Prognostic value of post-operative serum procalcitonin in gastric adenocarcinoma patients undergoing radical gastrectomy: propensity score matching analysis of extended cohort from a prospective bi-center study

Background The aim of this study was to investigate the predictive value of procalcitonin (PCT) on post-operative day (POD) 3 and 5 for the prognosis of gastric adenocarcinoma (GA) patients who underwent radical gastrectomy surgery in extended cohort from a prospective bi-center study. Methods Consecutive GA patients who received surgery in the Hunan Cancer Hospital were enrolled as the training cohort, and those from Wuhan Union Hospital were included as external validation cohort. The optimal cutoff concentration of PCT for overall survival (OS) in the training cohort was determined by X-tile. The independent predictive factors for OS were identified using univariate and multivariate Cox regression analyses. Furthermore, the predictive value of elevated PCT was clarified in the validation cohort and propensity score matched cohort, respectively. Results The optimal cutoff concentrations of PCT for OS were 0.67 ng/mL at POD 3 and 0.39 ng/mL at POD 5 in the training cohort (n = 906). Patients with higher PCT concentrations (≥ 0.39 ng/mL) at POD 5 had a significantly worse prognosis whether developing post-operative infections or not. Moreover, a synergistic influence was confirmed in those with elevated PCT concentration and infections. Multivariate analyses confirmed that PCT concentration ≥ 0.39 ng/mL at POD 5 was significantly associated with poorer survival in training cohort (HR: 1.422, 95% CI 1.041–1.943, P = 0.027), validation cohort (n = 297, HR: 2.136, 95% CI 1.073–4.252, P = 0.031) and matched cohort (n = 901, HR: 1.454, 95% CI 1.104–1.914, P = 0.008), separately. Conclusions PCT concentration ≥ 0.39 ng/mL at POD 5 was a reliable predictor for poorer prognosis in GA patients undergoing radical gastrectomy. Supplementary Information The online version contains supplementary material available at 10.1007/s10120-023-01422-0.


Introduction
Gastric adenocarcinoma (GA), with an estimated incidence of over one million new cases in 2020 worldwide, is one of the most common malignancies, with gastrectomy offering the only possible curative treatment to date [1,2].Unfortunately, the long-term outcomes of GA patients are still dismal given that only about 30% of patients are diagnosed at an early stage in the West and in China.Although the pathological tumor, node and metastasis (pTNM) stage have been the most widely used factors to predict the prognosis of malignancies, there has been increasing evidence that nutritional and inflammatory statuses are significantly associated with the oncological outcomes of numerous cancers.Several systemic inflammation-based variables have been confirmed to influence survival of GA patients, such as lymphocyte count, C-reactive protein (CRP), the neutrophil to lymphocyte ratio (NLR), Glasgow Prognostic Score (GPS), prognostic nutritional index (PNI), etc. [3][4][5][6][7].
Procalcitonin (PCT), the precursor of calcitonin consisting of 116 amino acids, has been well established as an inflammatory indicator [8][9][10].In order to clarify the predictive value of PCT for post-operative infection following gastrectomy in GA patients, we conducted a prospective study in two high-volume tertiary hospitals in China from June 2018 to July 2019 (ClinicalTrails.govnumber: NCT03780439).A total of 552 patients were enrolled in the study and our research confirmed that PCT was a more accurate marker than white blood and neutrophil counts to indicate infection [11].Since then, PCT concentrations have been routinely measured at post-operative day (POD) 3 and POD 5 in patients undergoing gastrectomy in our centers.
In recent years, more and more studies have reported that various types of cancers secret PCT and that PCT is a reliable predictor for long term survival, including for medullary thyroid cancer, colorectal cancer, non-small cell lung cancer (NSCLC), and others [12][13][14][15].But whether the serum concentration of PCT was related to the oncological outcomes of GA patients has never been investigated.Therefore, for the first time, we conducted this study to clarify whether the serum PCT concentration could predict the prognosis of GA patients following radical gastrectomy, using retrospectively collected data obtained from a large cohort of patients in two tertiary hospitals in China.

Study design
The medical records of consecutive patients with pathologically confirmed GA who underwent operations in Hunan Cancer Hospital from June 2018 to November 2020 and in Wuhan Union Hospital from June 2018 to November 2021 were reviewed.The demographic and clinicopathological data, operative variables and follow-up data were prospectively collected and retrospectively analyzed.The inclusion and exclusion criteria for this study and the flowchart of patients are shown in Fig. 1.Patients from Hunan Cancer Hospital were enrolled into the training cohort (Fig. 1A), whereas those from Wuhan Union Hospital following the same inclusion and exclusion criteria were included as the validation cohort (Fig. 1B).This study was approved by the ethics committee of our institutions (No. 41 in 2023) and written informed consent was obtained from all enrolled patients.

