Cost-effectiveness analysis of non-surgical periodontal treatment to prevent strokes and myocardial infarctions in the United Kingdom Authors: Edward Coote, Nitesh Patel, Rajesh Vijayanarayanan Published: 2026 ABSTRACT AbstractBackgroundPeriodontal disease (PD) is common in the UK and is associated with higher risks of stroke and myocardial infarction (MI). Non-surgical periodontal treatment (NSPT) may therefore generate downstream cardiovascular benefits, but its value for money as a CVD-prevention strategy for the NHS is unclear.MethodsA Markov cohort model simulated 65-year-olds with severe PD and no prior CVD over 10 years (with a 25-year scenario). Eight health states captured acute and chronic phases of stroke, MI, combined events, and death. Baseline CVD risks were derived from published sources; NSPT effects (stroke HR 0.55, MI HR 0.70) were taken from large international cohort studies. Costs (2024 € , NHS payer perspective) and QALYs (EQ-5D) were discounted at 3.5%. Deterministic and probabilistic sensitivity analyses assessed parameter and structural uncertainty and identified effect sizes required to meet NICE thresholds.ResultsOver a 25-year horizon, NSPT increased costs by € 9,755 and QALYs by 0.50 per patient (ICER € 19,322 per QALY). NSPT reduced non-fatal stroke and MI by 41% and 22%, respectively, with 22% of treated individuals event-free at 25 years and a 12% reduction in cumulative mortality. Probabilistic analysis resulted in a PSA mean ICER of € 19,129. The 10-year scenario produced an ICER of € 49,847, and a five-year scenario produced an ICER of € 197,614, demonstrating that NSPT is an intervention with a long payback period.ConclusionsUsing a lifetime horizon and given current available evidence, NSPT potentially provides cost-effective prevention against CVD events at current NICE thresholds. Conclusions are sensitive to the magnitude of NSPT’s cardiovascular effect, which is the main source of parameter uncertainty. We advocate for high-quality UK longitudinal evidence on NSPT’s effect on hard CVD endpoints.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12962-026-00786-1.Keywords: Cost-effectiveness, Periodontal disease, Cardiovascular disease, NHS, Markov model Background The economic burden of cardiovascular disease (CVD) in the UK is significant, despite falling incidence and mortality rates in recent years [1], . The total costs of coronary heart disease (CHD) and stroke in England are forecast to reach € 19.6 billion and € 15.9 billion by 20,250, respectively [2]. CVDs cause approximately 26% of all deaths in the UK and significantly raise an individual’s likelihood of developing vascular dementia [3]. Despite a 30% fall in CVD incidence between 2000 and 2019, the economic burden remains high [4]. As CVD affects a large proportion of the population, preventive measures must be cost-effective and deliver health gains beyond CVD, leading to downstream cost reductions. Periodontal disease (PD) affects at least 50% of people in the UK, with severe periodontal pocketing expected to increase by 56.7% by 2050 [5, 6]. A growing number of longitudinal cohort studies and randomised controlled trials supports evidence that treating PD can lower risk of CVD. Non-surgical periodontal treatment (NSPT) has been found to reduce cardiovascular biomarkers in patients with stable coronary artery disease, and to improve endothelial function in patients with a recent myocardial infarction and severe PD [7–9]. A 2024 meta-analysis found significant reductions in interleukin-6 and systolic blood pressure levels in the treated groups [10]. In patients that have suffered a recent stroke, both standard and intensive PD treatment reduced later stroke and myocardial infarction (MI) mortality [11]. Limitations of trial studies include short follow-up times and small patient pools, therefore longitudinal evidence has to be used to establish stronger causal links. Another meta-analysis of longitudinal studies calculated that the risk of all incident CVD was 22% higher in individuals with PD, and the risk of incident CVD increased ascendingly with PD severity [12]. A further review showed consistent findings when stratified by sex and by specific CVD event (CHD; OR = 1.19, 95% CI: 1.09–1.3 and stroke; OR = 1.29, 95% CI: 1.09–1.53) [13]. A US-based study with 21-years of follow-up found PD and caries associated with an increased risk of ischemic stroke (HR 1.86; 95% CI 1.32–2.61) [14]. The observed associations are subject to confounding variables and risk factors such as smoking, socioeconomic status, physical inactivity, and although the discussed trial evidence for short-term improvements in vascular biomarkers strengthens biological plausibility, it needs to be understood