Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Invited Commentary: Pancreas Cancer—We Know about Smoking, but Do We Know Anything Else?

Invited Commentary: Pancreas Cancer—We Know about Smoking, but Do We Know Anything Else? What we know with any certainty about pancreas cancer is quite limited. It is a disease of the developed world. It is more common in men than in women, and it has a case-fatality proportion that approaches 90 percent within 12 months of diagnosis (1). The only well-established etiologic factor is cigarette smoking. There are data showing that, as with many other epithelial cancers, a diet high in vegetables and fruit—and perhaps specifically high in folate—is associated with a lower risk, though not consistently (2). Genetic/familial predisposition is well-described but relatively rare (3–5). Chronic pancreatitis is a predisposing condition with perhaps a 20-fold excess risk (6), but the population attributable fraction is quite unclear—not least because there is underdiagnosis of pancreatitis. There are no screening tests for pancreas cancer. It is often difficult to diagnose, and it is almost always a painful, miserable way to die. Thus, new insights into its etiology are always useful, because they may suggest ways in which some of these major problems might be solved. There has been longstanding evidence of an association between diabetes mellitus and pancreas cancer, but whether these diseases are consequences of a common exposure or are causally connected, and if so, in what direction, remains unknown (1). A meta-analysis of epidemiologic studies of pancreas cancer showed a twofold increased risk among persons with diabetes mellitus (7). More recently, Gapstur et al. (8) have provided data showing that a higher postload plasma glucose level predicted elevated pancreas cancer mortality. This was a useful direction in which to take etiologic thinking, and it built on some earlier observations in an astute and productive manner, particularly upon knowledge about several anatomic and evolutionary oddities that characterize the pancreas. First, although the pancreas is an integral part of the digestive tract, unlike the rest of the tract it is never exposed either directly (the hollow organs) or indirectly (liver) to ingested food. Therefore, systemic exposure is almost certainly the mode of carcinogen delivery (compare this with, for example, the lung or esophagus). Second, although it is responsible for the production of an array of secreted digestive enzymes, the pancreas has very little endogenous metabolic capacity (9)—unlike the liver, with which it has strong evolutionary and developmental links. This suggests that neither local activation nor solubilization of a carcinogen is probable (although N-acetyltransferase 2 is one of the few enzymes that the pancreas possesses (9)). Therefore, if a bloodborne carcinogen is important, there are only a few candidates to consider. Third, and most intriguing of all, the pancreas comprises two very distinct parts—the exocrine pancreas, which manufactures digestive enzymes, and the endocrine pancreas, which is key to the control of blood sugar and, more broadly, energy homeostasis. This partnership is evolutionarily ancient; it is found in fish (10). In birds, which have a dorsal and a ventral pancreas, both elements appear in each organ. It has also been established experimentally that compounds which are toxic to the endocrine pancreas can damage the exocrine pancreas (11). As noted above, the association between diabetes mellitus and pancreas cancer has long been described. The exposure of exocrine duct cells, where the cancers arise, to very high levels of insulin as a result of proximity may be relevant. The major problems with most epidemiologic studies of pancreas cancer are consequences of the early and high case-fatality associated with the disease. This has resulted in the inclusion of only survivors in case-control studies or in the collection of surrogate data (surrogate dietary data are particularly problematic) or both (2). It is fair to say that, for this cancer especially, studies with prospectively collected data may be the only ones we can regard as interpretable. Several such studies have recently been undertaken, but the article appearing in this issue of the Journal (12) reports findings from one of the first to have prospectively examined macronutrient intake in relation to risk. The findings suggest an elevated risk for higher fat consumption and a reduced risk with greater intake of carbohydrate. A variety of questions arise. First, the cohort studied comprised only individuals with a history of cigarette smoking. Given that smoking is almost the only common risk factor we know of for this cancer, is it