Data from Sweden1 as well as from the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) program2 indicate an improvement in relative survival rates (RSRs) for patients diagnosed with both adenocarcinoma and squamous cell esophageal cancer over the last 3 decades. To provide further information on the issue, we considered trends in survival by histotype of esophageal cancer for cases diagnosed between 1974 and 1998 from the cancer registries of Vaud and Neuchâtel in the French-speaking part of Switzerland.3, 4 Information on survival has been published only for the Vaud Cancer Registry until 1993 for all esophageal cancers combined. Overall, 5-year relative survival rates increased in males from 0.08 in 1974–8 to 0.10 in 1989–93, and in females from 0.06 to 0.08.5 We decided therefore to consider trends in survival for both histologic types of esophageal cancer to 1998 for the registries of Vaud and Neuchâtel combined. The data were abstracted from the Vaud and Neuchâtel Cancer Registries files, which include incident cases of malignant neoplasms in the cantons, whose populations, according to the December 2000 census, were about 620,300 and 165,700, respectively. Information collected comprises general demographic characteristics of each case (age, sex, municipality of residence), site and histologic type of the tumor according to standard International Classification of Diseases for Oncology (ICD-O)6 and time of diagnostic confirmation. The present report includes 810 squamous cell carcinomas (ICD-O-1 M codes 8070-8076) and 221 adenocarcinomas (ICD-O-1 M 8140-8570) of the esophagus registered from 1974 to 1998.7 Information on survival is integrated from mortality statistics into the incidence data file and, for patients who are apparently alive, through an active follow-up based on verification of vital status from registries of current residence. The vital status of each case has been verified up to 30 December 2002. Relative survival rates were computed after allowance for the general life tables of the cantons. No information was available on tumor stage or treatment. Table I gives the number of registered cases and 1-, 3- and 5-year RSRs in 5 subsequent calendar periods (from 1974–8 to 1994–8) for squamous cell carcinomas, and in 4 calendar periods for adenocarcinomas, due to small absolute numbers in earlier calendar years. For squamous cell cancer, RSRs steadily increased from 25.1% to 46.7% at 1 year, from 7.6% to 19.7% at 3 years and from 5.2% to 13.9% at 5 years. Due to smaller absolute numbers, the pattern was less consistent for adenocarcinomas, but an improvement over calendar years (from 27.7% to 37.0% at 1 year, from 16.4% to 22.2% at 3 years, from 12.9% to 21.0% at 5 years) was observed for adenocarcinomas, too. Thus, the present analyses confirm the existence of appreciable improvements in RSRs for both histologic types of esophageal cancer over the last 3 decades.1, 2 Possible explanations of these findings include earlier diagnosis and detection through wider adoption of endoscopy for squamous cell and also for adenocarcinomas potentially related to Barrett's esophagus and gastroesophageal reflux,8 as well as potential advancements in (surgical) treatment of the disease.9, 10 Yours sincerely, The contribution of the Vaud and Neuchâtel Cancer Registry's staff is gratefully acknowledged. Fabio Levi, Van-Cong Te, Lalao Randimbison, Carlo La Vecchia

Sir In a combined analysis of the Nurses' Health Study and the Health Professional Follow-up Study, on a total of 206 women and 143 men with cancer of the pancreas, Schernhammer et al, (2002) did not find an increased risk of pancreatic cancer in relation to history of gallstones or cholecystectomy after adjusting for potential confounding factors. The issue of a possible association between gallstones or cholecystectomy and cancer of the pancreas is, however, still open to discussion, because several investigations reported an excess pancreatic risk in patients with gallstones. Apart from the papers quoted in Schernhammer et al, (2002), some excess risks were found in cohort studies from the United States (Bansal and Sonnenberg, 1996), Denmark (Johansen et al, 1996), and Sweden (Ye et al, 2001), and case–control studies from the United Kingdom (Cuzick and Babiker, 1989), Greece (Kalapothaki et al, 1993), and Israel (Schattner et al, 1997). The strength of the