that evidence extending this to hard CVD endpoints is more scarce. None of these studies explicitly assess the cost-effectiveness of treating PD to prevent CVD events, and since healthcare budgets are expected to become more constrained as the population ages, cost-effectiveness will continue to be important to health policy decision makers [15]. The consistent causal link between CVD and PD highlights the potential benefit of population-level PD interventions on reducing CVD risk and mortality. A prior UK study found NSPT cost-effective for improving glycated haemoglobin in Type 2 diabetes [16], and a US analysis suggested PD therapy is cost-effective in reducing diabetes-related microvascular disease [17]. Whether periodontal therapy is cost-effective as a CVD prevention strategy remains unknown. A Markov model approach was used to estimate to what extent NSPT for adults with severe PD will have a downstream effect on reducing the costs of and number of later CVD events. Long-term costs and effects, using quality-adjusted life years (QALYs), provided a valuation of NSPT’s role in CVD burden management. The base arm was not treated with NSPT. Ischaemic stroke and MI were modelled as the CVD outcomes arising from PD risk. The primary objective of this study was to evaluate, from the perspective of the NHS payer, the cost-effectiveness of NSPT for adults with severe PD in preventing MI and stroke. Methods Model structure A Markov model with 1-year cycles over a lifetime horizon of 25 years was constructed as the base case scenario to estimate the incremental costs and health outcomes associated with severe PD treated with NSPT to prevent CVD events. Five-year and ten-year scenarios were used as alternative exploratory scenarios. We modelled at an aggregate population level rather than at an individual level as a population-level policy question was evaluated which does not require individual-level heterogeneity. The model has eight mutually exclusive health states: No prior CVD (base), Post-Stroke (year 1 tunnel), Post-Stroke (year 2+), Post-MI (year 1 tunnel), Post-MI (year 2+), Post-Both (year 1 tunnel), Post-Both (year 2+), and Death (absorbing) (Supplementary Material Figure S1). Individuals enter the model with no prior history of CVD and can experience stroke, MI, both events or death during each cycle. A tunnel system was chosen to represent temporary health states to allow us to capture acute-phase costs and utilities. CVD event costs and mortality are much higher in the initial onset year compared to subsequent years, therefore, tunnel states provide a more accurate estimation of costs [18, 19]. No recurrent events in Post-both year 1 was allowed. Tunnel states relax the memoryless property while retaining time-since-event effects; tunnel duration was one cycle with death as the only competing exit. Half-cycle correction was not modelled. The base utilities for a 65-year old male were used. No sex-specific differences in NSPT effects have been found [20]. The decision to model at aged 65 was justified by the expected onset, survival rates and recurrent events for an individual after a stroke or MI [21, 22]. The rate of fatal stroke and MI events was given as 10 and 14%, respectively, from recent studies [23, 24]. The background death probability was calculated from Office of National Statistics life tables [25]. Baseline stroke and MI hazards in the base state were calculated from literature that provided the risk of onset events for individuals over 60 years old with severe PD [26, 27]. This was cross checked with a retrospective study that assessed by how much severe PD increased the incidence of the respective CVD event [28]. State-specific hazard multipliers captured elevated risks of subsequent CVD events after a prior event, reflecting disease progression. NSPT treatment effects were applied to the treatment arm using hazard ratios (HR) obtained from existing literature from international data. There have been no randomised controlled trials or long-term cohort studies testing the effect of NSPT on preventing CVD events within a UK population [29]. Therefore, we have used international cohort estimates which represent the best available evidence, and parameter uncertainty has been addressed through one-way sensitivity analysis and PSA. Threshold analysis identifies the effect sizes required for NSPT to meet NICE thresholds, allowing for the plausibility of HRs used to be judged against the underlying evidence base. There are varied estimates for the size of the impact that NSPT has on CVD events. Retrospective and prospective cohort studies of adults with PD who have had regular NSPT treatment have found lower hazards of stroke of 0.40–0.78 (95% CI 0.29–0.81) relative