Lung cancer remains the first cause of oncological death, and in recent years, LUAD is becoming more and more frequent. Although the proportion of IMAs in LUAD is relatively low, it was believed that people with IMA had worse prognosis.
IMAs was different from other LUAD, characterized by goblet or columnar tumor cells with abundant intracytoplasmic mucin and basally located nuclei. In some cases, IMAs showed the mixture of different pathological types[14, 15]. IMAs have special genetic signatures. Studies found that many genes unexpectedly enriched in mucin-producing gastrointestinal, pancreatic, and breast cancer showed significant differences in IMAs, including FOXA3, SPDEF, etc.. And there was evidence that B7-H4 expressed in IMAs, which was considered as a therapeutic target for immune checkpoint therapy. Besides, Kadota K et al. and Righi L et al. found that IMA was connected with KRAS mutation[17, 18], while NRG1 fusion looked frequent in IMAs even without KRAS mutations[19-21]. These unique pathological features may affect pathological diagnosis.
In recent years, there have been relatively few studies on systematic reviews of IMAs treatment. Therefore, we decided to constructed a nomogram to predict the prognosis for IMAs and helped to provided new sight for treatment.
In this research, patients diagnosed with IMA was included into our analysis. There are just over 1000 patients, and we included 774 patients with complete clinical information (Supplementary Fig. S1). These patients had a reasonable age distribution, and most had received surgery.
In univariate analysis, gender, age, differentiation grade, TNM stage, and treatments including surgery, radiation, chemotherapy were all related to IMAs progression (Table 1, Fig. 1, Fig. 2). As we can see, surgery treatment would decrease the HR, while radiation treatment and chemotherapy would not.
We conducted multivariate analysis using these significant variables in univariate analysis (Table 2). Except for laterality and gender, all the factors were statistically significant for OS. This result verified that older age, poorly differentiated grade, bilateral laterality, higher TNM stage, no surgery, radiation and chemotherapy were independent prognostic factors and improved the HR.
IMAs were mainly found in lower lobes and presented with multifocal consolidation and lung-to-lung or pleural metastasis. Many researches indicated tumor size and invasive size might be the independent factor influencing the prognosis of IMAs[9, 23]. Apart from surgery, non-TKI chemotherapy was used in many IMA patients, while OS seemed no improvement. As we could see, more than 70% patients received surgery, and for both OS and LCSS, surgery looked like the only treatment that would improve survival. Consistent with previous reports, chemotherapy does not promote prognosis, and so is radiation therapy, suggesting that chemotherapy and radiation therapy might not bring survival benefits for IMAs. Therefore, for patients with a clear diagnosis of IMA, we still choose surgical treatment as the first choice. But the effect of surgery combined with chemoradiotherapy remains to be seen.
We plotted nomograms based on independent prognostic factors suggested in multiple factors (Fig. 3). For OS, age was the main factor that influenced prognosis, and T stage for LCSS. The accuracy of this model was measured via ROC curves and calibration plots. The training cohort AUC was 0.834(95%CI: 0.791-0.876) and 0.830(95%CI: 0.789-0.872) for 3- and 5-year OS, respectively, with 0.839(95%CI: 0.794-0.884) and 0.839(95%CI: 0.796-0.882) for 3- and 5-year LCSS (Fig. 4). Furthermore, we compared the model with 8th edition AJCC TNM staging system and it showed that our model had a better predictive value for IMAs than TNM staging system (Fig. 5, Fig. 6), and IDI was 0.098(95%CI: 0.059-0.149), 0.105(95%CI: 0.068-0.162) for 3-/5- year OS and 0.105(95%CI: 0.068-0.162), 0.105(95%CI: 0.068-0.162) for 3-/5- year LCSS (Fig. 6).
To our knowledge, this is the first review model for both clinical characteristics and treatment of IMAs based on SEER database. The comprehensive clinical information of the SEER database provided great support for the study. However, there are many limitations that must be considered. IMA is difficult to diagnose till now, and many patients are classified as "adenocarcinoma" without specific pathological types. At the same time, the number of patients with a clear diagnosis of IMA in the databases around us is also very small, and we have not been able to verify the accuracy of this model in other databases. However, this model comprehensively evaluates the clinical characteristics and treatment, and provides ideas for improving the prognosis of IMA.