Predicting recurrence patterns and outcomes following resection of pancreatic ductal adenocarcinoma: a contemporary real-world data cohort study

Abstract

Introduction:

The prognosis in pancreatic ductal adenocarcinoma (PDAC) remains poor with 2-year recurrence rates around 80% despite curative-intent surgery. This study evaluated factors associated with early recurrence in a contemporary, multi-center, Australasian population. A better understanding of patients at risk of early recurrence (ER) could improve selection of operative candidates.

Methods:

De-identified data were extracted from 23 hospitals participating in the PURPLE pancreatic cancer registry between 2016 and 2024. Clinicopathological features, treatment, recurrence patterns, and survival were examined. ER was defined as cancer recurrence within 12 months of surgery.

Results:

Of 3041 patients identified, 539 completed curative intent surgery. At a median follow-up of 21.8 months, 74% of resected cases had recurred. The median recurrence free survival was 14.3 months. ER occurred in 41% of patients and was associated with advanced age (P = 0.04), worse performance status (P < 0.01), T3–T4 stage (P = 0.01), CA19-9 > 300 U/ml (P < 0.01), involvement of more than two lymph nodes (P < 0.01), lymphovascular invasion (P = 0.01), and no neoadjuvant and/or adjuvant treatment (P < 0.01). There was a significant association between the primary tumor location and recurrence site (P = 0.046). The median OS of those with isolated lung (31.4 months) and locoregional (29.2 months) recurrences was longer than those with liver (21.3 months) or peritoneal (20.7 months) recurrences (P = 0.01). The use of neoadjuvant and adjuvant therapy was associated with lower ER rates (P < 0.01).

Conclusion:

A high proportion of resected PDAC patients experienced ER. Multiple pathological factors can predict ER. Primary tumor location was associated with site of recurrence. Neoadjuvant and adjuvant therapy significantly reduced ER and increased OS.

Introduction

Pancreatic cancer is the third leading cause of cancer related deaths[1] and its global burden is predicted to double over the next 40 years[2]. Over 90% of pancreatic cancers are pancreatic ductal adenocarcinomas (PDAC)[3]. Less than a third of patients have a resectable tumor at the time of diagnosis[4], and despite curative intent resection many cancers recur. The current 5-year survival of all patients with PDAC is only 4.2%[5].

As patients with post-operative early recurrence (ER) derive minimal benefit from surgery, better identification of patients who are at a higher risk of ER is warranted, as other management strategies may be more appropriate in high-risk cases[6]. Previous studies have identified some perioperative risk factors for ER of pancreatic cancer in patients after surgical resection with curative intent. These include elevated preoperative and postoperative CA19-9 levels, larger tumor size, poor differentiation, high lymph node ratio, lymphovascular invasion, perineural invasion and advanced comorbidity index^[7–12^]. Data on Australasia-Pacific patients are limited.

This study utilizes data from the Pancreatic cancer: Understanding Routine Practice and Lifting End results (PURPLE) registry to identify perioperative clinicopathological features of ER, recurrence patterns and post recurrence survival within the Australasian population. This study was conducted in accordance with the TITAN Guidelines 2025 governing the declaration and use of artificial intelligence in research[13].

Methods

Data collection

Prospectively collated de-identified data from consecutive patients with pancreatic cancer was extracted from 23 centers participating in the PURPLE Pancreatic Cancer Registry (ACTRN12617001474347). PURPLE is an international translational registry that partners with multiple tertiary, metropolitan, regional, public, and private healthcare centers across Australia, New Zealand, and Singapore. Standardized key datapoints regarding patient demographics, disease characteristics, treatment and outcomes are collected. The dataset is validated against state level outcomes. Here, we explore PURPLE data to provide insights into the perioperative risk factors for pancreatic cancer recurrence and post recurrence outcomes within a large, longitudinally monitored population. This study was approved by the human ethics committees at each contributing hospital.

Study design and variables of interest

From the aggregated de-identified patient data extracted, patients were excluded if they had non-curative intent surgery, metastatic disease at the time of attempted resection, an initial treatment plan of palliation or bypass, non-PDAC histology, were lost to follow-up, or had oligometastatic disease despite undergoing curative-intent resection (Fig. 1). Patients were divided into two groups: early recurrence (ER) and non-early recurrence (non-ER) (Supplemental Digital Content File 1, available at: http://links.lww.com/MS9/B114). The ER group included patients who had a recurrence free survival (RFS) of less than 12 months. The non-ER group included patients who did not recur within 12 months. The 12-month cut-off point was chosen based on studies by Groot et al[9] and Nappo et al[14], who defined 12 months as the optimal threshold for differentiating between early and late recurrence, based on subsequent prognosis.

