Lung cancer (LC) is one of the most common types of cancer and carries the highest mortality burden, and as such is a significant health problem. Overall survival at 5 years is under 16%, and prognosis is directly related to disease stage at time of diagnosis. Primary prevention of smoking is key to reducing the incidence of LC, and the largest number of cases are diagnosed among former smokers. Even with this primary prevention and recent therapeutic advances, early disease diagnosis is essential if mortality is to be reduced.

The International Early Lung Cancer Action Program (I-ELCAP), a non-randomized study, already reported in 2006 that 85% of cancers were diagnosed in stage I, with an estimated 10-year survival of 88%1 (rising to 92% if the LC was resected within 1 month of diagnosis). The National Lung Screening Trial (NLST) is the first randomized screening study in high-risk individuals to show a reduction in LC mortality of 20.3% in the group that underwent screening with low-dose computed tomography (LDCT) compared to chest X-ray, and a 6.7% reduction in overall mortality.2 These results led the US Preventive Services Task Force (USPSTF) in 2014 to recommend screening in high-risk individuals, with a 2B degree of evidence.3 In 2015, the European Society of Radiology (ESR) and the European Respiratory Society (ERS) voiced their support for screening programs.4 The International Association for the Study of Lung Cancer (IASLC) has recently recommended the implementation of screening based on the results of the NELSON study (NEderlands-leuvens Longkanker Screenings ONderzoek) presented at the 19th World Conference of IASLC Lung Cancer in Toronto. This report confirms that screening with LDCT reduces LC mortality by 26% in men and 61% in women.

In Spain, the Clínica Universitaria de Navarra,5 a member of I-ELCAP, has accumulated extensive experience and made important contributions to screening strategies, including evidence for the use of positron emission tomography (PET),6 the value of emphysema and chronic obstructive pulmonary disease (COPD),7 and the creation of risk prediction models.7,8

In line with opinions published by the Spanish Society of Pulmonology and Thoracic Surgery and other societies on the implementation of screening in 2017,9 we intend to demonstrate that LDCT integrated in a structured protocol is useful for the detection of early-stage, potentially curable LC.

We included 4951 subjects in the study (Table 1). Of these, 65.4% were men; mean age was 56.89±5.26 years, and median age 56 years, with a range of 50–79. Mean tobacco consumption was 37.45±23.41 pack-years for men and 29.06±16.73 for women. In the baseline assessment, 2238 (45.8%) were former smokers.

In the baseline assessment, CT was negative in 4401 (88.89%) subjects, and 550 (11.11%) showed 1 or more nodules that were identified as positive. Of 3891 nodules detected, 692 met the criteria of positivity according to the I-ELCAP protocol. The follow-up schedule was determined on the basis of these results. Most cases (80.20%) were referred for a repeat LDCT as the first course of action, and in 89.36% the scan was performed at 3 months. Histological studies were performed in 23 cases (3.32%), PET-CT was performed in 2 (3.03%) and bronchoscopy in 5 (Fig. 1). Biopsy was indicated in 2 cases, despite a negative PET, given the small size of the nodules and the possibility of FP in the PET.

Fig. 1.

Decision flowchart for baseline screening.

(0.21MB).

In the 1-year scan, excluding the 36 subjects diagnosed with pulmonary tumors, LDCT was negative in 4691 (95.44%) participants, and 224 (4.56%) were positive; 4081 nodules were detected, of which 288 were positive. In 79.86% of the positive cases, LDCT follow-up at 1 month was indicated as a first option in 46.08%, in 47.83% at 6 months, and in 44 cases at 1 year. FNAB was indicated in 6 patients, PET-CT in 3, and bronchoscopy in 2. Surgical resection was indicated in only 1 patient (Fig. 2).Forty biopsies were carried out at baseline (32 FNAB, 6 bronchoscopies, 1 video-assisted surgery, and 1 exploratory thoracotomy), and 21 at the 1-year scan (12 FNAB, 7 bronchoscopies, and 2 thoracotomies). In the 44 FNABs performed, 6 patients developed pneumothorax as the only complication, of which only 2 required drainage. Forty (5.8%) of the 692 nodules classified as positive at baseline required invasive procedures. At the 1-year scan, 7.3% (21/288) of nodules were positive. We diagnosed 38 pulmonary tumors at baseline screening, including 36 LCs, and 14 in the 1-year scan, of which 13 were LC. Of these, 5 patients were diagnosed on the basis of newly identified nodules. The prevalence rate was estimated at 0.89%, and incidence at 0.1%.

