Prophylactic removal of impacted mandibular third molars: a systematic review and economic evaluation

Aetiology, pathology and prognosis

Impacted third molars are classified on the basis of their location (mandibular or maxillary), eruption status, nature of impaction, angulation of impaction and depth of impaction relative to the adjacent tooth. An impacted tooth can be visible in the mouth, can be explored with a periodontal probe or may be observed only through radiographic assessment. 5 Eruption status is described in Table 1.

The nature of the impaction can be when the tooth is covered only by soft tissue and is referred to as ‘soft-tissue impaction’. The tooth can also be covered by bone; this is known as ‘partial bony impaction’ when partially erupted or as ‘complete bony impaction’ when unerupted and not communicating with the mouth.

The nature of angulation can be based on Winter’s classification,6 whereby the 3M could be:

• mesioangular [angled towards the second molar (2M)]

• distoangular (angled away from the 2M)

• horizontal

• vertical

• buccal (angled towards the cheek)

• lingual (angled towards the tongue).

Based on Pell and Gregory’s7 classification relating to depth, the I3M can be class 1, 2, or 3 according to the amount of tooth covered by the mandibular ramus, or A, B, or C depending on the depth of the impacted tooth compared with the 2M.

Natural history with no treatment

Little is known about the natural history of I3Ms left in situ. This is due, in part, to the historical routine extraction of I3Ms, which means that we have limited data on which to make reliable estimates of the onset of pathology when the asymptomatic teeth are left in place. 3 Collecting the required data is also problematic in the UK, as clinical reporting systems are not sensitive enough to capture information relating to 3M management. 16 In addition, it would be costly to conduct a non-interventional/observational study to gather data on untreated I3Ms, as it would take decades because of the size of the study cohort needed to determine the occurrence of pathological conditions. 4

Epidemiology

The prevalence of I3Ms in the UK is unknown. Internationally, the prevalence of I3Ms is reported to range from 18% to 68%. 17 According to the results of a 2016 meta-analysis18 of 49 studies (83,484 individuals), the prevalence of 3M impaction worldwide in individuals aged > 17 years is 24.4% (95% CI 18.97% to 30.80%). The authors18 also found that the risk of having IM3Ms was higher than having impacted maxillary 3Ms (57.6%, 95% CI 43.3% to 68.3%; p < 0.0001), and that there was no difference in the incidence of impaction between men and women (18.6%, 95% CI –4.9% to 48.0%; p = 0.12). The most common angulation of impaction was found to be mesioangular (41.2%, 95% CI 33.8% to 49.0%).

The UK National Third Molar project8 was a cross-sectional survey that was set up in 1997 to assess the management of 3Ms in UK clinical practice. Clinical data were collected prospectively from all of the patients who were referred for assessment of 3Ms to oral and maxillofacial consultant surgeons during July 1995. 8 Completed questionnaires were returned from 181 consultants and 8298 patients (with 25,001 3Ms) who were referred to a hospital for assessment. Details of the eruption and symptom status of all 3Ms at the time of presentation are shown in Table 2. In the data available/recorded, the majority of mandibular third molars (M3Ms) (69.4%) were impacted, whereas maxillary 3Ms were more likely to be classified as ‘present and functional’ or ‘absent’.

The authors of the study8 reported that, after assessment, a total of 19,971 (80%) of the 25,001 3Ms were extracted and M3Ms were more likely to be extracted than maxillary 3Ms (87% vs. 71%, respectively). The most frequent indication for extraction was prophylactic removal (n = 8772, 44%), followed by pericoronitis (n = 7896, 40%). There were differences in the rates of and reasons for extraction between mandibular and maxillary 3Ms: 22% of M3Ms were extracted prophylactically compared with 79% of maxillary 3Ms, whereas 60% of M3Ms were removed as a result of pericoronitis, compared with 8% of maxillary 3Ms.

The results of the UK National Third Molar project8 showed that the most common age for the removal of 3Ms was between 21 and 25 years. However, in 2012, McArdle and Renton19 reported that the mean age of patients having 3Ms removed had increased from 25 years in 2000 to 32 years in 2010, with the most common age increasing from 26 to 29 years (CIs not reported).

Impact of the health problem

Prior to the introduction of the NICE guidance in 2000,2 the removal of 3Ms was one of the most common of all surgical procedures performed in the UK, with > 36,000 inpatient and 60,000 day-case admissions for ‘surgical removal of tooth’ in the period 1994–5. 20 During this period, the cost to the NHS in England of 3M surgery was estimated to be £30M per year, with additional estimated costs of £20M in the private sector. 20

The authors of a 2012 study19 investigating the effects of the NICE guidance2 on the management of 3Ms reported that, since the introduction of the NICE guidance,2 the number of 3M removals in secondary care (inpatient/day case) had reduced from ≈ 60,000 in the 1990s to ≈ 40,000 in 2003. However, since 2003, the number of removals appeared to have increased to ≈ 65,000 during 2009/10 (inpatient/day case only).

