An economic evaluation to determine the cost-effectiveness of different recall intervals and to compare NHS and patient-incurred costs and benefits is included in Chapter 5.

Ethics approval and protocol amendments

Favourable ethical opinions were granted for the INTERVAL Dental Recalls Trial by the Fife and Forth Valley Research Ethics Committee (feasibility study Research Ethics Committee reference number 09/S0501/1; main study Research Ethics Committee reference number 09/S0501/1).

The trial was registered with the International Standard Randomisation Controlled Trial Register (ISRCTN), reference number 95933794.

Amendments to the protocol were made after recruitment of practices and participants, and on conclusion of the feasibility study. These included an increase of more than one dentist per practice able to participate in consenting, recruiting, randomising and establishing risk-based recall intervals for participants in the risk-based arm and the assistance of dental postgraduate research networks to identify and recruit potential dentists and identify and approach potential participants.

Additional amendments, notified to the funder, included an increase in the number of practices recruited and increased numbers of participants per practice, extension of the recruitment period, changes of Trial Steering Committee (TSC) and Data Monitoring Committee (DMC) members, lengthening from 3 to 4 months plus or minus the 4-year anniversary of participant randomisation for final year assessments, and adaptions to study administrative processes. All changes were in accordance with approved contract variations.

Recruitment and consent of dental practices

The trial sought to recruit general dental practitioners/practices from across the UK (i.e. England, Wales, Northern Ireland and Scotland), representing a cross-section of practitioners in terms of urban/rural areas, community-level sociodemographics and fluoridated or non-fluoridated communities.

Dentists were recruited through local postgraduate dental research networks, by advertising in professional dental publications and through presentations at dental conferences and dental events. Trial information and recruitment evenings were organised in Birmingham and Cardiff and across Scotland.

The Trial Office in Dundee (TOD) sent potential dentist participants a personalised invitation letter for the dentist and their staff to attend a local information and recruitment session, at which the reasons for and design of the trial and practice involvement were described. Dental professionals were given the opportunity to discuss participation with the trial team. For dentists and teams that could not attend, information packs about the trial were posted/e-mailed from the TOD.

Trial team members telephoned dental practices to follow up the notes of interest of involvement. A site briefing/training session was arranged with the dentist, practice staff and TOD staff (see Training of dentists). Following the site briefing, dentists who were interested in the trial were asked to provide written consent to participate, a signed declaration agreeing to adhere to the trial protocol and a completed clinician beliefs questionnaire (see the NIHR project web page for details: www.journalslibrary.nihr.ac.uk/programmes/hta/063599/#/documentation; accessed October 2020). Original signed and dated dentist consent forms and declarations were held securely as a part of the trial site file at the TOD. Copies were made and returned to dentists.

Inclusion/exclusion criteria: dental practices

The inclusion criteria were:

• NHS provider for adult patients

• primary care provider: salaried service, corporate and independent operators

• willingness to follow trial protocol.

The exclusion criteria were:

• providing only private dental services to adults

• unwilling to follow trial protocol.

Recruitment and consent of participants

Recruitment of patient participants was achieved through standard procedures and agreements for primary care research in the four nations. In some areas of England, Wales and Northern Ireland, regional Clinical Local Research Networks assisted dental practice staff to identify eligible patients and facilitate an approach by including information about the trial in the appointment letter for their routine dental examination. In Scotland, co-ordinators from the Scottish Primary Care Research Network, when invited, provided a similar service. The appointment letter included an invitation to participate, the patient information leaflet and the baseline patient and cost questionnaires (see the project web page for copies of questionnaires: www.journalslibrary.nihr.ac.uk/programmes/hta/063599/#/documentation).

In instances where the Clinical Local Research Networks and Scottish Primary Care Research Network were not available at an agreeable time, or practices did not request or require assistance, practice staff undertook the duties of identifying and contacting eligible patients and inviting them to participate in the trial using the same paperwork.

At the routine appointment, discussion about the trial was held between the dentist and patient. If agreeable, potential participants were screened for suitability prior to their routine dental examination. Patients who contacted the practice to advise that they were not interested in taking part in the trial were reassured that they would still receive a dental examination appointment with their dentist as per practice policy.

There are a variety of patient recall appointment management strategies utilised within dental practices across the UK. Some dental practices arrange routine examination appointments for their patients up to 6 months or a year in advance. Some practices send letters, e-mail, telephone or text reminders to their patients when their routine dental examinations are due, asking them to contact the dental practice to make an appointment, whereas other practices pre-allocate the date and time of appointments and ask patients to contact the practice if the appointment is not suitable. The INTERVAL Dental Recalls Trial utilised a flexible and pragmatic participant recruitment strategy that aimed to be suitable for each practice’s usual recall procedure.

Eligibility of those who expressed an interest in taking part was confirmed against the trial inclusion and exclusion criteria. The dentist confirmed consent with those eligible and willing to participate in the trial. A signed participant consent form was obtained in triplicate. The participant retained a copy, the practice retained a copy in the patient’s notes in the site file, and the original copy was sent to the TOD.

Following consent, the dentist clinically examined the participant to establish suitability for randomisation to the 24-month arm (see Randomisation). If participants had not completed the questionnaires provided with the appointment letter, they were asked to complete the baseline questionnaire and cost questionnaire in the waiting room of the dental practice before placing them in a sealed opaque envelope and returning them to practice staff. Questionnaires were returned to the TOD by the dental practice in a sealed envelope.

The TOD staff did not have access to any participant data prior to the participants consenting to take part in the trial.

Inclusion/exclusion criteria: participants

The inclusion criteria were adult patients (≥ 18 years of age) who:

• were dentate

• had visited their dentist in the previous 2 years

• received their dental care in part or fully as an NHS patient, including dental examination.

The exclusion criteria were:

• patients who had a medical condition indicating increased risk of bleeding

• immunocompromised patients.

Participants whose medical condition changed during the follow-up period were not prohibited from continuing in the trial. Provision was made for dentists to record changes and rationale to the length of the recall interval, on the patient attendance data (PAD) form, but such participants remained within the allocated stratum.

Training of dentists

In England, Wales, Northern Ireland and Scotland, the process of training recruited dentists took the same format.

Training in trial procedures

Trial staff visited the practice by arrangement for a 1- to 2-hour site briefing/training session at an agreed and convenient time, attended by the participating dentist and practice staff. After a brief review of the trial aim and objectives, trial procedures were described and discussed.

Recruited dentists were defined as local investigators within the dental practice, and were responsible for recruiting, consenting and protecting the personal data of trial participants within the dental practice. Local Investigators signed an agreement to conduct the trial in compliance with applicable legislation including (1) the International Conference on Harmonisation Good Clinical Practice guideline73 and (2) the Department of Health and Social Care’s Research Governance Framework for Health and Social Care (April 2005)74 or the Scottish Executive Health Department’s Research Governance for Health and Community Care (2nd edition 2006),75 whichever was relevant.

Dentists and practice staff were advised at the trial briefing/training and in monthly practice newsletters that the International Conference on Harmonisation Good Clinical Practice guideline training was available in their local area and that attendance at these sessions could be arranged through the TOD. The TOD also signposted to the online NIHR Introduction to Good Clinical Practice e-learning (primary care) module. 76

Training in determining risk-based recall intervals

Following the site briefing/training session, each dentist was sent a link to an online training package. The online training package presented risk-based recall interval determination according to the NICE guideline,60 with written instruction, audio and video components, examples and test assessments. It described a systematic approach dentists could follow to consider setting and review of individualised patient recall intervals and to enable discussion and explanation between patient and dentist of the risk-based recall interval. Dentists were instructed to complete this training before screening any potential patient participants for the trial, and reminders via letter, e-mail, telephone and trial newsletter requested that dentists complete it on an annual basis during follow-up.

On completion of the online training package, and before being awarded with a certificate of training and 2 hours continuing professional development (CPD), dentists were required to complete an evaluation form that asked users to measure to what extent they felt that the training programme met the learning objectives, and how easy it was to access and understand. They were also asked to suggest improvements regarding content, design, navigability and length.

Feedback from users was mixed; some found it difficult to access and to navigate, whereas others found it user-friendly and were reassured that they could use the tool with patients. Some users found the content appropriate, easy to understand and a good support for participating in the trial, whereas others found some aspects to be basic to a practising dentist or felt that it could be shortened for the experienced practitioner.

For additional reference on risk-based interval determination, practices in England, Wales and Northern Ireland were provided with a link to the NICE guideline Dental Recall: Recall Interval Between Routine Dental Examinations. 60 Practices in Scotland were supplied with an electronic link and hard copy of the SDCEP Oral Health Assessment and Review (OHAR) guidance. 60,61 Both documents contained templates for checklists to record variables identified as potential modifying factors (risk variables) that influence the setting of recall intervals (see Trial interventions for more detail on risk-based templates).

Clinical outcome assessor training

The clinical outcome training was delivered by trial collaborators with expertise and experience of training assessors in periodontal and caries measures. The clinical outcome assessors and scribes for this trial were qualified, General Dental Council-registered dentists, dental hygienists/therapists and dental nurses employed by the trial.

The emphasis of the training was consistency of the examination process and agreement of scoring criteria. Following the didactic face-to-face and online training, the outcome assessors, with their research nurse, examined 15 patient volunteers in a clinical setting similar to that of a dental practice. The cohort of patient volunteers were similar in age and dental attendance behaviour to those recruited to the trial. The clinical outcome assessments were conducted at Dundee Dental Hospital and School, and each participant was examined by all outcome assessors.

The processes of clinical outcome assessment were agreed in advance, including the order of outcome measure assessment, time allocation, sequence around the mouth and moisture control. The primary clinical outcome of gingival bleeding on probing is a measure of gingival inflammation. It is described as ‘gingival inflammation/bleeding on probing’ in the study protocol;1 for clarity within this report it will be described as gingival bleeding on probing, the definition and outcome measurement remaining the same as outlined in the protocol. This clinical outcome does not allow for repeat assessment; therefore, neither intra-assessor nor interassessor reliability measurements were possible. Training77 for the primary outcome involved face-to-face discussion with the assessors and scribes about the assessment technique and scoring criteria. Prior to the assessment of the cohort of volunteer patients by the outcome assessment teams, a slide presentation developed for training in commercial clinical trials was presented by a periodontal clinical trial expert, and this was supplemented with group discussion about clinical photographs and clinical cases. Periodontal training included positioning, angulation of instrument and pressure of the University of North Carolina-15 (UNC15) periodontal probe, to ensure a standardised approach by the outcome assessors. Periodontal probing depths were recorded at six sites on erupted teeth using a probing force of approximately 25 g.

Training in caries assessment consisted of several components that have been developed and used in other clinical trials and epidemiological studies. This included an online training programme for the International Caries Detection and Assessment System (ICDAS), a half-day of slide presentations and discussions of the ICDAS codes and protocol for the clinical examination. The assessor training included both theoretical aspects and discussions regarding patient participants within the clinical trial setting. Practical training included simulation of the assessment protocol on extracted carious teeth representing carious lesions at all stages of lesion progression included in the ICDAS scale, as well as clinical assessment of a cohort of volunteers similar to the trial population, who had been specifically recruited for trial assessor training. All training in the use of ICDAS was completed under the supervision of a trial collaborator experienced in the use of ICDAS in clinical research.

The caries detection elements of the ICDAS criteria are now well tested and are advocated for general use as well as for use in the clinical trials and in dental epidemiology. 78,79 The ICDAS criteria measure both early stages of caries and more advanced stages of caries. For early caries, ICDAS measures the surface changes and potential histological depth of carious lesions by relying on surface characteristics related to the optical properties of sound and demineralised enamel prior to cavitation. Advanced stages are recorded when cavitation is evident. The trial utilised a modified ICDAS as clinical data were collected only on the caries experience. Restorations and non-carious tooth loss were not recorded.

The intensive face-to-face and online training was provided a month before the first trial outcome assessment to provide sufficient time for additional training if required. The training was repeated mid-way through the INTERVAL trial clinical outcome assessment period, to reinforce standardisation in the process and clinical measures. Throughout the clinical outcome collection period, the assessment team met regularly to confirm the outcome assessment processes and data collection methods to achieve the highest level of standardisation possible.

Randomisation

Eligible and consenting patient participants were clinically examined by their dentist to determine suitability for randomisation to the 24-month recall arm (yes/no). The decision that a patient was eligible for a 24-month recall was based on routine clinical examination and risk assessment. Dentists were instructed not to apply the detailed risk-based variable assessment unless randomised to the risk-based arm in either stratum.

