Increased incidence of pancreatic neoplasia in pernicious anemia

World J. Surg. 12, 866-870, 1988

World Journal of Surgery 9 1988 by the Soci~t~ Internationale de Chirurgie

Increased Incidence of Pancreatic Neoplasia in Pernicious Anemia K u r t B o r c h , M . D . , P h . D . , Eric K u l l m a n , M . D . , Stefan Hallhagen, M . D . , T o r b j 6 r n Ledin, M . S c . , and I n g e m a r Ihse, M . D . , P h . D . Departments of Surgery and Mathematical Statistics, University Hospital of Link6ping, LinkOping, Sweden. In a Swedish population, 361 patients with pernicious anemia were followed closely during a 7-year period with regard to the occurrence of all forms of malignant neop|asia. In addition to an increased incidence of gastric neoplasia (0.6%/year), there seemed to be an increased incidence of pancreatic neoplasia (0.3%/year) in the series. With reference to ageand sex-specific incidence rates in the general population, the number of malignant pancreatic tumors observed was significantly higher than expected (p < 0.02, Poisson analysis). Pancreatic malignancy was the primary cause of death in 5 (4%) of 134 patients who died during follow-up. Among 127 unselected patients with malignant pancreatic neoplasia, the prevalence of pernicious anemia was 3%. These findings indicate that there is a linkage between the two diseases which has not been previously recognized. Atrophic gastritis with hypergastrinemia may be the key to this linkage.

It is well established that achlorhydric atrophic gastritis associated with pernicious anemia (PA) is a premalignant condition with regard to gastric neoplasia development [1-6]. In addition to metaplastic and proliferative mucosal changes, pathologic colonization with bacterial species able to enhance N-nitrosamine formation are factors considered to be important in the carcinogenesis of the achlorhydric stomach [7-9]. In theory, such carcinogenic gastric content would pass relatively unchanged into the upper intestinal tract during the fasting state. The amount of gastric carcinogenic compounds entering the splanchnic circulation is unknown. However, any clinically relevant effect of such carcinogenic material on organs other than the stomach would be reflected in an increased incidence of neoplasia of that organ among patients with PA. None of several thorough studies, of which most are of an early date, has indicated an increased incidence of cancer of digestive organs other than the stomach in PA. In light of these considerations, the lack of recent follow-up studies concerning the incidence of all forms of neoplasia in PA, and the fact that the incidence pattern for cancer of digestive organs is changing [10--12], we found it relevant to extend a gastroscopic survey in patients with PA to include a close follow-up concerning all types of malignant neoplasia. Reprint requests: Kurt Botch, M.D., Department of Surgery, University Hospital, S-581 85 Link6ping, Sweden.

Material and Methods

Patients All patients with PA, alive by January l, 1979, in two neighboring Swedish health care districts were identified (in one district with the aim of initiating a gastroscopic survey) [5, 13]. By scrutinizing the registers of local chemists for vitamin BI2 prescriptions, hospital files, questionnaires, and, in cases of doubt, submitting patients to supplementary examination with a Schilling test, individuals without verified vitamin BI2 deficiency and those with vitamin Bi2 deficiency other than PA (gastric resection, ileal resection, malabsorption, etc.) were excluded. Patients with PA who were previously treated for gastric neoplasia were not included. The diagnostic criteria for PA included macrocytic anemia, characteristic bone marrow, subnormal levels of vitamin B~2 in serum, and improvement of symptoms on vitamin B~2 treatment. The diagnosis was further strengthened with a Schilling test, a pentagastrin test, or both, in a large proportion of the series [13]. By January 1, 1979, a total of 253 patients with PA were living in the two districts. From 1979 to 1984, a total of 108 new cases of PA were included. In the whole series of 361 patients, the mean age at the time of entry in the study was 73 years (range, 27-96 yr). The male to female ratio was 0.51. One hundred and thirty-four patients died during the period 1979-1985. Malignant neoplasia had been diagnosed prior to follow-up in 19 of the 361 patients (chronic lymphatic leukemia: 1, tonsil: 1, urinary bladder: 2, kidney: 1, uterus: 5, uterus and malignant melanoma: 1, prostate: 1, breast: 3, colon-rectum: 4) of whom all except one with chronic lymphatic leukemia and one with tonsillary cancer seemed to have been radically treated. In the statistical analysis concerning any specific type of cancer, patients who previously had that type of cancer were excluded. Of the 361 patients, 139 were screened with gastroscopy. The latter group was not included in statistical evaluations involving neoplasia of the esophagus or stomach. Registration of Death or Neoplasia During follow-up, until December 31, 1985, the registers of the Regional Oncologic Center and the National Cancer Registry (National Board of Health and Welfare, Stockholm) were