HIGHLIGHTS

• This study examines a large real-world cohort of 3041 patients with pancreatic ductal adenocarcinoma (PDAC), to identify patterns and risk factors for early recurrence (ER) in patients who underwent curative intent surgery for early-stage disease.

• Early recurrence (ER), within less than 12 months of resection occurred in 41% of cases, with sites of recurrence associated with differences in overall survival (OS) including median OS of 21.3 months with liver only recurrence, 29.2 months with locoregional recurrence, 20.7 months with omental recurrence and 31.4 months with lung only recurrence (P < 0.01).

• Importantly, independent perioperative risk factors for ER included ECOG performance status of ≥1 (P < 0.05), elevated CA19-9 > 300 U/ml (P < 0.008), T3/4 tumor size (P < 0.009), more than 2 involved lymph nodes (P < 0.001).

• The use of neo-adjuvant and adjuvant therapies was associated with reduced risk of ER, and improved RFS (P < 0.01).

• Future studies could incorporate these findings into a risk-scoring tool to improve selection and outcomes for early-stage PDAC resection and survival.

Figure 1.

CONSORT diagram of patient selection illustrating selection of patients from the PURPLE registry for inclusion in the retrospective cohort analysis. PDAC, pancreatic ductal adenocarcinoma.

Outcomes

The primary outcome measures for this study were ER and overall survival (OS), which was defined as time from diagnosis of PDAC to death. Patients alive at time of data extraction were censored at the date of last medical review. RFS as secondary endpoint, was also measured as the time from surgery to cancer recurrence.

Statistical analysis

The demographics, clinicopathological features, recurrence patterns and adjunct therapy of the patients were compared in the ER and non-ER groups (Supplemental Digital Content File 1, available at: http://links.lww.com/MS9/B114). A chi-square test was used to test for differences between levels of categorical variables. Univariate analyses were conducted to identify prognostic indicators associated with ER and OS for inclusion in multivariable analysis. Multivariable analysis of ER was conducted using a logistic regression model. Cox regression model was considered for multivariable analysis of OS. Backwards stepwise regression techniques were used, including all variables with a P-value of <0.2 from the univariate analysis. Two-sided P-values less than 0.05 were considered statistically significant. Statistical analysis was conducted using Stata version 19.0 (Stata Corp, College Station, Texas, USA).

Results

Between January 2016 and May 2024, de-identified data on a total of 3041 patients diagnosed with PDAC were extracted. Of these, 1109 (36%) had resectable and borderline resectable PDAC at time of diagnosis, and 593 (20%) subsequently successfully underwent surgical resection with curative intent (Fig. 1). Full demographic and clinicopathological characteristics are described in Supplemental Digital Content File 1, available at: http://links.lww.com/MS9/B114. Out of the 593 patients analyzed, 52% (n = 307) were male and the cohort’s mean age at diagnosis was 66 years [standard deviation (SD): 10.3]. An Eastern Cooperative Oncology Group (ECOG) performance status of 0 was observed in 49% (n = 289) of patients at diagnosis. The median preoperative CA19-9 serum concentration at diagnosis was 137.0 U/ml [Interquartile range (IQR): 32.00–461.00]. In 74% (n = 437) of patients, the tumor was located in the head of the pancreas with 23% (n = 134) patients having a tumor in the body or tail of the pancreas. Most resected cases presented with T stage of 1 or 2 (62%, n = 370), with well to moderately differentiated tumor grades identified in 62% (n = 369). The R0 resection status was confirmed in 61% (n = 359) of patients, R1 resection status in 24% (n = 144), and R2 resection status in 2% (n = 9). A third of cases had more than two positive nodes involved (n = 206). Perineural invasion was recorded in 71% (n = 423) and lymphovascular invasion in 50% (n = 297) of patients.

