Low-dose CT for lung cancer screening: position paper from the Italian college of thoracic radiology

Lung cancer (LC) is the leading cause of cancer-related death in Italy and worldwide, with over 2 millions cases in 2018 [1]. LC is associated with smoking, it is usually diagnosed in advanced stage in variable association with clinical symptoms and has 85% five-year mortality [1]. In Italy, LC is the third most common neoplasm (11% of alla cancers in 2018) while ranking first cause of cancer death (20% of all cancer deaths in 2018) [2].

Smoking cessation is the intervention for primary prevention of LC [3]. It is estimated that 85–90% of LC are associated with cigarette smoking in Italy, where prevalence of smoking people is 23% [2]. Smoking cessation decreases LC risk after 10 years since quitting [4]. Age is the second established risk factor for LC, and environmental and professional exposures represent additional risk factors [5‐10].

Screening with low-dose computed tomography (LDCT) is the main intervention for secondary prevention of LC and decreases the LC mortality by 20–30%, especially in women [11‐17]. A number of Italian trials contributed to the screening literature. In particular, over 8,000 subjects were enrolled in the three randomized controlled trials that took place in Italy since the early 2000^S. The Multicentric Italian Lung Detection (MILD, 4,099 participants) showed a statistically significant 39% reduction of lung cancer mortality after 10 years of screening, whereas the ITALUNG Trial (3,206 participants) and the Detection And screening of early lung cancer with Novel imaging Technology (DANTE, 2,450 participants) provided similar results in shorter screening periods and with lower statistical power. Non-randomized trials, including COSMOS and BioMILD in Milan, recruited about 10,000 participants [18, 19], and the bioMILD study (4,119 participants) prospectively investigated the integration of LDCT and blood biomarkers for optimized prolonged screening interval at 3 years [20]. An European Committee for Health Technology Assessment concluded that “screening for lung cancer with LDCT may have a mortality benefit” [21] and the United Stated Preventive Services Task Force (USPSTF) has recently broaden inclusion criteria for LDCT screening [5]. Although COVID-19 infection slowed cancer screening interventions [22], an Italian study demonstrated that LDCT screening can be safely performed during the COVID-19 pandemic [23]. The Cochrane Database of Systematic Reviews is encouraging pilot studies, notably with a short list of necessary outcome measures for continuous quality assurance in data collection and future meta-analyses [24] (Table 1).

In 2015, LCS by LDCT started in the USA and is reimbursed by Medicare. In Europe, despite availability of several shared guidelines [25‐27], population LCS has not started yet. In Italy, an ongoing discussion is supposed to prepare for inclusion of LCS by LDCT into the governmental healthcare supply (e.g. Livelli Essenziali di Assistenza, LEA), which already includes other population-based cancer screening (e.g. breast, colon-rectum and uterine cervix). These screening interventions are part of the National Prevention Plan, which has purportedly to include high-risk and socially or economically disadvantaged individuals [28].

Several initiatives and studies of LDCT for LCS have been funded and have already started or are next to start in Italy in 2021–2022 (Table 2). They feature some differences in method, thus implying heterogeneity of the proposed screening models, including enrolment criteria and modalities, annual or biennial frequency, strategies to promote smoking cessation, and use of biomarkers. However, taking into account the fundamental role of the radiologist in LCS, we trust sharing fundamentals of radiology practice in LCS is mandatory to warrant quality assurance and allow comparison and/or meta-analyses for informing the next step towards population-based LCS. According to such an objective of harmonization, the Italian College of Thoracic Radiologists discussed and elaborated the present document, which aims to be a practical support for radiologists involved or approaching LCS with LDCT in Italy. The following paragraphs present the current standard of reference for LDCT scan protocol, reading method, classification of findings, reporting and major areas of further management.

The screening LDCT is a simple and fast chest examination, which does not require administration of contrast agent. The technical details about hardware, acquisition, and reconstruction of LDCT for LCS were indicated by an expert panel of chest radiologists from the European Society of Thoracic Imaging (ESTI, link to online resource) [29]. Routine LDCT for LCS is hereafter described.