to untreated PD [30, 31]. NSPT has been found to lower the hazard of MI by 0.54–0.90 (95% CI 0.44–0.95) relative to untreated PD [32, 33]. The included studies primarily involved older adults with later stage PD who received regular NSPT. Given the heterogeneity in study populations and follow-up durations, the median of reported HRs was used as a summary estimate for base-case analysis. Therefore, stroke treatment HR was 0.55 and MI treatment HR was 0.7. Treatment effects only had an effect on the chance of stroke and MI, background death hazard remained constant throughout. Expected annual periodontal resource use per patient was estimated by combining procedure-specific unit costs with expected frequencies (probability of each procedure per year) from a previous study [16]. This is in line with NHS and dental periodontal treatment protocol [34, 35]. We have assumed fixed periodontal treatment intensity and procedures as severe PD is a chronic illness that will require consistent care over the horizon. Health utilities and costs The NHS payer perspective was used and reported costs and QALYs in accordance with National Institute for Health and Care Excellence (NICE) guidance [36]. This does not capture the true costs to clinicians and society, as it omits opportunity costs and non-medical costs. The NHS payer perspective is most relevant for reimbursement decisions and has been the most common perspective taken in dental health economics [37]. Base costs for non-surgical treatment were taken from a previous cost-effectiveness study, inflated to 2024 costs, and cross referenced with current NHS Dental band costs and procedure to ensure accuracy [16, 38, 39]. The per-cycle cost of a patient with severe PD was calculated as € 1,274.82. The use of pre-calculated costs from previous cost-effectiveness analysis ensured replicability and comparability of our results. full calculations are found in Supplementary Material Table 3. Stroke and MI costs for each health state were obtained from contemporary literature and applied to the relevant model states [18, 40–42]. All individuals experiencing a stroke or MI were assumed to receive treatment within the NHS. Costs were inflated to 2024 prices, with long-term costs assigned to chronic post-event states. These included NHS social care costs. A one-off acute cost was applied during the cycle in which the event occurred. These estimates were cross-checked for consistency against the National Stroke Audit Programme [43]. Where costs were stratified by age and sex chosen the costs most applicable to a 65-year old male. For individuals who experienced both a stroke and a MI, the higher of the two event costs was assigned (post-stroke) to ensure full capture of healthcare resource use. This approach avoids underestimation of costs when overlapping conditions occur, given the lack of a clear consensus on combining costs for multimorbidity in health economic evaluations [44]. Full calculations are found in the Supplementary Material Table 4. All QALY values were derived from EQ-5D sources. The decision to use 1 year as a cycle length was based on the available data for costs and QALYs associated with CVDs [19, 45]. There was found to be minimal difference between the health state of an individual 3 months after a stroke or MI, compared to 9 to 12 months after, which supports the decision to use a year for the Markov states. The health utility values for post-stroke and post-MI for the immediate post-event state were obtained from a single study reporting quality of life impacts for CVD on individuals in the UK [19]. Values for stroke events were cross referenced with values estimated in a meta-analysis [45]. MI utility values were verified with those used in a lipid-lowering cost-effectiveness analysis which we also used costs for [42]. The multiplicative approach was taken to calculate the utility values for post-both events for first year state and subsequent years state. This approach has been found to perform best for combined health states of two conditions simultaneously [46]. A one off disutility effect for each acute event was attached to the event states [45, 47]. The model purposefully excludes any direct oral-health utility gains from improved periodontal status; therefore, estimated QALY gains reflect downstream CVD effects only and are likely conservative. All cost and QALY outcomes were discounted at the suggested rate of 3.5% for NHS evaluations [36]. Full costing, QALY utility values and parameter breakdown can be found in Supplementary Material Table 1. Outcomes Model results included total discounted QALYs, costs and the incremental cost-effectiveness ratio (ICER). The intervention was considered cost-effective if the ICER