Abstract
Background.
Carbohydrate antigen (CA) 19-9 is a biomarker to monitor treatment effect. A threshold to predict prognostic significance remains undefined. We evaluated the impact of CA19-9 on overall survival (OS) in patients with early-stage pancreatic cancer (PC) utilizing the National Cancer Database (NCDB).
Methods.
The NCDB was queried from 2010 to 2014 to identify patients with clinical stage I–II PC. Patients who had undocumented pretreatment CA19-9 were excluded. Patients were stratified into two cohorts: CA19-9 < 98 U/mL and CA19-9 ≥ 98 U/mL, and further categorized into surgery versus no surgery. Twelve- and 24-month OS rates are reported.
Results.
Overall, 32,382 patients (stage I: 12,173; stage II: 20,209) were included. The majority of stage I (52.1%) and II (60%) patients had CA19-9 ≥ 98 U/mL. Stage I–II patients with CA19-9 < 98 U/mL had improved OS rates (stage I: 67.5%, 42.6%; stage II: 59.8%, 32.8%) compared with stage I and II patients with CA19-9 ≥ 98 U/mL (stage I: 50.7%, 26.9%; stage II: 48.1%, 22%). Among resected stage I patients, CA19-9 <98 U/mL was associated with improved OS (< 98: 80.5%, 56%; ≥ 98: 70.2%, 42.8%), and a similar trend was seen in resected stage II patients (< 98: 77.6%, 49.9%; ≥ 98: 71%, 39.2%). Unresected stage I patients with lower CA19-9 had improved OS (< 98: 42.1%, 17.5; ≥ 98: 29.9%, 10%), with similar findings in unresected stage II patients (< 98: 41.1%, 15.3%; ≥ 98: 33.4%, 10.6%).
Conclusions.
Our study demonstrated the prognostic value of CA19-9 in patients with clinical stage I–II PC, with a value < 98 U/mL demonstrating improved survival. Surgery significantly improved survival at 12 and 24 months irrespective of CA19-9.
Cancer-related death was the second leading cause of death in 2019, with pancreatic cancer (PC) representing the third most common cause of cancer death.1 This neoplasm was previously not among the most common cancers, but its increasing incidence in the Western world, its lethality, and poor outcomes are of major clinical concern. PC imposes significant disease burden and the 5-year survival rates of localized disease (39%) and regional disease (13%) remain poor.2 Further research into a biomarker to help aid clinicians and patients in understanding a patient’s prognosis is critical. Carbohydrate antigen 19-9 (CA19-9), a tumor-associated antigen, is one such example of a biomarker that emerged to help screen and diagnose patients prior to symptom onset.3
Prior studies concluded CA19-9 was an ineffective screening tool in asymptomatic patients. Nearly 7% of patients lack the sialylated Lewis blood group antigen for detection and may be nonsecretors.4 Instead, studies suggest CA19-9 is better at detecting PC in symptomatic patients with gastrointestinal symptoms or icterus.5–8 Various clinical scenarios, including chronic pancreatitis, cholestasis, extrapancreatic malignancies and normal values despite malignancy, create difficulties in CA19-9’s interpretation and application.5,8 Despite these challenges, the use of CA19-9 shifted from a screening marker to a prognostic marker.
The standard of care for clinical stage I and II PC is multimodal therapy including surgical resection and chemotherapy with or without radiotherapy. Among the patients who undergo curative resection, 5-year overall survival (OS) remains limited to 20%, with 30% of those patients developing early recurrence within 1 year.9 Predicting prognosis and monitoring treatment effect prior to and throughout a patient’s treatment course would be beneficial. Elevated pretreatment CA19-9 in both resectable and unresectable pancreatic disease has been demonstrated to be an independent prognostic biomarker of shorter OS.10 There is an increasing body of evidence suggesting CA19-9 may predict postoperative survival and can be used to monitor treatment response after both resection and/or chemoradiotherapy.6,10–15
In clinical practice, serum CA19-9 levels are routinely obtained for patients diagnosed with PC, yet no clear consensus exists with respect to a threshold that may predict improved survival. While studies elsewhere have suggested CA19-9 may be better suited as a marker to monitor treatment effect, it is still overall regarded as a prognostic marker.6,10–13,15 We aimed to evaluate the prognostic ability of a CA19-9 threshold of 98 U/mL in patients with early-stage PC who do and do not undergo surgery. We hypothesized that patients who underwent surgical resection with a pretreatment CA19-9 value < 98 U/mL would have improved survival compared with a pretreatment CA19-9 value ≥ 98 U/mL.
METHODS
Data Source
An observational study was performed using the National Cancer Database (NCDB), an oncology registry sponsored by the American Cancer Society and American College of Surgeons that includes patient data from over 1500 Commission on Cancer (CoC)-accredited United States facilities.16 The data are reported by participating institutions based on the CoC Facility Oncology Registry Data Standards Manual and represent approximately 70% of all newly diagnosed cancer cases nationwide and 75% of all PC diagnoses.17,18 The data used in the study were derived from a de-identified Participant User File report. The American College of Surgeons and the CoC have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigators.