Peri-operative management and follow-up
Surgeons with sufficient experience performed or supervised all operations in line with the guidelines and stages according to the eighth edition of the TNM system [16][17][18].A second-or third-generation cephalosporin was given as a prophylactic antibiotic lasting for about 72 h to all patients.From June 2018, the serum PCT concentration has been routinely measured at POD 3 and 5 in patients who underwent gastrectomy for GA in our centers, using an automatic electrochemiluminescence immunoassay analyzer as described in our previous study [11].The normal reference range of PCT concentrations was < 0.050 ng/mL.The prognostic nutritional index (PNI) was calculated as serum albumin level (g/L) + 0.005 × total lymphocyte count in the peripheral blood (per mm 3 ) [6].In line with previous studies [19,20], a 1 3 PNI value of at least 50 was defined as normal, with values between 40 and 50 regarded as mild-moderate malnutrition, and values lower than 40 as serious malnutrition.Neutrophil-to-lymphocyte ratio (NLR) was calculated as dividing neutrophil count by lymphocyte count in the peripheral blood.
Post-operative complications were diagnosed and staged according to the Clavien-Dindo system and only grade II or greater morbidities were analyzed [21].Post-operative infections, such as intra-abdominal infection, pulmonary infection, wound infection, sepsis, urinary tract infection and catheter-related infections, was diagnosed according to the definition of the Centers for Disease Control and Prevention [22] and briefly described in our previous study [11].Patients with pTNM stage II/III GC were recommended to receive adjuvant chemotherapy about 4 weeks after surgery and were treated with fluorouracil-and platinum-based regimens, such as XELOX and SOX combinations [23][24][25].
Patients were followed-up every 3 months by outpatient review or telephone in the first 2 years and then every 6 months from the 3rd to 5th year and once a year thereafter, until February 2023.Overall survival (OS) was calculated as the month from surgery to death or the last follow-up time, whichever occurred first.

Statistical analysis
A χ 2 test, Fisher's exact test or a Student's t-test were employed to evaluate potential differences between categorical and continuous variables when appropriate.The optimal cut-off concentration of PCT for OS was affirmed by X-tile (3.6.1 software 20, http:// medic ine.yale.edu/ lab/ rimm/ resea rch/ softw are.aspx) when it reached the maximum χ2 logrank value, as described in previous studies [4,6].Potential predictive factors for OS were first investigated using univariate analysis.Variables with a P-value < 0.05 in the univariate analysis were further enrolled into multivariate Cox regression analyses.Kaplan-Meier curves and log-rank tests were applied to compare the OS of patients in different groups.To minimize the potential influence of selective bias, patients with elevated PCT concentrations were matched to those with relatively low PCT concentration using a propensity score as previously described [26,27].Nearest neighbor matching with a caliper width of a 0.05 standard deviation was performed at a 1:3 ratio without replacement.Data analyses were performed using R version 4.0.1 and a correlated package or SPSS version 27.0.A two-tailed P-value < 0.05 was considered to be statistically significant.

Characteristics of patients
As shown in Table 1, 906 consecutive patients from the Hunan Cancer Hospital were enrolled into the training cohort and another 297 consecutive patients from the Wuhan Union Hospital were included as the validation cohort.Among the entire 1203 patients, the mean age was 57.59 years (range 22-84), with a mean body mass index (BMI) of 22.28 kg/m 2 (range 14.53-37.13),and a mean postoperative hospital stay of 10.35 days (range 4-87).More than half of the patients were male (64.6%), who underwent distal subtotal gastrectomy (72.4%) by a laparoscopic procedure (59.2%).In line with the 8th version of the staging   1, patients in the validation group had significantly lower ratios of comorbidities, lesser American Society of Anesthesiologists (ASA) scores, lower concentrations of pre-operative albumin and less intra-operative blood loss compared to the training group.But more patients with early-stage GA who underwent laparoscopic surgery, developed post-operative complications, had longer operation times and post-operative hospital stays (all P < 0.05).In addition, the PCT concentration at POD 3 was significantly higher in the validation group, but comparable at POD 5.