difficult to draw any conclusions about other etiologic factors in this setting? It could be argued either way: that the best population in which to identify other factors would be this one, where everyone is a smoker, or that a nonsmoking cohort would be better. The primary issue is whether the influence of other risk factors varies by smoking status. Since the majority of people who get pancreas cancer are either current smokers or former smokers, this population may indeed be the best kind in which to look for other modulators of risk. Smoking may be regarded as necessary but insufficient to explain the differences we see between persons with pancreas cancer and persons without cancer. Furthermore, consider lung cancer: It is possible to detect other risk factors (e.g., alcohol, vegetables, fruit) that appear to modify risks in both smokers and nonsmokers (12). However, if these associations are largely explained by uncontrolled confounding by smoking, as some have argued, then all bets are off. The association between pancreas cancer risk and a poor diet (e.g., a diet high in fat (as reported here) or low in vegetables and fruit and high in alcohol consumption (as reported elsewhere (2))) could all be explained by residual confounding. Second, a major issue in nutritional epidemiology is defining exposure. Many of us have increasingly taken the position that, since nutrient levels are an abstraction—a score derived from data on food and a generalized nutrient table—they do not provide enough data on real exposures. At most, they may tell us about the foods that are the major contributors to the nutrient score. In that case, an analysis which focuses on food items themselves (what is actually recorded) may be the only truly defensible approach. Furthermore, frustration with the small size of, and the inconsistency of nutrient-associated relative risks in, epidemiologic studies might finally lead us to consider how to do better than nutrient tables—particularly by developing approaches that do not attribute a single value (or a few values) of some nutrient to what is, in reality, a broad variety of foods. For instance, most nutrient calculation systems have few values over a small range for the fiber content of bread, even though this content varies extensively and the kinds of bread available in different communities are similarly variable. A better approach might be to identify all the kinds of bread that are available to a study population and ascribe nutrient values that are based on an availability distribution crossed with bread subtype rather than use, for example, two single archetypal values for “whole-grain bread” and “white bread.” (I am grateful to Dr. Tom Louis for earlier extensive discussions on this issue.) This problem may be greater for fat-containing and sugar-containing foods—even when attempts are made to allow for “low-fat” and “sugar-free” versions of some foods. Indeed, perhaps the oddest aspect of the data presented in the paper by Stolzenberg-Solomon et al. (12) is found in the comparison between their tables 4 and 5 (or tables 2 and 3). Table 5 shows that carbohydrate is inversely associated with risk but that starch and fiber do not explain this. Does that mean that sugar is associated with reduced risk? As table 3 shows, “starch” accounts for less than 50 percent of total carbohydrate. The puzzle deepens with an examination of the risks associated with the plausible sources of carbohydrate: The hazard ratios for rye and wheat products and for potatoes and legumes do not explain the carbohydrate finding. It is possible that vegetables or root vegetables account for some of this finding, but then it seems reasonable to ask why starch is not inversely associated with risk; the confidence limits for the fourth quintile of starch intake (hazard ratio = 1.7) actually exclude 1.0. Furthermore, the total weight of all vegetables and root vegetables is less than 50 g/day, thus accounting for only a small proportion of total daily carbohydrate intake. At best, this reinforces the problematic nature of studies of the relation between foods and nutrients. In any case, it raises some doubt about the interpretability of the carbohydrate findings. They can be contrasted with the fat findings, where the major subset of data that explains the fat association is saturated fat—and both butter and cream are associated with increased risk. Third, it seems reasonable to ask whether the fat and carbohydrate data reported here (taking these at face value) are consistent with the findings (discussed above) on diabetes, glucose, obesity, and physical activity. There is still considerable controversy about the direct relation between specific dietary patterns and the etiology of obesity and