Study Population
The NCDB Participant User File report was queried to identify patients ≥ 18 years of age with clinical stage I or II PC (pancreas primary site codes = C25.0–25.3, C25.7–25.9) between 2004 and 2015. The International Classification of Diseases for Oncology (edition 3.2) histology codes were used to identify patients with PC (8012, 8021, 8035, 8050, 8140, 8211, 8255, 8260, 8290, 8310, 8450, 8453, 8470, 8471, 8480, 8481, 8490, 8500, 8503, 8521, 8576). CA19-9 values in the NCDB were reported from 0–98 U/mL and those in excess as 98 U/mL, and for this reason we dichotomized CA19-9 values into < 98 U/mL and ≥ 98 U/mL. We excluded patients prior to 2010 and after 2015 from our analysis since CA19-9 was rarely reported in the NCDB before 2010 or after 2015. Patients with a missing or unknown pretreatment CA19-9 value (nearly 45% of patients in the initial cohort), clinical stage, clinical stage III or IV disease, incomplete or unknown treatment information including surgery, chemotherapy or radiotherapy, and missing survival status were excluded. Other data of interest included patient (age, sex, race), clinical (CA19-9 [U/mL], facility type, facility location, Charlson–Deyo Comorbidity Score [CCS], year of diagnosis, insurance type, surgical approach), oncologic (primary site [head, body and tail, other], grade, American Joint Committee on Cancer [AJCC] clinical tumor [T], nodal [N], and metastasis [M] staging system [6th and 7th edition]), and outcomes measures. The AJCC staging system for PC is the same for the 6th and 7th editions.19
Outcomes and Definitions
The primary outcome was OS at 12 and 24 months. OS is available in the NCDB and is a reasonable surrogate for disease-specific survival. Patients were stratified into two pretreatment CA19-9 cohorts: < 98 U/mL and ≥ 98 U/mL, and then further stratified into treatment cohorts, surgery versus no surgery. For each clinical stage, outcomes were compared for patients with CA19-9 < 98 U/mL compared with patients with CA19-9 ≥ 98 U/mL. We then compared outcomes between patients in the CA19-9 cohorts who had surgery compared with those who did not have surgery. The open surgical approach included procedures converted to open. The CCS summarized patient comorbidity burden. The pancreatic tumor primary site was stratified into three categories, i.e. pancreatic head, body and tail, or other, including overlapping lesions, sites not otherwise specified, other specified parts of the pancreas, and the pancreatic duct.
Statistical Analysis
The primary independent variable was CA19-9, which was stratified as < 98 U/mL and ≥ 98 U/mL. Statistical analysis was conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and SAS macros developed at the Biostatistics Shared Resource at the Winship Cancer Institute.20 Univariate association between each covariate and CA19-9 were assessed using the Chi-square test for categorical covariates and analysis of variance (ANOVA) for numerical covariates. The univariate association of each covariate with OS was assessed using Cox proportional hazards models. Multivariable Cox proportional hazards models for OS were carried out by a backward variable selection approach with a removal criterion of 0.2. All variables included in the models are detailed within the tables. Kaplan–Meier curves were generated to compare OS between study cohorts. Statistical significance was defined as p < 0.05. This study used de-identified data and was therefore exempt from Institutional Review Board approval.
RESULTS
A total of 380,524 patients with PC between 2004 and 2015 were identified and 32,382 met the inclusion criteria (Fig. 1). Among the included patients, 12,173 had clinical stage I disease and 20,209 had clinical stage II disease. Table 1 highlights the demographics and clinicopathologic characteristics of the included patients. A majority of patients had tumors located in the head of the pancreas (72 and 75%) and received single-agent (35.5 and 31.1%) or multi-agent (23.2 and 36.5%) chemotherapy for clinical stage I and II, respectively. Over half of the clinical stage I (6345 patients, 52.1%) and II patients (12,063 patients, 62.4%) had a CA19-9 value ≥ 98 U/mL.
FIG. 1.
Derivation of the study cohort. CA carbohydrate antigen
TABLE 1.
Characteristics of patients with clinical stage I and II pancreatic cancer diagnosed from 2004 to 2015 in the National Cancer Database
| Clinical stage I [n = 12,173] |
Clinical stage II [n = 20,209] |
|||||
|---|---|---|---|---|---|---|
| CA19-9 < 98 U/mL [n = 5828] |
CA19-9 ≥ 98 U/mL [n = 6345] |
p-value | CA19-9 < 98 U/mL [n = 8146] |
CA19-9 ≥ 98 U/mL [n = 12,063] |
p-value | |
| Median age, years | ||||||
| < 70 | 2966 (50.9) | 2743 (43.2) | < 0.001 | 4218 (51.8) | 5794 (48) | < 0.001 |
| ≥ 70 | 2862 (49.1) | 3602 (56.8) | 3928 (48.2) | 6269 (52) | ||
| Sex | ||||||
| Male | 2840 (48.7) | 3024 (47.7) | 0.238 | 4093 (50.2) | 6085 (50.4) | 0.782 |
| Female | 2988 (51.3) | 3321 (52.3) | 4053 (49.8) | 5978 (49.6) | ||
| Race | ||||||
| White | 4822 (82.7) | 5459 (86) | < 0.001 | 6661 (81.8) | 10,349 (85.8) | < 0.001 |
