Article Text
PDF
Abstract
Purpose To determine the cost-effectiveness of annual renal imaging surveillance (RIS) in hereditary leiomyomatosis and renal cell cancer (HLRCC). HLRCC is associated with a 21% risk to age 70 years of RCC. Presentations with advanced renal cell cancer (RCC) are associated with poor outcomes whereas RIS detects early-stage RCC; however, evidence for the cost-effectiveness of RIS is lacking.
Methods We developed a decision-analytic model to compare, at different age starting points (11 years, 18 years, 40 years, 60 years), the costs and benefits of lifetime contrast-enhanced renal MRI surveillance (CERMRIS) vs no surveillance in HLRCC. Benefits were measured in life-years gained (LYG), quality-adjusted life years (QALYs) and costs in British Pounds Sterling (GBP). Net monetary benefit (NMB) was calculated using a cost-effectiveness threshold of £20 000/QALY. One-way sensitivity and probabilistic analyses were also performed.
Results In the base-case 11-year age cohort, surveillance was cost-effective (Incremental_NMB=£3522 (95% CI −£2747 to £7652); Incremental_LYG=1.25 (95% CI 0.30 to 1.86); Incremental_QALYs=0.29 (95% CI 0.07 to 0.43)] at an additional mean discounted cost of £2185/patient (95% CI £430 to £4144). Surveillance was also cost-effective in other age cohorts and dominated a no surveillance strategy in the 40 year cohort [Incremental_NMB=£12 655 (95% CIs −£709 to £21 134); Incremental_LYG=1.52 (95% CI 0.30 to 2.26); Incremental_QALYs=0.58 (95% CI 0.12 to 0.87) with a cost saving of £965/patient (95% CI −£4202 to £2652).
Conclusion Annual CERMRI in HLRCC is cost-effective across age groups modelled.
- health care economics and organizations
- genetics
- surgical oncology
- medical oncology
- genetic predisposition to disease
Data availability statement
Data are available on reasonable request. The data analysed and example code for the decision model used in this study are available from the corresponding author on request.
Statistics from Altmetric.com
- health care economics and organizations
- genetics
- surgical oncology
- medical oncology
- genetic predisposition to disease
Introduction
Hereditary leiomyomatosis and renal cell cancer (HLRCC, #150800) is a tumour predisposition syndrome characterised by age-related penetrance (ARP) to the development of cutaneous and uterine leiomyomata, and renal cell carcinoma (RCC). RCC penetrance is 21% by age 70 years with the youngest reported diagnosis at age 11 years.1
Surgery, by radical or partial nephrectomy, is the mainstay for curative treatment of RCC, with 5-year survival in patients diagnosed with early-stage disease greater than 75%.2 Where metastatic disease is present, immunotherapies and targeted therapies, with or without nephrectomy, are the main interventions.3
HLRCC-associated RCC is particularly aggressive. Recently, we reported >60% of symptomatic presentations being with stage 3/4 disease where mean survival was 15.8 months.1 For unaffected individuals, annual surveillance by contrast-enhanced renal MRI (CERMRI) is recommended with additional imaging of lesions suspicious for RCC; unlike other hereditary RCC predisposition syndromes, surgical excision is advised on detection.4 We have shown the RCCs detected by imaging surveillance are stage 1, treatable by surgical resection, with individuals all alive and disease free at censoring.1
While there is a clinical case for renal imaging surveillance in HLRCC, this has not been formally evaluated for cost-effectiveness especially considering a potential of 50–60 years of imaging for an ARP of 21% to age 70 years. Here, we present a cost-effectiveness analysis of a once-yearly CERMRI surveillance strategy at different ages in individuals with HLRCC. This surveillance is compared with a no surveillance strategy whereby symptomatic RCC is diagnosed through the healthcare system.
Methods
Populations were simulated in a state-transition semi-Markov model. A UK National Health Service (NHS) perspective was adopted with a lifetime time horizon used for the collection of relevant costs and benefits. Evaluation methods were aligned to the ‘reference case’ for health technology assessments, as used by the National Institute for Health and Care Excellence (NICE).
