Cost-effectiveness of volume computed tomography in lung cancer screening: a cohort simulation based on Nelson study outcomes

Abstract

Objectives

This study aimed to evaluate the cost-effectiveness of lung cancer screening (LCS) with volume-based low-dose computed tomography (CT) versus no screening for an asymptomatic high-risk population in the United Kingdom (UK), utilising the long-term insights provided by the NELSON study, the largest European randomized control trial investigating LCS.

Methods

A cost-effectiveness analysis was conducted using a decision tree and a state-transition Markov model to simulate the identification, diagnosis, and treatments for a lung cancer high-risk population, from a UK National Health Service (NHS) perspective. Eligible participants underwent annual volume CT screening and were compared to a cohort without the option of screening. Screen-detected lung cancers, costs, quality-adjusted life years (QALYs), and the incremental cost-effectiveness ratio (ICER) were predicted.

Results

Annual volume CT screening of 1.3 million eligible participants resulted in 96,474 more lung cancer cases detected in early stage, and 73,825 fewer cases in late stage, leading to 53,732 premature lung cancer deaths averted and 421,647 QALYs gained, compared to no screening. The ICER was £5,455 per QALY. These estimates were robust in sensitivity analyses.

Limitations

Lack of long-term survival data for lung cancer patients; deficiency in rigorous micro-costing studies to establish detailed treatment costs inputs for lung cancer patients.

Conclusions

Annual LCS with volume-based low-dose CT for a high-risk asymptomatic population is cost-effective in the UK, at a threshold of £20,000 per QALY, representing an efficient use of NHS resources with substantially improved outcomes for lung cancer patients, as well as additional societal and economic benefits for society as a whole. These findings advocate evidence-based decisions for the potential implementation of a nationwide LCS in the UK.

Keywords:

• Cost-effectiveness
• lung cancer
• screening
• computed tomography
• NELSON

JEL CLASSIFICATION CODES:

• D61
• D6
• D
• D60

Introduction

Lung cancer remains the leading cause of cancer-related death worldwide. In the United Kingdom (UK), approximately 35,157 lung cancer deaths occur every year, accounting for 21% of all cancer deaths¹. Among lung cancer patients, approximately 50% are diagnosed at stage IV, with a 5-year survival rate of 2.9%, while only 15% are identified at stage I, with a corresponding 5-year survival rate of 56.6%². Given the availability of curative-intent treatments for early-stage lung cancer, the shift in identification from a late stage to an early stage could confer a significant improvement in health outcomes for lung cancer patients³. Such shift in disease stage detection has been realized in various lung cancer screening (LCS) studies with the proportion of lung cancers detected at stage I ranging from 58.3% to 75.9%, and only 3.4% to 15.4% at stage IV^4–7.

The NELSON study is the largest European randomized LCS trial, reporting a lung cancer mortality reduction of 24-33% for high-risk individuals⁷. Moreover, the NELSON study characterizes lung nodule findings into negative, positive, and indeterminate-sized, i.e. a follow-up CT scan after a 3-month observation period is performed to classify the nodule as positive or negative, based on the volumetric doubling time. With this stratification approach, the false positive rate drops to only 1.2% in the NELSON study and is now a feature of British trials and pilot programs. For example, the UK Lung Cancer Screening trial (UKLS) showed a slightly higher but similar low rate (3.6%)^4,8. Various new LCS projects have been initiated since then, such as screening projects in Croatia, Poland, and the European Commission funded implementation study 4-IN-THE-LUNG-RUN^8–10. One of the most advanced national programs is the targeted Lung Health Check program in England, which used considerable insights gained from the NELSON study for its volumetric protocol and quality assurance standard¹¹. Furthermore, based on NELSON study outcomes, the European Respiratory Society has recently submitted an open letter urging the European Union to further promote LCS implementation¹².

Though the cost-effectiveness of LCS compared to no screening has been intensively investigated, the cost-effectiveness of volume CT screening based on the NELSON study and its long-term outcomes is yet to be explored in the UK^4,13–19. This study aimed to inform policy makers on the health and economic impact of implementing a national LCS program based on the volumetric protocol and NELSON study outcomes in the UK. A de novo health economic model was developed, incorporating the latest results from the NELSON study, to evaluate the health and costs impact, as well as the cost-effectiveness of the volume CT screening compared to no screening.