There were separate, identical algorithms in the trial design for the two strata. Eligible participants were randomised in equal numbers within each of the two strata according to a minimisation algorithm including:

• dentist

• participant age (18–40 years/≥ 40 years)

• filled teeth (n ≤ 8/n > 8)

• absence of gingival bleeding on probing (yes/no)

• exempt from dental charges (yes/no).

Random allocation occurred via telephone, after the decision by the dentists about the patient’s suitability for a 24-month recall. The trial utilised the automated central randomisation service at the Centre for Healthcare Randomised Trials (CHaRT), University of Aberdeen, which had 24-hour telephone access. The service prompted dental practice staff to enter and confirm details by entering numbers (i.e. 1 = yes, 0 = no) on the telephone touch pad.

The dentist communicated the allocation outcome and confirmed trial details with the participant. Participants randomised to a fixed-period recall interval, either 24 or 6 months, were managed according to routine practice regarding the practice recall management system. For participants randomised to receive a risk-based recall, further history taking, examination and assessments were undertaken, if required, to determine the appropriate variable risk-based recall interval. This was discussed and agreed with the patient participant prior to the recall interval being entered into the routine practice management system (see NICE guideline60).

Owing to the nature of the interventions, it was not possible to blind participants and dentists to allocated recall intervals. TOD staff received an e-mail notification when a successful randomisation had taken place, providing practice and participant ID numbers, and trial arm allocation. Following randomisation, dental practice staff were asked to send the original signed patient consent form, baseline patient questionnaire and cost questionnaire, and screening/patient case report form (PCRF) to the TOD.

Trial interventions

The trial interventions recall intervals were a fixed-period 24-month recall interval, a risk-variable-adjusted-length recall interval based on the NICE guideline60 and a fixed-period 6-month recall interval.

Risk-variable-adjusted-length recall interval (National Institute for Health and Care Excellence guideline)

Patient participants allocated to the risk-based recall interval group attended their dentist at time intervals determined by the evidence-based process outlined in the 2004 NICE guideline on dental recall. 60 The essential steps of the procedure and the risk factors collected at recall examinations are outlined (from the guideline) in Figure 2.

FIGURE 2.

The NICE risk-based dental recall procedure and risk factors. Reproduced with permission. © NICE 2004 Dental Recall – Recall Interval Between Routine Dental Examinations. Available from www.nice.org.uk/guidance/cg19/evidence/full-guideline-pdf-193348909. 80 All rights reserved. Subject to Notice of rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

Reproduced with permission. © NICE 2004 Dental Recall – Recall Interval Between Routine Dental Examinations. Available from www.nice.org.uk/guidance/cg19/evidence/full-guideline-pdf-193348909. 80 All rights reserved. Subject to Notice of rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication.

The recommended steps in establishing the appropriate recall interval were:

1. Consider the age range – in the case of this trial, all patients were adults of ≥ 18 years.

2. Consider risk variables – identification of the pertinent risk and protective factors present for each patient from both the checklist and a comprehensive oral health assessment, leading to the evaluation of the impact of these factors in the context of the patient’s past levels of oral health and current disease experience, and then consideration of a likely range of recall intervals.

3. Integrate prediction of recall need – use of all the information obtained by the dental team in order to predict the potential level of threat to maintaining oral health and controlling disease for this patient and, from this, judge the most appropriate next recall interval.

4. Discuss with patient – to explicitly discuss the recommended recall interval with the patient, explain the influencing factors in setting the recall and record the agreed interval (or any reason given by a patient in disagreement).

5. Review – at each check-up review (oral health review), the appropriateness of the preceding interval is reviewed by the dentist and patient and the recall interval is reset according to the experience from the last period along with any change in the risk and protective variables identified at re-examination.

The frequency of recall interval appropriate for an individual patient depends on the likelihood that specific diseases or conditions may develop or progress beyond the control of secondary prevention. The selection of an appropriate recall interval for a patient is a multifaceted clinical decision that involves judgement and cannot be decided mechanistically.

The NICE guideline60 was developed using extensive consensus methods and the limited evidence available. The recommendation was that the recall interval range for adults should vary from 3 to 24 months according to risk.

The NICE guideline checklist60 was intended to be used as a guide to assist the dentist in setting an appropriate recall interval. It is not an exhaustive list of all factors that may influence the choice of a recall interval for a patient. There is insufficient evidence to assign a ‘weight’ to individual factors in the checklist and dentists must use their clinical judgement to weigh the risk and protective factors for each patient. The same checklist is in the SDCEP Oral Health Assessment and Review guidance61 and, therefore, for dentists across the UK guidance is consistent with the content of the trial training material including the online resource.

It was anticipated that by taking a comprehensive history and carrying out a comprehensive oral health assessment the dentist would be better informed to provide an accurate risk assessment and more appropriate preventative and interventive treatment recommendations including advice.

It was envisaged that, once trained, the time taken to complete this process would be 20 minutes for the first risk-setting visit and 15 minutes for subsequent recall examinations (oral health reviews).

Outcome measures

All primary and secondary outcome measures were measured at 4 years’ follow-up and are outlined below.

Data collection

Final/fourth-year follow-up

Dentists completed the end-of-trial clinician belief questionnaire after all clinical assessments had been completed in their practice.

Patient-centred outcomes were collected at 4-year follow-up using the annual patient questionnaire. To maximise return, the final questionnaire was sent to participants at least 6 weeks before the date of their final year clinical assessment appointment, with a reminder sent 2 weeks before the appointment. Participants who had not returned a questionnaire by the time of their follow-up assessment appointment were asked to complete the questionnaire at that appointment prior to being assessed. A summary of the data collection items and time point is presented in Figure 3.

FIGURE 3.

Data collection schedule summary.

All participants were invited to attend a trial final-year clinical assessment appointment by their dental practice either at the time of their routine check-up or as a separate appointment. A brief medical history was undertaken about bleeding disorders or immunocompromised disorders. The assessor also confirmed continuing consent with the participants. Gingival bleeding on probing scores, periodontal probing depths, calculus and caries detection were measured by the assessors and recorded on a clinical chart by the dental research nurse, who was a member of the trial team.

Participants who could not attend were contacted and given the option of attending at least one other day or time. Patients who were no longer registered at the practice were offered the opportunity to return to the practice (with practice permission) for an assessment examination, or at another INTERVAL practice (if appropriate) or at a suitable satellite location.

All participants who attended the follow-up assessment received a letter of appreciation for their participation in the trial, details of where the trial results would be published and a final £15 gift voucher in recognition of their contribution. Letters and vouchers were subsequently posted, where possible, to participants who did not attend a follow-up assessment.

Sample size

An exploratory trial in a similar population reported that 35% of gingival sites were bleeding on probing (SD 25%). 87 The Cochrane review of periodontal instrumentation (PI) suggested that 6-month PI versus no PI reduces bleeding sites by 15%. 88 The recall interval was expected to produce an effect lower than this given that the majority of participants in all arms would still receive PI at some time during follow-up. Assuming that either risk-based versus 24-month recall or 6-month versus 24-month recall could reduce/increase the percentage of sites bleeding by 7.5%, the study with 235 participants in each arm could detect such a difference with 90% power at 5% significance, and, likewise, detect a difference of 0.3 of the SD of the OHIP-14 score or any other global measure of OHRQoL. For the caries clinical outcome, assuming a SD of 3.5, the study with 235 participants per arm would detect a shift in white spot lesions (from 3.3 to 4.2) at 80% power and 5% significance. 79 We combined the two strata, without introducing bias, to estimate this comparison. We anticipated smaller effect sizes for the 6-month versus risk-based recall comparison than 6-month versus 24-month recall given that many of the participants in the risk-based group would be seen more frequently than 24 months. A study with 750 participants in each arm could detect a difference in bleeding scores of 4.5% with at least 90% power and a 5% significance level,89 and likewise detect a difference of 0.17 of the SD of the OHIP-14 score. For the caries clinical outcome, assuming a SD of 3.5, a study with 750 participants per arm could detect a 20% relative shift in white spot lesions from 3.3 to 3.9 at 90% power and 5% significance. 79 Although there was no reason to be concerned about contamination effects in this trial or clustering by dentist, the sample size was conservatively estimated such that if contamination occurred with 15% of the control participants or the intracluster correlation was 0.03, the study would still have 80% power to detect the hypothesised changes in the bleeding score. Our sample size calculations indicated that we needed to randomise 705 participants to stratum 1 (235 in each arm) and 1030 to stratum 2 (515 in each arm).

Contamination effects

There were no obvious mechanisms for contamination to occur in this trial. Our experience of simultaneously conducting an educational cluster and patient randomised trial in dentistry suggested that contamination occurred in at most 15% of participants (if any);89 therefore, fewer participants are required to perform a patient randomised trial than a cluster randomised trial. 90 There is no perceived direct influence of skill on the patient outcomes and, even if that were hypothesised, the intracluster correlation would be very low (< 0.03).

The sample size has been conservatively estimated such that, if contamination occurred with 15% of the control participants or the intracluster correlation was 0.03, the study would still have 80% power to detect the hypothesised changes in the bleeding score.

Statistical analyses of outcomes

Baseline data were explored to better understand dentists’ decision-making process when allocating participants to different strata (eligible to be randomised to 24-month recall vs. not eligible). The primary analysis used an intention-to-treat framework and all participants with available data remained in their allocated groups. The participant outcomes listed above were compared between 24-month, risk-based and 6-month recall groups (for the stratum eligible for a 24-month recall) and risk-based versus 6-month recall (for the stratum ineligible for a 24-month recall). Outcomes collected at year 4 (clinical outcomes, behaviour, knowledge and PBC scores) were analysed using a generalised linear model with a random effect for dental practice. Outcomes collected at years 1, 2, 3 and 4 were analysed using a mixed-effects model with two random effects: participant and practice. A time by treatment interaction term was included in the models. The appropriate effect sizes and 95% confidence intervals were derived. All analyses were adjusted for the protocol minimisation variables [age, dentist, filled teeth (≤ 8 or > 8) and absence of gingival bleeding on probing]. Stata 15 (StataCorp LP, College Station, TX, USA) was used to undertake the analysis.

Missing data

Missing items in scales were dealt with as recommended in the literature by their authors, when recommendations were available. Otherwise, a complete-case approach was used where, in the presence of any missing items in a patient’s score, the score was considered missing. Continuous missing data at baseline were imputed for modelling purposes as recommended in the literature,91 and categorical missing data at baseline used the missing indicator method. 92 The primary intention-to-treat analysis was on observed data.

Sensitivity analysis

We explored differences between responders and non-responders to inform our missing data model. As a sensitivity analysis, missing primary outcome data were imputed using a predictive mean matching multiple imputation approach. 92

Subgroup analysis

A pre-planned subgroup analysis explored the effect modification of age (< 45 years, 45–64 years, ≥ 65 years) and social class (exempt from payment and not exempt from payment) by including a subgroup-by-treatment interaction term in the primary clinical outcome model described in Statistical analyses of outcomes. Further pre-planned subgroup analyses included residence in a fluoridated area, and dentist characteristics were included in the protocol; however, there were too few participants (2%) in fluoridated areas to make a subgroup analysis meaningful. The dentist characteristics subgroup analysis was an error in the protocol. A post hoc subgroup analysis by country was also included.

Trial oversight

The University of Dundee acted as the sponsor for the study. The trial was co-ordinated from the TOD in the Dental Health Services Research Unit, Dundee, which provided day-to-day support for the dental practices and outcome assessors/research nurses. The TOD was responsible for issuing and collecting trial documentation from practices and participants (including annual questionnaires and reminders) and co-ordination of participant follow-up. The TOD was also responsible for the entry of collected data into the database, including screening logs/PCRFs, PAD forms, baseline, costs and annual patient questionnaires, baseline and follow-up dentist questionnaires and clinical data from outcome assessments. Clinical data were entered by the dentally qualified assessment team.

CHaRT, in the Health Services Research Unit, University of Aberdeen, provided the database applications and information technology programming for the trial, and hosted the randomisation system, provided experienced trial management guidance, and took responsibility for all statistical aspects of the trial (including interim reports to the TSC and the DMC).

Trial Operational Committee meetings were held weekly and attended by Co-Chief Investigators, the Trial Manager and key TOD staff.

The Operational Management Committee met monthly, chaired by one of the Co-Chief Investigators, and comprised co-investigators in TOD and CHaRT.

