Adjuvant chemotherapy after surgery for pancreatic ductal adenocarcinoma: retrospective real-life data

For inclusion, curative intend had to be stated and PDAC had to be histopathologically proven. All subtypes and localizations of ductal adenocarcinoma as well as all kind of surgical procedures were considered. Patients, who received neoadjuvant chemotherapy or radiation, were not enrolled as well as patients, who died during the hospital stay.

Data was obtained from in-house medical records and from the database of the comprehensive cancer center, as well as follow-up reports from oncologists and general practitioners. Examined parameters included patients’ demographics; overall, 5-year, and disease-free survival; postoperative complications; and tumor and treatment characteristics such as type and course of operation. Furthermore, the application of adjuvant chemotherapy or reasons for not applying and premature termination of adjuvant chemotherapy were captured. Primary aim was the impact of number of adjuvant chemotherapy cycles on long-term survival of patients after pancreatic resection to clarify the hypothesis whether more cycles might improve survival.

Parameters were documented and analyzed using IBM SPSS for Windows (Version 22.0), and statistical significance was tested using Mann–Whitney U test for continuous and the chi-square tests as well as the Fisher’s exact test for categorical variables. Overall survival was analyzed using the Kaplan–Meier method with post hoc log-rank tests. Multivariate survival analysis was performed with the Cox proportional hazard model. p <  0.05 was considered significant. Results are presented as median values unless otherwise specified.

In 223 (89%) patients, adjuvant chemotherapy was recommended (suggested by interdisciplinary tumor board), but only 62% (n = 137) of the patients finally received adjuvant chemotherapy. Two patients (7%) without recommendation for adjuvant chemotherapy received it leaving 139 (56%) patients with adjuvant chemotherapy after pancreas resection. Application of adjuvant chemotherapy had an impact on overall survival (see Fig. 1, p = 0.001). In 33% (n = 27) of patients with recommended but not delivered adjuvant chemotherapy, tumor recurrence or metastases were detected prior to the initiation of adjuvant chemotherapy. Further reasons for not delivering adjuvant chemotherapy were poor general condition (n = 17, 21%) or personal reasons (n = 13, 16%). Postoperative complications were only in 9% (n = 7) of the patients the cause for not applying adjuvant chemotherapy. On average, chemotherapy was delivered later than recommended. The median for starting adjuvant chemotherapy was 8.5 weeks after the operation while studies recommended 6–8 weeks postoperatively. In contrast to not applying chemotherapy, postoperative complications were frequently the reason for delaying chemotherapy (p = 0.011). To that respect, delayed removal of abdominal drains (> 7 days, p = 0.011) and prolonged hospital stay (> 18 days, p <  0.001) were found responsible. In line with the recommendation in the treatment period, gemcitabine (GEM) was applied as monotherapy in 95% (n = 130) of patients. Three (2%) patients had a therapy with the combination of GEM and 5-fluorouracil (5-FU) and two (2%) patients had a monotherapy with 5-FU. Additional two (2%) patients had a therapy with the combination of GEM and erlotinib. More than 60% (n = 84) of treated patients received 6 cycles of chemotherapy (Fig. 2). Overall, 54 (39%) out of 139 patients, who received adjuvant chemotherapy, terminated the therapy prior to finishing 6 cycles. Reasons for premature termination of adjuvant chemotherapy were toxicity of the applied medication (n = 13, 24%) or recurrence (n = 18, 33%). Forty-one percent (n = 22) premature termination of chemotherapy was undertaken due to patient wish. When adjuvant chemotherapy was terminated early, recurrence-free survival was decreased from 12.2 to 6.9 months (p = 0.005) even when patients with recurrence less than 1 month postoperatively were excluded. In 40 (29%) patients, the anticipated dose of GEM had to be reduced, cycles elongated or, medication had to be changed. Patients with 6 cycles of chemotherapy showed a median overall survival of 27.7 months. Twenty-one (15%) patients that received 4–5 cycles showed a median survival of 26 months, and 33 (24%) of patients receiving only 1–3 cycles had a median survival of 14 months (Fig. 3). Median overall survival was not affected by delayed initiation of chemotherapy (> 8 weeks, n = 71, p = 0.510) or dose reduction, conversion of medication, or prolonged interval between chemotherapy applications (n = 40, p = 0.449).

