Sage Journals: Discover world-class research

Abstract

French

PurposeScreening for lung cancer is recommended to reduce lung cancer mortality, but there is no consensus on patient selection for screening in Canada. Risk prediction models are more efficacious than the screening recommendations of the Canadian Task Force on Preventive Health Care (CTFPHC), but it remains to be determined which model and threshold are optimal. MethodsWe retrospectively applied the PLCO_m2012, PLCO_all2014 and LLPv2 risk prediction models to 120 lung cancer patients from a Canadian province, at risk thresholds of ≥ 1.51% and ≥ 2.00%, to determine screening eligibility at time of diagnosis. OncoSim modelling was used to compare these risk thresholds. ResultsSensitivities of the risk prediction models at a threshold of ≥ 1.51% were similar with 93 (77.5%), 96 (80.0%), and 97 (80.8%) patients selected for screening, respectively. The PLCO_m2012 and PLCO_all2014 models selected significantly more patients for screening at a ≥ 1.51% threshold. The OncoSim simulation model estimated that the ≥ 1.51% threshold would detect 4 more cancers per 100 000 people than the ≥ 2.00% threshold. All risk prediction models, at both thresholds, achieved greater sensitivity than CTFPHC recommendations, which selected 56 (46.7%) patients for screening. ConclusionCommonly considered lung cancer screening risk thresholds (≥1.51% and ≥2.00%) are more sensitive than the CTFPHC 30-pack–years criterion to detect lung cancer. A lower risk threshold would achieve a larger population impact of lung cancer screening but would require more resources. Patients with limited or no smoking history, young patients, and patients with no history of COPD may be missed regardless of the model chosen.

Keywords

lung cancer risk prediction models screening guidelines risk threshold

Introduction

Lung cancer has one of the lowest 5-year survival rates of all cancers in Canada at 22%.¹ This can be attributed to the high proportion of lung cancers diagnosed at late stages,² which are associated with reduced survival rates.³ Fortunately, earlier diagnosis with low-dose computed tomography (LDCT) screening programs reduce lung cancer mortality.⁴

The Canadian Task Force on Preventive Healthcare (CTFPHC) recommends lung cancer screening with annual LDCT for three consecutive years in patients 55–74 years old with ≥ 30 pack-years of smoking history who continue to smoke or have quit within the past 15 years.⁵ These recommendations were based on the National Lung Screening Trial (NLST), an American study which demonstrated a 20% reduction in lung cancer mortality when patients were screened with LDCT compared to chest radiography.⁶ Despite the CTFPHC recommendations, Ontario is currently the only Canadian province with a formal lung cancer screening program, though this program is not yet province-wide.⁷ Informal screening occurs on a limited basis in Canada, but many physicians are not following CTFPHC guidelines.⁸

While NLST-based guidelines are prevalent in Canada and elsewhere, other proposed screening methods include risk calculators. Age and smoking are the sole criteria used for NLST-based screening; however, other important lung cancer risk factors include air pollution and exposures to industrial chemicals, asbestos, and radon, as well as low socioeconomic status, a personal history of chronic obstructive pulmonary disease (COPD) or cancer, and a family history of lung cancer.⁹ Tammemägi et al¹⁰ incorporated some of these factors into a prediction model entitled PLCO_m2012, and determined that patients with a risk score ≥ 1.51% had consistently lower mortality rates in the LDCT arm compared to the chest radiography arm of NLST cohorts. Furthermore, when applied to the participants of the Prostate, Lung, Colorectal, and Ovarian (PLCO) Screening Trial, and more recently, an international prospective cohort study,¹¹ this ≥ 1.51% risk threshold achieved better sensitivity, specificity and positive predictive value than NLST-based criteria.¹² Tammemägi and colleagues also developed the PLCO_all2014 risk calculator to incorporate never-smokers; however, as no PLCO never-smokers met a ≥ 1.51% risk threshold, they concluded that never-smokers should not be screened.¹²