K. Borch et al.: Pancreatic Carcinoma

867

Table 1. Distribution,

according to primary site, of malignant neoplasias of digestive organs as the cause of death among 134 patients with pernicious anemia (PA) from 1979to 1985 (age at death, 55 years or more), and among 82,781 deaths in the general Swedish population during 1982 (age at death, 55 years or more) [13]. Patients with PA

Table 2. Distribution, according to primary site, of malignant

neoplasias diagnosed among 361 patients with pernicious anemia (age, 59 years or more). Corresponding distribution, according to primary site, in the general Swedish population aged 55 years or more is given parenthetically [11].

General population

Cause of death (neoplasia o f . . . )

N

% of total

N

% of total

Esophagus Stomach Small intestine Colon Rectum Liver (primary) Biliary passages (including gallbladder) Pancreas Digestive organ (unknown primary site) Total digestive organs

0 5 0 3 0 1 1 5 2 17

0 3.7 0 2.2 0 0.8 0.8 3.7 1.5 12.7

300 1,445 37 1,498 764 338 644 1,303 72 6,401

0.4 1.8 0.0 1.8 0.9 0.4 0.8 1.6 0.1 7.8

screened annually for histologically verified cancer registrations in the series. All deaths were recorded from the Regional Bureau of Statistics, and the cause of death and autopsy findings were extracted from death certificates and autopsy reports. Data concerning unselected patients with histologically verified gastric or pancreatic neoplasia in the largest of the two districts were collected from hospital registers and the Regional Oncologic Center. These data were cross-checked with the local PA series. The mean incidence rates in the local area for these neoplasias during the follow-up period were identical to those for the whole of Sweden during 1982 [12].

Statistical Analysis The general population of the whole of Sweden was used as reference group in the statistical evaluation [14]. The age(5-year groups) and sex-specific cancer incidences in the latter during 1982 were used as a basis for estimating individual patient risks of acquiring specific types of cancer. The total risk for each patient was calculated by summing the risks for each year of participation in the study. In cases of entry after January 1, 1979, it was assumed that the patient had participated half the year of entry. The principle of half-year correction was also adopted with regard to the year of death in patients who died of causes unrelated to the specific neoplasia studied. If, on the other hand, the patient died of the specific neoplasia studied, the total follow-up period (until December 31, 1985) was considered, irrespective of the year of death. The number of specific neoplasias expected to occur in the series were calculated by summing the individual risks of acquiring that neoplasia. Poisson analysis was used to evaluate the probability that the observed number of cases with neoplasia was due to chance only, and not to an association with PA. The Poisson analysis is valid considering the small risks for each patient and the sufficiently large number of patients studied.

All patients

Nonscreened patients~

Malignant neoplasia o f . . .

