The study shows an important prevalence of newly diagnosed SPN 2.1 % (95 % CI 1.9 – 2.3) for chest radiograph and 17.0 % (95 % CI 15.5 – 18.5) for CT in the imaging tests of the 25,529 patients included in the study. For both types of imaging tests, this prevalence varied according to age, reason for test, and clinical department. Moreover, in patients undergoing a chest radiograph, the prevalence varied according to sex and setting. Patients older than 50 years or those who were undergoing chest imaging in the course of a preoperative examination for extrapulmonary neoplasm were more likely to show an SPN. Men and outpatients having a chest radiograph were also more likely to show an SPN. In relation to SPN characteristics, nodules detected by radiograph were larger than 8 mm, as opposed to nodules detected by CT. However, for both types of tests, detected nodules were mainly located in the upper lobes and with irregular borders. In those patients who had an SPN detected by chest radiograph, irregular border was associated with smoking (current and former). In patients who had a CT, a size exceeding 8 mm was associated with smoking. SPNs appeared to be associated with age ≥ 60 years, diagnosis of a respiratory illness, or male gender. In addition, an irregular border was also associated with sex (men).
Recent screening studies in high-risk populations using low-dose CT reported a higher SPN prevalence than in our study (73.7 % in the PanCan Study [9] and 27.9 % in the National Lung Screening Trial [8], vs. 17.0 % in our study). However, according to previous screening studies using CT, the prevalence of SPNs ranged from 8 to 51 % [3]. The SPN prevalence in the screening studies using radiograph was also higher than that of our study (6.2 % in the National Lung Screening Trial [8] vs. 2.1 % in our study). Previous hospital-based studies using chest radiograph to detect SPN also showed a higher prevalence (23 to 75 %); however, as we previously mentioned, the lack of description of patients’ features or the highly selected population could have affected the results of these studies [6, 13]. The lower SPN prevalence shown in our study could be explained by the unselected nature of our clinical population by including patients having chest imaging from all clinical departments including the accident and emergency department.
For both, patients having a chest radiograph and those having a CT, the prevalence of SPNs was higher in those patients with a diagnosis of neoplasm (4.3 % for chest radiograph and 18.8 % for CT) or those undergoing a preoperative examination (3.2 % for chest radiograph and 53.8 % for CT) compared to patients with other diagnoses. According to a previous study [19], 75 % of patients with extrapulmonary carcinoma or sarcoma who had a CT showed a SPN, whereas our study showed a lower percentage, probably due to the inclusion of an unselected population in our study.
Most of the detected SPNs had characteristics that have been associated with malignancy according to previous quantitative models [7]. A systematic literature review carried out in 2007 by the ACCP [3] described that irregular, spiculated, and lobulated nodule borders were more predictive of malignancy than smooth edges. Nodule size (SPN bigger than 8 mm, according to Fleischner recommendations [17],) and nodule location (those SPN located in upper lobe are more likely to be malignant [20]) are also related with malignancy. In our study, for patients who had both a chest radiograph and a CT, and taking into account only those patients with a nodule characterization, less than half of them showed smooth edges. Of the SPN detected by chest radiology, 59.7 % were larger than 8 mm. This percentage is lower than the data shown in the National Lung Screening Study, where more than 70 % of the SPN detected by chest radiograph were larger than 8 mm. Regarding SPN observed by CT, 44.9 % were larger than 8 mm, which is similar to the 43 % found in the National Lung Screening Study. Thus, according to nodule size, the pretest probability of malignancy could be lower in our study than in screening studies.
Both ground-glass and partially solid nodules are considered subsolid nodules in contrast with solid nodules [17]. Regarding the malignancy of nodules according to the different types of opacity, different opinions exist in the literature; previously some studies [21] found ground-glass opacities to be associated with malignancy, but other studies did not include these characteristics in their prediction models [9]. However, when partially solid and ground-glass nodules were pooled [22], the prevalence of malignancy was higher than for solid nodules. In any case, according to Fleischner recommendations [17], these types of nodules should be managed conservatively. In our study, the percentage of ground-glass nodules was 6.1 % and 3.4 % for partially solid nodules. Although the frequency of ground-glass nodules is lower than those shown, for example, in the Pan Can Study [9] (15.8 %), the frequency of subsolid nodules is higher than those shown in other screening studies, such as the Nelson study [23]. Thus, the likelihood of malignancy in our population could be high.