| Black/African American | 732 (12.6) | 609 (9.6) | 1151 (14.1) | 1198 (9.9) | ||
| Other | 274 (4.7) | 277 (4.4) | 334 (4.1) | 516 (4.3) | ||
| Facility type | ||||||
| Community Cancer Program | 227 (3.9) | 283 (4.5) | < 0.001 | 302 (3.7) | 517 (4.3) | < 0.001 |
| Comprehensive Community Cancer Program | 1712 (29.6) | 2116 (33.5) | 2265 (28) | 3656 (30.5) | ||
| Academic/Research Program | 3061 (52.9) | 3006 (47.5) | 4494 (55.6) | 6283 (52.3) | ||
| Integrated Network Cancer Program | 791 (13.7) | 919 (14.5) | 1021 (12.6) | 1550 (12.9) | ||
| Facility location | ||||||
| Northeast | 1214 (21) | 1300 (20.6) | 0.913 | 1905 (23.6) | 2798 (23.3) | < 0.001 |
| South | 1959 (33.8) | 2143 (33.9) | 2888 (35.7) | 3902 (32.5) | ||
| Midwest | 1709 (29.5) | 1864 (29.5) | 1986 (24.6) | 3267 (27.2) | ||
| West | 909 (15.7) | 1017 (16.1) | 1303 (16.1) | 2039 (17) | ||
| Charlson–Deyo score | ||||||
| 0 | 3629 (62.3) | 3702 (58.4) | < 0.001 | 5290 (64.9) | 7350 (60.9) | < 0.001 |
| 1 | 1598 (27.4) | 1899 (29.9) | 2116 (26) | 3449 (28.6) | ||
| ≥ 2 | 603 (10.3) | 744 (11.7) | 740 (9.1) | 1264 (10.5) | ||
| Primary payor | ||||||
| Not insured/unknown | 190 (3.3) | 211 (3.3) | < 0.001 | 325 (4) | 428 (3.6) | < 0.001 |
| Private | 1852 (31.8) | 1649 (26) | 2725 (33.4) | 3587 (29.7) | ||
| Medicaid/Medicare/other government | 3786 (65) | 4485 (70.7) | 5096 (62.6) | 8048 (66.7) | ||
| Year of diagnosis | ||||||
| 2010–2012 | 2710 (46.5) | 2673 (42.1) | < 0.001 | 4086 (50.2) | 5365 (44.5) | < 0.001 |
| 2013–2015 | 3118 (53.5) | 3672 (57.9) | 4060 (49.8) | 6698 (55.3) | ||
| Primary site | ||||||
| Head | 3951 (67.8) | 4811 (75.8) | < 0.001 | 5924 (72.7) | 9208 (76.3) | < 0.001 |
| Body and tail | 1222 (21) | 880 (13.9) | 1259 (15.5) | 1573 (13) | ||
| Othera | 655 (11.2) | 654 (10.3) | 963 (11.8) | 1282 (10.6) | ||
| Grade | ||||||
| Well differentiated | 521 (14) | 323 (9.4) | < 0.001 | 503 (11.6) | 503 (9.7) | 0.001 |
| Moderately differentiated | 1958 (52.4) | 1798 (52.4) | 2180 (50.4) | 2572 (49.5) | ||
| Poorly differentiated/undifferentiated | 1259 (33.7) | 1310 (38.2) | 1641 (38) | 2122 (40.9) | ||
| AJCC clinical TMN | ||||||
| I | 263 (4.5) | 213 (3.4) | < 0.001 | – | – | 0.059 |
| IA | 1642 (28.2) | 1174 (18.5) | – | – | ||
| IB | 3923 (67.3) | 4958 (78.1) | – | – | ||
| II | – | – | 191 (2.3) | 225 (1.9) | ||
| IIA | – | – | 4416 (54.2) | 6601 (54.7) | ||
| IIB | – | – | 3539 (43.4) | 5237 (43.4) | ||
| Surgery | ||||||
| Yes | 3238 (6.9) | 2720 (46.1) | < 0.001 | 4148 (55.2) | 3760 (33.2) | < 0.001 |
| No | 2077 (39.1) | 3184 (53.9) | 3364 (44.8) | 7556 (66.8) | ||
| Radiotherapy | ||||||
| No radiotherapy | 4646 (79.7) | 5310 (83.7) | < 0.001 | 6475 (79.5) | 10,225 (84.8) | < 0.001 |
| Neoadjuvant | 241 (4.1) | 251 (4) | 514 (6.3) | 734 (6.1) | ||
| Adjuvant | 941 (16.2) | 784 (12.4) | 1157 (14.2) | 1104 (9.2) | ||
| Chemotherapy | ||||||
| No chemotherapy and/or surgery | 3126 (53.6) | 3962 (62.4) | < 0.001 | 4838 (59.4) | 8325 (69) | < 0.001 |
| Neoadjuvant | 304 (5.2) | 339 (5.3) | 684 (8.4) | 942 (7.8) | ||
| Adjuvant | 2205 (37.8) | 1846 (29.1) | 2277 (28) | 2270 (18.8) | ||
| Both | 193 (3.1) | 198 (3.1) | 347 (4.3) | 526 (4.4) | ||
| Chemotherapy regimen | ||||||
| None | 1941 (35.8) | 2251 (38) | 0.001 | 1963 (26) | 3306 (29.1) | < 0.001 |
| Single agent | 2161 (39.9) | 2164 (36.5) | 2734 (36.2) | 3547 (31.2) | ||
| Multi-agent | 1319 (24.3) | 1511 (25.5) | 2864 (37.9) | 4509 (39.7) | ||
Data are expressed as n (%)
AJCC American Joint Committee on Cancer, CA carbohydrate antigen
Other primary site includes overlapping lesions, sites not otherwise specified, other specified parts of the pancreas, and the pancreatic duct
Open approach includes minimally invasive procedures converted to open
Clinical Stage I
Patients with clinical stage I PC had a median OS of 15.4 months. Among stage I patients, those with CA19-9 < 98 U/mL had a median survival of 19.6 months, compared with 12.2 months for patients with a CA19-9 value ≥ 98 U/mL (Fig. 2). On multivariable analysis, surgery compared with no surgery among patients with CA19-9 < 98 U/mL (hazard ratio [HR] 0.36, p < 0.001) and CA19-9 ≥ 98 U/mL (HR 0.41, p < 0.001) was independently associated with improved OS (Table 2). Among patients with CA19-9 < 98 U/mL, older age (HR 1.24, p < 0.001) and CCS 1 (HR 1.13, p = 0.001) and 2 (HR 1.38, p < 0.001) were associated with worse survival, whereas tumor location in the body/tail (HR 0.81, p < 0.001) compared with the head of the pancreas, and receipt of single- (HR 0.81, p < 0.001) and multi-agent chemotherapy (HR 0.73, p < 0.001) were associated with improved OS (Table 2). For patients with CA19-9 ≥ 98 U/mL, older age (HR 1.22, p < 0.001) and CCS 1 (HR 1.12, p = 0.001) or 2 (HR 1.40, p < 0.001) compared with CCS 0 were associated with worse survival, whereas receipt of single- (HR 0.61, p < 0.001) and multi-agent (HR 0.51, p < 0.001) chemotherapy compared with no chemotherapy was associated with improved OS (Table 2). A subanalysis of patients with body and tail tumors only demonstrated similar results, with worse OS for patients with CA19-9 ≥ 98 U/mL compared with CA19-9 < 98 U/mL (results not shown).
FIG. 2.
