International Journal of Epidemiology 1999;28:626630

Thyroid cancer is a relatively rare form of cancer, causing lessthan one per cent of cancer deaths, but with higher incidence inyounger women, and, together with gallbladder cancer, it is oneof the non sex-related site cancers whose rates are higher forfemales than for males.13

An established risk factor for thyroid cancer is a history ofbenign thyroid disease, mostly goitre, adenomas and hyper-thyroidism.49 Other recognized risk factors include exposure toradiotherapy2 and radioactive iodine,1012 mainly during

childhood,1315 and long-term residence in regions with iodine imbalances, including high endemic goitre areas,9 suchas iodine-deficient regions in the Alps, or areas with high iodinelevels, like Iceland and Hawaii.1619 Furthermore, a diet poor invegetables and fruit2023 and rich in starchy food,24 which areindicators of a poor diet in several populations, has beenassociated with an increased risk of thyroid cancer.6,7,25

Between 1986 and 1992 we conducted a case-control studyof thyroid cancer in Italy, whose results confirmed the strongassociation with history of benign thyroid disease.9 Other deter-minants for thyroid cancer risk were a history of radiotherapy,long-term residence in endemic goitre areas9 and selected aspectsof diet.24 The role of each of these factors, or their combination,depends not only on the strength of the association, but also onthe prevalence of each factor in the population. To address theseissues, we estimated the population attributable risk (PAR) ofthyroid cancer, i.e. the proportion of thyroid cancer cases in

International Epidemiological Association 1999 Printed in Great Britain

Case-control study of thyroid cancer in Northern Italy: attributable risk Francesca Fioretti,a Alessandra Tavani,a Silvano Gallus,a Silvia Franceschi,b Eva Negria and Carlo La Vecchiaa,c

Background The percentage of thyroid cancer cases attributable to specific risk factors can becalculated to focus preventive strategies. The per cent population attributablerisks (PAR) for thyroid cancer were estimated in relation to history of benignthyroid diseases, history of radiotherapy, residence in endemic goitre areas andselected indicators of a poor diet, using data from a case-control study conductedbetween 1986 and 1992 in Northern Italy.

Methods Cases were 399 histologically confirmed incident thyroid cancers and controlswere 617 patients, admitted to hospital for a wide range of acute, non-neoplastic,non-hormone-related diseases. The PAR were computed on the basis of multi-variate odds ratios (OR) and on the distribution of risk exposure among cases,assuming they are representative of the general population of cases.

Results A history of benign thyroid disease accounted for 18.9% of cases, radiotherapyfor 1.2%, residence for >20 years in endemic goitre areas for 2.4% of cases, andtheir combination for 21.7% of thyroid cancer cases; selected indicators of a poordiet accounted for 40.9% of thyroid cancer cases in this population. The com-bination of all factors considered explained over 57% of thyroid cancer cases inboth sexes. The estimates for thyroid-related conditions were higher in womenthan men, whereas the opposite was true for dietary indicators. The overall PARwere somewhat higher in people aged >45 years (63.8%) than in youngersubjects, and for follicular (69.1%) rather than papillary (53.7%) cancers.

Conclusions Exposure to a few simply identified and potentially modifiable risk factors orindicators (benign thyroid disease, residence in endemic goitre area and a poordiet) explained about 60% of thyroid cancer cases in this Italian population,indicating the theoretical scope for prevention.

Keywords Attributable risk, diet, radiotherapy, risk factors, thyroid cancer, benign thyroiddiseases

Accepted 3 February 1999

a Istituto di Ricerche Farmacologiche Mario Negri, Via Eritrea 62, 20157Milan, Italy.

b Centro di Riferimento Oncologico, Via Pedemontana Occidentale, 33081,Aviano (PN), Italy.

c Istituto di Statistica Medica e Biometria, Universit degli Studi di Milano,Via Venezian 1, 20133 Milan, Italy.

Reprint requests to: Francesca Fioretti, Istituto di Ricerche FarmacologicheMario Negri, Via Eritrea 62, 20157 Milano, Italy. E-mail: tavani@irfmn.mnegri.it

626

Northern Italy attributable to selected benign thyroid diseases,radiotherapy, long-term residence in endemic goitre areas andselected indicators of a poor diet.

