2023 Volume 70 Issue 11 Pages 1069-1075
CA19-9 is a tumor marker for pancreatic cancer (PC), and the nondiabetic cut-off level is 37 U/mL. CA19-9 levels are said to rise in patients with tumors like PC and intraductal papillary mucinous neoplasm (IPMN). CA19-9 levels have also been shown to be related to HbA1c levels. We hypothesized that the CA19-9 cut-off levels would differ between patients with poorly controlled diabetes. This real-world trial was designed to test our hypotheses. This was a retrospective cohort study. All inpatients with poorly controlled diabetes had mean HbA1c levels of 10.0% and were divided into three groups: those with pancreatic cancer (PC group, N = 20), those with IPMN (IPMN group, N = 55), and those with neither (NC group, N = 985). Serum CA19-9 levels in the PC group were significantly higher than in the IPMN and NC groups (p < 0.001). CA19-9 levels did not differ statistically between the IPMN and NC groups. According to the receiver operating characteristic (ROC) analysis, serum CA19-9 levels of 98.4 U/mL had the highest sensitivity and specificity to detect PC, when comparing PC to IPMN + NC groups. Using this cut-off, the sensitivity and specificity of CA19-9 for PC were 70.0% and 96.5%, respectively, with a 0.81 area under the ROC curve. CA19-9 levels in two inpatients were >98.4 U/mL, most likely due to hepatocellular carcinoma and esophageal cancer. CA19-9 cut-off levels were thought to be 98.4 U/mL. However, we should keep in mind that the sensitivity and specificity were not 100%.
EARLY PANCREATIC CANCER (PC) is difficult to detect because patients with early PC rarely exhibit symptoms. PC is rarely curable and has a high mortality rate. The overall 5-year survival rate for patients with PC in Japan was only 10.2% [1].
Diabetes patients have a higher incidence of PC than non-diabetics [2-4], and the development of PC diseases is associated with elevated blood glucose levels [4, 5]. It is critical to determine whether patients with poorly controlled diabetes have PC.
Carbohydrate antigen 19-9 (CA19-9) was a cell surface glycoprotein complex derived from a colorectal carcinoma cell line that was overexpressed in PC and thus used as a tumor marker for PC [6]. CA19-9 levels were also found to be elevated in patients with cancers other than PC, such as cholangiocarcinoma, colorectal cancer, esophageal cancer, and hepatocellular carcinoma [7-12]. The CA19-9 cut-off levels for detecting PC without diabetes were set as 37 U/mL [7, 8].
Intraductal papillary mucinous neoplasm (IPMN) is a benign pancreatic cyst that can occasionally become malignant. Abdominal ultrasonography or plain abdominal computed tomography (CT) was frequently used to detect IPMN. CA19-9 levels were found to be elevated in some IPMN patients [13].
CA19-9 had a sugar chain attached to it, and CA19-9 levels were reportedly higher in diabetic patients than in non-diabetics [14]. CA19-9 levels have also been found to rise in poorly controlled diabetic patients [15]. A cross-sectional study revealed that CA19-9 levels were proportional to HbA1c levels [16]. Therefore, diabetic patients were thought to have higher CA19-9 cut-off values for detecting PC than nondiabetic individuals.
In a case-control study comparing diabetic patients with PC and those without PC, matched for age, gender, and PC risk factors, the CA19-9 cut-off values for the detection of PC was 75 U/mL [16]. However, the results of a case-control study do not reflect the prevalence of PC in the real world.
The purpose of this study was to compare patients with PC versus those without PC to determine the proposed CA19-9 cut-off values for the detection of pancreatic cancer in patients with poorly controlled diabetes. We also assessed the prevalence of malignancy other than PC in such uncontrolled diabetic inpatients.
This was a retrospective cohort study endorsed by the University Hospital Medical Information Network (UMIN) (UMIN ID: 000049985). All data was obtained from electronic records. This study followed the Declaration of Helsinki and was approved by the ethics committees of Saiseikai Yokohamashi Nanbu Hospital and Yokohama Sakae Kyosai Hospital. The approval numbers are 2019-D31 for the former and 20210419-2 for the latter. Informed consent was obtained through a website opt-out (https://www.nanbu.saiseikai.or.jp/media/tounai2019D31.pdf).
ParticipantsWe enrolled inpatients with diabetes ranging in age from 19 to 92 years in both hospitals from April 2015 to March 2018. All these people were hospitalized to learn how to manage diabetes and optimize their treatments. Besides, their CA19-9 levels were checked, and they had abdominal ultrasonography or a plain abdominal CT test during their hospitalization to screen for malignancies. All patients were divided into three groups: those with PC group, which included one patient with intraductal papillary mucinous cancer, those with IPMN group, and those with neither (NC group). PC was diagnosed based on the results of a biopsy or the progression of the disease. The abdominal symptoms and jaundice of patients in the IPMN and NC groups were examined six months after hospitalization to detect PC in the two groups. Patients with a history of PC and those with other malignancies who required treatment were excluded; however, those who did not require treatment were included.
