Lung cancer screening for never smokers: current evidence and future directions

Abstract

Screening for lung cancer using low-dose computed tomography is an established means for early lung cancer detection in smokers, but the role of screening for never smokers is unclear. In the 13 lung cancer screening studies involving unselected never smokers, detection rates ranged from 0.1% to 1.1% and positive predictive values ranged from 0.4% to 4.5%. In three lung cancer screening studies involving selected never smokers, selection was based primarily on occupational asbestos exposure, environmental radon exposure and family history of lung cancer in first-degree relatives. Detection rates ranged from 0.3% to 2.6%, and positive predictive values ranged from 0.7% to 15%. Also, 80%–100% of lung cancer cases discovered by screening were early stage, with limited data suggesting survival benefit. Lung cancer screening for never smokers with selected risk factors may achieve detection rates similar to those of screening for high-risk ever smokers. However, further research on optimal subject selection, alternative screening methods and clinical/economic outcomes is needed.

INTRODUCTION

Lung cancer was the second most diagnosed cancer worldwide in 2020, constituting 2.2 million (11.4%) of 19.3 million new cancer cases.[1] It was also the leading cause of cancer death, with 1.8 million deaths in the same period. Given that most patients are diagnosed with advanced cancer,[2] which limits curative treatment, an opportunity exists to improve lung cancer survival via early lung cancer detection. Compared to the treatment of advanced lung cancer, treatment of early-stage lung cancer is both more effective and less costly and it results in excellent survival.[3]

Screening for lung cancer using low-dose computed tomography of the chest (LDCT) has been recommended for high-risk individuals based on age and smoking history.[4] Using a risk-based approach fulfils several criteria for screening, as described by Wilson and Jungner in 1968:[5] lung cancer is an important health concern; the natural history of lung cancer is well understood; early-stage lung cancer is often asymptomatic; a method for screening is widely available; and effective treatment for early-stage lung cancer exists. Low-dose computed tomography of the chest can diagnose lung cancer early and has reasonable detection rates among current and former heavy smokers globally: 2.4% in North America,[6,7] 2.6% in Europe[7,8] and 3.7% in Asia.[9] Survival benefit has also been demonstrated in North American and European LDCT lung cancer screening trials, with relative lung cancer-related mortality reduction of 20%–39% among current and former heavy smokers.[6,8,10] Estimates from these trials indicate that about 300 screens were required to prevent one lung cancer death.[11]

Unlike lung cancer in smokers, lung cancer in never smokers is almost exclusively non-small-cell lung cancer and, rarely, small-cell lung cancer.[12] Nonetheless, like lung cancer in smokers, lung cancer in never smokers can be just as aggressive (according to positron emission tomography maximal standardised uptake values and tumour volume doubling time)[13] and may share similar prognosis upon cancer diagnosis.[14] However, in contrast to lung cancer screening for smokers, the yield and benefits of lung cancer screening for never smokers are less certain.[15] This review, therefore, aims to discuss the current status of lung cancer screening for never smokers and uncover the knowledge gaps for further research.

METHODS

A search of PubMed^® (pubmed.ncbi.nlm.nih.gov) from inception to 23 October 2022, updated on 4 June 2023, was performed using the search string: lung and (cancer or carcinoma) and (screen or screening) and (never-smoker or never-smokers or “never smoker” or “never smokers” or “never smoked” or non-smoker or non-smokers). This was done to supplement the author’s personal library of articles. Articles relevant to the considerations covered in this narrative review were included.

PROS AND CONS OF LUNG CANCER SCREENING

The benefits of lung cancer screening need to be balanced against the downsides associated with lung cancer screening. Factors supporting lung cancer screening include a high mortality rate, the possibility of detecting early stage lung cancer with LDCT, and avoidance of the high mortality and morbidity associated with advanced disease when treatment is started at the early stages.