Materials and MethodsThe data were derived from a case-control study of cancer of the thyroid, conducted between 1986 and 1992 in NorthernItaly (the Greater Milan area, Friuli Venezia Giulia and Venetoregions), whose general design has already been described.26

Cases were patients older than 15 years and under age 75with incident (diagnosed within 2 years before the interview),histologically confirmed thyroid cancer, admitted to the eightmajor teaching and general hospitals in the areas under surveil-lance. A total of 399 patients (108 men and 291 women, medianage 44 years, range 1672) were interviewed. Of these, 274 hadpapillary carcinomas (or mixed papillary/follicular), 69 follicular,and 56 anaplastic or other or undefined histological types.

Controls were patients from the same geographical areas,admitted to the same network of hospitals, during the samecalendar period as cases for acute, non-neoplastic, non-hormonal-related diseases, unrelated to known or likely riskfactors for thyroid cancer. A total of 617 controls (190 men and427 women, median age 46 years, range 1674 years) wereinterviewed. Of these, 15% were admitted for traumatic con-ditions, 17% for non-traumatic orthopaedic disorders (mainlylow back pain and disc diseases), 28% for acute surgical con-ditions (such as acute appendicitis or strangulated hernia), and 40% for other miscellaneous illnesses (such as ear, noseand throat, eye, dental or skin disorders). Over 95% of casesand controls approached for interview participated.

Interviews were conducted by trained interviewers in ahospital setting, using a structured questionnaire to collect dataon socio-demographic and anthropometric characteristics,smoking habits, and other lifestyle details, history of radio-therapy (unrelated to the present thyroid disease), and time ofresidence in endemic goitre areas. Information on medicalhistory included questions on episodes of thyroiditis, benignnodules (adenomas), goitre, hyper- or hypothyroidism and otherunspecified thyroid disease, diagnosed more than one yearbefore the index diagnosis, and history of any radiotherapy.Benign pathologies were combined in a single variable (anybenign thyroid disease) for the purpose of the present analysis.The food frequency questionnaire comprised 29 items, includ-ing questions on intake of refined cereals, such as bread, pasta,rice and polenta; the frequency of their total intake was com-puted by adding single weekly frequencies. Details of the foodfrequency questionnaire have been given elsewhere.24 Thereproducibility of the questionnaire was satisfactory.27 Beta-carotene intake was computed from the frequency of weeklyintake of the main vegetables and fruit, using standard Italianfood composition tables.28 In this population a high intake ofrefined cereals and low consumption of vegetables and fruitmay be considered indicative of a generally poorer diet, i.e. alsopoor in several other aspects.

Data analysis

Odds ratios (OR) of thyroid cancer, and the corresponding 95% confidence intervals (CI), were derived using unconditionalmultiple logistic regression.29 The model included terms for age

at diagnosis, sex, education, history of benign thyroid disease,radiotherapy, residence in endemic goitre areas, and intake ofrefined cereals and beta-carotene.

Using the multivariate OR and the distribution of the riskfactors in the cases, percentage PAR were computed, i.e. theproportion of thyroid cancers that would be avoided if all sub-jects were unexposed, or exposed to the level associated withthe lowest risk. The method described by Bruzzi et al.30 impliesknowledge of the risk estimates and of the joint distribution ofthe risk factors in the population of cases only, and thus can beused for hospital-based case-control studies. The PAR wereexpressed as per cent terms, and their variance calculations and95% CI were obtained as described by Benichou and Gail31 foreach separate risk factor and for various combinations of themwhenever computationally possible.32

ResultsTable 1 gives the distribution of thyroid cancer cases and thecomparison group according to age, sex and other risk factors.