Kaplan–Meier plot for overall survival of National Cancer Database patients with clinical stage I and II pancreatic cancer, by CA19-9 cohorts. Overall stage-specific adjusted 12- and 24-month survival rate of CA19-9 < 98 U/mL versus CA19-9 > 98 U/mL patient cohorts. (a) Clinical stage I; (b) clinical stage II. Data from the National Cancer Database 2004–2015. CA carbohydrate antigen, CI confidence interval
TABLE 2.
Multivariable Cox proportional hazard overall survival models for patients with clinical stage I pancreatic cancer, by CA19-9 cohort, in the National Cancer Database
| CA19-9 < 98 U/mLa,b |
CA19-9 ≥ 98 U/mLc,d |
|||
|---|---|---|---|---|
| Hazard ratio | p-value | Hazard ratio | p-value | |
| Surgery | ||||
| Yes | 0.36 (0.34–0.39) | < 0.001 | 0.41 (0.38–0.44) | < 0.001 |
| No | Reference | Reference | ||
| Median age, yearse | ||||
| < 69 | Reference | < 0.001 | – | < 0.001 |
| ≥ 69 | 1.24 (1.15–1.33) | – | ||
| < 72 | – | Reference | ||
| ≥ 72 | – | 1.22 (1.15–1.30) | ||
| Sex | ||||
| Male | Reference | 0.131 | – | – |
| Female | 0.95 (0.89–1.02) | – | – | |
| Facility type | ||||
| Community Cancer Program | 1.47 (1.22–1.75) | < 0.001 | 1.21 (1.04–1.41) | 0.012 |
| Comprehensive Community Cancer Program | 1.16 (1.07–1.26) | < 0.001 | 1.15 (1.07–1.23) | < 0.001 |
| Academic/Research Program | Reference | < 0.001 | Reference | – |
| Integrated Network Cancer Program | 1.28 (1.16–1.41) | 1.18 (1.08–1.29) | < 0.001 | |
| Facility location | ||||
| Northeast | 0.83 (0.76–0.92) | < 0.001 | 0.85 (0.78–0.92) | < 0.001 |
| South | Reference | – | Reference | – |
| Midwest | 0.91 (0.83–0.99) | 0.021 | 1.01 (0.94–1.09) | 0.821 |
| West | 0.80 (0.72–0.88) | < 0.001 | 0.91 (0.84–1.00) | 0.047 |
| Charlson–Deyo score | ||||
| 0 | Reference | – | Reference | – |
| 1 | 1.13 (1.05–1.22) | 0.001 | 1.12 (1.04–1.19) | 0.001 |
| ≥ 2 | 1.38 (1.24–1.53) | < 0.001 | 1.40 (1.28–1.53) | < 0.001 |
| Year of diagnosis | ||||
| 2010–2012 | Reference | 0.122 | – | – |
| 2013–2015 | 1.06 (0.99–1.13) | – | – | |
| Primary site | ||||
| Head | Reference | – | Reference | – |
| Body and tail | 0.81 (0.74–0.89) | < 0.001 | 0.97 (0.89–1.06) | 0.524 |
| Otherf | 0.93 (0.83–1.03) | 0.169 | 0.87 (0.79–0.96) | 0.006 |
| Chemotherapy regimen | ||||
| Single agent | 0.81 (0.74–0.87) | < 0.001 | 0.61 (0.57–0.66) | < 0.001 |
| Multi-agent | 0.73 (0.67–0.80) | < 0.001 | 0.51 (0.47–0.55) | < 0.001 |
| None | Reference | – | Reference | – |
CA carbohydrate antigen
Number of patients with pancreatic cancer diagnosed between 2004 and 2015 used in the CA19-9 < 98 U/mL model, n = 4910
The following variables were considered for the CA19-9 < 98 U/mL model after using backward selection with a criterion of 0.2 (removal of race): surgery (reference = no), age (median 69 years), sex (reference = male), facility type (reference = academic/research program), facility location (reference = South), Charlson–Deyo score (reference = 0), year of diagnosis (reference = 2010–2012), primary site (reference = head), and chemotherapy (reference = none)
Number of patients with pancreatic cancer diagnosed between 2004 and 2015 used in the CA19-9 ≥ 98 U/mL model, n = 5487
The following variables were considered for the CA19-9 ≥ 98 U/mL model after using backward selection with a criterion of 0.2 (removal of sex, race, year of diagnosis): surgery (reference = no), age (median 72 years), facility type (reference = academic/research program), facility location (reference = South), Charlson–Deyo score (reference = 0), primary site (reference = head), and chemotherapy (reference = none)
Age was dichotomized as the median age for a specific cohort
Other primary site includes overlapping lesions, sites not otherwise specified, other specified parts of the pancreas, and the pancreatic duct
Further stratifying clinical stage I CA19-9 cohorts by treatment, 5958 patients underwent surgical resection, while 5261 patients did not (electronic supplementary Table 1). Among those patients who had surgery, the majority had a CA19-9 value < 98 U/mL (3238 patients, 54.3%). In this cohort, patients with CA19-9 <98 U/mL compared with patients with a CA19-9 value ≥ 98 U/mL who underwent surgical resection were less likely to have tumors in the head of the pancreas (66.3% vs. 75%, p < 0.001), stage IB disease (63% vs. 75.4%, p < 0.001), and receive neoadjuvant chemotherapy (7.7% vs. 9.7%, p = 0.002). Similar trends were noted among patients with a CA19-9 value < 98 and ≥ 98 U/mL who did not undergo surgical resection (electronic supplementary Table 1). Among patients who underwent surgery, those with CA19-9 < 98 U/mL compared with CA19-9 ≥ 98 U/mL had a 12-month survival rate of 80.5% versus 70.2% (p < 0.001) and a 24-month survival rate of 56% versus 42.8% (p < 0.001) (Fig. 3a). Patients with CA19-9 < 98 U/mL, compared with those with CA19-9 ≥ 98 U/mL, who did not undergo resection had a 12-month survival rate of 42.1% versus 29.9% (p < 0.001) and a 24-month survival rate of 17.5% versus 10% (p < 0.001) (Fig. 3b).
FIG. 3.
