Introduction

Top Abstract Introduction Patients and methods Results Discussion References

Pancreatic cancer is difficult to diagnose and unsatisfactory to treat, with most patients dying within 6 months of diagnosis and virtually all by 1 year. ^{1 2} At present surgery offers the only prospect of longer survival.

Testosterone may have a positive effect on the growth of pancreatic carcinoma. This would be supported if specific androgen receptor blockade improved survival. The concept is supported by the presence of androgen receptors within human pancreatic cancer tissue,³ together with the enzymes, aromatase and 5-reductase, which convert testosterone into either oestradiol or a more active androgen, 5-dihydrotestosterone, respectively.⁴ Furthermore, all patients with pancreatic cancer have low serum testosterone concentrations. ^{5 6}

Confirmatory evidence for this central role of testosterone came with the demonstration of its growth potentiating action on human pancreatic adencocarcinoma xenografts grown in nude mice, together with the inhibiting action of an antiandrogen.⁷

	Patients and methods

Top Abstract Introduction Patients and methods Results Discussion References

The trial was conducted at Hinchingbrooke Hospital, Huntingdon, between October 1993 and January 1997 and was approved by the appropriate ethical committee. This is a district general hospital serving a population of about 170 000. Normally, such a hospital would see about 15 patients a year with pancreatic cancer. This hospital, however, is a referral unit for endoscopic retrograde cholangiopancreatography (ERCP) for a wide area of East Anglia, Hertfordshire, and Bedfordshire, which enabled the inclusion of 49 patients in the study over a 3 year period. All patients informed about the trial and who met the inclusion criteria readily accepted the opportunity to participate. Only one patient stopped treatment during the conduct of the trial because he wanted to know to which treatment he had been allocated; he is still included in the analyses because of intention to treat. The pancreatobiliary unit is relatively new, and normally I would perform a pancreatoduodenectomy for either carcinoma of the pancreas or ampulla of Vater at a rate of about one every 2 months. The criterion for resection for carcinoma of the pancreas was that the tumour was about 1 cm in size, on the basis of computed tomography, with no signs of spread at laparotomy.

Forty nine consecutive patients (26 men, 23 women; median age 68 (range 47-85) years) with a clinical diagnosis of pancreatic carcinoma, whatever the tumour load, who satisfied the inclusion criteria were entered into the study. They were numbered from 1 to 49 after independent randomisation. Before the initiation of the study a random code was generated by an independent company (Til Occam). This code was used to assign treatment on the basis of the order of patients admitted to the trial; study medications were blinded to both investigator and patient through the use of placebo. As the trial progressed, however, some men in the flutamide treatment group became aware of their treatment group because of side effectsnamely, breast tingling and gynaecomastia; this had been discussed with all patients at the start of the trial.

Diagnosis and treatment
The diagnosis of cancer was based on the typical but non-specific history of jaundice, weight loss, and upper abdominal or back pain, or both, in the absence of a significant intake of alcohol with two or more of the following findings at investigation: isolated pancreatic mass demonstrated on enhanced computed tomography; typical low bile duct stricture at endoscopic retrograde cholangiopancreatography (40 patients), usually with an isolated pancreatic duct stricture, the double duct sign; and histological confirmation by open biopsy or analysis of ascitic fluid (17 patients). Patients with macroscopic tumour at the ampulla of Vater or duodenum, which could be mistaken for primary tumours of these structures, were excluded from the study. Forty patients had a tumour within the head of the gland and nine in the body (table 1).

Power and analysis
The study was designed to test the null hypothesis that there would be no difference in survival time according to treatment against an alternate hypothesis of longer survival with flutamide treatment. This trial had 90% power to detect a difference if the 1 year survival rate with flutamide is between 40% and 50% (with 25 in each group, keeping the risk of a false positive at 5%, and knowing that the 1 year survival for patients with cancer of the pancreas is <10% at present). Any increases in survival would improve the power of the trial to determine a difference. A previous pilot study with six patients showed a 1 year survival of 100%; none of those patients were included in this study.

	Results

Top Abstract Introduction Patients and methods Results Discussion References

An intention to treat analysis was carried out which included all 49 patients recruited to the study. A per protocol analysis was also carried out for the 36 patients who received more than 6 weeks of study treatment. This cut off period, which applied to all patients, was chosen to exclude those 13 (eight in the flutamide group, five in the placebo group) with a heavy tumour load, including liver metastases and ascites, who died within this period and who would probably not have taken sufficient medication for an effect.