Predictors for OS in the training cohort
With a median follow-up time of 31.0 months (range, 4-58), neither the median recurrence free survival (RFS) time nor the median OS time were reached in the training cohort.A total of 251 cases (27.7%) of recurrence and 231 cases (25.5%) of death were recorded, with 1-, 2-, 3-and 4-year OS rates of 92.3%, 80.5%, 75.0% and 68.2%, respectively.The cut-off concentrations of PCT at POD 3 and POD 5 for OS were set at 0.67 ng/mL and 0.39 ng/mL by X-tile, when they reached the maximum χ 2 values of 8.8624 and 14.3769, with the minimum log-rank test P-value of 0.0604 and 0.0050, separately (Supplementary Fig. 1).Similarly, the best cutoff value of NLR was set as 2.84 by X-tile, when it reached the maximum χ 2 value of 17.4698 and minimum log-rank test P-value of 0.0012.As presented in Table 2, univariate analyses revealed that age, the hemoglobin and albumin concentrations, lymphocyte count, PNI, NLR, intra-operative blood loss, tumor stage, peri-operative blood transfusion, post-operative complication, PCT concentrations at POD 3 and POD 5 were potentially associated with OS (all P < 0.05).Further multivariate Cox regression analyses containing the above mentioned factors confirmed that older age (≥ 65 years), larger intra-operative blood loss (≥ 300 mL), advanced tumor stage, post-operative complication and higher PCT concentrations (≥ 0.39 ng/mL) at POD 5 were independent predictors for unsatisfactory oncological outcomes.

Association among PCT, infection and prognosis in the training cohort
To explore the association with PCT concentrations, postoperative infection and prognosis in the training cohort, we further classified the 906 patients into infection (n = 51, 5.6%) and non-infection (n = 855, 94.4%) subgroups.As shown in Fig. 2, non-infection patients with higher PCT concentrations at POD 5 were associated with significantly more unfavorable OS than those with lower PCT concentrations (P = 0.003).Similarly, patients with infections and higher PCT concentrations seemed to have worse prognosis, but the difference did not reach statistical significance (P = 0.062).
In addition, to investigate the potential additive impact of an increased PCT concentration and infection on survival, we further stratified the 906 patients into 4 subgroups based on whether the PCT concentration was ≥ 0.39 ng/mL at POD 5 or the patient was infected or not: PCT low/infection (−) (n = 737), PCT high/infection (−) (n = 118), PCT low/ infection ( +) (n = 38) and PCT high/infection ( +) (n = 13), respectively.Not surprisingly, patients with elevated PCT concentrations and infections had the worst oncological outcomes (Table 3 and Supplementary Fig. 2).A synergistic influence was confirmed in the PCT high/infection ( +) group comparing to the PCT high/infection (−) group (hazard ratio [HR]: 2.916, 95% confidence interval [CI] 1.460-5.824,P = 0.002).

Oncological outcomes in the validation cohort
Among the 297 patients in the validation group, 23 experienced recurrence (7.7%) and 43 (14.5%)died with a median follow-up of 25.0 months (range, 6-56).The 1-, 2-, 3-and 4-year OS rates were 98.3%, 93.1%, 80.9% and 70.1%, separately.Given the relatively short duration of the followup time, neither the median RFS time nor the median OS time could be determined.As presented in Supplementary Table 1, univariate analyses revealed that age, BMI, tumor stage and the PCT concentration at POD 3 and POD 5 were potentially predictors for OS (all P < 0.05).After using these 5 variables in further multivariate Cox regression analyses, a higher PCT concentration (≥ 0.39 ng/mL) at POD 5 along with an older age (≥ 65 years), lower BMI (< 18.5 kg/m 2 ) and advanced tumor stage were confirmed to be independently related to OS.

Propensity score matching (PSM) analysis
Propensity scores were conducted using variables such as age, BMI, intra-operative blood loss, post-operative complications and the TNM stage, which had been confirmed as significant predictors for OS in the training and/or validation group.The matched cohort was comprised of 901 patients after one-to-three PSM, with 250 patients in the PCT high concentration group (≥ 0.39 ng/mL at POD 5) and 651 in the PCT low concentration group (< 0.39 ng/mL at POD 5), respectively.As described in Table 4, the baseline characteristics such as gender, age, BMI, post-operative complications and TNM stage differed significantly before matching, but became comparable after matching.Further univariate and multivariate Cox regression analyses also confirmed that a PCT concentration ≥ 0.39 ng/mL at POD 5 was significantly associated with OS in the matched cohort (HR: 1.454, 95% CI 1.104-1.914,P = 0.008, Table 5).As shown in Supplementary Fig. 3, patients with elevated PCT concentrations at POD 5 had significantly worse oncological outcomes no matter before and after matching (P < 0.001 and P = 0.033, respectively).