The population informing the model were a hypothetical patient cohort with HLRCC but who had not developed RCC. In the base-case, an 11-year-old population (youngest age of reported HLRCC-associated RCC)1 was modelled to understand the downstream costs and benefits of a lifetime strategy of renal imaging surveillance (RIS) versus no surveillance. Other population subgroups (ages 18 years, 40 years, 60 years) were also simulated to assess the impact of RIS starting age for cost-effectiveness.
These particular ages were used so as to reflect the main ages at which the calculated ARP of RCC changes as per previously published data pertaining to patients with HLRCC.1 Thus, until age 11 years the risk is 0%, rising to 1.7% by age 18 years, 3.6% by age 40 years and 20.8% by age 60 years. In the current study, no actual patients were involved; the outcomes are modelled on published data, guidance and expert opinion as detailed below.
Natural history model
Model conceptualisation was informed by: (1) systematic review of economic studies for RCC complemented with a targeted review for newer studies;5 (2) published data and expert opinion to map surveillance pathways;6 (3) guidance on RCC diagnosis and management3 and (4) NICE technology appraisals for advanced/metastatic RCC treatment.6–8 A semi-Markovian structure was selected due to the long-term nature of the intervention and differential risk profile associated with RCC disease at different stages of the life cycle. The model iterated over 12-monthly cycles, using half-cycle corrections for the different populations, over a lifetime. Within the model, the population cohort could reside in any number of health-states (figure 1), with the model starting in the no active RCC health-state, ‘HLRCC’.
Schematic of the semi-Markov model structure. RCC, renal cell cancer; ARCC, advanced RCC; HLRCC, hereditary leiomyomatosis and renal cell cancer; PRCC, post RCC.
RCC risk was set to begin age 11 years with a 21% cumulative risk to age 70 years.1 RCC was conceptualised as two discrete health-states representing early (stage 1/2) or late (stage 3/4, referred to as advanced RCC, ‘ARCC’).9 A series of ‘tunnel’ health-states (1–10 years post-RCC) were included following the RCC health-state reflecting the increasingly lower risk of moving to ARCC following successful surgical resection of RCC.10
An absorbing health-state of death was included to capture mortality arising from RCC and other competing causes. All-cause mortality could occur from all health-states within the model with transition probabilities estimated from age and sex-adjusted national life tables.11 Patients could only progress to RCC-cause mortality via the ARCC health-state. Transitions from ARCC to death were sourced from recent technology appraisals assessing systemic therapies for advanced disease.7 12
Surveillance model
RIS consisted of annual CERMRI conducted over the remaining lifetime of individuals. For suspected RCCs seen on surveillance CERMRI, CT chest/abdomen/pelvis was incorporated for characterisation and staging. RIS, informed by Forde et al,1 altered the proportions of early/late-stage RCC diagnoses (90% early-stage, 10% late-stage), although reported surveillance detected RCCs in HLRCC are all early-stage.1 Sensitivity analysis varied the proportions of early/late stage RCCs detected.
Although we previously demonstrated non-surveillance detected RCCs in HLRCC are mostly late-stage,1 we elected to use a conservative assumption whereby individuals not undergoing RIS had the same RCC staging proportions as the UK general population (56% early-stage and 44% late-stage).2
Annual RIS with CERMI was assumed to be 100% sensitive, given no interval RCCs have been reported,1 but with an assumed 5% false-positive rate, which was varied in the sensitivity analysis. False positive results were assumed to cause patients disutility as well as imposing a healthcare cost.
RCC treatment model
Management of surveillance and non-surveillance detected RCCs, consisted of open partial/radical nephrectomy dependent on RCC size, surgical complexity and considering HLRCC guidance.4 13 Different surgical modalities altered treatment costs.
Treatment for ARCC is typically systemic which aims to extend life but is not curative. This aspect was informed by expert opinion and relevant clinical guidelines/technology appraisals and consisted of first line therapies (combination of PD-1/PD-L1 checkpoint inhibitors with tyrosine kinase inhibitors), second-line therapies and palliative care as appropriate.12 14 Frequency of adverse events for these therapies were incorporated in the model.
