Biliary tract cancer (BTC) is a group of biliary malignancies mainly comprising intrahepatic cholangiocarcinoma (ICC), extrahepatic cholangiocarcinoma (ECC), and gallbladder cancer (GBC). ICC is estimated to account for about 0.59% of all new cancer cases in 2018, while GBC containing ECC accounts for 1.2% worldwide [1]. Despite the low incidence, the mortality of BTC is relatively high, comprising up to 2% of total cancer deaths per year [2]. BTC is acknowledged to be aggressive with a dismal prognosis; according to the RARECAREnet project, based on cases between 2000 and 2007 in Europe, survival for ICC was 25% at 1 year, 8% at 3 years, and 5% at 5 years, while the 5-year survival rate of GBC was less than 5% [3]. Survival for ECC was marginally better, albeit dramatically reducing from 48% at 1 year to 23% at 3 years and 17% at 5 years [4]. Increased incidence has been observed in cholangiocarcinoma and GBC [5] worldwide. Unfortunately, the only available curative treatment for BTC is surgery.

The prognosis of BTC is highly dependent on tumor stage, as the estimated 5-year survival is 50% for American Joint Committee on Cancer (AJCC) stage I, 30% for stage II, 10% for stage III, and 0% for stage IV [6]. Both novel biomarkers and staging systems are warranted, as no prognostic biomarkers are currently used clinically, and the AJCC and Bismuth-Corlette staging system are not accurate in predicting survival [2]. Currently, nodal involvement, tumor differentiation, preoperative and postoperative serum carbohydrate antigen 19-9 (CA19-9) [7] have been confirmed as important prognostic factors in cholangiocarcinoma patients undergoing surgery. Furthermore, lymph node status and the presence of R0 curative resection are closely related to survival in GBC [8]. During the search for novel factors, inflammation-related indexes have been extensively studied as prognostic factors, but their roles in predicting overall survival (OS) for BTC are still under discussion. The neutrophil-to-lymphocyte ratio, one of the most popular inflammation-related indices, cannot maintain its role as a dependent prognostic factor of OS in BTC [9]; thus, more accurate predictors need to be identified.

It has been widely accepted that platelets contribute to tumor progression through a variety of mechanisms, including angiogenesis, assisting metastasis, and regulating immunity. Platelets can protect tumor cells from immune attack by modulating innate and adaptive immune responses [10]. The role of platelets as regulators of inflammation has been recognized, and as a result, the prognostic ability of platelet-related indices has been investigated in many cancers. Platelet count (PLT) and morphologic indices, including the mean platelet volume (MPV), plateletocrit (PCT), and platelet distribution width (PDW), have been frequently studied in a number of cancers, including pancreatic cancer [11], invasive breast cancer [12], lung cancer [13], and esophageal cell carcinoma [14]. However, the relationship between platelet-related indices and prognosis in BTC remains unknown. Therefore, we investigated the prognostic performance of platelet indices in patients with TNM stages I–III BTC after surgical resection and found two potential prognostic factors. A new prognostic model and staging system were developed based on these two factors, which may benefit patients with BTC in surgery selection and postoperative management.

BTC is a type of entity tumor with distinctive characteristics with respect to high heterogeneity, poor diagnosis, and quick progression without diverse effective treatment. Radical surgery is the only acknowledged curative treatment, as the role of adjuvant chemotherapy in treating BTC remains to be confirmed. Consequently, a broader and more effective prognostic model is required for the prognostication of patients with BTC undergoing surgical resection. Therefore, we constructed and internally validated a new prognostic model to predict OS and to stratify patients with BTC into different stages after surgery. The model was presented as a nomogram based on six independent prognostic factors, including two platelet indices, nodal involvement, CA19-9, radical surgery, tumor differentiation, PDW/PCT, and MPV. By integrating tumor features, hemostatic parameters, and operation correlated factors, our novel staging system based on the model showed good discrimination and accuracy.

