The prognostic factors in EC patients are known to be complicated, thus, an accurate and comprehensive prognostic method is of great significance for the evaluation of esophageal cancer and the optimization of treatment strategies. So far, TNM staging system proposed by AJCC has been widely accepted for the assessment of cancers including EC. For patients who are adapted to surgery, determining explicit T and N staging is the most crucial task throughout.(20) However, the TNM staging system mainly represents anatomical relevance, sometimes does not fully reflect the prognosis. Therefore, a type of model called nomogram that can incorporate multiple prognostic factors is maturing. Recently, nomograms are popularly investigated in cancer-related survival, recurrence and metastasis with considerable potential.(21–24) Overall survival (OS) is regarded as a most commonly used outcome for cancer-related survival prognosis with confounding factors while the nomogram studies on EC-related OS prediction are less of quantity. We involved some classical and novel factors which are clinically available to establish a nomogram which is unprecedented and innovative.
By means of literature review and clinical experience, a set of factors were included for the initial analysis in EC patients underwent radical esophagectomy. In addition to some conventional demographic and clinicopathological factors, such as age, gender, tumor location, tumor differentiation, T and N staging, etc., the verified cancer-related prognostic factors, including Albumin to globulin ratio (AGR), Neutrophil to lymphocyte ratio (NLR), Platelet to lymphocyte ratio (PLR), Positive lymph nodes ratio (PRL), Low-density lipoprotein (LDL), and Prognostic Nutritional Index (PNI) are candidates of nomogram establishment.(25, 26) Among them, six factors including age, gender, N stage (with modified), AGR, PRL, PNI were identified as independent prognostic factors by statistical Cox hazard regression analysis, and the Hazard Ratio (HR) of each factor displays the weight of score in the nomogram.(9)
For demographic characteristics, we found that two factors, age and gender, were associated with OS of EC patients after radical esophagectomy, and thus were included into the establishment of nomogram. As a matter of genetic difference, males have been found with higher incidence and mortality in a variety of malignancies, as found in various studies and confirmed in ours.(1, 27, 28) In addition to cancer-related factors, it cannot be ignored in the prognosis of elderly cancer patients that the pathological mechanisms induced by aging may cause more nutritional and metabolic diseases and disorders such as amyotrophy, metabolism damage and neurological disease, contributing to the impediment of longevity.(29–31) In the result of our study, the age over 65 years old was certified as an independent factor for OS prediction, conforming to the interpretation mentioned above.
Further in comparison with some nomogram studies in resectable EC patients, Shao’s study focuses on the predictive ability of inflammation-related factors for OS, and accordingly building a nomogram model. His study is partially similar to our study, and some factors, including PLR and NLR, are selected to be used in our initial analysis.(32) However, with no statistical difference found, they were removed from multivariable Cox hazard regression analysis in our study. The above situation indicates that the factors affecting the OS of radical resection EC patients are very complicated. PLR and NLR are not only related to the prognosis of malignant tumors, but also determined in inflammatory diseases such as Rheumatic Disease and cardiovascular diseases.(33, 34) EC is dependent on the complex interaction between the tumor and the hosts’ inflammatory response, so the distinction of PLR and NLR shown between Shao’s study and ours imply that the independence of factors affecting prognosis are unseparated from the integrity of the individual physiology.(35) As aforementioned, a low AGR is associated with increased cancer mortality in cancer patients, a generally healthy screened population study proves AGR as a risk factor for cancer incidence and mortality, in both short- and long-term cohorts.(36) There is biological plausibility for the link between low AGR and increased cancer incidence and mortality that an increase of cytokines on account of tumor microenvironment may elevate the total protein levels with induced albumin synthesis suppression in the liver.(37, 38) As for EC patients, a malnutrition status relevant decreased serum albumin and AGR level could be another mechanism of poor survival. PNI is composed of serum albumin level and lymphocyte count as a calculative index which is widely utilized for prognostic evaluation of various malignancies. As a factor reflects to tumor-related nutritional status and system inflammation response, the role of PNI in EC patients’ prognosis is highly desirable for study. Kazuo’s study demonstrated the predictive role of PNI in EC patients outcome and its inseparable relationship with Tumor-infiltrating lymphocytes (TILs).(15) As proved, PNI was significantly associated with OS in Cox hazard regression analysis with a negative correlation of HR 0.676 (95%CI: 0.463–0.987) in our study, and it's firstly included in the nomogram for OS prediction in resectable EC patients. Moreover, the laboratory-sourced factors we selected are all preoperative to eliminate the inflammatory and metabolic bias caused by surgery.
For resectable EC patients, though there is a gap in skills between surgeons, the lymph node dissection degree is considered as an effective adjudication, extended lymph node dissection proves a significant amelioration on the prognosis of EC patients.(39, 40) Recently, the ratio of positive retrieved lymph nodes to total number of retrieved lymph nodes—known as the positive lymph node ratio (abbreviation as PRL or LNR)—has been shown to be a superior indicator of survival in EC.(41, 42) Compared with N stage according to AJCC, it not only reflects the quantity but the extent of the lymphatic metastasis, especially for the EC patients with less than two-field lymph node dissection or underestimated N staging due to insufficient retrieve of positive lymph nodes.(43) Evidently, PRL showed a higher HR of 2.970 (95%CI: 1.902–4.638) and greatest weight in nomogram compared with N stage in our study.
Furthermore, we used the method of calibration curve and C-index to verify our nomogram model internally and externally with a considerable accuracy and consistency. Additionally, ROC curves of the nomogram were performed for the prediction of OS, and area under curve (AUC) is the crucial index to discriminate the accuracy of different prediction methods. The larger value of the AUC (no less than 0.5) manifests more outstanding predictive ability.(44) Compared with AJCC T and N staging strategy, we observed that the AUC of the nomogram in the primary cohort reached 0.801 (95%CI: 0.744–0.859), which was significantly higher than the T and N stages. Despite the fact that the AUC of nomogram in validation cohort dropped to 0.727 (95%CI: 0.626–0.829), it’s still superior to AJCC T and N staging in predicting OS. Eventually, by dividing the primary cohort into three equal risk groups according to the total nomogram score (TNS) of each patient, the OS of the three groups showed a significant difference. Similar result was also observed in validation cohort. This indicates that the nomogram model, if well-established, can demonstrate favorable clinical applicability for the prediction of OS in EC patients underwent radical esophagectomy.
In the end, we performed analysis based on the risk classification system in total cohort, comparing with an authoritative tumor staging system AJCC TNM staging to identify its comprehensive application clinically. Eventually, the risk classification system showed strong correlation with TNM staging (r2 = 0.647, P < 0.001), and a better efficacy in OS prediction than TNM staging was informed by log-rank test and ROC curves.
In contrast with studies on nomogram in EC patients, PNI, which is scarcely investigated in nomogram of EC patients, was integrated into our study, together with internal and external validation of nomogram. Besides, an applicable risk classification system based on nomogram is developed in our study.(11, 16, 32)
Several limitations exist in our study. Firstly, it’s a retrospective analysis with probably inherent bias. Secondly, the primary and validation cohorts of this study are from a single center. Thirdly, the threshold index of factors mentioned in our study is heterogeneous. Therefore, a large-sample, multi-center randomized trial needs to be further verified.