Information provided to NICE in the British Dental Association (BDA) (McArdle LW, British Dental Association, 2016, NICE submission) and the Faculty of General Dental Practice (FGDP) (Renton T, FGDP, 2016, NICE submission) submissions suggested that the prophylactic removal of 3Ms prevents future harm to patients. These organisations argue that the introduction of the NICE guidance2 initially resulted in a reduction in the number of 3Ms extracted; however, this figure has since increased. It is argued that, irrespective of the NICE guidance in 2000,2 the need for surgical extraction was not negated but postponed until a later date. It is further argued that patients aged > 25 years are at a higher risk of surgical morbidity relating to 3M extraction (Renton T, NICE submission). Another possible explanation for the increase in 3M extractions could be that patients who may have more than one I3M undergo multiple treatment episodes as and when other 3Ms become problematic (McArdle LW, NICE submission).

Management of disease

The treatment options for people with I3Ms are surgical removal or standard care without prophylactic removal of 3Ms.

Variation in services and/or uncertainty about best practice

Internationally, there is a vast quantity of published literature relating to the management of 3Ms and many published international guidelines with recommendations for best practice relating to asymptomatic or disease-free 3Ms. However, there is still debate and it remains a contentious subject. According to the FGDP submission (Renton T, NICE submission), there are differences of opinion between professionals in the UK relating to best practice. However, the submission authors assert that most UK dentists believe that erupted, non-functional, low-risk M3Ms should be removed at a young age to prevent increased surgical morbidity in older age and to prevent future harm to the patient (Renton T, NICE submission).

There is significant geographical variation in current practice when international guidelines are examined. 21,27–34 The American31 guidelines recommend a more interventional approach to 3M management. In the UK NHS setting, there is a ‘no-intervention’ policy unless there are distinct therapeutic indications, although there are differences of opinion between professionals.

There is variation between services relating to the use of either general anaesthesia or local anaesthesia and sedation. There are published data35 that illustrate that only 3% of IM3M cases in a London teaching hospital required general anaesthetic, with 40% of cases requiring intravenous sedation. However, our clinical advisor has pointed out that not all district general hospitals offer sedation services for dental extractions; therefore, the proportion of patients who receive higher-risk general anaesthetic is greater. In terms of service provision, many dental practices in the UK do not provide intravenous sedation, which results in patients of these practices being referred to hospital to undergo surgical extraction under general anaesthetic. There is also considerable variation in the perioperative care provided; for example, the provision of informed consent, the provision of patient information, the provision of pre-operative mouth rinses, the provision of analgesia and the rates of antibiotic prescription (Renton T, NICE submission).

Relevant UK guidelines

The NICE technology appraisal 1 (TA1)20 was completed in 2000; the resultant NICE guidance2 informed that the prophylactic removal of pathology-free I3Ms was not recommended (Box 1).

A review of the existing NICE guidance,2 via a review proposal in 2014, concluded that no new trial data on this topic were available. As a result, a decision was made that the NICE guidance2 did not need to be revisited and that the topic should remain on the static list. However, as the recommendations set out in the NICE guidance2 were increasingly being perceived as controversial by the dental profession, a NICE consultation with relevant stakeholders was then undertaken. Consultation responses highlighted that additional pertinent trial data were available, and therefore should be assessed. In response, NICE instructed that the current guidance2 should be partially updated (i.e. prophylactic indications only) via the multiple technology appraisal (MTA) process.

Summary of the intervention

The surgical extraction of IM3Ms with evidence of pathology (see Box 1) can be undertaken in primary care, secondary care and specialist clinics. The NHS commissioning oral surgery pathway36 clearly outlines social, medical and dental factors that dictate the optimal setting (Renton T, NICE submission).

Specialist radiographic equipment and assessment may be required for risk assessment of IM3Ms, including panoral radiography and cone beam computerised tomography, which requires the input of a radiologist. For a patient who requires sedation (primary care), specialist nursing is required. Intravenous sedation services require additional staff training, the correct facilities and indemnity costs.

Identification of important subgroups

There is intrapatient variance in the presentation of I3Ms, that is a single patient can have multiple 3Ms (i.e. maxillary as well as mandibular and bilateral presence) with different types of impaction (i.e. vertical, horizontal, distoangular and mesioangular). These are the most common impaction types considered as subgroups, although a smaller proportion of patients may have ectopic impactions. The variability of 3M impactions results in different secondary disease distribution, which is dependent on the nature of the impaction (McArdle LW, NICE submission).

Patients with high-risk M3Ms (the roots cross the inferior dental canal) could be 10 times more likely to develop temporary or permanent inferior alveolar nerve injury (Renton T, NICE submission).

