According to the Global Cancer Statics 2020, lung cancer is the leading cause of cancer mortality worldwide, with an estimated 2.2 million new cases and 1.8 million new deaths in 2020 (1). SCLC is characterized as one of the most lethal and aggressive types which accounts for around 15% of lung cancer, while NSCLC accounts for nearly 80%. It is widely known that SCLC is a recalcitrant carcinoma with remarkable metastatic and recurrent proclivity. As previous studies have reported, only one-third of SCLC patients are initially diagnosed with early-stage, while close to 70% of the cases are advanced stage at diagnosis (2, 3). In a comprehensive analysis of 358 SCLC patients with extensive-stage, 43.8% of patients died within 12 months (4). Another study from China also reported that the median OS was just 11 months and 58.4% of patients with extensive-stage SCLC died within 1 year (5). To sum up, patients with SCLC have a worse prognosis which would be influenced by many factors such as age, lymph node metastasis, distant metastasis, and so on. Nonetheless, the influence of pleural invasion on prognosis remains unknown.

Pleural invasion (PI) has been established as a negative prognostic factor in NSCLC and was first included as a non-size based T2 descriptor in the 5th edition AJCC staging system in 1997 (6). Pathological PI is classified into the following subgroups according to the International Association for the Study of Lung Cancer: PL0, the tumor grows within the parenchyma or does not completely penetrate the elastic layer; PL1, the tumor extends beyond the elastic layer; PL2, tumor invades into the surface of the visceral pleura; PL3, tumor invades into or through the parietal pleura (7, 8). Generally speaking, PL0 indicates no evidence of PI, while PL1, PL2, and PL3 stand for tumor invasion of pleura. In the 8th TNM staging system, tumors ≤ 3 cm (T1a or T1b) with PL1 or PL2 are upgraded to T2a while tumors with PL3 are defined as T3. Previous research has reported that NSCLC patients with visceral pleural invasion (VPI) are associated with a higher incidence of malignant pleural effusion, mediastinal lymph node metastasis, and postoperative recurrence (9–11). In a retrospective study of 2,657 patients with T1-4N0-2M0 NSCLC, they demonstrated that VPI was the strongest significant independent predictor of recurrence in patients with pathological stage I treated without adjuvant chemotherapy (12). In another series of 16,315 NSCLC patients with stage I–II, VPI occurred in 3,389 patients (21%) and it was a prevalent finding associated with worse prognosis, even among patients with tumors > 3 cm (13). Although there were some studies investigating prognostic factors for SCLC, most of them ignored the effect of PI on survival or were limited to a small number of cases (14–17). However, the effect of PI on OS in patients with SCLC is seldom reported.

As a widely used method, nomograms can accurately predict OS of cancer patients. Most existing nomograms, however, are derived from NSCLC patients with VPI. Nomograms for the survival of SCLC patients with PI have not been published until now.

Therefore, we aimed to determine whether the presence of PI could influence OS in SCLC patients and construct a novel nomogram to predict OS in SCLC patients with PI based on the demographic and clinicopathologic variables from the SEER database. As a result, this could facilitate individualized patient care as well as medical therapy.

We conducted this retrospective study based on SCLC patients with PI and without PI to investigate its prognostic value via the SEER database. Besides, independent prognostic factors for SCLC patients with PI were identified. By utilizing multivariate Cox regression and PSM analyses to balance the confounding factors, our results demonstrated that SCLC patients with PI were correlated with a significantly worse prognosis. Furthermore, to efficiently predict OS for SCLC patients with PI, we established a prognostic nomogram with reliable accuracy and discriminative ability which were validated by ROC, calibration, and DCA curves. This nomogram can serve as a practical tool for clinicians to identify patients with a high risk of poor survival and to determine the optimal clinical treatment for patients diagnosed with PI.

Pleural invasion was first reported in 1958 by Brewer et al. to be a poor negative survival predictor for lung cancer (18). Recently, the studies concentrated on PI about its incidence and prognostic effect are increasing gradually. The incidence of VPI is variable, accounting for approximately 11.5%–46.6% of total NSCLC cases (18–21). It is widely accepted that the presence of VPI is confirmed to be an adverse prognostic factor in NSCLC, especially in patients with early stage. In a series of 886 NSCLC patients, there was a significant difference between the patients with VPI and without VPI in the 5-year OS rates, which were 80.8%, 63.7%, and 49.6% in PL0, PL1, and PL2, respectively (22). In line with another study consisting of 1,488 patients with surgically resected non-small cell carcinoma, the OS of patients with PI was worse than those without PI and the 5-year OS rates with PL0, PL1, PL2, and PL3 tumors were 80%, 60%, 55%, and 52%, respectively (23). Other comparative studies also revealed the same poor prognosis in NSCLC patients with VPI or PI (11, 21, 24). However, there are some shortcomings in these studies. For example, some studies did not adjust for baseline confounders to assess the prognostic value of PI accurately. Other studies had a limited number of cases or were based on NSCLC. The prognostic value of PI in SCLC patients had less attention.