Table 2 gives the corresponding multivariate OR according to various risk factors in the overall dataset and in strata of sex, age, and major histological types (papillary and follicularcarcinomas). History of any benign thyroid disease was consist-ently associated with cancer risk (OR = 7.6), with an apparentlystronger association in men (OR = 37.1) than women (OR = 6.2),

THYROID CANCER IN NORTHERN ITALY 627

Table 1 Distribution by sex of 399 thyroid cancer cases and 617 controlsaccording to age and other risk factors. Italy, 19861992

Thyroid cancer cases Controls

Women Men Women Men

Age at diagnosis (years)

,40 115 37 157 74

4054 95 39 130 65

>55 81 32 140 51

Education (years)

,7 133 45 228 69

>7 158 63 199 121

Benign thyroid diseases

No 220 92 407 189

Yes 71 16 20 1

Radiotherapy

No 276 105 412 186

Yes 15 3 15 4

Residence in endemic goitre area (years)

Never or ,20 266 101 405 178

>20 25 7 22 12

Refined cereal intake (portions/week)a

,14.5 72 9 162 30

14.5,22 130 38 158 67

>22 89 61 107 93

Beta-carotene intake (m g/week)a

>36 380 85 32 152 54

24 370,36 380 93 32 148 58

,24 370 113 44 127 78

a Approximate tertiles of intake.

while there was no real heterogeneity across strata of age andhistological types. History of radiotherapy and residence for>20 years in endemic goitre areas also increased thyroid cancerrisk (OR = 1.4 for both). The association with radiotherapy wasapparently stronger at younger age (OR = 2.0 for patients aged,45 years). The OR was 2.0 for the highest tertile of refinedcereal intake compared to the lowest, and 1.4 for the lowestintake of beta-carotene compared to the highest, with noheterogeneity across strata of the selected covariates.

The PAR for selected risk factors and their combinations areshown in Table 3, assuming all subjects could be moved to thelowest exposure levels. History of any benign thyroid diseaseaccounted for 18.9% of thyroid cancer cases, history of radio-therapy for 1.2% and long-term residence in endemic goitre areafor 2.4%. The latter two estimates, however, were not significant.The selected indicator of a poor diet explained 40.9% of cases.Although the OR for benign thyroid diseases was higher in menthan women, the proportion of thyroid cancer cases attributableto any benign thyroid disease was higher among women (PAR20.5% compared to 14.4% in men), due to the higher fre-quency of the diseases among women; the PAR for any benignthyroid disease were higher among younger subjects (21.0%)and among patients with follicular carcinoma (26.5%). The PARfor the combination of a high intake of refined cereals and lowbeta-carotene was higher in men (81.1%) than women (33.2%)and at age >45 years (57.7%, compared to 28.4% at age ,45).The combination of radiotherapy, residence in endemic goitreareas and any benign thyroid diseases accounted for 21.7% ofcases. The combination of all variables accounted for 57.5% ofthyroid cancer cases and was very similar in women (58.6%)and men (58.3%), and apparently higher at age >45 (63.8%)

than in younger subjects (52.7%). With reference to the mainhistological types considered, the overall PAR were somewhathigher for follicular (69.1%) than papillary carcinomas (53.7%).

DiscussionThis study indicates that in this Italian population, more thanhalf of thyroid cancer cases may be accounted for by a combina-tion of a few identifiable factors or risk factor correlates such asa history of any benign thyroid disease, history of radiotherapy,long-term residence in endemic goitre areas and some basicindicators of a poor diet.

History of residence in endemic goitre areas, although a wellrecognized factor for the disease,24 accounted only for a smallproportion of thyroid cancers, and the corresponding estimatewas not significant. Such a proportion, however, may have beenunderestimated by the fact that current residence was com-parable for cases and controls, and current residence in Italy is strongly correlated to past one. Likewise, the PAR for radio-therapy was not significant. A minimum latency of 3 years wasconsidered between radiation and diagnosis of thyroid cancer,but our definition of radiotherapy included not only conditionsrequiring heavy exposure (angioma of the neck, thyroid nod-ules, enlarged thymus) but also degenerative arthritis (mainlyof the neck) which implied relatively low-dose irradiation of theanterior wall.26 History of any benign thyroid disease, althoughstrongly associated in terms of OR, accounted for less than 20%of thyroid cancer cases. Although the OR for the combination of selected indicator foods was only moderately elevated, thisexplained over 40% of thyroid cancer cases, suggesting thetheoretical scope for dietary intervention on thyroid cancer.

