Case Control Example

8/3/2019 Case Control Example

1/6

Int. J . Cancer: 44, 611-616 (1989)0 989 Alan R. Liss, Inc.

Publ ication of the International Union Agalnst CancerPublication de I'Union lnternationale Contre le Cancer

A CASE-CONTROL STUDY OF GASTRIC CANCER AND DIET IN ITALYEva B UIATTI ' .~" ,omenico PALLI' Adriano DE C A R LI ~,in o A MA D O R I~ ,laudio A VEL LINP , imonetta B I A N C H I ~ ,Roberta BIS ERN I~,rancesco CIPRIAN I'Pierh igi COCCOh, Attilio GIACOSA', Ettore M A R U B IN I~ ,iccardo PU NT ON I~,Carla VIN DIGN I~,oseph FRA UM ENI,R . ~nd W illiam BLOT^I Unita di Epidemiologia, Centro per lo Studio e la P revenzione O ncologica , Florence; 21stituto di Statistica Med ica e Biometria,Universita di M ilano, M ilan; 3Servizio di Onco logia, Forli; 4Servizio di Anatomia P atologica, Imola; 51stituto di Anato miaPatologica, Universita di Firenze, F lorence; 61stituto di M edicina del Lavoro, Universita di C agliari, Ca gliari; 71stituto Nazionaleper la Ricerca sul Cancro, Genoa ; %tituto di Anutomia Patologica I I , Universita di Siena, Sienu, Italy; and 9National CancerInstitute, Bethesda , M D , USA.

A case-control study was conducted in high- and low-r iskareas of Ita ly t o evaluate reasons for the striking geographicvariation in gastric cancer (GC)mortality within the country.Personal interviews w ith 1,016 histologically confirmed GCcases and 1,159 population controls of similar age and sexrevealed that the patients were more often of lower socialclass and resident in ru ral areas and more frequently reporteda familial history of gastric (but not other) cancer. After ad-justing for these effects, case-control differences were foundfo r several dietary variables, assessed by asking about theusual frequency of consumption of 146 food items and bever-ages. A significant trend of increasingGC risk was found withincreasing consumption of traditiona l soups, meat, salted/dried fish and a combination of cold cuts and seasonedcheeses. The habit of adding salt and the preference for saltyfoods were associated with elevated GC risk, while more fre-quently storing foods in the refrigerator, the availability of afreezer and use of frozen foods lowered r isk. ReducedGC r i s kwere associated wi th increasing intake of raw vegetables,fresh fr ui t and citrus fruits. Lowered r i s k was also related toconsumption of spices, olive oi l and garlic. Ne ithe r cigarettesmoking nor alcoholic beverage drinking were significantlyrelated to GC risk. The case-control differences tended to beconsistent across geographic areas, despite marked regionalvariations in intake levels of certain foods. The high-risk areastended to show higher consumption of food associated wi thelevated r i s k (traditional soups, cold cuts) and lower con-sumption of foods associated wi th reduced risks (raw vegeta-bles, citrus fruits, garlic). Our findings indicate that dietaryfactors contribute t o he regional variation of stomach cancerOccurrence in Italy, and offer clues for further etiologic andprevention research.

Gastric cancer (GC) is estimated to be the most commoncancer worldwide, and the second leading cause of cancerdeath (Kurihara et al . , 1984; Parkin et al., 1988). AmongEuropean countries, Italy has one of the highest rates of GCmortality, although there is substantial geographic variationwithin the country (C islaghi et al., 1986). Despite recent de-clines, GC still ranks first in cancer mortality in parts of northcentral Italy, where some of the w orld's highest GC rates arefound, while rates are com paratively lo w in the south (ISTA T,1989; Decarli et al., 1986). To investigate reasons for theregional GC differences, a large multi-center case-controlstudy was conducted in high- and low-risk areas. We nowreport the results of this investigation, focusing on patterns infood intake that may influence GC risk and contribute to thegeographic variation.SUBJECTS AN D METHODS