A prototype of the PLCO_m2012 model was used in the prospective Pan-Canadian Early Detection of Lung Cancer (PanCan) Study that showed a risk threshold of ≥ 2.00% can lead to superior efficacy in lung cancer screening with earlier cancer detection compared to the NLST model.¹³ While PanCan did not assess a ≥ 1.51% risk threshold, an Australian study on ever-smokers determined that sensitivity of a ≥ 1.51% risk threshold was 11% greater than a ≥ 2.0% threshold, while the specificity decreased by 8.5%.¹⁴

The PLCO_m2012 model is also being used for lung cancer screening in the UK¹⁵ alongside the Liverpool Lung Project (LLPv2) model which was developed from a case-control study in Liverpool, UK.¹⁶ However, the National Health Service (NHS) has proposed a higher risk threshold at ≥ 2.5% for the LLPv2 to limit unnecessary imaging.¹⁷ LLPv2 ≥ 2.5% selected an equal number of patients for screening as PLCO_m2012 ≥ 1.51% in a Manchester, UK cohort of ever-smokers, and both of these criteria selected more patients than NLST criteria.¹⁸

Although there are rapid assessment review panels that expedite diagnosis and treatment of lung cancer, such as our Thoracic Triage Panel (TTP),¹⁹ there are few formal screening programs and trials in Canada. While there are screening programs under development, there is not yet an official consensus on whether risk prediction models should be used in lieu of CTFPHC guidelines. Therefore, the purpose of this study was to retrospectively compare the sensitivities of three well-known lung cancer risk prediction models, (1) PLCO_m2012, (2) PLCO_all2014 and (3) LLPv2, against CTFPHC recommendations and determine which risk model and risk threshold selects the most patients in a cohort who have confirmed lung cancer.

Methods

We performed a retrospective chart review on patients diagnosed with lung cancer between 2015–2019 in the Eastern Health region of Newfoundland and Labrador (NL) to determine whether they would have been eligible for screening at time of diagnosis. We randomly selected 300 of 1000 patients with confirmed lung cancer from the pre-existing TTP database.

Three risk-prediction models were applied to these patients to retrospectively calculate the risk of developing lung cancer: (1) PLCO_m2012, calculated through the online ‘Should I Screen’ webpage provided by the University of Michigan^10,20; (2) PLCO_all2014, a Canadian model developed by Tammemägi and colleagues¹²; and (3) the Liverpool Lung Project (LLPv2) Risk Model, developed by Cassidy et al.¹⁶ Each model relies on various predictive factors, and these variables were not available for all our patients. To ensure high-quality and reliable data, we included only 120 patients who had sufficient data for risk calculations using all three models.

Data were collected from the TTP database and electronic medical records. Patients’ smoking history, age at diagnosis, sex, weight, height, level of education and ethnicity were recorded. If the level of education was not clearly stated, it was assumed to be the lowest education required for the patient’s employment. Additional variables were collected for medical history such as a self-reported 1st- or 2nd-degree family history of lung cancer; history of malignancy or asbestos exposure; and previous diagnosis of pneumonia, emphysema, bronchitis, tuberculosis or COPD, determined based on physician charting.

The above information was used to determine the number of patients who would have been selected based on CTFPHC lung cancer screening guidelines. Additionally, the number of patients who had a risk prediction of ≥ 2.00% and ≥ 1.51% in all three models were tallied, representing patients who would have been eligible for screening according to the risk threshold of the PanCan study,¹³ and the risk threshold recommended by Tammemägi et al,¹² respectively.