was below the NICE threshold of € 20,000-€ 30,000 per QALY gained. We also tested at the new NICE threshold range of € 25,000-€ 35,000 to determine if this resulted in different conclusions [48]. Net monetary benefit was also calculated. This study adhered to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) 2022 checklist. The model was built in Microsoft Excel. Sensitivity analysis One-way sensitivity analysis was used to evaluate the impact of changing single parameter values for costs, utilities and transition probabilities. The maximum and minimum values within a range were used to test uncertainties when provided, or ±10% where confidence intervals were not reported in the source. PSA was used to evaluate the robustness of the results, especially given treatment effect data is not from the UK. The β-distribution was used for utility and transition probabilities. The γ-distribution was used for costs. 10,000 Monte Carlo simulations were ran with the value of each model input being randomly drawn from the assigned parametric distribution. A cost-effectiveness plane and cost-effectiveness acceptability curve were drawn using these results. Full distribution parameters for PSA are in Supplementary Tabdeltemple 5. We used 5-year and 10-year horizons to assess the time-horizon sensitivity of our outcomes. Threshold analysis was used to identify the treatment HR effect sizes associated with each cost-effectiveness threshold. Results Base case analysis Table 1 provides the deterministic results across the 25-year base case and the 5-year and 10-year scenarios, and the probabilistic 25-year results. Over the 25-year horizon, NSPT was associated with a € 9,755 increases in costs and 0.50 additional QALYs per patient, which resulted in an ICER of € 19,322. NSPT had a positive net monetary benefit of € 5,400 at λ = € 30,000. Table 1Deterministic and probabilistic results over 5-, 10-, and 25-year horizonsHorizonBaseTreatedIncrementalIncrementalICERcost (€ )cost (€ )cost (€ )QALYs(€ /QALY)5 years5,0079,0224,0150.020197,61410 years13,36619,2985,9320.11949,84725 years (base case)36,21645,9719,7550.50519,32225 years, PSA mean——13,4390.76019,129 Deterministic and probabilistic results over 5-, 10-, and 25-year horizons Using a 10-year horizon produced an ICER of € 49,847, and the 5-year horizon produced an ICER of € 197,614. Over the 25-year horizon, NSPT reduced cumulative non-fatal stroke incidence by 41% (0.434 events per person in the base arm versus 0.257 in the treated arm), and non-fatal MI by 22% (0.607 versus 0.474). At 25 years, 22% of individuals in the treated arm were CVD-free compared to 11% in the base arm. Cumulative all-cause mortality was 12% lower in the treated arm. Final transition probabilities are found in Supplementary Material Table 2. Lower mortality and more event-free survivors has translated into health gains, but mostly in avoiding morbidity rather than extending lifespan. Sensitivity analysis Figure 1 shows the results of our one-way sensitivity analysis for the 25-year scenario. Both treatment effect HRs drove the largest ICER changes. Varying the stroke HR across its range of 0.29–0.81 produced ICERs of € 3,046-€ 46,815, and varying the MI HR across 0.44–0.95 produced ICERs of € 11,275-€ 38,085. The per-cycle NSPT cost was the third most influential on the ICER, with a ±10% variation resulting in an ICER range of € 15,015 and € 23,629. Case fatality rates, the hazard multipliers for recurrent events, and acute event costs all moved the ICER by less than 2%. All other parameters changed the ICER by less than 10%. Tornado plot and table of one-way sensitivity analysis for selected variable Figure 2 and Table 2 shows the results of our threshold analysis using the 25-year horizon. Using the base case MI HR of 0.70, stroke HRs up to 0.56 keep the ICER under € 20,000 and up to 0.67 keep it under € 30,000. With the base case stroke HR of 0.55, MI HRs up to 0.71 keep the ICER under € 20,000. One-way sensitivity analysis of treatment effects Table 2NSPT hazard ratios required for the 25-year ICER to reach successive NICE thresholds. Base-case HRs from the source literature are 0.55 (stroke) and 0.70 (MI)NICE thresholdStroke HR neededMI HR needed(€ /QALY)(MI HR fixed at 0.70)(Stroke HR fixed at 0.55)20,0000.560.7225,0000.620.8130,0000.670.8735,0000.720.92 NSPT hazard ratios required for the 25-year ICER to reach successive NICE thresholds. Base-case HRs from the source literature are 0.55 (stroke) and 0.70 (MI) Figures 3 and 4 reports the results of our PSA. A mean incremental cost of € 13,439 and mean incremental QALY gain of 0.703 gave a PSA ICER of € 19,129. In 99% of simulations, NSPT was both more effective and more costly than no treatment. The probability