Gastric cancer remains one of the deadliest cancers worldwide, mainly due to its nonspecific clinical presentation and advanced stage at diagnosis [4]. The improvement of surgical technique and tailored systemic treatment increased overall survival and is presently considered the only potentially curative treatment [1, 3]. Disease-free survival (DFS) in locally advanced GC, estimated to be 30–60%, confirms that a high percentage of patients treated with curative intent end up with recurrence of disease [9]. A key step to achieve long-term survival seems to be related to early diagnosis, given the aggressiveness of GC. This fact has been confirmed by the improved survival in Asian countries where population screening is implemented [4]. Attempting to find a cheap, practical and non-invasive method of screening for poor outcome at the time of diagnosis would allow to select patients who might benefit from a more aggressive, or different, type of treatment [13, 16, 17].

Some studies have tried to establish a relationship between serum and ascitic TMs and their role in different settings. They have been compared in terms of discrimination of benign and malignant causes of ascites, and their usefulness as predictors of the course of disease after curative surgery [11, 13, 18]. Shibata et al. reported that after curative surgery, when comparing CEA and CA 19.9, serum CA 19.9 showed a higher predictive value for recurrence of disease [9]. When tested in ascites, Du et al. found that CEA, CA 15.3, and CA 19.9 predicted PC with 94.6% accuracy [12]. Their isolated value could also be optimized when used in combination according to other series [11].

In the presented study, the populations’ demography followed the global epidemiologic data on GC, with an increased incidence of 65.8% in male patients and a median age at diagnosis of 70 years. In western countries with no screening, GC is a disease of the elderly, with more than 90% of the diagnosis being made in patients with over 55 years of age and in advanced stages of disease, consistent with our own data [19].

When checked for PC and mortality after 1 year, cT status proved to correlate significantly (p = 0.032), thus emphasizing the importance of an early diagnosis in order to increase DFS[4]. cN status on the other hand did not significantly correlate with either outcome (p = 0.177 for PC; p = 0.239 for mortality).

At the time of SL, serum and peritoneal TMs were measured. A significant correlation between both samples of the same TM was found for CA 19.9 and CA 72.4. However, the serum and peritoneal concentration of the other TMs did not correlate significantly which makes the peritoneal washing analysis not replaceable by the serum assay. To date, there have been conflicting data regarding this relationship, given that most of the studies were conducted with small population sizes. Tuzun et al. managed to correlate significantly serum and peritoneal TMs in patients with malignant ascites [18]. On the other hand, another comparative study concluded that peritoneal TMs were of increased value in terms of sensitivity in determining malignant ascites [11].

A ROC curve analysis was conducted in order to determine the predictive power of peritoneal TM assays. According to our data, peritoneal CA 125 above 107.6 U/mL has high yield for the prediction of PC (p = 0.019). Also, CEA was able to independently predict CP when above 2.0 ng/mL (p = 0.020). In similar studies, Yang et al. measured CEA in GC-associated malignant ascites and concluded that for values above 2.3 ng/mL, it had diagnostic value for malignant vs. benign ascites [16]. Another study by Taobo et al. concluded that serum CA 125 was significantly higher in patients with peritoneal metastasis in GC and that its measurement is useful in predicting curability [3]. Moreover, serum CEA has also been proven to be an independent risk factor for poor prognosis [20]. CA125, although primarily used in ovarian cancer, is frequently positive in cases of peritoneal recurrence, thus being considered an independent predictor of poor outcome [3].

Regarding PC, positive, and negative predictive values were calculated and, although both TMs had low PPVs (CEA: 52.8%, CA 125: 53.4, CA 125 + CEA: 52.7%), the NPV was significant for both, thus establishing that below the determined cutoffs, a GC patient is not likely to have disease progression in the form of PC within a year of the SL. In this setting, combining both TMs could provide a higher yield for a positive outcome (NPV of CA 125 + CEA: 92.8%).

Choosing to associate the peritoneal washing harvest for TM measurement to the cytology during SL makes this analysis cost effective, reproducible, and does not add any other invasive procedure to the patients’ treatment. In fact, doing so before starting any type of treatment could be another useful tool to select patients with a predictable, more aggressive, course of disease, and potentially tailor treatment options such as extending indications for hyperthermic intraperitoneal chemotherapy to patients with negative peritoneal washing cytology.

The presented study has, however, some limitations. Although widely used for the follow-up of digestive tract malignancies, CEA production varies according to cancer location, histologic subtypes in different disease stages and its concentration can be influenced by non-malignant conditions and inflammation-inducing external factors such as smoking [21]. Despite the statistically significant predictive value obtained, tumor markers have low diagnostic yield, and sensitivity. CEA and CA125 are not gastric cancer-specific, nor have a linear variation with PC, which can limit their widespread use as a predictive tool. This is supported by the fact that not all patients who developed PC after one year had TM measurements above the upper limit of the reference range at the time of the staging laparoscopy. Nevertheless, their positive predictive value in the peritoneal washing reached 70% in similar studies, which in turn makes them useful when above the determined threshold [16]. The analysis was made with a pool of patients of a single center and, due to the aggressiveness of GC and late diagnosis, the study’s population was reduced. Secondly, the patients were not all submitted to the same systemic treatment and the response to treatment was not taken into account. Also, some of the patients had disease progression and were not submitted to curative surgery. Further studies with larger sample sizes should be conducted to provide further information on TMs and their predictive value.