In the present study, we confirmed that PI was associated with statistically significantly deteriorated OS among SCLC patients in both original and matched cohorts. In addition, the current multivariate Cox regression analysis showed that PI was an independent worse prognostic factor before and after matching, consistent with previous studies showing VPI led to poor OS. Stratified analysis indicated that OS difference was not existent between the non-PI group and PI group in SCLC patients across all subgroups, except for younger age, female, white race, right laterality, lower lobe and lung, T2 stage, without distant metastasis, without radiotherapy, surgery, and chemotherapy. The number of SCLC patients reviewed who were located in the lower lobe and lung site (295/774), non-right (357/784), and non-white (129/784) was relatively low. Given the small sample size, the results cannot be accurately evaluated. Age was regarded as a prognostic factor in SCLC patients as well as our results. SCLC patients with older age may have a poor physical condition, have a high risk of metastasis, and are more likely to die from other diseases. Whether in the 7th or 8th TNM classification, PI was included in the T stage. In addition, VPI was described as T2, while parietal PI was described as T3. A retrospective analysis published by Qi et al. showed that the 5-year OS of patients without VPI was significantly better than those with VPI in NSCLC with pT2 stage (25). Our results demonstrated that PI exhibited a significant impact on survival in T2 stage, whereas had no effect in T3 and T4 stage. Maybe higher T stage is often accompanied by a high risk of metastasis or more extents of invasion which impairs the effect of PI in SCLC patients. It deserves further investigation to illuminate the phenomenon. Most studies said in lung cancer patients without node lymph node metastasis, VPI led to a worse survival than those without VPI, while no significant difference existed in patients with lymph node metastasis (12, 21). Though our study showed that the OS of patients with PI and without PI was the same regardless of lymph node metastasis, the OS seemed to favored non-PI. Perhaps, a relatively small number of patients underestimated the impact of PI in N0 diseases and N⁺ diseases. Besides, we didn't distinguish the degree of PI so their actual effect may be confused. Generally, the incidence of distant metastasis at the time of initial diagnosis of SCLC is more than 60% and overall survival and median survival are worse in patients with metastatic SCLC (26). Therefore, these may be the reason why the impact of PI in metastatic SCLC patients was not obvious in this study.

The reason for the worse prognosis caused by PI is still unclear. Perhaps, one possible reason is a high probability of lymph node metastasis in patients with PI. Because of the abundance in lymphatic vessels of the pleura, lung cancer cells in the subpleural tend to invade the pleural layer rapidly through the flow of pleural effusions in the pleural cavity. Once lymph nodes have been involved, malignant cells will further migrate by invading mediastinal lymphatic vessels (27). Like previous studies reported, VPI was significantly correlated with more extensive hilar or mediastinal lymph node metastasis (22, 28). Thus, PI may be the first step for further distant metastasis through minimal hematogenous dissemination, regardless of lymph node involvement.

Nomograms can predict prognosis efficiently in other tumors. Tang et al. combined age, sex, grade at diagnosis, number of metastatic organs, histology, and chemotherapy to predict cancer-specific survival in metastatic esophageal cancer (29). In metastatic cutaneous melanoma, a nomogram was established based on age, sex, race, marital status, insurance, number of metastatic organs, T stage, N stage, surgery, and chemotherapy (30). To date, there are many studies on the prognosis and nomograms of SCLC, nonetheless, there is still no study concentrated on the prognosis in SCLC patients with PI based on the clinical characteristics, which leads to a worse cancer prognosis. Furthermore, a novel prognostic nomogram was constructed in this study.

In a retrospective study of 1,374 NSCLC patients with stage pT1-2N2M0, Zhang et al. exhibited that the presence of VPI is a poor prognostic factor, even in patients III/N2 NSCLC (31). And beyond that, they also revealed that the OS could be improved significantly by chemotherapy, especially in non-VPI patients. What's more, Wang et al. constructed a nomogram based on age, gender, race, histology, T stage, N stage, and M stage to predict survival in NSCLC patients with VPI (27). As well as in SCLC, VPI was an indicator of a poor prognosis for SCLC with surgery, especially in those N0 diseases, besides, age, N stage, and chemotherapy were recognized as independent prognostic factors in patients with VPI (32). Although chemotherapy plus immune checkpoint inhibitors are recommended for patients with extensive-stage SCLC, platinum plus etoposide has been recommended for many years and is a common treatment of advanced SCLC in practice (33, 34). In line with these reports, age, N stage, and chemotherapy were associated with OS in our study. Distant metastasis was a strong independent factor for OS and was incorporated into the prognostic nomogram (35, 36), consistent with the nomogram in our study. In a propensity score-matched analysis of SCLC patients from America and China, surgical resection could significantly improve overall survival and lung cancer-specific survival of stage III SCLC patients (37). In another propensity-score analysis of early-stage patients with SCLC, Wang et al. demonstrated a statistically significant benefit for surgery (38). But, considering the limited number of patients with surgery in the present study, it should be cautious to make the decision on surgery in SCLC patients with PI based on the patients' conditions. Radiotherapy has been demonstrated to improve median overall survival in early stage of SCLC (39, 40), as well as in patients with PI in our study. Further investigation on benefits of radiotherapy in SCLC patients with PI is required.

However, there are still some limitations that need to pay attention in this study. First, prospective randomized controlled studies are required to confirm our results because of the selection bias of the retrospective study. Second, due to the lack of external validation in the present study, an inherent bias can not avoid. Third, because of high mortality and low survival, the small number of SCLC patients with pleural invasion may have contributed to the possible error. Besides, it may reduce the accuracy of the nomogram without histological data in SCLC with PI. Finally, the data on tumor grade and the specific degree of pleural invasion was incomplete which may underestimate their impact on overall survival. The survival effect of specific degree of tumor grade and pleural invasion should be further investigate in the future.

Our study demonstrated that the existence of PI in SCLC patients had a statistically significant adverse impact on survival. At the same time, we comprehensively demonstrated that age, N stage, M stage, surgery, radiotherapy, and chemotherapy were the independent prognostic factors for SLCLC patients with PI. The nomogram may beapplied as a clinically useful to assess the prognosis of SCLC patients with PI and could facilitate clinical decision-making.