What is not included in the assessment

It is beyond the remit of this assessment report to comment on or to draw conclusions relating to the wider topic of the management of 3Ms; this assessment report focusses primarily on summarising the relevant evidence relating to the surgical extraction or retention of asymptomatic IM3Ms.

It is worth noting that the aims of the original assessment report conducted by Song et al. ,20 which contributed to the NICE guidance2 issued in 2000, were not exactly the same as the aims of this assessment, which is a partial update of TA1. Song et al. 20 aimed to ‘provide a summary of existing evidence on prophylactic removal of impacted wisdom teeth, in terms of the incidence of surgical complications associated with prophylactic removal, and the morbidity associated with retention.’

Identification of studies

Inclusion and exclusion criteria

Two out of the three reviewers (JH, GP and RD) independently screened all of the titles and abstracts identified by the initial search using Covidence (Melbourne, VIC, Australia). Full-text copies of any titles/abstracts that may have been eligible were obtained and assessed for inclusion by two reviewers (JH and GP) according to the inclusion and exclusion criteria listed in Table 4. Discrepancies were resolved by consultation with a third reviewer/clinical advisor. Studies that did not meet the inclusion criteria were excluded and the reasons for exclusion were summarised. For studies that were identified as not meeting the criteria at the data abstraction stage, the bibliographic details and reasons for exclusion were summarised.

The eligibility criteria reflected the decision problem from NICE;11 additional criteria were that studies were based in settings with a similar dental system to that in the UK (i.e. Europe, North America, Australasia), English-language papers (owing to a lack of resources to translate non-English-language papers) and the time frame of 1999 onwards, because this was a partial update of the previous NICE guidance. 2 The inclusion criteria for SRs were necessarily broader and were not restricted to those looking at IM3Ms only.

Data abstraction strategy

Data relating to study characteristics and outcomes were extracted by one of two reviewers (Joanne Fisher or Juliet Hounsome) and were independently checked for accuracy by a second reviewer (Joanne Fisher or Juliet Hounsome). Disagreement was resolved through consensus; when necessary, a third reviewer was consulted. Study data that were reported in multiple publications were extracted and reported as a single study.

Critical appraisal strategy

The quality of the included studies was assessed by one reviewer (Juliet Hounsome or Joanne Fisher) and independently checked for agreement by a second reviewer (Juliet Hounsome or Joanne Fisher). Disagreements were resolved through consensus. The quality of the cohort studies was assessed using an adapted version of the Newcastle–Ottawa Scale for assessing the quality of cohort studies37 and SRs were assessed according to criteria outlined by the Centre for Reviews and Dissemination. 38

Methods of data synthesis

The results of the data extraction and quality assessment for each included study are presented in structured tables (see Tables 5–9 and Appendices 3 and 4) and as a narrative summary (see following section).

Quantity and quality of research available

The results of the electronic searches (conducted on 4 April 2016) and the application of the inclusion criteria are shown in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram in Figure 1.

FIGURE 1.

The PRISMA flow diagram: clinical evidence review.

In total, 22 citations,20,39–59 reporting the results of nine SRs20,40,42–46,50,56 and four cohort studies,47,48,58,59 were included in the review. No randomised controlled trials (RCTs) were identified.

The reasons for excluding papers at the full-text review stage are summarised in Appendix 2.

As shown in Figure 1, a total of 72 papers were initially included at full-text review. However, on further inspection, no relevant data for the specific population of interest for this review were available for 50 papers; these papers were subsequently excluded. The bibliographic details, with the reasons for exclusion, for these 50 studies are also reported in Appendix 2.

One study58 reported on outcomes both for standard care with and standard care without prophylactic removal of IM3Ms. Four papers39,47,48,57 reported on two studies that assessed the outcomes of standard care without prophylactic removal of IM3Ms, and one paper59 reported on a study that assessed prophylactic removal of IM3Ms. A further 16 papers20,40–46,49–56 reported on nine SRs that assessed whether or not I3Ms should be removed prophylactically.

Assessment of the effectiveness of the included studies

Additional evidence

Search strategy

The search strategy developed for the clinical searches (see Appendix 1), with the addition of an economics filter, was used to identify studies reporting the costs and benefits associated with extracting/retaining I3Ms. As part of the search strategy, NHS EED, which is located within The Cochrane Library, and EconLit (EBSCOhost) were also searched. All databases were searched on 29 April 2016. The results were entered into an EndNote X7.4 library, de-duplicated and exported into Covidence.

Informal searching activities were carried out to identify economic evaluations relevant to the decision problem. These included contacting experts in the field and a search of Google Scholar (Google Inc., Mountain View, CA, USA). The Google Scholar search was updated on 1 February 2017 and revealed no relevant results.

The two clinical submissions from professional stakeholders that were submitted to NICE as part of the MTA process were also checked for cost-effectiveness data.