628 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

Table 2 Odds ratios with 95% confidence interval (CI) of 399 cases of thyroid cancer and 617 controls according to selected risk factors. Italy,19861992

Odds ratios (95% CI)a

Sex Age (years) Histological type

All Women Men ,45 >45 Papillary Follicular

Any benign thyroid diseases

No 1c 1c 1c 1c 1c 1c 1c

Yes 7.6 (4.6, 12.6) 6.2 (3.6, 10.6) 37.1 (4.7, 291.3) 9.3 (4.4, 19.7) 6.5 (3.2, 12.9) 7.6 (4.5, 13.1) 11.6 (5.6, 23.8)

Radiotherapy

No 1c 1c 1c 1c 1c 1c 1c

Yes 1.4 (0.7, 2.8) 1.4 (0.6, 3.0) 1.3 (0.3, 6.2) 2.0 (0.6, 6.8) 1.1 (0.5, 2.6) 1.5 (0.7, 3.3) 2.0 (0.6, 6.4)

Residence in endemic goitre areas (years)

No or ,20 1c 1c 1c 1c 1c 1c 1c

>20 1.4 (0.8, 2.5) 1.7 (0.9, 3.2) 1.2 (0.4, 3.5) 2.2 (0.9, 5.7) 1.2 (0.6, 2.4) 1.4 (0.8, 2.7) 2.3 (0.9, 5.7)

Refined cereal intake (portions/week)b

,14.5 1c 1c 1c 1c 1c 1c 1c

14.5,22 1.9 (1.4, 2.7) 2.0 (1.3, 2.9) 2.3 (0.9, 5.7) 1.6 (0.9, 2.6) 2.5 (1.5, 4.1) 1.8 (1.2, 2.7) 2.6 (1.2, 5.4)

>22 2.0 (1.4, 2.9) 2.0 (1.3, 3.1) 2.5 (1.0, 6.0) 2.0 (1.2, 3.6) 2.2 (1.3, 3.7) 2.2 (1.4, 3.3) 2.3 (1.0, 5.0)

Beta-carotene intake (m g/week)b

>36 380 1c 1c 1c 1c 1c 1c 1c

24 370,36 380 1.0 (0.7, 1.4) 1.0 (0.6, 1.4) 1.0 (0.5, 1.9) 1.0 (0.6, 1.5) 1.0 (0.6, 1.7) 0.8 (0.6, 1.2) 0.9 (0.4, 1.8)

,24 370 1.4 (1.0, 1.9) 1.6 (1.1, 2.4) 0.9 (0.5, 1.6) 1.3 (0.8, 2.1) 1.5 (0.9, 2.4) 1.2 (0.8, 1.8) 1.6 (0.8, 3.0)

a Estimates from multiple logistic regression equations, including terms for age, sex, education and all the variables included in the Table.b Approximate tertiles of intake.c Reference category.

An inverse relation with vegetable and fruit intake wasconsistently observed in several studies.6,7,24,26 In this Italianpopulation, high starch intake may simply be an indicator of adiet poor in several aspects, possibly also iodine, whose defici-ency is related to both benign and malignant thyroid diseases.2

A diet poor in vegetables and fruit and rich in starches may alsobe correlated to non-dietary risk factors, although allowing foreducationas an indicator of socioeconomic statusdid notnotably modify the observed association.

Our general definition of benign thyroid disease includesconditions strongly related to cancer risk (goitre and nodules)and others less strongly related (hyperthyroidism andthyroiditis).2,5,6,9 The absolute number of cases for each benigndisease was too low to allow separate PAR estimations. Theoverall estimate in any case is a weighted average of thestrength of the association and prevalence of various diseasesand should therefore provide, at least in first approximation, ameasure of the role of all benign thyroid disease at the popu-lation level.