The Trial Management Committee (TMC) met approximately annually, chaired by the Co-Chief Investigators, and comprised co-investigators, key members of the TOD and CHaRT and a lay representative.

The TSC comprised an independent chairperson and two further independent members, plus a member of the public acting as a lay patient representative. The TSC met approximately annually.

The DMC comprised an independent chairperson and two further independent members. The DMC met approximately annually.

Patient and public involvement

Prior to the start of the trial, advice on the design and conduct of the study was sought from members of the public partnership groups and from similar patient groups in other parts of the UK sourced under guidance from INVOLVE (UK National Advisory group).

These independent public partnership groups comprise volunteers who work in partnership with NHS Tayside and aim to provide a conduit for the views of people about their local services.

Patient advisors were a valuable resource at the outset of the trial and helped to ensure good conduct and patient-friendly practice throughout the duration of the trial. Patient advisors were involved with the trial design and provided invaluable feedback on trial recruitment and communication strategies. Patient advisors also contributed to the content and layout of trial invitation, trial newsletters and the design of patient participant questionnaires. This ensured that trial participants could understand and easily complete these materials.

As quality of life was a primary outcome of the trial, patient advisors’ input to the proposed questionnaire design was essential. Qualitative work with patients was carried out to ensure that the outcome measures were patient centred.

Lay representatives on the TSC and TMC actively contributed to trial oversight, processes and procedures, including helping to interpret the trial findings, preparation of the monograph and assisting with the review of the Plain English summary.

Introduction

In this chapter we report the methods used to conduct the within-trial health economic analysis comparing different dental recall strategies as follows: eligible for 24-month recall – 24-month versus risk-based versus 6-month recall; and ineligible for 24-month recall – risk-based versus 6-month recall. All analyses are completed according to the intention to treat principle at 4 years’ follow-up. The health economic analysis compares the costs and benefits of different dental care recall strategies. The results are presented using different perspectives. Economic evaluations to inform UK health-care decision-making, for example through NICE, typically take the form of cost–utility (i.e. cost per QALY) analyses, reporting incremental cost-effectiveness ratios (ICERs). However, in the context of dentistry, there are concerns that generic EQ-5D-3L-based QALYs lack the sensitivity to capture the processes and outcomes of care that are of value to patients and decision-makers. 93 Furthermore, it is argued that the time dimension embedded in QALY calculations may underestimate the impact of acute health effects such as painful caries. 94,95

Similarly, from a cost point of view, dentistry is unique within health care as the majority of patients pay a contribution towards their NHS care. It is, therefore, crucial to consider both the costs falling directly on the NHS dental budget and the impact on patients in terms of co-charges for NHS care and the opportunity cost of time and travel to dental care appointments when making recommendations about the most efficient dental recall strategy.

For these reasons, it is necessary to consider different frameworks of evaluation that (a) are sensitive to changes in important dental health outcomes such as bleeding or caries, (b) value outcomes that are relevant to service users and (c) capture the full cost burden to both patients and the NHS of different recall strategies. Providing results from a range of perspectives of benefits and cost is essential to equip decision-makers with all the relevant information necessary to reach informed decisions about the efficient allocation of scarce dental care resources.

In terms of the benefits, the scopes are WTP for dental recall interval and associated outcomes [cost–benefit analysis, including all benefits (health and non-health) that are important to individuals], QALYs and WTP for dental health outcomes (caries and self-reported bleeding gums). The preferred analysis depends on what outcome a decision-maker wishes to maximise: social welfare, QALYs or (WTP for) defined dental health (caries and bleeding). In terms of costs, the perspectives are NHS dental budget, NHS total budget and societal perspective (NHS and participant).

Resource use and costs

Resource use and cost data are collected from an NHS and dental patient participant perspective. NHS costs include provision of dental care services (obtained from data linkage to routinely collected claims data) and costs of attending non-dental health professionals for dental problems (obtained from participant questionnaires). Patient participant costs include co-payments for dental care, purchase of dental care products, and the travel costs and the opportunity cost of time spent attending dental appointments obtained from a combination of routinely collected data and participant self-reported questionnaires.

NHS dental costs

Resource use and costs associated with the use of NHS dental services are obtained through data linkage for each randomised trial participant to national-level dental claims data held by the ISD (Scotland), NHSBSA (England) and BSO (Northern Ireland). For the 13 participants recruited from a single Welsh practice, resource use is obtained from practice note data extraction. The actual cost to the NHS of providing treatments, the split of cost burden between the NHS and the patient, and the level of granularity of data available for analysis is dependent on different remuneration systems and contracting arrangements in place for payment of NHS dental care across different UK countries. For example, dentists in Scotland and Northern Ireland are reimbursed on the basis of fee for service, whereas in England contract payments are based on banded categories of treatment complexity. This means that there is a greater granularity of data available for Scotland and Northern Ireland to inform resource use and costing. There are also differences in patient co-charges for dental check-ups ranging from £0.00 (Scotland) to £19.70 (England), which have implications for the interpretation of NHS versus patient perspective costs across the regions. Tables 1 and 2 describe the NHS and patient breakdown of treatment charges across the UK regions for fee-paying adults. All cost data are reported in 2016/17 values based on the regional dental payment contracts. Regional specific variation in the dental payment systems is discussed in the following paragraphs.

Participant costs

Three sources of participant cost are included in the analysis: (a) participant co-charges for NHS dental care, (b) direct out-of-pocket expenses for private dental care and (c) the opportunity cost of participant’s (and any companion’s/carer’s) time spent attending dental appointments.

Costs of time and travel

There is a direct cost to the participant for travelling to every dental appointment and an opportunity cost of their time (and their companion’s/carer’s time, if accompanied) spent travelling when they are away from their usual activities (e.g. work, leisure time and child care). Routine dental claims are used to identify the number of attendances for NHS dental treatment. It is conservatively assumed that each unique treatment’s start/acceptance date (rather than each item of service claimed in Scotland/Northern Ireland) relates to a single dental visit. This approach probably underestimates the true participant perspective costs if more than one visit was required for a course of treatment. The unit time and travel cost to participants and companions of a return journey to the dental practice is calculated using data provided by respondents in the baseline questionnaire, using a subset of questions obtained from a standardised patient costs questionnaire. 105 The cost of a visit is assumed to remain constant at the individual level across the time horizon of the trial to avoid overburdening respondents with multiple similar questions in each annual questionnaire. Unit costs of travel and the opportunity cost of time commitment of travelling to and attending dental appointments are detailed in Table 3. The total opportunity cost of time and travel is obtained by multiplying the number of treatment courses by the unit opportunity cost of time and travel for each participant.

Statistical analysis and reporting of cost data

Cost models are estimated using generalised linear models, specifying a log link function and gamma family to account for a skewed distribution with a minimum of zero cost, and a long tail reflecting a small proportion of participants with high costs. The chosen family (gamma) is based on the results of a modified Park test, and the chosen link function is based on a combination of Akaike information criterion (AIC) score and p-values from the Pearson correlation test, the Pregibon link test and the modified Hosmer–Lemeshow test110,111 obtained using the glmdiag program for Stata. Costing models are adjusted for the trial minimisation variables, region and baseline EQ-5D-3L score, and a random effect for centre to control for practice-specific variation in treatment decisions. For the comparison of treatment options across all trial participants, a fixed effect is included to account for randomised stratum in the trial. The cost models are estimated using Stata’s meglm command. Recycled predictions of the coefficient on randomised treatment determine the incremental costs compared with standard care (currently 6-month recall). Two sets of results are reported. First, results are reported for all data comparing risk-based recall with 6-month recall. Second, a three-way comparison is presented for those participants randomised to the stratum of participants who are eligible for 24-month recall. Costs associated with risk-based and 24-month recall are compared against each other, and with the assumed standard care (i.e. 6-month recall).

Descriptive statistics for cost data (mean, SD, N) are reported for all resource use items from both an NHS and a patient participant perspective based on complete-case data for information. NHS perspective costs to the dental budget are relatively complete because of a high match rate across the regions between the routine dental claims data and the recruited trial population. However, participant perspective costs (based on questionnaire response data) are subject to significant missing data. For the statistical analysis of costs, the mechanism of data missingness was investigated using logistical regression analysis, where the dependent binary variable (missing or not) was regressed on sociodemographic characteristics (age, gender), minimisation variables, region, baseline EQ-5D-3L utility score, and baseline measures of clinical outcome. Statistically significant explanatory variables were assumed to indicate that data were missing at random as opposed to missing completely at random. Best practice was applied, imputing data using multiple imputation of missing data following Stata’s MI procedure. 112 Missing cost data are imputed by component of cost (e.g. dental product costs) for each questionnaire, using predictive mean matching (average of five closest values), accounting for the repeated measures nature of the costs from the annual participant questionnaires. M = 5 imputed data sets are used to ensure the stability of results, and the data sets were combined using Rubin’s rules to obtain a single estimate of costs. 113 All data imputation models account for randomised group and adjust for explanatory variables as described above.

Benefits

Three different measures of benefit are included in the economic evaluation as alternative analyses. The analyses differ in terms of the scope and definition of the benefits included. Benefits are considered as (a) WTP for both dental recall frequency and associated dental health outcomes to inform a cost–benefit analysis (CBA) that includes all perceived benefits, health and non-health, to the general population, (b) QALYs to inform a cost–utility analysis (CUA) and (c) WTP for dental health outcomes, specifically ICDAS and bleeding gums on brushing, where the utility of dental health outcomes is measured in terms of WTP. The preferred analysis depends on what outcome a decision-maker wishes to maximise: social welfare, QALYs or (WTP for) dental health outcomes. All benefits are valued using general population preferences.

Willingness to pay: discrete choice experiment

Willingness to pay is calculated based on responses to an online DCE conducted with a nationally representative sample of the UK population, with oversampling in Scotland to determine any subgroup effects on preferences driven by different, regional-specific payment systems. A DCE is a survey designed to elicit respondents’ preferences through the choices they make between different hypothetical goods or services. In the context of this study, respondents make choices between different packages of dental care that vary in terms of their recall frequency, health outcomes (bleeding gums when brushing and caries experience over a 4-year period, as informed by ICDAS classification groups) and cost. Each package comes at an annual cost over 4 years, reflecting the trial’s time horizon.

Discrete choice experiments assume that a treatment or service’s value depends on its component attributes and the levels of those attributes. By including the price proxy within the DCE (i.e. the annual cost of each dental package), we can obtain a monetary valuation for any given dental recall package. These WTP estimates are used as a measure of benefit within the CBA, and can be used to value all outcomes or a subset of those outcomes (e.g. health related only). The DCE received ethics approval from the College Ethics Review Board at the University of Aberdeen (ref: 2015/12/1278). The main considerations for the DCE design are as follows.

Discrete choice experiment selection of attributes and levels

The trial protocol sets the primary research question, and the DCE therefore seeks to value general population preferences for different frequencies of dental recall (24 month, 12 month, 6 month, or risk based). The selection of outcome attributes and levels was partially determined by the trial (e.g. valuing avoidance of caries), but the selection of outcomes for valuation was also based on engagement with key stakeholders (in formal and informal discussion), literature review and primary qualitative (focus group) work with members of the general population to derive a set of attributes and levels meaningful and important to respondents. It became clear from the focus group work that respondents would have difficulty valuing and understanding the primary clinical trial outcome (gingival bleeding on probing). The DCE attribute for bleeding is, therefore, bleeding on brushing (this is mapped to a post hoc trial outcome that asks respondents to report on the frequency of bleeding when they brush their teeth). The final list of attributes and levels included in the DCE is provided in Table 4 and an example choice set can be found in Figure 4.

TABLE 4 - Attributes and levels included in the DCE

Attributes	Levels
You have a dental check-up	Every 2 years Every year Every 6 months It varies depending on your risk
In 4 years’ time, your gums will bleed	Never Hardly ever Occasionally Fairly often Very often
Over the next 4 years, you will have	No dental decay Early dental decay Moderate dental decay Advanced dental decay
The annual cost to you	£20 per year (total cost: £80 over 4 years) £50 per year (total cost: £200 over 4 years) £100 per year (total cost: £400 over 4 years) £200 per year (total cost: £800 over 4 years)

FIGURE 4.

Example choice task. Note: the example choice set shown in Figure 4 is for a respondent categorised as having moderate dental health. If this hypothetical respondent chose dental check-up package B, they would be willing to pay £200 per year for that package.