of cost-effectiveness was 52% at the € 20,000 threshold, 67% at € 25,000, 77% at € 30,000 and 84% at € 35,000. Cost-effectiveness plane Cost-effectiveness acceptability curve Discussion This study provides a structured economic assessment of the downstream cardiovascular implications of NSPT in a UK-relevant setting over a lifetime 25-year horizon. The analysis quantifies potential reductions in MI and stroke events under published effect sizes, demonstrates that conclusions are highly time-horizon dependent for chronic preventive interventions, and identifies the treatment-effect magnitude required for NSPT to meet NICE thresholds. The 41% reduction in cumulative non-fatal stroke and 22% reduction in non-fatal MI over the horizon reflect an intervention that holds clinical relevance. It aligns with epidemiological consensus evidence that treating PD could significantly help to prevent CVD [29]. These relative reductions translated to 0.505 QALYs gained per person, indicating that the absolute health benefit was modest relative to the incremental cost. This stems from the fact that despite many non-fatal events being averted, those events carry limited per-person QALY gains. At an incremental discounted cost of € 9,755 per patient for 0.505 QALYs, the cost-effectiveness of NSPT compares unfavourably with well-established CVD prevention methods [49]. At the calculated incremental costs, the better value of established CVD prevention methods has to be considered. Well researched cost-effectiveness of methods such as statins and smoking cessation will likely buy more QALYs, especially in a crowded priority-setting space such as the NHS. An important finding of ours is the strong time-horizon dependence of the results. Given our results, NSPT is an intervention with a long payback period, typical of chronic preventative therapies, with the majority of QALY gains accruing in the later years of the horizon. We find that policy decisions based on short time frames may undervalue NSPT. We use threshold analysis to make the economic case for NSPT explicit. Even at moderately high HRs, which produce less of a treatment effect, the ICER remains cost-effective under a lifetime horizon. Current findings highlight that evidence for CVD benefits of periodontal interventions remains low-certainty [50], making high-quality RCTs or longitudinal studies essential to address this gap. However, a recently published major UK randomised trial has shown that intensive periodontal treatment can slow carotid artery atherosclerosis progression and improve vascular function [51]. This provides a degree of plausibility for our base-case HRs as it is the strongest causal evidence to date for NSPT’s mechanism of action on vascular outcomes. Previous economic studies have focused on diabetes-related outcomes, limiting comparability. Choi et al. found that expanding periodontal treatment in a US population was cost-saving in averting tooth loss and microvascular disease, with the majority of healthcare cost saving coming from averting tooth loss [17]. This approach simulated individuals rather than an aggregate population due to its testing of correlations between demographic characteristics and chronic disease risks obtained from the National Health and Nutrition Examination Survey (NHANES) (2009–2014). A UK model by Solowiej-Wedderburn found periodontal treatment in a 58-year old man with type 2 diabetes (HbA1c 7%-7.9%) to have an ICER of € 28,000, with health gains larger in patients with higher HbA1c [16]. These attributed savings helped to improve glycaemic control, suggesting that periodontal therapy’s value is greater when considering the benefits beyond CVD alone. This highlights that the value of NSPT depends heavily on the range of outcomes included, and that CVD-only analyses provide a narrower perspective. We kept the scope of our study to cardiovascular endpoints, meaning our QALY estimates exclude the direct oral-health utility gains and other systemic benefits, such as reductions in hypertension and improvements in endothelial function. We predict that the reported ICERs are therefore conservative estimates. A multi-outcome framework modelling the effects of NSPT on relevant outputs such as CVD events, glycaemic control and direct oral-health benefits would likely produce markedly different ICERs in favour of NSPT. Therefore, the 25-year ICER produced here should be taken as a lower bound for the case of NSPT. Limitations The most significant limitation is the uncertainty surrounding NSPT’s effect on CVD events. To address this uncertainty, a wide range of NSPT effect sizes were examined. We used time-constant HRs for NSPT treatment effects, although in clinical reality NSPT would be expected to