Dietary glycemic load, added sugars, ad carbohydrates as risk factors for pacreatic cacer: the Multiethic Cohort Study 1 4 Ute Nöthligs, Suzae P Murphy, Lye R Wilkes, Bria E Hederso, ad Laurece N Koloel ABSTRACT Backgroud: Because elevated blood glucose cocetratios have bee show to be associated with greater risk of pacreatic cacer, a high dietary glycemic load, which is based o a empirical measure of blood glucose respose after food cosumptio, has bee hypothesized as a pacreatic cacer risk factor. However, results so far are scarce ad icosistet. Objective: We aalyzed data for 162 150 participats i the Hawaii- Los Ageles Multiethic Cohort Study to ivestigate associatios betwee glycemic load, dietary carbohydrates, sucrose, fructose, total sugars, ad added sugars ad the risk of pacreatic cacer. Desig: Dietary itake was assessed at baselie by usig a quatitative food-frequecy questioaire. Durig 8 y of follow-up, 434 icidet pacreatic cacer cases occurred. Results: Glycemic load ad added sugars were ot sigificatly associated with pacreatic cacer risk. The risk icreased with higher itakes of total sugars, fructose, ad sucrose, ad the associatio with fructose was sigificat whe the highest ad lowest quartiles were compared (relative risk: 1.35; 95% CI: 1.02, 1.80; P for tred 0.046). A sigificat associatio was foud with fruit ad juices itake (1.37; 1.02, 1.84; P for tred 0.04) but ot with soda itake. Statistical evidece of a sigificat iteractio with body mass idex was preset oly for sucrose itake (P 0.04). A compariso of the highest ad lowest quartiles of sucrose itake i overweight or obese participats gave a relative risk of 1.46 (0.95 2.25; P for tred 0.04), but the compariso was ot sigificat i ormal-weight participats. Coclusios: High fructose ad sucrose itakes may play a role i pacreatic cacer etiology. Coditios such as overweight or obesity i which a degree of isuli resistace may be preset may also be importat. Am J Cli Nutr 2007;86:1495 501. KEY WORDS Glycemic load, pacreatic cacer, added sugars, cohort studies, diet, soft driks INTRODUCTION Pacreatic cacer is the most fatal cacer i adults, with a 5-y survival rate of 5% (1). More tha 33 000 ew pacreatic cacer cases were expected i the Uited States i 2006 (2). Because of the poor progosis ad the miimal effect of covetioal treatmet methods (3), it is importat to focus o prevetio of this disease. Cigarette smokig is the most importat etiologic factor yet idetified curret smokers have a pacreatic cacer risk approximately double that of osmokers (4). The risk attributable to smokig has bee estimated at 25% (5). Obesity ad a family history of pacreatic cacer have also bee associated with the disease (3, 6 13). Other risk factors iclude icreasig age, male sex, ad Native Hawaiia or Africa America race-ethicity (14). Dietary factors may also be importat. We recetly reported a positive associatio betwee the itakes of red meat ad processed meat ad pacreatic cacer i our cohort (15). Elevated fastig (16) ad postload (17) glucose cocetratios ad diabetes mellitus (18) have bee associated with greater risk of pacreatic cacer. The hypothesized mechaism ivolves isuli (19 21), ad this hypothesis has recetly bee supported by fidigs likig higher isuli cocetratios ad isuli resistace to pacreatic cacer risk (22). Postpradial blood glucose cocetratios are iflueced by food cosumptio. Differet foods cause differet absolute peaks i blood glucose ad differet rates of chage i blood glucose cocetratios durig the period after cosumptio (23). A empirical measure of blood glucose respose after cosumptio of a specific food, the glycemic idex (GI), has bee developed to classify foods accordig to their postpradial glycemic effects ad, hece, accordig to their effects o blood isuli cocetratios (24). Because the amout of carbohydrate i a diet is a major determiat of blood glucose cocetratios, the GI of a food item is multiplied by its carbohydrate cotet to derive the glycemic load (GL) per 100-g itake of the food. The GL of a diet ca the be calculated from the amouts ad types of foods cosumed. O the basis of the assumptio that glucose metabolism plays a role i the developmet of pacreatic cacer, we hypothesized that a high dietary GL is positively associated with the risk of pacreatic cacer. Both GI ad GL have bee ivestigated with respect to pacreatic cacer i 3 prospective studies (25 27), ad 1 From the Cacer Research Ceter of Hawaii, Uiversity of Hawaii, Hoolulu, HI (UN, SPM, LRW, ad LNK); the Uiversity of Souther Califoria, Los Ageles, CA (BEH); ad the Germa Istitute of Huma Nutritio Potsdam-Rehbruecke, Nuthetal, Germay (UN). 2 The cotets of this report are solely the resposibility of the authors ad do ot ecessarily represet the official views of the Natioal Cacer Istitute. 3 Supported by grat o. R37 CA054281 from the Natioal Cacer Istitute. 4 Address reprit requests to U Nöthligs, Departmet of Epidemiology, Germa Istitute of Huma Nutritio Potsdam-Rehbruecke, Arthur-Scheuert- Allee 114-116, 14558 Nuthetal, Germay. E-mail: ute.oethligs@dife.de. Received Jauary 16, 2007. Accepted for publicatio July 12, 2007. Am J Cli Nutr 2007;86:1495 501. Prited i USA. 2007 America Society for Nutritio 1495