Methods

Evidence available on the cost-effectiveness of LCS was investigated to support the modelling approach and possible inputs for this study. The most common methodological approaches in modelling cancer screening are individual-level models, also known as microsimulation, and cohort-level Markov model. The utilization of these two approaches were reported to be 34% versus 41% in a review examining cancer screening simulation models²⁰. The microsimulation model is able to capture more detailed dynamics within the population. However, these models can be highly complex, computationally demanding, and difficult to evaluate and validate²¹. Additionally, the effectiveness of personalized LCS strategies are still being explored, while robust cohort results are available from the NELSON study. Therefore, our analysis used a decision tree to simulate the identification and diagnoses of lung cancer patients, then a state-transition Markov model was employed to simulate treatments and long-term survival for lung cancer patients, and in combination to assess the cost-effectiveness of LCS with volumetric CT.

The base-case analyses were conducted from the UK National Health Service (NHS) perspective using a lifetime horizon. The cost-effectiveness analysis followed the National Institute for Health and Care Excellence (NICE) technology appraisal methodology²². The primary health outcomes were the quality-adjusted life years (QALYs) and life years gained (LYG). The summary outcome was the incremental cost-effectiveness ratio (ICER), calculated by dividing incremental costs by incremental effects (i.e. QALYs). A 3.5% discount rate was used for both health outcomes and monetary outcomes.

Model structure

A decision tree, shown in Figure 1, was utilised to simulate the screening approach based on the NELSON study at a population cohort level. In the no screening arm, lung cancer patients were diagnosed through clinical presentation, and asymptomatic patients with pre-clinical disease were included as missed individuals. In the screening arm, the eligible population would either go through a volume CT scan or decide not to participate. Non-participant patients followed clinical presentation diagnoses, while screen-detected patients benefited from early detection. Screening participants who eventually had a negative baseline scan (either directly, after an indeterminate follow-up scan, or being a confirmed false positive scan after diagnostic work-ups) entered the next screening round (R2) in the next year, as annual screening is recommended for LCS currently²³. Up to a total of 17 annual screens were modelled, which reflected the mean age of participants in the NELSON study (58 years), while the maximum inclusion age for a scan being 74 years (i.e. up to 17 annual screening rounds)²⁴.

Figure 1. Decision tree for lung cancer screening with volume computed tomography.

Additionally, a state-transition Markov model was created to simulate disease progression and treatments for lung cancer patients from stage at diagnosis, as shown in Figure 2. Diagnosed lung cancer patients entered the model in a pre-progression state, and they either transited to the post-progression state if disease progressed or directly to the death state. Patients with progressed disease transited to the post-progression state, and eventually to the death state. The transition probabilities between states were based on the overall survival and the disease/progression-free survival per lung cancer stage at diagnosis. Background mortality was also considered in the model to adjust for other causes of death. Diagnosed individuals were assumed to be in one of the three mutually exclusive health states at any given time: 1) pre-progression; 2) post-progression; or 3) death (both lung cancer mortality and other causes mortality). The missed individuals with undetected disease were assumed to follow the transition as stage II lung cancer patients based on expert opinions, and likewise, they would transit to the death state in the end. The cycle length was set at 3 months, based on the current lung cancer treatment and follow-up scheme²⁵.

Figure 2. State-transition markov model based on the natural history of lung cancer.

Model inputs

Model inputs consisted of (1) eligible population, (2) screening performance and lung cancer epidemiology, (3) survival data, (4) utilities, and (5) costs. The main model parameters used in the base-case analysis are presented in Table 1. The assumptions made for the model structure and inputs are summarised in Supplementary Table S1.

Table 1. Input parameters for the base-case analysis.

