Authors: Xiaodan Zhu, Changxing Shen, Jingcheng Dong
Categories: Research
In recent years, clinicians often encounter patients with multiple pulmonary nodules in their clinical practices. As most of these ground glass nodules (GGNs) are small in volume and show no spicule sign, it is difficult to use Mayo Clinic Model to make early diagnosis of lung cancer accurately, especially in large numbers of nonsmoking women who have no tumor history. Other clinical models are disadvantaged by a relatively high false-positive or false-negative rate. Therefore, there is an urgent need to establish a new model of predicting malignancy or benignity of pulmonary GGNs for the sake of making accurate and early diagnosis of lung cancer.
Included in this study were GGNs surgically resected from patients who were admitted to Yiwu Central Hospital from January 2018 to March 2024, including both male and female patients, there is no gender specific issue. The nature of all these GGN tissues was confirmed pathologically. The case data were statistically analyzed to establish a mathematical prediction model, the prediction performance of which was verified by the pathological results.
Altogether 261 GGN patients met the inclusion criteria. Using the results of logistic regression analysis, a mathematical prediction equation was established as Malignant probability (mP) = e^x^/ (1 + e^x^); when mP was > 0.5, the GGN was considered as malignant, and when mP was ≤ 0.5, it was considered as benign. x= -2.46 + 1.032gender + 1.85mGGN + 1.40VCS-0.0027mean CT value of the nodule + 0.078*maximum diameter of the nodule, where e represents the natural logarithm; if the patient was a female, gender = 1 (otherwise = 0); if the pulmonary nodule was a mixed GGN, mGGN = 1 (otherwise = 0); if the pulmonary nodule had vascular convergence sign, VCS = 1 (otherwise = 0). The prediction performance of the mathematical prediction model was verified as the negative prediction value was 0.97156 and the positive prediction value was 0.3800 in the model group versus 1 and 0.25 in the verification group.
In this study, we identified female gender, mGGN, VCS, mean CT value and maximum nodule diameter as five key factors for predicting malignancy or benignity of pulmonary nodules, based on which we established a mathematical prediction model. This novel innovation may provide a useful auxiliary tool for predicting malignancy and benignity of pulmonary nodules, especially in women patients.
The online version contains supplementary material available at 10.1186/s12885-024-13004-z.
Lung cancer is a malignant tumor, ranking first in both the incidence and mortality worldwide, fundamentally because of the lack of effective early prediction and interventional methods [1, 2]. The Mayo Clinic Model is known as the classic clinical model for predicting the malignant probability (mP) of pulmonary nodules by mainly using age, smoking history, extra-thoracic surgery history, nodule diameter, the presence or absence of spicule sign, and whether the nodule is located in the upper lung lobe as six major factors to make a comprehensive prediction. However, the main problem that clinicians encounter in their clinical practice is that patient’s pulmonary ground glass nodules (GGNs) are often multiple and small in size. In recent years, large number of young patients with GGNs who had no smoking and/or tumor history were diagnosed as having early lung cancer. In addition, more studies have reported the association between comorbidity-chronic inflammation and carcinogenesis, and influences of natural environments, food safety and psychic stress on carcinogenesis. More cases of GGNs complicated with other chronic tissue or organ diseases in sub-healthy individuals, such as intestinal inflammation, diabetes, insomnia, and constipation have been reported [3–6]. Therefore, we need to re-think influences of these previously ignored comorbidity factors on the incidence of lung cancer and their impact on the prediction value, based on which to establish a new and more clinical applicable model for predicting malignancy or benignity of pulmonary nodules for the sake of improving the accuracy of clinical diagnosis of lung cancer appropriate for modern society. Judgement of pulmonary nodules by simply relying on CT imaging features may produce a relatively high false-positive rate, thus redundantly adding some unnecessary invasive examinations, which may increase the patient’s psychosomatic sufferings [7–9]. In this study, we intended to establish a more effective and clinically applicable prediction model by analyzing the clinical data of patients with pathologically confirmed pulmonary nodules, based on which to screen out risk factors that have the highest value of predicting malignancy or benignity of pulmonary nodules from GGN patients with complete and multiple clinical data for the sake of improving the screening efficiency of pulmonary nodules.
All pulmonary nodules ≤ 30 mm in diameter surgically resected from patients who received surgery in Yiwu Central Hospital (Yiwu, China) between January 2018 and March 2024 were collected and analyzed.
The inclusion criteria were (1) all solitary GGNs in the lung parenchyma as shown on CT imaging of the said hospital; (2) nodules ≤ 30 mm in diameter; (3) patients with complete clinical data; (4) the availability of pathological diagnosis after surgery; (5) the availability of tumor markers of CEA, SCC, NSE and CYFRA21-1; and (6) no anti-tumor therapy administered before surgery.
The exclusion criteria were patients (1) with lung metastatic tumors; (2) with nodules > 30 mm in diameter; (3) with incomplete clinical data; (4) whose preoperative CT examination was not performed in our hospital; and (5) who had received chemotherapy, radiotherapy, or other treatments before surgery.