1496 NÖTHLINGS ET AL the results have bee icosistet. Therefore, we aalyzed 8-y prospective data from the Multiethic Cohort Study to ivestigate associatios betwee dietary GL ad pacreatic cacer risk. Furthermore, we examied various carbohydrates ad sugars, especially added sugars, to fully ivestigate the associatios. SUBJECTS AND METHODS Study desig ad populatio The Multiethic Cohort Study i Hawaii ad Los Ageles was established to ivestigate lifestyle exposures, especially diet, i relatio to cacer outcomes. The desig of the Multiethic Cohort Study was detailed elsewhere (28). I brief, the cohort is composed of 215 000 me ad wome who were 45 75 y old at cohort creatio ad who were erolled i the study betwee 1993 ad 1996. Africa Americas, Japaese Americas, Latios, Native Hawaiias, ad whites were the 5 targeted racialethic groups. All study participats iitially completed a selfadmiistered questioaire icludig a detailed dietary assessmet ad sectios o body weight ad height, physical activity, smokig behavior, history of medical coditios, reproductive history, ad family history of cacer. We excluded study participats who did ot belog to 1 of the 5 targeted racial-ethic groups, those with implausible diets as described previously (15), those whose body mass idex (BMI; i kg/m 2 ) iformatio was missig or implausible (ie, 15 or 50), those with missig iformatio o smokig status or itesity or duratio of smokig, ad those with prevalet pacreatic cacer at cohort etry. We also excluded participats with self-reported prevalet diabetes mellitus because they may have reduced their cosumptio of added sugars ad carbohydrates. These exclusios left 162 150 participats (72 966 me ad 89 184 wome) for this aalysis. As explaied i the ivitatio to participate, retur of a completed questioaire idicated a subject s writte iformed coset. The istitutioal review boards of the Uiversity of Hawaii ad the Uiversity of Souther Califoria approved the study proposal. Dietary assessmet The quatitative food-frequecy questioaire (FFQ) was especially desiged ad validated for use i this multiethic populatio (28, 29). I brief, food items for the quatitative FFQ were selected from 3-d measured dietary records i 60 subjects i each racial-ethic group. The miimum set of food items cotributig 85% of the itake of a specific list of utriets for each racial-ethic group was chose ad supplemeted by the iclusio of food items that were commo i the diet of each particular racial-ethic group. The quatitative FFQ asks about cosumptio frequecies ad portio sizes. I the quatitative FFQ calibratio study (29), average correlatio coefficiets for carbohydrate itake as a percetage of eergy itake were 0.57 for me ad 0.58 for wome betwee three 24-h recalls ad the quatitative FFQ. The Cacer Research Ceter of Hawaii food compositio table cotais both GL ad added sugars. GL values are calculated with the use of published GI values (23). Added sugars iclude all sugars used as igrediets i processed ad prepared foods, such as breads, cakes, soft driks, jam, ad ice cream, ad sugars eate separately or added to foods at the table (30). The food compositio table icludes a large recipe database ad may uique foods cosumed by a multiracial ad multiethic populatio. For FFQ items coverig 1 food, utriet profiles of the items were calculated by usig a weighted average of the specific foods based o the frequecy of use i the 24-h recalls obtaied as part of the calibratio study (29). Food mixtures were disaggregated ito their igrediets by usig a customized recipe database before food group itake was calculated. The itakes of total carbohydrates, sucrose, fructose, total sugars, added sugars, ad GL ad the itakes of odiet sodas, fruit ad juices, ad subgroups of citrus fruit ad yellow-orage fruit were aalyzed i this study. Case ascertaimet Icidet exocrie pacreatic cacer cases (ICD-02 codes C25.0-C25.3 ad C25.7-C25.9) were idetified by record likages to the Hawaii Tumor Registry, the Cacer Surveillace Program of Los Ageles Couty, ad the Califoria State Cacer Registry. All 3 registries are members of the Natioal Cacer Istitute s Surveillace, Epidemiology ad Ed Results Program. Case ascertaimet was complete through 31 December 2002. Likages to the Natioal Death Idex ad death certificate files i Hawaii ad