have a varied impact over the horizon. For example, treatment effect could diminish with an individual’s age since competing mortality risks rise and a modifiable exposure becomes harder to treat. However, there are no published estimates of time-varying NSPT effects on incident CVD events. A wide range of plausible HRs were tested in one-way sensitivity analysis and PSA. We understand that this assumption may result in the overestimation of QALY gains over a lifetime horizon, therefore, quantifying time-varying NSPT effects on CVD events should be a research priority when designing any future UK longitudinal study or trial. The scope of this study was to evaluate the cost-effectiveness of NSPT on fatal and non-fatal CVD events. Therefore, these results do not include the patient-level benefits of NSPT on severe PD, and the QALY estimates underestimate the total health benefits. Direct utility gains from reduced tooth pain, tooth retention, and better chewing function and nutrition are not captured in this analysis. Including these gains would improve the cost-effectiveness of periodontal treatment, but widen the scope beyond CVD outcomes. For individuals with severe PD and co-existing comorbidities, NSPT may offer broader systemic health improvements, such as reductions in hypertension and enhancements in endothelial function [52]. These potential benefits were not modelled and should not be inferred as contributing to the results. Modelling at an aggregate Markov level does not account for individual patient-level factors that could affect treatment effectiveness, despite the evaluation targeting a population-level policy. However, no UK-based longitudinal study exists that estimates the size of NSPT while controlling for individual characteristics, and therefore, we did not use a microsimualtion evidence-based risk model. Such a study would also reduce uncertainty surrounding NSPT’s effect on cardiovascular risk. Conclusions Our results demonstrate that, given the available evidence and using a 25-year horizon, NSPT is cost-effective as a CVD-prevention strategy for adults aged 65-years and older with severe periodontal disease in the UK. The result is sensitive to the magnitude of NSPT’s cardiovascular effect, and the principal driver of decision uncertainty is the absence of UK longitudinal evidence on NSPT’s effect on incident stroke and MI. The causal link between PD and CVD needs to be established with greater evidence. Considering chronic disease interventions over horizons shorter than a lifetime horizon will underestimate cost-effectiveness, and sufficient scenario analysis can help to mitigate this. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements N/A. Abbreviations Cardiovascular disease Coronary heart disease Periodontal disease Non-surgical periodontal treatment Myocardial infarction Quality-adjusted life year Incremental cost-effective ratio Probabilistic sensitivity analysis Authors contribution EC and NP conceived the study, collected and curated relevant literature. RV supervised the study and made substantial contributions to the manuscript. EC collected the data, built the Markov model, ran analyses, and drafted the initial manuscript. NP contributed to manuscript structuring, framing, and substantive revisions to the manuscript. All authors approved the final version, and agree to be accountable for all aspects of the work. Funding EC, NP and RV are employees of 21D Clinical Limited, which RV founded, but no specific funding was allocated specifically to this research article. Data availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request (email). All data used was from previously published articles or national sources. Declarations Ethical approval Not applicable. Consent for publication All authors have agreed to submit this research article for submission. Consent to participate Not applicable. Competing interests EC and NP work for 21D Clinical Limited. This is a for-profit company that is limited to long-term dental prosthetics. The purpose of this study was the provide evidence on the impact of treating oral health conditions non-surgically through the National Health Service, in order to increase holistic understanding about the benefits of oral health and the methods used to investigate. Funding: The funders had no role in the study design, collection of data, presentation of results or writing of the article. The authors declares no conflicts of interest with this study. The authors does not have any direct or indirect financial interests in the information listed or data used in the study. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request (email). All data used was from previously published articles or national sources.