Discrete choice experiment experimental design

A pivoted, segmented design was generated using Ngene™ experimental design software (ChoiceMetrics Pty Ltd, Sydney, NSW, Australia) to select the combination of treatment descriptions, and to create the choice sets, which minimised the design’s D-error. First, respondents were assigned to one of three segments (good, moderate or poor dental health) based on their self-reported dental health as measured using stated caries experience and bleeding frequency.

The opt-out scenario described the implications of choosing the ‘no dental check-up’ package (i.e. receiving none of the services within the package and at no cost). Outcome attributes in the reference case were tailored to each segment, reflecting a modest reduction in the outcome attributes if no dental package was selected. The levels of the outcome attributes were then pivoted around these baseline (opt-out) levels, to ensure that only realistic choice tasks were presented to respondents. The pivoting ensured that respondents were not presented with situations in which their health could get worse by opting into one of the dental packages.

The DCE has a single five-level attribute with three four-level attributes, resulting in 320 (51 × 43) possible combinations and (320 × 319)/2 = 5040 unique choice sets. The resultant chosen main effects D-optimal design consisted of 24 choices, split equally across three blocks of eight choices. Two further choice tasks were added to test for consistency and dominance to each block of choices, meaning that each respondent completed a total of 10 choice tasks. The responses to the internal validity choice tasks were not included in the analysis models. The experimental design was based on vague priors, specifying only the expected coefficient sign for outcome attributes and null priors for the recall frequency attribute.

Discrete choice experiment questionnaire development

There are three sections to the survey. Section 1 asks about respondents’ experiences of dental care, current dental health, how often they attend their dental practice and by whom they are treated. Responses to dental health questions are used to inform the pivoting of the design as described in Discrete choice experiment experimental design. Section 2 asks the respondents to complete the DCE, and section 3 concludes the survey with a series of demographic and attitudinal questions that may help explain choices. In addition, there are debriefing questions about ease of understanding, realism and appropriateness of the choice task questions, as well as questions designed to understand respondents’ experiences of taking online surveys more generally. The questionnaire was designed using Qualtrics (Qualtrics, Provo, UT, USA) online survey design platform.

Discrete choice experiment data collection

Data were collected using Qualtrics online panels. Data collection for the DCE was conducted in parallel with the trial follow-up period and responses to the survey were provided between 7 July and 5 August 2016. The survey was nationally representative, with population census quotas sought for age (among adult population), sex and region. We oversampled in Scotland (n = 183) to enable regional-specific subgroup analysis split by contract type (UDA, England/Wales; fee for service, Scotland/Northern Ireland). At the pilot stage, 30% of responses were sought from non-regular attenders (defined as not having seen the dentist in the previous 2 years),114 but it was not possible to achieve this target and, instead, a relaxed quote of 10% was sought for the main data collection phase. All responses were anonymised and respondents could leave the survey at any time and did not need to provide a reason for doing so. Those who completed the survey were reimbursed in a manner determined by the survey panel.

Discrete choice experiment data analysis

The DCE data are analysed using best practice methods and followed random utility maximisation theory. 115 In brief, all data are analysed using a mixed logit model, specifically an error components logit model, allowing for multiple choices per respondent. The observable, individual specific systematic component of the utility function (V_njt) for individual n, choosing dental check-up package j, in each choice task t is specified as a linear additive function of the attributes and levels presented to the respondents:

⁽¹⁾ $$\begin{gathered} V njt = α + β 1 Recall annual + β 2 Recall six − monthly + β 3 Recall risk − based + + β 4 Bleed hardly − ever + β 5 Bleed occasional \\ + β 6 Bleed fairly often + β 7 Bleed very often + β 8 Decay early + β 9 Decay moderate + β 10 Decay advanced + β 11 Annual Cost , \end{gathered}$$

where α is the alternative specific constant (ASC), included as a random (normally distributed) parameter in the model. The ASC accounts for latent or unobserved utility associated with choosing any package, regardless of the attribute levels, and β represents the marginal utilities associated with the attributes and levels.

All categorical variables are effects coded where the β coefficients represent the effect of a unit change in an attribute level (away from the grand mean) on utility. The advantage of effects coding is that the reference level of an attribute is uncorrelated with the ASC. The reference categories for the effects coded attributes are 24-month recall, never bleeding and having teeth that have no decay. The utility of the reference category, relative to the grand mean of the attributes, is obtained as the negative sum of the estimated coefficients for that attribute. The coefficient on cost indicates the marginal change in utility for a £1 increase in annual cost. The error term is split into two, where µn is the individual specific random effect and ϵj, the unobserved component of the error term, is independent Type I extreme value distributed. The final model is estimated using the maximum simulated likelihood method, with 200 Halton draws.

The marginal WTP of an attribute level is equal to the negative of the marginal rate of substitution between that level and the cost attribute. For the purposes of result reporting, confidence intervals (CIs) are based on the delta method, calculated using Stata’s nlcom command which computes point estimates and CIs for potentially non-linear combinations of parameter estimates, implemented after running the corresponding error components logit models.

The models are re-estimated separately using interaction terms for several predetermined subgroups, including (1) sex; (2) region, to determine if system of reimbursement, and patient co-charge (£0 in Scotland) has an impact on preferences; (3) smoking status, to determine if current or previous smokers’ preferences differ from the overall group; (4) household income, to determine if lower ability to pay (annual income < £20,800) has an impact on preferences; (5) experience of dental check-ups (attending at least every 6 months); and (6) experience of dental decay, to determine if having experience of dental decay has an impact on preferences for its avoidance. The impact of subgroups on main attribute-level effects is assessed by the significance of interaction terms between subgroup identifier and attribute level coefficient. Likelihood ratio tests are used to test the joint significance of the interaction terms.

Statistical analysis of benefits data

The approach for the analysis of benefits data is similar for calculations of both WTP and QALYs. All benefits, regardless of measurement approach, are discounted at a rate of 3.5% per annum in line with NICE guidance. 103 Total WTP and QALY data are both analysed using a generalised linear regression model, adjusting for minimisation covariates, region and baseline EQ-5D-3L utility score, and include a random effect for centre. For the comparison of all data in a single model, a fixed effect for stratum is added. For each benefit measure, models are estimated using Stata’s meglm command with a Gaussian family and identify link. The chosen family (Gaussian) is based on the results of a modified Park test, and the chosen link function is based on a combination of AIC score, and p-values from the Pearson correlation test, the Pregibon link test and the modified Hosmer–Lemeshow test.

Missing data exist for benefits because participants may not have attended their final year clinical assessment appointment or may not have returned questionnaires detailing experience of bleeding on brushing. As with costs, the mechanism of missingness was explored using logistical regression models to determine any predictors of missingness. Presence of significant explanatory variables was taken as an indication that data were missing at random and could not be assumed missing completely at random, and so multiple imputation would be required. Missing data are imputed using Stata’s MI procedure112,117 (by randomised group) using predictive mean matching (five closest values), and accounting for minimisation variables and baseline EQ-5D-3L utility score. M = 5 imputed data sets are used and results obtained from the combined data using Rubin’s rules. 113

Economic evaluation

Introduction

This chapter describes the study groups at trial entry, followed by a description of the interventions received, the results at the annual follow-up points and a statistical analysis of the primary and secondary outcomes. Randomised groups are presented separately dependent on eligibility stratum: whether the patient participant was eligible or not to be randomised to a 24-month recall. If the participant was eligible, three treatment options were available and are presented in this chapter (i.e. risk-based recall, 24-month recall and 6-month recall). If the participant was not eligible to be recruited to the 24-month recall stratum, two randomised groups are presented (risk-based recall and 6-month recall).

Recruitment to the study

Participants were recruited between July 2010 and July 2014. Data were closed to follow-up on 13 August 2018. The flow of participants for each INTERVAL eligibility stratum is shown separately in Figure 5 (eligible to be randomised to 24-month recall) and Figure 6 (ineligible to be randomised to 24-month recall), with information on primary outcome collection at the different follow-up time points in the form of a Consolidated Standards of Reporting Trials (CONSORT) flow diagram.

FIGURE 5.

The CONSORT flow diagram for participants eligible for the 24-month recall stratum.

FIGURE 6.

The CONSORT flow diagram for participants ineligible for the 24-month recall stratum.

Participants

Participants were recruited from 51 practices (see Appendix 1, Table 28). A total of 2372 participants were recruited to the trial. Of the 648 participants recruited to the 24-month recall stratum, 217 were allocated to risk-based recall, 216 to 24-month recall and 215 to 6-month recall. Of the 1724 participants ineligible for 24-month recall stratum, 861 were allocated to risk-based recall and 863 were allocated to 6-month recall. In total, 1078 participants received a risk-based recall and 1078 received a 6-month recall. Half of all participants (1188) were recruited from Scotland, 1031 were recruited from England and Wales (43%) and 153 were recruited from Northern Ireland (7%).

Description of the groups at trial entry

Trial follow-up

Statistical analyses

Patient-reported outcomes at 4 years

Table 14 shows the patient-reported outcomes at 4 years. Participants in the eligible for 24-month recall group had lower OHIP-14 scores than ineligible participants, as observed at baseline. Participants remained positive about dental services, with a high level of satisfaction on average (around 5 out of 7, where 7 represents the maximum satisfaction with the service). Anxiety levels remained similar, and on average around 10 out of 25 on the MDAS, ranging from 5 to 25 (maximum anxiety). Outcomes measured only at 4 years included PBC, attitude, oral health behaviour and knowledge. The first two ranged from 1 to 7 (where 7 means maximum PBC and most positive attitude); participants had, on average, a score of 4.8 for PBC and 4.2 for attitude.

TABLE 14 - Patient-reported outcomes at 4 years

Patient-reported outcome	Eligible for 24-month recall			Ineligible for 24-month recall
	Risk based	24 month	6 month	Risk based	6 month
Primary patient-reported outcome, mean (SD), n
OHIP-14 (score 0–56)	4.1 (5.7), 145	4.8 (6.4), 153	4.8 (6.2), 152	5.5 (6.8), 624	5.8 (8.3), 630
Secondary patient-reported outcomes, mean (SD), n
Satisfaction (score 1–7)	5.2 (0.7), 151	5.0 (0.7), 153	5.1 (0.7), 155	5.3 (0.6), 647	5.2 (0.6), 654
Dental anxiety (score 5–25; MDAS)	10.5 (4.5), 150	10.9 (4.7), 153	11.0 (4.8), 155	10.3 (4.5), 645	10.5 (4.7), 649
Perceived behaviour control (score 1–7)	4.8 (1.4), 149	4.6 (1.4), 153	4.5 (1.5), 155	4.7 (1.5), 647	4.7 (1.4), 654
Attitude (score 1–7)	4.3 (0.9), 151	4.3 (0.9), 153	4.1 (0.8), 155	4.1 (0.9), 647	4.1 (0.8), 654
Behaviour (score 1–9)	5.2 (1.8), 151	4.9 (1.7), 153	5.3 (1.6), 155	5.5 (1.7), 648	5.4 (1.7), 654
Knowledge (score 1–9)	6.5 (1.5), 150	6.6 (1.3), 153	6.8 (1.2), 155	6.6 (1.4), 648	6.7 (1.4), 651
Self-reported bleeding (score 1–5)	2.0 (1.0), 148	2.0 (0.9), 151	2.2 (1.0), 154	2.1 (1.0), 590	2.1 (1.0), 603
EQ-5D-3L utility (score –0.594 to 1)	0.893 (0.208), 145	0.884 (0.209), 151	0.856 (0.268), 157	0.866 (0.230), 637	0.867 (0.237), 645

Patient-reported outcomes means and standard errors per year and per stratum are presented in Figures 7–10. Overall, the scores remained unchanged during the trial and are similar across groups. Patient-reported outcomes by year and stratum are presented in Appendix 1, Tables 54–56.

FIGURE 7.

Primary participant reported outcome: OHRQoL (OHIP-14; mean and standard error) by randomised allocation and risk stratum. (a) Eligible for 24-month recall; and (b) ineligible for 24-month recall.

FIGURE 8.

Anxiety (mean and standard error) by randomised allocation and risk stratum. (a) Eligible for 24-month recall; and (b) ineligible for 24-month recall.

FIGURE 9.

Attitude scale (mean and standard error) by randomised allocation and risk stratum. (a) Eligible for 24-month recall; and (b) ineligible for 24-month recall.

FIGURE 10.

Satisfaction scale (mean and standard error) by randomised allocation and risk stratum. (a) Eligible for 24-month recall; and (b) ineligible for 24-month recall.