Conversely, factors against lung cancer screening include radiation exposure,[16] false-positive test results (as high as 96% in the National Lung Screening Trial)[6] and complications (as frequent as 16.6% in real-world settings) from invasive biopsy procedures (e.g., bronchoscopic, percutaneous and surgical biopsies).[17] Overdiagnosis and overtreatment of lung cancer that would otherwise not affect future health status and survival may occur, particularly if indolent cancers are being detected without any stage-shift from late-stage to early-stage disease.[18] This leads to a high burden of evaluating these findings, given the need for follow-up, repeated imaging and invasive investigations. Moreover, patients who have abnormal but indeterminate findings on imaging may suffer from worry and distress. As LDCT often covers parts of the neck, heart and upper abdomen, non-lung-related incidental findings (e.g., thyroid nodules, coronary calcification, liver cysts, kidney cysts) may trigger additional testing, increasing healthcare utilisation.[19]

LUNG CANCER AND ASSOCIATED RISK FACTORS IN NEVER SMOKERS

Lung cancer is not just a smoker’s disease, and screening only smokers for lung cancer would miss nearly half of all lung cancer patients.[14] Among never smokers, defined as persons each of whom has smoked less than 100 cigarettes in their lifetime,[20] the lung cancer burden is increasing in absolute and relative terms. The number of lung cancer cases among never smokers is higher due to increasing incidence; the relative proportion among all lung cancer cases is higher as tobacco smoking rates drop and as non-smoking-related risk factors become relatively more prominent [Table 1].[2,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41]

Table 1.

Selected non-smoking-related risk factors for lung cancer.

Non-smoking-related risk factor	Magnitude of risk	[Ref]
Air pollution (PM2.5)	HR 1.43, 95% CI 1.11–1.84 for each 10 mcg/m3 increment in PM2.5	[21]
Air pollution (PM10)	HR 1.14, 95% CI 1.13–1.15 for each 10 mcg/m3 increment in PM10 for male never smokers; HR 1.06, 95% CI 1.05–1.07 for female never smokers	[22]
Asian ethnicity	HR 2.83, 95% CI 1.64–4.89 in a prospective cohort; OR 3.78, 95% CI 1.19–12.05 in a case-control study	[23]
BMI (kg/m2)	HR 0.95, 95% CI 0.93–0.98 Note: Inverse relationship between BMI and lung cancer risk	[24]
BMI changes (kg/m2) in male never smokers	Major loss (<−1.0 kg/m2/year): HR 1.97, 95% CI 1.12–3.45 Gain (≥1.0 kg/m2/year): HR 2.15, 95% CI 1.15–4.0 Above compared to stable BMI−0.1 to <0.1 kg/m2/year	[25]
Consumption of internal organs of animals	OR 1.85, 95% CI 1.06–3.22	[26]
Emphysema	OR 6.3, 95% CI 2.4–16.9	[27]
Male sex (vs. female sex)	HR 1.49, 95% CI 1.30–1.72	[24]
Gastro-oesophageal reflux disease, presumably linked to chronic micro-aspiration and lung inflammation	OR 1.86, 95% CI 1.26–2.73	[28]
Genetics (MDM2 SNP309 TT genotype)	OR 2.1, 95% CI 1.01–4.36	[29]
History of chronic respiratory disease (pulmonary tuberculosis, chronic bronchitis, emphysema, asthmatic bronchiectasis, silicosis, pneumoconiosis)	HR 1.34, 95% CI 1.11–1.63	[24]
History of tuberculosis among patients with COPD	HR 1.24, 95% CI 1.03–1.50	[30]
History of lung cancer in first-degree relatives	HR 1.42, 95% CI 1.13–1.80	[24]
	HR 1.8, 95% CI 1.0–3.2	[31]
	HR 2.48, 95% CI 1.27–4.84	[32]
Metabolic syndrome	OR 1.11, 95% CI 1.05–1.16	[33]
Non-cystic fibrosis bronchiectasis	HR 1.28, 95% CI 1.17–1.41	[34]
Occupational dust exposure	OR 2.47, 95% CI 1.21–5.03	[26]
Poor ventilation at workplace	OR 4.02, 95% CI 1.74–9.29	[26]
Radon exposure	OR 2.06, 95% CI 1.61–2.64	[35]
	OR 2.19, 95% CI 1.44–3.33	[36]
Second-hand smoke exposure	HR 2.04, 95% CI 1.31–3.16	[37]
	RR 1.28, 95% CI 1.10–1.48	[38]
Seropositive rheumatoid arthritis	HR 1.77, 95% CI 1.06–2.97	[39]