Califoria provided iformatio o vital status ad causes of death. Statistical aalysis Differeces across quartiles of GL were tested with the Cochra-Armitage test for tred for categorical variables ad with the t test for slope i liear regressio models of mea values o GL for cotiuous variables. Cox proportioal hazards models usig age as the time metric were calculated to derive relative risks (RRs). Perso-times were determied by begiig with the date of cohort etry, defied as the date of questioaire completio, ad edig at the earliest of the followig dates: date of pacreatic cacer diagosis, date of death, or 31 December 2002, the closure date of the study. Tests based o Schoefeld residuals showed o evidece that proportioal hazards assumptios were violated for ay aalysis. We preset models icludig both sexes, because there was o evidece of iteractio by sex, after adjustmet for sex ad follow-up time o study from baselie ( 2 y, 2 5 y, or 5 y) as strata variables, to allow for differet baselie hazard rates. A separate aalysis suggested some differeces betwee RRs across 3 follow-up time strata, although the differeces were ot statistically sigificat (data ot show). Quartiles of utriet ad food itakes were based o the distributio of the variable i the overall cohort. Media values for quartiles by sex ad race-ethicity were used i the respective models to test for tred. Race-ethicity, age at cohort etry, smokig status, pack-years of smokig, family history of pacreatic cacer, eergy itake (logarithmically trasformed), itake of red ad processed meats, ad BMI were used as adjustmet factors i the disease risk models. The latter 3 factors have bee idetified i the Multiethic Cohort Study as risk factors for pacreatic cacer (15, 31). To reduce measuremet error i the dietary assessmets, utriets ad foods were aalyzed i terms of desities ie, by 100 or 1000 kcal/d. Our group (29) ad others (32) have show that correlatios with referece measuremets were higher for desities tha for absolute itake. Give the strog evidece for the
GLYCEMIC LOAD AND PANCREATIC CANCER RISK 1497 hypothesis of a differet associatio by overweight or obesity that ca be associated with isuli resistace (26, 33, 34), we also calculated all models separately for ormal-weight (BMI 25) or overweight or obese (BMI 25) participats. The likelihood ratio test was used to determie the sigificace of the iteractio betwee BMI ad mai exposure variables with respect to pacreatic cacer. The test compares a mai effects, o-iteractio model with a fully parameterized model cotaiig all possible iteractio terms for the variables of iterest. RESULTS Participat characteristics are give i Table 1. The percetage of me icreased across quartiles of dietary GL. Mea BMI was slightly elevated i the last quartile of GL oly. All dietary variables of iterest were positively associated with GL. Because all of these dietary variables are likely to be highly associated with each other, we calculated Spearma correlatio coefficiets betwee GL ad the itakes of total carbohydrates, sucrose, fructose, total sugars, ad added sugars. Correlatio coefficiets were 0.93, 0.40, 0.45, 0.46, ad 0.23, respectively, which idicates that the strogest correlate of GL was the itake of total carbohydrates. Durig follow-up, 434 icidet pacreatic cacer cases occurred i the cohort. GL was ot associated with pacreatic cacer risk i the overall cohort (Table 2). However, the RR for fructose was sigificatly elevated i the highest quartile (P for tred 0.046). Although the risks for the itakes of total sugars TABLE 1 Characteristics of participats i the Multiethic Cohort Study by quartile (Q) of dietary glycemic load 1 Q1 ( 40 463) as well as sucrose were highest i the fourth quartile ad were suggestive of a associatio, either the poit estimates or the test for tred were sigificat. Because obesity ca be a determiat of isuli resistace, we stratified our aalysis to examie whether the effects of GL or carbohydrate itakes varied by BMI. Higher (but ot sigificatly higher) risks of pacreatic cacer were see i the overweight ad obese group (BMI 25) tha i the ormal-weight group (BMI 25) i the top quartiles of itakes of all dietary variables, ad there were stroger treds across quartiles (data ot show). However, a sigificat (P 0.04) iteractio was evidet for sucrose oly (Table 3). Amog overweight ad obese participats, the RR (95% CI) for the fourth quartile of sucrose itake compared with the first was 1.46 (0.95, 2.25; P for tred 0.04); amog ormal-weight participats, the respective values were 1.07 (0.71, 1.60; P for tred 0.85). The use of calibratio-adjusted utriet itakes gave similar results (data ot show). Additioal adjustmet for total physical activity did ot alter the fidigs, although a further