Forty percent (n = 56) of patients received adjuvant chemotherapy in our hospital, and 60% (n = 83) were treated by an oncologist in private practice. External oncologists performed dose reductions approx. 2.5 times more frequent than oncologists associated to the university hospital. Patients with dose reduction (n = 40) were more frequently treated external (n = 29, 73%) than in our hospital (n = 11, 27%, p = 0.097).

Until the end of follow-up, only eight patients (3%) were alive and in complete remission (CR). In all cases, CR lasted for more than 5 years and they did not develop any recurrence. Only four of these patients had full recommended chemotherapy regimen (6 cycles of GEM, full dosage). Patients’ characteristics are shown in Table 3. Due to small sample size, we waived further statistical analysis of these patients.

The pancreas-specific tumor marker CA 19.9 was increased in 165 (66%) patients preoperatively. The median concentration of CA 19.9 was 113 (95% CI 509–1304) U/mL at the time of diagnosis. Preoperative CA 19.9 of more than 500 U/mL (n = 61, 27%) was associated with decreased overall survival (12.8 vs. 22.6 months, p = 0.012) as well as with decreased 3-year survival rate (8 vs. 27%, p = 0.001). The recurrence-free survival of patients with a preoperatively CA 19.9 of more than 500 U/mL was 5.5 (range 5–9) months compared to 10.1 (range 10–14) months of patients with lower values (p = 0.475). CA 19.9 decreased to 25 (95% CI 220–1021) U/mL after pancreas resection. Patients (n = 80, 38%) without normalization of CA 19.9 (< 37 U/I) postoperatively had a significantly lower overall survival than patients (n = 131, 62%) with normalization (p = 0.001). Overall survival of patients with normalized CA 19.9 postoperatively was 26.4 (range 7–87) months compared to 17.2 (range 1–82) months of patients without normalization (p <  0.001). The recurrence-free survival of postoperatively normalized patients was 12.2 (range 0–82) months compared to 5.3 (range 0–28) months of non-normalized patients (p < 0.001). The higher CA 19.9, the shorter was survival (preoperative: r = − 0.177, p = 0.007, postoperative: r = − 0.147, p = 0.32). Upon recurrence, CA 19.9 increased to 147 (95% CI 724–3523) U/mL. CA 19.9 of more than 500 U/mL (n = 55, 33%) at the time of recurrence diagnosis showed a slightly but non-significant decreased overall survival (13.0 vs. 23.6 months, p = 0.081) as well as a significant decreased 3-year survival rate (7 vs. 23%, p = 0.01).

Diagnosis of recurrence was made in 193 (83%) of 232 follow-up data patients with a median disease-free survival of 7.8 months (range 0–83 months) after resection of the pancreatic tumor. Median disease-free survival was significantly affected by adjuvant chemotherapy and found to be 4.1 months after operation in patients without adjuvant chemotherapy (n = 84) and 10.9 months in patients with chemotherapy (n = 139, p = 0.01). The majority of patients (n = 85, 44%) suffered from distant metastases, followed by the combination of distant metastases and local recurrence (n = 84, 43%). Sole local recurrence was only found in 13% (n = 25) of patients. Distant metastases were found in the liver (n = 95, 49%), lymph nodes (n = 75, 39%), peritoneum (n = 71, 37%), and lung (n = 50, 26%). Subgroup analysis of influence of site-specific metastases on survival showed that lung metastases were associated with a better survival compared to metastases of the liver (p = 0.043, see Fig. 4). Patients with lung metastases (n = 11) had an overall survival of 31.0 (range 7–46) months compared to 22.0 (range 4–47) months of patients with liver metastases (n = 21).

The median overall survival was 18.5 (range 1–147) months. One-year survival rate was 69%, decreasing to 21% after 3 years and to 6% after 5 years. Almost 90% (n = 171, 87%) of the patients died of tumor progression or tumor-related complications. The cause of death of 23 (12%) patients is unknown. Only three patients (1%) died not to tumor-related causes (myocardial infarction, vasovagal asystole, and stroke). The most common tumor-related cause of death was a worsening of general condition presenting as increasing weakness, fatigue, and immobility in 135 patients (69%). Twenty-eight percent (n = 55) suffered from tumor cachexia and 12% (n = 24) from infections. Nearly half of the deceased (n = 92, 46%) had an organ failure such as a respiratory insufficiency (n = 36, 18%), renal insufficiency (n = 17, 9%), or a liver failure (n = 38, 19%). Ascites occurred in 30% (n = 58) of the deceased.