BMI: body mass index, CI: confidence interval, COPD: chronic obstructive pulmonary disease, HR: hazard ratio, OR: odds ratio, RR: relative risk, PM: particulate matter

Of all lung cancer cases, 10%–40% now occur among never smokers, particularly in Asian populations.[24] Among never smokers with lung cancer, about two-thirds have identifiable risk factors (e.g., passive smoking, air pollution, radon, occupational carcinogens, cooking/heating fumes, respiratory infection, ionising radiation, hereditary genetic mutations) and one-third do not (apart from age).[42] Although each of these non-smoking-related risk factors is less strongly associated with lung cancer than heavy smoking (≥40 pack-years) (hazard ratio 17.89, 95% confidence interval [CI] 15.31–20.91, compared to never smokers),[43] a combination of these non-smoking-related risk factors may still increase an individual’s cumulative risk substantially.

FINDINGS OF LUNG CANCER SCREENING STUDIES IN NEVER SMOKERS

In the 13 lung cancer screening studies involving unselected never smokers [Table 2], detection rates ranged from 0.1% to 1.1% and positive predictive values (PPV) ranged from 0.4% to 4.5%.[24,44,45,46,47,48,49,50,51,52,53,54,55,56] Given that the definition of a positive result and the distribution of risk factors vary by study, comparing PPV across studies is not possible. Apart from two studies arising from the same randomised controlled trial (Prostate, Lung, Colorectal, and Ovarian [PLCO] Cancer Screening Trial) done in the USA, all other studies were observational ones performed in East Asia (China, Japan, South Korea). Population characteristics were similar, with mean ages ranging from 40 to 75 years, and females usually comprising >50% of the screened cohort. LDCT was the predominant screening method used, with the PLCO trial being the sole exception (chest X-rays were used for screening). In a subset of 5,483 never smokers in the Cancer Screening Program in Urban China (CanSPUC), LDCT screening failed to show mortality reduction among screened versus non-screened subjects.[57] The authors postulated that the lack of mortality reduction was due to inadequate sample size, limited follow-up time and inaccurate risk stratification strategies.

Table 2.

Lung cancer screening studies for unselected never smokers.