stratificatio by level of physical activity amog the overweight or obese participats suggested a higher RR for those with a higher level of physical activity tha for those with a low level of physical activity (data ot show). Fruit ad juices combied were the largest cotributor to fructose itake i the cohort, followed by odiet sodas. To cofirm our fidigs for fructose, we aalyzed the itakes of several food groups soda, fruit juices, total fruit, citrus fruit, ad yelloworage fruit i disease risk models (Table 4). The combied Q2 ( 41 353) Q3 ( 41 134) Q4 ( 39 200) Glycemic load 78.3 18.0 2 124.3 11.6 169.5 15.6 271.9 72.3 Me (%) 35 41 48 56 Age (y) 60 9 60 9 60 9 59 9 Race-ethicity (%) Africa America 23 15 13 13 Japaese America 20 31 35 31 Latio 21 19 19 25 White 31 30 27 19 Native Hawaiia 4 6 7 11 Follow-up (y) 8.4 1.6 8.4 1.5 8.3 1.5 8.3 1.5 BMI (kg/m 2 ) 25.6 4.7 25.4 4.6 25.5 4.6 26.3 4.9 Physical activity (METs/d) 1.6 0.3 1.6 0.3 1.6 0.3 1.7 0.3 Smokig status (%) Never 44 46 47 45 Former 37 38 39 39 Curret 19 15 14 17 Pack-years of smokig 10.4 15.1 9.9 14.8 9.7 14.6 10.1 14.9 Family history of pacreatic cacer (%) 1.6 1.8 1.9 1.6 Eergy itake (kcal/d) 1238 416 1753 412 2274 492 3476 1057 Carbohydrates (% of eergy) 48.3 9.1 52.2 8.4 54.1 8.3 55.9 8.5 Sucrose (g/d) 20.8 8.1 31.8 10.1 43.2 13.4 69.4 30.4 Fructose (g/d) 12.9 6.8 20.3 9.7 28.4 13.7 50.4 35.8 Total sugar (g/d) 57.7 21.7 88.3 27.8 120.3 37.6 197.4 93.6 Added sugars (teaspoos/d) 5.9 3.4 9.1 4.9 12.8 6.9 23.3 18.4 1 METs, metabolic equivalet tasks. Treds across quartiles were sigificat (P 0.001) for all variables except for ever smokers ad family history of pacreatic cacer, accordig to the Cochra-Armitage tests for categorical variables ad the t test for slope i liear regressio models of mea values o glycemic load for cotiuous variables. 2 x SD (all such values).
1498 NÖTHLINGS ET AL TABLE 2 Multivariate-adjusted relative risks (RRs) (ad 95% CIs) for glycemic load ad itakes of carbohydrates i relatio to pacreatic cacer i the Multiethic Cohort Study 1 Dietary itake Cases/ ocases Adjusted RR (95% CI) Glycemic load (g 1000 kcal 1 d 1 ) 63.3 95/40 604 1 63.3 72.7 115/40 720 1.24 (0.94, 1.64) 72.7 82.3 111/40 693 1.14 (0.85, 1.53) 82.3 113/39 699 1.10 (0.80, 1.52) P for tred 2 0.65 Carbohydrates (g 1000 kcal 1 d 1 ) 46.7 103/41 045 1 46.7 52.6 104/40 686 1.05 (0.79, 1.39) 52.6 58.7 116/40 408 1.14 (0.85, 1.53) 58.7 111/39 577 1.04 (0.75, 1.46) P for tred 2 0.75 Sucrose (g 1000 kcal 1 d 1 ) 13.7 83/37 284 1 13.7 17.7 107/40 254 1.16 (0.87, 1.54) 17.7 22.1 122/41 826 1.25 (0.94, 1.66) 22.1 122/42 352 1.23 (0.91, 1.65) P for tred 2 0.21 Fructose (g 1000 kcal d 1 ) 3 7.3 93/39 383 1 7.3 10.8 113/40 397 1.16 (0.88, 1.53) 10.8 15.4 101/40 806 1.04 (0.78, 1.39) 15.4 127/41 130 1.35 (1.02, 1.80) P for tred 2 0.046 Total sugar (g 1000 kcal 1 d 1 ) 40.0 97/38 763 1 40.0 49.4 105/40 403 1.03 (0.78, 1.37) 49.4 62.6 104/40 993 1.01 (0.75, 1.35) 62.6 128/41 557 1.28 (0.95, 1.73) P for tred 2 0.09 Added sugars (teaspoos 1000 kcal 1 d 1 ) 3.2 87/35 443 1 3.2 4.6 107/40 591 1.08 (0.81, 1.43) 4.6 6.5 132/42 530 1.28 (0.97, 1.68) 6.5 108/43 152 1.08 (0.81, 1.44) P for tred 2 0.61 1 Models were cotrolled for sex ad time o study ( 2 y, 2 5 y, or 5 y) as strata variables. Multivariate adjustmet for race-ethicity, age at cohort etry, smokig status, pack-years of smokig, family history of pacreatic cacer, eergy itake (log-trasformed), itakes of red meat ad processed meat, ad BMI. Quartiles were based o the distributio of each variable i the etire cohort. 2 Media values for the overall quartiles i each sex ad racial-ethic group were aalyzed as a cotiuous variable i the respective model to test for tred. 3 Free moosaccharide. itake of fruit ad juices ad the itake of fruit aloe were associated with a 37 42% icrease i the risk of pacreatic cacer i the overall cohort. Amog overweight or obese participats, total fruit ad juice itake was associated with a 51% higher risk, but weaker associatios were see whe fruit ad fruit juices were separated (data ot show). Whe total fruit itake was divided ito citrus fruit or yellow-orage fruit, the itake of yelloworage fruit was positively associated with pacreatic cacer risk amog ormal-weight participats. However, o statistical evidece of a iteractio of ay fruit or juice variable with BMI was preset. A high itake of regular sodas was ot associated with pacreatic cacer risk i the overall cohort or i the subgroup of obese or overweight participats. To determie whether the associatio was limited to fruit, we also ivestigated joitly the treds for fructose from fruit, soda, ad other sources. All treds were positive ad osigificat. Therefore, the associatio with fruit may be due to the fructose cotet