The case-control study involved 7 centers grouped into 4areas, 2 with high (1: Forli/Cremona/Imola and 2: Florence/Siena) and 2 with low (3: Genoa and 4: Ca gliari) death rates forGC. The location of centers and their age-adjusted death ratesfor GC are presented in Figure 1 . Mortality is highest in area1 and lowest in area 4, with about a 3-fold difference in ratesbetween the areas among both males and females.All patients with histologically confirmed GC first diag-

nosed between June 1985 and December 1987 among residentsin the study areas aged 75 or less were eligible as cases andwere sought for interview. Cases were identified in surgery andgastroenterology departments and outpatient gastroscopic ser-vices of private and public hospitals. Ascertainment of caseswas compared in each center with the local cancer registry(CR) wherever available (Florence, Forli and Genoa) or pa-thology department files to evaluate completeness of reporting.Slides were sought from each case for review and diagnosticclassification according to the system of Lauren (1965) previ-ously utilized in two large histopathologic series in these samestudy areas (Amadori et al . , 1986; Amorosi et al . , 1988).Non-epithelial neoplasms of the stomach, primarily lympho-mas, were excluded from analysis.

Controls were randomly selected from 5-year age and sexstrata of the general population of each center, approximatelyin the ratio of 1:1 to the cases in each stratum. For sampling,municipal computerized lists of residents (which existed for60% of the population) or National Health Service computer-ized files (for the remaining 40%) were used. Both sourcesprovide comprehensive coverage of the resident population inthe age classes considered.A structured questionnaire, developed and tested during apilot phase, was used to obtain demographic, socio-economic,residential, occupational, smoking, medical, family and di-etary information. Dietary patterns of cases and controls wereassessed by asking the usual frequency of intake and portionsize (categorized as sma ll, medium or large) of 146 food itemsand beverages, as consumed in a 12-month period approxi-mately 2 years before the interview. For some items it wasasked whether foods were consumed preserved or fresh,cooked or raw, and whether they were prepared at home orpurchased. Questions on the habit of adding salt, on preferencefor salty foods, and on storing methods were also included.Limited data on past diet were obtained by asking about fre-quency of consumption for 17 major food groups, referring tothe time when the subject was aged 15-20 years.A group of professional interviewers was trained centrally inthe use of the questionnaire, which was administered with theaid of an instruction manual an d an atlas containing pictures ofthe more frequently consumed foods represented in 3 portionsizes. All cases and controls were interviewed personally.Cases were interviewed at the hospital (94.1%) or at theirhomes (5 .9% ); controls w ere interviewed at their homes(63 .2% ), at the local Health Department (30.2%) or elsewhere(6.6%). Further details of the data-collecting procedures arepresented elsewhere (B uiatti et al., 1989).Intakes of individual food items and of food groups were

'OTowhom reprint requests should be sent, at: Vid e Volta 171, Florence50131, Italy.Received: April 12, 1989 and in revised form May 31, 1989.


2/6

612

categorized into tertiles defincd by wcckiy frequency of con-sumption among all controls, Cumulative intakc for each foodgroup was obtained by sum ming thc frcquency of consumptionfo r individual food items in thc group. The food groups, theirconstituent foods, and thc mc an, 33rd and 67th percentile levelof consumption are shown in Appcndix 1 . Weighting the frc.qucncics of consumption by portion sire was also conducted.bu t results varied littlc and are thus not presented.ociation bctwecri GC risk and the dietaryand other exposure variables wa s the odds ratio ( OR ). AdjustedOR estimates and corresponding 95'3 confidence inten als (C1)were obtaincd by the Mantel-Haenszel technique and signii'i-cance fo r trends (ove r the tertiles) w as evaluated using theMantel extension ch i-squarc test (Breslow and Da y. 19x0).Multiva riate logistic regression analyses were also conducted .Always included in the regression mod els were age (actual agcin years, age squared) and categorical terms for sex, area (4study areas) and placc (ruraliurban) of residence, migrationfrom the south (,yesino), socio-economic status (low . rncdiurn,high, based on a combination of occupation and educationalIcvel), familial history of ( iC (0 , 1 , 2 + first-degrec familymembers with G C ) , and the Q uetclet index (tertile catcgoriesof weightiheight squa red). In addition, logistic regression mod-els including all these terms plus categorical variables fo r tcr-tile levels of consumption of one or more dietary variableswere used in ordcr to estimate the effect of food items sepa-rately or adjusted each one for the others. Models including aterm for each of thc 7 centers (instead of the 4 study arecis)were also used for analysis, but no relevant differences wercfound and these results are not prcsentcd.