diabetes, even as the relation of these conditions to chronic disease is becoming increasingly clear: Mismatches between energy intake and physical activity appear to be implicated in the etiology of many diseases of the developed world. Accordingly, it may be that, rather than indicate fat as a risk-increasing nutrient or carbohydrate as protective (contrary to those who think high-glycemic-index food is the root of all evil), we should find better ways to understand the relation of eating patterns (13) (more than nutrient intakes) to obesity, diabetes, and other chronic diseases. Finally, we need to consider whether it is possible, as these authors have reported, to measure total energy intake—and if so, what it means. In the paper by Stolzenberg-Solomon et al., higher energy is reported as being inversely related to risk (12). All things being equal, higher energy intake ought to predict greater obesity. Willett, however, has argued that the reverse is the case—namely, that higher energy intake is a marker of higher energy output (14). Accordingly, it seems likely that what is being described here is higher energy intake's being associated with greater physical activity—and thus, the possibility that it is higher energy output which is associated with reduced risk. This seems both plausible and consistent with the insulin, diabetes, and glucose findings noted above. However, it is not always the case that higher energy intake predicts higher physical activity; there seems to be a subset of the population who have both a high energy intake and body mass index and a low energy output—and a particular dietary pattern involving high intakes of energy-dense and micronutrient-poor foods (13). If we can find better ways to identify these individuals, we may be able to establish the environmental and genetic factors that predispose them to such patterns. Furthermore, by distinguishing better between the two kinds of individuals with high energy intake, we can take a new approach to quantifying the increased and decreased risks associated with deleterious and protective behaviors. Nonetheless, it may be that, as with obesity, coronary heart disease, colon cancer, and diabetes, avoiding dietary excesses and maintaining physical activity may have an impact on the incidence of this, perhaps the most deadly of human cancers. (Reprint requests to Dr. John D. Potter at this address). REFERENCES 1. Anderson KE, Potter JD, Mack TM. Pancreatic cancer. In: Schottenfeld D, Fraumeni J Jr, eds. Cancer epidemiology and prevention. New York, NY: Oxford University Press, 1996:725–71. Google Scholar 2. World Cancer Research Fund Panel. Food, nutrition and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research, 1997. Google Scholar 3. Lynch HT, Fusaro L, Lynch JF. Familial pancreatic cancer: a family study. Pancreas  1992; 7: 511–15. Google Scholar 4. Ghadirian P, Boyle P, Simard A, et al. Reported family aggregation of pancreatic cancer within a population-based case-control study in the Francophone community in Montreal, Canada. Int J Pancreatol  1991; 10: 183-96. Google Scholar 5. Lindor NM, Greene MH, Mayo Familial Cancer Program. The concise handbook of family cancer syndromes. J Natl Cancer Inst  1998; 90: 1039–71. Google Scholar 6. Lowenfels AB, Maisonneuve P, Cavallini G, et al. Pancreatitis and the risk of pancreatic cancer. N Engl J Med  1993; 328: 1433–7. Google Scholar 7. Everhart J, Wright D. Diabetes mellitus as a risk factor for pancreatic cancer: a meta-analysis. JAMA  1995; 273: 1605–9. Google Scholar 8. Gapstur SM, Gann PH, Lowe W, et al. Abnormal glucose metabolism and pancreatic cancer mortality. JAMA  2000; 283: 2552–8. Google Scholar 9. Anderson K, Hammons G, Kadlubar F, et al. Metabolic activation of aromatic amines by human pancreas. Carcinogenesis  1997; 18: 1085–92. Google Scholar 10. Cubilla AL, Fitzgerald PJ. Tumors of the exocrine pancreas. In: Hartmann WH, Sobin LH, eds. Atlas of tumor pathology. Washington, DC: Armed Forces Institute of Pathology, 1984. Google Scholar 11. Walker AM. Diabetes and pancreatic cancer. In: Zatonski W, Boyle P, Tyczynski J, eds. Cancer prevention: vital statistics to intervention. Warsaw, Poland: PA Interpress, 1990:152–4. Google Scholar 12. Stolzenberg-Solomon RZ, Pietinen P, Taylor PR, et al. Prospective study of diet and pancreatic cancer in male smokers. Am J Epidemiol  2002; 155: 783–92. Google Scholar 13. Slattery M, Boucher K, Caan B, et al. Eating patterns and risk of colon cancer. Am J Epidemiol  1998; 148: 4-16. Google Scholar 14. Willett WC. Nutritional epidemiology. 2nd ed. New York, NY: Oxford University Press, 1998. Google Scholar http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png American Journal of Epidemiology Oxford University Press