OncoSim was used to estimate the effect of changing risk thresholds on the number of patients screened. OncoSim is a microsimulation model that simulates cancer in the Canadian population using Canadian demography data (birth, immigration, migration and death) and cancer incidence and mortality data from the Canadian Cancer Registry.^21,22 Briefly, the OncoSim-Lung model simulates lung cancer risk in smokers and non-smokers using the age- and smoking-related coefficients in the PLCO₂₀₁₄ risk equation. It also accounts for increased risk of lung cancer with radon exposure and has a residual term that takes into account other factors; the residual term was calibrated to match Canadian lung cancer incidence data by age, sex and province.²³ OncoSim simulates the smoking behaviour of the Canadian population (smoking initiation starting from age 15 years, quit rate and smoking intensity) using data from three large Canadian health surveys.^24-26 Smoking trajectories were externally validated against other survey years and tobacco manufacturers’ data,^27,28 and smoking initiation rates were also calibrated to match the smoking prevalence by 5-year age group, sex and province.²⁹ To project the impact of lung cancer screening, we assumed lung cancer screening started in 2022; 40% of individuals eligible would attend screening, and the target participation rate would be reached in the next 3 years. This analysis is based on the Canadian Partnership Against Cancer’s OncoSim model. OncoSim has been made possible through a financial contribution from Health Canada, through the Canadian Partnership Against Cancer. The assumptions and calculations underlying the simulation results were prepared by Eastern Health and the responsibility for the use and interpretation of these data is entirely that of the authors.

Data were presented as mean and standard deviations for continuous data and frequency and percentages for categorical data. Student t-tests were used for continuous data, χ² tests for categorical data and Mann–Whitney U test for ordinal data. McNemar’s χ² tests were used to determine whether there were significantly more patients that would have been selected for screening by one model compared to the others, or by the ≥ 1.51% threshold compared to the ≥ 2.00% threshold. Statistical analyses were performed using SPSS version 27.0 (IBM SPSS Statistics) and R version 3.6.1. A P-value < .05 was considered significant.

This study was approved by the NL Health Research Ethics Board (HREB) and the Research Proposals Approval Committee (RPAC) of Eastern Health.

Results

Most patients were diagnosed at ≥ 60 years of age with equal numbers of men and women (Table 1). Only 5.0% of patients never smoked, while the rest had a 39.8 ± 15.0 pack-year history. 33.3% of patients were ex-smokers and quit 14.6±11.6 years before their lung cancer diagnosis. Notably, 40.0% of patients had COPD and 23.3% had a family history of lung cancer. Most patients were diagnosed at an advanced stage, with 39.2% having metastatic disease at presentation.

Table 1.

Patient characteristics which were used in the calculation of risk of lung cancer with any of PLCO_m2012, PLCO_all2014 and LLPv2 models.

Characteristic	Patients (N = 120) no. (%) or mean ± sd
Age group (yrs)
< 50	2 (1.7)
50–59	16 (13.3)
60–69	50 (41.7)
> 70	52 (43.3)
Sex
Female	59 (49.2)
Male	61 (50.8)
Estimated education
Less than high school	0 (0.0)
High school	62 (51.7)
Post high school training	18 (15.0)
Some college	12 (10.0)
College	13 (10.8)
Postgraduate/professional	15 (12.5)
Race/Ethnicity
White	120 (100.0)
Non-white	0 (0.0)
BMI	26.5 ± 5.8
Duration of smoking (yrs)	39.8 ± 15.0
Cigarette packs/day	1.1 ± 0.6
Quit time (former smokers) (yrs)	14.6 ± 11.6 (n = 40)
Smoking status
Current	74 (61.7)
Former	40 (33.3)
Never	6 (5.0)
Medical history
COPD	48 (40.0)
Pneumonia	6 (5.0)
Emphysema	10 (8.3)
Bronchitis	1 (0.8)
Tuberculosis	0 (0.0)
Past history of malignancy	8 (6.7)
Exposure to asbestos	2 (1.7)
Family history of lung cancer	28 (23.3)

In retroactively applying the CTFPHC criteria, only 56 (46.7%) of lung cancer patients would have been selected for screening (Table 2), which was a significantly lower proportion compared to those selected by each of the three risk models (P < .001). Using a ≥ 2.00% risk threshold, the numbers selected for screening by the PLCO_all2014, PLCO_m2012 and LLPv2 models were 82 (68.3%), 91 (75.8%) and 90 (75%), respectively. The PLCO_m2012 model selected significantly more patients than the PLCO_all2014 model (P = .046; Table 2). Lowering the risk threshold to ≥ 1.51%, the PLCO_all2014 model selected 14 more patients for a significant increase to 96 (80.0%; P < .001; Table 2), the PCLO_m2012 model selected 6 more patients for a significant increase to 97 (80.8%; P = .041; Table 2), and the LLPv2 model selected 3 more patients which did not significantly alter patient selection (P = .25; Table 2).