Benefits: life-years and QALYs
Benefits were measured in life-years and quality-adjusted life-years (QALYs) which incorporate both length and health-related quality of life (‘utility’) into a single measure. Mean utility values for the health-states are reported in online supplemental table 1 and were sourced from the literature, with EQ-5D derived values being selected as the preferred measure.15
Supplemental material
HLRCC-state patients were assumed to be asymptomatic and to have utility values equivalent to same age individuals in the general population. For individuals undergoing RIS there was an ongoing disutility, applied additively, associated with annual CERMRI. Utility values for those in the advanced disease health-state were sourced from clinical trials assessing the use of novel immunotherapies with a mean utility multiplier of 0.65 selected for a population at age 60 years.16 17 There is little evidence for the utility impact of RCC at early-stage disease so patients were assumed to have a utility value of 0.75 which did not differ depending on surgery type.18 In the post-RCC health-states, patients were assumed to have a utility close to the general population at 0.95. All mean utility values for non-HLRCC health-states were adjusted to account for age using general population scores. An additional disutility of 0.01 was applied for a false positive result.19
Healthcare resource-use and costs
Resource-use and costs were quantified using sources identified from the literature and mapped to model health-states (online supplemental tables 2–5). If necessary, existing cost estimates were updated to the 2018–2019 price year using the NHS cost inflation index (NHSCII).20 Resource-use that was directly quantified were costed by attaching relevant UK unit costs sourced from national sources.20–22 Annual CERMI surveillance was calculated to cost £176 per-patient per-year (online supplemental table 2). No surveillance was assumed to incur zero yearly costs.
Model analysis, sensitivity analysis and verification
Lifetime costs and benefits were discounted at 3.5%. Results were quantified as incremental costs, benefits, incremental cost-effectiveness ratios and net monetary benefit (NMB). An upper cost-effectiveness threshold of £20 000/QALY was used to determine cost-effectiveness.
One-way sensitivity analysis was used to explore key potential drivers of cost-effectiveness. For probabilistic sensitivity analysis (PSA), costs were parameterised using a log-normal distribution with utilities and probabilities using a beta distribution. For model inputs where the underlying distribution was unknown, the SE was assumed to be 10% of the mean value. PSA generated 1000 simulations for each modelled starting age and was used to construct cost-effectiveness acceptability curves varying the cost-effectiveness threshold. 95% credible intervals were calculated using the 2.5% and 97.5% percentiles of the probabilistic outputs.
All model analysis was conducted in Microsoft Excel with the final model checked against the TECH-VER verification checklist.23 The Consolidated Health Economics Reporting Standards (CHEERS) was followed throughout (online supplemental table 6).24
Results
Base-case: 11-year-old cohort
Compared with no surveillance, a lifetime of surveillance commencing in an 11-year-old population was highly cost-effective with a mean incremental NMB of £3522 per patient (95% CI −£2747 to £7652) at a cost-effectiveness threshold of £20 000/QALY (table 1). Surveillance generated more health gains at a higher cost producing 1.25 (95% CI 0.30 to 1.86) additional life-years and 0.20 QALYs (95% CI 0.07 to 0.43) at an additional mean discounted cost of £2185 (95% CI £430 to £4,144) per patient. At a threshold of £20 000/QALY, there would be a 91.5% chance of active surveillance being cost-effective, rising to 95.3% chance at a threshold of £30 000/QALY (figure 2).
Cost-effectiveness results
Incremental cost-effectiveness plane (left panel) and cost-effectiveness acceptability curve (right panel). QALYs, quality-adjusted life years.
Alternate subgroups
In the other age subgroups (18-year-old, 40-year-old, 60-year-old), surveillance was also very cost-effective (table 1). In the 18-year-old subgroup, mean NMB was £5144 per patient (95% CI −£2022 to £10189), generating 0.34 QALYs (95% CI 0.08 to 0.52) at an additional mean discounted cost of £1692 per patient (95% CI £−312 to £3810). In the 60-year-old subgroup, mean NMB was £2507 per patient (95% CIs −£934 to £5233), generating 0.16 QALYs (95% CI 0.05 to 0.24) at an additional mean discounted cost of £623 per patient (95% CI £698 to £1968).