Assessments of blood profiles and characteristicsTo assess glucose metabolism, all patients had overnight fasting blood samples drawn on the second day after admission. HbA1c levels were determined using high-performance liquid chromatography (Adams A1c HA-8160; Arkray Inc., Kyoto, Japan in Saiseikai Yokohamashi Nanbu Hospital and Adams A1c HA-8160 or Adams A1c HA-8181; Arkray, Inc., Kyoto, Japan in Yokohama Sakae Kyosai Hospital). Fasting plasma C-peptide and CA19-9 levels were measured in a central clinical laboratory (SRL, Inc., Tokyo, Japan). Baseline characteristics and other information such as age, gender, height, and weight, smoking habits were obtained from electronic medical records.
Statistical analysisWhen continuous variables followed a normal distribution, they were presented as mean ± SD and median [first quartile, third quartile], respectively. Age, body mass index, and HbA1c levels were compared between the three groups using analysis of variance. Sex and smoking habits were compared using the chi-square test, or Fisher’s exact test, depending on the data distribution. Fasting serum C-peptide immunoreactivity and CA19-9 levels were compared using the test among the three groups. The Spearman R correlation coefficient was used to examine the relationship between pancreatic adenocarcinoma stages and CA19-9 levels, as well as the associations between CA19-9 and HbA1c levels. Multiple regression analysis was used to determine the correlations of log CA19-9. A p-value <0.05 was deemed statistically significant. The optimal serum CA19-9 levels for predicting PC were determined using receiver operating characteristic (ROC) curves. The sensitivity and specificity of CA19-9 for the cut-off values were then calculated by maximizing the sum of sensitivity and specificity.
All statistical analyses were carried out using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan) and the graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria), a modified version of the R commander that includes with statistical functions commonly used in biostatistics [17].
From April 2015 to March 2018, 1,211 diabetic inpatients were assessed for eligibility. Among these, 151 were excluded for reasons such as not having had abdominal ultrasonography or an abdominal CT test, leaving 1,060 patients enrolled (Fig. 1). Table 1 shows the baseline characteristics; patients with PC and IPMN were older than those suffering neither of them (NC group). One of the patients was diagnosed with PC six months after discharge. Of 55 patients with IPMN, 52 underwent imaging follow-up and showed no development of PC; two of them refused imaging test and remaining one got missed. The serum CA19-9 levels in the PC group were significantly higher than those in the IPMN and NC groups (p < 0.001, respectively), whereas the IPMN group did not differ from the NC group (p = 0.84). There were no significant differences in the three groups regarding sex, BMI, smoking habit, HbA1c, or fasting serum CPR levels. The current study’s patients had a mean HbA1c level of 10.0%.
Flowchart of this study
Features of the patients
| PC group (N = 20) | IPMN group (N = 55) | NC group (N = 985) | p value | |
|---|---|---|---|---|
| Age (years) | 73.1 ± 7.8 | 75.3 ± 7.7 | 65.4 ± 13.6 | <0.001 |
| Sex (Male/Female) | 11/9 | 27/28 | 597/388 | 0.213 |
| BMI (kg/m2) | 23.5 ± 5.8 | 24.6 ± 4.3 | 25.4 ± 5.0 | 0.115 |
| Smoking (never/past/current) | 13/3/4 | 30/20/5 | 506/283/196 | 0.157 |
| FPG (mg/dL) | 164.0 ± 50.7 | 150.3 ± 55.8 | 160.9 ± 71.5 | 0.541 |
| HbA1c (%) | 10.9 ± 1.9 | 9.7 ± 1.8 | 9.9 ± 2.2 | 0.083 |
| sCPR (ng/mL) | 0.94 [0.60–1.99] | 1.50 [0.89–2.20] | 1.42 [0.90–2.11] | 0.136 |
| CA19-9 (U/mL) | 274.0 [32.6–1,470] | 17.9 [8.65–26.7] | 14.0 [7.0–27.5] | <0.001 |
Date are represented as median [first quartile, third quartile], means ± standard deviation, or numbers (percentages). BMI, body mass index. FPG, fasting plasma glucose. HbA1c, glycated hemoglobin. sCPR, serum C-peptide immunoreactivity. CA19-9, carbohydrate antigen 19-9.
