Development and external validation of a nomogram for predicting the conditional probability of survival after D2 lymphadenectomy for gastric cancer: A multicentre study

Abstract

Background

Previous studies have elucidated that on average, long-term cancer survivors have better prognoses than newly diagnosed individuals. This study aimed to devise a nomogram to predict the conditional probability of cancer-specific survival (CPCS) in gastric cancer (GC) patients after D2 lymphadenectomy.

Methods

Clinicopathological data for 2,596 GC patients who underwent D2 lymphadenectomy in an Eastern institution (the training cohort) were retrospectively analysed. Cancer-specific survival (CSS) was predicted using Cox regression models. A nomogram was constructed to predict CPCS at 3 and 5 years post-gastrectomy. Two external validations were performed using a cohort of 2,198 Chinese patients and a cohort of 504 Italian patients.

Results

In the training cohort, the 5-year CPCS was 59.2% immediately post-gastrectomy and increased to 68.8%, 79.7%, and 88.8% at 1, 2, and 3 years post-gastrectomy, respectively. Multivariate Cox regression analyses showed that age; tumour site, size and invasion depth; numbers of examined and metastatic lymph nodes; and surgical margins were independent prognostic factors of CSS (all P < 0.05) and formed the nomogram predictor variables. Internal validation showed that the conditional nomogram exhibited good discrimination ability at 3 and 5 years post-gastrectomy (concordance index, 0.794 and 0.789, respectively). External validation showed a 3- and 5-year concordance index of 0.788 and 0.785, respectively, in the Chinese cohort, and 0.792 and 0.787, respectively, in the Italian cohort. Calibration of the nomogram predicted that survival corresponded closely with actual survival.

Conclusions: we developed a robust nomogram to predict CPCS after D2 lymphadenectomy for GC based on survival duration.

Introduction

The incidence of gastric cancer (GC) in China is much higher than that in any other country [1]. Moreover, in China, GC is the second leading cause of cancer-related death [2]. Gastrectomy with D2 lymph node (LN) dissection is the standard treatment for curable GC in eastern Asia [3]. Recent reports revealed a trend towards improved cancer-specific survival (CSS) rates in patients who have undergone D2 lymphadenectomy, especially when lymphadenectomy was performed in a high-volume center [[4], [5], [6], [7]].

Death risk assessments for GC patients are customarily based on the American Joint Committee on Cancer Tumour Node Metastasis (AJCC TNM) staging system or on more detailed nomograms [8,9]. Most reports evaluate the risk of death only from the time of surgery. However, recent studies have shown that the risk of death in patients with GC is not constant [10,11]. The conditional probability of CSS (CPCS) represents the probability of remaining life expectancy for a specific number of postoperative years based on the length of time that the patient has already survived. The benefits of CPCS estimates have already been demonstrated in various types of cancers [[12], [13], [14]]. For cancer survivors, CPCS estimates might allow patients to quantify improvements in their prognosis over time. CPCS estimates are equally necessary for clinicians because they might allow adjustment of the frequency and type of follow-up over time. In contrast to conventional survival estimates, CPCS estimates can provide cancer survivors and clinicians with a more accurate evaluation of the changes in the risk of death.

A nomogram is a useful tool for predicting the probability of survival of an individual patient with a simple way and for improving personalized decision-making in cancer treatment accordingly [15]. Recently, nomograms based on different databases have been constructed for predicting individual survival of GC [9,16,17]. Few studies have focused on the correlation of D2 lymphadenectomy to CSS; more importantly, these models cannot be applied to the prediction of CPCS. In assessing the prognosis of cancer survivors, CPCS estimates may drastically differ from baseline CSS predictions, especially after several years of cancer mortality-free follow-up.

Because of the high incidence of GC in eastern Asia, surgeons in these areas have accumulated substantial experience in GC-related surgical techniques, and gastrectomy with D2 lymphadenectomy is routinely performed with low morbidity and mortality rates [18,19]. The previously reported static GC nomograms did not provide an accurate prediction of the dynamic outcomes for cancer survivors because the prediction of the previous nomograms was only based on conventional CSS. Therefore, to date, the dynamic prognosis of GC patients who have undergone D2 lymphadenectomy has not been precisely predicted. Considering the sustainable surveillance requirement, this study aimed to construct a new conditional survival nomogram for predicting CPCS in GC patients from a large-volume institution after D2 lymphadenectomy. The performance of the nomogram model was validated using a cohort of 2,198 patients from China and another cohort of 504 patients from Italy.