K E ~ I J L . IA total of 1,229 patients with histologically confirme d GCwere identified as eligible for thc study. Of these, 5 0 (4. 1 % )refused to participate and 163 (13.2%) had died or were too il lfor interview . The analysis thus included 1,016 GC cascs. ?l'heGC diagnosis was based on surgical specimen review for 7 4 2

arid biopsy specimen examination for the remainder. The can-e r s arose primarily in the lower parts of the stomach, withonly 7 , S % in the gastric cardia or gastro-esophageal junction.Using the Lauren classification system. SS% were intestinal-type carcinomas, 23% diffuse-type, and 2 2 4 mixed or unclas-sified.Among the I , 159controls sampled from rcsidents' lists, 140(12 .1%) refused intcrview and 126 (10.9%) were no longerrcsident, were deceased, or w ere unavailable because of mentalor other health conditions. These subjects were replaced withadditional residents randomly selected from the same age andsex stratum, so the total number of interviewed controls wasj . l S 9 .Table I shows the distribution of cases and con trols by studyarea, sex, and age. About 80% of the subjects came fromhigh-risk areas and about 60% were males. The median agewas nearly 65. The G C patients more often (34% v s . 23% forcontrols) lived in rural places within each of the study areas,and tended to be of lower socio-economic status. In the 2high-risk areas in northern Italy, those who were born in thesouth of Italy and had migrated north (about 8% of the sub-jects) experienced a lower risk of GC than did natives of thearca . The cases in all areas also tended to have low er values of

itsual adult weight-for-height. The cases also more often re-ported GC in a first-degree family mcmber; across all areas,16% had 1 and 5 % ha d 2 or more relatives with GC, in contrastto 1 1 % 1 and 1 % among controls. Cases and controls did notdiffer with respect to fam ily size or history of other digestive-tract cancers in family members.

Therc was little overall case-control difference in tobaccos imking ; ORs for non-, ex-, and current cigarette smokers inlow and high pack-year catcgories were, respectively, 1 .O , 0.9(95%'CI = 0.7- 1 . 1 , 1 .0 (95% CI = 0.8-1.4) and 1.2 (95%CI = 0.9--1.7). Use of tobacco products other than cigaretteswas uncommon. There was also little trend in risk with in-creasing alcohol (mostly wine) intake. After adjusting forsmoking and the othcr factors mentioned above, the OR s fo rnon-drinkers, drinkers of wine less than 2 timesiday, 2 times/day and 2 + timedday were 1 0: 0 .9 (95% CI = 0.&1.2), 1.2(95% CI -= 0.9-1.S), nd 1.3 (95% C1 = l .(b1.8).

Table 11 presents ORs according to tertile of recent consu mp-tion for the 17 major food groups considered. Although thelevels of consum ption of these foods tended to differ among the4 study areas, the OR s were generally similar within each area.Intake of several food cate gories was associated w ith increasedGC risk. S trong rising trends in risk with increasing consump-tion were observed for traditional soups and meats, with ORsequal to or exceeding 1.8 n thc highest tertiles. More moderaterising trends were obs erved for saltedidried fish, cold cuts, andseasoned chceses.There was little or no association between G C risk and in -

'I A B L t 1 - NIJMBLKS OF CASES .4ND CONTROLS BY STUDY AKEA, S E X ,A N D AG E

354 (34)4h 7 (46)122 (12)x2 (8)640 (63)376 (37)46 ( 5 )140 (14)292 (29)538 (53)l__l__

ControlsN %_ _ _ ~ ~

371 (32)543 (47)137 (12)108 (9 )705 (61)454 (39)

70 (61155 (13)322 (28)612 (53)