Study selection and inclusion criteria

Studies were selected based on their relevance to the decision problem and on the specific economic criteria displayed in Table 10. Two reviewers (AB and SB) independently examined the titles and abstracts of all of the studies identified by the search to find potentially eligible publications (stage 1). In the next stage (stage 2), two reviewers (AB and SB) examined the full texts of studies that were identified as being potentially relevant at stage 1. During stage 2, two modifications were made to the inclusion criteria:

1. Owing to limited information about UK costs being available, studies that included any costs were included in the review.

2. Papers reporting short-term health-related quality-of-life (HRQoL) outcomes were excluded from the review (i.e. papers with only long-term HRQoL outcomes were included) to align HRQoL outcomes with the outcomes reported in the clinical papers.

Disagreements about inclusion were resolved through discussion and, in all cases, a consensus was reached; it was, therefore, not necessary to consult a third reviewer during the screening and selection process.

Quantity of included evidence

From the main searches, the AG identified 493 potentially relevant papers for inclusion in the review of economic evidence. Of these, 34 papers were considered for inclusion after stage 1. As shown in Figure 2, eight studies17,20,71–76 were initially included at stage 2. However, on further inspection, five71–75 of the eight studies did not include information that was relevant to the population of interest; these papers were, therefore, subsequently excluded from the review. Bibliographic details and summary data from these five studies71–75 are available in Appendix 2. Of the 34 papers considered for inclusion after stage 1, 31 were excluded during stage 2, which left three papers17,20,76 to be included in the review. The reasons for excluding the 31 studies are listed in Table 11.

FIGURE 2.

The PRISMA flow diagram: economic evidence review.

The three papers17,20,76 that were included in the review are listed in Table 12. Two studies17,20 provided information on costs and one study76 provided information on patient HRQoL. The characteristics of these studies are presented in Table 13.

Quality of the included evidence

Contrary to the review protocol, the AG made the decision not to quality assess the papers included in the review of cost-effectiveness evidence using a cost-effectiveness checklist. This decision was made because only one paper20 directly considers the cost-effectiveness of prophylactic removal of I3Ms in a UK setting [and a summary of this paper and its quality are located in the clinical evidence section of this report (see Tables 8 and 9 and Appendix 4)].

Economic review: overview of included papers

The AG concludes that relevant data on I3Ms are limited to three studies. 17,20,76 Two of the papers report details about the cost-effectiveness of the prophylactic removal of I3Ms. The review by Song et al. 20 includes details about the cost-effectiveness from a UK NHS perspective, whereas the material presented in the study by Anjrini et al. 17 is of less direct relevance as estimates are based on the Australian health-care system and the results are presented in Australian dollars. The third paper76 reports findings that relate to an assessment of oral HRQoL after the removal of I3Ms.

Key results: cost-effectiveness of prophylactic removal of impacted third molars

Cost-effectiveness review: conclusions

As there is very limited clinical effectiveness evidence comparing the prophylactic removal of I3M with a ‘watchful waiting’ strategy, it is unsurprising that economic evidence relating to this comparison is also limited. There are only two published cost-effectiveness studies17,20 that directly consider this comparison and, in both cases, the authors conclude that there is currently no economic evidence to support the prophylactic removal of I3Ms. However, Song et al. 20 restrict their conclusion to a short- to medium-term time frame.

Model pathways

The elements of both the intervention (prophylactic removal) and the comparator (watchful waiting) pathways were determined through consultation with clinical experts and examination of clinical data identified via the AG’s SR of clinical effectiveness evidence. Clinical consultation was conducted through a combination of face-to-face meetings, telephone conversations and comments on draft versions of the report.

In the NHS watchful waiting pathway, there are many different points at which an IM3M may, or should, be extracted. For example, the decision to extract an IM3M may be determined by a specific number of instances of pericoronitis, the degree of severity of pericoronitis, a decayed adjacent tooth or the amount of tooth pain. Following several searches of the literature and consultation with experts, the AG concluded that data to populate the current watchful waiting pathway with extraction driven by, for example, a decayed adjacent tooth or number of instances of pericoronitis are absent and clinical advice to the AG was that patient preference would probably play a significant role in the decision to extract.

The AG has, therefore, chosen to design an economic model that is based on what actually happens in terms of the current extraction rates of IM3Ms (i.e. use available data) rather than on what might or should happen under a watchful waiting pathway following symptom development. This approach minimises the use of assumptions around when extraction occurs and allows analysis of the true current situation in the NHS against a change in extraction rates if prophylactic removal was recommended. However, how far the results can be interpreted as a comparison of a watchful waiting strategy with strict adherence to extraction only after certain symptom-related criteria is met, against a prophylactic removal strategy, is unclear.