The acquisition starts with a bidimensional scout scan over the chest and is followed by a volume acquisition from apex to lung bases. Deep inspiratory breathhold is mandatory. Any external object must be removed from the chest to minimize radiation exposure and avoid artefacts that would impair nodule measurements. For the volume acquisition, the tube setting is set at low current (typically below 40 mAs) and 120 kVp (or 140) voltage. It is of paramount importance that the tube voltage (and acquisition and reconstruction parameters) is kept consistent through subsequent LDCT examinations of the same subject to allow reliable evaluation of nodule features, especially for subsolid nodules. The above setting aims to minimize the radiation exposure, while maintaining appropriate image quality for volume segmentation [30]. The reduction of radiation exposure is also available via filtering the X-ray beam by tin filter installed between the tube and the aluminium bowtie filter, his technique was proposed by voltage 100 kVp and current 100 mAs with automatic exposure control [31, 32].

The slice collimation must be thin (≤ 1 mm) to grant optimal data quality for image reconstruction. The number of detector rows of the spiral CT scanner is not per se a limitation for LCS. However, CT scanners with few (e.g. 4–16) rows of detectors are becoming obsolete for the purpose of LCS, since they typically require higher radiation doses owing to lower efficiency of old detector technology. Therefore, although screening LDCT has been performed also with few-row CT scanners, the technological development justifies recommendations for up-to-date (64-row or higher) CT scanners. Software development also assists in dose reduction, namely by controlling the noise: iterative reconstructions or deep learning algorithms outstand the old filtered back projection. Phantom studies indicate that advanced reconstruction algorithms together with careful tube setting allow radiation exposures similar to that of chest X-ray (also known as ultra-low dose CT), while being far more sensitive to lung nodules detection [33].

The radiation dose of a screening chest LDCT varies depending on the biometric features (height and weight) of the subject. The ideal threshold of volume computed tomography (CT) dose index (CTDI_vol) is set below 2.0 mGy by the America College of Radiology. It is noteworthy that the Italian law (number 101 released 31 July 2020) requires to report the radiation exposure associated with every examination on each radiology report. While setting an optimized LDCT protocol for LCS, the radiologist should also bear in mind that the image quality should also allow assessment of pulmonary emphysema and coronary artery calcifications [34], which along with pulmonary nodule represent the so-called BIG-3 [35].

The time required for reading a screening LDCT examination by an experienced radiologist is generally below 10 min, but it can be less than 5 min in case of a negative test [36]. Usually, the LDCT is read by two independent radiologists. This method is similar to that recommended in breast screening and undoubtedly increases the costs of LC (along with the need of dedicated CT scanner spaces and of acquisition and maintenance of software for volumetric assessment of lung nodules size). Not surprisingly, great attention is paid to the possible implementation of Computer Assisted Diagnosis (CAD) systems. In a study in the Netherlands, LDCT reading performed by a single radiologist supported by CAD system replaced double reading [15]. The time required for LCS test reporting varies according to the use of CAD as support to the reader and type of report (free or structured). Adoption of a certified CAD is suggested to help reducing variability in detection rate between readers, while classification might still vary substantially depending on the manual correction often required for nodule segmentation [37]. Overall, the use of CAD for reading LDCT in LCS can be endorsed, but requires an appropriate education on its function, strength, and limits, that have been outlined in specific recommendations by the ESTI [38]. Webinars on the theoretical backgrounds of CAD and practical hands-on workshops by the same expert panel are available to promote CAD use in LCS with LDCT [39]. Nonetheless, the LDCT for LCS should be read by radiologists educated in lung imaging and with specific skills in LCS. For this purpose, since 2019 the ESTI is providing certification courses for theoretical and practical education of radiologists to the practice of LCS [39].

The work-up of suspicious or highly suspicious (Lung-RADS 1.1 categories 4A-4X) nodules with a solid component exceeding 8 mm can be performed with 18F-fluoro-2-deoxy-glucose -positron emission tomography (FDG-PET)/CT, CT-guided fine needle aspiration or core biopsy, and Video Assisted Thoracic Surgery (VATS). The choice of work-up strategy usually reflects local availability and expertise and is usually accompanied by a staging contrast enhanced CT at standard dose.

Differently, follow-up LDCT is the management tool for indeterminate (Lung-RADS 1.1 category 3) nodules at baseline or annual repeat LDCT with the goal to ascertain possible nodule growth. The latter is trusted at ≥ 25% increase of lesion volume [26], or an increase of the mean diameter > 1.5 mm (Fig. 1). Integrated description of growth dynamics is anticipated for the next Lung-RADS 2.0 release [43, 44]. The dynamics of growth in solid lesion can be estimated by serial LDCT and calculation of the Volume Doubling Time (VDT) (Fig. 1). Notably, the VDT can be calculated based on either mean diameter or segmented volume: a VDT ≤ 400 days is associated with a malignant nodules [25] (Fig. 1). So far, the VDT was validated for solid lung nodules, but not for part-solid or non-solid nodules. For the latter two types of nodules, increase of a solid component in mixed nodules or appearance of new solid component in a former non-solid nodule is considered a sign of significant growth and potential malignancy [58] (Fig. 2, 3).