Table 2.

Comparisons of the number of patients who would have been selected for screening, between each of three risk predication models, at each of two risk thresholds, and the CTFPHC guidelines (N = 120).^a

Model	Threshold, %	PLCO_all2014		PLCO_m2012		LLPv2		CTFPHC
Model	Threshold, %	1.51%	2.00%	1.51%	2.00%	1.51%	2.00%
PLCO _all2014	1.51	--	96 vs 82 (p < .001^b*)	96 vs 97 (p = 1.000)	--	96 vs 93 (p = 1.000)	--	96 vs 56 (p < .001^b*)
PLCO _all2014	2.00	82 vs 96 (p < .001^b*)	--	--	82 vs 91 (p = .046^b*)	--	82 vs 90 (p = .765)	82 vs 56 (p < .001^b*)
PLCO_m2012	1.51	97 vs 96 (p = 1.000)	--	--	97 vs 91 (P = .041^b*)	97 vs 93 (p = 1.000)	--	97 vs 56 (p < .001^b*)
PLCO_m2012	2.00	--	91 vs 82 (p = .046^b*)	91 vs 97 (p = .041^b*)	--	--	91 vs 90 (p = 1.000)	91 vs 56 (p < .001^b*)
LLPv2	1.51	93 vs 96 (p = 1.000)	--	93 vs 97 (p = 1.000)	--	--	93 vs 90 (p = .248)	93 vs 56 (p < .001^b*)
LLPv2	2.00	--	90 vs 82 (p = .765)	--	90 vs 91 (p = 1.000)	90 vs 93 (p = .248)	--	90 vs 56 (p < .001^b*)
CTFPHC		56 vs 96 (p < .001^b*)	56 vs 82 (p < .001^b*)	56 vs 97 (p < .001^b*)	56 vs 91 (p < .001^b*)	56 vs 93 (p < .001^b*)	56 vs 90 (p < .001^b*)	--

^aThe first value in each comparison cell corresponds to the selection criteria of that cell’s row, whereas the second value in each comparison cell corresponds to the selection criteria of that cell’s column.

^bIndicates significance at the P < .05 level.

OncoSim modelling projections show that risk thresholds of ≥ 1.51% and ≥ 2.00% would increase the number of individuals screened compared to choosing patients based on CTFPHC criteria (30 pack-years, Table 3); however, it would also increase the number of cancers detected and number of cancer deaths avoided. The total number of screens and false positives over three years would be similar between the ≥ 2.00% risk threshold and the CTFPHC criteria. Comparing the two risk thresholds, the lower risk threshold ≥ 1.51% would screen 300 per 100 000 more people, detect more cancers (2 more cancer per 100 000 population) and avoid more cancer deaths.

Table 3.

OncoSim projected impact of lung cancer screening in Canada in 2022–2024.

	30 Pack-years	≥ 2.00%	≥ 1.51%
(per 100 000 people)
Number of individuals screened	1300	1500	1700
Number of screens	2200	2400	2900
Number of false positives	220	230	280
Number of cancers detected	24	26	28
Number of cancer deaths avoided	1.3	1.5	1.6

While there was an overlap in the patients selected by all three models, 60.8% and 70.8% for risk thresholds of ≥2.00% and ≥1.51%, respectively, there were 16 patients who would not have been selected for screening by any model at a risk threshold of ≥ 1.51%, with only 1 of these patients being eligible for screening following the CTFPHC guidelines. Patients who were not selected were younger, had a lesser smoking history, had greater body mass index (BMI) and had lower rates of COPD (Table 4).

Table 4.

Differences in patient characteristics between those who were selected by at least one risk prediction model and those who were missed by all three risk prediction models, at a threshold of > 1.51%.