N

%

N

Esophagus Stomach Small intestine Colon Rectum Liver (primary) Biliary passages (including gallbladder) Pancreas Digestive organ (unknown primary site) Total digestive organs except esophagus and stomach Digestive organs Other organs ~ Total

-

-

%

0 5

0 (1.0) 20.8 (5.7)

0 4 1 1 1

0 (0.5) 13.8 (9.0) 3.4 (5.3) 3.4 (1.6) 3.4 (2.4)

-

-

5 2

17.2 (4.4) 6.9 (1.1)

-

-

14

48.3 (24.3)

-

-

15 9 24

62.5 (29.4) 37.5 (70.6) 100.0(100.0)

15 29

51.7 (75.7) 100.0(100.0)

aOf the 361 patients, 139 were screened with gastroscopy. Frequencies involving esophageal or gastric neoplasia were considered in the 222 nonscreened patients. ~ verae, bone, skin (melanoma), kidney, lung, breast, uterus, ovary, prostate.

Results

Mortality In 24 (18%) of the 134 patients who had died by December 3i, 1985, the primary cause of death was histologically-verified malignant neoplasia which, in 17 cases, emanated from a digestive organ (Table 1) and, in 7 cases, from other organs [skin (melanoma), tonsil, lung, kidney, uterus, ovary]. As seen in Table 1, there was an increased frequency of deaths caused by neoplasia of digestive organs in PA. Deaths due to gastric and pancreatic neoplasia were the main cause of this increase.

Incidence of Malignant Neoplasia The distribution, according to primary site, of neoplasias diagnosed among all 361 patients is given in Table 2. It is apparent that the fraction of patients with pancreatic neoplasia (4 adenocarcinomas and l malignant carcinoid) as well as the fraction of patients with gastric neoplasia (all adenocarcinomas) is increased in comparison with the general population. With knowledge of all deaths and newly diagnosed cases (1979-1984), the size of the patient population at the entry of each year and, thus, the annual incidence of malignant neoplasia of digestive organs could be calculated (Table 3). The table illustrates that even the maximum incidence rates in the general population (age, 85 yr or more) were far below those in patients with PA. Using age- and sex-specific risks derived from the general population [12] and accumulated for the individual number of years studied in the whole series of 361 patients, the expected number of malignant pancreatic neoplasias was 1.3. The differ-

868

World J. Surg. Voi. 12, No. 6, December 1988

Table 3. Incidence of malignant neoplasias of digestive organs in unselected patients with pernicious anemia (PA) (age, 28-96 yr) alive at the beginning of each year from 1979 to 1984." Incidence in the general population in 1982 Malignant

Annual incidence in PA d u r i n g . . .

neoplasia o f . . .

1979

1980

1981

1982

1983

Stomach b Pancreas Digestive organs r

1/150 0/275 2/148

1/159 1/289 2/156

1/159 0/301 3/156

0/157 1/289 2/155

1/147 1/283 2/146

1984

Mean annual incidence in PA (%)

Age -~ 25 years (%)

Age -> 85 years (%)

1/134 2/267 2/133

0.56 0.30 1.45

0.03 0.02 0.18

0.19 0.12 0.88

qncidence rates in the general population of Sweden during 1982 (age, 25 yr or more; and age, 85 yr or more) [1 l] are given for comparison. blncidence among patients who were not screened with gastroscopy. CAs in b and excluding 3 patients treated for colorectal cancer prior to follow-up.

ence between the latter and the 5 observed was statistically significant (p < 0.02). Considering the 253 patients who were followed from January 1, 1979, during the whole 7-year period, the difference between the expected and observed number of cases was also significant (p < 0.005). When the patient with malignant pancreatic carcinoid was excluded, the difference between the observed and expected numbers of pancreatic adenocarcinomas was also significant, both in the total series (p < 0.05) and among patients followed during the whole 7-year period (p < 0.03). Among the 222 patients who were not endoscopically screened, the expected number of malignant gastric neoplasias was 1.1, which differed significantly from the 5 observed (p < 0.005). The number of malignant neoplasias of digestive organs expected to occur among 219 nonscreened patients without previous neoplasia of digestive organs was 5.4, which differed significantly from the 15 observed (p < 0.001). From 1979 to 1985, a total of 127 new cases of pancreatic malignancy were diagnosed in the largest of the 2 districts studied. Four of the 5 patients with PA and pancreatic neoplasia were living in this district and, accordingly, the prevalence of PA among unselected patients with pancreatic malignancy was 3%. Discussion