At median follow-up of 21.75 months (IQR: 12.83–35.30), 74% (n = 436) of patients experienced recurrence with 40.91% (n = 243) recurring within the first 12 months after curative intent resection. The median RFS was 14.3 months (IQR: 8.74–24.35) and death within 12 months of diagnosis was noted in 22.8% (n = 135) of the cohort.

Statistically significant differences in the following variables were observed between the ER and non-ER groups; age, ECOG, preoperative CA19-9, T stage, positive nodes, perineural invasion and lymphovascular invasion (Supplemental Digital Content File 1, available at: http://links.lww.com/MS9/B114).

At time of recurrence, metastatic disease was identified in more than one site in almost a third of cases. The most common isolated site of recurrence was locoregional (29%), followed by liver (21%), lung (8%), and peritoneal recurrences (4%). There were higher rates of liver recurrence in the ER group, compared with higher rates of locoregional and lung recurrences in the non-ER group (P < 0.01) (Table 1). A correlation between primary tumor location and the site of recurrence was also observed with 32% incidence of locoregional recurrence occurring in tumors at the head of pancreas compared with 17% and 14% for tumors in the body or tail of pancreas respectively. Primary tumors in the body and tail of the pancreas were associated with higher rates of peritoneal, omental, and multisite recurrences (Table 2).

Table 1.

Recurrence patterns according to RFS, n (%).

Factor	All patients	Early recurrence	Non-early recurrence	P-Value
N	360	171	189
Recurrence Location				< 0.01
Locoregional	103 (28.6%)	38 (22.2%)	65 (34.4%)
Peritoneal/Omental	16 (4.4%)	5 (2.9%)	11 (5.8%)
Liver	74 (20.6%)	41 (24.0%)	33 (17.5%)
Lung	30 (8.3%)	8 (4.7%)	22 (11.6%)
Two or more sites	116 (32.2%)	68 (39.8%)	48 (25.4%)
Other	21 (5.8%)	11 (6.4%)	10 (5.3%)

Table 2.

Recurrence patterns according to primary tumor location, n (%).

Factor	Head	Body	Tail	P-Value
N	270	41	37
Recurrence Location				0.046
Locoregional	85 (31.5%)	7 (17.1%)	5 (13.5%)
Peritoneal/Omental	9 (3.3%)	4 (9.8%)	2 (5.4%)
Liver	57 (21.1%)	6 (14.6%)	9 (24.3%)
Lung	23 (8.5%)	2 (4.9%)	3 (8.1%)
Two or more sites	80 (29.6%)	21 (51.2%)	14 (37.8%)
Other	16 (5.9%)	1 (2.4%)	4 (10.8%)

The median OS across the entire cohort was 21.8 months (IQR: 12.84-35.34) and the 1-year and 2-year survival rates were 77% (n = 458) and 46% (n = 270), respectively (Fig. 2). Comparison between ER and non-ER cohorts demonstrated that median OS was 11.4 months vs 31.3 months, with 2-year survival rates of 14% vs 67%, respectively (P = 0.01). The 1-year survival rate of the ER cohort was 44%. Unsurprisingly, site of metastasis had an impact on OS with better median survival outcomes associated with isolated lung (31.4 months) and locoregional (29.2 months) recurrence compared to isolated liver (21.3 months) and peritoneal/omental (20.7 months) recurrence (P < 0.01) (Fig. 3). Evaluation of OS according to site of primary tumor showed no significant difference (P = 0.31) (Fig. 4).

Figure 2.

Kaplan–Meier survival analysis showing overall survival in the entire cohort following curative-intent resection of PDAC. Median overall survival and survival probabilities at key time points are described. OS, overall survival; PDAC, pancreatic ductal adenocarcinoma.

Figure 3.

Kaplan–Meier survival curves comparing overall survival in patients with recurrence, categorized according to recurrence location (liver, locoregional, lung, peritoneal/omental, or ≥2 sites). OS, overall survival.

Figure 4.

Kaplan–Meier survival curves showing overall survival following curative-intent resection, categorized by primary tumor location (head, body, or tail of the pancreas). OS, overall survival.