Subgroup analyses

Bleeding on probing outcome

Figures 11 and 12 show the means and 99% CIs for the differences in bleeding at 4 years in the subgroups for recall frequency and strata, respectively. There was no evidence of treatment heterogeneity by age category or by payment for treatment at the 1% level (see Appendix 1, Table 52). There was also no evidence of treatment heterogeneity for country in the overall sample (eligible for 24-month recall stratum and ineligible for 24-month recall stratum); however, in the eligible for a 24-month recall stratum, participants in England who were randomised to a 6-month recall showed a significant improvement compared with those randomised to a risk-based recall (mean difference 4.98, 95% CI 1.14 to 8.83; p < 0.001).

FIGURE 11.

Subgroup results for recall allocation in the eligible for 24-month recall stratum for bleeding on probing: difference between arms by subgroup.

FIGURE 12.

Subgroup results for recall allocation in the overall sample (eligible and ineligible strata) for bleeding on probing: difference between arms by subgroup. RB, risk-based.

Caries outcome (post hoc subgroup analyses)

Figures 13 and 14 show the odds ratios and 99% CIs for caries at 4 years, comparing by recall frequency and by stratum. For this purpose, caries were grouped into two categories: sound surfaces/initial lesions (n = 303) versus moderate or serious lesions (n = 1313). There was no evidence of treatment heterogeneity by age category, country or payment for treatment at the 1% level; in the overall sample in England, patients randomised to 6-month recall were less likely to have serious carious lesions than those randomised to risk-based recall (odds ratio 1.66, 95% CI 0.98 to 2.81; p = 0.01). Interaction coefficients are shown in Appendix 1, Table 53.

FIGURE 13.

Subgroup results for recall allocation in the eligible for 24-month recall stratum for caries: odds ratio by subgroup (CIs with cap have been truncated to fit the plot). RB, risk based.

FIGURE 14.

Subgroup results for recall allocation in the overall sample (eligible and ineligible strata) for caries: odds ratio by subgroup. RB, risk based.

Service providers’ measures

Introduction

This chapter reports the results of the within-trial health economic analysis (4-year follow-up). Descriptive statistics for resource use, costs and outcomes (QALYs, WTP) are presented separately by randomised stratum as follows: [eligible for 24-month recall (24 month; risk based; 6 month) and ineligible for 24-month recall (risk based; 6 month)] using complete-case data. The chapter then reports the findings of the online general population DCE. The economic evaluation results, combining incremental costs and benefits, are then reported under the three different evaluation frameworks (see Chapter 3) based on multiple imputation of missing data. A range of alternative scenario analyses are provided to investigate the impact of different costing perspectives and methodological issues on the results.

Resource use and costs

Resource use and costs are reported for NHS dental care, NHS other care and participants. This section details descriptive statistics from complete-case data. The reader is referred to Appendix 2, Table 59, for descriptive statistics based on multiple imputation of missing data.

Benefits

Benefits are assessed in this study using a DCE with the general population to obtain WTP tariffs to map to clinical trial outcomes (see Chapter 3) and services received and in terms of EQ-5D-3L-based QALYs.

Quality-adjusted life-years

Table 20 details descriptive statistics for the EQ-5D-3L and calculated QALYs from the trial. Across all time points, full EQ-5D-3L profile response data were available for only 35% (836/2372) of the randomised trial population. However, there were only 151 (6%) members of the trial population missing all EQ-5D-3L time points (including baseline). Missing data were evenly distributed across the randomised groups. Additional regression analysis found no evidence that missingness was determined by any participant characteristics or dental health outcomes. QALY data were, therefore, assumed to be missing at random.

Respondents were generally in good health, with greater than 3.5 undiscounted QALYs gained over 4-year time horizon across the randomised arms and strata.

Discrete choice experiment results

Economic evaluation results

This section combines costs and benefits data, using multiple imputation of missing data, under three evaluation frameworks: (1) cost per QALY gained, (2) INB (intervention receipt + dental health benefits) and (3) INB (dental health outcomes: bleeding gums and caries experience).

Economic evaluation results for risk-based versus 6-month recall (whole trial population)

Table 24 reports the results and Figures 15–17 illustrate the joint uncertainty in costs and benefits for the comparison of risk-based versus 6-month recall using the data set pooled across strata. The figures illustrate joint uncertainty in costs and benefits using scatterplots of the cost-effectiveness (benefit) plane and cost-effectiveness (benefit) acceptability curves for each analysis framework. The preferred recall strategy is highlighted in each of the analyses in Table 24 for ease of reading.

For the comparison of risk-based versus 6-month recall, there was no evidence of a difference in total costs to the dental budget, QALYs gained or dental health outcomes. On the balance of probabilities, 6-month recall has a greater chance of being the most efficient strategy when considering costs from only a dental health budget perspective. There is substantial uncertainty in this conclusion for the CUA and cost per monetary valuation of dental outcomes, with a much higher chance of net benefit associated with 6-month recall when adopting a societal approach to benefit valuation in the CBA. The results for the CUA and CBA remain robust to the range of deterministic sensitivity analyses undertaken. The results of the incremental cost per unit of (monetary) health benefit (i.e. framework 3) are generally robust, with two exceptions:

1. Including participant costs of attending dental appointments (i.e. the opportunity cost of time and travel, and the co-charge payments incurred) reduces the cost-effectiveness case for 6-monthly recalls. In this scenario, there is no evidence to suggest that 6-month recall is an efficient use of resources when society is only willing to pay for health benefits; however, broadening the costing perspective does not change the findings from the CBA that 6-monthly recalls have the greatest probability of maximising societal net benefit.

2. The most efficient recall strategy when measuring benefits in terms of WTP for dental health (bleeding and caries) outcomes is sensitive to the UK region analysed. Considering data from Scotland (fee for service) only suggests a substantially higher chance that risk-based recall is associated with positive net (dental health) benefits, whereas in England (treatment banding), 6-month recall is the most likely net beneficial strategy. There are two reasons for these divergent region-specific results. First, in England there are more band 2 treatments (specifically fillings) provided in the risk-based group (ineligible stratum only), generating higher costs relative to 6-month recall than in Scotland. Second, the larger number of fillings is also counted on the benefits side (as a negative) in the WTP calculation in England; therefore, the combined effect of these drivers is a substantial difference in the probability of each strategy being an efficient use of resources across the regions (simultaneously accounting for the joint uncertainty in both costs and dental health benefits).

TABLE 24 - Economic evaluation results (risk based vs. 6 months)

For the complete-case analysis, number of observations was as follows: CUA (n = 742); CBA (n = 961); WTP dental health framework (n = 961).

Note that the estimates of incremental costs and benefits are obtained from mixed-effects regression using MI estimate. Ratios are calculated parametrically and the probability of cost–benefit is obtained through bootstrapped loops of the chosen multiple imputation model (n = 2000 repetitions with m = 5 imputed data sets). All analyses are reported on the multiply imputed data set unless otherwise stated. All models are based on the combined trial sample randomised to 6 month or risk based, with adjustment for stratum. Purple shaded text indicates the most efficient strategy for each comparison in each analysis framework.

Analysis	Incremental cost: mean difference, risk based vs. 6 month (95% CI)	Framework 1: CUA		Framework 2: CBA		Framework 3: WTP (dental outcomes)		Probability of being the most efficient recall strategy
		Incremental QALY: mean difference, risk based vs. 6 month (95% CI)	ICER (£/QALY)	Incremental benefits: mean difference, risk based vs. 6 month (95% CI)	INB	Incremental benefits: mean difference, risk based vs. 6 month (95% CI)	INB	Recall strategy	CUA (@ £20,000 per QALY)	CBA (@ benefit/cost = 1)	WTP (dental outcomes) (@ benefit/cost = 1)
Costing perspective: NHS dental	+£1.82 (–£6.66 to +£10.30)	–0.006 (–0.042 to +0.031)	Dominated	–£39.39 (–£58.40 to –£20.38)	–£41.21	–£4.05 (–£11.76 to +£3.66)	–£5.87	Risk based	30%	0%	17%
								6 month	70%	100%	83%
Costing perspective: NHS all	–£1.32 (–£19.22 to +£16.58)	–0.006 (–0.042 to +0.031)	£220	–£39.39 (–£58.40 to –£20.38)	–£38.07	–£4.05 (–£11.76 to +£3.66)	–£2.73	Risk based	31%	0%	34%
								6 month	69%	100%	66%
Costing perspective: NHS + participant-incurred costs	–£13.02 (–£68.09 to +£42.05)	–0.006 (–0.042 to +0.031)	£2170	–£39.39 (–£58.40 to –£20.38)	–£26.37	–£4.05 (–£11.76 to +£3.66)	+£8.97	Risk based	32%	22%	69%
								6 month	68%	78%	31%
Costing perspective: direct costs to NHS and participants of attending NHS dental care	–£2.87 (–£20.31 to +£14.57)	–0.006 (–0.042 to +0.031)	£478	–£39.39 (–£58.40 to –£20.38)	–£36.52	–£4.05 (–£11.76 to +£3.66)	–£1.18	Risk based	30%	0%	51%
								6 month	70%	100%	49%
Map WTP for caries outcome to untreated caries only	+£1.82 (–£6.66 to +£10.30)	–0.006 (–0.042 to +0.031)	Dominated	–£41.79 (–£61.10 to –£22.49)	–£43.61	–£6.45 (–£17.07 to +£4.17)	–£8.27	Risk based	30%	0%	12%
								6 month	70%	100%	88%
WTP mapped to trial outcomes using per protocol analysis	+£1.82 (–£6.66 to +£10.30)	–0.006 (–0.042 to +0.031)	Dominated	–£171.66 (–£159.92 to –£183.41)	–£176.48	–£6.45 (–£17.07 to +£4.17)	–£8.27	Risk based	30%	0%	17%
								6 month	70%	100%	83%
Scotland only	–£3.57 (–£12.61 to +£5.48)	–0.008 (–0.057 to +0.041)	£446	–£65.54 (–£99.29 to –£31.78)	–£61.97	+£6.52 (–£10.72 to +£23.77)	+£10.09	Risk based	37%	0%	91%
								6 month	63%	100%	9%
England only	+£11.42 (–£3.91 to +£26.74)	–0.003 (–0.058 to +0.052)	Dominated	–£23.18 (–£48.03 to +£1.67)	–£34.60	–£13.21 (–£23.95 to –£2.46)	–£24.63	Risk based	35%	0%	1%
								6 month	65%	100%	99%
Undiscounted costs and benefits	+£1.74 (–£7.23 to +£10.70)	–0.007 (–0.064 to +0.050)	Dominated	–£11.83 (–£17.12 to –£6.54)	–£13.57	–£1.15 (–3.58 to +£1.28)	–£2.89	Risk based	31%	0%	25%
								6 month	69%	100%	75%
Complete-case analysisa	CUA: –£1.48 (–£13.01 to +£10.05) CBA: –£7.00 (–£20.90 to +£6.89) WTP (dental): –£7.00 (–£20.90 to +£6.89)	–0.022 (–0.067 to + 0.020)	£67.27	–£53.43 (–£83.25 to –£23.61)	–£46.00	–£4.42 (–£14.19 to +£5.35)	+£2.58	Risk based	25%	0%	62%
								6 month	75%	100%	38%

FIGURE 15.

Risk-based vs. 6-month recall (CUA): (a) scatterplot of incremental costs and incremental QALYs; and (b) CEAC.

FIGURE 16.

Risk-based vs. 6-month recall (CBA): (a) scatterplot of incremental costs and incremental benefits (WTP); and (b) CBAC. NB, net benefit.

FIGURE 17.

Risk-based vs. 6-month recall (WTP for dental health outcomes): (a) scatterplot of incremental costs and incremental benefits (WTP); and (b) CBAC. NB, net benefit.

Economic evaluation results for risk-based versus 6-month versus 24-month recall (stratum eligible for 24-month recall)

Table 25 and Figures 18–20 report the corresponding findings from the three-way comparison of risk-based versus 6-month versus 24-month recall for those deemed eligible for 24-month recall.

TABLE 25 - Economic evaluation results (participants eligible for 24-month recall)

Percentages are rounded to the nearest whole number for presentation in the table, and may, therefore, occasionally sum to 0.99 or 1.01.

For the complete-case analysis, number of observations was as follows: CUA (n = 202), CBA (n = 285), WTP dental health framework (n = 285).

BCR, benefit–cost ratio; P (C/B), probability of positive net benefit at a BCR = 1; P (C/E), probability of cost-effectiveness.