The relationship between PA and gastric neoplasia has been studied extensively and, no doubt, these patients are at special risk of developing gastric neoplasia [1-6, 15]. The majority of studies concerning the relationship between PA and gastric neoplasia were published from the 1930's to the 1960's. During the latter part of this period and further on, the incidence of gastric cancer has gradually decreased and the incidence of pancreatic cancer gradually increased in the Western world [10-12]. According to recent fiberendoscopic surveys, the increased prevalance of gastric neoplasia in PA seems to be unchanged [5, 16, 17], despite a decreasing incidence in the corresponding general population [11, 12]. This is in agreement with the observation in nonscreened patients in the present study. None of many thorough studies has indicated the existence of an increased incidence of malignant neoplasia of digestive organs other than the stomach in PA [15]. Demmler found that carcinoma of the pancreas was the cause of death in 3 (1.3%) of 238 patients with PA [18], and Strandell and Jansson observed one case of pancreatic cancer among 686 patients [19]. In the present series, pancreatic and gastric neoplasias constituted

29% (10/34) of all symptomatic malignant neoplasias observed. In the general population (age, 55 years or more), these 2 neoplasias accounted for 10% of all malignant neoplasias [12]. It is unlikely that this increased incidence of pancreatic neoplasia in PA is a local phenomenon, considering that the local incidence of pancreatic neoplasia did not differ from that in the whole country [12]. Since it seems unlikely that an association between PA and pancreatic neoplasia has been previously missed or misinterpreted, it seems that the incidence of pancreatic neoplasia in PA has increased concomitantly with the increase in the general population, but to a significantly higher level. In other words, pancreatic carcinogenesis may be facilitated in PA. In this respect, it seems natural to speculate whether the severe atrophy of the stomach is a factor of importance. Carcinogenic compounds generated in the achlorhydric stomach [7-9] may enhance or add to factors involved in pancreatic carcinogenesis in general. Furthermore, it cannot be excluded that the access of carcinogenic substances to the splanchnic circulation and, thus, to the pancreas is facilitated in atrophic gastritis. Patients with antrum-sparing atrophic gastritis have hypergastrinemia of antral origin [20-22]. Whether these patients have deranged circulating levels of other relevant peptide hormones, as for instance cholecystokinin, is not known. Gastrin is a t r o p h i c hormone for gastric fundic epithelial and endocrine tissue [22-24]. It is structurally and phylogenetically related to cholecystokinin [25, 26]. Like cholecystokinin and its analogue, caerulein [24, 27-31], gastrin may exert a trophic effect on the pancreas [30, 32-33]. It is a matter of controversy whether caerulein or cholecystokinin, and, thus, perhaps gastrin, enhance pancreatic carcinogenesis [31, 34-36]. One may, however, speculate whether chronic hypergastrinemia associated with PA is a promoting factor with regard to pancreatic neoplasia development. Clarification of the possible relationship between hypergastrinemic atrophic gastritis and pancreatic neoplasia may contribute to the understanding of pancreatic carcinogenesis and, if so, also to the identification of individuals at special risk of acquiring pancreatic neoplasia. Note added in proof: With the kind aid of Professor J. Rehfeld

(Department of Clinical Chemistry, Rigshospitalet, Copenhagen, Denmark), we have analyzed fasting concentrations of cholecystokinin in plasma by radioimmunoassay in 68 unselected patients with pernicious anemia (included in the present study) and in 13 controls with endoscopically and histologically normal gastric mucosa. With reservation for