A visual representation of the intervention pathway (prophylactic removal) is shown in Figure 3. The pathway is modelled as a combination of a Markov model process to move people between different IM3M health states, with decision trees to determine the probability of complications and severity of complications at each cycle and to determine the outcomes from the extraction of an IM3M.

FIGURE 3.

Prophylactic removal of the IM3M model pathway. M2M, mandibular second molar; RoCT, root canal treatment.

The cycle length is 1 year. In each cycle, a person can develop IM3M symptoms (the severity of which is determined by decision trees) but not have the IM3M extracted, have the IM3M extracted (either with or without complications, determined by decision trees), remain in a post-extraction state (with or without permanent nerve damage) or die from any cause.

For every person with an IM3M in situ, there is a probability that, in each cycle, the person will die (from any cause), their IM3M will develop symptoms and/or their IM3M will be extracted. The probability of extraction is independent of symptom development. The possible symptoms are pericoronitis, mild pain and severe pain, which is determined in the model by a decision tree. If a person has only symptoms in a cycle without extraction, they remain in the ‘IM3M in situ’ state.

When a tooth is extracted, the person might be found by the dental surgeon to have developed a DCC in the adjacent mandibular second molar (M2M). The AG has assumed that the probability of extraction of an IM3M is independent of the development of a M2M DCC. This assumption has been made because the data that are available describe the proportion of people with M2M DCC in a population who had an IM3M extracted only, and not on whether or not this was the reason for the extraction. 104 If M2M DCC is present, the tooth can be extracted, be simply restored or have more complex restoration, including root canal treatment.

Extraction of an IM3M can be complication free or result in mandibular fracture, temporary nerve damage, permanent nerve damage or alveolar osteitis (‘dry socket’). If a person has IM3M extraction in a given cycle, these events are determined by a decision tree. After extraction, people can enter either an ‘extracted with permanent nerve damage’ state or an ‘extracted with no nerve damage’ state. They will then remain in either of these states for the lifetime of the model or until death.

Based on the pathway, a model was constructed in Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) that captured the cost and benefits of 1000 patients transitioning through the intervention and comparator pathways.

Model transition probabilities

The transition probabilities used in the AG’s model have been drawn from published studies identified in the clinical evidence review and from evidence identified by clinical experts.

Evidence presented by Fernandes et al. 48 has been used to represent the annual rate of tooth extraction and the development of symptoms for people who have asymptomatic IM3Ms. In this 1-year prospective cohort study, the authors collected data from patients with at least one IM3M. All of the participants were registered with general dental practices in Scotland, where SIGN guidelines34 recommending a watchful waiting pathway have been in place since 2000. The AG considered that these data were generalisable to an English NHS setting.

In the paper by Fernandes et al.,48 the annual rate of extraction (5.47%) was found to be independent of age. The inclusion criterion in terms of the age of the cohort studied by Fernandes et al. 48 was 18–70 years, but only those aged 20–63 years were actually recruited to the study. This determined the model start age of 20 years. It was assumed that the annual rate of extraction continued at 5.47% after the age of 63 years. A scenario analysis explored a scenario in which no extractions were undertaken after the age of 63 years.

Although the Fernandes et al. 48 study is a prospective cohort study, the AG considered that it was well designed with a large sample and provided a satisfactory evidence base to estimate the annual rate of extraction of IM3Ms in a UK NHS setting. The annual rate of extraction of 5.47% that was reported by Fernandes et al. 48 gives a 5-year extraction rate of 24.5%, which is lower than, but of a similar order to, that reported by Hill and Walker47 (31.4%). Given that Hill and Walker47 report on extraction from a different part of the UK to Fernandes et al. 48 and the two studies suggest similar annual extraction rates, this supports the generalisability of the findings from Fernandes et al. 48 to the wider NHS. To check the effect of varying the annual rate of extraction, the upper (7.39% per annum, equivalent to a 5-year extraction rate of 32.0%) and lower (3.94% per annum) bounds of the 95% CI quoted in the paper by Fernandes et al48 were used in the AG’s sensitivity analysis.

For the prophylactic removal pathway, the AG’s base-case assumption is that not all people will accept the recommendation that the IM3M(s) should be extracted and, therefore, these teeth will remain in situ. Fernandes et al. 48 reports that 45.9% of patients who had their IM3Ms extracted did not know why the tooth had been extracted, and so the extraction could have been prophylactic (despite the SIGN guideline34 recommendation). Based on this finding, the AG chose a base-case value of 46.95% to represent the proportion of people who would accept prophylactic removal of asymptomatic IM3Ms, if it were offered to them. The AG recognises that it is likely that this figure overestimates the true rate of non-prophylactic removal, as it is unlikely that all of the 45.9% of patients who could not recall why their IM3M was extracted had their tooth extracted when the IM3M was asymptomatic. However, it could also be an underestimate if future guidelines were to suggest that prophylactic removal is recommended. Owing to these uncertainties, the AG carried out sensitivity analyses to explore the impact of different levels of acceptance of prophylactic removal (10%, 25%, 50%, 75% and 100%).