Kaplan–Meier plot for adjusted overall survival of National Cancer Database patients with clinical stage I and II pancreatic cancer, by treatment group and CA19-9 cohorts. Overall stage, treatment-specific (surgery vs. no surgery), and CA19-9 adjusted 12- and 24-month overall survival rate. Blue line indicates CA19-9 < 98 U/mL and red line indicates CA19-9 ≥ 98 U/mL. (a) Clinical stage I, surgery, pretreatment CA19-9 < 98 U/mL versus CA19-9 > 98 U/mL; (b) clinical stage I, no surgery, pretreatment CA19-9 < 98 U/mL versus CA19-9 > 98 U/mL; (c) clinical stage II, surgery, pretreatment CA19-9 < 98 U/mL versus CA19-9 > 98 U/mL; (d) clinical stage II, no surgery, pretreatment CA19-9 < 98 U/mL versus CA19-9 > 98 U/mL. Data from the National Cancer Database 2004–2015. CA carbohydrate antigen, CI confidence interval
Clinical Stage II
Median OS of patients with clinical stage II PC was 12.9 months. Among patients with CA19-9 < 98 U/mL, median survival was 15.5 months, compared with 11.5 months for those patients with CA19-9 ≥ 98 U/mL (Fig. 2). In clinical stage II patients, multivariable analysis identified surgery was associated with improved OS (HR 0.39, p < 0.001; and HR 0.42, p < 0.001), irrespective of CA19-9 (Table 3). For patients with CA19-9 < 98 U/mL, older age (HR 1.16, p < 0.001) and CCS 1 (HR 1.08, p = 0.012) and 2 (HR 1.31, p < 0.001) were associated with worse OS, whereas tumor location in the body/tail (HR 0.90, p = 0.007) compared with the head of the pancreas, and receipt of single- (HR 0.52, p < 0.001) and multi-agent (HR 0.44, p < 0.001) chemotherapy, were independently associated with improved OS (Table 3). A similar trend was noted for patients with CA19-9 ≥ 98 U/mL (Table 3). A subanalysis of patients with body and tail tumors only demonstrated similar results, with worse OS for patients with CA19-9 ≥ 98 U/mL compared with CA19-9 <98 U/mL (results not shown).
TABLE 3.
Multivariable Cox proportional hazard overall survival models for patients with clinical stage II pancreatic cancer, by CA19-9 cohort, in the National Cancer Database
| CA19-9 < 98 U/mLa,b |
CA19-9 ≥ 98 U/mLc,d |
|||
|---|---|---|---|---|
| Hazard ratio | p-value | Hazard ratio | p-value | |
| Surgery | ||||
| Yes | 0.39 (0.37–0.42) | < 0.001 | 0.42 (0.40–0.44) | < 0.001 |
| No | Reference | Reference | ||
| Median age, yearse | ||||
| <68 | Reference | < 0.001 | – | – |
| ≥ 68 | 1.16 (1.09–1.22) | – | – | < 0.001 |
| <69 | – | Reference | ||
| ≥ 69 | – | 1.18 (1.13–1.23) | ||
| Race | ||||
| White | – | – | Reference | – |
| Black/African American | – | – | 0.96 (0.89–1.02) | 0.199 |
| Other | – | – | 0.84 (0.76–0.94) | 0.002 |
| Facility type | ||||
| Community Cancer Program | 1.30 (1.12–1.52) | 0.001 | 1.05 (0.94–1.17) | 0.363 |
| Comprehensive Community Cancer Program | 1.14 (1.07–1.22) | < 0.001 | 1.16 (1.11–1.22) | < 0.001 |
| Academic/Research Program | Reference | – | Reference | – |
| Integrated Network Cancer Program | 1.23 (1.13–1.33) | < 0.001 | 1.15 (1.07–1.22) | < 0.001 |
| Facility location | ||||
| Northeast | 0.87 (0.81–0.94) | < 0.001 | 0.86 (0.81–0.91) | < 0.001 |
| South | Reference | – | Reference | – |
| Midwest | 0.99 (0.93–1.07) | 0.872 | 1.02(0.96–1.07) | 0.553 |
| West | 0.94 (0.87–1.02) | 0.124 | 0.94 (0.89–1.01) | 0.071 |
| Charlson–Deyo score | ||||
| 0 | Reference | – | Reference | – |
| 1 | 1.08 (1.02–1.15) | 0.012 | 1.10 (1.05–1.15) | < 0.001 |
| ≥ 2 | 1.31 (1.20–1.43) | < 0.001 | 1.30 (1.22–1.39) | < 0.001 |
| Primary site | ||||
| Head | Reference | – | – | – |
| Body and tail | 0.90 (0.84–0.97) | 0.007 | – | – |
| Other | 0.96 (0.88–1.04) | 0.321 | – | – |
| Chemotherapy regimen | ||||
| Single agent | 0.52 (0.49–0.56) | < 0.001 | 0.50 (0.48–0.53) | < 0.001 |
| Multi-agent | 0.44 (0.41–0.47) | < 0.001 | 0.39 (0.37–0.41) | < 0.001 |
| None | Reference | – | Reference | – |
CA carbohydrate antigen
Number of patients with pancreatic cancer diagnosed between 2004 and 2015 used in the CA19-9 < 98 U/mL model, n = 6910
The following variables were considered for the CA19-9 < 98 U/mL model after using backward selection with a criterion of 0.2 (removal of sex, race, year of diagnosis): surgery (reference = no), age (median 68 years), facility type (reference = academic/research program), facility location (reference = South), Charlson–Deyo score (reference = 0), primary site (reference = head), chemotherapy (reference = none)
Number of patients with pancreatic cancer diagnosed between 2004 and 2015 used in the CA19-9 ≥ 98 U/mL model, n = 10,603
The following variables were considered for the CA19-9 ≥ 98 U/mL model after using backward selection with a criterion of 0.2 (removal of sex, primary site, year of diagnosis): surgery (reference = no), age (median 69 years), race (reference = White), facility type (reference = academic/research program), facility location (reference = South), Charlson–Deyo score (reference = 0), primary site (reference = head), chemotherapy (reference = none)
Age was dichotomized as the median age for a specific cohort
Among patients with clinical stage II disease, the minority of patients (n = 7124, 35.3%) underwent surgical resection. Those who had surgery differed in the distribution of age, race, insurance type, CCS, tumor location, and receipt of neoadjuvant radiotherapy and/or chemotherapy when comparing patients with a CA19-9 value < 98 U/mL with patients with a CA19-9 value ≥ 98 U/mL (electronic supplementary Table 2). Patients who did not undergo resection differed in distribution of age, race, tumor location, and radiotherapy/chemotherapy treatment when comparing CA19-9 < 98 U/mL with CA19-9 ≥ 98 U/mL in this population. Patients with CA19-9 < 98 U/mL, compared with patients with CA19-9 ≥ 98 U/mL, who underwent surgery had a 12-month survival rate of 77.6% versus 71% (p < 0.001) and a 24-month survival rate of 49.9% versus 39.2% (p < 0.001). Among patients who did not undergo surgery, those with CA19-9 < 98 U/mL, compared with patients with CA19-9 ≥ 98 U/mL, had a 12-month survival rate of 41.1% versus 33.4% (p < 0.001) and a 24-month survival rate of 15.3% versus 10.6% (p < 0.001) (Figs. 3c, d).