K. Borch et ai.: Pancreatic Carcinoma

well-known problems concerning cross-reactivity between antigastrin and anticholecystokinin antibodies, the levels of cholecystokinin seemed to be increased in pernicious anemia (median, 1.1 pmol/1; range, 0-11.9 pmol/1) as compared with normal controls (median, 0.2 pmol/1; range, 0-5.8 pmol/l) (p < 0.02, Mann-Whitney test). Considering possible promoting factors with regard to pancreatic carcinogenesis in severe fundic atrophic gastritis, these preliminary results should focus further research on the profile of hormones with trophic effects on the pancreas. R~sum~

Trois cent soixante et nn patients Su6dois pr6sentant une an6mie pernicieuse ont 6t6 suivis de pros pendant une p6riode de 7 ans pour d6celer l'apparition de cancer, quelque soit son type. Dans cette s6rie, en plus d'une incidence augment6e de n6oplasmes gastriques (0.6% par an), il semble que l'incidence des n6oplasmes pancr6atiques soit 61ev6 aussi (0.3% par an). Le nombre de tumeurs pancr6atiques malignes 6tait sup6rieur b, celui qu ' o n attendait (p < 0.02 par l'analyse de Poisson) compte tenu de l'fige et du sexe des contr61es de la population g6n6rale. Une tumeur maligne 6tait la cause de mort de 5 (4%) des 134 patients d6c~d6s par la suite. Parmi 127 patients non sfilectionn6s ayant une tumeur pancr6atique maligne, la prfivalence d'an6mie pernicieuse 6tait de 3%. Ces donn6es sugg~rent qu'i! existe un rapport entre des deux maladies non reconnues auparavant. La gastrite atrophique avec hypergastrin6mie pourrait ~tre le chainon manquant. Resumen

Ninguno de diversos estudios comprensivos, la mayoria de los cuales no son de reciente publicaci6n, ha indicado una aumentada incidencia de cfincer de los 6rganos del tracto digestivo, diferente de la reconocida mayor frecuencia del cfincer g~strico, en la anemia perniciosa. Trescientos sesenta y un pacientes con anemia perniciosa, diagnosticados en dos vecinos distritos de salud suecos, fueron investigados durante un periodo de seguimiento de 7 afios en relaci6n a la presencia de toda forma de neoplasia maligna. Se encontr6 que ademfis de una aumentada incidencia de neoplasia gfistrica (0.6% por afio)i hay una aparente mayor incidencia de neoplasia pancre~ttica (0.3% pot afio) en esta serie. En comparaci6n con las tasas especfficas de incidenci a por edad y sexo en la poblaci6n general, se encontr6 un mayor nt~mero de tumores malignos del p~mcreas (p < 0.02 en el anfilisis de Poisson). La neoplasia maligna del pancreas constituy6 la primera causa de muerte en 5 (4%) de 134 pacientes fallecidos en el transcurso del seguimiento. En 127 pacientes no seleccionados con neoplasia pancreAtica, la incidencia de anemia perniciosa fue de 3%. Estos hallazgos indican que existe una relaci6n no previamente reconocida entre las dos enfermedades. La gastritis atr6fica con hipergastrinemia bien puede ser la clave de tal relaci6n. References

1. Strandell, B.: Cancer of the stomach and pernicious anemia. Acta Med. Scand. 10[Suppl. 40]:89, 1931 2. Rigler, L.G., Kaplan, H.S.: Pernicious anemia and tumors of the stomach. J. Natl. Cancer Inst. 7:327, 1947