Notes

Estimates of incremental costs and benefits are obtained from mixed-effects regression using MI estimate. Ratios are calculated parametrically and the probability of cost–benefit is obtained through bootstrapped loops of the chosen multiple imputation model (n = 2000 repetitions with M = 5 imputed data sets). All analyses are reported on the multiply imputed data set unless otherwise stated. Purple shaded text indicates the most efficient strategy for each comparison in each analysis framework.

In this case, the ICER is reported for an intervention in the south-west quadrant of the cost-effectiveness plane; therefore, a lower value of the ICER indicates that the cost savings achieved for the risk-based intervention are not sufficient to justify a QALY loss. In such scenarios, lower ICERs indicate that an intervention is less cost-effective than its comparator.

Analysis	Comparison	Incremental cost, mean difference (95% CI)	Framework 1: CUA		Framework 2: CBA		Framework 3: WTP (dental outcomes)		Probability of being the most efficient recall strategya
			Incremental QALY, mean difference (95% CI)	ICER	Incremental benefits, mean difference (95% CI)	INB	Incremental benefits, mean difference (95% CI)	INB	Recall strategy	CUA P (C/E) @ £20,000	CBA P (C/B) @ BCR = 1	WTP (dental health) P (C/B) @ BCR = 1a
Costing perspective: NHS dental	24 month vs. risk based	–£6.46 (–£18.03 to +£5.11)	+0.007 (–0.102 to +0.117)	Dominant	–£109.22 (–£151.34 to –£67.09)	–£102.76	+£0.57 (–£18.28 to +£19.43)	£7.03	Risk based	16%	0%	12%
	24 month vs. 6 month	–£12.32 (–£30.94 to +£6.29)	+0.021 (–0.052 to +0.095)	Dominant	–£172.43 (–£241.82 to –£103.04)	–£160.11	–£1.22 (–£20.62 to +£18.18)	£11.10	6 month	25%	100%	3%
	Risk based vs. 6 month	–£5.86 (–£21.85 to +£10.12)	+0.014 (–0.082 to +0.110)	Dominant	–£63.21 (–£110.24 to –£16.18)	–£57.35	–£1.80 (–£19.72 to +£16.13)	£4.06	24 month	59%	0%	86%
Costing perspective: NHS all	24 month vs. risk based	–£3.61 (–£40.11 to +£32.89)	+0.007 (–0.102 to +0.117)	Dominant	–£109.22 (–£151.34 to –£67.09)	–£105.61	+£0.57 (–£18.28 to +£19.43)	£4.18	Risk based	17%	0%	20%
	24 month vs. 6 month	–£9.33 (–£50.44 to +£31.77)	+0.021 (–0.052 to +0.095)	Dominant	–£172.43 (–£241.82 to –£103.04)	–£163.10	–£1.22 (–£20.62 to +£18.18)	£8.11	6 month	25%	100%	10%
	Risk based vs. 6 month	–£5.73 (–£43.81 to +£32.35)	+0.014 (–0.082 to +0.110)	Dominant	–£63.21 (–£110.24 to –£16.18)	–£57.48	–£1.80 (–£19.72 to +£16.13)	£3.93	24 month	59%	0%	70%
Costing perspective: NHS + participant–incurred costs	24 month vs. risk based	–£37.92 (–£178.37 to +£102.53)	+0.007 (–0.102 to +0.117)	Dominant	–£109.22 (–£151.34 to –£67.09)	–£71.30	+£0.57 (–£18.28 to +£19.43)	£38.49	Risk based	17%	6%	26%
	24 month vs. 6 month	–£25.22 (–£126.33 to +£75.89)	+0.021 (–0.052 to +0.095)	Dominant	–£172.43 (–£241.82 to –£103.04)	–£147.21	–£1.22 (–£20.62 to +£18.18)	£24.00	6 month	25%	94%	9%
	Risk based vs. 6 month	+£12.70 (–£120.90 to +£146.31)	+0.014 (–0.082 to +0.110)	£907	–£63.21 (–£110.24 to –£16.18)	–£75.91	–£1.80 (–£19.72 to +£16.13)	–£14.50	24 month	59%	0%	65%
Costing perspective: direct costs to NHS and participants of attending NHS dental care	24 month vs. risk based	–£41.84 (–£64.87 to –£18.82)	+0.007 (–0.102 to +0.117)	Dominant	–£109.22 (–£151.34 to –£67.09)	–£67.38	+£0.57 (–£18.28 to +£19.43)	£42.41	Risk based	16%	1%	1%
	24 month vs. 6 month	–£57.82 (–£106.55 to –£9.09)	+0.021 (–0.052 to +0.095)	Dominant	–£172.43 (–£241.82 to –£103.04)	–£114.61	–£1.22 (–£20.62 to +£18.18)	£56.60	6 month	23%	99%	0%
	Risk based vs. 6 month	–£15.98 (–£58.60 to +£26.64)	+0.014 (–0.082 to +0.110)	Dominant	–£63.21 (–£110.24 to –£16.18)	–£47.23	–£1.80 (–£19.72 to +£16.13)	£14.18	24 month	62%	0%	99%
Map WTP for caries outcome to untreated caries only	24 month vs. risk based	–£6.46 (–£18.03 to +£5.11)	+0.007 (–0.102 to +0.117)	Dominant	–£108.03 (–£153.00 to –£63.07)	–£101.57	+£1.76 (–£23.33 to +£26.84)	£8.22	Risk based	17%	0%	8%
	24 month vs. 6 month	–£12.32 (–£30.94 to +£6.29)	+0.021 (–0.052 to +0.095)	Dominant	–£174.98 (–£248.18 to –£101.78)	–£162.66	–£3.78 (–£27.71 to +£20.16)	£8.54	6 month	26%	100%	5%
	Risk based vs. 6 month	–£5.86 (–£21.85 to +£10.12)	+0.014 (–0.082 to +0.110)	Dominant	–£66.95 (–£118.77 to –£15.13)	–£61.09	–£5.53 (–£32.39 to +£21.23)	£0.33	24 month	57%	0%	88%
WTP mapped to trial outcomes using per protocol analysis	24 month vs. risk based	–£6.46 (–£18.03 to +£5.11)	+0.007 (–0.102 to +0.117)	Dominant	–£304.53 (–£345.87 to –£263.20)	–£298.07	+£0.57 (–£18.28 to +£19.43)	£7.03	Risk based	16%	0%	12%
	24 month vs. 6 month	–£12.32 (–£30.94 to +£6.29)	+0.021 (–0.052 to +0.095)	Dominant	–£468.59(–£511.45 to –£425.72)	–£456.27	–£1.22 (–£20.62 to +£18.18)	£11.10	6 month	25%	100%	3%
	Risk based vs. 6 month	–£5.86 (–£21.85 to +£10.12)	+0.014 (–0.082 to +0.110)	Dominant	–£164.06 (–£196.36 to –£131.76)	–£158.20	–£1.80 (–£19.72 to +£16.13)	£4.06	24 month	59%	0%	86%
Scotland only	24 month vs. risk based	–£9.98 (–£17.44 to –£2.51)	+0.000 (–0.163 to +0.163)	Dominant	–£126.67 (–£184.99 to –£68.35)	–£116.69	–£11.92 (–£55.69 to +£31.86)	–£1.94	Risk based	34%	0%	30%
	24 month vs. 6 month	–£16.27 (–34.02 to +£1.48)	–0.000 (–0.095 to +0.095)	£162,700	–£229.56 (–£332.09 to –£127.03)	–£213.29	+£6.34 (–£33.22 to +£45.89)	£22.61	6 month	43%	100%	0%
	Risk based vs. 6 month	–£6.30 (–£25.79 to +£13.20)	–0.001 (–0.154 to +0.153)	£6,300	–£102.89 (–£196.35 to –£9.44)	–£96.59	+£18.26 (–£24.47 to +£60.98)	£24.56	24 month	23%	0%	70%
England only	24 month vs. risk based	–£9.62 (–£40.96 to +£21.72)	+0.048 (–0.098 to +0.194)		–£131.30 (–£190.82 to –£71.79)	–£121.68	+£12.51 (–£15.94 to +£40.95)	£22.13	Risk based	5%	22%	10%
	24 month vs. 6 month	–£15.67 (–£47.99 to +£16.64)	+0.063 (–0.064 to +0.190)	Dominant	–£170.20 (–£269.40 to –£70.99)	–£154.53	–£4.14 (–£36.90 to +£28.63)	£11.53	6 month	10%	78%	14%
	Risk based vs. 6 month	–£6.05 (–£31.95 to +£19.85)	+0.015 (–0.156 to +0.187)	Dominant	–£38.89 (–£111.66 to +£33.88)	–£32.84	–£16.64 (–£48.90 to +£15.62)	–£10.59	24 month	86%	0%	76%
Undiscounted costs and benefits	24 month vs. risk based	–£6.43 (–£18.23 to +£5.38)	+0.023 (–0.071 to +0.118)	Dominant	–£31.38 (–£43.57 to –£19.19)	–£24.95	+£0.91 (–£3.81 to +£5.62)	+£7.34	Risk based	17%	16%	15%
	24 month vs. 6 month	–£13.97 (–£33.15 to +£5.20)	+0.002 (–0.089 to +0.094)	Dominant	–£48.81 (–£68.23 to –£29.40)	–£34.84	+£1.21 (–£4.53 to +£6.95)	+£15.18	6 month	29%	84%	2%
	Risk based vs. 6 month	–£7.55 (–£24.51 to £9.42)	–0.021 (–0.109 to +0.067)	£359.52	–£17.43 (–£30.76 to –£4.10)	–£9.88	+£0.30 (–£5.19 to +£5.79)	+£7.85	24 month	54%	0%	83%
Complete–case analysisb	24 month vs. risk based	CUA:–£10.58 (–£31.51 to +£10.35) CBA: –£15.23 (–£36.23 to +£5.77) WTP (dental):–£15.23 (–£36.23 to +£5.77)	+0.036 (–0.101 to +0.173)	Dominant	–£111.63 (–£171.69 to –£51.56)	–£96.40	+£4.95 (–£19.78 to +£29.67)	+£20.18	Risk based	6%	0%	11%
	24 month vs. 6 month	CUA: –£14.49 (–£36.47 to +£7.50) CBA: –£26.45 (–£59.45 to +£6.54) WTP (dental): –£26.45 (–£59.45 to +£6.54)	–0.048 (–0.152 to +0.056)	£301.88	–£218.40 (–£305.18 to –£131.62)	–£191.95	–£3.02 (–£30.36 to +£24.32)	+£23.43	6 month	76%	100%	8%
	Risk based vs. 6 month	CUA: –£3.91 (–£25.54 to +£17.73) CBA: –£11.22 (–£45.81 to + 23.36) WTP (dental): –£26.45 (–£59.45 to +£6.54)	–0.084 (–0.210 to +0.042)	£46.45	–£106.77 (–£173.03 to –£40.52)	–£95.55	–£7.97 (–£31.22 to +£15.28)	+£18.48	24 month	17%	0%	81%

FIGURE 18.

Eligible for 24-month recall: (a) scatterplot of incremental costs and incremental QALYs; and (b) CEAC, for risk-based vs. 6-month recall (CUA).

FIGURE 19.

Eligible for 24-month recall: (a) scatterplot of incremental costs and incremental benefits (WTP); and (b) CBAC, for risk-based vs. 6-month recall (CBA). NB, net benefit.

FIGURE 20.

Eligible for 24-month recall: (a) scatterplot of incremental costs and incremental benefits (WTP); and (b) CBAC, for risk-based vs. 6-month recall (WTP for dental health outcomes). NB, net benefit.

For the analysis of trial data for those eligible for the 24-month recall, the 24-month interval is the least costly interval overall, and is significantly less costly than 6-month recall. For the CUA and WTP for dental outcomes frameworks, there is a 59% and a 86% probability that 24-month recall is the most efficient strategy when considering costs to the dental budget, driven by the potential for cost-savings without adversely affecting health outcomes (generic or dental). However, as DCE respondents highly value more frequent check-ups, the wider societal measure of benefit (CBA) remains favourable to 6-month recall.

In the eligible stratum, analysis frameworks focused on generic health (CUA) and dental health (WTP for dental outcomes) and both indicate that 24-month recall is the most likely recall strategy to deliver an efficient use of resources. The conclusion remains robust to the range of scenario analyses undertaken; however, broadening the scope of benefit valuation to include society’s WTP for services received, the optimal treatment decision would remain 6-month recalls. For the CBA, sensitivity analyses remain robust to this conclusion.

