Combining CRP and CA19-9 in a novel prognostic score in pancreatic ductal adenocarcinoma

Inflammation promotes tumor progression, induces invasion and metastatic spread. This retrospective study explored CRP, CA19-9, and routine laboratory values as preoperative prognostic factors in pancreatic cancer patients. Between 2000 and 2016, there were 212 surgically treated pancreatic cancer patients at Helsinki University Hospital, Finland. Out of these, 76 borderline resectable patients were treated with neoadjuvant therapy (NAT); 136 upfront resected patients were matched for age and sex at a 1:2 ratio. We analyzed preoperative CRP, CA19-9, CEA, leukocytes, albumin, bilirubin and platelets. CRP and CA19-9 were combined into a prognostic score: both CRP and CA19-9 below the cut-off values (3 mg/l and 37 kU/l, respectively), either CRP or CA19-9 above the cut-off value, and finally, both CRP and CA19-9 above the cut-off values. Among all patients, median disease-specific survival times were 54, 27 and 16 months, respectively (p < 0.001). At 5 years, among patients with CRP and CA19-9 levels below the cut-off values, 49% were alive and 45% were disease-free. Among NAT patients the corresponding survival rates were 52% and 45% and among those undergoing upfront surgery 45% and 40%, respectively. This novel prognostic score combining CRP and CA19-9 serves as a useful preoperative tool estimating survival.

Neoadjuvant therapy. NAT consisted of FOLFIRINOX, gemcitabine alone or in combination with capecitabine, cisplatin or nab-paclitaxel. Additional radiotherapy was administered to 31 (35%) patients. NAT was administered to borderline resectable patients, defined as superior mesenteric vein, portal vein or superior mesenteric artery contact upon diagnosis. Supplementary Table 1 describes in detail the treatment regimens along with the administered adjuvant therapy regimens.
High-sensitivity CRP and other laboratory values. CRP was determined from preoperatively collected plasma samples (n = 212). Samples were stored at − 80 °C until assayed at our research laboratory at the University of Helsinki. A high-sensitivity CRP method was used given its ability to measure lower levels of circulating CRP than possible with standard methods. A monoclonal antibody (anti-hCRP, code 6405, Medix Biochemica, Espoo, Finland) was used to capture CRP and as a tracer in a sandwich assay, a method described in more detail before 17 . Other preoperative laboratory values consisted of routine laboratory tests analyzed at the clinical laboratory of Helsinki University Hospital: albumin (g/l, n = 211), CA19-9 (kU/l, n = 212), CEA (μg/l, n = 210), leukocytes (E9/l, n = 212), platelets (E9/l, n = 212) and bilirubin (μmol/l, n = 211). Samples for CRP (96%, n = 204) and routine laboratory tests (97%, n = 206) were primarily collected 1-4 days preoperatively. Tumor marker levels were in 90% of patients (n = 191) determined 1-7 days preoperatively (range 1-60 days). Among NAT patients, all samples were collected following NAT. Table 1 lists the cut-off values along with a comparison of the median values between NAT patients and those undergoing upfront surgery. For laboratory values, we used the manufacturer's recommended cut-off values for CEA, platelets, bilirubin, and leukocytes, for CA19-9 based on the literature 26 , and for albumin based on the Glasgow prognostic score (GPS) 27 . For GPS and the modified Glasgow prognostic score (mGPS) 28 , cut-off values were based on the literature. The cut-off value for CRP (3 mg/l) was based on the cut-off value used in Finland. We explored higher cut-off values for CRP, finding that they did not differentiate survivors as well as using 3 mg/l as the cut-off value (data not shown).