of fruit or aother factor. Because smokig is the most well-established risk factor for pacreatic cacer, we stratified our aalysis accordig to smokig status. Overall, there was little evidece of effect modificatio by smokig status (ever, former, or curret), although the greater risk associated with fruit itake was most apparet amog ever smokers (data ot show). DISCUSSION I the Multiethic Cohort Study, high sugar itake specifically, fructose itake was associated with a greater risk of pacreatic cacer. This associatio was ot reflected i the cosumptio of sodas, but we did observe a greater risk with a higher itake of fruit ad juices. Statistical evidece for a iteractio of sucrose itake with BMI was preset, ad it showed a higher risk of pacreatic cacer i overweight or obese study participats with higher sucrose cosumptio. To date, 4 prospective studies have ivestigated dietary GL ad various carbohydrates i relatio to pacreatic cacer risk (25 27, 35). Two of these studies foud o associatio for GL, GI, total carbohydrates, total sugar, sucrose, or fructose (25, 27), ad oe study foud a greater risk with higher dietary GL, GI, ad fructose itake i sedetary wome with a BMI 25, but ot i the overall cohort (26). I the preset study, fructose itake was associated with the highest risk of pacreatic cacer. A subsequet aalysis of high cosumptio of soft driks i this same study populatio showed sigificatly greater risks i wome but ot i me (33). The preset study also foud a positive associatio betwee fructose itake ad pacreatic cacer risk, but we did ot see a greater risk with higher itake of sodas; these results did ot differ sigificatly betwee me ad wome. A sigificat iverse associatio with the itake of carbohydrates was reported amog smokers i Filad i the fourth prospective study (35). A recet prospective study i Swede reported high cosumptio of sugar added to coffee, tea, cereals ad other foods, ad soft driks to be sigificatly associated with a greater risk of pacreatic cacer (36), but the study did ot cofirm a effect modificatio by BMI or physical activity. Our results supported a modifyig effect of BMI o the associatio betwee the itake of sucrose ad pacreatic cacer risk; the associatio was positive i overweight ad obese participats, but there was o sigificat associatio i ormal-weight participats. We previously reported that physical activity was ot associated with pacreatic cacer i the Multiethic Cohort Study (31), ad stratificatio o this variable did ot support the hypothesis that low physical activity associated with a high BMI will lead to a eve larger RR. I fact, the opposite was true the RRs were greater i overweight or obese participats with higher physical activity. Differeces i eergy metabolism may explai this observatio,
GLYCEMIC LOAD AND PANCREATIC CANCER RISK 1499 TABLE 3 Multivariate-adjusted relative risks (RRs) (ad 95% CIs) for sucrose ad pacreatic cacer stratified by BMI i the Multiethic Cohort Study 1 BMI 25 BMI 25 Dietary itake Cases/ocases Adjusted RR (95% CI) Cases/ocases Adjusted RR (95% CI) Sucrose (g 1000 kcal 1 d 1 ) 13.7 44/16 775 1 39/20 509 1 13.7 17.7 64/19 200 1.27 (0.86, 1.87) 43/21 054 1.04 (0.67, 1.61) 17.7 22.1 55/21 497 0.97 (0.64, 1.45) 67/20 329 1.65 (1.10, 2.48) 22.1 63/22 254 1.07 (0.71, 1.60) 59/20 098 1.46 (0.95, 2.25) P for tred 2 0.85 0.04 1 BMI was calculated by kg/m 2. Models were cotrolled for sex ad time o study ( 2y, 2 5 y, or 5 y) as strata variables. Multivariate adjustmet for race-ethicity, age at cohort etry, smokig status, pack-years of smokig, family history of pacreatic cacer, eergy itake (log-trasformed), itakes of red meat ad processed meat, ad BMI. The BMI sucrose iteractio was sigificat, P 0.04. 2 Media values for the overall quartiles i each sex ad racial-ethic group were aalyzed as a cotiuous variable i the respective model to test for tred. ad, if that observatio is cofirmed i other studies, more detailed ivestigatios should be udertake. The fidig of a greater risk with higher fruit ad juices itake was surprisig, ad it merits some discussio. Fruit, most ofte i combiatio with vegetables, geerally is thought to have beeficial effects i terms of cacer prevetio at various sites, icludig the pacreas (37). To date, 7 prospective studies have reported o fruit itake ad pacreatic cacer risk, ad oe detected a sigificat associatio, either positive or iverse (10, 35, 38 42). A recet study from Swede icluded 135 pacreatic cacer cases i a cohort of 81 922 me ad wome (38). Whe the highest ad lowest quartiles were compared, overall fruit itake was ot sigificatly associated with pacreatic cacer risk (RR: 1.10; 95% CI: 0.64, 1.88), or was citrus fruit itake. Participats i the highest quartile cosumed 2.5 servigs of fruit/d. A