The multivariate analysis showed that the absence of lymph node invasion (p = 0.012), of lymph vessel or vein invasion (p = 0.014 and p = 0.016), and distant metastases (p < 0.001) were associated with longer overall survival. A higher tumor grading (G1/G2, p = 0.019) as well as R0 resection (p < 0.001) improved survival.

Prognosis of PDAC is, due to late diagnosis, rapid tumor progression, and frequent recurrence, limited. Pancreatic surgery followed by adjuvant chemotherapy improves survival and is the only chance for curative treatment. Despite a defined and guideline-recommended therapeutic strategy, the prognosis remains poor. It is known that postoperative complications and a delayed recovery after pancreatic surgery lead to adjuvant chemotherapy omission and treatment delays [7]. Initiation of adjuvant chemotherapy is recommended 6–8 weeks postoperatively, which was, due to various factors, frequently not able in our patients. Delayed initiation of chemotherapy has been reported previously. Merkow et al. observed an initiation of adjuvant chemotherapy with a median time to adjuvant therapy of 52 days (7.4 weeks) in patients without postoperative complications. In patients with complications, the delay was even 70 days [7]. The retrospective analysis of clinical data reveals that the recommended duration between surgery and initiation of chemotherapy is likely to be met only in a subset of patients. Nevertheless, survival analysis of our data showed no association between delayed treatment and decreasing survival, which is confirmed by other studies [14]. Timing of chemotherapy is not as crucial as an adequate number of cycles [14, 15]. Our results showed that more than 3 cycles after curative pancreas surgery were associated with better survival. Interestingly, the benefit of more than 5 cycles was smaller and patients with 4–5 and with 6 cycles had a comparable survival. Contrary to that, Epelboym et al. found that survival is increased after 6 cycles or more cycles of adjuvant chemotherapy [15]. However, they only differentiated between 1 and 5 and 6 and more cycles, which is different to our analysis. For other cancers, it has been reported that more cycles do not necessarily lead to an increased survival [16, 17]. Nevertheless, it must be considered that the results might be influenced by patients’ comorbidities. Only patients in good condition are able to be treated with chemotherapy. It is even more surprising that patients treated with 4–5 cycles had nearly the same survival time than patients with 6 cycles as it is assumable that the patients with fewer cycles quitted therapy due to poor condition or intolerable side effects.

Subgroup analysis of lung metastatic patients compared to liver metastatic patients showed site-specific influence on survival. Despite small sample size, we found an improved overall survival of patients with lung metastases compared to patients with liver metastases, confirming the results of a previously published database analysis of metastatic pancreatic cancer patients with real-life data of initially curative treated patients [18]. Further analysis of our patients showed just tendencies of demographic and clinical differences prior to diagnosis of pancreatic cancer, but due to the small sample size, we are not able to draw firm conclusions. However, future research should include molecular genetic data [19], which is not available in the patients of our study, who were treated between 2001 and 2012, where molecular genetic analysis was not performed as standard.

Due to the retrospective nature, GEM had been applied to 95% of the patients which was in line with past guidelines for chemotherapy treatment of PDAC [4, 20]. Based on novel findings and significantly improved survival, this recommendation will change to a modified treatment with FOLFIRINOX [21, 22]. Our retrospective analysis, however, indicated that already a substantial number of patients are not able to receive full guideline-recommended chemotherapy. For these patients, the beneficial toxicity profile of GEM will remain a valid option for adjuvant chemotherapy. Additionally, our data questions the general applicability of adjuvant mFOLFIRINOX in the overall cohort of resected pancreatic cancer patients since 62% of our patients received adjuvant chemotherapy at all and only 20% received all cycles without dose reduction. Our data is supported by another publication reporting adjuvant chemotherapy in slightly more than half of the patients after curative intended pancreatic surgery [7].

Our retrospective analysis is limited by its monocentric study design and a small sample size as well as by lack of documentation and documentation errors, but only 8% of patients were lost to follow-up. The large time span for collecting the data and change of guidelines affected the treatment recommendations given and the intensity by which the indication for adjuvant therapy was explained to the patient. For identifying reasons for terminating chemotherapy, heterogeneous documentation received from different sources had to be used.