Although the prevalence of SPN observed in our study in much lower than in screening studies, the proportion of cases with irregular nodules (both for patients with a chest radiograph and those with a CT), is similar to that observed in screening studies [3]. In our study, taking into account only those patients with a nodule characterization, 36.8 % of the patients with a CT showed smooth nodules, in comparison with previous studies using CT, where the prevalence of smooth nodules varied from 20 to 44 %. However, we were unable to obtain data related to SPN morphology in some cases, making it difficult to compare the results.
Some patient characteristics such as sex, advanced age, or history of smoking have been associated with a higher risk of malignant SPN [24]. In our study, men, advanced age, respiratory disease, and current or former smoking were associated with SPN characteristics related to malignancy, such as nodule diameter or irregular border. In our study, in patients who had a chest radiograph, smokers were more likely to show nodules with irregular edges than non-smokers. Swensen [4] reported an increased likelihood of malignancy with the higher number of cigarettes per day. Although we did not quantify the smoking habit, it seems to be associated with SPN characteristics related with malignancy, as would be expected. In patients having a CT, men were more likely to show nodules bigger than 8 mm and nodules with irregular edges than women. Moreover, patients ≥60 years old were more likely to show nodules bigger than 8 mm. According to previous studies [7, 25], the patient’s age is one of the most important clinical factors associated with malignancy. In fact, Swensen [4] described a 2.1 likelihood ratio of malignancy in patients 60 to 69 years of age, and a 5.7 in patients 70 to 83 years of age. In multivariable analysis, the patient’s previous malignancy was not associated with any SPN variables related with malignancy. In previous studies, the relationship between lung cancer and previous malignancy is not clear [6]. However, some morphologic variables such as spiculation or irregular borders may dilute the association in multivariable analysis given that metastases are usually well defined and have no spiculation.
This study presents new data about SPN detected in a clinic-based population for both chest radiograph and CT, but some limitations should be addressed. The observer variability in the determination of the presence of nodules and their characteristics was less than perfect and should be acknowledged as one limitation of the study. This limitation in imaging studies has been shown in other studies [26]. All the radiologists who participated in the study followed the same criteria for the detection and characterization of nodules. Moreover, the evaluation of observer agreement, its causes and the determination of a consensus with the participation of two experts helped to limit the impact of this problem in the overall study findings.
The lack of electronic medical records meant that we were unable to retrieve complete information for a relatively high proportion of cases. Incomplete information on some patients’ data, such as smoking habit, occupational exposure, or family history, could lead to information bias. However, there were no significant differences with respect to other patient’s characteristics (such as age, sex, diagnostic test, reason for test, or nodule characteristics) between patients with the available data and those without.
Despite the similar populations, and the use by radiologists of similar protocols to detect SPN, we observed a different SPN prevalence between the two hospitals included in the study. The reason for this difference could be due to several factors; in addition to the usual inter-reader variation observed when ordering a diagnostic test [27], there were some differences related to the type of clinical departments included in each hospital. Only the hospital that showed the lowest prevalence of SPN included the 4,830 imaging tests carried out in the emergency department (the prevalence of SPN in this department is 1.1 %). Moreover, 86 % of the imaging tests ordered from primary care were included in the hospital that showed the highest prevalence (the prevalence of SPN in the primary care department was 4.6 %).
As opposed to previous studies [28], we did not exclude patients with a malignancy outside the chest because we wanted to show the whole spectrum of patients with an SPN in a clinic-based population. This study differs in many ways from previous evaluations of SPNs. Our study is based on information from consecutive radiology reports, representative of the general clinic-based population in two health districts and it is not biased by patient selection. It should be underlined that in order to establish the predictive value of these, SPN and patient characteristics for lung cancer will be ascertained once follow-up of the cohort, now under investigation, will be accomplished.
In conclusion, we have shown that SPNs may be observed in one of every six patients undergoing CT and in one of every 47 patients undergoing a chest radiograph in a routine clinical setting. This was lower than that shown in screening studies using a highly selected population. Some significant differences related to SPN characteristics have also been shown between our population and those included in screening studies, such as nodule size (smaller in our study than in screening studies). Moreover, some patient characteristics such as age, sex, respiratory disease, or smoking habit were associated with nodule characteristics that are known to be associated to malignancy (bigger nodule size or irregular border).
Given the difference in nodule characteristics and variables associated with the nodule prevalence for patients who have had a CT or a chest radiograph, it is important to study these techniques separately in the general population. Once long-term risk of cancer is determined, these results will be applicable to estimate the clinical ‘pretest’ probability of malignancy of a SPN detected in a clinic-based population for both chest radiograph and CT.