Fig. 2.

Decision flowchart for 1-year scan.

(0.15MB).

To evaluate the screening program and LDCT as a diagnostic tool, we calculated S, SP, PPV, NPV, PPR, and NPR for individuals and for nodules (Tables 2 and 3).

In the baseline screening, 692 (19.57%) were radiologically positive, and of these, 656 (94.8%) were considered FP, after the exclusion of LC. Taking into account all noncalcified nodules (NCN) (3536), the rate of FP was 18.55%. In the 1-year scan, 96.52% of the nodules classified as positive, and 7.64% of all NCNs, were FPs.

We identified a total of 52 tumors: 48 LC (Table 4), 1 metastatic melanoma (unknown primary), and 3 benign tumors (tuberculoma, histiocytosis X, aspergilloma). A total of 52.09% of the tumors were diagnosed in stage I (Table 5). Forty (76.92%) LCs were treated surgically. The median follow-up was 3.67 years (0.01 to 7.99).

Overall survival of the detected LCs was 58.5% at 5 years, with 67.1% cancer-specific survival (75.8% in the surgical patients and 41% in the non-surgical patients). Five-year survival in stage IA LCs was 89.4% and 80% in stage IB tumors.

This study shows that LDCT is a valid tool for LC screening as part of a comprehensive program for early detection of lung nodules.

Existing studies vary widely in terms of sample selection criteria and the design of protocols for evaluating the nodules. Like us, many research groups included voluntary participants. The NLST13 authors speak of the “healthy volunteer effect”, with individuals who are more committed to both their health and the program.14 It is clear that selecting higher-risk patients and cohorts with greater disease prevalence would increase the yield of screening. For this reason, the NELSON15 investigators selected their population from individuals who responded to a questionnaire; other studies took into account the existence of COPD/emphysema,7 and others used risk prediction models evaluating multiple factors that without doubt improve sensitivity and PPV.16,17

The proportion of nodules considered radiologically positive in our study was 19.57% in the baseline screening and 7.91% in the 1-year scan. Newly identified nodules are less frequent than dominant nodules, but are more likely to progress to LC. Henschke et al.,1 using the same criteria, found a rate of positive nodules of 13%. In the NLST study,18 the rate of positive nodules in the first round of screening was 27.3%. The NELSON study15 estimated a rate of 2.6%, but if we take into account the nodules defined as indeterminate which underwent LDCT follow-up at 3 months, the rate is 19.2%. In our study, 94.8% and 96.52% of the nodules defined as positive in the baseline and 1-year scans, respectively, could be considered FP, although with respect to the total number of NCNs, the FP rate was 18.55% and 7.64%, respectively. The NLST study found an FP rate of 96%.18,19 In the NELSON study, in the first 3 rounds, excluding indeterminate nodules, 200 LC were identified among the 493 positive tests, i.e., 64.3% of positive nodules can be considered FPs. The authors estimated a rate of 3.86%.20 It is clear, then, that the rate of FP depends largely on how a positive nodule is defined. In the opinion of the I-ELCAP and NELSON investigators, a nodule that does not grow should not be regarded as positive.