Author (Year)[Ref]	Study type	Population characteristics	Screening method	Screening outcomes
Guo (2022)[44], Wang (2023)[57]	Obs	214,764 never smokers, mean age 55.19 years, 70.70% female, in China	One-off LDCT done under CanSPUC from October 2013 to October 2019	Over 6 years of study, 344 lung cancer cases detected (0.2% of screened subjects) Subset of 5,483 never smokers studied failed to show mortality reduction among screened versus non-screened subjects
Hamaguchi (2022)[45]	Obs	13,159 never smokers, median age~61 years, ~42% female, in Japan	Chest CT screening by a mobile LDCT screening unit in the 10-year period from April 2009 to March 2019	Detection rate of 0.3% (41 cases), 92.7% in stages I–II
Hocking (2010)[46], Oken (2005)[47]	RCT	34,936 never smokers, 60.8% female, in the USA	Part of the PLCO randomised trial Subjects received a baseline CXR, followed by three annual single-view CRXs CXRs were defined as positive when a radiologist identified a mass, nodule, infiltrate or any other abnormalities considered ‘suspicious’ for cancer	Detection rate 0.1% (24 cases) among 6,574 positive screens (PPV 0.4%). Annual screening with CXR did not reduce lung cancer mortality compared to usual care[58]
Kakinuma (2020)[48]	Obs	6,021 never smokers, mean age 57.9 years, 64.4% female, in Japan	LDCT screening for lung cancer between February 2004 and November 2011	Detection rate of 1.1% (66 cases) among 2,515 positive nodules (non-calcified, ≥5 mm) (PPV 2.6%)
Kang (2019)[49]	Obs	12,176 never smokers, median age 50–59 years, 63.6% female, in South Korea	LDCT screening between May 2003 and June 2016 A positive nodule was defined as any non-calcified nodule >3 mm in any diameter	Detection rate of 0.5% (55 cases) among 1,218 nodules (PPV 4.5%) Of the 55 cases, 51 (92.7%) were in stage 0 (carcinoma in situ) or stage I, and all patients had adenocarcinomas
Kim (2020)[50]	Obs	17,968 never smokers, mean age 50.5 years, 68.6% female, in South Korea	LDCT screening for lung cancer between January 2009 and December 2018 at a tertiary centre	Detection rate 0.5% (84 cases) among 2,908 positive nodules (PPV 2.9%); 75 (89.3%) of 84 cases were in stage I
Kondo (2011)[51]	Obs	218 lung cancer patients, mean age 65.1 years, 78% female, in Japan	Annual LDCTv Compared to 160 patients who underwent CXR	Bronchioloalveolar carcinoma 65.1%; Stage IA 88.5%, Stage IB 6.0%, Stage II 1.8%, Stage IIIA 1.4%, Stage IIIB 0.5%, Stage IV 1.8%; 5-year survival rate 95.0% for patients who underwent LDCT, compared to 73.0% for patients who underwent CXR
Li (2003)[52]	Obs	4,251 never smokers, mean age 64.2 years, in Japan	LDCT at least once in a 3-year period from May 1996 to March 1999	Detection rate 0.6% (24 cases), 22 (92%) of 24 cases were in Stage IA
Nawa (2019)[53]	Obs	9,751 never smokers, aged 50–74 years, in Japan	LDCT, interval not fixed Subjects had an average of two or three CT screens over an average of 9.85 years	In never smokers only, using LDCT versus no screening, hazard ratio for lung cancer incidence was 1.79, hazard ratio for lung cancer mortality was 0.41 and hazard ratio for all-cause mortality was 0.57
Shan (2021)[54]	Obs	4,102 never smokers, median age 63 years, 46.6% female, 96 (2.3%) with positive family history of cancer, in China	Population-based screening programme using LDCT from 1 June 2014 to 31 May 2017	Detection rate 0.5% (20 cases), all at stages I–II, among 2,172 nodules (PPV 0.9%)
Shao (2022)[55]	Obs	9,784 never smokers, aged>40 years, ~2/3 female, in China	One-off LDCT using mobile CT with a deep learning system to detect high-risk lung nodules	At 1-year follow-up, 68 lung cancer cases detected (0.7% of screened subjects) among>7,000 nodules (PPV 1%), >80% at Stage I, AUC for lung cancer 0.636–0.647 in validation and testing sets, respectively
Sone (1998)[56]	Obs	3,014 never smokers, aged 40–74 years, in Japan	LDCT using mobile CT unit in a van	Detection rate 0.5% (14 cases)
Wang (2022)[24]	Obs	1,081,603 never smokers, aged 40–75 years, 70.3% female, in China	One-off LDCT	2,269 lung cancer cases detected (0.2% of screened subjects)

AUC: area under the receiver operating characteristic curve, CanSPUC: Cancer Screening Program in Urban China, CI: confidence interval, CXR: chest X-ray, LDCT: low-dose computed tomography of the chest, Lung-RADS: Lung Imaging Reporting and Data System (Lung-RADS 3–4 meant at least one non-calcified nodule 6 mm or larger, or a pure ground-glass nodule 20 mm or larger), Obs: observational study, PLCO: Prostate, Lung, Colorectal, and Ovarian, PPV: positive predictive value, RCT: randomised controlled trial

In three lung cancer screening studies (all observational) involving selected never smokers [Table 3], selection was based primarily on the following risk factors: occupational asbestos exposure, environmental radon exposure and family history of lung cancer in first-degree relatives.[59,60,61] Detection rates ranged from 0.3% to 2.6%, though these results cannot be directly compared to those from randomised trials for high-risk ever smokers due to different study methodologies. Furthermore, the range of PPV from observational studies is very wide (0.7%–15%), and the reasons behind such variability need to be explored. Also, 80%–100% of lung cancer cases discovered by screening were in stages I–II, with limited data suggesting survival benefit.[51]

Table 3.