In summary, for those eligible for longer recall intervals, 24-month recall is, on average, the least costly interval for both the NHS and patients, and delivers significant cost savings when combining all costs of attending dental appointments (NHS and patient incurred). These cost savings are achieved with no meaningful differences in generic or dental health outcomes. From these perspectives, 24-month recall is the most likely strategy to generate positive net benefit; however, the general population places a significant value on more frequent dental checks, despite their additional cost and lack of clinical benefit. Taking a more holistic view of benefits, 6-month recall generates the greatest net benefits despite the potential for additional patient and NHS costs.

Comparison with other randomised clinical trials and studies

The Cochrane systematic review of dental recall, which was updated in 2013,72 reported insufficient evidence to determine the effect of different recall intervals, with only one study included. 125 The trial population of the only included study125 comprised children or adolescents aged 3, 16 or 18 years at trial entry. Wang et al. 125 included 185 patients and compared recall intervals of 12 months and 24 months over a 24-month follow-up period for the outcomes caries, as measured by decayed, missing and filled surfaces of primary teeth (dmfs) increment, and decayed, missing and filled surfaces of permanent teeth (DMFS) increment, and total time taken for examination and treatment. The authors of the Cochrane review concluded that a very low-quality body of evidence from one RCT was insufficient to reach any conclusions regarding the potential beneficial and harmful effects of varying recall intervals between dental check-ups. The Cochrane review Recall intervals for oral health in primary care patients was updated in 2020 by authors of the INTERVAL trial monograph, incorporating results from this study. 126 The updated search of 17 January 2020 identified the INTERVAL trial as the only new study eligible to be included in the updated review and the only trial to include patient participants > 18 years at trial entry. Given that different outcomes were reported in the two studies included in the updated Cochrane review, it was not possible to synthesise the data. The forest plots from the INTERVAL trial data are included to present the treatment effects across trial arm comparisons for the INTERVAL trial-reported outcomes (Figures 21–23).

FIGURE 21.

Forest plot for (a) the continuous outcomes; and (b) the outcome prevalence of extensive caries: risk-based vs. 6-month recall. M–H, Mantel–Haenszel.

FIGURE 22.

Forest plot for (a) the continuous outcomes; and (b) the outcome prevalence of extensive caries: 24-month vs. 6-month recall.

FIGURE 23.

Forest plot for (a) the continuous outcomes; and (b) the outcome prevalence of extensive caries: risk-based vs. 24-month recall. M–H, Mantel–Haenszel.

To our knowledge, this is the first RCT to fully integrate costs and benefits of different dental recall intervals and, therefore, provides the best available evidence on the short-term cost-effectiveness of different recall strategies proposed in the NICE guidance. 127 The NICE guidance was informed by a HTA report that developed a lifetime decision model to assess the cost-effectiveness of 3-, 6-, 12-, 18-, 24- and 36-month recall from a UK NHS perspective. 71 The report found that the incremental cost per decayed, missing, filled (DMF)-free tooth gained increased as recall intervals were narrowed; however, given that there is no evidence or precedence regarding society’s WTP for a DMF-free tooth gained, the results are difficult to interpret. Furthermore, the scope of costs was limited, and included only check-ups and treatment for decay, omitting any other dental treatments that may be triggered by recall. Finally, the results are of limited value to current decision-making as the payment system for dentistry has been updated in England and Wales, since this review was carried out, to the current banding system. To our knowledge, there are no other studies directly relevant to the cost-effectiveness of routine dental checks or oral health reviews in a UK setting.

Strengths

The INTERVAL trial was a pragmatic trial in primary care dental practice designed to provide evidence for the clinical and cost-effectiveness of variable recall intervals. The traditional view and ‘mantra’ of 6-month check-ups still exists among the majority of patients and dental professionals despite the introduction of recommendations decades ago for a risk-based and variable approach to reviewing patients. 3 The design of INTERVAL was complex and unique in order to answer the commissioning brief. A strength of the design was being able to compare three recall intervals and demonstrate the ability of dentists to distinguish between high- and low-risk routine dental attenders. INTERVAL included outcomes that were relevant and important to patients, clinicians and policy-makers. The design enabled and ensured that the most robust measures were collected including independent clinical assessments and access to national routine data.

Surprisingly, there have been few high-quality studies comparing different recall intervals, whether fixed or variable, according to risk. This is at a time when NHS resources are constrained, and there are demands for realistic and patient-centred care across health care. The published studies evaluating different recall intervals have involved few providers and most are in specialist settings that evaluate the impact of complex interventions usually directed at preventing caries in high-risk children. The only other trial undertaken was in a population of children and adolescents who received regular dental care in a public dental clinic in Norway from one dentist and one dental hygienist. 125 The trial inclusion criteria were less pragmatic than in INTERVAL, with individuals considered at high risk of dental caries excluded from the study. The commissioned call that INTERVAL answered presented a complex question that required a novel trial design. The strength of the design was to include two strata to avoid allocating participants to a 24-month recall for whom it was not clinically appropriate. A misinterpretation of INTERVAL may come about if the design is not understood. A significant strength is that dentists in INTERVAL applied their professional and clinical judgement before allocating a patient participant to the stratum that included randomisation to 24-month recall.

A strength of INTERVAL was that it was a pragmatic trial allowing the dentists and patient participants to respond to changes in circumstance. Changes in circumstance could have been logistics for allocating a recall appointment, or not attending as scheduled because of holidays, or responses to change in clinical condition. We predicted that the participants allocated to a 6-month recall would be most likely to attend at 9 months, thereby taking into account the added duration of any course of treatment following the previous check-up date and adjustments for the logistics or general life. A strength of INTERVAL is that dentists did implement a different recall interval for the patient participants randomised to the risk-based group in both strata of the trial. This overcame a potential risk at the outset that we might not observe a difference in the number of check-ups experienced; however, a surprising observation was that the time interval for the risk-based group was so close to the 6-month recall group. In addition, for patients whom dentists felt were suitable not to be seen, unless needed, for 24 months, dentists allocated a risk-based recall interval in between 6 months and 24 months, which was, however, not closer to the latter. A further strength of INTERVAL is that because of its pragmatic nature it has demonstrated that dentists can accurately and reliably assess risk as shown by the difference in the recall behaviours of the two strata.

The strengths of the INTERVAL trial include the recruitment and retention of a large number of centres (n = 51) and, to our knowledge, is the first dental trial to involve all the nations of the UK. As the nations operate in different contractual systems, INTERVAL provides insight to the potential impact that the system has on clinical decision-making and practice systems. The dental practices represent a wide range of geographical locations including remote and rural settings, and a range of characteristics including the number of dentists (average three), and whether or not they were a training practice, or employed at least one hygienist (70%). INTERVAL was not powered to detect differences across contractual systems and the numbers from Northern Ireland are too small to make any conclusions, although there appears to be a difference in the determination of risk-based recall interval for dentists in England and Scotland.

Recruitment and retention in trials is a frequent challenge, but an achievement for the INTERVAL trial was the recruitment of 2372 of the potentially eligible patients, and their retention with provision of data at 4 years from 1765 participants with 69% providing clinical data. This may be as a result of the participants being routine dental attenders and, hence, interested in dental health, and it was possibly their first invitation to participate in research. The randomised groups in both strata were balanced/similar at baseline and the reasons for loss to follow-up were related to inability of the practice to contact its patients; therefore, we were confident in the robustness of the results regardless of the missing data. The dental behaviour and clinical characteristics of the participants mean that the findings of the INTERVAL trial are generalisable to regular dental attenders across the UK in the NHS, and, therefore, also in similar third-party funding systems. Applying a risk-based determination of recall interval is advocated in many dental health-care systems and hence the findings of INTERVAL have relevance internationally.

The INTERVAL trial found no evidence of a difference in oral health from a risk-based recall interval compared with the traditional 6-month check-up. This also applied to patient participants whom the dentist assessed as low risk and suitable for a check-up at 24 months. For this group, the INTERVAL trial found no evidence of a difference between 24-month, risk-based or 6-month recall. The results were robust to sensitivity analyses exploring the impact of missing data.

The pragmatic design of the trial did not prevent fidelity to the interventions with separation in the frequency of dental check-ups between groups. We have evidence of intervention fidelity though monitoring intended check-up appointments. Dentists participated in training and were asked to revisit the e-learning tool, although compliance was variable. Conducting the trial in a primary care NHS setting with a design that involved minimal requirements for the dental practices, blind outcome assessment and access to routine data are strengths. The process undertaken to inform the determination of a risk-based recall was not independently observed (i.e. information collection, assessment and consultation with the patient); however, we believe the dentists performed this intervention as instructed in a way with which they were confident. The evidence from the dentist experience questionnaire supports the view that dentists who randomised participants to the 24-month stratum increased in confidence and were more likely to adopt/maintain that behaviour. In contrast to those who did not, their concern/anxiety increased by the end of the trial.

There was no evidence that different recall intervals made a difference to measures of patient quality of life, anxiety and satisfaction with care. The strength of INTERVAL was that these measures were collected annually during the 4 years of the trial. The reaction of some to them being considered suitable to have a recall interval of 24 months was an indication of confidence in their ability to maintain their oral health.

The health economic analysis is based on best practice methodology for conducting economic evaluation alongside RCTs. The analysis provides the only evidence that is relevant to decision-making in the UK NHS regarding the most efficient dental recall interval, and is based on the current, most up-to-date payment systems across the UK regions. The analysis is novel in implementing a CBA considering both WTP for dental health outcomes and wider societal benefits. The use of a DCE to capture the value placed on both the trial interventions and important dental health outcomes by the general population is a distinct advantage and overcomes the problems with interpreting economic evaluation results from other studies. The approach ensures consistency of methodology with recent economic evaluations of routine dental interventions (e.g. the IQuaD study101). A further advantage of CBA is the simultaneous valuation of multiple dental health outcomes within a single outcome measure (i.e. bleeding gums and dental decay), representing an important advantage over more traditional, single outcome measure cost-effectiveness studies (e.g. decayed teeth only) typically conducted in dentistry.

In terms of the within-trial analysis, we used best practice methodology, incorporating the most advanced recommendations for analysis with the appropriate use of missing data models to minimise the potential for bias.

Limitations

The INTERVAL trial experienced challenges throughout its course, some of which were anticipated and others of which could not have been predicted. Recruitment of dentists and participants was a challenge which was overcome, but this led to a considerable delay in the trial. The lack of research-experienced practices and the fact that the dentists were responsible for recruiting and obtaining consent from patient participants had an impact on our ability to reliably predict recruitment. INTERVAL was the first in a series of trials that the research team have been involved with, and experience has shown that a weakness was not having the research nurses supporting practices in person at dedicated recruitment sessions to overcome this. We did not collect information on the total number of routine patients approached as practices operated a process that complemented their practice management system; therefore, the reasons why patients chose not to participate in INTERVAL are unknown.

We experienced delays in analysis due to challenges encountered obtaining routine data. The process and mechanisms for establishing permission for access to routine data did not exist for England and Wales. Retrieving the routine data delayed analysis; the reasons for this included there not being a system for information retrieval or sharing of dental data at NHS services in England and Northern Ireland prior to INTERVAL.

A weakness was not collecting information on what factors were considered to inform the risk-based interval and the interaction with the patient to both determine risk and negotiate the interval. A possible weakness was the choice of the primary outcome as gingival bleeding on probing for which calibration is not possible. The rationale for this decision was that it is a measurable and modifiable outcome that indicates oral health in general. INTERVAL evaluated different recall intervals on maintaining oral health and no difference in this disease measure would reflect the absence of either improvement or deterioration. The other clinical outcomes were recorded, but the team’s experience and previous research confirmed that the progression of caries and periodontal disease was going to be too slow in this population.

INTERVAL had a drop-out rate of 25–30% in the questionnaire data and 30–37% in attendance at appointments, a higher value than expected at the design stage of the trial. However, the 95% CIs for risk-based versus 6-month recall and risk-based versus 24-month recall precluded our predefined clinical minimally important differences in gingival bleeding of 4.5% and 7.5%, respectively. In addition, the drop-out rates were balanced between the arms, and sensitivity analyses using multiple imputation showed that the results were robust.