	Selected by at Least One Model (N = 104)	Not Selected by Any Model (N = 16)
Characteristic	no. (%) or mean ± sd.	no. (%) or mean ± sd.	P value
Age	69.1 ± 6.7	59.1 ± 8.1	< .001
Sex (male)	54 (51.9)	7 (43.8)	.543
Estimated education
Less than high school	0 (.0)	0 (0)
High school	55 (52.9)	7 (43.8)
Post high school training	16 (15.4)	2 (12.5)	.509
Some college	9 (8.7)	3 (18.8)
College	11 (10.6)	2 (12.5)
Postgraduate/professional	13 (12.5)	2 (12.5)
Current smoker	69 (66.3)	5 (31.3)	.007^a
BMI	26.1 ± 5.8	29.6 ± 5.1	.025
Duration of smoking (yr)	43.4 ± 11.1	16.3 ± 15.6	< .001
Cigarette packs/day	1.2 ± 0.6	.6 ± 0.7	.001
Quit time (former smokers) (yr)	14.2 ± 11.3 (n = 35)	16.8 ± 14.5 (n = 5)	.649
Medical history
COPD	48 (46.2)	0 (0)	< .001^a
Pneumonia	6 (5.8)	0 (0)	1.000
Emphysema	10 (9.6)	0 (0)	.355
Bronchitis	1 (1.0)	0 (0)	1.000
Tuberculosis	0 (.0)	0 (0)	N/A
Past history of malignancy	7 (6.7)	1 (6.3)	1.000
Exposure to asbestos	2 (1.9)	0 (0)	1.000
Family history of lung cancer	25 (24.0)	3 (18.8)	.761

^aP-value for Pearson’s chi-square test.

Discussion

We compared CTFPHC screening guidelines to three lung cancer risk prediction models and found that all three models were more sensitive in selecting known lung cancer patients for screening, with the ≥ 1.51% risk threshold selecting up to 80% of the study cohort. This risk threshold would increase the number of individuals screened but detect more cancers in our population. The 13.3% of patients who were not selected for screening by any model at the ≥ 1.51% risk threshold had higher BMI, were younger and smoked less.

The choice of a risk threshold is a balancing act between sensitivity and specificity. We recommend a ≥ 1.51% risk threshold, as all three models had approximately 80% sensitivity in our study population. This is concordant with the sensitivity (80.1% [95% CI, 76.8%–83.0%]) reported by Tammemägi et al¹² when applying a ≥ 1.51% threshold to PLCO patient data. As our cohort had already been diagnosed with lung cancer, we could not directly assess specificity of the models. However, Tammemägi et al¹² reported that with a ≥ 1.51% threshold, the PLCO_m2012 model would have achieved greater specificity than the original criteria recommended by the United States Preventive Services Task Force (USPSTF)³⁰ (66.2% [95% CI, 65.7%–66.7%] vs 62.7% [95% CI, 62.2%-63.1%], P < .001), which were similar to the selection criteria of the NLST.

Our modelling analysis using OncoSim demonstrated how the clinical benefits (cancer detected and cancer death avoided), resources required (number of screens) and harms (false positives) vary between CTFPHC guidelines and different risk thresholds of the PLCO_all2014 model. Required resources are similar between the ≥ 2.00% threshold and the CTFPHC criteria. Lowering the risk threshold from ≥ 2.00% to ≥ 1.51% would increase the number of individuals eligible for screening and the clinical benefits of lung cancer screening to detect more cancers, with the hope that earlier detection would increase survival and improve quality of life. However, it would also both increase the number of LDCT scans, thereby increasing costs and waiting times for other imaging, and increase the number of false positives that would increase patient anxiety, repeat imaging and unnecessary invasive procedures. Therefore, a cost–benefit analysis is necessary to determine the ideal risk threshold for lung cancer screening in Canada.

At a threshold of ≥ 1.51%, we did not observe a significant difference in the sensitivity of the PLCO_m2012, PLCO_all2014 and LLPv2 models. Thus, our results do not support the use of one model over another. However, the PLCO_m2012 is already being implemented in some regions within Canada, including the screening program in Ontario³¹ and in a clinical trial in Alberta.³² These practical considerations might motivate the use of the PLCO models in Canadian provinces in preference to other risk models.