Recently, an increasing number of prognostic factors for cancer have been defined, including platelet indices. The platelet indices selected in the current study were indices depicting platelet morphology including PLT, PCT, MPV, and PDW, and the ratios in between, including PDW/PCT, PDW/PLT, MPV/PLT, and MPV/PCT. PLR, an indicator of inflammation, which is drawing increasing attention, was also included. Prognostic values of most of these indices have not been investigated for BTC. Univariate Cox analysis revealed that eight out of 10 indices were associated with OS in BTC, suggesting a marked impact of platelets on tumor progression. According to time-AUC curves of these eight indices (Fig. 1), we found that MPV and PDW/PCT have the highest potential to be independent prognostic factors of BTC, which was subsequently confirmed by the final multivariable Cox analysis. MPV, representing platelet size, has been demonstrated to be an independent predictor for OS in many cancers, and a high MPV has been reported as a poor indicator of OS in pancreatic cancer [11], invasive breast cancer [12], lung cancer [13] and esophageal carcinoma [14]. Our findings add to the accumulating evidence that high MPV is an independent negative predictor of OS. PCT as determined by MPV and PLT, reflects the number and size of platelets simultaneously. In our study, high PLT and high MPV were both observed in patients with poor survival, resulting in a reasonably high PCT. Furthermore, a previous study found that increased PCT levels were indicative of poor prognosis in metastatic colorectal cancer [15]. Although the relationship between PCT and OS in cancers remains to be investigated, based on the evidence that MPV and thrombocytosis are associated with poor survival in many cancers [16], we can speculate that PCT has the potential to be a good predictor of tumor progression. The role of PDW in predicting OS is controversial. Low PDW was found to be correlated with poor OS in non-small cell lung cancer [17], colorectal cancer [18], and gastric cancer [19]. The mechanism underlying high PDW with poor tumor stage remains to be elucidated. An emerging notion is that the interplay between PDW and MPV should be considered when evaluating their prognostic performance. Overall, the correlation between PDW and tumor progression remains to be investigated, and whether the underlying mechanisms are unique for heterogeneous tumors is unknown. In our study, by combining the PDW and PCT. PDW/PCT showed better performance in terms of predicting prognosis than PDW or PCT alone, while the prognostic performance of PDW/PCT has not yet been revealed in cancers.

An increasing body of evidence has highlighted the role of platelets in various stages of tumorigenesis, including tumor growth, invasion, metastasis, and chemotherapy resistance. The interaction between platelets and tumor cells is mutual, and recent studies have demonstrated that cancer can influence platelet’ s, physiology, and activation, even in the RNA profile; thus, providing potential markers for cancer prognosis [20]. Mechanisms of cancer-stimulated thrombocytosis have been elucidated, and primary tumors can influence megakaryopoiesis and thrombopoiesis by modulating a variety of cytokines directly or indirectly. Granulocyte colony-stimulating factor and macrophage colony-stimulating factor secreted by tumor cells can stimulate megakaryopoiesis [20], while elevated interleukin-6 released from cancer cells can increase the ploidy of megakaryocytic nuclei, resulting in an elevated platelet volume [21]. The ploidy of megakaryocytes has been observed to correlate with the MPV [22]. Moreover, platelets with a high MPV have been confirmed to be a potential indicator of platelet activation [23]. Larger platelets contain higher absolute amounts of proteins and more receptors on the surface, which enables them to interact with tumor cells [22]. For example, during the process of platelet activation, granules containing ADP released by platelets [20] interact with P2Y₁₂ receptors on platelets, triggering epithelial-mesenchymal transition, followed by invasion and metastasis [24]. Platelets with high MPV can also release more growth and prothrombotic factors, promoting angiogenesis and tumor progression [21]. Furthermore, activated platelets can crosslink, forming clots, and protecting circulative tumor cells against shear stress or immune attack [24]. In addition, increased platelet turnover is often observed in cancer patients, possibly arising from increased platelet consumption or sequestration. When platelet turnover increases, new platelets are produced with a larger size [22]. Whether larger platelets are more prothrombotic remains unclear. Evidence clarifying the role of small platelets in cancer-associated thrombosis is limited. Different mechanisms for platelets with different sizes interfering with cancer cells may exist, considering the diversity of cancer and the extensive functions of platelets. Lower PDW represents lower heterogeneity of platelets, which can be interpreted by a higher proportion of large new platelets under the stimulation of cancer cells. In contrast, platelets with a high MPV have more potential to interfere with cancer cells, assisting tumor progression in many stages. Based on these evidences and our findings, we propose that high MPV and low PDW/PCT are indicators of poor prognosis in BTC.

