A subgroup of patients with metastatic or recurrent pancreatic cancer have outcomes that are significantly better than the average patient population. This study sample represents only approximately 5% of patients treated with palliative gemcitabine at this institution. This proportion of long-term survivors is relatively low when compared with most randomized clinical trials of gemcitabine [11
]. Potential explanations for this low 1-year survival rate include incomplete ascertainment of long-term survivors, an overall worse prognosis of patients referred for treatment at our institution (referral bias), and the selection of patients with better performance status in clinical trials (selection bias). It is also possible that some patients from our initial population (n = 435) never received gemcitabine due to rapid tumor progression.
The median OS and TTF on gemcitabine were 26.9 and 11.5 months, respectively, while most of the randomized trials with single agent or gemcitabine-based combinations report OS and TTP between 3.8 to 6.7 months and 2.2 to 3.5 months, respectively [13
]. The striking differences in survival outcomes between this group of patients and patients on randomized trials suggest two possible explanations: 1) selected patients receive significant benefit from gemcitabine or 2) this selected population has a better prognosis independent of treatment modality. Clearly, some combination of these might also be true. Given the retrospective nature of this analysis, no firm conclusion differentiating these two possibilities can be made. However, several findings suggest that treatment with gemcitabine accounts, at least in part, for the prolonged survival of these patients. The 25% response-rate to gemcitabine seen in this study is considerably higher than the 5–10% described in the literature [11
]. Although the differences in OS between CT scan responders and non-responders (24 and 18.8 months, respectively) did not reach statistical significance, there was a trend towards longer survival in responders. The TTF on gemcitabine for the entire group of patients was 11.5 months, significantly longer than that in unselected patients. However, this retrospective study does not allow us to attribute the observed long-term outcomes to either increased gemcitabine responsiveness in selected patients or the presence of prognostic factors associated with prolonged survival.
There were no clinical characteristics that predicted long-term survival within this group. Interestingly, commonly considered prognostic factors in metastatic pancreatic cancer did not have a significant impact on OS in our analysis. ECOG PS, a significant clinical factor for OS in previous studies, was not correlated with survival [5
]. This was somewhat expected, because there were no patients with a PS greater than 2 in this study. Likewise, age, gender, number and location of initial sites of metastasis, and pathologic grade did not correlate with OS. The presence of liver metastasis is also considered a poor prognostic factor. The fact that only 3 patients (15%) had liver metastasis at presentation could have contributed to the relative prolonged survival outcomes of our study sample.
The only two factors that significantly influenced OS were total bilirubin and the serum log CA19-9 at baseline in our study. Elevated serum bilirubin had a negative impact on OS, with a HR = 1.31 (p = 0.021). Elevated total bilirubin probably reflects the severity of initial biliary obstruction, incomplete drainage from biliary decompression procedures, hepatic dysfunction due to prolonged cholestasis or the presence of liver metastasis. Cholestasis due to tumor biliary obstruction only partially explains the detrimental contribution of serum bilirubin on OS, as 3 of the 5 patients with elevated bilirubin had a successful biliary decompression procedure. Since only 5 patients had elevated bilirubin levels, the association of bilirubin levels and overall survival should be interpreted with caution. CA19-9 levels correlated with OS, although in a logarithmic rather than in a linear fashion. The decrease in CA19-9 values after gemcitabine did not correlate with the length of OS in our study. Since the patients included in this analysis were selected for at least one year survival after initiation of gemcitabine therapy, and the group as a whole had a relatively high proportion of CA19-9 responders (63%), it is perhaps not surprising that there was not a detectable effect of CA19-9 response on duration of survival.
Previously, Heinemann et al
described a correlation of decreasing levels of CA19-9 and clinical response in patients treated with gemcitabine and cisplatin [23
]. However, they did not describe the correlation of CA19-9 response to survival and did not perform any statistical analysis. Saad et al
demonstrated a significant correlation between the pretreatment serum CA19-9 levels, CA 19-9 response and OS in 28 patients with advanced pancreatic cancer treated with gemcitabine. In their multivariate analysis, they found that both baseline and CA19-9 response correlated to OS (p = .0005 and .0497, respectively) [24
]. Similarly, in a series of 43 patients, Halm et al
showed a significant higher OS for patients with a greater than 20% decrease in CA19-9 values from baseline after treatment with gemcitabine [25
]. In this analysis, CA19-9 responses were the strongest predictor of OS. Ueno et al
have performed a retrospective study of 103 patients with metastatic disease treated with systemic chemotherapy. In their report, serum C-reactive protein ≥ 5 mg/dL, PS of 2 or 3 and CA19-9 above 10,000 U/mL correlated significantly to shorter OS after a multivariate analysis [22
]. These results suggest prognostic and predictive value of both baseline and changes in post treatment CA19-9 for OS in patients with metastatic pancreatic cancer.
Because only six patients had a surgical resection, we could not address the impact of margins and involvement of lymph nodes on length of survival. Six patients with 1 to 7 positive nodes had an OS of 13.1 to 25.0 months after recurrence. In addition, 4 patients with positive margins showed an OS of 17.7 to 25 months after recurrence. Thus, occasionally patients with positive lymph nodes or margins at the time of resection can have survival durations up to two years. Although the difference in OS between patients with well and moderately differentiated tumors compared to those with poorly differentiated tumors did not reach statistical significance, there were only 12 patients with available information on histological grade, which might limit the power to detect smaller differences. Among resected cases, only one tumor measured less than 3 cm, all patients had perineural extension, 83% had positive lymph nodes and 66% had positive surgical margins, all features of aggressive behavior and poor prognosis [26
]. Based on this study, no histopathological characteristics correlated with OS.
Currently, there is no universally accepted standard second-line chemotherapy after gemcitabine failure for metastatic pancreatic cancer although a 5-FU (or capecitabine) based regimen would be most commonly used. We demonstrated that in a selected subgroup of patients, median TTF on gemcitabine and OS were 11.5 and 26.9 months, respectively. Therefore, median intervals greater than 1 year between gemcitabine withdrawal and death can be expected in these cases. In this study, 10 patients (50%) received at least a second-line treatment and the median number of chemotherapeutic regimens or experimental therapy in patients receiving more than one regimen was 3. These data suggest that a significant number of patients who survive more than 1 year with metastatic pancreatic cancer will still be candidates for further therapies after gemcitabine failure. Future trials of second-line therapies for this selected population seem to be warranted.
Although gemcitabine is a well-tolerated chemotherapeutic agent with a favorable toxicity profile, long-term use of gemcitabine is associated with potential development of HUS and other uncommon toxicities, as shown in this and our previous study [30
]. This highlights the importance of continued monitoring for HUS and other side effects for patients undergoing prolonged treatment with gemcitabine.