The admission number, gender, age, pathologic results, tumor markers, pulmonary CT imaging features, smoking history, personal tumor history, family tumor history, diabetes history, chronic underlying disease history, insomnia and constipation were collected form patients with pulmonary nodules detected in physical examinations who did not receive any treatment before surgery.
Criteria of judging the pathological results of the pulmonary nodule tissues.
Malignant malignant pulmonary GGNs that were surgically resected and pathologically confirmed.
Benign benign pulmonary GGNs that were surgically resected and pathologically confirmed.
Laboratory tumor marker indexes obtained within a month before surgery.
The imaging information was collected, including the pulmonary nodule mixed-GGN (mGGN) or pure-GGN (pGGN); whether the nodular lesion was accompanied with VCS and short spicule sign; whether the boundary was smooth; whether the morphology was regular; whether the nodule was accompanied with pleural indentation; whether there was vacuolar sign; the maximum diameter of the nodule (mm); and the CT value of the nodular lesion. Lung cancer vascular convergence sign refers to the presence of multiple blood vessels that nourish and gather towards lung cancer. The vascular convergence near the pulmonary hilum are mostly composed of blood vessels or bronchi, and the blood vessels are mostly dilated small arteries with thickened vessel walls, indicating abundant blood supply to lung cancer; The vascular convergence on the distal hilum side is composed of dilated venules, which may be related to obstructed venous return. CT shows that one or several small pulmonary blood vessels are pulled and displaced towards the lesion, interrupting or penetrating the lesion at the site of the lesion [10].
The pulmonary nodules were grouped according to the pathological results. Comparison between two groups was performed using R language for data analysis. The above risk factors were analyzed by single-factor logistic regression analysis (whether the laboratory indexes were elevated or not, the presence or absence of underlying diseases, and imaging presentations were defined as binary “yes” was defined as 1, and “no” as 0; “elevated” as 1, and “normal” as 0). Significant results obtained in unilateral analysis were subjected to multi-factor logistic regression analysis. p < 0.05 was considered statistically significant. The mathematical prediction equation is as mP = e^x^/(1 + e^x^), where p > 0.5 was considered malignant, and p ≤ 0.5 was considered benign.
Of the 261 surgically resected pulmonary nodular lesions, 211 were pathologically confirmed as malignant, and 50 as benign. As shown in Table 1, the female/male (F/M) ratio in the malignant group was 143/68, and 20/30 in the benign group, indicating that the female gender was the predominant risk factor for lung cancer. In addition, the smoking history was a statistically significant risk factor in the malignant group, and absolutely a main factor for lung cancer in the male population. There was no significant difference in age and body weight (BW) between the malignant and benign groups. Pathology of the 211 lung cancer cases, 140 cases (66.4%) were adenocarcinoma, 56 cases (26.5%) were in situ cancer, and squamous cell carcinoma (SCC) and small cell lung cancer (SCLC) accounted for a very small proportion (Table 2). Single-factor regression analysis of factors related to malignancy of pulmonary nodules showed that gender, VCS, mean CT value of the nodule, maximum diameter of the nodule, and smoking history were significant factors for predicting malignancy or benignity of the pulmonary nodules (Table 3). Multi-factor regression analysis showed that gender, mGGN, VCS, mean CT value of the nodule, and maximum diameter of the nodule were independent factors of predicting malignancy or benignity of the pulmonary nodules (Table 4) and could be used as parameters of predicting malignancy or benignity of the pulmonary nodules. An overview of the above findings concluded the mP = e^x^/ (1 + e^x^); when mP was > 0.5, the nodule was considered malignant, and when mP was ≤ 0.5, the nodule was considered benign. x=-2.46 + 1.032gender + 1.85mGGN + 1.40VCS-0.0027nodule mean CT value + 0.078*nodule maximum diameter, where e is a natural logarithm; if the patient was a female, gender = 1 (otherwise = 0); if the pulmonary nodule was a mixed GGN, mGGN = 1 (otherwise = 0); if the pulmonary nodule had vascular convergence sign, VCS = 1 (otherwise = 0). The prediction performance of the mathematical prediction model was verified and the results are shown in Tables 5 and 6. The specificity of the prediction model in the cohort with pulmonary nodules was 0.86864, the sensitivity was 0.76000, the negative prediction value was 0.97156, and the positive prediction value was 0.3800 versus 1.0, 0.9, 0.25, 1 and 0.25 in the verification group, respectively. AUC = 0.814 in the ROC curve graph of the model group, as shown in Fig. 1 below. AUC = 0.952 in the ROC curve graph of the verification group, as shown in Fig. 2 below.
Fig. 1 ROC curve graph of the prediction model in the model group, AUC=0.814
Fig. 2 ROC curve graph of the prediction model in the verification group, AUC=0.952
The surgically resected pulmonary nodules were pathologically diagnosed as malignant in 28 patients, 28 GGNs were removed in total, and a total of 4 patients with GGNs underwent surgical resection, and pathological examination confirmed benign lesions. Four GGNs were removed. The data of the verification group were substituted to the model developed in this study. The prediction performance of the prediction model was verified by the pathological results as the gold standard and the ROC curve drawn by the R language software.