View larger version (15K): [in this window] [in a new window]   Fig 1.   Survival time of all patients with pancreatic cancer according to treatment group

View larger version (15K): [in this window] [in a new window]   Fig 2.   Survival time of patients with pancreatic cancer who received over 6 weeks' treatment according to treatment group

The survival times were calculated and missing data were analysed as censored at the time of the latest available follow up data. Kaplan-Meier estimates of the times at which half patients remained alive were calculated. Treatments were compared by the log rank and generalised Wilcoxon tests.

Survival time for all patients was analysed first, and at final analysis two patients were still alive in the flutamide group (8%) and none in the placebo arm (table 2). The median survival for the treatment group was 226 days (8/12) compared with 120 days (4/12) for the placebo arm (fig 1). Comparison of treatment by the generalised Wilcoxon test gave P=0.079 and by the log rank test P=0.01. The slight difference between the results from the two tests was because the log rank test gives greater emphasis to treatment differences that occur at later time points, which is where the most pronounced differences are seen in these data.

For those patients receiving more than 6 weeks' treatment the median survival times were 350.5 days (12/12) for those treated with flutamide versus 165.5 days (5/12) for those receiving placebo (fig 2). Again, treatment comparisons by generalised Wilcoxon test gave a value of P=0.001, while the log rank test gave P=0.001 (table 3).

Statistical analysis of changes in tumour size on the basis of computed tomography did not show a significant difference between the groups when they were assessed from 0 to 3 months. Later analysis was more difficult as several patients had died, reducing numbers for comparison. There was no significant difference in survival between men and women in the treatment group.

At final analysis the side effects of treatment experienced by the patients on flutamide could be grouped into three. Firstly, there was the effect on the male breast, which was mild tingling in 55%, with overt gynaecomastia in three patients. Secondly, there was gastrointestinal disturbancenamely, diarrhoeawhich was self limiting and did not necessitate stopping treatment in the eight patients affected. Thirdly, one patient developed paraesthesianamely, a tingling sensation in all limbs which stopped on cessation of treatment.

	Discussion

Top Abstract Introduction Patients and methods Results Discussion References

This study has shown a doubling of survival for patients taking flutamide and provides support for the central role of testosterone as a growth factor in pancreatic cancer.

In a previous study with the antiandrogen cyproterone acetate no clinical effect was shown.⁸ This may be because it has inherent androgenic activity.⁹ Animal and human studies with analogues of luteinising hormone releasing hormone, which reduce serum testosterone concentrations, however, have shown promising effects in inhibiting tumour growth. ^{10 11} After the demonstration of oestrogen receptors within pancreatic cancer tissue¹² several studies were performed with tamoxifen, but none showed a consistent positive effect. ^{13 14} In this study the sex difference was hardly noticeable at 1.1:1 (men:women). Normally this is somewhere about 1.25-1.4:1. This probably reflects the wide referral nature of the patients.

There was no significant difference with regard to changes in tumour size assessed at baseline and at 3 months on computed tomography, the main problem being the reduction in patient numbers as time progressed. It was notable on the scans of several patients treated with flutamide, however, especially in those who survived longest, that the size of the tumours did not change until a terminal stage was reached, when the tumour would rapidly expand. I have interpreted this as the tumour possibly escaping from androgen blockade; this effect is seen with prostatic carcinoma.¹⁵

There were no consistent findings with regard to weight changes or performance status, many patients gaining weight terminally when ascites appeared. The Karnofsky index was used but the main problem is its objective bias. At the start of treatment 80% of patients were considered as between able to carry on normal activities and self caring but unable to do active work (Karnofsky scale 90-70). A fifth required considerable assistance and frequent medical care (Karnofsky scale 50-40). No patient was initially considered as terminal, even though 13 patients deteriorated and died within 6 weeks. All were enthusiastic about participating in the trial. Protocols of measurements of quality of life were not available at the start of this trial.

Reliability of diagnosis
Although histological confirmation of diagnosis was obtained in 17 patients, computed tomography¹⁶ and endoscopic retrograde cholangiopancreatography¹⁷ each give a diagnostic accuracy in excess of 90%, the latter also excluding carcinomas of the ampulla of Vater and periampullary region. The differential diagnosis of pancreatic carcinoma may include cholangiocarcinomas of the lower third of the common bile duct, but this is a distinction that can often be made only on careful dissection of resection specimens. Secondary carcinomas and lymphomas are extremely uncommon, and other solid (that is, non-cystic) pancreatic malignancies account for less than 3-5% of unselected series. Fine needle aspiration biopsy sampling, which we introduced to the unit after the start of this study, may provide a less traumatic and relatively accurate option for diagnostic confirmation, notwithstanding a false negative rate and difficulty in classification of tumours of this region.