3/6

G A S T RI C C A N C E R A N D D I E T IN I T A L Y 613TABLE I1 - RELATIVE RISKS OF GASTRIC CANCER ACCORDING TO LEVELOF CONSUMPTION O F GROUPS OF FOODS. RELATIVE RISKS ADJUSTED FORAGE. SEX. STUDY AREA. SOCIAL CLASS. RESIDENCE. MIGRATION FROMSOUTH, FAMILY HISTORY OF GASTRIC CANCER, QUETELET INDEX, BU TNOT FOR OTHER GROUPS OF FOODS

Food group

Bread and pastaTraditional soupsMeatsCold cutsSalted and dried fishOther fishMilk and dairy productsSeasoned cheesesRaw vegetablesCooked vegetablesBeansSpicesOniodgarlicCitrus fruitOther fresh fruitDried and preserved fruitDesserts

Tertile- _ _ _ ~_ _(low) 2 3-__

1.0 1.1 1.01.0 1 .6 2.41.0 1.3 1.81.0 1 .1 1. 21.0 1.1 1.41.0 1.0 1.21.0 1.0 1.11.0 1.2 1.21.0 0.8 0. 61.0 0.9 1. 11.0 0.8 0.81.0 0.7 0.71.0 1.0 0.81.0 0.7 0.61.0 0.6 0.41.0 0.8 1.01.0 0.7 0.8

Trendp-value0.99


4/6

614 IiSCIJSSIOU'This casc-control study revealed significant aswciationswith dietary factors that appear to contribute to the markedregional variation in GC w ithin Italy. The relatively wid e vari-ation in the levels of consumption of specific foods probablyenhanced our ability to detect dietary risk factors, despite thedifficulties inherent in retrospective studies of diet and cancer(Block, 1982 ). Perhap s foremost among the limitations of such

E'l ;\ Iinvestigations is the problem of recall of food intake and thelikelihood of misclassification of consumption levels. If thisoccuned randomly, the odds ratios would be biased towardsthe null value of 1.0 and we would have underestimated dietboth as a risk factor for GC and as a contributor to the geo-graphic patterns of GC in Italy. Since this is the largest case-control study of gastric cancer reported to date, the sample sizeovercomes some of th e loss of power due to random exposuremisclassifica tion. To reduce th e possibility of differential (non-random) recall between cases and controls due to changes indiet related to the onset or treatme nt of gastric cance r, we askedabout food intake 2 years prior to interview; the patterns offood consumption suggest no systematic over- or under-reporting by thc cases. There also seems little chance for mis-diagnosis of cance r, since all cases in this study were identifiedthrough histologic tissue examination. Nor did the cases orcontrols represent selected subsets of the population, since wesought to enroll all patients, and randomly selected controls inthe study areas, achieving a high response rate for both.Our data are qualitatively similar to those of several otherGC investigations, which found lower risk associated with in-take of fruit and vegetables ( Nom ura, 1982; Correa et ul . ,1983. 1985; Risch et ul., 1985; Trichopoulos et al., 1985; LaVecchia et ul . , 1987: Hu et a l . , 1988; You et a l . , 1988). Wefound that those consu ming high lev els of both fre sh fruits andvegetables had only 30% of the GC risk of those consuminglow levels of both. R eductions in risk we re associa ted with rawbut not cooked vegetables and with fresh but not dried orpreserved fruits, suggesting that cookingiprocessing may alterany cancer-inhibiting properties of these foods. Protective ef-tects associated with fruit and/or vegetable intake have beenobscrved in Europe. North and South America and A sia, eventhough intake patterns vary greatly, indica ting that compone ntscommon to fresh fruits and vegetables worldwide are involved.The mechanisms are not clear, although micronutrients havebeen suspected, particularly vitamin C, which can inhibit en-dogenous formation of N-nitroso compounds (Mirvish, 1983),thought to be detemiinants of GC in several areas of the world(Cornea et u l . , 1975 ), and also beta-carotene. O ne of the stron-gc r associations in ou r study was with consumption of toma-toes. a source of Iycopene, a carotenoid not efficiently con-verted to retinol in rivo (Simpson and Chichester, 1981).Fur-thermore, there were no protective effects associated withliver, dairy products and other sources of retinol. We are cur-rently obtaining data on nutrient conten t (including retinol andcarotcno ids) of various Italian foods to estimate the participantsintake of specific vitam ins and mine rals (Fidan za and Versigli-oni, 19 8l ), but the patterns reported here suggest that protec-tive effects may be due to carotenoids and not vitamin A ,similar to the situation that app ears to hold fo r lung and perhapsother cancers (Ziegler, 1989).Several investigations in experimental animals have shownthat extracts from garlic and onions have strong cancer inhib-itory properties (B clma n, 1983; Sparnin et a l . , 1986; Wargo-vich, 1987; Wargovich et u l . , 1988). Garlic also demonstratesanti-fungal and anti-bacterial pr operties (Bloc k, 1985), andmay l imit bacterial growth in the s tomach and bacteria-catalyzed conversion of nitrate to nitrite and thus reduce thepossibility for in vivo formation of N-nitroso compounds. In-vestigations in Hawaii and Greece have noted lower GC riskassociated with higher consumption of onions (Haenszel et ul . ,1972; Trichop oulos et a l . , 1985). In a case-control study of GCin a high-risk area of northeastern Chin a, reduc ed risks wereassociated with intake of each of 5 Allium vegetables, includ-ing garlic (You et d.,988, 1989). When we becam e aware ofthe Chinese findings, we added a question to our questionnaireto ascertain fre quency of in take of garlic (in addition to use ofcondiments containing onionigarlic). An inverse associationbetween cooked garlic consumption and GC risk was pro-nounced in this sample, with those in the highest one-third of