study i Fiish smokers foud a osigificat RR of 0.85 for the highest cosumptio of all fruit ad berries i a aalysis of 163 cases ad 26 948 ocases (35). Media cosumptio for cases ad ocases was 100 g/d. Four studies with mortality as the outcome did ot fid ay associatios (39 42). Oe of these studies (39) icluded the largest umber of pacreatic cacer deaths thus far ( 3751), although the dietary assessmet was very limited ad icluded oly oe questio about cosumptio of citrus fruit or juices. Our study, i cotrast to these previously published studies, examied the largest umber of icidet pacreatic cacer cases by usig a comprehesive ad detailed dietary assessmet, which eabled us to detect statistically sigificat associatios of smaller magitude. Because the itake of fruit ad juices combied is the largest cotributor to fructose itake i our cohort, the icrease i risk with high fruit itake may be explaied by fructose ad total sugars, both of which are highly correlated with fruit itake (0.69 ad 0.65, respectively). However, models with fructose separated by sources showed similar associatios for all sources. A effect of fruit other tha through sugars, therefore, caot be ruled out at this poit. Furthermore, although the ull associatio with soda does ot seem to support this cojecture, uderreportig of the itake of low-utriet-desity beverages such as soda may atteuate the true associatio (43). The hypothesized mechaism likig a impaired glucose metabolism or diabetes mellitus to pacreatic cacer ivolves isuli. Isuli ca promote tumor developmet by ihibitig apoptosis ad stimulatig cell proliferatio, ad it has bee argued that isuli acts as a promoter for pacreatic carciogeesis (19 21). Our fidigs support this hypothesis to some extet, because slightly elevated risks, especially those with sucrose, were see i overweight or obese participats who may already have a uderlyig degree of isuli resistace (21). We reported a higher pacreatic cacer risk i obese me but ot i obese wome i the Multiethic Cohort Study, ad evidece i the literature for a positive associatio betwee BMI ad pacreatic cacer risk is fairly cosistet (31). As discussed i the literature, the GI is a cotroversial cocept (44). It has bee poited out that the GIs of foods vary by types, by processig or preparatio, ad by combiatios of foods cosumed together, which reders the actual determiatio of a GI value for a specific food difficult. I additio, the glycemic respose after igestio of a food does ot ecessarily predict the isuli respose (44). Most importat for our study, it has bee debated whether the GI or GL is accurate whe usig dietary itakes collected with a FFQ, because some questioaire items may group foods with differig GL values (44). However, the quatitative FFQ used for the Multiethic Cohort Study is legthy (almost 200 items) ad thus less likely tha most to group dissimilar foods. I our study, we used GL rather tha GI because the blood glucose respose after igestio of a food is determied by both its carbohydrate cotet ad its GI, ad GL combies the 2. We further explored the relatios betwee variables by calculatig correlatio coefficiets. Ideed, the correlatio betwee GL ad carbohydrate itake was very high ( 0.9), ad we foud almost idetical RRs for pacreatic cacer. We suggest that dietary GL i geeral may ot add importat ew iformatio about the quality of carbohydrates i the diet of our cohort participats. Several limitatios have to be take ito accout i iterpretig this aalysis. Our populatio was from Hawaii ad Califoria oly. However, the study was populatio-based i desig to maximize geeralizability to the US populatio (28). Furthermore, dietary measuremet error is certaily preset i our data. However, we attempted to miimize the error by aalyzig utriets ad foods as desities ie, per 100 or 1000 kcal/d (29). Fially, GI values have ot bee determied for may local food products, ad values had to be estimated from similar foods.
1500 NÖTHLINGS ET AL TABLE 4 Multivariate-adjusted relative risks (RRs) (ad 95% CIs) for sodas, juices ad fruit i relatio to pacreatic cacer i the Multiethic Cohort Study 1 Dietary itake Cases/ ocases Adjusted RR (95% CI) Nodiet sodas (g 1000 kcal 1 d 1 ) 0 168/63 718 1 0.1 18.2 77/32 027 0.91 (0.69, 1.20) 18.2 75.7 105/32 773 1.25 (0.97, 1.61) 75.7 84/33 198 1.07 (0.82, 1.41) P for tred 2 0.54 Fruit ad juices (g 1000 kcal 1 d 1 ) 77.4 94/41 491 1 77.4 144.7 105/40 391 1.10 (0.83, 1.47) 144.7 238.2 102/40 187 1.05 (0.78, 1.40) 238.2 133/39 647 1.37 (1.02, 1.84) P for tred 2 0.04 Fruit juices (g 1000 kcal 1 d 1 ) 4.7 106/38 731 1 4.7 20.2 105/40 341 1.02 (0.78, 1.34) 20.2 60.0 103/40 778 0.98 (0.74, 1.29) 60.0 120/41 866 1.08 (0.83 1.41) P for tred 2 0.56 Fruit (g 1000 kcal 1 d 1 ) 53.3 88/42 400 1 53.3 104.1 