Most positive results only needed to be followed up with imaging procedures. Invasive procedures were required in 40 (5.8%) of the 692 nodules classified as positive on baseline LDCT, and in 20 (7.3%) of the 288 classified as positive in the 1-year scan. Results were positive in 95% and 66% of biopsies performed, showing that the evaluation protocol can accurately select patients who really require a biopsy. In the 1-year scan, biopsy was indicated when there was growth of existing nodules or appearance of new nodules (the majority of them, 275). Newly identified nodules were less common than dominant nodules, but are more likely to be malignant. This was probably the reason why a greater number of biopsies were negative (or not useful for diagnosis) in the 1-year scan. In any case, iatrogenic complications were rare, and none were severe.

As in other observational series, the prevalence rate is much higher than the incidence rate. In our study, the prevalence rate (0.89%) was close to that reported in the P-ELCAP study (1%),5 in Veronesi et al.21 (1.1%), and in the I-ELCAP study (1.3%).1 We found a low incidence rate (0.1%) compared to other studies (P-ELCAP,5 1.4%), that was closer to that of I-ELCAP (0.3%)1 or the 1-year detection rate of P-ELCAP (0.41%). These differences may be explained by the different inclusion criteria, or because we estimated the incidence rate only from newly occurring nodules.

S and SP of both the evaluation protocol, taking individuals into account, and the LDCT, taking nodules into account, are very high. NPV is a key parameter in any screening program, and in our case it was 99.9%. PVV, defined as the probability of disease when the test is positive, was around 6.55%, in line with other studies.19,22–25 Only Henschke et al.1,26 found a PPV of 11.5% in the initial screening, that increased to 25% in subsequent rounds. The number of FPs determines a low PPV rate. Given our high S and high NPV, a negative result makes it highly unlikely that the individual has cancer.

DP, understood as the percentage of patients who were diagnosed correctly, was 89.56% and 95.65% in the baseline and 1-year scans, respectively. We used likelihood ratios to determine diagnostic power. Our PPR suggests that it is 9–18 (baseline/1-year) times more likely to obtain a positive result in a patient than in a healthy individual. This PPR in the 1-year scan shows that it is a very powerful strategy that strongly supports the diagnosis. We found an NPR of 0.09 in the baseline screening, indicating that the power of screening for ruling out a diagnosis is high. No similar data have been published from other studies.

Prevalence is, in short, the probability of having disease before screening is performed. PPV would be the post-screening probability of having disease, and lies at around 5%-6.5%. We know the prevalence rate of LC in the population (approximately 45/100000 inhabitants). The resulting ratio is 111–144, meaning that we have multiplied our ability to detect patients with disease by 111–144.

Distribution by LC stages is slightly lower than in other studies: in the NLST and in the NELSON studies, 61.6% and 70.8%, respectively, were stages IA-IB.27 In P-ELCAP,5 73% of LCs were stage I. The percentage of early stage disease increases in successive rounds (results not shown).

Surgery was performed in 76.92% of our cases. In the I-ELCAP, 85% of patients underwent resection, with a 10-year survival of 92% when the intervention was performed within 1 month of diagnosis.1 Cancer-specific survival in our study was 67.1%, and 75.8% in the group undergoing surgery. Other European studies found similar data: Bellomi et al.28 found that 72% of their patients with a diagnosis of LC were in stage I. This group had a 5-year survival rate of 89% and an overall survival rate of 63%. Blanchon et al.29 found an 80% survival rate for LC detected in the group undergoing CT screening, and 88% for LC in stage I. These data are promising if we remember the overall survival of LC in the general population.

One of the main limitations of screening is overdiagnosis, which we cannot evaluate in the absence of a control group. Overdiagnosis is often confused with FPs. FP is a diagnostic error, and overdiagnosis is a prognostic error.30

LDCT integrated into a standardized evaluation protocol with a carefully designed and defined algorithm (with specific inclusion criteria determining the duration and intervals of screening) is a valid tool for screening because it helps diagnose LC at earlier stages. It is important to minimize all limitations before it can be implemented as a populational screening tool.

The authors state that they have no conflict of interests.