Discussion
In this post hoc analysis with an extended cohort from our previously prospective study, for the first time it has been revealed that a higher serum PCT concentration (≥ 0.39 ng/ mL) at POD 5 was an independent predictor for unfavourable prognosis in stage I-III GA patients who underwent curative resection.Subgroup analyses confirmed that an increased PCT concentration significantly adversely impacted the prognosis in patients no matter whether they developed post-operative infections or not.In addition, a synergistic effect was shown given that patients with elevated PCT concentrations and suffering from infection had an even worse prognosis, compared to those with elevated PCT concentrations but non-infected (HR: 2.916, P = 0.002).In the external validation cohort of 297 patients, a PCT concentration ≥ 0.39 ng/mL at POD 5 was also confirmed to be significantly associated with a poorer OS.Further PSM analysis, which aimed to cancel out the potential influence of selective bias between patients with elevated PCT or not, also demonstrated that a PCT concentration ≥ 0.39 ng/mL at  POD 5 was an independent predictor for poorer survival in the matched 901 patients.Taken together, the present study revealed that an elevated PCT concentration at POD 5 significantly influenced the prognosis of GA patients undergoing radical gastrectomy.External validation and the matched cohort confirmed our findings, which offered statistical power to improve the reliability of the conclusions.
The PCT concentration is considered to be an accurate predictor for systemic inflammation, especially for the diagnosis of bacterial infections [8][9][10][11].In patients with bacterial infection, PCT levels increased 2-4 h following bacterial stimulus and peaked at 12-24 h, with a half-life about 24 h [28].But in patients receiving elective colorectal surgery, PCT levels reached peak at POD 2 and a subsequent decrease in the following 2 days [29].Meanwhile, infection and chronic inflammation has long been considered to play an essential role in the development of varied types of cancers and to be involved in promoting all stages of tumorigenesis, such as tumor growth, progression and metastasis [30].There is mounting evidence that the suppressed immunity status of cancer patients not only increases the incidence of post-operative complications, but also adversely impacts the long-term oncological outcomes.
In recent years, many studies have highlighted the predictive role of PCT on the prognosis of several cancers, such as medullary thyroid cancer, NSCLC, colorectal cancer, neuroendocrine neoplasms, and so on [12][13][14][15]31].In a prospective population-based study that enrolled 3322 patients [32], the baseline PCT was found to be significantly associated with all-cause mortality and cancer-related mortality in men after a median follow-up time of 16.2 years.In a retrospective study of 277 colorectal cancer patients who received surgery, the pre-operative PCT concentration was identified as a significant predictor for cancer-specific survival in stages I to III of the disease [15].In another retrospective study consisting of 95 metastatic NSCLC patients who received immune-check point inhibitors, a baseline PCT concentration of > 0.1 mg/L was confirmed to result in significantly worse outcomes in terms of OS [33].In contrast, Booka and colleagues [13] reported that in 105 esophageal cancer patients who underwent esophagestomy, a low PCT concentration (< 1 ng/mL) at POD 2 was an independent predictor of poorer OS.Possible explanations for the contradictory conclusions might be the inconsistency of patients inclusion criteria, varied PCT measurement times and the PCT cutoff concentration, relatively small sample sizes and the retrospective nature of the studies.
As far as we are aware, whether the PCT concentration is associated with the prognosis of GA patients has never been investigated.Therefore, we conducted the present study with a large sample size from an extended cohort of a prospective study.As described in Table 2, although lymphocyte count, NLR and PNI was considered as potential predictor for OS in univariate analysis in the training cohort, all of them lost their significance after multivariate analyses.In contrast, elevated PCT level at POD 5 was confirmed as an independent predictor for prognosis, which showed advantages comparing with other inflammation factors.In order to examine the generalizability of our findings, we further performed external validation and PSM analysis.
As shown in Table 1, the baseline characteristics differed significantly between the training and validation groups, including the values for intra-operative bleeding, tumor stage and the incidence of post-operative complications, which are well-established predictors for the prognosis of GA patients [25,34].External validation using the dataset with significantly different baseline characteristics ascertained the robust prediction power of PCT for OS.In addition, as described in Table 4, some important variables, such as BMI, tumor stage and post-operative complications, were significantly different among patients with elevated PCT concentrations or not before matching.In order to eliminate potential selective bias and to mimic a randomized trial, we further conducted a PSM analysis.After matching, all baseline characteristics were comparable between the two groups.Further univariate and multivariate Cox regression analyses also confirmed that an elevated PCT concentration at POD 5 was independently associated with OS in the matched cohort.Taken together, external validation and PSM analyses added additional statistical power to make our final conclusions more reliable.
Although few research groups have reported the relationship between the PCT concentration and prognosis of cancer patients, the underlying mechanism(s) has not been unequivocally clarified.Baseline PCT was generally secreted by thyroid C-cells and other neuroendocrine cells in the lung and bowel, and predominantly by the cells in the lung and bowel when inflammation occurred [12].In the hypoxic tumor micro-environment, pro-inflammatory factors such as tumor necrosis factor-α (TNF-α) and interleuin-6 (IL-6) were produced as a result of tissue damage and/or tumor necrosis, which might further activate the systemic inflammatory response [35].As a result, increased concentrations of serum PCT levels facilitated tumor metastasis and led to poor oncological outcomes [36].In addition, bacterial toxins including endotoxins could directly stimulate the production of PCT in patients with developing infections, which might also weaken the immune system and facilitate tumor progression [13].In the present study, a patient with a higher PCT concentration had a significantly worse prognosis no matter whether they developed an infection or not.Although the difference in the infection group was slightly outside a statistically significant level (P = 0.062), most likely due to the relatively small sample size (n = 51).Additionally, an additive effect was confirmed in patients with elevated PCT concentrations and suffering from an infection.Thus, it seemed that the adverse impact of the PCT concentration on prognosis was independent of infection.Therefore, further studies are still needed to investigate the exact underlying mechanisms.
The present study had several limitations.First, patients' pre-operative PCT concentrations were not measured.Thus, the association between the baseline PCT concentration and prognosis, and the impact of the dynamic change of the PCT concentration before and after surgery on prognosis could not be investigated.Second, the median follow-up time of 30 months for the entire cohort was insufficient to analyze later relapses and deaths of patients.Third, patients undergoing neo-adjuvant chemotherapy were excluded from the present study.As a result, whether our findings could be used in these patients needs further external validations.Last, but by no means least, given that different laboratory examination methods used to measure PCT concentrations, the PCT concentration might differ significantly among institutions and therefore it seemed inappropriate to cite our cutoff values directly.
In conclusion, for the first time, our study has revealed that an elevated PCT concentration (≥ 0.39 ng/mL) at POD 5 was a significant predictor for poorer prognosis in GA patients following radical gastrectomy.The extended cohort from a prospective study with a large sample-size of patients, the combination use of external validation and PSM analysis improved the reliability of our final conclusions.But further studies are still needed to clarify the exact underlying mechanisms.