association, however, was variable across studies, and different potential confounding factors were not always taken into account. In order to provide further information on the issue, we updated the analysis of a case–control study conducted in Italy between 1983 and 1992 (La Vecchia et al, 1990). Briefly, the study included 362 patients from the major teaching and general hospitals in Greater Milan with incident, histologically confirmed pancreatic cancer (229 men, 133 women, median age 59 years), and 1552 controls (1141 men, 411 women, median age 55 years) admitted to the same network of hospitals for acute, non-neoplastic conditions, unrelated to alcohol or tobacco consumption (33% traumas, 17% nontraumatic orthopedic conditions, 36% acute surgical diseases, and 14% other miscellaneous disorders). Less than 3% of cases and controls approached refused the interview. Trained interviewers identified and questioned cases and controls using a structured questionnaire, including information on education and other socioeconomic factors, anthropometric measures, general lifestyle habits, such as tobacco and alcohol consumption, and a few selected indicator foods. The patients were also asked if they had a diagnosis of selected medical conditions, and the age at first diagnosis was recorded. Odds ratios (OR) and corresponding 95% confidence intervals (CI) were estimated using unconditional multiple logistic regression models, including terms for age, education, tobacco consumption, body mass index, and history of diabetes. Table 1 gives the distribution of pancreatic cancer cases and controls, and the corresponding ORs, according to history of cholelithiasis. Subjects with a history of cholelithiasis showed no increased risk of cancer of the pancreas (OR=1.03, 95% CI=0.67–1.59). The OR was 1.33 (95% CI=0.74–2.40) for subjects with a diagnosis of cholelithiasis less than 10 years before interview, and 0.80 (95% CI=0.43–1.50) for diagnosis 10 or more years before. Table 1 Relation between pancreatic cancer and history of cholelithiasis among 362 cases and 1552 controls (Milan, Italy, 1983–1992) Thus, our findings are consistent with those of the Nurses' Health Study and the Health Professional Follow-up Studies (Schernhammer et al, 2002), and indicate that cholelithiasis is not materially associated with pancreatic cancer risk after major identified confounding factors have been considered. A modestly increased risk was observed 10 years after a diagnosis of cholelithiasis, but no greater excess risk can be found 10 or more years after. Thus, if any association exists, it is unlikely to be causal. The apparent association reported from several case–control studies can at least in part be because of a more accurate recall of gallbladder disease by pancreatic cancer patients. In our study, however, information on medical history proved satisfactorily reproducible (Bosetti et al, 2001), indicating that recall bias is unlikely to have played a major role. Other potential biases of this study should be limited, given the almost complete response rate, the administration of a standard questionnaire under similar conditions, and the same catchment area for cases and controls.

Dietary fibre has been reported to protect from several neoplasms, but the issue remains controversial. No previous study considered in depth the topic of fibres and prostate cancer. A multicentre case-control study was conducted in Italy from 1991 to 2002, including 1,294 men with incident, histologically confirmed prostate cancer and 1,451 controls admitted to the same network of hospitals as cases with acute nonmalignant conditions. Multivariate odds ratios (OR) and 95% confidence intervals (CI) were obtained after allowance for major identified confounding factors, including total energy intake. Compared to the lowest quintile, the OR of prostate cancer for the highest quintile of total fibre intake was 0.93 (95% CI 0.71-1.22). The risk was inversely related with soluble fibre (OR = 0.89, 95% CI 0.78-1.02, for a difference between 80th and 20th percentile), cellulose (OR = 0.88, 95% CI 0.78-1.01) and vegetable fibre (OR = 0.82, 95% CI 0.73-0.93). These relationships were consistent across strata of age, family history of prostate cancer, body mass index and education. Vegetable fibres appear, therefore, to have a favourable association with prostate cancer risk.