We undertook an additional analysis in an 8-year-old cohort as there may be instances where CEMRIS starts between the ages of 8 years and 11 years (online supplemental table 7), (online supplemental figures 1,2). Here, mean NMB was £2550 per patient (95% CI −£2897 to £6175), generating 0.25 QALYs (95% CI 0.06 to 0.37) at an additional mean discounted cost of £2437 per patient (95% CI £845 to £4423).
The most cost-effective strategy was in the 40-year-old subgroup which dominated a no surveillance strategy with a NMB of £12 655 per patient, producing 0.58 QALYs (95% CI 0.12 to 0.87) with a cost saving of £965 per patient (95% CI £−4202 to £2652). Here, surveillance had a 98% probability of being cost-effective at a threshold of £20 000/QALY (figure 2).
One-way sensitivity analysis
The proportion of patients diagnosed with stage 1/2 RCC was an important driver of cost-effectiveness. Annual CERMRI scan needs to detect 76% of RCCs at stage 1/2 for surveillance to be cost-effective versus no surveillance, at a threshold of £20 000/QALY (table 2). The base-case assumption was that surveillance would detect 90% of cancers at stage 1/2. Important determinants of NMB are displayed in figure 3 along with the ranges used for the model parameters. Lower surveillance costs led to higher NMB and vice versa; if these costs exceeded £350 per year, surveillance was not expected to be cost-effective at either end of the model cohort starting age 11 or 60 years. Yet, even at £350 per year, surveillance would still be cost-effective in both 18-year-old and 40-year-old cohorts. Varying the discount rate for both costs and benefits between 0% and 6% (as is practice for public health interventions) also had a similar effect on cost-effectiveness with lower discount rates improving cost-effectiveness and higher-rates reducing cost-effectiveness. Irrespective of the discount rate selected, surveillance was cost-effective in all age-cohorts. Lower ARCC costs (£25 000/year) reduced NMB of surveillance in all age-cohorts while higher ARCC costs (£150 000/year) increased the cost-effectiveness of surveillance. Neither parameter input changed whether surveillance would be considered cost-effective. Among the other 43 parameter inputs (online supplemental table 1) to be varied in one-way sensitivity analyses, only a few had a material impact on NMB and none changed whether surveillance would be considered cost-effective at a £20 000/QALY cost-effectiveness threshold.
One-way sensitivity varying the proportion of cases identified through active surveillance for an 11-year-old cohort
Tornado diagrams: deterministic sensitivity analyses for 11-year-old cohort (top left panel), 18-year cohort (top right panel), 40-year-old cohort (bottom left panel) and 60-year-old cohort (bottom right panel). ARCC, advanced RCC; HLRCC, hereditary leiomyomatosis and renal cell cancer.
Discussion
In this decision-analytic model, populated using recently published ‘real-world’ data,1 we demonstrate that, in HLRCC, annual CERMRI surveillance in the modelled age groups is cost-effective.
In both the younger cohorts (commencing age 11 years and 18 years) and an older cohort age 60 years, the cost of surveillance was more expensive than a no surveillance strategy, but produced gains in expected health (life-years and QALYs) making it a cost-effective strategy at thresholds typically considered for resource allocation.25 Modelling of an older cohort, commencing at age 60 years, was still cost-effective although with a lower NMB (£2366 per patient) than for the 11-year and 18-year cohorts; this was despite having lived through much of the increased RCC risk and considering the competing risk of other cause mortality.
International guidance recommends, for HLRCC, consideration of predictive genetic testing from the age of 8–10 years onwards with decisions made on an individual basis.4 26 We therefore included an additional analysis for an 8-year-old cohort to encompass the very earliest age at which CEMRIS might commence. This was still cost-effective although less so than for the other age groups modelled, reflecting the ARP data used to inform the model where RCC risk commences age 11 years.