With reference to OR estimates, this study has some limita-tions as well as several strengths typical of hospital-based case-control studies.33 This is one of the largest case-control studiesof thyroid cancer to date; only acute conditions, unrelated toknown or likely risk factors for thyroid cancer, were included inthe comparison group; the distribution of cases and controls was comparable in terms of area of current residence, and theparticipation of cases and controls was almost complete, thuslimiting the extent of any potential selection bias. However, the areas under investigation were not covered by cancerregistry and total numbers of incident cases are unidentified.The upper age limit at 75 years, introduced to increase reliabilityof information collected, has led to the exclusion of less than15% of cases in this population.34 The potential confounding

effect of several covariates was allowed for in the analysis,although allowance for confounding may have been incom-plete, on account of computational constraints in the size of thematrix for estimation of variance.31,32 Any such underadjust-ment, however, should be limited, since the PAR estimates werepractically unchanged when different terms were introduced inthe models. Furthermore, the PAR for dietary factors are verylikely underestimated, since in the absence of an unexposedcategory, these are influenced by the definition of quantiles ofintake or any other arbitrary cut-off, and consequently by theshape of the dose-risk relationship.

Computation of PAR implies that the cases are representativeof all cases in the population studied.30 Such an assumption isreasonable, since cases were identified in the major teachingand general hospitals of the areas under surveillance, althoughin the absence of a cancer registry the exact proportion of casesinterviewed is not known. To improve the reliability andvalidity of the information collected,27 we included only sub-jects below age 75 years, so the PAR applied to this age grouponly.

Although the causal relationship for some of the risk factorsconsidered remains open to discussion, since they may re-present indicators of risk, and although their modification is not obvious, intervention on a few identified and potentiallymodifiable conditions is likely to be relevant on a public healthscale. In 1994 in Italy there were about 400 deaths from thyroidcancer in women and 200 in men.35 Surveillance of benignthyroid diseases, supplementing iodine in the diet in endemicgoitre areas and some modification in the diet (useful also in theprevention of other more common chronic diseases) are likelyto avoid over 300 deaths per year in Italy, mainly in youngerwomen, thus indicating the scope and importance of preventivehealth measures for this disease.2,3

THYROID CANCER IN NORTHERN ITALY 629

Table 3 Per cent attributable risksa with 95% confidence interval (CI), according to selected risk factors in strata of selected covariates. Italy,19861992

Per cent attributable risks (95% CI)a

Sex Age (years) Histological type

All cases Women Men ,45 >45 Papillary Follicular

Benign thyroid diseases 18.9 20.5 14.4 21.0 17.0 19.4 26.5

(14.6, 23.3) (15.0, 26.0) (7.6, 21.2) (14.7, 27.3) (10.9, 23.1) (14.1, 24.6) (15.4, 37.6)

Radiotherapy 1.2 1.4 0.7 2.0 0.4 1.7 2.8

(1.5, 3.9) (2.0, 4.8) (3.1, 4.5) (1.3, 5.4) (3.8, 4.6) (1.5, 5.0) (3.3, 8.9)

Residence in endemic goitre areas 2.4 3.4 1.2 3.5 1.7 2.3 6.6

(1.2, 6.0) (0.9, 7.7) (4.9, 7.4) (0.6, 7.7) (4.1, 7.5) (1.8, 6.4) (1.9, 15.0)

Indicator of poor dietb 40.9 33.2 81.1 28.4 57.7 24.5 67.2

(14.4, 67.5) (2.6, 63.8) (42.2, 120.0) (12.1, 68.9) (25.4, 90.1) (10.7, 59.8) (27.1, 107.2)

Benign thyroid diseasesand goitre area 20.7 22.8 15.5 23.3 18.6 21.0 31.9

(15.5, 25.8) (16.6, 29.1) (11.0, 26.2) (15.0, 27.1) (19.3, 44.5)

Radiotherapy, goitre areaand benign thyroid diseases 21.7 24.0 16.2 24.7 18.7 22.3 34.4

Benign thyroid diseases

and indicator of poor dietb 56.0 56.7 57.5 50.4 63.0 51.8 66.1

All the above 57.5 58.6 58.3 52.7 63.8 53.7 69.1

a Based on OR reported in Table 2, after allowance for age, sex and all variables included in the Table.b Poor diet is defined as a high intake of refined cereals and low intake of beta-carotene.

630 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

AcknowledgementsThis work was conducted with the contribution of the ItalianAssociation for Cancer Research, Milan. The authors thank MrsJ Baggott, MP Bonifacino and the GA Pfeiffer Memorial Librarystaff for editorial assistance.