Fig. 1
Fig. 1 Flow diagram of the training cohort A and the validation cohort B in the present study

Fig. 2
Fig. 2 Over survival curves of the 906 patients in the training cohort who underwent curative resection for stage I-III gastric cancer stratified by procalcitonin (PCT) level at post-operative day (POD) 5 (< 0.39 or ≥ 0.39 ng/mL) in the non-infection group A and infection group B

Table 1
Clinicopathological characteristics of the training and validation cohort of patients with stage I-III gastric cancer (n = 1203) Data are presented as mean ± SD or n (%).ASA, American Society of Anesthesiologists.POD, post-operative day * Tumor stages are based on 8th edition of the Union for International Cancer Control TNM classification † Defined as Clavien-Dindo grade II or greater.Patients who developed both infectious and non-infectious complications were classified into the infectious group , there were 377 (31.3%) stage I, 278 (23.1%) stage II and 548 (45.6%) stage III cases.The mean surgical time was 221.35 min (range, 55-640), and the mean intra-operative bleeding was 175.53 mL (range,.A total of 189 patients (15.7%) developed grade II or greater post-operative complications and there were 87 cases (7.2%) of infection.As shown in Table system

Table 2
Univariate and multivariate analyses of prognostic factors for overall survival following radical gastrectomy of gastric adenocarcinoma in the training cohort (n = 906) CI, confidence interval; HR, hazard ratio; UV, univariate analysis; MV, multivariate analysis † Tumor stages are based on 8th edition of AJCC TNM classification ‡ Defined as Clavien-Dindo grade II or greater

Table 3
Subgroup analysis of overall survival (OS) of stage I-III gastric adenocarcinoma following radical gastrectomy in the training cohort (n = 906)

Table 4
Clinicopathological characteristics of the entire study cohort stratified by procalcitonin (PCT) level at post-operative day 5, before and after propensity score matching Data are presented as mean ± SD or n (%) † Defined as Clavien-Dindo grade II or greater Tumor stages are based on 8th edition of the Union for International Cancer Control TNM classification *

Table 5
Univariate and multivariate analyses of prognostic factors for overall survival following radical gastrectomy of gastric adenocarcinoma in the propensity score matched cohort (n = 901)