Lung cancer screening studies for selected never smokers.

Author (Year)[Ref]	Study type	Population characteristics	Screening method	Screening outcomes
Fasola (2007)[59]	Obs	360 never smokers, in Italy	LDCT screening for asbestos-exposed workers and former workers	Detection rate 0.3% (one case of bronchioloalveolar carcinoma) among 148 non-calcified nodules (PPV 0.7%)
Panina (2022)[60]	Obs	2,198 never smokers, aged 40–75 years, in Kazakhstan	One-off LDCT in radon-contaminated regions of Kazakhstan	At 1-year follow-up, 32 lung cancer cases detected (1.5% of screened subjects) among 334 positive (Lung-RADS 3–4) nodules (PPV 9.5%)
Yang (2021)[61]	Obs	12,011 never smokers, mean age 61.2 years, 73.8% female, in Taiwan	One-off LDCT: a solid or part-solid nodule >6 mm or pure ground-glass nodule >5 mm in diameter was considered positive Screened subjects had ≥1 risk factor: family history of lung cancer within third-degree relatives, passive smoking, tuberculosis, chronic obstructive pulmonary disease, ‘cooking index’ ≥110, not using ventilator during cooking	311 lung cancer cases (2.6% of screened subjects) among 2,094 LDCT-positive nodules (PPV 15%); prevalence of lung cancer was 3.3%, 1.6% and 1.7% in those with first-degree, second-degree and third-degree relatives with lung cancer, respectively; 96.5% patients had stage 0 (carcinoma in situ) or stage I lung cancer and 18% of patients had stage 0 lung cancer

LDCT: low-dose computed tomography of the chest, Lung-RADS: Lung Imaging Reporting and Data System, obs: observational study, PPV: positive predictive value

FUTURE DIRECTIONS OF LUNG CANCER SCREENING FOR NEVER SMOKERS

Enhancing subject selection

As mentioned, the studies on unselected never smokers demonstrate widely variable and relatively low detection rates. In contrast, the studies on selected never smokers suggest that if non-smoking-related risk factors are used to select high-risk never smokers for lung cancer screening, it is likely that detection rates could be improved. High detection rates would then translate into a lower number needed to screen to prevent one cancer-related death and lessen healthcare resource utilisation. On a related note, identifying the more impactful risk factors would help to prioritise who really needs to be screened to enhance the value and cost-effectiveness of lung cancer screening.

Risk models that include multiple risk factors have been constructed using data from screening studies involving unselected never smokers. Such separate risk models are required for never smokers, as models such as PLCOm2011 and PLCOm2012 prediction calculators are not suitable for non-smokers. One example is the CanSPUC risk model based on age, male gender, low education attainment, family history of lung cancer, history of tuberculosis and absence of hyperlipidaemia.[44] During validation, the model showed a moderate predictive discrimination for lung cancer risk, with the area under the receiver operating characteristic curve (AUC) being 0.668, 0.678 and 0.685 for 1-, 3- and 5-year lung cancer risk, respectively.

Another example is the development of the China National Cancer Center Lung Cancer model 2021 prediction tool based on age, sex, body mass index, history of lung cancer in first-degree relatives and history of chronic respiratory diseases (pulmonary tuberculosis, chronic bronchitis, emphysema, asthmatic bronchiectasis, silicosis, pneumoconiosis).[24] The AUC for lung cancer was 0.673–0.705 in various cohorts within the study, at a median follow-up of 3.7 years. At the ≥0.47% model risk threshold, lung cancer screening achieved a sensitivity of 24%, specificity of 90%, PPV of 0.6% and a number needed to screen of 784 to avoid one lung cancer death. Note that the number needed to screen of 784 remains substantially higher than the estimated number needed to screen of 300 among ever smokers.[11] This implies that the risk threshold among ever smokers may need to be increased to improve detection rates and lower the number needed to screen.