Statistics.
We tested the distribution of continuous variables for normality using the Kolmogorov-Smirnov test. To compare laboratory values between groups, the Mann-Whitney U and Jonckheere-Terpstra tests were used for continuous variables. The Fisher's exact test and linear-by-linear association were used for categorical variables. We estimated survival using the Kaplan-Meier method (log rank). Primary end-point was diseasespecific survival (DSS) which was determined from surgery to cancer-specific death. Secondary end-point was disease-free survival (DFS), which was determined from surgery to the first recorded disease progression or death from PDAC. We calculated multivariate analyses using the Cox proportional hazards method: age, sex, stage, tumor grade, adjuvant therapy, perivascular invasion, resection margin status, albumin, and CRP and CA19-9 were considered clinically relevant and included in the model. Variables were examined for possible interactions. For each variable, the assumption of a constant proportional hazard rate over time was tested using a time-dependent variable; all variables except tumor grade met the assumption. Thus, the model was stratified by tumor grade. Additionally, the area under the receiver operating characteristic (ROC) curve at 5 years postoperatively was calculated. Patients with missing data were censored from survival analyses. The end of follow-up was January 14, 2020, with a minimum follow-up of 3 years or until death. We calculated all statistical analyses with SPSS (version 24, IBM, New York, NY, USA), and considered p < 0.05 as statistically significant. All tests were two-tailed. Table 1. Cut-off values with the number of patients and comparison of median values between NAT patients and those undergoing upfront surgery. We used the manufacturer's recommended cut-off values for CRP, CEA, platelets, bilirubin, and leukocytes, and for CA19-9 based on the literature 26 and for albumin as in the Glasgow prognostic score 27 . For the Glasgow prognostic score (GPS) 27 and the modified Glasgow prognostic score (mGPS) 28

Results
After exclusions, 76 patients were treated with NAT and subsequent surgery and 136 patients underwent upfront surgery (total n = 212). The median follow-up time was 2.2 years. Table 2 shows the comparison of clinicopathological characteristics between NAT patients and those undergoing upfront surgery.

