A number of previous studies have provided justification for measuring CA19-9 levels in patients with advanced pancreatic cancer receiving either chemotherapy or radiation, both for prognostic and for monitoring purposes. Patients with higher CA19-9 levels prior to initiation of chemoradiation for locally advanced disease have poorer outcomes in terms of both response rates and overall survival (Ikeda et al, 2001
; Micke et al, 2003b
). A lower CA19-9 level following completion of chemoradiation also appears to correlate with survival (Micke et al, 2003a
). Furthermore, the degree of change in CA19-9 levels during radiotherapy may indicate how well a patient will fare, with two separate studies demonstrating that either a 50 or a 75% biomarker decline correlates with improved median survival time (Katz et al, 1998
; Okusaka et al, 1998
With the widespread use of gemcitabine as the mainstay of treatment in patients with advanced pancreatic cancer, recent reports have begun to examine CA19-9 response to gemcitabine-based treatments and whether kinetics of CA19-9 can serve as a predictor of response to such treatments. A small retrospective study by Saad et al (2002)
of 28 patients with advanced pancreatic cancer treated with gemcitabine found that lower pretreatment levels of CA19-9, as well as a
50% decline in CA19-9 anytime during treatment, correlated with better survival rates. Halm et al (2000)
examined 43 unresectable patients receiving gemcitabine treatment and found that those with a >20% decrease of their baseline CA19-9 level after 8 weeks of treatment had a longer median survival than those with a rise or a decrease <20% (268 vs
110 days), a finding confirmed by Ziske et al (2003)
. This biomarker response was in fact the only independent predictor of survival in a multivariate analysis, showing a greater level of significance than either objective tumour response or clinical benefit response (Ziske et al, 2003
). Heinemann and colleagues, meanwhile, collected CA19-9 data from patients with advanced pancreatic cancer receiving treatment on a study protocol using a combination of cisplatin and gemcitabine (Stemmler et al, 2003
). CA19-9 responders, defined as those with a
50% decrease in CA19-9 levels within 2 months after the start of treatment, survived significantly longer than CA19-9 nonresponders (295 vs
174 days, P
Our study represents the first to examine whether changes in CA19-9 correlate with clinical outcomes in patients with advanced pancreatic cancer treated with FDR gemcitabine. Infusion of gemcitabine at an FDR is a strategy intended to optimise pharmacokinetics of the drug (Grunewald et al, 1990
). In a recently published randomised phase II study, Tempero et al (2003)
reported that administration of gemcitabine by FDR infusion resulted in superior outcomes in patients with advanced pancreatic cancer compared to administration of this drug by standard 30-min infusion.
We were specifically interested in examining how changes in CA19-9 during treatment with FDR gemcitabine correlate with clinical outcomes. Our response criteria were purposely broad in that a decline from baseline in CA19-9 at any point in time after initiation of treatment, whether early or late, was counted as a biomarker response, although a significant proportion of patients who responded did so within the first 2 months of therapy. While we specifically did not require a sustained biomarker response for the purposes of this study, all but seven patients did have a sustained decline in CA19-9 confirmed over two consecutive measurements. Analysis of our data even recategorising those seven patients as nonresponders did not affect the strong correlations found in this study.
Additionally, seven patients (9.2%) had only a solitary CA19-9 measurement at baseline. While we recognise the inherent bias in including these patients in our analysis (follow-up measurements were generally not obtained on these patients because of rapid disease progression and/or clinical deterioration), we ultimately decided to count them as biomarker nondecliners. Again, reanalysis of our data if we excluded these patients did not affect our study results in any way. Based on the results of our analysis, a rising/nondeclining CA19-9 appears to be a clear indicator of early progressive disease and to correlate with very poor clinical outcomes. This finding may potentially be used in clinical practice as justification for discontinuing systemic therapy early on.
One limitation of our analysis was the lack of uniformity in treatment and in methodology for assaying CA19-9 levels between different patients, given the retrospective and multi-institutional nature of this analysis. However, we did intentionally select three trials in which gemcitabine was consistently administered by FDR infusion, and we believe that the addition of low-dose cisplatin in two of the three trials should not have any substantial impact on our findings. Additionally, each clinical trial did attempt to maintain intrasubject consistency in terms of the laboratory where each subject's serial CA19-9 levels were measured. In the future, clinical studies incorporating CA19-9 measurements should use a standardised laboratory assay to ensure reliable results. Ideally, approval of this test by the Food and Drug Administration for monitoring patients on systemic therapy would further enhance these efforts.
As all the clinical trials from which data were extracted did not include objective response as an end point, we were only able to use unconfirmed response data recorded by study investigators in two of the three trials to correlate CA19-9 responses to radiographic responses. Some might argue that objective responses are the truest indicator of therapeutic activity. Nonetheless, the outcome variables we chose for this study, particularly overall survival, are the most relevant and clinically meaningful in terms of deciding whether a particular agent is worthy of further study or approval. We chose TTF rather than time to time to tumour progression as the other major outcome variable to examine because a number of patients on these trials discontinued study treatment for reasons other than disease progression; thus, censoring these data at those time points would have diluted our numbers substantially.
This analysis demonstrates that declines in serum CA19-9 levels of at least 25, 50, and 75% during treatment with a FDR gemcitabine-containing regimen all correspond with improved patient outcomes. Despite the relatively small numbers in this analysis, the highly statistically significant findings indicate a strong correlation between CA19-9 response and both survival and TTF. When clinical trials report biomarker response data, they generally use 50% as the threshold as an indicator of success. Our analysis suggests that a 25% decline in CA19-9 may provide adequate evidence for the clinical efficacy of a new therapeutic agent or combination treatment strategy. Furthermore, grouping the biomarker response by quartile demonstrates that greater declines in CA19-9 are associated with improved clinical outcomes. While this may be an intuitively obvious concept, ours is the first study to provide conclusive evidence to support a clear and direct correspondence between the degree of biomarker decline and how well patients fare with their disease.
In conclusion, serial CA19-9 measurement represents a useful prognostic tool in patients receiving chemotherapy for advanced pancreatic cancer, with at least a 25% threshold in decline from baseline correlating well with improved patient outcomes. While our analysis was limited to patients receiving an FDR gemcitabine-based regimen, there is little reason to suspect that our findings would not be broadly applicable to other systemic therapies. Thus, monthly CA19-9 measurements appear to be justified for predicting outcomes to therapy and tailoring treatment decisions, and should be considered as a surrogate end point in clinical trials for the selection of new treatments.