However, risk thresholds that identify high-risk subjects to increase detection rates tend to reduce the number of eligible participants. An optimal threshold depends on not only the statistics of screening, but also society’s willingness to pay for screening and the healthcare system’s capacity for screening. In health systems at risk of overload, a high LDCT screening burden may ironically aggravate imaging delays already present for patients with confirmed lung cancer.[62] To mitigate screening burden, machine learning methods may be used to analyse clinical parameters and refine candidate selection, such that lower-risk subjects are excluded without affecting the detection of early-stage lung cancer cases.[63] Further research is currently needed to construct and validate risk stratification models, especially in settings outside of Asia [Table 4].

Table 4.

Further research on lung cancer screening for never smokers.

General research area	Specific topics
Subject selection	• Risk-prediction models incorporating non-smoking-related risk factors
	• Identification of risk thresholds to optimise sensitivity and specificity
	• Lung cancer studies involving populations outside of East Asia
Screening method	• Artificial intelligence to enhance screening efficiency
	• Nodule management strategy that maximises detection of lung cancer and minimises the need for invasive diagnostic tests
	• Identification of optimal screening intervals and duration
	• Alternative screening technology, for example, measurement of serum RNAs and RNA fragments, other serum biomarkers, exhaled volatile organic compounds
	• Recruitment strategies incorporating methods to overcome hesitancy and to maximise uptake of lung cancer screening
Screening outcomes	• Establishing survival outcomes, particularly from randomised trials
	• Uncovering lead-time bias (testing increases the perceived survival time without affecting the course of the disease) and length-time bias (testing preferentially detects slower progressing tumours with better prognosis)
	• Determining cost-effectiveness, willingness-to-pay thresholds and numbers needed to screen for various screening strategies

RNA: ribonucleic acid

Improving current screening methods and exploring alternative screening methods

The standard method of lung cancer screening is LDCT, which may be improved in several ways. Deep learning methods using convolutional neural networks can use LDCT images as input data and help human readers identify high-risk nodules. This has the potential to improve the speed, efficiency and accuracy of nodule detection. Although the probability of a detected nodule being high risk for lung cancer does not differ significantly between never smokers and current smokers,[64] more detailed nodule characterisation using artificial intelligence may help differentiate malignant from benign nodules. Better discrimination of malignant and non-malignant aetiologies, in turn, reduces false-positive results, improves the PPV of screening (current data being highly variable, ranging from 0.7% to 15%), lessens the need for invasive biopsy[55] and decreases the requirement for surveillance scans.[65]

Given that lung cancer screening has more false positives due to the presence of tuberculosis sequelae (e.g. lower specificity of 80% in subjects with evidence of prior tuberculosis vs. 85% in subjects without a prior history of tuberculosis),[66] it is possible that lung cancer screening may have better specificity in countries where pulmonary tuberculosis is uncommon. Higher specificity, in turn, translates to a better benefit-to-cost ratio for screening and follow-up management. Nonetheless, efforts to mitigate tuberculosis-related false positives include categorising certain nodules (e.g. non-calcified nodules with adjacent fibrotic change and calcified nodules) as a separate category indicating benign-appearing nodules.[66,67] Addition of epidemiological data to the imaging features of nodules has also been attempted to refine nodule classification and reduce the false-positive rate.[68]

The optimal lung nodule follow-up strategy for never smokers has not been well defined. Ideally, the frequency and duration of lung cancer screening should be minimised, without missing incident lung cancer cases. Among female never smokers without known risk factors for lung cancer, it seems unnecessary to repeat annual LDCT screening for at least 5 years or even longer, unless the initial LDCT showed Lung Imaging Reporting and Data System (Lung-RADS) category 4 findings.[69] Whether this screening interval is optimal for higher-risk never smokers requires further elucidation.