Discussion
In this retrospective study, we demonstrate that combining CRP and CA19-9 in a prognostic score predicts postoperative outcomes in PDAC patients. A 5-year survival rate of 20-25% has been previously reported among resected PDAC patients 2,3 . However, in the present study, we identified a subgroup of NAT patients with a 5-year survival rate of 52% and a subgroup of patients undergoing upfront surgery with 5-year survival rate of 44%. These subgroups were characterized by low CRP and CA19-9 levels irrespective of disease stage and tumor differentiation. Additionally, combining CRP and CA19-9 differentiated patients with a poor survival. To the best of our knowledge, this represents the first study combining CRP and CA19-9 levels into a prognostic score. The primary advantages of this prognostic score are the preoperative availability of CRP and CA19-9, in addition to their affordable price. CRP and CA19-9 represent routine markers and no other marker with this level of accessibility has previously been reported to predict survival like the combination of CRP and CA19-9 in this study.
Resulting from an inflammatory state, the liver increases CRP production 16 ; elevated CRP levels reflect the systemic inflammatory response. Cancer is known to cause an inflammatory response, which associates with factors affecting survival, such as cachexia, fatigue, malnutrition, and a slower clearance of anticancer drugs and, thus, higher treatment-related toxicity 12 . Additionally, high preoperative CRP and CA19-9 levels could indicate micrometastatic disease, predict inadequate postoperative recovery, and interruptions or poor delivery of adjuvant therapy, resulting in a poor prognosis. Therefore, combining CRP and CA19-9 was logical and reasonable. Moreover, NAT appears to affect the tumor and its microenvironment, and the cancer-related inflammatory response 25 . Based on these aspects we hypothesized that the examined markers impact the survival of NAT patients and those undergoing upfront surgery differently. The results of this study support our hypothesis. Both CRP alone and CRP and CA19-9 combined differentiated long-term survivors better among NAT patients than those undergoing upfront surgery.
Many previous reports indicated the prognostic value of CRP in advanced PDAC patients receiving palliative treatment with cut-off values ranging from 1 to 20 mg/l 18,19,[21][22][23][24] . Similarly, CRP levels have correlated with tumor burden, indicating the aggressiveness of advanced pancreatic cancer 25 . In resected PDAC, without a separate analysis of NAT patients, high CRP associated with a poor prognosis: a median survival of 21 (CRP ≤ 3 mg/l) versus 13 months in 51 patients (CRP > 3 mg/l) 20 and median survival times of 32 months (CRP ≤ 5 mg/l), 24 months (CRP > 5-15 mg/l) and 14 months (CRP > 15 mg/l) 17 in 256 patients were reported. In another study on 474 patients (27% resected), a low CRP (< 4.5 mg/l) associated with a good prognosis 22 . Our results agree with these findings. Additionally, we managed to demonstrate the prognostic value of CRP in NAT patients and compare NAT patients to patients who underwent upfront surgery in a quite large cohort of resected patients. In contrast, Pine et al. 19 found that a high CRP (> 5 mg/l) did not associate with outcome in resected patients (n = 58) and Garcea et al. 29 found no correlation between CRP levels and disease recurrence (n = 74).
With our high-sensitivity method, we successfully explored the effect of low-grade inflammation on PDAC survival with slightly elevated levels of CRP. The results obtained by our high-sensitivity CRP method are comparable with standard CRP assays, however, the high-sensitive method is able to measure levels of even < 1 mg/l. In our study, we were able to show a correlation between slightly elevated CRP levels-as a marker of low-grade inflammation-and survival. In fact, among NAT patients, the median CRP of patients with disease recurrence within 12 months (4.4 mg/l) was only slightly higher than the reference limit (3 mg/l) routinely used in Finnish laboratories. Furthermore, in GPS and mGPS a CRP level of 4.4 mg/l would represent a low value. We must note that only one NAT patient with CRP ≥ 3 mg/l survived for more than 5 years, thereby emphasizing the role of inflammation in cancer survival. The CRP level of this patient was 3.93 mg/l. Interestingly, this agrees with previous results showing a correlation between low-grade inflammation and an increased risk of cardiovascular risk death 30 .
Different prognostic scores combining CRP and albumin (GPS, mGPS, and the CRP-to-albumin ratio), have been shown to predict survival in PDAC [31][32][33][34][35] . A median survival of 17 months has been reported for a GPS score of 2, 26 months for a score of 1, and 28 months for a score of 0; for mGPS, median survival reached 17, 28, and 26 months, respectively, for the scores of 2, 1, 0 35 . Our simple prognostic score combining CRP and CA19-9 proved superior to these prognostic scores, with a median survival of up to 54 months. In our study, a low preoperative albumin level associated with poor postoperative survival among NAT patients but did not reach statistical significance among patients who underwent upfront surgery. Combining albumin to our prognostic score along with CRP and CA19-9 did not prove feasible nor improve our results.
Previous PDAC studies on CA19-9 demonstrated that patients with low levels of CA19-9 (< 37 kU/l) had a longer median survival (32-36 months) than those with high levels (> 37 kU/l, 12-15 months) 36 . Additionally, a preoperative CA19-9 > 100 kU/l predicted early recurrence (at 6 months) and overall a poor postoperative prognosis 37 . These findings agree with ours, although in our study both NAT patients and those undergoing upfront surgery with CA19-9 ≤ 37 kU/l exhibited a longer median survival (NAT 43 months; upfront surgery 46 months).
Because 5-10% of the Caucasian population cannot synthesize CA19-9, we could not determine whether patients with CA19-9 ≤ 37 kU/l had low values resulting from the cancer biology or due to Lewis antigen-negativity. This represents one limitation of CA19-9 and thereby of our study. However, patients with undetectable www.nature.com/scientificreports/ CA19-9 (levels below 2 kU/l) appear to experience a similar survival to those with levels ≤ 37 kU/l, and a better prognosis than those with elevated levels 38 . Other limitations to our study include the lack of high-sensitivity CRP determined before NAT, the limited number of NAT patients, and the lack of a validation cohort. The fact that only operated patients were included in the study might cause bias, however, the main purpose of the study was to find preoperative prognostic factors. We additionally acknowledge the fact that the retrospective nature of the study might cause selection bias. However, the strengths of our study include the ability to compare NAT patients to those undergoing upfront surgery and to determine low levels of CRP using the high-sensitivity CRP assay. The main strength is that we could demonstrate the positive prognostic value of a low CRP level in addition to the negative prognostic value of an elevated CRP level and that we were able to introduce a new prognostic score for PDAC. The results of this study show that the combination of two preoperative prognostic biomarkers provides an essential improvement in the preoperative evaluation of the clinical outcome of PDAC patients. Interestingly, CA19-9 is a typical PDAC biomarker whereas CRP is thought to reflect the patient's systemic reaction to the tumor. Thus, the prognosis of PDAC appears to rely on both tumor-and patient-related factors. Even a low-grade systemic inflammation reflected by CRP had a significant impact on patient prognosis. The wide availability and low cost of these laboratory tests render them valuable in the preoperative evaluation of PDAC patients. Considering the overall poor prognosis of PDAC, even among those resected, the extraordinary prognosis of patients with a low CRP in combination with a low CA19-9 should be noted. These markers have not been studied together before, making this finding notable. Investigating CRP and CA19-9 on operated NAT patients is especially interesting, adding new valuable information to pancreatic cancer research. This new and simple prognostic score deserves to be further validated in other patient cohorts.

Data availability
The data analyzed and generated within the study are not publicly available because they contain data that have not been published as such yet.