The incidence of pulmonary nodules is high, but as most of them grow slowly and their volume doubling time varies individually, the diagnostic value of depending on long-term imaging observation of their changes is limited, which may cause misdiagnosis or mal-diagnosis [11]. In recent years, various models for predicting the nature of pulmonary nodules based on imaging or biological characteristics have been developed successively, such as the prediction model by combining ctDNA with imaging features, DNA methylation technique, tsRNA technique, and Imaging omics [12–16]. Their disadvantages lie in their prediction of malignancy or benignity simply depending on unilateral features of pulmonary nodules, and therefore their false-positive rate is relatively high. The human body is a unified whole and needs to be considered from all aspects including the gender, smoking history, underlying disease, related tumor history, comorbidities, laboratory parameters, and pulmonary CT imaging omics from which to screen out the most valuable prediction factors. Our study revealed that gender, mGGN, VCS, mean CT value and maximum nodule diameter were independent factors for predicting malignancy or benignity of pulmonary nodules. As shown in Table I, the female gender is the predominant factor in the lung cancer cohort, and the smoking history is the predominant factor in the male lung cancer cohort. This finding gives us an alert reminder that we need to enhance research on factors contributing to lung cancer in non-smoking women and propose protective and interventional measures. In addition, science popularization education on smoking cessation remains an imperative task. Current research maintains an indisputable connection between smoking and lung cancer. Although tobacco-derived carcinogens and enzyme polymorphism have been confirmed to increase the risk of cancer occurrence in smokers, the latest epidemiological data have documented gender specificity also as an extra new factor of lung cancer. The probability of developing cancer in smoking women is three-fold high as that in non-smoking women, and the risk is even higher in non-menopausal women who use estrogen or progesterone therapies. Other multiple studies have demonstrated that there exist significant gender-related differences in cancer, such as differences in histological distribution, the incidence in non-smokers, EGFR mutation, and DNA adduct accumulation. Whether women are more likely to be affected by smoking-related carcinogenic factors or obtain more benefits from cancer screening needs to be further investigated [17–21]. Many previous studies have found that with respect to CT imaging features, mGGN, VCS, CT value and maximum nodule diameter are four parameters that have been demonstrated as the key factors in predicting malignancy of pulmonary nodules. It is reported that the nodule volume and CT value are closely related to the degree of infiltrative growth of mGGN, and two key imaging markers as well [22–27]. Pulmonary nodules have strong growth inertia, and the appearance of VCS often signifies the initiation or even infiltration of malignant transformation of the pulmonary nodules. Many studies have reported VCS as an independent factor of malignant pulmonary nodules, knowing that VCS is a very classic imaging parameter of predicting malignancy or benign of pulmonary nodules, the CT image of vascular convergence sign is shown in Fig. 3 [28–31]. As shown by our statistical analysis, the mathematical prediction model established by combining the above four most critical CT features and gender had good negative prediction performance, especially in women patients with pulmonary nodules.Therefore, if a pulmonary nodule shows changes in density, maximum diameter, CT features, and age during follow-up, the probability of malignancy will also change accordingly, just as the Fig. 4. This is a great medical contribution to the vast majority of women with pulmonary nodules.
Fig. 3 This is a CT image of vascular convergence sign in a patient (woman, 37 years old) with malignant pulmonary nodules in this study, The red arrow indicates that blood vessels are gathering towards pulmonary nodules
Fig. 4 An 88 year old female was diagnosed with pulmonary GGN in 2017, and a chest CT scan in 2024 showed a significant increase in size and density of the lesion, accompanied by vascular convergence sign, the yellow arrow indicates the pulmonary nodule. The malignant probability of this GGN has significantly increased
Below is the link to the electronic supplementary material.
CXS contributed substantially to the study design. XDZ contributed substantially to data obtainment. CXS and XDZ had full access to all data in the study, took responsibility for data integrity and the accuracy of data analysis, including adverse effects in particular, and contributed substantially to the writing of the manuscript. JCD provided design guidance for this study.
This research project is supported by the Science and Technology Bureau of Yiwu City, Zhejiang Province, with funding number 21-3-100.
All data has been provided within the manuscript and supplementary information files.
This study is a retrospective clinical data analysis study that did not expose any patient privacy. The Ethics Committee of Yiwu Central Hospital agreed to waive the signing of informed consent forms. All research materials were reviewed and approved by the Ethics Committee of Yiwu Central Hospital, Zhejiang Province, China. Approval H2021-IRB-045.
Not Applicable. All datas, images, and tables publicly published in this article do not involve patients’ privacy, and we have not compromised patients’ anonymity.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Changxing Shen, Email: changxing9737@126.com.
Jingcheng Dong, Email: jcdong2004@126.com.
All data has been provided within the manuscript and supplementary information files.