Findings from the Fernandes et al. 48 study suggest that symptom development is age dependent (at least up to the age of 63 years), declining as a patient ages. In the base case, the AG assumes that the rate of symptom development follows a linear trend from the ages of 53 to 63 years, and that this trend is applied beyond the age of 63 years. A scenario analysis explores the impact of symptom development no longer occurring past the age of 63 years. The probabilities, by age, of experiencing symptoms can be found in Appendix 6.

The probability of a person who, on having an IM3M extracted, is found to have associated M2M DCC, and the likelihoods of the M2M being extracted, simply restored or undergoing complex restoration, have been taken from the a study by McArdle et al. 104 In this study, the investigators undertook a retrospective review of 339 people in England, across two cohorts from 2006 to 2014, who had IM3Ms removed as a result of having M2M DCC.

The probabilities of specific complications that are associated with tooth extraction were derived from Chuang et al. 80 Chuang et al. 80 also reported an odds ratio (OR) for all complications (OR 1.46) for patients aged ≥ 25 years compared with those aged < 25 years. The AG has adjusted the individual complication rates reported by Chuang et al. 80 using this OR to estimate the probability of specific complications for those aged < 25 years and for those aged ≥ 25 years.

One complication for which there is no evidence available in the study by Chuang et al. 80 is the rate of permanent nerve damage. The AG has used a value reported by Valmaseda-Castellón et al. 105 This value has been adjusted for age (aged < 25 years and aged ≥ 25 years) using the OR reported by Chuang et al. 80

Table 14 shows a full list of probabilities used in the AG’s model. Rates for death from any cause have been taken from the Office for National Statistics’ life tables. 106

The use of annual transition probabilities for symptom development in the AG’s model results in a person being able to experience IM3M symptoms only once per year for each tooth. Given that people may develop symptoms more than once per year in each IM3M, the model may underestimate the actual annual symptom burden for people with IM3Ms. This will result in the AG’s model underestimating the annual cost of treating IM3M symptoms and the impact of IM3M symptoms on HRQoL.

As a watchful waiting strategy leaves people with more IM3Ms in situ than prophylactic removal does, the results of the AG’s model will underestimate the potential reduction in costs and gains in HRQoL from reductions in IM3M symptoms when these two pathways are compared. The incremental cost-effectiveness ratio (ICER) per QALY gained generated by the model for prophylactic removal compared with watchful waiting will therefore be an overestimate, as people can develop IM3M symptoms more than once per year.

By designing the model around the available data and excluding the complicated pathology that can occur if IM3Ms are left in situ, the model is biased towards generating more favourable results for a watchful waiting strategy. For example, with an IM3M left in situ, pericoronitis can develop into a severe infection that can spread to the throat and lead to severe cellulitis, which causes airway blockage resulting in hospitalisation, intensive care unit admission and, in some cases, death. Pericoronitis can also lead to abscess formation and is accompanied by a potential risk of developing sepsis.

No published data could be found to estimate the risk of these serious events occurring. However, one of the AG’s clinical experts reported treating one person per month who had been admitted to the intensive care unit with life-threatening cellulitis that was due to having had an IM3M. Although there is no way to generalise this experience into a probability that can be included in the AG’s model, if such a complication were to be included, it would result in a reduction in QALYs and an increase in costs for the watchful waiting strategy.

Resource use and unit cost estimation

The total number of patients with IM3M extractions each year in the NHS is unknown. However, McArdle et al. 104 have estimated that, in 2014/15, there were 152,000 people with IM3M extractions: 67,000 (44.1%) extractions were carried out during inpatient admissions, 38,000 (25.0%) during outpatient attendances and 47,000 (30.9%) during a primary care appointment. The AG has used these estimates as the basis for estimating the cost to the NHS of IM3M extractions. The unit costs of extraction in an acute setting have been taken from the NHS Reference Costs 2015 to 2016. 107 The cost of an extraction in primary care is a band 2 treatment and is charged to the NHS as 3 units of dental activity (UDAs).

The AG has assumed that 75% of M2Ms with DCC are extracted at the same time as the IM3M, at no additional cost. For those M2Ms extracted independently, the cost of extraction is assumed to be the same as the estimated cost of IM3M extraction. Sensitivity analyses have been used to explore the impact of this assumption on model results. Values of concurrent IM3M and M2M extractions of between 0% and 100% have been used in these analyses.