Parameter		Base-case value	PSA distribution
Screening effectiveness ^24
NELSON round 1
Regular scan likelihood	Negative	79.21%	Dirichlet
	Indeterminate	19.20%	Dirichlet
	Positive	1.59%	Dirichlet
Indeterminate scan likelihood	Negative	94.57%	Dirichlet
	Positive	5.43%	Dirichlet
True negative		99.93%	Dirichlet
False negative		0.07%	Dirichlet
True positive^a		38.67%	Dirichlet
False positive^a		61.33%	Dirichlet
Stage distribution at diagnosis
Stage I		64.86%	Dirichlet
Stage II		9.46%	Dirichlet
Stage III		18.92%	Dirichlet
Stage IV		6.76%	Dirichlet
NELSON round 2 and onwards
Regular scan likelihood	Negative	92.17%	Dirichlet
	Indeterminate	6.58%	Dirichlet
	Positive	1.25%	Dirichlet
Indeterminate scan likelihood	Negative	91.23%	Dirichlet
	Positive	8.77%	Dirichlet
True negative		99.90%	Dirichlet
False negative		0.10%	Dirichlet
True positive		44.35%	Dirichlet
False positive		55.65%	Dirichlet
Stage distribution at diagnosis
Stage I		75.86%	Dirichlet
Stage II		6.90%	Dirichlet
Stage III		13.79%	Dirichlet
Stage IV		3.45%	Dirichlet
Epidemiology and demography
Stage distribution at diagnosis (no screening) ^2^,^29
Stage I		16.31%	Dirichlet
Stage II		8.56%	Dirichlet
Stage III		22.58%	Dirichlet
Stage IV		52.53%	Dirichlet
Total population ^61		67,866,004	Gamma
Lung cancer incidence aged 50–74 years ^28		28,105	Gamma
Smoking rate ^26^,^62		14.17%	Beta
Costs
Recruitment costs ^49
Pre-invitation notice letter		£0.66	Gamma
Invitation letter for participation		£0.79	Gamma
First invitation reminder (as needed)		£0.84	Gamma
Second invitation reminder (as needed)		£0.84	Gamma
Telephone triage call		£7.62	Gamma
Invitation letter for screening appointment		£0.70	Gamma
Screening costs ^42
CT-scan		£49.00	Gamma
Report reading		£21.00	Gamma
Diagnostic costs per person ^5^,^6^,^24^,^42
Screening (NELSON study)		£411	Gamma
No screening		£626	Gamma
Treatment costs ^13
For the first 90 days (a 3-month cycle)
Stage I		£6,240	Gamma
Stage II		£7,440	Gamma
Stage III		£6,874	Gamma
Stage IV		£5,862	Gamma
For the first year excluding the first 90 days (per 3-month cycle)
Stage I		£2,279	Gamma
Stage II		£2,489	Gamma
Stage III		£1,934	Gamma
Stage IV		£1,108	Gamma
For the second year onwards (per 3-month cycle)
Stage I		£432	Gamma
Stage II		£490	Gamma
Stage III		£419	Gamma
Stage IV		£309	Gamma
End of life costs (per patient) ^45		£3,817	Gamma
Survival
Overall survival (5-year survival rate) by stage at diagnosis ^35
Stage I		78.63%	NA.
Stage II		54.90%	NA.
Stage III		29.24%	NA.
Stage IV		5.91%	NA.
Disease/progression-free survival (1-year disease/progression-free survival rate) ^,^32^,^33^,^34^,^31
Stage I		87.80%	NA.
Stage II		87.80%	NA.
Stage III		41.81%	NA.
Stage IV		34.16%	NA.

NA Not applicable; PSA: Probabilistic sensitivity analysis.

^a true positive refers to the proportion of true positive scans among the total positive results, while false positive refers to the proportion of false positive scans among the total positive results, and the sum of them equals to 100%.

Eligible population

UK national population was selected by age (50–74 years) and smoking history to define the screening eligible population, according to the inclusion criteria in the NELSON study²⁶. In total, 2,810,666 people were considered eligible for LCS, and the actual screening population was estimated based on the LCS uptake rate observed in the UKLS (46.5%), equalling 1,306,960 people²⁷.

Screening effectiveness and lung cancer epidemiology

The LCS effectiveness was based on the NELSON study. In the NELSON study, the baseline screening (Round 1) detected relatively more stage III and IV lung cancer patients than future screening rounds, as patients who had missed the opportunity to be detected earlier through screening were cumulated in the first screening round. This made the population cohort in Round 1 heterogenic compared to future rounds. Therefore, the baseline screening effectiveness in the model was calibrated against NELSON Round 1 results, while the subsequent rounds in the model followed NELSON Round 2 results, conducted 1 year after Round 1, which indicated the annual incidence for screen-detected lung cancer patients. Interval cancers were also considered in the model and followed the clinical presentation diagnoses per stage. For the no screening arm, the lung cancer incidence and stage distribution were based on the data from the World Health Organization (WHO), Office for National Statistics (ONS), and Cancer Research UK (CRUK)^2,28,29.

Survival data

The disease-free survival curve for early stage lung cancer (I-II) was derived from a retrospective study conducted in a British healthcare centre³⁰. The progression-free survival curve for stage III lung cancer was obtained from a meta-analysis³¹. For stage IV lung cancer, survival was estimated from several studies, including LUX-Lung 3, KEYNOTE-189, and Impower 133, reflecting different common treatments and lung cancer types^32–34. Disease- and progression-free survival curves reflect the time to disease progression or time to death. To determine the time to disease progression without death, lung cancer stage-specific overall survival (OS) rates were subtracted from the corresponding disease and progression-free survival rates. These progression rates were used to inform the transition from the pre-progression state to the post-progression state in the model (Supplementary Table S2). Overall survival for lung cancer by stage at diagnosis was informed by the Kaplan-Meier (KM) survival curves published by the International Association for the Study of Lung Cancer (IASLC)³⁵. These OS data were used to inform the transitions from the pre- and post-progression state to the lung cancer death state in the model (Supplementary Table S3). For the missed individuals, it was assumed that they had underlying early stage lung cancer, hence the survival of stage II lung cancer patients was applied.