DISCUSSION
We utilized the NCDB to evaluate pretreatment CA19-9 as a prognostic biomarker in clinical stage I and II PC. This study aimed to identify whether a CA19-9 value < 98 U/mL compared with a CA19-9 value of ≥ 98 U/mL was associated with improved survival. We further delineated stage and CA19-9-specific OS for patients with both resected and unresected PC. Patients with clinical stage I and II PC and a pretreatment CA19-9 < 98 U/mL have significantly improved survival compared with their same clinical stage counterparts with CA19-9 ≥ 98 U/mL. Regardless of stage, patients who undergo surgical resection have significantly improved survival irrespective of pretreatment CA19-9, although this effect is more pronounced in patients with CA19-9 < 98 U/mL. Single- or multi-agent chemotherapy confers a significant survival benefit for stage I and II patients regardless of pretreatment CA19-9 value, with a greater benefit noted in patients with CA19-9 > 98 U/mL. These results enrich current literature and suggest that a pretreatment CA19-9 threshold of 98 U/mL may be useful when counseling patients with early-stage PC.
Prior studies have sought to identify various thresholds of CA19-9, ranging from 37 U/mL (a normal serum level) to 2000 U/mL, which may be associated with increased mortality in patients with PC.21–24 These studies were single-center, limited in sample size, and restricted only to patients who underwent surgical resection or those with advanced disease (clinical or pathological stage III–IV). These factors limit the generalizability of prior study thresholds and results. Similarly, a large NCDB study included patients with pathologic stage I–IV pancreatic ductal adenocarcinoma (PDAC) and found patients with elevated pretreatment CA19-9 > 37 U/mL had worse OS compared with patients with levels < 37 U/mL or nonsecretors.25 While this well-powered study demonstrated poor OS in early-stage PDAC, it excluded patients who did not undergo curative intent resection. Surgical resection does not come without accepted morbidity and mortality, and understanding how pretreatment CA19-9 may be associated with survival in unresected patients is equally as important. Our study included only patients with early clinical stage disease and demonstrated that nearly 44% of stage I and 58% of stage II patients do not undergo surgical resection. Reasons for not undergoing resection are multifactorial, but reasonable given the associated risks, and are ultimately beyond the scope of this study. The applicability of survival rates informed by populations undergoing resection to a subset of patients who do not undergo resection is limited. This study fills a void in the current literature by providing clinicians with more generalizable and well-powered survival information for patients with early-stage PC. Specifically, not only does it provide survival estimates by stage but also survival estimates for patients who do and do not pursue surgical resection. This information is valuable to clinicians who can use this noninvasive means to better understand patient prognosis and to effectively counsel patients. The survival estimation of patients with early-stage disease who do not undergo resection is not provided elaborately elsewhere and is crucial to inform the shared decision making discussion between patients and providers and better align patient goals and treatment options.
Our finding that irrespective of CA19-9 value, surgical resection was an independent predictor of improved OS in patients with stage I and II PC is consistent with prior literature.26,27 Similar to studies elsewhere, we identified that a lower CA19-9 coupled with surgical resection was associated with improved OS at 12 and 24 months.21,24,25 However, direct comparison of median OS measured in other studies is limited due to the different CA19-9 thresholds and inclusion of patients with advanced clinical-or pathologic-stage disease. In one of the largest prospective cohorts, Ferrone et al. analyzed patients with stage II–IV PC and compared preoperative CA19-9 thresholds of 37, 200, 1000, and 2000 U/mL. They found significantly improved survival for patients with values below each indicated threshold (200, 1000, and 2000 U/mL) compared to patients with values above the threshold. Survival was similar and approximately 1.8–2.3 years for patients below their suggested thresholds.24 While this includes a significant number of patients with advanced disease, it suggests an optimal CA19-9 threshold for predicting improved survival may be between 37 and 200 U/mL. Bergquist et al. applied a CA19-9 threshold of 37 U/mL to a cohort of NCDB patients with pathologic stage I and II PDAC, and found that regardless of the level of elevation (37–98 U/mL), patients had stage-specific decreased OS.25 Our study’s pretreatment value of 98 U/mL falls within the range these studies explored. In contrast to Bergquist et al., we found significantly improved OS below 98 U/mL in clinical stage I and II PC. A CA19-9 threshold of 98 U/mL may be a reasonable threshold to use for triaging and counseling patients with early PC.