To estimate the cost associated with pericoronitis and severe pain, in the absence of any published information or clinical advice, the AG has assumed that 25% of people will self-medicate and that symptoms will resolve without the need for dental or medical intervention. For the 75% of people who require treatment, dental care will be required. Based on clinical advice, this dental care would comprise band 2 treatment (3 UDAs) plus an antibiotic prescription for erythromycin (a further 0.75 UDAs for issuing a prescription). For the 75% of patients with severe pain, the AG has assumed that an emergency dental appointment will be required (1.2 UDAs) and that people will be prescribed an analgesic (codeine, which incurs a further 0.75 UDAs for issuing a prescription). The AG has undertaken sensitivity analyses to explore the impact on model results of varying the proportion of people self-treating from 0% to 100%.

The cost of a UDA varies across England. In the base case, the AG has used a figure of £25. The BDA quotes this figure as being the mean UDA cost across England. 108 The AG has undertaken sensitivity analyses to explore the impact of lower and higher UDA costs on model results. The figures used in the sensitivity analyses are based on a freedom of information request1 made in 2009 and ranged from £11.08 to £105.58.

Antibiotic and analgesic prescription costs have been sourced from Prescription Cost Analysis, England 2015 data. 109 The AG has assumed that the cost of all other aspects of treatment is covered either by the relevant NHS reference cost or by the payments received for UDAs.

The AG has assumed that mild pain does not result in any cost being incurred by the NHS and that the treatment of alveolar osteitis is included in the cost of the initial extraction. For more serious complications following extraction, the AG has assumed that everyone with permanent nerve damage will receive surgery to try to correct the damage and that no one with temporary nerve damage will receive corrective surgery. No litigation costs from permanent nerve damage are included in the model. However, as it is likely that some people with permanent nerve damage will receive compensation and that some people with temporary nerve damage will receive surgery (indeed, the nerve damage may be only temporary because of corrective surgery), these assumptions mean that the AG’s model will underestimate the true cost to the NHS that arises from nerve damage. For mandibular fracture, the NHS Reference Cost107 for fixation of jaw following fracture was used.

With the exception of the costs of antibiotics and analgesia, the AG has undertaken sensitivity analyses to explore the impact of varying these costs on model results. For those costs based on the NHS Reference Costs 2015 to 2016,107 upper and lower quartile figures have been used in these analyses.

A full list of costs used in the AG’s model is provided in Table 15.

Health measurement and valuation estimation

No IM3M-specific utilities, or utility values related to IM3M symptoms or extraction complications, could be identified from a targeted search of the published literature (see search strategy in Appendix 7). Therefore, the AG used values from a working paper by Ara and Brazier110 to populate these parameters in the model. It is reported in this paper110 that people with teeth, mouth or tongue conditions in the UK have a EuroQol-5 Dimensions utility score that is 0.345 less than those who do not report having any teeth, mouth or tongue conditions. This is comparable to a utility decrement for level 3 pain based on the EuroQol-5 Dimensions, three-level version, questionnaire (‘I have extreme pain or discomfort’) of 0.386. 110

The AG’s model includes a 0.345 utility decrement to represent the experience of people with any IM3M complication(s) or any complication(s) following extraction. For people experiencing mild pain, the decrement was assumed to be 50% of 0.345 (i.e. 0.1725). Sensitivity analyses were undertaken to explore the impact of varying these figures on model results. The decrements were varied individually over the 95% CI (0.102 to 0.549) reported in the paper by Ara and Brazier. 110 The AG has assumed that M2M restoration is not associated with any loss of utility and that extraction of the M2M would result in a loss of utility only if it occurred independently of an IM3M extraction.

To generate a QALY loss when a person experiences a complication, the duration of symptoms from the complication is required. The AG was not able to identify any evidence that described the duration of symptoms associated with extraction complications. Thus, it was necessary to make a number of assumptions about the durations of symptoms; the AG’s clinical experts were contacted to verify these assumptions. The AG has undertaken sensitivity analyses to assess the impact of varying these estimates of duration on model results (duration varied by ± 50%). To reflect the declining HRQoL as people age in the model, utility declines with age in line with age-related population norms described in the paper by Ara and Brazier. 110

The details of the values used in the model to represent utility decrement, the duration of symptoms assumed and the resulting QALY loss associated with symptoms and complications are shown in Table 16.

Analysis of uncertainty

The AG explored the uncertainty surrounding the model assumptions using deterministic sensitivity analyses and scenario analyses. Probabilistic sensitivity analysis was not undertaken, as the only parameter for which a distribution could be drawn based solely on published evidence was the annual rate of extraction. Creating distributions around the central value of other parameters would not be meaningful either because the parameter values are essentially fixed (e.g. in the case of costs) or because the central values are, at least partly, based on assumptions as a result of the lack of data (e.g. the actual utility values associated with symptoms or the duration of symptoms). Furthermore, a probabilistic sensitivity analysis could potentially lead to confounding results, as the uncertainty around most of the assumptions used in the AG’s model is not known; therefore, from a statistical perspective, this cannot be confidently modelled.