To reflect a lifetime horizon, survival extrapolation was needed, based on the statistical method supported by the NICE Technical Support Document and Guyot et al.^36,37. First, a simulated patient level dataset was derived, then parametric extrapolation techniques were applied to predict longer term survival. Details about the distribution functions fitted per extrapolated survival curve are presented in Supplementary Table S4. Background mortality was used to account for all-cause mortality in the model, based on life tables for England and Wales³⁸.

Utilities

Health status was estimated based on the utility values for different health states in the model. For the pre-progression state, utility values were derived from a study investigating 2344 lung cancer patients using the SF-6D questionnaire with a UK population value set. The utility estimates were 0.71, 0.68, 0.67, and 0.66 for lung cancer patients stage I to IV, respectively³⁹. For the post-progression state, utility values were adopted from the corresponding pre-progression state. For first-diagnosed stage I and II patients, a Stage III utility value of 0.67 was applied; while for first-diagnosed stage III, a Stage IV value of 0.66 was applied; and the utility value for Stage IV patients was conservatively assumed to remain at 0.66. For lung cancer free participants and missed individuals, age-dependent utility values based on the population norms were applied⁴⁰.

Costs

To calculate the costs from the UK healthcare system perspective, direct costs were categorised into recruitment costs, screening costs, diagnostic costs, and treatment costs. Recruitment costs were incurred by sending invitation letters and reminders, making telephone triage calls, and scheduling screening appointments, which followed the recruitment strategy reported by the Yorkshire Lung Screening Trial⁴¹. Screening costs consisted of the CT scan itself plus report reading costs, and the values were derived from the National Tariff Workbook⁴².

Diagnostic costs referred to the costs incurred after positive scan results in the screening arm, or costs incurred after clinical presentation with lung cancer-related symptoms in the no-screening arm. The procedures and their corresponding frequencies for diagnostic work-up were derived from the NELSON study²⁴. The unit costs for each procedure, such as further imaging, bronchoscopy, percutaneous cytologic analysis or biopsy were informed by the National Tariff Workbook⁴². Diagnostic costs per person were calculated by weighting the utilization per diagnostic procedure with the corresponding costs for each procedure (Supplementary Table S5).

The computation of treatment expenses was based on a systematic review encompassing the costs incurred during the first year, which covered initial treatments, and it also reported the costs for subsequent years¹⁴. The initial treatment costs were based on a retrospective 1-year cohort study, which consisted of all emergency, inpatients and outpatient costs from the records of 3274 lung cancer patients⁴³. The costs for subsequent years were estimated to be 13% of the first-year costs, aligning with the front-loaded allocation of resources following a lung cancer diagnosis⁴⁴. Additionally, the end-of-life treatments costs were applied to all patients who died from lung cancer⁴⁵. All costs were converted and inflated to represent 2022 UK prices⁴⁶.

Sensitivity analyses

One-way (univariate) sensitivity analysis (OSA) was conducted with deterministic changes of ±20% to parameter values, and the results were used to identify the main drivers for the cost-effectiveness of LCS, presented in a Tornado diagram. Probabilistic sensitivity analysis (PSA) was conducted by simulating the whole cohort with a set of parameter values sampled probabilistically from suitable distributions reflecting parameter uncertainties (Table 1).

Scenario analyses

Several scenarios were explored. Firstly, the cost-effectiveness of LCS was evaluated from a societal perspective, which included indirect costs such as productivity loss, informal care, and transportation costs, in addition to the direct healthcare costs from a healthcare system perspective. Productivity loss was calculated based on the human capital approach, and consisted of two elements, the premature death of patients before retirement age and the absence from the labour forces due to illness. Secondly, during the last decade, the introduction of novel medicines, particularly in later stage lung cancer, are increasingly used in clinical practice with improved outcomes for patients, but potentially at an increased treatment costs in this later stage⁴⁷. The influence of this trend on the cost-effectiveness of LCS was explored in a scenario analysis as well. Additionally, other scenario analyses were conducted to investigate the cost-effectiveness of LCS under various conditions, including LCS uptake rates for the target population, different time horizons and discounting rates. All the parameters used in scenario analyses are summarised in Supplementary Table S6.

Results

Base-case results

Based on the UK population, a cohort of 1,306,960 participants were screened among 2,810,666 eligible individuals, utilising the volumetric protocol and NELSON study outcomes. Screening resulted in 96,474 more early stage (stage I-II) lung cancer diagnoses and a decrease of 73,825 late stage (stage III-IV) diagnoses. The 274,853 lung cancer deaths in the no screening arm decreased to 239,660 lung cancer deaths when implementing LCS, indicating that 53,732 premature lung cancer deaths would be averted. There were 592,668 life years gained, and 421,647 additional QALYs, equivalent to 1.18 additional QALYs gained per lung cancer patient diagnosed through LCS.