Notably, all 48 (40%) patients with a diagnosis of COPD were selected for screening by at least one risk model at a ≥ 1.51% threshold. This finding suggests that a diagnosis of COPD might be a particularly specific indicator of whether someone would be selected for screening by these models. Family doctors should become acquainted with the model’s input criteria, such as a diagnosis of COPD, to help them in referring appropriate patients to official screening centres for evaluation of eligibility for screening.

Our study cohort had a similar proportion of stage I (22.5% vs 21%) and II cancers (5.8% vs 8%), with a higher proportion of stage III (32.5% vs 20%) and lower proportion of stage IV cancers (39.2% vs 49%) compared to the national average.³ Over half of lung cancers in Canada are being diagnosed at late stages, with 3-year survival rates of 5–22% compared to 71% and 49% for stage I and II lung cancers, respectively.³ As the PanCan study showed that the PLCO_m2012 risk prediction model can detect cancers at earlier stages,¹³ implementing screening programs with risk prediction models is critical to increasing patient survival rates.

Thus far, lung cancer screening in Canada has mostly constituted clinical trials and pilots. These initiatives have not yet converged on shared selection criteria, and beyond these academic centres and trials, Canadian physicians might be more likely to follow the better-known, simpler CTFPHC guidelines. The CTFPHC initially published their lung cancer screening guidelines in 2016 based on the findings of the NLST,⁵ as did many other professional associations. These guidelines have evolved independently over time between different healthcare groups, researchers and countries. In 2021, the USPSTF released expanded recommendations for LDCT lung cancer screening for patients between 50–80 years old who are current smokers or quit < 15 years ago and have a 20+ pack-year smoking history.³⁰ Considerable variation is apparent in Canadian approaches to screening. In Ontario’s screening pilot, patients had to be between 55–74 years old, a current or former smoker for ≥ 20 years’ duration and had to have a PLCO_m2012 risk ≥ 2.00%.³¹ The ongoing Alberta Lung Cancer Screening Program seeks to compare the effectiveness of screening between a group of patients who meet NLST criteria and a group of patients with a PLCO_m2012 risk ≥ 1.51%,³² which matches the comparison drawn by Tammemägi et al.¹² Risk prediction models have not been used in participant selection for the ongoing British Columbia Lung Screen Trial, in which eligible participants must be 55–80 years old, be a former or current smoker and have smoked for ≥ 20 years.³³ The same is true of Quebec’s Lung Cancer Screening Demonstration Program, where participants must be 55–74 years old, have smoked for ≥ 20 years duration, and are current smokers or have quit < 15 years ago.³⁴ Evidently, there is not yet a consensus among Canadian provinces regarding the use of risk prediction models in selecting candidates for lung cancer screening, nor which risk threshold to use. This lack of standardization may contribute to patient selection and quality issues.⁸ Our results demonstrate the benefit which can be expected if the risk threshold is lowered from ≥ 2.00% to ≥ 1.51%, all other factors remaining equal.

Limitations of this study include a relatively small sample size, retrospective design and some missing data requiring assumptions. We collected data for patients who had already been diagnosed with lung cancer and thus could not calculate specificity. The data corresponded to the time of diagnosis and so there may be a slight skew of some risk factors such as age and pack-years, though each screening regimen was affected similarly, and so comparative value is retained. The diagnosis of COPD was based on physician records, as opposed to the results of standard pulmonary function tests, which limits the reliability of this important input variable. However, we feel this practice more accurately reflects real-world implementation of screening protocols. In the region of study, ethnic variation was low, and thus, our results have limited applicability for more ethnically diverse Canadian subpopulations and Canada’s indigenous peoples and should be validated in such groups. The PLCO_m2012 and PLCO_m2014 use an American-centric classification system of race/ethnicity, which presents a practical challenge when implementing these models in Canada, though Pasquinelli et al³⁵ have developed a risk model with the ethnicity parameter removed. A formal screening program applied to patients of all backgrounds would hopefully collect this key patient data.