The factors predicting ER were analyzed using univariable and multivariable logistic regressions. Independently significant perioperative risk factors for ER in the multivariate analysis include ECOG performance status (P < 0.05 for ECOG 1 compared to ECOG 0), elevated preoperative CA19-9 levels >300 U/ml (OR: 2.101, P = 0.008), advanced T stage (T3-T4 compared to T1-T2, OR: 1.701, P = 0.009), involvement of more than two lymph nodes (OR: 2.146, P < 0.001), and lymphovascular invasion (OR: 1.710, P = 0.009). Neoadjuvant therapy, adjuvant therapy and the use of both neoadjuvant and adjuvant therapies significantly reduced the likelihood of ER (all P-values < 0.01) (Supplemental Digital Content File 2, available at: http://links.lww.com/MS9/B115). Similar results were obtained when neoadjuvant and adjuvant therapy status were not included in the model, with the exception of the inclusion of age as a significant variable (>85 years: OR: 6.115, P = 0.043).

Significant independent risk factors for OS in the multivariable analysis include male gender (HR: 1.298, P = 0.028), preoperative CA19-9 levels >300 U/ml (HR: 1.733, P = 0.002), advanced T stage (T3-T4 compared to T1-T2: HR: 1.436, P = 0.004), poorly differentiated tumor grade (HR: 1.617, P < 0.001), more than two positive lymph nodes (HR: 1.865, P < 0.001), presence of lymphovascular invasion (HR: 1.462, P = 0.005) and R1 resection status (compared to R0: HR: 1.389, P = 0.0017). Neoadjuvant therapy (P = 0.028), adjuvant therapy (P < 0.001) and the use of both neoadjuvant and adjuvant therapies (P < 0.001) significantly increased OS (Supplemental Digital Content File 3, available at: http://links.lww.com/MS9/B116. Similar to the ER multivariable analysis, age was found to be significantly associated with OS (>85 years: HR: 4.730, P = 0.001) when neoadjuvant and/or adjuvant therapy was not considered in the multivariable analysis. Resection status also dropped out of the model as non-significant.

There was a significant association between systemic therapy and RFS in patients with resected PDAC. Neoadjuvant therapy was administered to 133 patients. A significantly higher proportion of the non-ER group (21%) received neo-adjuvant therapy in comparison to the ER group (16%) (P < 0.01). Similarly, adjuvant therapy was more frequently administered in the non-ER group (79%) than the ER group (56%, P < 0.01) (Table 3).

Table 3.

Association between systemic therapy and RFS, n (%).

	Total (n = 593)	ER (n = 243)	Non-ER (n = 350)	P-Value
Neoadjuvant therapy	133 (22.4%)	38 (15.6%)	95 (21.1%)	<0.01
Adjuvant therapy	411 (69.3%)	136 (56.0%)	276 (78.9%)	<0.01

Discussion

This study provides a contemporary, large-scale Australasian analysis of recurrence patterns following curative-intent surgery in patients with PDAC. The findings of this study provide new insights on the association of CA19-9 levels with post-operative outcomes, the relationship between primary tumor location, RFS and metastatic pattern, and the impact of metastasis location on OS. Additionally, it highlights the influence of adjunct therapy on RFS and reinforces current understanding of risk factors for poor postoperative outcomes. This finding aligns with recent findings from the DYNAMIC-Pancreas clinical trial, reported by Lee et al[15].

The overall recurrence rate, median RFS and median OS observed in this study were 73.5%, 14.3 months, and 21.8 months, respectively. This aligns closely with existing literature^[9,11,16^]. Our study identified an ER rate of 41%, which is slightly lower than the 58% ER rate reported by Ge et al[17] and the 51% reported by Groot et al[18]. It is possible that the improvement in ER rate in this contemporary study may reflect increased uptake of adjuvant and neoadjuvant chemotherapy approaches, greater use of combination chemotherapy and advances in surgical techniques over time.

This study found that significant perioperative risk factors for ER and poor OS included advanced age at primary diagnosis, pre-operative CA19-9 > 300 U/ml, T stage 3–4, more than two positive lymph nodes and lymphovascular invasion, which corroborates other studies that identified similar predictors for ER and poor prognosis^[7,8,11,17,19^]. Interestingly, an ECOG value of >0 was a significant risk factor for ER but not OS. Male gender, poorly differentiated tumor grade and R1 resection status are significant risk factors for poor OS but not ER. Neoadjuvant therapy, adjuvant therapy and the use of both neoadjuvant and adjuvant therapies were identified as protective factors.