Over the 17 years, recruitment costs and screening costs amounted to approximately £18.30 million and £1.14 billion. On an annual basis these were the equivalent of £1.08 million and £66.81 million, respectively. Over a lifetime, screening led to an increase in total treatment costs of £1.10 billion (£26.08 million per annum). However, treatment costs in stage III and IV decreased by £754.04 million (£17.95 million per annum). Total incremental costs were around £2.30 billion (£54.76 million per annum). The equivalent ICER was £5,455 per QALY gained, for an annual LCS programme in the UK from a healthcare system perspective. The net monetary benefit (NMB) was approximately £6.13 billion with a WTP threshold of £20,000 per QALY over a lifetime horizon (Table 2).

Table 2. Main results from the base-case analysis.

	Screening	No screening	Incremental
Clinical and health outcomes
Lung cancer diagnoses
Total diagnoses	356,414 (100%)	333,765 (100%)	22,649
Stage I	151,317 (42%)	54,449 (16%)	96,868
Stage II	28,249 (8%)	28,642 (9%)	−393
Stage III	67,191 (19%)	75,356 (23%)	−8,165
Stage IV	109,658 (31%)	175,317 (53%)	−65,660
Missed individuals	NA.	22,649	NA.
Stage III and IV averted	73,825
Lung cancer deaths
Total	239,660	293,392	−53,732
Stage I	57,296	20,606	36,690
Stage II	13,476	13,657	−181
Stage III	61,558	68,997	−7,439
Stage IV	107,330	171,593	−64,263
Missed individuals^a	NA.	18,539	NA.
Life years
Total	44,548,981	43,956,312	592,668
Stage I	1,248,689	448,351	800,338
Stage II	187,909	188,939	−1,030
Stage III	185,745	207,336	−21,591
Stage IV	123,783	197,785	−74,002
Missed individuals	NA.	111,046	−111,046
Lung cancer free participants^b	42,802,855	42,802,855	0
Quality-adjusted life years
Total	34,810,730	34,389,083	421,647
Stage I	893,554	320,801	572,753
Stage II	134,250	134,908	−657
Stage III	123,531	137,890	−14,359
Stage IV	81,697	130,538	−48,842
Missed individuals	NA.	87,248	−87,248
Lung cancer free participants	33,577,698	33,577,698	0
Costs
Total	£6,807,106,541	£4,507,006,217	£2,300,100,325
Recruitment costs	£18,302,102	NA.	£18,302,102
Screening costs	£1,135,716,437	NA.	£1,135,716,437
Diagnostic costs	£214,240,181	£163,456,583	£50,783,598
Treatment costs	£5,438,847,821	£4,343,549,634	£1,095,298,187
Stage I	£2,893,262,559	£1,039,595,392	£1,853,667,167
Stage II	£492,477,752	£496,810,756	−£4,333,005
Stage III	£984,048,195	£1,098,919,579	−£114,871,384
Stage IV	£1,069,059,315	£1,708,223,906	−£639,164,591
Health economic outcomes
ICER	£5,455
NMB	£6,132,840,946

NA, Not applicable; ICER, Incremental cost-effectiveness ratio; NMB, Net monetary benefit.

^aThe 18,539 lung cancer deaths for missed individuals in the no screening arm indicated that some asymptomatic lung cancer patients went undiagnosed, of which the deaths were lung cancer related.

^bLung cancer free participants refer to individuals who either do not have lung cancer, or have not been identified with lung cancer.

Sensitivity analyses

One-way sensitivity analysis resulted in small changes in the ICER, all within UK WTP thresholds. The main variation came from the changes in the discount rate applied to the health effect, the unit costs for a CT scan, and the utility values for stage I (Figure 3). After 1,000 iterations, probabilistic sensitivity analysis resulted in an average ICER of £5,728 per QALY. Figure 4 shows the spread of probabilistic simulations in the form of an incremental cost-effectiveness scatterplot, in which 100% of simulations resulted in an ICER under £20,000 per QALY.

Figure 3. Tornado diagram from one-way sensitivity analysis.

Figure 4. Incremental cost-effectiveness scatterplot from probabilistic sensitivity.

Scenario analyses

Results for scenario analyses are presented in Table 3. Incorporating a societal perspective revealed a lower ICER (£5,444 per QALY), while including screening associated disutility and increasing background mortality by 100% resulted in higher ICERs, £5,822 and £6,495 respectively. All scenarios resulted in an ICER well below the UK’s WTP threshold (£20,000 per QALY), except for the scenarios with a time horizon of 5 years, indicating an ICER of £61,266 per QALY.