While radiation exposure from LDCT is not prohibitive, it is always preferable to limit the cumulative radiation exposure of individual subjects. Besides increasing the screening interval, another way to avoid radiation is to develop non-CT-based methods of lung cancer screening [Table 4]. Serum ribonucleic acids (RNAs) and RNA fragments have been demonstrated to predict lung cancer up to a decade before diagnosis in smokers,[70] and similar studies should be done in never smokers. Serum carcinoembryonic antigen (CEA) and cytokeratin 19 fragment (Cyfra 21-1) are other promising biomarkers for lung cancer detection, as these are significantly elevated in lung cancer patients compared to healthy controls, and among never smokers living in residential areas with radon levels.[71] In addition, exhaled volatile organic compounds were able to correctly classify lung cancer patients with an accuracy, sensitivity and specificity of 88.79%, 89.58% and 88.23%, respectively.[72] Measurement of these volatile organic compounds using electronic noses may provide another non-invasive method for lung cancer screening. Such technological advances can then mitigate screening hesitancy attributable to radiation concerns, improve acceptability and accessibility of lung cancer screening, and ultimately uncover more early-stage lung cancer cases.[73]

Obtaining unbiased and long-term screening outcomes

Most of the cancer cases were detected early when lung cancer screening was applied to never smokers. Given that identification of early-stage lung cancer among ever smokers was associated with improved mortality,[74] lung cancer screening among never smokers may yield survival benefit. A relatively small observational study among never smokers demonstrated that the 5-year survival rate was 95.0% for patients who underwent LDCT compared to 73.0% for patients who underwent chest X-ray, which corresponded to an 81% mortality reduction.[51] This remarkable mortality reduction may nonetheless overestimate the true benefit of lung cancer screening, as observational studies are susceptible to both lead-time bias (testing increases the perceived survival time without affecting the course of the disease) and length-time bias (testing preferentially detects slower progressing tumours with better prognosis). Similarly, even though a recent systematic review of 14 observational studies of lung cancer screening involving both smokers and never smokers suggested that the relative risk of lung cancer mortality with LDCT screening remained favourable in never smokers,[75] such a result may be affected by lead-time and length-time bias.

Lead-time bias was suggested by a study using LDCT screening for lung cancer among Taiwanese women, of whom <5% smoke, between 2004 and 2018.[76] After the introduction of LDCT screening, increasing early-stage disease incidence (occurrence of new cases of lung cancer) was not accompanied by a decrease in late-stage disease incidence. This suggests that the substantial change in 5-year survival from 18% to 40% during the study period was biased by increased LDCT detection of indolent early-stage lung cancers, and that the increased proportion of early-stage disease may not represent true stage shift from late-stage to early-stage disease.[77]

Except for the PLCO trial, existing studies are all observational and prone to lead-time and length-time biases. Long-term randomised trials of lung cancer screening for never smokers are thus required to obtain unbiased estimates of survival benefit, similar to those for ever smokers [Table 4]. Using a randomised trial design, lead-time bias is avoided by setting zero time as the time of randomisation, instead of the time of cancer detection.[78] Length-time bias and overdiagnosis bias are also avoided since randomised groups are compared. Unbiased outcome data can then be fed into modelling studies for cost-effectiveness, which inform policymaking for lung cancer screening programmes.[79,80,81]

CONCLUSION

When high-risk never smokers undergo LDCT screening, detection rates for incident lung cancer may be several-fold higher than when unselected never smokers undergo screening. Therefore, current observational studies support further development of risk model-aided lung cancer screening for never smokers. To optimise subject selection, promising risk factors include occupational asbestos exposure, environmental radon exposure and family history of lung cancer in first-degree relatives. Nonetheless, existing observational studies are prone to lead-time and length-time biases, and lung cancer screening for never smokers is not yet ready for widespread implementation. Long-term randomised trials of lung cancer screening for never smokers, either with LDCT or with alternative screening methods, will eventually be required to obtain unbiased estimates of survival benefit.

Conflicts of interest

See KC is a member of the SMJ Editorial Board and was thus not involved in the peer review and publication decisions of this article.

Funding Statement

Nil.