This study highlights an association between the location of the primary tumor and the site of recurrence, with pancreatic head PDACs exhibiting higher locoregional recurrence rates compared to body or tail PDACs, which had higher rates of peritoneal, omental, and multisite recurrences. This observation was not specifically addressed in prior studies. Although primary tumor location was associated with recurrence site, it was not a risk factor for ER or poorer OS, a finding consistent with previous studies[20].

Our study also reports a significant association between recurrence location and ER, with higher rates of liver recurrence observed in the ER group and higher rates of locoregional and lung recurrences in the non-ER group. These findings align with reports from Kolbeinsson et al[21], Murakawa et al[22], and Zhang et al[11], who similarly observed a higher proportion of liver-only recurrence in patients with ER.

Unsurprisingly, our study demonstrated that ER is associated with a significantly shorter median OS of 11.4 months compared to 31.3 months in the late and no recurrence (non-ER) group, which is consistent with other reports^[8,9,11^]. Additionally, our study found that isolated lung and locoregional recurrences are associated with better OS compared to isolated liver or peritoneal recurrences, supporting the findings of Asakura et al[23] and Suenaga et al[16].

While this study contributes valuable data to the literature, it is important to acknowledge its limitations. As an observational cohort study, it cannot establish causation. Inherent clinician and patient biases and variation in treatment protocols across centers may also exist due to the non-randomized observational nature of registry collated data.

In this comprehensive, contemporary, multi-center, real-world population study, a high proportion of resected PDAC patients experienced early recurrence. The study’s findings underscore the variable outcomes following surgery and the need for personalized approaches to PDAC management, including for patients identified as high-risk for ER and poor OS. Such a personalized approach is being investigated in the DYNAMIC-pancreas clinical trial that investigates the utility of circulating tumor DNA as a biomarker to aid stratification of adjuvant chemotherapy after resection of early-stage pancreatic cancer (trial registration number removed for peer review)[24]. The insights gained from this study can inform the design of future clinical trials, predictive tools, and treatment strategies aimed at reducing ER rates and improving long-term outcomes for PDAC patients.

Conclusion

Our study offers insights into recurrence patterns and survival outcomes following curative-intent surgery for pancreatic ductal adenocarcinoma (PDAC). Its results suggest that advanced age, pre-operative CA19-9 > 300 U/ml, T stage, involvement of more than two lymph nodes, lymphovascular invasion and absence of chemotherapy treatment are important risk factors for early recurrence. In addition to confirming established recurrence patterns in a large Australasian cohort, we identify novel associations between primary tumor location and recurrence site. Our study also showed that isolated lung and locoregional recurrences are linked to better survival compared to liver or peritoneal recurrences. These findings contribute to the ongoing effort to improve the management and outcomes of PDAC patients, emphasizing the need for multidisciplinary, and novel approaches in this challenging malignancy.

Ethical approval

Human Research Ethics Committees at each of the participating institutions approved the study. HREC/16/MH/216, Melbourne Health site reference number: 2016.200.

Consent

Not applicable.

Sources of funding

BL is the senior author and receipent of the Hemstritch Centenary Fellowship, and philanthropic support from the Hemstritch Foundation, the Segal Foundation & the Hemstritch Pancreatic Cancer Centre of Excellence, Walter & Eliza Hall Institute (WEHI), which has supported this work. These funding bodies had no involvement in the data collection, analysis or interpretation of the data.

Author contributions

B.L., E.M., and M.T.: concept design; B.L., E.M., J.S., B.K., S.-A.M.L., R.W., R.Z., M.N., A.R., F.D., V.D., M.B., S.C., C.E C., and D.T.: data acquisition; E.M., B.L., M.T., B.L., B.K., M.B., J.S., R.Z., and V.D.: data interpretation and drafting of manuscript; E.M., M.T., B.L., J.S., M.M., B.K., S.-A.M.L., R.W., R.Z., M.N., A.R., F.D., V.D., M.B., S.C., C.E.C., D.T., and B.L.: approval of manuscript.

Conflicts of interest disclosure

The authors declare that they have no conflicts of interest relevant to this manuscript.

Research registration unique identifying number (UIN)

PURPLE Pancreatic Cancer Registry (ACTRN1261700147 4347).

Guarantor

Belinda Lee.

Provenance and peer review

This paper was not commissioned/invited.

Data availability statement

De-identified aggregated data are available upon reasonable request for this article.