Conclusion

This study demonstrates the effective application of risk prediction models in the selection of patients from a Canadian province for lung cancer screening. A risk threshold of ≥ 1.51% is more sensitive compared to a ≥ 2.00% threshold in identifying lung cancers. The PLCO_m2012, PLCO_all2014 and LLPv2 models were equally sensitive at the ≥ 1.51% risk threshold and were all superior to the CTFPHC guidelines. Patients who are younger, who have limited or no smoking history, or do not have COPD may be missed regardless of the model chosen.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Richard J. Smith

Victoria Linehan

Hensley H Mariathas

References

Statistics Canada and the Public Health Agency of Canada . Canadian Cancer Statistics 2021 Report. Canada: Cancer.Ca; 2021. https://cancer.ca/en/research/cancer-statistics/canadian-cancer-statistics. Accessed 3 December, 2021.

Bryan

Masoud

Weir

, et al. Cancer in Canada: Stage at diagnosis. Health Rep. 2018; 29(12): 21-25.

Statistics Canada and the Public Health Agency of Canada . Canadian Cancer Statistics: A 2020 Special Report on Lung Cancer. Canada: Cancer.Ca; 2020. https://cancer.ca/en/research/cancer-statistics/canadian-cancer-statistics. Accessed 3 December, 2021.

Field

Vulkan

Davies

MPA

, et al. Lung cancer mortality reduction by LDCT screening: UKLS randomised trial results and international meta-analysis. Lancet Reg Health – Eur 2021; 10: 100179.

Canadian Task Force on Preventive Health Care . Recommendations on screening for lung cancer. CMAJ. 2016; 188(6): 425-432.

Aberle

Adams

Berg

, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med. 2011; 365(5): 395-409.

Cancer Care Ontario. Ontario Lung Screening Program, https://www.cancercareontario.ca/en/cancer-care-ontario/programs/screening-programs/ontario-lung-screening-program. (2021, Accessed 7 November 2021).

Linehan

Harris

Bhatia

. An audit of opportunistic lung cancer screening in a Canadian province. J Prim Care Community Health. 2021; 12: 215013272110514.

De Groot

Carter

Munden

The epidemiology of lung cancer. Transl Lung Cancer Res. 2018; 7(3): 220-233.

10.

Tammemägi

Katki

Hocking

, et al. Selection criteria for lung-cancer screening. N Engl J Med. 2013; 368(8): 728-736.

11.

Tammemägi

Ruparel

Tremblay

, et al. USPSTF2013 versus PLCOm2012 lung cancer screening eligibility criteria (International Lung Screening Trial): Interim analysis of a prospective cohort study. Lancet Oncol. 2022; 23(1): 138-148.

12.

Tammemägi

Church

Hocking

, et al. Evaluation of the lung cancer risks at which to screen ever- and never-smokers: Screening rules applied to the PLCO and NLST cohorts. PLoS Med. 2014; 11(12): e1001764.

13.

Tammemägi

Schmidt

Martel

, et al. Participant selection for lung cancer screening by risk modelling (the Pan-Canadian Early Detection of Lung Cancer [PanCan] study): A single-arm, prospective study. Lancet Oncol. 2017; 18(11): 1523-1531.

14.

Weber

Yap

Goldsbury

, et al. Identifying high risk individuals for targeted lung cancer screening: Independent validation of the PLCO m2012 risk prediction tool: Targeting individuals for lung cancer screening. Int J Cancer. 2017; 141(2): 242-253.

15.

Crosbie

Balata

Evison

, et al. Implementing lung cancer screening: baseline results from a community-based ‘Lung Health Check’ pilot in deprived areas of Manchester. Thorax. 2019; 74(4): 405-409.

16.

Cassidy

Myles

van Tongeren

, et al. The LLP risk model: an individual risk prediction model for lung cancer. Br J Cancer. 2008; 98(2): 270-276.

17.

NHS England - National Cancer Programme . Targeted Screening for Lung Cancer with Low Radiation Dose Computed Tomography: Standard Protocol Prepared for the Targeted Lung Health Checks Programme. 2019. https://www.england.nhs.uk/wp-content/uploads/2019/02/targeted-lung-health-checks-standard-protocol-v1.pdf Accessed 7 November, 2021.