Table 3. Results from the scenario analyses.

Scenario name	Screening		No screening		Incremental costs	Incremental QALYs	ICER
	Total costs	Total QALYs	Total costs	Total QALYs
Base-case analysis	£6,807,106,541	34,810,730	£4,507,006,217	34,389,083	£2,300,100,325	421,647	£5,455
Societal perspective	£8,387,309,306	34,810,730	£6,092,058,373	34,389,083	£2,295,250,933	421,647	£5,444
Screening uptake rate (25%)	£5,760,540,037	34,664,488	£4,515,464,698	34,437,796	£1,245,075,340	226,692	£5,492
Screening uptake rate (75%)	£8,194,415,627	35,004,586	£4,495,793,811	34,324,511	£3,698,621,816	680,076	£5,439
Number of screening rounds (3 rounds)	£1,761,592,486	36,300,298	£1,123,109,385	36,178,732	£638,483,101	121,565	£5,252
Number of screening rounds (5 rounds)	£2,746,821,075	35,972,294	£1,782,257,145	35,783,076	£964,563,930	189,218	£5,098
Number of screening rounds (10 rounds)	£4,792,218,741	35,353,094	£3,152,860,955	35,035,808	£1,639,357,786	317,285	£5,167
Time horizon (5 years)	£2,933,968,281	10,570,019	£1,472,488,052	10,546,164	£1,461,480,229	23,855	£61,266
Time horizon (10 years)	£4,618,047,939	18,676,772	£2,835,696,855	18,581,651	£1,782,351,083	95,121	£18,738
Time horizon (15 years)	£5,979,670,158	24,721,666	£3,923,237,061	24,529,562	£2,056,433,097	192,104	£10,705
Discount rates (cost: 0%, health outcome: 0%)	£9,110,781,637	52,403,724	£6,020,769,542	51,644,472	£3,090,012,095	759,252	£4,070
Discount rates (cost: 6%, health outcome: 6%)	£5,686,379,843	27,601,126	£3,765,395,804	27,308,399	£1,920,984,039	292,727	£6,562
OS for missed individuals follows general population	£6,807,106,541	34,810,730	£4,507,006,217	34,503,800	£2,300,100,325	306,930	£7,494
OS for missed individuals follows stage I patients	£6,807,106,541	34,810,730	£4,507,006,217	34,423,678	£2,300,100,325	387,052	£5,943
OS for missed individuals follows stage III patients	£6,807,106,541	34,810,730	£4,507,006,217	34,352,936	£2,300,100,325	457,795	£5,024
OS for missed individuals follows stage IV patients	£6,807,106,541	34,810,730	£4,507,006,217	34,323,000	£2,300,100,325	487,731	£4,716
Include screening associated disutility	£6,807,106,541	34,784,144	£4,507,006,217	34,389,083	£2,300,100,325	395,060	£5,822
Increase background mortality by 100%	£6,300,547,894	29,843,934	£4,174,940,899	29,516,681	£2,125,606,995	327,253	£6,495
Include novel treatments for stage IIIB and IV lung cancer^a	£6,987,150,224	34,814,859	£4,724,820,024	34,394,983	£2,262,330,199	419,876	£5,388

ICER, Incremental cost-effectiveness ratio; QALYs, Quality-adjusted life years; OS, overall survival.

^athis scenario analysis took into account both the treatment effects and associated costs of the novel treatments for IIIB and IV lung cancer patients.

Discussion

This study evaluated the cost-effectiveness of volume-based low-dose CT screening versus no screening in the UK, based on the NELSON study outcomes, resulting in an ICER of £5,455 per QALY and 421,647 additional QALYs across eligible participants. The results were robust as indicated by sensitivity analyses. All analyses remained within the threshold of £20,000 per QALY, a commonly used WTP in the UK, providing further confidence in the results and the underlying model. In addition, the study demonstrated that implementing annual volume CT screening over 17 rounds would detect 96,474 more lung cancer cases in stage I and II, and 73,825 fewer cases in stage III and IV, leading to 53,732 premature lung cancer deaths averted, compared to no screening, as early detection provides patients with the opportunity to benefit from curative-intent treatments and better prognosis. In our modelling study, the lung cancer mortality reduction was estimated to be around 18% over a lifetime horizon, a figure notably consistent with outcomes observed in extensive clinical trials utilising low-dose CT for LCS. Specifically, the NELSON study demonstrated a 24 − 33% reduction in lung cancer mortality over a 10-year follow-up period⁷, while the National Lung Screening Trial conducted in the United States reported a 20% reduction in the same context⁴⁸.