The ARP for RCC in HLRCC shows risks increasing dramatically from age 40 years to age 60 years.1 This was reflected in our modelling whereby surveillance in a cohort age 40 years was cheaper than a no surveillance strategy and would be expected to produce the greatest incremental health gains across all age cohorts.
Cancer prevention and early detection strategies are central to improving survival rates.27 The introduction of population-based screening is often challenging in terms of false positive/negative findings, the upfront investment costs for deferred benefit and has therefore only been implemented for a couple of specific scenarios.28 However, the implementation of personalised prevention and early detection strategies in genetically predisposed high-risk population groups enables the directed use of resources in a clinically and cost-effective manner.27
There are established clinical guidelines for renal imaging surveillance in hereditary RCC predisposition, estimated to account for ~3%–5% of RCC.29 To our knowledge, this is the first study to investigate, and demonstrate, the cost-effectiveness of such a strategy quantifying the costs and health gains over a lifetime of surveillance. The main reasons for these findings likely reflect the: (i) high ARP of RCC in HLRCC; (ii) markedly different survival rates by stage at diagnosis; (iii) high therapeutic costs of treating advanced disease; (iv) high proportion of surgically curable disease detected by surveillance and (v) low patient and healthcare burden from false positive findings. Based on these findings and those of our previous study,1 we would recommend that annual renal imaging surveillance is offered in HLRCC across all age groups, starting age 11 years.
The main limitation from this study is the inherent, but necessary, simplicity when disease processes are modelled. The biological complexities, for example, arising from different histological subtypes, their detection and behaviours, are often not incorporated. We have countered this by utilising observational data from patients with HLRCC; while preferable to estimates of incidence and outcomes, there is often unavoidable bias when considering rare disease with limited, often historical, datasets. Even with modelling using outcome data from sporadic RCC which is more favourable than for HLRCC-associated RCC, we still demonstrated cost-effectiveness. It is likely with the advent of new therapeutic strategies in advanced disease, outcomes will improve further for this devastating cancer.
Our model is based on the current standard of surveillance care, annual CERMRI. There have been recent concerns regarding gadolinium retention and potential neurotoxicity, although no defined consequential clinical condition has been demonstrated.30 While contrast is needed to detect early RCCs/renal cysts with septal enhancement and so is indicated for early RCC detection in these high-risk patients, its use should follow current national guidance.31 There are also often limitations in MRI accessibility in view of the hardware and highly skilled radiological expertise required. Therefore, there is a need to develop novel clinically and cost-effective means of early RCC detection in these high-risk patient groups. Furthermore, there is also the challenge of ensuring that genetically predisposed high-risk patients are identified through appropriate referral and genetic testing to enable early detection strategies to be put in place.
In conclusion, we demonstrate that annual surveillance by CERMRI to enable early RCC detection is a cost-effective strategy in HLRCC, across all age groups modelled.
Data availability statement
Data are available on reasonable request. The data analysed and example code for the decision model used in this study are available from the corresponding author on request.
Ethics statements
Patient consent for publication
Ethics approval
This study does not involve human participants.
Acknowledgments
ERM acknowledges support from the NIHR Cambridge Biomedical Research Centre, ERW from Cancer Research UK (C64726/A30910), and ERW and DGE from the Manchester NIHR Biomedical Research Centre (IS-BRC1215-20007). The University of Cambridge has received salary support (ERM) from the NHS in the East of England through the Clinical Academic Reserve.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @ER_Woodward
Collaborators N/A.
Contributors AJT, DGE, ERM and ERW conceptualised the project and ERW is its guarantor. AJT and ERW curated the data. AJT undertook the formal data analysis with assistance from YMA, VACR and ERW. All authors contributed to the interpretation of the results. AJT and ERW drafted the manuscript. All authors critically revised the manuscript.
Funding This work was funded by the NIHR Manchester Biomedical Research Centre (IS-BRC-1215-20007).
Disclaimer The views expressed are those of the authors and not necessarily those of the NHS or Department of Health.
Competing interests AJT declares conflicts with Perspectum. DGE declares consultancies with AstraZeneca, SpringWorks and Recursion and declares no conflict of interest specifically related to this study. The other authors declare no conflicts of interest.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.