5/6

GASTRIC CANCER AND DIET IN ITALY 615intake having only 40% of the G C risk of those in the lowest.Although both garlic and onions are used mostly as flavoringagents for more complex dishes and absolute quantities aresmall, the findings provide a further incentive for evaluatingthe effects of garlic and other Allium vegetables in additionalpopulations.An interesting finding is the inverse association of GC witholive-oil consumption, which appeared to be independent ofthe protective effect of vegetable intake. Experimental databased on mammary and intestinal tumor models have sug-gested no promotional effect for olive oil, which containsmono-unsaturated fatty acids, in contrast to other types of fats(Cohen et al . , 1986; Reddy and M aeura, 1984). Whether oliveoi l or other mono-unsaturated fats have protective effects forgastric or other cancers deserves further study.

We found that more GC patients than controls reported ataste for salty foods and more often added table salt to theirfood. O ther investigations (Nomu ra, 1982; Correa et al . , 1983;Tuyns, 1983; Montes et a l . , 1985; Lu and Qin, 1987) have alsoreported that high salt intake increases GC risk. Irritation bysalt may lead to gastric atrophy and thus generate lesions whichmay eventually progress to cancer (Kodama et ul . , 1984; Mon-tes et a l . , 1985). It is difficult to adequately measure totalsodium intake because of its varying and sometimes unknowncontent in m any processed foods, so our data on salt consump-tiodpreference must be interpreted as crude approximations ofintake. Nevertheless, the temporal decline in use of salt-preserved foods and the increase in use of refrigeration tend tocoincide with the decline in stomach cancer mortality in manyparts of the world. We found direct support of this hypothesis,since risks of GC w ere higher among those who had a shorterduration of access to a refrigerator, who had never had afreezer, and who used frozen foods less frequently. In areas ofthe world where refrigerated storage is still uncom mon, such asparts of rural China, GC rates have not yet begun to decline(You et a l . , 1988).Frequent consumption of meats and cold dishes (cold cutsand seasoned cheeses) was related to increased GC risk. The

effect for cold dishes was mainly d ue to preserved meats, mostcommon of which are cooked hams and salami. Meats pre-served by nitrates have been linked to increased GC risk inother studies (Nomura, 1982; Risch et al . , 1985), with suspi-cion centering on the potential for endogenous nitrosamineformation . Italian cured meats are often preserved with nitrates