The cost-effectiveness of LCS has been analysed in various studies, and the results were inconsistent^13–19. In the UK, three main economic evaluations were conducted. One study concluded that the once-only CT screening under the UKLS protocol could be cost-effective with an ICER of £8,466 per QALY^4,19. However, the UKLS trial was not powered to evaluate mortality reduction in a relatively short follow-up timeframe, thus the observational element of the economic evaluation was restricted to those events and findings that occurred within the active trial period⁴. Another study used the economic evaluation method reported in the UKLS trial, applying Manchester, UK specific evidence where possible, concluding that the Manchester program was a cost-effective use of limited NHS resources, based on an ICER of £10,069 per QALY¹⁹. The third analysis was a microsimulation based economic evaluation by Snowsill et al. resulting in an ICER of £28,000 per QALY gained for LCS with a single CT scan in smokers aged 60 − 75 years with a ≥ 3% risk of lung cancer¹³. This estimate is higher compared to the results from our model, which may be explained by various factors. Firstly, we used the NELSON age inclusion criteria (50 − 74 years versus 55 − 80 years in Snowsill et al.), which might lead to a better-defined target cohort, and a more cost-effective approach due to improved outcomes. Secondly, our model utilised the NELSON screening outcomes, which benefits from the volumetric nodule management algorithm that provides a high sensitivity at a relatively lower false-positive rate, compared to other LCS trials. Thirdly, the calibration of the microsimulation model produced lung cancer stage distributions that were remarkably different from the observed data in multiple large-scale LCS trials. For example, the NELSON study showed a proportion of 64% diagnosed at stage I versus 7% at stage IV in the baseline screening, while these numbers were 32% (stage I) and 54% (stage IV) for annual LCS in the Snowsill et al. analysis, indicating a lower impact of LCS on stage shift. An interim report incorporating recent data and expert opinions to update the previously-mention model was published by the UK National Screening Committee recently, and suggested that the ICER for the most cost-effective strategy was £1,529 per QALY⁴⁹.

The one-way sensitivity analysis (OSA) demonstrated that the discount rate on health effects had the biggest impact on the ICER (Figure 3). In the study, a discount rate of 3.5% was used for both health effects and costs according to the NICE guidelines²². However, this equal discount rate approach has been debated frequently in literature⁵⁰. Dutch and Belgian guidelines now recommend a lower discount rate for benefits than costs, as to avoid a too strong penalization of interventions that generate most of their benefits in the future, such as screening and vaccination programs⁵¹. Hence, LCS has the potential to yield more healthcare benefits and a lower ICER if a lower discount rate was used. Additionally, it is noteworthy that conducting a scenario analysis with a 2% discount rate for the health effects resulted in an ICER of £4,290 per QALY, reinforcing the aforementioned argument.

The OSA concurrently revealed that the CT scan costs also exert a notable influence on the ICER (Figure 3). The unit cost of a CT scan was £70 in the model, derived from the NHS Tariff Workbook⁴². It consisted of costs for CT scan itself (£49) and report reading costs (£21). The artificial intelligence (AI) assisted image reading has the potential to support radiologists with image reading and interpretation, reducing time spent, thus lowering the CT scan costs and further improve the cost-effectiveness of LCS⁵². Furthermore, scenario nalyses revealed a notable decrease in the ICER to £5,032 and £4,839 per QALY, when the image reading costs were reduced to £10 and £5, correspondingly. This implies that efforts should be dedicated to containing CT costs to ensure the long-term sustainability of implementation LCS in the UK.

Scenario analyses were conducted to explore the impact of LCS uptake rates. Under the uptake rates of 25% and 75%, the ICERs were £5,492 and £5,439, respectively, which were similar to the base-case figure (£5,455). It was mainly explained by the proportional changes in costs and QALYs following the variation in screening population size. However, increasing LCS uptake rate provides more clinical benefits. For example, additional 79,381 late-stage lung cancer cases and 57,777 premature lung cancer deaths would be averted if the uptake rate increased from 25% to 75%. In addition, evidence shows that disadvantaged groups are at a greater risk of experiencing poor survival due to lung cancer’s late presentation⁵³. The access to LCS could partially address this health inequality through early detection and promote a fair distribution of health resources across society, which is also on the priority list of Europe’s Beating Cancer Plan⁵⁴.

Furthermore, another scenario analysis incorporating indirect costs estimated an ICER of £5444 per QALY from the societal perspective, with additional cost savings of approximately £138.77 million from productivity gains. It considered LCS's potential impact on the ability of working-age individuals diagnosed with lung cancer to sustain employment by preventing premature mortality and improving prognoses through early detection. Additionally, from a broader societal and economic view, it is reported that in the UK, resource allocation criteria used by the government or its agencies in the health sector values life and health significantly lower than the other non-health departments⁵⁵. Therefore, LCS could possibly provide values beyond the current estimates if this potential imbalance in the value of the same attribute (health and life) across public sectors was to be corrected.