Invited Commentary: Pancreas Cancer—We Know about Smoking, but Do We Know Anything Else?

Potter, John D.
American Journal of Epidemiology , Volume 155 (9) – May 1, 2002
Download PDF
3 pages

Loading next page...
 
/lp/oxford-university-press/invited-commentary-pancreas-cancer-we-know-about-smoking-but-do-we-pwvuKRDN52

References (16)

Publisher
Oxford University Press
ISSN
0002-9262
eISSN
1476-6256
DOI
10.1093/aje/155.9.793
Publisher site
See Article on Publisher Site

Abstract

What we know with any certainty about pancreas cancer is quite limited. It is a disease of the developed world. It is more common in men than in women, and it has a case-fatality proportion that approaches 90 percent within 12 months of diagnosis (1). The only well-established etiologic factor is cigarette smoking. There are data showing that, as with many other epithelial cancers, a diet high in vegetables and fruit—and perhaps specifically high in folate—is associated with a lower risk, though not consistently (2). Genetic/familial predisposition is well-described but relatively rare (3–5). Chronic pancreatitis is a predisposing condition with perhaps a 20-fold excess risk (6), but the population attributable fraction is quite unclear—not least because there is underdiagnosis of pancreatitis. There are no screening tests for pancreas cancer. It is often difficult to diagnose, and it is almost always a painful, miserable way to die. Thus, new insights into its etiology are always useful, because they may suggest ways in which some of these major problems might be solved. There has been longstanding evidence of an association between diabetes mellitus and pancreas cancer, but whether these diseases are consequences of a common exposure or are causally connected, and if so, in what direction, remains unknown (1). A meta-analysis of epidemiologic studies of pancreas cancer showed a twofold increased risk among persons with diabetes mellitus (7). More recently, Gapstur et al. (8) have provided data showing that a higher postload plasma glucose level predicted elevated pancreas cancer mortality. This was a useful direction in which to take etiologic thinking, and it built on some earlier observations in an astute and productive manner, particularly upon knowledge about several anatomic and evolutionary oddities that characterize the pancreas. First, although the pancreas is an integral part of the digestive tract, unlike the rest of the tract it is never exposed either directly (the hollow organs) or indirectly (liver) to ingested food. Therefore, systemic exposure is almost certainly the mode of carcinogen delivery (compare this with, for example, the lung or esophagus). Second, although it is responsible for the production of an array of secreted digestive enzymes, the pancreas has very little endogenous metabolic capacity (9)—unlike the liver, with which it has strong evolutionary and developmental links. This suggests that neither local activation nor solubilization of a carcinogen is probable (although N-acetyltransferase 2 is one of the few enzymes that the pancreas possesses (9)). Therefore, if a bloodborne carcinogen is important, there are only a few candidates to consider. Third, and most intriguing of all, the pancreas comprises two very distinct parts—the exocrine pancreas, which manufactures digestive enzymes, and the endocrine pancreas, which is key to the control of blood sugar and, more broadly, energy homeostasis. This partnership is evolutionarily ancient; it is found in fish (10). In birds, which have a dorsal and a ventral pancreas, both elements appear in each organ. It has also been established experimentally that compounds which are toxic to the endocrine pancreas can damage the exocrine pancreas (11). As noted above, the association between diabetes mellitus and pancreas cancer has long been described. The exposure of exocrine duct cells, where the cancers arise, to very high levels of insulin as a result of proximity may be relevant. The major problems with most epidemiologic studies of pancreas cancer are consequences of the early and high case-fatality associated with the disease. This has resulted in the inclusion of only survivors in case-control studies or in the collection of surrogate data (surrogate dietary data are particularly problematic) or both (2). It is fair to say that, for this cancer especially, studies with prospectively collected data may be the only ones we can regard as interpretable. Several such studies have recently been undertaken, but the article appearing in this issue of the Journal (12) reports findings from one of the first to have prospectively examined macronutrient intake in relation to risk. The findings suggest an elevated risk for higher fat consumption and a reduced risk with greater intake of carbohydrate. A variety of questions arise. First, the cohort