or nitrites, but we do not have precise information on thenitratehitrite concentrations of the cooked meats, which varygreatly in number and variety throughout Italy. Other Europeanstudies suggest that meat consumption could b e associated withincreased GC risk (Trichopoulos et al., 1985; Jedrychowski eta l . , 1986). In other areas of the world at high risk for GC , meatconsumption is very low and the main sources of protein arerepresented by cereals or fish.We also found a strong positive association with traditionalsoups. Foods in this broad category, heavily consumed mainlyin the high-risk north-central part of Italy, are traditionallyprepared once or twice per week and consumed as leftoversafter reboiling or heating. Since reboiling reduces bacterialcounts, these soups w ere often left unrefrigerated. Their maincomponents are broth, beans, beef/pork stuffing, bread, cornflour, rice and other starchy foods.No association between GC risk and intake of bread andpasta was observed, in contrast to a previous investigation innorthern Italy w hich implicated starchy foods, including riceand pasta (La Vecchia et al . , 1987). Some of these foods areincluded in the traditional-soups group in our study; however,for bread and pasta there wa s nearly uniformly high frequencyof intake, with even those in the lowest tertile of intake con-suming them up to twice per day. We also could not confirmthe moderate (4&50%) excess risk of stomach cancer amongcigarette smokers, as reported in several cohort and case-control studies (Nomura, 1982; Yo u et a l . , 1988; SurgeonGeneral, 1982), although there was a small (20%) excessamong the heaviest smokers. Wine drinking also was not in-dependently associated with elevated risk, contrary to reportsfrom case-control studies in Louisiana (Corre a et a l . , 1985)and France (Hoey et al . , 1981). A recent review of epidemi-ologic evidence of relation of alcohol intake to cancer con-cluded that there is little to suggest a causal role for drinkingof alcoholic beverages in stomach cancer (IARC, 1988).In conclusion, ou r study indicates that dietary habits areimportant risk factors for gastric can cer and contribute to thegeographic variation within Italy. The series of causative and

protective factors in the diet offer clues for further research toclarify the origins of gastric canc er and to help develop pre-ventive strategies.AC KNOWLEDGEM ENTS

This work w as supported by: the Italian Consiglio Nazionale

APPENDIX 1 - CONSTITUENT FOODS, TERTILE LEVELS OF WEEKLY CONSUMPTION, A N D M EA N VALUES WITHIN TERTILES FOR THE FOOD GROUPSTertiles

1 2 3_______ood group Constituent foods 2 3~~~

Mean 33% Mean 67% Mean.-Bread and pasta Al l breads, commerc ial pasta 15.1 18.1 20.7 23.0 21.0Traditional soups Meat, bean, bread and rice soup, stuffed pasta, polenta 1.5 2.6 3.7 4.9 1.5Meat Beef, mutton, pork, chicken, rabbit, offal, liver 3.8 5.2 6.1 7.0 9.3Cold cuts All kinds of preserved meats 0.9 2.2 3.5 5.2 8.9Salted and dried fish Anchovies, cod, herring 0.0 0.1 0.2 0.4 1 oOther fish Fresh and frozen fish, canned tuna 0.2 0.5 0.9 1.2 2.1Milk and dairy products Milk, yogurt, dairy cheeses 0.5 3.0 6.4 7.0 10.8Seasoned cheeses Al l seasoned cheeses 1.1 2.1 3.4 4.7 1.6Raw vegetables Salad, cucumber, tomatoes, carrots, onions, fennel 2.9 4.1 6.0 1.5 10.7Cooked vegetables All cooked vegetables 3.8 5.7 7. 1 8.5 11.9Beans Garbanzo, fava, other beans, peas, lentils 0.1 0.3 0.7 1.0 1.9Spices Peppers, chili, cloves, cinnamon, nutmeg 0.2 0.8 1.9 3.2 6.7Oniodg arlic Onion and garlic as condiments 1.2 4.6 4.6 4.6 7.2Citrus fruit Orang es, grapefruits, citrus juice 0.2 1 o 1.8 3.0 4.9Other fresh fruit Apricots, peaches, plums, cantaloup e, figs, grapes 1.2 2.4 3.2 4.2 6.9Dried and preserved fruit Dried figs, grap es, dates, cooked and canned fruit, nuts 0.0 0.1 0.6 1.0 3.6Desserts Cakes, chocolate , pastry, ice cream 0.2 0.9 1.9 4.5 8.7Weekly frequency of consurnption.-2Dividing point between first and second te~tile.-~Dividingoint between second and third tertile