The analysis is based on 4-year outcomes from the trial. It is important to extrapolate trial results over a longer-term time horizon in order to fully capture all the costs and outcomes associated with different interventions. This was outwith the scope of the current study. There is no evidence from the trial to suggest differences in caries experience or bleeding across the different recall intervals, and it is therefore unlikely that a decision model would lead to different conclusions regarding the most efficient allocation of resources. There is a lack of good-quality information in dentistry to inform decision analysis models, for example estimates of long-term baseline transition probabilities to determine the life course of a tooth through different stages (well, decayed, root canal treatment required, missing). Further research is urgently required to bridge this gap in the evidence base and determine the economic value of long-term caries prevention.

There are different payment systems across the different UK regions. Dental check-ups are free of charge in the fee-for-service system in Scotland, there are patient co-payments in the Northern Ireland fee-for-service system, whereas England and Wales adopt a treatment banding approach. The method of payment and the cross-region variation raise three specific issues for the economic evaluation:

1. The payment for dental care services is not necessarily a true reflection of the opportunity cost of the resource required to deliver dental services. Owing to the practical barriers to conducting a detailed micro-costing exercise, and the vast number of different possible treatments in primary dental practice, we were not able to directly elicit opportunity costs in this study.

2. The different payment systems across the regions, and the process of submitting dental claims means that the resource use data collected from the practices varies substantially, which limits comparability. The quality of the data, in particular the detailed treatment information (e.g. check-ups and fillings) obtained from the routine claims data, could also potentially vary across countries.

3. The different systems may induce different provider and consumer behaviours, and different regional-specific incentives make direct comparability of results across the regions difficult. However, results are unlikely to be biased at the average UK level because the proportion of practices randomising to each group was approximately equal across the different regions.

Clinical assessment

Overcoming the practical challenge of arranging the final clinical assessment visits for participants at 51 dental practices across the four nations of the UK relied on trust, respect and partnership working between the research team and the general dental practices involved. Patient participants volunteered around 30 minutes of their time for the clinical assessment by one of the blinded assessment teams. The attendance of 1624 participants for their clinical assessments demonstrates their interest, commitment and the value placed on seeking evidence for routine oral health care and the role of practice-based dental research. The robust training undertaken prior to data collection and throughout the assessment period ensured consistency between assessors. The fact that calibration was not possible does not limit the findings, and we believe that no more could have been done to ensure reliable measures.

Patient and public involvement and engagement

Prior to the start of the trial, advice on the design and conduct of the study was sought from members of the Public Partnership Groups and from similar patient groups in other parts of the UK sourced under guidance from INVOLVE.

These independent public partnership groups comprise volunteers who work in partnership with NHS Tayside and aim to provide a conduit for the views of people about their local services.

Patient advisors were a valuable resource at the outset of the trial and they helped to ensure good conduct and patient-friendly practice throughout the duration of the trial. Patient advisors were involved with the trial design and provided invaluable feedback on trial recruitment and communication strategies. Patient advisors also contributed to the content and layout of the trial invitation, trial newsletters and the design of patient participant questionnaires. This ensured that trial participants could understand and easily complete these materials. This input was obtained remotely and from individuals. In future studies, we would recommend working with a group with some face-to-face interaction.

As quality of life was a primary outcome of the trial, patient advisors’ input to the proposed questionnaire design was essential. Qualitative work with patients was carried out to ensure that the outcome measures were patient-centred.

Lay representatives on the TSC and TMC actively contributed to trial oversight, processes and procedures, including helping to interpret the trial findings, preparation of the monograph, and assisting with the review of the Plain English summary.

The results were discussed with the four Chief Dental Officers of the UK to enable early consideration of the implications for policy and practice.

Generalisability

The INTERVAL trial was designed pragmatically to investigate the effectiveness and cost–benefits of three different recall intervals. The 51 recruiting dental practices were situated in England, Wales, Northern Ireland and Scotland. On average, participating dentists had been qualified for 16 years and had an approximate total list size of 5000 patients. The average number of dentists in their practices was three and approximately 70% of practices employed at least one hygienist.

As at 30 September 2018, 94% of the adult population in Scotland were registered with an NHS dentist. Of these, 67% had attended their dentist at least once in the previous 2 years. The Adult Dental Health Survey 2009114 reported that over half of the population in England, Wales and Northern Ireland had attended a dental practice within the last 3 years. The INTERVAL trial recruited 2372 patient participants who were regular attenders and had attended the dentist at least once in the previous 2 years, with 43% recruited in England and Wales, 7% in Northern Ireland and 50% in Scotland. 128

At baseline, the average age of participants was 45 years, 55% were female, 15% identified as smokers and 55% used manual toothbrushes.

We are confident that the practices and participants recruited are a true representation of adults who attend NHS dental practices across the UK.

Recommendation for research

Further research is required to:

• enhance and support the communication of a variable risk-based recall interval to patients and the dental team, and the co-development of risk assessment tools

• develop recall interval quality management and improvement tools

• understand the role of risk-based recall and risk management of adults who are irregular dental attenders

• better understand public perceptions of the value of 6-month fixed-period recall intervals to help elucidate barriers to and facilitators of change

• inform methods for long-term decision modelling in dentistry utilising long-term dental cohort data where possible

• better utilise routine claims data sets, including tooth-level treatment data, to create cohorts of data to follow and predict individuals’ health and to understand and improve care pathways in dentistry.

Recruitment sites

We would like to thank the staff and participants of the following dental practices:

Aesthetics Dental & Implant Surgery, Amblecote Dental Care, Anita Belbin Dental Surgery, Bayview Dental Practice, Bridge Dental Centre, Brightons & Polmont Dental Practice, Brunswick Dental Practice, Bute Dental Surgery, Castle House Dental Practice, Chong Kwan Dental Centre, Church Road Dental Care, Colin C Yule & Associates, Collegiate Dental Practice, Cottam’s Dental, Implant & Orthodontic Practice, Craigmillar Dental Centre, Discovery Dental Care, Dunbar Dental Practice, Eastside Dental Practice, Green Lane Dental Surgery, Hazel House Dental Surgery, Horizon Dental Clinic (Blyth), Horizon Dental Clinic (Monkseaton), Kelvingrove Dental Care Ltd, Kingsway Dental Practice, Lochshell Dental Clinic, Loughside Dental Practice, M&S Dental Care, Montgomery Street Dental Care, Montrose Dental Practice, Mount Florida Dental, My Dental Care, N13 Dental Clinic, Nairn Dental Clinic, Number One Surgery, Park View Family Dental, PCK Wee Dental Surgery, PG Lawson & Associates, Platt & Common, Roseville Dental Practice, Salmon Lane Dental Care, Shotley Bridge Dental Care, Six Gables Dental Practice, Smiledent Dental Practice, Southside Dental Care, St Leonard’s Dental Practice, Walmley Dental Practice, Windsor Dental Practice (Manchester), Windsor Dental Practice (Salford), Woodside Dental Practice and Young Smile Dental Care.

Independent members of the Trial Steering Committee

Robert Elford, Trevor Johnson, Edwina Kidd (former chairperson), the late James Steele (chairperson) and Elizabeth Treasure (chairperson).

Members of the Trial Management Committee

Tony Anderson, Debbie Bonetti, Trevor Burke, Philip Dolan, Mark Forrest, Ronald Gorter, Richard Herbert, Penny Hodge, Dirk Mettes, Wendy McCombes, Ian Needleman, Margaret Ross and Debbie White.

Independent members of the Data Monitoring Committee

Martin Chalkley, Chris Deery and Simon Gates (chairperson).

Ethics approval

Ethics approval for the trial was given by the Fife and Forth Valley Research Ethics Committee on 13 January 2009 and 20 September 2011 (Research Ethics Committee reference 09/SO501/1).

The trial was registered with the ISRCTN (reference number ISRCTN95933794).

Funding

The University of Dundee agreed to act as sponsor for the research and the study was adopted by the Scottish Dental Practice Based Research Network and the relevant local research networks in England, Wales and Northern Ireland.

Service support was provided by a central subvention from the NIHR and the local primary care trusts and health boards.

Contributions of authors

Jan E Clarkson (https://orcid.org/0000-0001-5940-2926) (Professor and Joint Chief Investigator) contributed to the conception and design of the trial, the conduct of the trial, the recruitment of dentists, the interpretation of the results, and the drafting, editing and approval of the monograph.

Nigel B Pitts (https://orcid.org/0000-0001-6184-4213) (Professor and Joint Chief Investigator) contributed to the conception and design of the trial, the conduct of the trial, the recruitment of dentists, the interpretation of results, and the drafting, editing and approval of the monograph.

Beatriz Goulao (https://orcid.org/0000-0003-1490-7183) (Statistician) developed the statistical analysis plan, led the statistical analyses, contributed to the interpretation of the data, made a significant contribution to the drafting of Chapter 4, and reviewed and approved the monograph.

Dwayne Boyers (https://orcid.org/0000-0002-9786-8118) (Health Economist) conducted the economic analysis, contributed to the interpretation of the data, drafted Chapters 3 and 5, and reviewed and approved the monograph.

Craig R Ramsay (https://orcid.org/0000-0003-4043-7349) (Professor) contributed to the conception and design of the trial, the conduct of the trial, oversaw trial methods, the interpretation of the results, and the drafting, editing and approval of the monograph.

Ruth Floate (https://orcid.org/0000-0003-2424-4643) (Trial Manager) was responsible for the day-to-day management of the trial, contributed to the interpretation of results, and the drafting and editing of the monograph.

Hazel J Braid (https://orcid.org/0000-0002-2062-8546) (Trial Administrator) was responsible for the day-to-day management of the trial, contributed to the interpretation of results, and the drafting and editing of the monograph.

Patrick A Fee (https://orcid.org/0000-0002-7188-3416) (Clinical Research Fellow) was responsible for the conduct of the trial, contributed to the collection of clinical outcomes, the interpretation of results, and the drafting and editing of the monograph.

Fiona S Ord (https://orcid.org/0000-0002-9901-7739) (Research Hygienist) was responsible for the recruitment of participants, contributed to the collection of clinical outcomes, and contributed to the drafting and editing of the monograph.

Helen V Worthington (https://orcid.org/0000-0002-4851-7469) (Professor and Triallist) contributed to the design of the trial, the interpretation of results and reviewing and approving the monograph.

Marjon van der Pol (https://orcid.org/0000-0003-0636-1184) (Professor) contributed to the conception and design of the health economics analysis, oversaw health economic methods, and contributed to editing the monograph.

Linda Young (https://orcid.org/0000-0002-6004-3374) (Research Manager and Economist) contributed to the design of the trial, the interpretation of results, and drafting and editing of the monograph.

Ruth Freeman (https://orcid.org/0000-0002-8733-1253) (Professor) contributed to the design of the trial, the interpretation of results, and reviewing and approving the monograph.

Jill Gouick (https://orcid.org/0000-0002-1456-8241) (Research Dental Nurse) was responsible for the collection of clinical outcomes, and contributed to reviewing and approving the monograph.

Gerald M Humphris (https://orcid.org/0000-0002-4601-8834) (Professor) contributed to the design of the trial, the interpretation of results, and reviewing and approving the monograph.

Fiona E Mitchell (https://orcid.org/0000-0002-7922-0676) (Research Dental Nurse) was responsible for the collection of clinical outcomes, and contributed to reviewing and approving the monograph.

Alison M McDonald (https://orcid.org/0000-0002-0256-2889) (Senior Trials Manager) was responsible for overseeing management of the trial and contributed to the interpretation of data, and to reviewing and approving the monograph.

John DT Norrie (https://orcid.org/0000-0001-9823-9252) (Professor and Triallist) contributed to the design of the trial, the interpretation of the data, and reviewing and approving the monograph.

Kirsty Sim (https://orcid.org/0000-0002-2365-5013) (Research Hygienist) was responsible for the collection of clinical outcomes, and contributed to reviewing and approving the monograph.

Gail Douglas (https://orcid.org/0000-0002-0531-3909) (Professor) contributed to the design of the trial, the interpretation of results, and reviewing and approving the monograph.

David Ricketts (https://orcid.org/0000-0002-9851-1778) (Professor) contributed to the design of the trial, the interpretation of results, and reviewing and approving the monograph.

Publication

Clarkson JE, Pitts NB, Bonetti D, Boyers D, Braid H, Elford R, et al. INTERVAL (investigation of NICE technologies for enabling risk-variable-adjusted-length) dental recalls trial: a multicentre randomised controlled trial investigating the best dental recall interval for optimum, cost-effective maintenance of oral health in dentate adults attending dental primary care. BMC Oral Health 2018;18:135.

Data-sharing statement

All available data can be obtained by contacting the corresponding author.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.