studied comprised only individuals with a history of cigarette smoking. Given that smoking is almost the only common risk factor we know of for this cancer, is it difficult to draw any conclusions about other etiologic factors in this setting? It could be argued either way: that the best population in which to identify other factors would be this one, where everyone is a smoker, or that a nonsmoking cohort would be better. The primary issue is whether the influence of other risk factors varies by smoking status. Since the majority of people who get pancreas cancer are either current smokers or former smokers, this population may indeed be the best kind in which to look for other modulators of risk. Smoking may be regarded as necessary but insufficient to explain the differences we see between persons with pancreas cancer and persons without cancer. Furthermore, consider lung cancer: It is possible to detect other risk factors (e.g., alcohol, vegetables, fruit) that appear to modify risks in both smokers and nonsmokers (12). However, if these associations are largely explained by uncontrolled confounding by smoking, as some have argued, then all bets are off. The association between pancreas cancer risk and a poor diet (e.g., a diet high in fat (as reported here) or low in vegetables and fruit and high in alcohol consumption (as reported elsewhere (2))) could all be explained by residual confounding. Second, a major issue in nutritional epidemiology is defining exposure. Many of us have increasingly taken the position that, since nutrient levels are an abstraction—a score derived from data on food and a generalized nutrient table—they do not provide enough data on real exposures. At most, they may tell us about the foods that are the major contributors to the nutrient score. In that case, an analysis which focuses on food items themselves (what is actually recorded) may be the only truly defensible approach. Furthermore, frustration with the small size of, and the inconsistency of nutrient-associated relative risks in, epidemiologic studies might finally lead us to consider how to do better than nutrient tables—particularly by developing approaches that do not attribute a single value (or a few values) of some nutrient to what is, in reality, a broad variety of foods. For instance, most nutrient calculation systems have few values over a small range for the fiber content of bread, even though this content varies extensively and the kinds of bread available in different communities are similarly variable. A better approach might be to identify all the kinds of bread that are available to a study population and ascribe nutrient values that are based on an availability distribution crossed with bread subtype rather than use, for example, two single archetypal values for “whole-grain bread” and “white bread.” (I am grateful to Dr. Tom Louis for earlier extensive discussions on this issue.) This problem may be greater for fat-containing and sugar-containing foods—even when attempts are made to allow for “low-fat” and “sugar-free” versions of some foods. Indeed, perhaps the oddest aspect of the data presented in the paper by Stolzenberg-Solomon et al. (12) is found in the comparison between their tables 4 and 5 (or tables 2 and 3). Table 5 shows that carbohydrate is inversely associated with risk but that starch and fiber do not explain this. Does that mean that sugar is associated with reduced risk? As table 3 shows, “starch” accounts for less than 50 percent of total carbohydrate. The puzzle deepens with an examination of the risks associated with the plausible sources of carbohydrate: The hazard ratios for rye and wheat products and for potatoes and legumes do not explain the carbohydrate finding. It is possible that vegetables or root vegetables account for some of this finding, but then it seems reasonable to ask why starch is not inversely associated with risk; the confidence limits for the fourth quintile of starch intake (hazard ratio = 1.7) actually exclude 1.0. Furthermore, the total weight of all vegetables and root vegetables is less than 50 g/day, thus accounting for only a small proportion of total daily carbohydrate intake. At best, this reinforces the problematic nature of studies of the relation between foods and nutrients. In any case, it raises some doubt about the interpretability of the carbohydrate findings. They can be contrasted with the fat findings, where the major subset of data that explains the fat association is saturated fat—and both butter and cream are associated with increased risk. Third, it seems reasonable to ask whether the fat and carbohydrate data reported here (taking these at face value) are consistent with the findings (discussed above) on diabetes, glucose, obesity, and physical activity. There is still considerable controversy about the direct