After integrating MPV and PDW/PCT into the multivariable Cox model, we developed a new prognostic model and a novel staging system. The OS of patients belonging to different stages were significantly different, either limited to specific types of BTC or the entire BTC cohort. Meanwhile, the AJCC 7th TNM staging system presented modest prognostic power in patients with ECC (Fig. 3H), and has been shown to have poor prognostic performance in several subsets of patients with BTC [25]. The C-index of our novel prognostic model was 0.703 in the training cohort and 0.728 in the validation cohort. Compared to the AJCC TNM stage, with a C-index of 0.561 and 0.618, our prognostic model has superior discrimination to the AJCC TNM stage. DCA also implied increased clinical benefits with the utility of our new staging system. Based on these results, we can conclude that our new prognostic model outperformed the AJCC TNM staging system in many aspects. An obvious advantage of our new prognostic model is the universal application to patients with BTC without the need to adjust the prognostic standard for any specific type. The AJCC TNM staging system is applicable for all subclasses of BTC, whereas the staging standards are not uniform with respect to different types of BTC. For example, T stage in GBC represents the infiltration depth of the tumor, while in ECC, T stage depends on the extent of invasion. In contrast, our model performed well in predicting the survival of patients with all types of BTC, which allows for greater convenience and wider application in comparison to the TNM staging system. Prognostic models have been proposed for GBC [26] and ICC [27], respectively, while relatively few models have been developed for BTC independent of a tumor site. Schweitzer et al. [28] proposed a prognostic score for patients with BTC at the time of diagnosis before any intervention. Bridgewater et al. [29] developed a model to predict OS in advanced BTC with limited predictive prognostic accuracy. A suitable method for survival prediction of patients with BTC undergoing resection has not yet been determined. Our novel staging system had extraordinary discriminative capability in predicting the OS of patients with BTC independent of the tumor site. The nomogram cannot only inform patients after surgery about their prognosis but also help to identify candidates for aggressive surgery among patients with BTC beforehand. With respect to the involvement of platelet indices, our new prognostic model can be applied to stratify patients in clinical trials intended to explore the efficacy of anti-platelet therapy, since post-diagnosis aspirin use has been reported to be associated with prolonged BTC survival [30].

One thing needs to be explained is the relatively high proportion of non-radical surgery (38.3%). This could be attributed to several reasons. Symptoms of biliary tumors were of insidious onset and most patients were found to be in advanced stage during the first medical visit. In our research, 31.3% patients were classified as TNM stage III with diverse severity of perinephric tissue invasion which increased the difficulty of surgery. Another situation was that preoperative assessment underestimate the range of tumor invasion where only radical surgery had the chance to remove all lesions. However, due to patient intolerance or family disagreement, non-radical surgery was performed instead of radical surgery. Another reason was that surgeries were performed between December 2002 and December 2017, intraoperative frozen-section was not routinely performed until recent years. From 2002 to 2012, 111 patients out of 213 (52.1%) received radical surgery while from 2013 to 2017, 168 patients out of 245 (68.6%) received radical surgery. The development of new technology elevated the probability of radical surgery.

Our study has several limitations. First, our data need to be completed, as detailed features of tumors, including invasion or local extension, are absent. However, the prognostic ability of perineural invasion and papillary tumors are inconsistent in BTC due to correlation with other factors [6], the absence of these data probably had a slight influence on the outcome. Second, our study was a single-center retrospective study with inevitable selection bias; multi-center and prospective studies are needed before generalization to a broader population. Third, the lack of external validation may decrease the credibility of the model, although we performed internal validation and obtained a high C-index value in the validation cohort.

To the best of our knowledge, our study is the first to systematically evaluate platelet indices as prognostic factors for BTC. We found that MPV and PDW/PCT are both independent prognostic factors for OS in patients with BTC. Subsequently, a new prognostic model and novel staging system were proposed for patients with BTC undergoing surgical resection, whose prognostic ability outperformed TNM staging in many aspects.