Additional findings in screening LDCT are common, being observed in 4.4 to 40.7% cases. They are more frequent with increasing participants age and can imply further evaluations. The wide variability reflects inconsistent definition of such findings and especially their clinical relevance [59].

Recommendations about reporting of additional findings have evolved [60‐62]. To date, the growing experience and advances in knowledge on LDCT screening suggest a “granular interpretation”. The reference guidelines for such a critical interpretation are provided both by the American College of Radiology [50] and the National Health System England [63]. Findings unrelated to LC and pulmonary nodules can be distinguished in smoking-related and non-smoking-related. We shall label smoking-related findings as “collateral findings” which include calcifications of coronary arteries (CAC), pulmonary emphysema and interstitial pulmonary abnormality/disease. The remainder, namely the wide array of non-smoking-related findings will be interpreted as the true “incidental” findings (Table 4).

The recommended interval for LDCT screening is once a year [5]. However, several studies demonstrated that two-year interval allows efficient surveillance while reducing radiation exposure and costs [85‐87], whereas longer interval is associated with increased number of advanced LC [85]. The possibility of biennial LDCT screening test should be reserved to subjects with negative baseline by LungRADS 1.1, which represent over 70% of LDCT [47].

Screening should start at age 50 and interrupted at 80 [5]. Moreover, LDCT screening is not recommended for former smokers who quit smoking > 15 years or for people not eligible for or not willing to undergo LC surgery.

So far, the selection of subjects at risk of LC has been mainly based on age and smoking history measured by pack-years. However, the yield of these selection criteria seems relatively low compared to multifactorial profiling [19, 88] including pre-test (before LDCT) findings related to personal and family history [53, 89]. Moreover, several multifactorial risk models aim also to post-test risk refinement by inclusion of findings from LDCT [77, 90]. These factors contribute to stratification of LC risk, as well as of CV and respiratory morbidity and mortality that can ultimately hinder possibility of surgical treatment [5].

The above two general considerations underly the quest of LCS personalization [88, 91, 92] both for LDCT interval [75, 93] and LCS duration in the individual lifetime, also considering the theoretical risk from ionizing radiations [94].

As for every screening intervention, chest LDCT has its harms [95, 96]. In this manuscript we will mention harms directly related to the radiological practice. They include false-positive results with downstream unnecessary (or even harmful) investigations and invasive work-up, overdiagnosis, distress and anxiety due to indetermined test results, and radiation-induced cancer [97]. In particular, overdiagnosis is a topic of active debate in LCS. Subsolid nodule is probably the most prominent finding associated with slow and potentially clinically indolent growth [58, 98], whose progrostic weight may ultimately be overcome by competing causes of death in heavy smokers or former smokers [77].

On the other hand, also false-negative findings are to be accounted for in LCS practice and this is prone to different interpretations [99]. However, estimate of false-negative rates in an important metric for quality assurance in LCS. False-negative can derive from different moments of the LCS practice, including reading of the LDCT test and definition of nodule management protocols. Accordingly, in the NELSON trial the majority of false-negatives could be attributed to detection or interpretation error [100], whereas the risk of missed cancers increased significantly if screening interval was prolonged at 2.5 years [85].

Psychological distress related to LDCT findings is a further potential harm, which is directly linked to the radiology report [97]. The Danish Lung Cancer Screening Trial investigated the psychological distress in LCS and found that data so far available might be biased by selection of a more robust population [101]. This is probably due to the more favourable socio-demographic profile of people participating in the early LCS trials. Looking towards LCS implementation in the general population, the radiologist should strain towards improved communication of the findings to the screening participant, and this might include accounting for person-hours dedicated to the verbal communication of the report.

In conclusion, LDCT represents the standard of reference for LCS. The use of LDCT as preferred test in LCS is intended as optimal practice, yet not perfect. The Italian College of Thoracic Radiologists is convinced that the use of quality assurance references is mandatory to make population practice as accurate as LCS trial results. Technology update is mandatory to maintain appropriate quality of LCS practice, while continuous education is warranted to follow the most appropriate evidence, similarly to what was already witnessed in mammography screening. The near future of LDCT for LCS calls for preparedness in technology and medical skills, the next step is eventually foreseen in continuous optimization of resources [102].