869

3. Zamcheck, N., Grable, E., Ley, A., Norman, L.: Occurrence of gastric cancer among patients with pernicious anemia at the Boston City Hospital. N. Engl. J. Med. 252:1103, 1955 4. yon Knorre, G.,Pechau, K.-G.: Spfitschicksale vonPerniziosakranken. Zeitschr. Inn. Med. 30:701, 1975 5. Borch, K.: Epidemiologic, clinicopathologic, and economic aspects of gastroscopic screening of patients with pernicious anemia. Scand. J. Gastroenterol. 21:21, 1986 6. Botch, K., Renvall, H., Kullman, E., Wilander, E.: Gastric carcinoid associated with the syndrome of hypergastrinemic atrophic gastritis. A prospective analysis of 11 cases. Am. J. Surg. Pathol. 11:435, 1987 7. Sander, J+: Nitrosaminsynthese durch Bakterien. HoppeSeyler's Zeitschr. Physiol. Chem. 349:429, 1968 8. Ruddell, W.S.J., Bone, E.S., Hill, M.J., Blendis, L.M., Walters, C.L+: Gastric juice nitrite. A risk factor for cancer in the hypochlorhydric stomach? Lancet 2:1037, 1976 9. Ruddell, W.S.J., Bone, E.S., Hill, M.J., Walters, C.L+: Pathogenesis of gastric cancer in pernicious anemia. Lancet 1:521, 1978 10. Krain, L.S.: Cancer incidence. The crossing of the curves for stomach and pancreatic cancer. Digestion 6:356, 1972 11. Hayakawa, N., Kurihara, M.: International comparison oftrends in cancer mortality for selected sites. Soc. Sci. Med. 15D:245, 1981 12. National Board of Health and Weltare, The Cancer Registry: Cancer incidence in Sweden 1982, Stockholm, Liber, 1985 13. Botch, K., Liedberg, G.: Prevalence and incidence of pernicious anemia. An evaluation for gastric screening. Scand. J. Gastroenterol. 19:154, 1984 14. National Central Bureau of Statistics: Causes of Death 1982, Stockholm, Liber, 1984 15+ Chanarin, I.: The Megaloblastic Anemias, Oxford, Blackwell Scientific Publications, 1969 16. Stockbrfigger, R., Menon, G.G., Beilby, J.O.W., Mason, R.R., Cotton, P.B.: Gastroscopic screening in 80 patients with pernicious anemia. Gut 24:1141, 1983 17. Borch, K., Renvall, H., Liedberg, G.: Gastric endocrine cell hyperplasia and carcinoid tumors in pernicious anemia. Gastroenterology 88:638, 1985 18. Demmler, K.: Todesursachen behandelter Perniziosapatienten. Med. Klin. 61:575, 1966 19. Strandell, B+, Jansson, T.: Anemia perniciosa och cancer. Nord Med. 14:1316, 1937 20. McGuigan, J.E., Trudeau, W.L.: Serum gastrin concentrations in pernicious anemia. N. Engl. J. Med. 282:358, 1970 21. Fahrenkrug, J., Schaffalitzsky de Muckadell, O.B., Hornum, I., Rehfeld, J.F.: The mechanism of hypergastrinemia in achlorhydria. Effect of food, acid, and calcitonin on serum gastrin concentrations and component pattern in pernicious anemia, with correlation to endogenous secretin concentrations in plasma. Gastroenterology 71:33, 1976 22. Borch, K., Renvall, H., Liedberg, G., Andersen, B.N.: Relations between circulating gastrin and endocrine cell proliferation in the atrophic gastric fundic mucosa. Scand. J. Gastroenterol. 21:357, 1986 23. Hart-Hansen, O+, Pedersen, T., Larsen, J.K., Rehfeld, J.F.: Effect of gastrin on gastric mucosal cell proliferation in man. Gut 17:536, 1976 24. Johnson, L+R.: New aspects of the trophic action of gastrointestinal hormones. Gastroenterology 72:788, 1977 25. Mutt, V+: Further investigations on intestinal hormonal peptides. Clin. Endocrinol. 5[Suppl.]:175s, 1976 26. Larsson, L.I., Rehfeld, J.F.: Evidence for a common evolutionary origin of gastrin and cholecystokinin. Nature 269:335, 1977 27. Mainz, D.L., Black, O., Webster, P.D.: Hormonal control of pancreatic growth. J. Clin. Invest. 52:2300, 1973 28. Barrowman, J.A., Mayston, P.D.: The trophic influence of cholecystokinin on the rat pancreas. J. Physiol. 238:73P, 1973 29. Ihse, l., Arnesj6, B., Lundqvist, I.: Effects on exocrine and endocrine rat pancreas of long-term administration of CCK-PZ (cholecystokinin-pancreozymin) or synthetic octa-peptide-CCKPZ. Scand. J. Gastroenterol. /1:529, 1976