PC group: patients with pancreatic cancer
IPMN group: patients with intraductal papillary-mucinous neoplasm
NC group: patients with neither pancreatic cancer and intraductal papillary-mucinous neoplasm
There was a subtle positive correlation between HbA1c and serum CA19-9 levels in the NC + IPMN group (r = 0.259, p < 0.001, Fig. 2).
The relationship between CA19-9 and HbA1c levels
CA19-9, carbohydrate antigen 19-9. HbA1c, glycated hemoglobin.
When the CA19-9 cut-off values and area under the curve (AUC) in the ROC curve were calculated, in comparison PC to NC + IPMN groups, CA19-9 levels of 98.4 U/mL predicted the possibility of pancreatic cancer with a sensitivity of 70.0% and a specificity of 96.5% (AUC 0.81 [95%CI: 0.66–0.96], Fig. 3).
Receiver operating characteristic (ROC) curves of CA19-9 levels for predicting pancreatic cancer
Plot: CA19-9 cut-off levels (specificity and sensitivity)
CA19-9, carbohydrate antigen 19-9.
Fourteen patients with PC were diagnosed with pancreatic adenocarcinoma and one patient with IPMC based on biopsy, and five patients were suspected of having pancreatic adenocarcinoma due to the progression of their diseases. PC operations were performed on nine patients. There was no invasion of other tissues in an IPMC patient. Of the 19 patients with pancreatic adenocarcinoma, 2/8/2/7 were in stages I/II/III/IV. CA19-9 levels showed increasing tendency with advanced stage of pancreatic adenocarcinoma (p = 0.08). All IPMN patients did not require surgery.
Eleven were diagnosed with cancers other than PCs out of a total of 1,060: three colon cancers, two thyroid cancers, one ovarian cancer, one hepatocellular carcinoma, one esophageal cancer, one cholangiocarcinoma, one lung cancer, and one bladder cancer. CA19-9 levels in two patients with hepatocellular carcinoma and esophageal cancer exceeded 98.4 U/mL. One patient with thyroid cancer demonstrated CA19-9 levels of 85.7 U/mL.
As shown in Table 2, regarding two CA19-9 cut-off levels (37.0 U/mL; used as usual, and 98.4 U/mL; proposed in this study), the number of patients with PC is displayed in Table 2. Out of 870 patients with CA19-9 levels less than 37.0 U/mL, six patients had PC. Among the 140 patients with CA19-9 levels ranging from 37.0 to 98.4 U/mL, there were no PC patients. Among 50 patients whose CA19-9 levels were over 98.4 U/mL, 14 had PC. Patients with PC having CA19-9 levels <37.0 U/mL showed lower HbA1c levels than those with CA19-9 levels ≥37.0 U/mL (9.5% ± 1.6% vs. 11.5% ± 1.8%, p = 0.032).
CA19-9 levels of the patients
| CA19-9 (U/mL) | CA19-9 < 37.0 | 37.0 ≤ CA19-9 < 98.4 | 98.4 ≤ CA19-9 |
|---|---|---|---|
| PC group (N) | 6 | 0 | 14 |
| IPMN group (N) | 47 | 6 | 2 |
| NC group (with other malignancy than PC) (N) | 8 | 1 | 2 |
| NC group (without any malignancies) (N) | 809 | 133 | 32 |
| Total | 870 | 140 | 50 |
CA19-9, carbohydrate antigen 19-9. PC, pancreatic cancer.
PC group: patients with pancreatic cancer
IPMN group: patients with intraductal papillary-mucinous neoplasm
NC group: patients with neither pancreatic cancer and intraductal papillary-mucinous neoplasm
This was the first trial to examine the CA19-9 cut-off level to determine PCs in patients with poorly controlled diabetes in the real world. CA19-9 cut-off levels were 37.0 U/mL in nondiabetic patients with sensitivity and specificity ranging from 70% to 90% [7, 8, 18, 19]. Whereas among patients with mean HbA1c levels ≥10%, the CA19-9 cut-off level was 98.4 U/mL with a sensitivity of 70.0% and specificity of 96.5% in the current study. Due to its high specificity, CA19-9 may be a useful biomarker for screening for PCs.
Murakami et al. [16] discovered PC with a sensitivity of 69.5% and specificity of 98.2% using a CA19-9 cut-off level of 75U/mL in diabetic patients with mean HbA1c levels of 9%. According to the findings of this study, CA19-9 can be used to screen for PC with high specificity. In the current real-world study, the CA19-9 cut-off levels of 98.4 U/mL, which was higher than Murakami’s study, were proposed presumably because patients in this study had worse glycemic controls than those in his report. CA19-9 levels were proportional to HbA1c levels in diabetic patients [9, 20], as shown in Fig. 2 of this study. In addition to the incidence of PC among correlates in Table 1 (p < 0.001), multiple regression analysis revealed that only HbA1c levels contributed to log CA19-9 values. Higher CA19-9 cut-off levels could mainly be because of higher HbA1c levels, though unexpected factors might be found as well.