Chemotherapy has known survival benefits for patients with early clinical stage PC.28–31 Our study corroborates this and demonstrates that patients with clinical stage I or II PC have an associated survival benefit from single- or multi-agent chemotherapy irrespective of the CA19-9 value, although a greater benefit is seen in patients with CA19-9 ≥ 98 U/mL. Clinical trials are examining the use of neoadjuvant therapy versus upfront surgery and various combinations of chemotherapeutic formulations in patients with resectable PC.24,28,30,32 The timing of chemotherapy is an important consideration since 40% of patients after resection are unable to complete adjuvant therapy.28,30 Bergquist et al. stratified a cohort of patients identified in the NCDB with pathologic stage I–IV PC and pretreatment CA19-9 > 37 U/mL by treatment groups and demonstrated that those patients who received neoadjuvant therapy + surgery had improved OS compared with surgery + adjuvant therapy and surgery alone.25 While important, the application of our study’s thresholds to treatment sequence is outside the scope of this study and may be a direction for future investigation.
This study has strengths and limitations that should be acknowledged. Although this is the first large study utilizing the NCDB to analyze a CA19-9 threshold of 98 U/mL among patients with clinical stage I and II PC, it is a retrospective study and prone to selection bias. Additionally, the stringent inclusion criteria for pretreatment CA19-9 excluded a significant number of patients, which introduces sampling bias. However, the demographic distribution of included patients is diverse, which minimizes this risk. The NCDB does not contain certain data variables that may be of interest when drawing conclusions, including bilirubin level, disease-specific survival, or more granular information regarding specific treatments such as chemotherapy regimens, dose, or duration. We attempted to control for these issues and relevant confounding factors with multivariable models. Subanalysis of patients with body and tail tumors only demonstrated similar results. While bilirubin value is albeit useful to ensure it is normal when the CA19-9 value is drawn, it does not change the results presented in this manuscript. The NCDB does capture nearly 70% of all new cancer diagnoses nationwide and about 75% of all PC diagnoses, providing the opportunity to perform adequately powered analyses, which is a major gap in the current literature on this topic.17,18 The time period of this study crossed both the 6th and 7th AJCC editions, which may have resulted in different treatment algorithms for similar stage presentations, although AJCC staging is the same between these editions.19 Despite these limitations, we believe that these data are robust, given the careful inclusion criteria, multivariable analyses accounting for hospital variation, and corroboration with other literature demonstrating that lower CA19-9, surgical resection, and chemotherapy are independent predictors of improved OS.21–25,28,29
CONCLUSION
In patients diagnosed with clinical stage I and II PC, pretreatment CA19-9 < 98 U/mL is associated with improved OS at 12 and 24 months. This study suggested the pretreatment CA19-9 threshold of 98 U/mL may be a reasonable cut-off to apply to patients with clinical stage I or II PC. Survival information, extrapolated by noninvasive means, may be useful for clinicians when counseling patients who do and do not undergo surgical resection, and to inform the shared decision-making process. Future studies should focus on defining the optimal treatment sequence associated with improved OS in patients with early-stage PC.
Supplementary Material
ACKNOWLEDGMENT
Research reported in this publication was supported in part by the Biostatistics Shared Resource of Winship Cancer Institute of Emory University and NIH/NCI under award number P30CA138292. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
FUNDING
This work received no direct funding.
Footnotes
Supplementary Information The online version contains supplementary material available at https://doi.org/10.1245/s10434-022-12497-x.
DISCLOSURES Darren Carpizo is the founder of Z53 Therapeutics Inc. Alexa D. Melucci, Alexander C. Chacon, Paul R. Burchard, Vasileios Tsagkalidis, Anthony S. Casabianca, Subir Goyal, Jeffrey M. Switchenko, David A. Kooby, Charles A. Staley, and Mihir M. Shah have no disclosures to declare.
REFERENCES
- 1.Centers for Disease Control and Prevention. An Update of Cancer Deaths in the United States 2021. Available at: https://www.cdc.gov/cancer/dcpc/research/update-on-cancer-deaths/index.htm#:~:text=What%20were%20the%20leading%20causes,intrahepatic%20bile%20duct%20(5%25). Accessed 10 Feb 2022.
- 2.American Cancer Society. Survival Rates for Pancreatic Cancer Available at: https://www.cancer.org/cancer/pancreatic-cancer/detection-diagnosis-staging/survival-rates.html. Published 2021. Accessed 10 Feb 2022.
- 3.Steinberg W The clinical utility of the CA 19–9 tumor-associated antigen. Am J Gastroenterol 1990;85(4):350–5. [PubMed] [Google Scholar]
- 4.Tempero MA, Uchida E, Takasaki H, Burnett DA, Steplewski Z, Pour PM. Relationship of carbohydrate antigen 19–9 and Lewis antigens in pancreatic cancer. Cancer Res 1987;47(20):5501–3. [PubMed] [Google Scholar]
- 5.Dumitra S, Jamal MH, Aboukhalil J, et al. Pancreatic cancer and predictors of survival: comparing the CA 19–9/bilirubin ratio with the McGill Brisbane Symptom Score. HPB (Oxford) 2013;15(12):1002–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Poruk KE, Gay DZ, Brown K, et al. The clinical utility of CA 19–9 in pancreatic adenocarcinoma: diagnostic and prognostic updates. Curr Mol Med 2013;13(3):340–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kim JE, Lee KT, Lee JK, Paik SW, Rhee JC, Choi KW. Clinical usefulness of carbohydrate antigen 19–9 as a screening test for pancreatic cancer in an asymptomatic population. J Gastroenterol Hepatol 2004;19(2):182–6. [DOI] [PubMed] [Google Scholar]
- 8.Homma T, Tsuchiya R. The study of the mass screening of persons without symptoms and of the screening of outpatients with gastrointestinal complaints or icterus for pancreatic cancer in Japan, using CA19-9 and elastase-1 or ultrasonography. Int J Pancreatol 1991;9:119–24. [DOI] [PubMed] [Google Scholar]
- 9.Lekka K, Tzitzi E, Giakoustidis A, Papadopoulos V, Giakoustidis D. Contemporary management of borderline resectable pancreatic ductal adenocarcinoma. Ann Hepatobiliary Pancreat Surg 2019;23(2):97–108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Schultz NA, Christensen IJ, Werner J, et al. Diagnostic and prognostic impact of circulating YKL-40, IL-6, and CA 19.9 in Patients with Pancreatic Cancer. PLoS One 2013;8(6):e67059–e67059. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kondo N, Murakami Y, Uemura K, et al. Prognostic impact of perioperative serum CA 19–9 levels in patients with resectable pancreatic cancer. Ann Surg Oncol 2010;17(9):2321–9. [DOI] [PubMed] [Google Scholar]
- 12.Barton JG, Bois JP, Sarr MG, et al. Predictive and prognostic value of CA 19–9 in resected pancreatic adenocarcinoma. J Gastrointest Surg 2009;13(11):2050–8. [DOI] [PubMed] [Google Scholar]
- 13.Humphris JL, Chang DK, Johns AL, et al. The prognostic and predictive value of serum CA19.9 in pancreatic cancer. Ann Oncol 2012;23(7):1713–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Park JK, Paik WH, Ryu JK, et al. Clinical significance and revisiting the meaning of CA 19–9 blood level before and after the treatment of pancreatic ductal adenocarcinoma: analysis of 1,446 patients from the pancreatic cancer cohort in a single institution. PLoS One 2013;8(11):e78977–e78977. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang GY, Malik NK, Chandrasekhar R, et al. Change in CA 19–9 levels after chemoradiotherapy predicts survival in patients with locally advanced unresectable pancreatic cancer. J Gastrointest Oncol 2013;4(4):361–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.American College of Surgeons. American College of Surgeons. National Cancer Database Available at: https://www.facs.org/quality-programs/cancer/ncdb. Accessed 10 Nov 2020.
