Lung cancer screening: current knowledge, opportunities, and challenges

Lung cancer (LC) is the leading cause of cancer-related morbidity and mortality worldwide [1]. Despite significant therapeutic advances, the 5‑year survival for LC stands at around 20% [2] due to advanced-stage diagnosis in many cases—when curative treatment is no longer feasible [1]. The rationale behind lung cancer screening (LCS) is that early detection of LC reduces mortality, as demonstrated by large randomized LCS trials [3, 4].

The global landscape of LCS is evolving, with many countries progressively implementing national LCS programs. In Europe, this renewed impetus is supported by EU4Health-funded initiatives, such as the Strengthening the Screening of Lung Cancer in Europe (SOLACE) project, designed to foster collaboration at the European level [5]. However, implementing LCS in population-based programs poses several challenges. These include ensuring widespread access for participant recruitment, standardized assessment of nodules detected on low-dose computed tomography (LDCT) scans, and appropriate management of incidental findings (IFs).

This review article aims to summarize current knowledge on LCS and to provide an overview of the opportunities and challenges associated with its implementation in everyday practice.

The use of LDCT-LCS has shown a reduction in LC-related mortality of 20–39% in large LCS cohorts, although only a minor reduction in all-cause mortality has been observed [3, 4]. The limited effect on reducing all-cause mortality may reflect either the limited severity of the detected conditions—insufficient to influence all-cause mortality—or a pre-existing reduction in participants’ life expectancy due to age and/or comorbidities [6]. Another relevant aspect of LCS is the opportunity of adopting a conservative management strategy for indolent lesions, such as subsolid nodules (SSNs; [6]). Indeed, their progression toward invasive disease is generally slow and marked by the appearance/growth of a solid component, which can be safely detected on serial LCS-LDCTs [6, 7]. This approach helps avoid overtreatment and limit unnecessary invasive procedures [7, 8]. Notably, LCS also provides a “teachable moment” to encourage smoking cessation interventions, and together, these strategies can lead to a significant reduction in LC-specific and overall tobacco-related mortality [6, 9, 10].

On the other hand, a major concern with LCS is the occurrence of false positives, whose rates vary widely across different studies. The “Updated Evidence Report and Systematic Review for the US Preventive Services Task Force” reported values ranging from 7.9% to 49.3% for baseline screening and from 0.6% to 28.6% for subsequent rounds [11]. This can lead to unnecessary downstream imaging and/or invasive diagnostic procedures, potentially resulting in complications and psychological distress for patients [6, 12]. Another potential harm of LCS is overdiagnosis (and consequently, overtreatment), which, in this context, refers to the detection (and possible treatment) of an indolent LC in an individual with an already reduced life expectancy [13, 14]. The risk of false positives and overdiagnosis can be mitigated through strict adherence to nodule management protocols and the adoption of standardized diagnostic workups.

Regarding radiation exposure, current LCS trials and simulation studies suggest that it represents a minor concern due to technological advancements. Thus, accurate use of optimized LDCT protocols remains of paramount importance to minimize the risk of radiation-induced malignancies across LCS rounds and additional imaging [15].

Proper selection of participants is a fundamental step for enhancing the effectiveness and sustainability of LCS programs at the population level. Traditionally, eligibility has been based on risk factors such as age and smoking history. However, more comprehensive risk models have been developed incorporating additional factors, such as family history, comorbidities, and occupational/environmental exposure [16‐18]. Recently, the TALENT study in Taiwan expanded eligibility by including never-smokers with a family history of LC and lowering the screening age for women to 45 years, reflecting the significant proportion of LCs in non-smokers, especially women, observed in that population [19].

Due to the stigma associated with smoking, the recruitment of high-risk populations faces several challenges, particularly in reaching vulnerable groups, who are less likely to engage with screening invitations [20]. In this regard, for example, the SOLACE initiative targeted four hard-to-reach populations (namely, ethnic minorities, disadvantaged socioeconomic individuals, those living in remote areas, and participants recruited via occupational medicine) and developed tailored strategies to improve their engagement [5]. These include training members from ethnic minorities or socially deprived groups on the importance of LCS and on how to disseminate this message, as well as using mobile CT-truck units in strategic locations to facilitate access to screening sites [5].

