Because pancreatic cancer has no early signs or symptoms, it tends to be diagnosed at an advanced stage; 80% to 90% of patients present with disease that is not curable by surgery alone due to spread of the cancer to adjacent tissues or beyond . Metastatic pancreatic cancer has proven to be one of the most chemotherapy-resistant tumors, and, until recently, it has been difficult to demonstrate any beneficial impact of systemic therapy on this disease. Sobering testimony to this is the fact that pancreatic cancer has the lowest 5-year survival rate (3%) of any cancer listed in the Surveillance, Epidemiology, and End Results (SEER) data base of the National Cancer Institute (NCI) . A brief review of clinical trial results prior to 1995 will serve to place more recent results in proper perspective.
Phase II Trials
A wide range of response rates can be found in single-agent phase II trials in patients with advanced pancreatic cancer (as reviewed by Rothenberg et al) . In evaluating these data, it is important to keep in mind that clinical trial methodology and criteria for objective response have evolved over time. Phase II trials conducted in the 1970s routinely included patients with different tumor types in a single trial. As a result,response rates reported in these studies were often based on a small number of patients with that particular cancer and were therefore subject to a high degree of statistical uncertainty. Trials performed prior to 1985 relied heavily on estimation of tumor size by physical examination, with responses defined as shrinkage of a palpable abdominal mass by 50% or more or a reduction in palpable liver span by 30% or more .
The inherent inaccuracy of these techniques, intraobserver and interobserver variability, and the influence of confounding factors on the size of the measured lesions all contributed to the initial reports of high response rates for drugs such as fluorouracil(Drug information on fluorouracil) (5-FU), chlorambucil(Drug information on chlorambucil) (Leukeran), and mitomycin(Drug information on mitomycin) (Mutamycin), as well as the failure to confirm these promising response rates in subsequent trials, especially when CT scans were used to determine tumor response. It is instructive to review how these factors have led to a substantial decrease in response rates reported for 5-FU during the past 35 years.
The earliest reports on the activity of 5-FU in pancreatic cancer showed response rates as high as 56% . Subsequent trials, involving larger numbers of patients and relying on CT imaging to determine response, reported response rates of 0% to 19%, even when biochemical modulation or infusional schedules of drug administration were utilized [5-9]. In effect, no drug has met the criteria for significant antitumor activity (ie, an objective response rate of 20% or more) using radiographic definitions of response in the post-CT era.
Phase III Trials
Two kinds of comparative studies have been conducted in patients with advanced pancreatic cancer: those that compared active treatment to best supportive care (to determine whether chemotherapy made any difference in the outcome of these patients) and those that compared multiagent regimens to single-agent chemotherapy (to determine whether combinations of drugs with distinct mechanisms of action could improve outcome over that achieved with single agents). Of the three trials that compared active treatment to best supportive care, two demonstrated no significant difference [10,11], while one generated very provocative results suggesting a substantial survival advantage in favor of a five-drug regimen (Mallinson regimen) . A confirmatory trial was never done, and subsequent trials of the regimen failed to replicate the impressive results of this trial .
In fact, phase III trials comparing this or other promising multiagent regimens (such as streptozotocin, mitomycin, and 5-FU [SMF] or 5-FU, Adriamycin, and mitomycin [FAM]) to single-agent 5-FU failed to demonstrate any advantage of multiagent chemotherapy over 5-FU alone [13,14]. Despite a number of promising preclinical leads and provocative phase II study results, virtually no progress has been made in chemotherapy for metastatic pancreatic cancer during the past 30 years.
Between 1991 and 1994, 25 investigational new drugs were evaluated in phase II trials for the treatment of pancreatic cancer. The median response rate in these trials was 0% (range, 0% to 14%) and the median survival was 3 months . One trial conducted during this period focused on gemcitabine(Drug information on gemcitabine) (2',2'-difluorodeoxycytidine [Gemzar]). Although the objective response rate to this drug was only 11% and median survival was 5.6 months, several important observations were made in this trial . The 1-year survival rate was surprisingly high (23%), as the responses observed were durable (more than 4 to more than 20 months). The most striking aspect of this study, however, was the impact of gemcitabine on tumor-related symptoms. Of the five patients who had an objective response to gemcitabine, four were able to resume normal daily activities. Three of the five patients were also able to reduce their daily consumption of analgesics. An additional 14 patients who did not meet radiographic criteria for response experienced disease stabilization for 4 months or more, and, of these, 9 had an improvement in performance status.
These results raised the question, how could a cytotoxic agent produce any beneficial effect without substantial tumor shrinkage? Several possible explanations have been proposed. First, gemcitabine could have been acting merely as an analgesic. This explanation was not felt to be likely since analgesic requirements tend to increase as the patient develops tolerance, the tumor grows, and the pain becomes more severe. Patients in this trial experienced significant alleviation of tumor-related symptoms on a stable dose of gemcitabine, and that effect was not lost during the 1 week out of every 4 in which the patient did not receive chemotherapy.
Another possible explanation was that gemcitabine may have interfered with neurotrophic growth factors that are essential for the spread and invasion of the tumor into the celiac plexus, which lies directly behind the pancreas. Evidence to support this theory is currently lacking.
Alternatively, minimal tumor shrinkage (ie, less than the 50% shrinkage required to qualify as an objective response) may have been sufficient to relieve the compression or invasion of the celiac plexus by the tumor. In other words, the standard paradigm for the determination of antitumor activity may not be well suited for pancreatic cancer.
Further examination suggests that tumor shrinkage may be an especially inappropriate way of determining the antitumor effect of cytotoxic therapy in pancreatic cancer. It is important to recognize our current limitations in the noninvasive imaging of pancreatic cancer. Standard CT scanning does not reliably distinguish a pancreatic tumor from normal pancreatic tissue, tumor desmoplasia, or pancreatitis. Incomplete opacification of the small intestine makes it difficult to distinguish tumor from gut, as well. As a result, inclusion of noncancerous tissue may significantly overestimate tumor shrinkage (as in the case of inflammatory tissue being included in the tumor measurement) or underestimate tumor shrinkage (if fibrotic tissue was included in the tumor measurement).