CA19-9 levels of 37.0 U/mL were found to be the best cut-off levels for diagnosing PC in the absence of diabetes [7, 8, 18, 19]. When patients showed CA19-9 levels ranging from 37.0 to 98.4 U/mL (the proposed cut-off levels for diagnosing PC in the current paper) should be re-examined after patients’ glycemic controls improved. According to some studies, CA19-9 levels decreased as glycemic control improved [9, 21, 22]. If the patients’ CA19-9 levels do not decrease, they should undergo magnetic resonance imaging (MRI), which is thought to detect PCs better than CT [23].
As shown in Table 2, 0.6% of patients with CA19-9 levels less than 98.4 U/mL had PC. When CA19-9 levels were greater than 98.4 U/mL, 28.0% of patients developed PC.
A few PC patients had low CA19-9 levels, which was due in part to the fact that a few people with Lewis blood group-negative showed false-negative of the CA19-9 test [9], due to difficulty synthesizing CA19-9 [24]. In this study, six PC patients had CA19-9 levels of <37.0 U/mL. Some of them may be Lewis blood group-negative.
In this real-world trial, 11 out of a total of 1,060 patients were diagnosed with other cancers than PC. CA19-9 levels exceeded 37.0 U/mL in 3 of 11 patients with thyroid cancer, hepatocellular carcinoma, or esophageal cancer. Patients with hepatocellular carcinoma and esophageal cancer have been reported to have elevated CA19-9 levels [7, 11, 12], and in the current study, patients with both of these cancers had CA19-9 levels greater than 98.4 U/mL. To our knowledge, few papers have examined the links between CA19-9 levels and thyroid cancer patients. These three patients may have a minor impact on our results because of their small number. We re-examined the CA19-9 cut-off levels for PC after excluding these three patients; the CA19-9 cut-off levels, sensitivity, and specificity remained unchanged (data not shown).
Some IPMN patients reportedly had elevated CA19-9 levels [13]; however, in the current study, there was no statistical difference in CA19-9 levels between the IPMN and NC groups, as shown in Table 1. CA19-9 cut-off values were considered appropriate to compare the PC group with the IPMN and NC groups.
History of diabetes (i.e., long-term exposure to high glucose and/or hyperinsulinemia) may be a risk factor for the development of PC [25-27]. Among the 957 cases with known disease durations, the PC group tended to have longer duration of diabetes than other groups, although the difference was not significant (10.7 ± 10.5 years vs. 6.2 ± 8.3 years (p = 0.055)).
This study had several limitations. First, screening for PC was conducted using abdominal ultrasonography or plain abdominal CT, which had lower sensitivity and specificity than MRI and endoscopic ultrasonography (EUS) [23]. We believe that some patients with PCs may have been missed by either of these imaging tests although we investigated the abdominal symptoms and jaundices of patients without PCs half a year after hospitalization. Second, the current study included 20 patients with PC. Selection bias should be considered because of the small number of patients with PC. Additionally, 12.5% of all patients were excluded due to missing data, which may have contributed to selection bias. However, our proposal of CA19-9 cut-off levels should be suitable because this was a multicenter trial and the cut-off level was similar to Murakami’s report. Third, this study did not investigate how many patients were negative for Lewis blood group. The CA19-9 cut-off values for PC might be affected by the proportion of patients with negative Lewis blood group. Fourth, some studies revealed an increased risk of PC among heavy alcohol drinkers [25, 28-30]. However, in this study, the amount of alcohol consumed was not checked. Finally, the CA19-9 levels increased as the PC stages progressed [31]. In the current study, of the 19 patients with PC, 2/8/2/7 were in stages I/II/III/IV. When our proposed CA19-9 cut-off levels are used in other hospitals or clinics, PC stages should be considered.
We propose that the CA19-9 cut-off values for PC detection in patients with diabetes having a mean HbA1c of 10.0% can be set at 98.4 U/mL. In this real-world trial, 11 patients out of a total of 1,060 were diagnosed with cancers other than PCs. Two of the eleven patients had elevated CA19-9 levels because of the cancers, namely hepatocellular carcinoma and esophageal cancer. CA19-9 was thought to be a useful marker for detecting certain cancers, particularly PCs. However, we should keep in mind that the sensitivity and specificity were not 100%.
Conceived and designed the study: YT MT YM TY. Performed the experiments: TM. Analyzed the data: YT. Provided clinical data: YT TM YM. Wrote the paper: YT MT YT.
None of the authors had any potential conflicts of interest associated with this study.