relation between specific dietary patterns and the etiology of obesity and diabetes, even as the relation of these conditions to chronic disease is becoming increasingly clear: Mismatches between energy intake and physical activity appear to be implicated in the etiology of many diseases of the developed world. Accordingly, it may be that, rather than indicate fat as a risk-increasing nutrient or carbohydrate as protective (contrary to those who think high-glycemic-index food is the root of all evil), we should find better ways to understand the relation of eating patterns (13) (more than nutrient intakes) to obesity, diabetes, and other chronic diseases. Finally, we need to consider whether it is possible, as these authors have reported, to measure total energy intake—and if so, what it means. In the paper by Stolzenberg-Solomon et al., higher energy is reported as being inversely related to risk (12). All things being equal, higher energy intake ought to predict greater obesity. Willett, however, has argued that the reverse is the case—namely, that higher energy intake is a marker of higher energy output (14). Accordingly, it seems likely that what is being described here is higher energy intake's being associated with greater physical activity—and thus, the possibility that it is higher energy output which is associated with reduced risk. This seems both plausible and consistent with the insulin, diabetes, and glucose findings noted above. However, it is not always the case that higher energy intake predicts higher physical activity; there seems to be a subset of the population who have both a high energy intake and body mass index and a low energy output—and a particular dietary pattern involving high intakes of energy-dense and micronutrient-poor foods (13). If we can find better ways to identify these individuals, we may be able to establish the environmental and genetic factors that predispose them to such patterns. Furthermore, by distinguishing better between the two kinds of individuals with high energy intake, we can take a new approach to quantifying the increased and decreased risks associated with deleterious and protective behaviors. Nonetheless, it may be that, as with obesity, coronary heart disease, colon cancer, and diabetes, avoiding dietary excesses and maintaining physical activity may have an impact on the incidence of this, perhaps the most deadly of human cancers. (Reprint requests to Dr. John D. Potter at this address). REFERENCES 1. Anderson KE, Potter JD, Mack TM. Pancreatic cancer. In: Schottenfeld D, Fraumeni J Jr, eds. Cancer epidemiology and prevention. New York, NY: Oxford University Press, 1996:725–71. Google Scholar 2. World Cancer Research Fund Panel. Food, nutrition and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research, 1997. Google Scholar 3. Lynch HT, Fusaro L, Lynch JF. Familial pancreatic cancer: a family study. Pancreas  1992; 7: 511–15. Google Scholar 4. Ghadirian P, Boyle P, Simard A, et al. Reported family aggregation of pancreatic cancer within a population-based case-control study in the Francophone community in Montreal, Canada. Int J Pancreatol  1991; 10: 183-96. Google Scholar 5. Lindor NM, Greene MH, Mayo Familial Cancer Program. The concise handbook of family cancer syndromes. J Natl Cancer Inst  1998; 90: 1039–71. Google Scholar 6. Lowenfels AB, Maisonneuve P, Cavallini G, et al. Pancreatitis and the risk of pancreatic cancer. N Engl J Med  1993; 328: 1433–7. Google Scholar 7. Everhart J, Wright D. Diabetes mellitus as a risk factor for pancreatic cancer: a meta-analysis. JAMA  1995; 273: 1605–9. Google Scholar 8. Gapstur SM, Gann PH, Lowe W, et al. Abnormal glucose metabolism and pancreatic cancer mortality. JAMA  2000; 283: 2552–8. Google Scholar 9. Anderson K, Hammons G, Kadlubar F, et al. Metabolic activation of aromatic amines by human pancreas. Carcinogenesis  1997; 18: 1085–92. Google Scholar 10. Cubilla AL, Fitzgerald PJ. Tumors of the exocrine pancreas. In: Hartmann WH, Sobin LH, eds. Atlas of tumor pathology. Washington, DC: Armed Forces Institute of Pathology, 1984. Google Scholar 11. Walker AM. Diabetes and pancreatic cancer. In: Zatonski W, Boyle P, Tyczynski J, eds. Cancer prevention: vital statistics to intervention. Warsaw, Poland: PA Interpress, 1990:152–4. Google Scholar 12. Stolzenberg-Solomon RZ, Pietinen P, Taylor PR, et al. Prospective study of diet and pancreatic cancer in male smokers. Am J Epidemiol  2002; 155: 783–92. Google Scholar 13. Slattery M, Boucher K, Caan B, et al. Eating patterns and risk of colon cancer. Am J Epidemiol  1998; 148: 4-16. Google Scholar 14. Willett WC. Nutritional epidemiology. 2nd ed. New York, NY: Oxford University Press, 1998. Google Scholar

Journal

American Journal of Epidemiology – Oxford University Press

Published: May 1, 2002

There are no references for this article.