- 17.National Cancer Database Facility Oncology Registry Data Standards (FORDS): Revised for 2016 Available at: https://www.facs.org/quality-programs/cancer/ncdb/call-for-data/fordsmanual. Accessed 12 Sep 2021.
- 18.Mallin K, Browner A, Palis B, et al. Incident cases captured in the national cancer database compared with Those in U.S. Population Based Central Cancer Registries in 2012–2014. Ann Surg Oncol 2019;26(6):1604–12. [DOI] [PubMed] [Google Scholar]
- 19.Bilimoria KY, Bentrem DJ, Ko CY, et al. (2007) Validation of the 6th edition AJCC Pancreatic Cancer Staging System: report from the National Cancer Database. Cancer 110(4): 738–744. [DOI] [PubMed] [Google Scholar]
- 20.Liu Y, Nickleach DC, Zhang C, Switchenko JM, Kowalski J. 2018. Carrying out streamlined routine data analyses with reports for observational studies: introduction to a series of generic SAS® macros. F1000Res 7:1955. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Dong Q, Yang XH, Zhang Y, et al. Elevated serum CA19-9 level is a promising predictor for poor prognosis in patients with resectable pancreatic ductal adenocarcinoma: a pilot study. World J Surg Oncol 2014;12:171. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Berger AC, Meszoely IM, Ross EA, Watson JC, Hoffman JP. Undetectable preoperative levels of serum CA 19–9 correlate with improved survival for patients with resectable pancreatic adenocarcinoma. Ann Surg Oncol 2004;11(7):644–9. [DOI] [PubMed] [Google Scholar]
- 23.Waraya M, Yamashita K, Katagiri H, et al. Preoperative serum CA19-9 and dissected peripancreatic tissue margin as determiners of long-term survival in pancreatic cancer. Ann Surg Oncol 2009;16(5):1231–40. [DOI] [PubMed] [Google Scholar]
- 24.Ferrone CR, Finkelstein DM, Thayer SP, Muzikansky A, Fernandez-delCastillo C, Warshaw AL. Perioperative CA19-9 levels can predict stage and survival in patients with resectable pancreatic adenocarcinoma. J Clin Oncol 2006;24(18):2897–902. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Bergquist JR, Puig CA, Shubert CR, et al. Carbohydrate Antigen 19–9 elevation in anatomically resectable, early stage pancreatic cancer is independently associated with decreased overall survival and an indication for neoadjuvant therapy: a national cancer database study. J Am Coll Surg 2016;223(1):52–65. [DOI] [PubMed] [Google Scholar]
- 26.Chakraborty S, Singh S. Surgical resection improves survival in pancreatic cancer patients without vascular invasion- a population based study. Ann Gastroenterol 2013;26(4):346–52. [PMC free article] [PubMed] [Google Scholar]
- 27.Wagner M, Redaelli C, Lietz M, Seiler CA, Friess H, Büchler MW. Curative resection is the single most important factor determining outcome in patients with pancreatic adenocarcinoma. Br J Surg 2004;91(5):586–94. [DOI] [PubMed] [Google Scholar]
- 28.Müller PC, Frey MC, Ruzza CM, et al. Neoadjuvant Chemotherapy in Pancreatic Cancer: an appraisal of the current high-level evidence. Pharmacology 2021;106(3–4):143–53. [DOI] [PubMed] [Google Scholar]
- 29.Janssen QP, O’Reilly EM, van Eijck CHJ, Groot Koerkamp B. Neoadjuvant treatment in patients with resectable and borderline resectable pancreatic cancer. Front Oncol 2020;10:41–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sohal D, Duong MT, Ahmad SA, et al. SWOG S1505: Results of perioperative chemotherapy (peri-op CTx) with mfolfirinox versus gemcitabine/nab-paclitaxel (Gem/nabP) for resectable pancreatic ductal adenocarcinoma (PDA). J Clin Oncol 2020;38(15 Suppl):4504. [Google Scholar]
- 31.Conroy T, Hammel P, Hebbar M, et al. Unicancer GI PRODIGE 24/CCTG PA.6 trial: A multicenter international randomized phase III trial of adjuvant mFOLFIRINOX versus gemcitabine (gem) in patients with resected pancreatic ductal adenocarcinomas. J Clin Oncol 2018;36(18 Suppl):LBA4001. [Google Scholar]
- 32.Reni M, Balzano G, Zanon S, et al. Safety and efficacy of preoperative or postoperative chemotherapy for resectable pancreatic adenocarcinoma (PACT-15): a randomised, open-label, phase 2–3 trial. Lancet Gastroenterol Hepatol 2018;3(6):413–23. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.