18.

Lebrett

Balata

Evison

, et al. Analysis of lung cancer risk model (PLCO M2012 and LLP v2 ) performance in a community-based lung cancer screening programme. Thorax. 2020; 75(8): 661-668.

19.

Common

Mariathas

Parsons

, et al. Reducing Wait Time for Lung Cancer Diagnosis and Treatment: Impact of a Multidisciplinary, Centralized Referral Program. Can Assoc Radiol J 2018;69(3):322-327.

20.

Lau

Caverly

Cherng

, et al. Development and validation of a personalized, web-based decision aid for lung cancer screening using mixed methods: A study protocol. JMIR Res Protoc. 2014; 3(4): e78.

21.

Goffin

Flanagan

Miller

, et al. Biennial lung cancer screening in Canada with smoking cessation—outcomes and cost-effectiveness. Lung Cancer. 2016; 101: 98-103.

22.

Evans

Gauvreau

Flanagan

, et al. Clinical impact and cost-effectiveness of integrating smoking cessation into lung cancer screening: a microsimulation model. CMAJ Open. 2020; 8(3): E585-E592.

23.

Statistics Canada . Canadian Cancer Registry. 2005. https://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&Id=31025 Accessed 5 March, 2022.

24.

Statistics Canada and Health and Welfare Canada . The Health of Canadians: Report of the Canada Health Survey. 1981. https://publications.gc.ca/collections/collection_2016/statcan/CS82-538-1981-eng.pdf Accessed 5 March, 2022.

25.

Statistics Canada . National Population Health Survey1994/1995 to 2010/2011. 2011. https://www12.statcan.gc.ca/census-recensement/2011/ref/92-135/surveys-enquetes/nationalhealth-nationalesante-eng.cfm Accessed 5 March, 2022.

26.

Statistics Canada . Canadian Community Health Survey 2001 to 2017. 2017. https://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3226 Accessed 5 March, 2022.

27.

Statistics Canada . Table 303-0062 - Production, Sales and Inventories of Tobacco Products, Monthly (Kilograms unless Otherwise Noted). 2014. https://www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1610004401 Accessed 5 March, 2022).

28.

Forey

Hamling

, et al. International Smoking Statistics Web Edition Canada. 2009. http://www.pnlee.co.uk/iss.htm Accessed 5 March, 2022.

29.

Statistics Canada . Canadian Community Health Survey. 2009. https://www23.statcan.gc.ca/imdb/p3Instr.pl?Function=assembleInstr&a=1&&lang=en&Item_Id=76867 Accessed 5 March, 2022).

30.

Krist

Davidson

Mangione

, et al. Screening for lung cancer: US Preventive Services Task Force recommendation statement. JAMA. 2021; 325(10): 962-970.

31.

Darling

Tammemägi

Schmidt

, et al. Organized lung cancer screening pilot: informing a province-wide program in Ontario, Canada. Ann Thorac Surg. 2021; 111: 1805-1811.

32.

Taghizadeh

Taylor

MacEachern

, et al. Tobacco use and motivation to stop smoking among long-term smokers who are ineligible for lung cancer screening. Lung Cancer. 2017; 111: 101-107.

33.

BC Cancer . The BC Lung Screen Trial. 2021. http://www.bccancer.bc.ca/our-research/participate/lung-health Accessed 31 October, 2021.

34.

Gouvernement du Québec . Lung Cancer Screening Demonstration Project. 2021. https://www.quebec.ca/en/health/advice-and-prevention/screening-and-carrier-testing-offer/lung-cancer-screening-demonstration-project Accessed 31 October, 2021.

35.

Pasquinelli

Tammemägi

Kovitz

, et al. Risk prediction model versus united states preventive services task force lung cancer screening eligibility criteria: Reducing race disparities. J Thorac Oncol. 2020; 15(11): 1738-1747.

Efficacy of Risk Prediction Models and Thresholds to Select Patients for Lung Cancer Screening

Abstract

Keywords

Introduction

Methods

Results

Discussion

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

ORCID iDs

References