6/6

616 l i i ! l A ' l ' l l E l A l . .dellc Riccrchc, Applied Project "Onco logia"; thc U .S Na-tional Cancer Institute; the Istituto Oncologico Rornagnolo inForli, Italy; the Regione Toscana, Italy; the Rcgione Emilia-Romagna, Italy: and the Italian Lega per la Lotta Contro iTumori. We are gratcful for the support of the medical, pa-thology and nursing staff in each Center for case recruitmentand diagnosis. W e arc indebted to Prof. A. Morettini (Unita diGastroenterologia, Florence), Dr. B. Lancia ( Istituto Nazio-nale della Nutrizione. Rom e), also Dr P Comba and Dr . R . secretarial assistance.

Pirastu (Istituto Superiore di Sanita, Rome) for suggestions instudy design and questionnaire development. W e thank Dr. A .Ershow, Dr. B . J . Stone and Dr. M . Gail (NCI) for advice andguidance in study design and analysis, and Dr. C. Dichter(Philade lphia) for helpful sug gestions regarding nutrient com-position. We also express appreciation to the interviewers Mrs.G . Barni, Mrs. G . Cordopatri and Mrs. D. Tanzini , also Mrs.P. Fallani (CSPO-Florence) and Mrs. H . Brown (NCI) forREFbK

AMAIXM. I ) . . PA1 1 1 , i>..ADOVAhl, f : . , IR A G O N I , A . a n d R A V A I O I, A.. Gastric caaccording to Lauren's classification in a higresidence. Tirnmri. 72 , 381-486 (1986).AMOROX.A , . B I A N C H I , . . B U I A T 7 1 , E . . ( ~ P R I A N I . r., ' A L l . 1 , D. antistric cancer in a high-risk area i n Italy. Histopathologicng to Lauren'+ classification. Conccr . 62 . 2191-2196

rlic i ) i l inhibit\ turnor promotioii. Curc !nogo-B1 O( K, F , l'hc chcini\try 0 1 garlic and onion,. .Ici. Amcr . . 251. 114- l V11985).BLO CK , . . A review ot validation\ of dietar! a.I. Epidemiol., 115, 40 2 50 5 11982).B R E S I O W . 2. and I ~ A Y ,.E.. Statistical methods in canccr research.Vo l I , 'The analysis o f case-control studies. IARC Scient(fk Pihiimioti?2 , IARC. L j o n :1980j.AVFLLINI.. , B I A N C H I ,., I P R I A N I .. , Cocc-0,P . , UE C R L I , N.. V1.h-D I C N I , C. and B to- r , W . . Methodological issues i n a iiiulticentric \tud y ofgastric canccr and diet in Italy: Ftudy design, data source\ and qualitycorrtrola. Tiitnoti. i n prcsh 119x0)~ ~ ~ S I . A ~ ~ i - 1 1 ,: I I)E( AR i 1. A , , 1.A VI.CCt1IA. {I., L A L I . X I ) A , N . , \ I t:ZZiN O T T E , G . and S M A N S . . , Dutu. sturisric.t ( l i d niups onm7cw mortcditi ,IFil/y IY7. i i i977. Pitagora, Bologna (19861.and ROSE. I)., Dietary fat and mammary cancer I . Promoting el'fects oldifferent dietary fats o il N-nitrosomethylurca-induced rat mammary t t i -m o r i g c n c h . J nut. Cunccr I rw .~7, 33 4?11986)c . . 'AJAKP0, L.F. . f i A P W S Z E 1 , W.. BOl.ANOS, 0.Diet and gastric cancer: nutrition survey i n a hiph-'cr I n s t . . 78. 673-67X (1083)

R l i i h i T1 . E . , PA1.1.1, 1) A M A r ) O R I , I ) . , M 4 K I B I N I ,

Case Control Example

Documents

Transcript of Case Control Example