Appropriate nodule management remains the cornerstone of LDCT-LCS. Nodule management guidelines offer clinicians structured recommendations for risk assessment, surveillance, and possibly diagnostic interventions [21]. Notably, a distinction should be made between guidance designed for incidental pulmonary nodules (i.e., Fleischner criteria) and those developed for screening purposes, which include the I‑ELCAP (Early Lung Cancer Action Program) recommendation [22], the Lung-RADS v2022 [23], and the recently proposed ESTI (European Society of Thoracic Imaging) recommendations [21, 24]. Finally, the British Thoracic Society guidelines are intended for both incidental and screen-detected nodules [25].

The premise of the different nodule management categories is to estimate the likelihood that a detected nodule represents LC. Despite some heterogeneity across existing guidelines, LDCT outcomes are generally stratified into negative, indeterminate, or positive categories, primarily based on nodule size thresholds. The volumetric approach, first established in the NELSON trial [26], has subsequently been adopted by the major international guidelines [22‐25]. Growth assessment is also crucial, with volume doubling time (VDT) serving as a measure of growth rate [21]. In addition, morphological descriptors are consistently considered across different guidelines: while suspicious features (e.g., spiculation, pleural indentation) may upstage even small nodules, larger nodules showing benign morphology (e.g., intrapulmonary lymph nodes, intranodal fat) can be downgraded [21].

To summarize, negative results lead to the next routine screening round (1- or 2‑year, according to the program); indeterminate findings require short-term follow-up (with timing specified in the radiology report according to guideline recommendations); and positive findings generally prompt referral to a multidisciplinary team for further evaluation, which may include functional imaging (e.g., FDG PET-CT), tissue sampling, or surgical resection depending on the estimated malignancy risk and patient factors. Among minimally invasive diagnostic options, transthoracic needle biopsy (TTNB) and conventional flexible bronchoscopy remain widely used, while novel methods such as radial endobronchial ultrasound (r-EBUS), virtual bronchoscopy (VB), electromagnetic navigation (EMN), and robot-assisted bronchoscopy (RAB) have further improved bronchoscopic sampling of peripheral pulmonary lesions [27].

Incidental findings are defined as abnormalities detected on LDCTs unrelated to the primary scope of LCS [28, 29]. Their reporting and management are still a matter of debate, and no evidence-based guidelines are currently established for either radiologists or clinicians. Notably, reporting IFs may provide an opportunity for early detection of serious conditions, such as chronic obstructive pulmonary disease (COPD) and cardiovascular disease (CVD), which together with LC represent the so-called Big 3 [28, 29]. Other common relevant findings include interstitial lung abnormalities (ILA) and extrapulmonary malignancies (e.g., mediastinal, liver, pancreatic, kidney; [28]; Fig. 1). However, reporting IFs often entails additional investigations and specialist consultations, with a substantial impact on radiologists’ workload and overall healthcare costs [30]. Moreover, the risk of overcalling and overdiagnosis may also generate psychological distress among LCS participants [28]. Taken together, and considering the associated ethical and legal implications, the reporting and management of IFs remain controversial, underscoring the need for standardized strategies that balance clinical benefits with potential harms and maintain cost-effectiveness.

Bild vergrößern

Moving toward an increasingly personalized approach, the use of artificial intelligence (AI)-based tools and blood biomarkers hold considerable promise for improving the LCS workflow.

In terms of nodule detection and volumetric assessment, computer-aided detection (CAD) systems have been widely adopted for LDCT reporting [31, 32]. Specifically, AI-based automated approaches have already been used as concurrent or second readers to improve efficiency and consistency in LDCT interpretation, while their potential use as primary readers is still under investigation [33, 34]. However, the potential of AI extends well beyond these tasks. Recent studies have demonstrated that deep learning models, which rely solely on imaging data, can outperform multivariable risk models (e.g., the Brock model) in estimating malignancy risk [35, 36].

Blood biomarkers are another active area of investigation in LCS, with several biomarkers currently under study [37]. They may contribute to refining eligibility criteria and support the management of indeterminate pulmonary nodules [38]. In this regard, the MILD trial demonstrated that combining a circulating microRNA (miRNA) with LDCT improved accuracy compared to LDCT alone [39], while the BIOMILD trial showed that the integration of baseline LDCT and blood miRNAs can predict individual LC incidence and mortality [40].

Implementing LCS at the population level represents a unique opportunity to enable early detection of LC, identify other serious conditions, and promote primary prevention. However, several issues still need to be addressed. While standardized approaches are essential, implementation strategies might ultimately be tailored to the specific resources and population characteristics of each context. Results from ongoing LCS programs and trials will provide further evidence to refine these strategies and continue exploring the role of AI in LCS.