107/40 718 1.18 (0.89, 1.58) 104.1 178.4 111/39 910 1.20 (0.90, 1.62) 178.4 128/38 688 1.42 (1.05, 1.93) P for tred 2 0.03 Citrus fruit (g 1000 kcal 1 d 1 ) 13.4 103/41 140 1 13.4 40.7 104/40 693 1.03 (0.78, 1.35) 40.7 93.9 108/40 128 1.02 (0.77, 1.34) 93.9 119/39 755 1.08 (0.82, 1.43) P for tred 2 0.63 Yellow-orage fruit (g 1000 kcal 1 d 1 ) 6.0 96/41 377 1 6.0 14.7 97/40 727 0.98 (0.74, 1.30) 14.7 35.1 111/39 975 1.10 (0.83, 1.46) 35.1 130/39 637 1.16 (0.87, 1.54) P for tred 2 0.19 1 Models were cotrolled for sex ad time o study ( 2 y, 2 5 y, or 5 y) as strata variables. Multivariate adjustmet for race-ethicity, age at cohort etry, smokig status, pack-years of smokig, family history of pacreatic cacer, eergy itake (log-trasformed), itakes of red meat ad processed meat, ad BMI. Quartiles were based o the distributio of each variable i the etire cohort. 2 Media values for the overall quartiles i each sex ad ethic group were aalyzed as a cotiuous variable i the respective model to test for tred. Because of the rapid fatality of pacreatic cacer, case-cotrol studies have geerally relied heavily o proxy iterviews. Recall bias, which is already a problem i case-cotrol studies, is of eve more cocer i the case of proxy iterviews. Hece, results from case-cotrol studies of pacreatic cacer must be iterpreted with great cautio. I cotrast, the relatively low icidece rate of pacreatic cacer leads to low umbers of cases i prospective studies uless the observatio period is log. I additio to the large sample size, the heterogeeity of participats i the Multiethic Cohort Study ad therefore the variability i dietary itake is a stregth of the preset study, because it facilitates the detectio of meaigful associatios. Our FFQ was especially desiged for the study populatio, ad it was comprehesive, which allowed us to adjust for eergy itake. I coclusio, the preset study adds to the evidece that itakes of fructose ad sucrose play a role i the developmet of pacreatic cacer. The stroger associatio with a high itake of sucrose i those who were overweight or obese implies that such coditios, i which a uderlyig degree of isuli resistace may be preset, are also importat. We foud some evidece for a greater pacreatic cacer risk with a high itake of fruit ad juices but ot with a high itake of sodas. The authors resposibilities were as follows UN, SPM, ad LNK: study cocept ad desig; LNK, LRW, ad BEH: data collectio; UN ad LW: statistical aalysis; UN, SPM, ad LNK: iterpretatio of results; UN: writig of the mauscript; UN, SPM, LRW, LNK, ad BEH: critical review of the mauscript; ad all authors: review ad approval of the fial mauscript. Noe of the authors had a fiacial or persoal coflict of iterest. REFERENCES 1. Jemal A, Clegg LX, Ward E, et al. Aual report to the atio o the status of cacer, 1975 2001, with a special feature regardig survival. Cacer 2004;101:3 27. 2. Cacer facts ad figures 2006. Atlata, GA: America Cacer Society, 2006. 3. Li D, Xie K, Wolff R, Abbruzzese JL. Pacreatic cacer. Lacet 2004; 363:1049 57. 4. Nilse TI, Vatte LJ. A prospective study of lifestyle factors ad the risk of pacreatic cacer i Nord-Trodelag, Norway. Cacer Causes Cotrol 2000;11:645 52. 5. Lowefels AB, Maisoeuve P. Epidemiology ad prevetio of pacreatic cacer. Jp J Cli Ocol 2004;34:238 44. 6. Klei AP, Brue KA, Peterse GM, et al. Prospective risk of pacreatic cacer i familial pacreatic cacer kidreds. Cacer Res 2004;64: 2634 8. 7. Vimalachadra D, Ghaeh P, Costello E, Neoptolemos JP. Geetics ad prevetio of pacreatic cacer. Cacer Cotrol 2004;11:6 14. 8. Calle EE, Murphy TK, Rodriguez C, Thu MJ, Heath CW Jr. Diabetes mellitus ad pacreatic cacer mortality i a prospective cohort of Uited States adults. Cacer Causes Cotrol 1998;9:403 10. 9. Everhart J, Wright D. Diabetes mellitus as a risk factor for pacreatic cacer. A meta-aalysis. JAMA 1995;273:1605 9. 10. Shibata A, Mack TM, Pagaii-Hill A, Ross RK, Hederso BE. A prospective study of pacreatic cacer i the elderly. It J Cacer 1994; 58:46 9. 11. Stolzeberg-Solomo RZ, Pietie P, Taylor PR, Virtamo J, Albaes D. A prospective study of medical coditios, athropometry, physical activity, ad pacreatic cacer i male smokers (Filad). Cacer Causes Cotrol 2002;13:417 26. 12. Batty GD, Shipley MJ, Marmot M, Smith GD. Diabetes status ad post-load plasma glucose cocetratio i relatio to site-specific cacer mortality: fidigs from the origial Whitehall study. Cacer Causes Cotrol 2004;15:873 81. 13. Berrigto de Gozalez A, Sweetlad S, Specer E. A meta-aalysis of obesity ad the risk of pacreatic cacer. Br J Cacer 2003;89: 519 23. 14. Ghadiria P, Lych HT, Krewski D. Epidemiology of pacreatic cacer: a overview. Cacer Detect Prev 2003;27:87 93. 15. Nothligs U, Wilkes LR, Murphy SP, Haki JH, Hederso BE, Koloel LN. Meat ad fat itake as risk factors for pacreatic cacer: the Multiethic Cohort Study. J Natl Cacer Ist 2005;97:1458 65. 16. Jee SH, Ohrr H, Sull JW, Yu JE, Ji M, Samet JM. Fastig serum glucose level ad cacer risk i Korea me ad wome. JAMA 2005;293:194 202.