870

World J. Surg. Voi. 12, No. 6, December 1988

30. Dembinski, A.B., Johnson, L,R.: Stimulation of pancreatic growth by secretin, caerulein, and pentagastrin. Endocrinology 106:323, 1980 31. Andr6n-Sandberg, A.., Dawiskiba, S., Ihse, I.: Studies of the effect of caerulein administration On experimental pancreatic carcinogenesis. Scand. J. Gastroenterol. 19:122, 1984 32. Mayston, P.D., Barrowman, J.A.: Influence of chronic administration of pentagastrin on the pancreas in hypophysectomized rats. Gastroenterology 64:391, 1973 33. Majundar, A.P.N., Goltermann, N.: Chronic administration of

pentagastrin. Effects on pancreatic protein and nucleic acid contents and protein synthesis in rats. Digestion 19:144, 1979 34. Townsend, Jr., C.M,, Franklin, R.B., Watson, L.C., Glass, E.J., Thompson, J.C.: Stimulation of pancreatic cancer growth by caerulein and secretin. Surg. Forum 32:228, 1981 35. Howatson, A.G., Carter, D.C.: Pancreatic carcinogenesis--Enhancement by cholecystokinin in the hamster-nitrosamine model. Br. J. Cancer 51:107, 1985 36. Townsend, C.M., Singh, P., Thompson~ J.CX Gastrointestinal hormones and gastrointestinal and pancreatic carcinomas. Gastroenterology 91:1002, 1986

Invited Commentary

There is both clinical and experimental evidence that adenocarcinoma of the stomach Occurs more frequently than expected when the gastric mucosa has been rendered achlorhydric by operations intended to cure peptic ulcer disease. This association has been reported in both long-term clinical reviews and animal models. The authors' hypothesis that hypergastrinemia may be similarly related to an increased incidence of PC has to be tempered by the realization that changes in the chemical interactions between foodstuffs in the achlorhydric stomach and the complex bacterial flora not present in the acidic stomach may produce endogenous nitrosamines. Gastrin would have to be implicated as a promoter or an inducer of PC. Any intestinal hormonal substance such as gastrin which stimulates either secretion or epithethial renewal is a priori classified as a "growth factor." Nearly all growth factors can promote the expression of epithelial gastrointestinal cancers without being the initiating carcinogen. If bacterial nitrosation in the achlorhydric stomach produced the initiating carcinogens, then ascorbic acid excess might serve as an antioxidant to block the process, thus reducing the risk of carcinogen production by bacteria. In this latter instance, the potential to reduce the incidence of PC in patients with PA by the daily administration of ascorbic acid would merit evaluation.

J o h n S. S p r a t t , M . D . Division of Surgical Oncology, University of Louisville, Louisville, Kentucky, U.S.A. The study of the etiology of pancreatic cancer (PC) remains a complex and incompletely defined area in the epidemiology of carcinogenesis. Botch et al. have addressed an even more narrow area of the problem by attempting to confirm a significant association between PC and pernicious anemia (PA). They hypothesize that atrophic gastritis with secondary hypergastrinemia might be responsible for the association. The authors have a statistical problem because they have a comparatively small number of cases, which were observed over 7 years for Poisson distributed events. Among 361 patients identified as having PA, 19 had had a neoplasm (not PC) before entry into the study. We are not told with clarity how many of the PC's occurred among these 19, a datum that might suggest a common carcinogen. Among all P A ' s , 3% had an accumulation of proven PC. A major problem with all epidemiological studies on PC is the great variability of diagnostic accuracy almost mandating the need for autopsies on patients dying among the study population. The accuracy of this study is enhanced by histopathological confirmation of PC.