The parameter ranges explored in the deterministic sensitivity analyses are summarised in Table 17 and the scenario analyses performed are summarised in Table 18.

Base-case results

Deterministic sensitivity analysis

Tornado diagrams summarising the results of the one-way deterministic sensitivity analyses detailed in Table 17 are shown in Figure 4.

FIGURE 4.

Deterministic sensitivity analyses results.

The tornado diagram shows that the most important parameters that affect model results are the discount rate, the cost of a UDA, the annual extraction rate for IM3Ms, the utility decrements applied to IM3M symptoms and the duration of the symptoms.

Although the results are most sensitive to changes in the discount rate, which is to be expected given that the prophylactic strategy effectively ‘front loads’ extraction costs in the model compared with the watchful waiting strategy, which spreads those costs over many years, the ICER is still under £20,000 per QALY gained even when the discount rate is 5% per annum. The AG considers that, although the 3.5% per annum discount rate is the correct one to apply, an argument could be made that prophylactic removal constitutes a public health intervention similar to vaccination, and so a 1.5% per annum discount rate should be applied instead. At this lower rate, the model predicts an ICER of £3377 per QALY gained for the prophylactic removal strategy.

Given that there is no direct evidence on the QALY loss from symptoms of IM3Ms, it is potentially concerning that the ICER per QALY gained is sensitive to values used in the AG’s model to estimate the IM3M symptom QALY loss. However, over all of the parameter ranges considered, only the lower bound for the utility decrement for severe pain (0.102 as opposed to 0.345 in the base case) results in an ICER of > £20,000 per QALY gained for prophylactic removal (£21,469 per QALY gained).

As stated previously, there is evidence from Hill and Walker47 that the annual rate of extraction of IM3Ms is closer to the upper bound of 7.39% per annum reported by Fernandes et al. ,48 rather than the 5.47% per annum used in the base case. If the annual rate of extraction was 7.39% per annum, the ICER would be £9944 per QALY gained. However, even if the annual rate of extraction was at the lower bound, suggested by Fernandes et al. ,48 of 3.94%, the ICER would be £13,847 per QALY gained, which is still below the £20,000 per QALY gained threshold.

As stated previously, the cost of a UDA varies widely across the country. The value used in the AG’s base case (£25) is at the lower end of the potential range used in England (i.e. from £11.08 to £105.58). At the lower UDA price (£11.08), the ICER for prophylactic removal is £12,925 per QALY gained and, thus, remains below the £20,000 per QALY gained threshold. For geographical areas where the UDA price is higher than the £25 base-case value, the ICER per QALY gained increasingly favours a prophylactic removal strategy.

The AG’s model results are insensitive to several parameters for which no information was available and assumptions had to be made, for example the duration of symptoms of complications from extraction, the percentage of patients seeking treatment for pericoronitis and the percentage of patients having a M2M extraction at the same time as an IM3M extraction. This finding suggests that the lack of robust information on these parameters does not affect on the conclusions that can be drawn from the AG’s model.

Scenario analyses

The total number of patients with IM3M extractions each year in the NHS is unknown. However, McArdle et al. 104 have estimated that, in 2014/15, 152,000 people underwent IM3M extractions: 67,000 (44.1%) extractions were carried out during inpatient admissions, 38,000 (25.0%) during outpatient attendances and 47,000 (30.9%) during a primary care appointment. The AG has used these estimates as the basis for estimating the cost of IM3M extractions to the NHS. Unit costs of extraction in an acute setting have been taken from the NHS Reference Costs 2015 to 2016. 107 The cost of an extraction in primary care is a band 2 treatment and is charged to the NHS as 3 UDAs. Although the number of extractions in an acute setting was derived from published statistics, McArdle et al. 104 estimated the number of extractions in primary care from a historical data source, as primary care IM3M extractions have not been recorded since 2004/5. McArdle et al. 104 state that they believe that their estimate of 47,000 extractions in primary care is probably an underestimate. However, given the potential uncertainty around the number of primary care extractions, scenario analyses were undertaken in which (1) no extractions occurred in primary care and (2) all extractions occurred in primary care.

Summary of cost-effectiveness results

The results generated by the AG’s economic model indicate that the ICER per QALY gained for the comparison of the cost-effectiveness of a prophylactic removal strategy with a watchful waiting strategy is £11,741 for people aged 20 years with asymptomatic IM3Ms. The incremental cost per person associated with prophylactic extraction is £55.71, with an incremental QALY gain of 0.005 per person. The base-case ICER per QALY gained was found to be robust when a range of one-way sensitivity analyses were carried out to test parameter uncertainty and when scenario analyses were carried out to test structural assumptions.

Search strategy

Search strategy

Search strategy

Search strategy

Search strategy

Systematic review characteristics

Search strategy