In our model, we observed more lung cancers detected in the screening arm than in the no screening arm, which might be due to under-reporting of true clinically presenting lung cancer cases or overdiagnosis. Overdiagnosis is commonly defined as screen-detected disease that, in the absence of screening, would not have become clinically evident within the participant’s lifetime. The overdiagnosis rate is 6.79% based on the 46.5% LCS uptake rate in the base-case analysis in our model, and this is consistent with what has been noted in multiple trials^56,57. With the additional lung cancers detected in the screening arm, results from our model demonstrates that volume CT screening is consistently cost-effective. In addition, this overdiagnosis is likely to flatten with longer follow-up, as it is reported that some lung cancer cases could eventually present clinically after more than 10 years, making it challenging to compare the lung cancer incidence between both arms in the study⁵⁸.

The lack of high-quality data is frequently referred to as a key limitation in health economic quantitative research. Specifically, up-to-date data on the local treatment costs are scarce. Consequently, treatment costs used in the model were sourced from a single teaching hospital in the UK and reflected 2013 costs. The recent advances in novel targeted therapy and immunotherapy improve survival outcomes for late-stage lung cancer patients, but also have led to dramatic increase in costs⁵⁹. Scenario analyses were conducted to investigate the specific impact of integrating novel drugs into the treatment landscape for lung cancer patients. The results showed that LCS exhibits the potential for greater cost-effectiveness with a ICER of £5388 per QALY. This improvement is attributed to the earlier detection of lung cancer patients, subsequently leading to reduced treatment costs in the new treatment paradigm with novel therapeutics. This implies that LCS could contribute to tackling the escalating worldwide burden associated with rising cancer-related expenditures, while simultaneously enhancing cancer prognosis through early detection.

Well-established randomized control studies have been utilised for the effectiveness of LCS to support the economic evaluations. Further research ought to prioritize the acquisition of high-quality, localized data for specific parameters, such as up-to-date costs to reflect advancements in treatment practice, benefits of returning to work, the long-term survival for lung cancer patients, among other relevant facets. In addition, LCS has the potential to provide and health promotion opportunities at the point of screening for other non-communicable diseases, including chronic obstructive pulmonary disease (COPD) and cardiovascular disease (CVD), in high-risk LCS participants. If these benefits would be quantified in clinical trials and effectiveness evidence were made available for decision modelling, volume CT screening is expected to be even more cost-effective than the current estimates, as it would generate better prognosis for patients with COPD and CVD through early detection without incurring additional screening costs⁵⁹. Despite the existence of data gaps and unquantified benefits, all sensitivity analyses in our study showed robustness in the results. Bridging data gaps would provide further precision in the prediction of cost-effectiveness for LCS.

Conclusion

Annual LCS with volume-based low-dose CT for a high-risk asymptomatic population is cost-effective in the UK, at a threshold of £20,000 per QALY, representing an efficient use of NHS resources with substantially improved outcomes for lung cancer patients, as well as additional societal and economic benefits for society as a whole. These findings advocate evidence-based decisions for the potential implementation of a nationwide LCS in the UK.

Transparency

Declaration of financial/other interests

X.P., D.R. and H.B. are employed by iDNA. E.D. and M.O have a financial interest in iDNA. J.R. is employed by AstraZeneca and owns shares in the company. D.B. reports lecture fees from MSD, BMS, Roche, and AstraZeneca; he is the advisor to UK National Screening Committee. H.JM.G reports grants from Cancer-id EU funding and Interreg grant outside the submitted work; consulting fees from Eli Lilly and Novartis. M.J.P reports EU grants outside the submitted work; personal fees from Asc Academics; he is the shareholder at Health-Ecore in Zeist (NL) and Pharmacoeconomics Advice in Groningen (NL).

Author contributions

X.P., E.D., D.R., and R.V. designed the model and the computational framework. M.P. verified the analytical method. X.P., D.R., and H.B. performed the modelling. J.R worked out the model technique critique. D.B., H.G., and M.O supported with clinical validation for the model parameters. X.P. took the lead in writing the manuscript. All authors provided critical feedback and helped shape the research, analysis and manuscript.

Reviewer disclosures

Peer reviewers on this manuscript have received an honorarium from JME for their review work but have no other relevant financial relationships to disclose.

Previous presentations

An abstract based on the research was presented as a poster in the conference ISPOR Europe 2022 conference in Vienna.

Acknowledgements

No assistance in the preparation of this article is to be declared.

Additional information

Funding

This work was funded by AstraZeneca PLC.