GLYCEMIC LOAD AND PANCREATIC CANCER RISK 1501 17. Gapstur SM, Ga PH, Lowe W, Liu K, Colagelo L, Dyer A. Abormal glucose metabolism ad pacreatic cacer mortality. JAMA 2000;283: 2552 8. 18. Huxley R, Asary-Moghaddam A, Berrigto de Gozalez A, Barzi F, Woodward M. Type-II diabetes ad pacreatic cacer: a meta-aalysis of 36 studies. Br J Cacer 2005;92:2076 83. 19. Fisher WE, Boros LG, Schirmer WJ. Isuli promotes pacreatic cacer: evidece for edocrie ifluece o exocrie pacreatic tumors. J Surg Res 1996;63:310 3. 20. McCarty MF. Isuli secretio as a determiat of pacreatic cacer risk. Med Hypotheses 2001;57:146 50. 21. Kaaks R, Lukaova A. Eergy balace ad cacer: the role of isuli ad isuli-like growth factor-i. Proc Nutr Soc 2001;60:91 106. 22. Stolzeberg-Solomo RZ, Graubard BI, Chari S, et al. Isuli, glucose, isuli resistace, ad pacreatic cacer i male smokers. JAMA 2005; 294:2872 8. 23. Foster-Powell K, Holt SH, Brad-Miller JC. Iteratioal table of glycemic idex ad glycemic load values: 2002. Am J Cli Nutr 2002;76: 5 56. 24. Jekis DJ, Kedall CW, Augusti LS, et al. Glycemic idex: overview of implicatios i health ad disease. Am J Cli Nutr 2002;76(suppl): 266S 73S. 25. Johso KJ, Aderso KE, Harack L, Hog CP, Folsom AR. No associatio betwee dietary glycemic idex or load ad pacreatic cacer icidece i postmeopausal wome. Cacer Epidemiol Biomarkers Prev 2005;14:1574 5. 26. Michaud DS, Liu S, Giovaucci E, Willett WC, Colditz GA, Fuchs CS. Dietary sugar, glycemic load, ad pacreatic cacer risk i a prospective study. J Natl Cacer Ist 2002;94:1293 300. 27. Silvera SA, Roha TE, Jai M, Terry PD, Howe GR, Miller AB. Glycemic idex, glycemic load, ad pacreatic cacer risk (Caada). Cacer Causes Cotrol 2005;16:431 6. 28. Koloel LN, Hederso BE, Haki JH, et al. A multiethic cohort i Hawaii ad Los Ageles: baselie characteristics. Am J Epidemiol 2000; 151:346 57. 29. Stram DO, Haki JH, Wilkes LR, et al. Calibratio of the dietary questioaire for a multiethic cohort i Hawaii ad Los Ageles. Am J Epidemiol 2000;151:358 70. 30. Sharma S, Murphy SP, Wilkes LR, Au D, She L, Koloel LN. Extedig a multiethic food compositio table to iclude stadardized food group servigs. J Food Compos Aal 2003;16:485 95. 31. Nothligs U, Wilkes LR, Murphy SP, Haki JH, Hederso BE, Koloel LN. Body mass idex ad physical activity as risk factors for pacreatic cacer: the Multiethic Cohort Study. Cacer Causes Cotrol 2007;18:165 75. 32. Subar AF, Kipis V, Troiao RP, et al. Usig itake biomarkers to evaluate the extet of dietary misreportig i a large sample of adults: the OPEN study. Am J Epidemiol 2003;158:1 13. 33. Scherhammer ES, Hu FB, Giovaucci E, et al. Sugar-sweeteed soft drik cosumptio ad risk of pacreatic cacer i two prospective cohorts. Cacer Epidemiol Biomarkers Prev 2005;14:2098 105. 34. Despres JP, Lemieux I. Abdomial obesity ad metabolic sydrome. Nature 2006;444:881 7. 35. Stolzeberg-Solomo RZ, Pietie P, Taylor PR, Virtamo J, Albaes D. Prospective study of diet ad pacreatic cacer i male smokers. Am J Epidemiol 2002;155:783 92. 36. Larsso SC, Bergkvist L, Wolk A. Cosumptio of sugar ad sugarsweeteed foods ad the risk of pacreatic cacer i a prospective study. Am J Cli Nutr 2006;84:1171 6. 37. Food, utritio, ad the prevetio of cacer: a global perspective. Washigto, DC, ad Lodo, Uited Kigdom: America Istitute for Cacer Research/World Cacer Research Fud, 1997. 38. Larsso SC, Hakasso N, Naslud I, Bergkvist L, Wolk A. Fruit ad vegetable cosumptio i relatio to pacreatic cacer risk: a prospective study. Cacer Epidemiol Biomarkers Prev 2006;15:301 5. 39. Coughli SS, Calle EE, Patel AV, Thu MJ. Predictors of pacreatic cacer mortality amog a large cohort of Uited States adults. Cacer Causes Cotrol 2000;11:915 23. 40. Sauvaget C, Nagao J, Hayashi M, Specer E, Shimizu Y, Alle N. Vegetables ad fruit itake ad cacer mortality i the Hiroshima/ Nagasaki Life Spa Study. Br J Cacer 2003;88:689 94. 41. Zheg W, McLaughli JK, Gridley G, et al. A cohort study of smokig, alcohol cosumptio, ad dietary factors for pacreatic cacer (Uited States). Cacer Causes Cotrol 1993;4:477 82. 42. Mills PK, Beeso WL, Abbey DE, Fraser GE, Phillips RL. Dietary habits ad past medical history as related to fatal pacreas cacer risk amog Advetists. Cacer 1988;61:2578 85. 43. Kat AK. Nature of dietary reportig by adults i the third Natioal Health ad Nutritio Examiatio Survey, 1988 1994. J Am Coll Nutr 2002;21:315 27. 44. Pi-Suyer FX. Glycemic idex ad disease. Am J Cli Nutr 2002; 76(suppl):290S 8S.