References1 La Vecchia C, Lucchini F, Negri E, Boyle P, Maisonneuve P, Levi F.

Trends of cancer mortality in Europe, 19551989: IV. Urinary tract,eye, brain and nerves, and thyroid. Eur J Cancer 1992;28A:121081.

2 Franceschi S, Boyle P, La Vecchia C, Burt AD, Kerr DJ, MacFarlane GJ.The epidemiology of thyroid carcinoma. Crit Rev Oncog 1993;4:2552.

3 Franceschi S, La Vecchia C. Thyroid cancer. Cancer Surveys 1994;19/20:393422.

4 McTiernan AM, Weiss NS, Daling JR. Incidence of thyroid cancer inwomen in relation to reproductive and hormonal factors. Am JEpidemiol 1984;120:42335.

5 Preston-Martin S, Bernstein L, Pike MC, Maldonado AA, HendersonBE. Thyroid cancer among young women related to prior thyroiddisease and pregnancy history. Br J Cancer 1987;55:19195.

6 Ron E, Kleinerman RA, Boice JD Jr, LiVolsi VA, Flannery JT,Fraumeni JF Jr. A population-based case-control study of thyroidcancer. J Natl Cancer Inst 1987;79:112.

7 Kolonel LN, Hankin JH, Wilkens LR, Fukunaga FH, Hinds MW. Anepidemiologic study of thyroid cancer in Hawaii. Cancer Causes Control1990;1:22334.

8 Goldman MB, Monson RR, Maloof F. Cancer mortality in womenwith thyroid disease. Cancer Res 1990;50:228389.

9 DAvanzo B, La Vecchia C, Franceschi S, Negri E, Talamini R. Historyof thyroid disase and subsequent thyroid cancer risk. Cancer EpidemiolBiomarkers Prev 1995;4:19399.

10 Wingren G, Hallquist A, Hardell L. Diagnostic X-ray exposure andfemale papillary thyroid cancer: a pooled analysis of two Swedishstudies. Eur J Cancer Prev 1997;6:55056.

11 Schneider AB, Recant W, Pinsky SM, Ryo UY, Bekerman C, Shore-Freedman E. Radiation-induced thyroid carcinoma: clinical courseand results of therapy in 296 patients. Ann Intern Med 1986;105:40512.

12 Archer VE. Risk of thyroid cancer after diagnostic doses of radio-iodine. J Natl Cancer Inst 1989;81:71315.

13 Ron E, Modan B, Preston D, Alfandary E, Stovall M, Boice JD Jr.Thyroid neoplasia following low-dose radiation in childhood. RadiatRes 1989;120:51631.

14 Williams ED. Relation between 131-I therapy for thyrotoxicosis anddevelopment of thyroid carcinoma. Lancet 1986;ii:45657.

15 Holm LE, Wiklund KE, Lundell GE et al. Thyroid cancer after diag-nostic doses of iodine-131: a retrospective cohort study. J Natl CancerInst 1988;80:113238.

16 Report of the Subcommittee for the Study of Endemic Goitre andIodine Deficiency of the European Thyroid Association: goitre andiodine deficiency in Europe. Lancet 1985;i:128993.

17 Hrafnkelsson J, Jonasson JG, Sigurdsson G, Sigvaldason H, Tulinius H.Thyroid cancer in Iceland 19551984. Acta Endocrinol 1988;118:56672.

18 Parkin DM, Whelan SL, Ferlay J et al. (eds). Cancer Incidence in FiveContinents, Vol. VII. IARC Scientific Publications No.143. Lyons,France: International Agency for Research on Cancer, 1997.

19 Goodman MT, Yoshizawa CN, Kolonel LN. Descriptive epidemiologyof thyroid cancer in Hawaii. Cancer 1988;61:127281.

20 Preston-Martin S, Jin F, Duda MJ, Mack WJ. A case-control study ofthyroid cancer in women under age 55 in Shanghai (PeoplesRepublic of China). Cancer Causes Control 1993;4:43140.

21 Ron E, Modan B. Thyroid. In: Schottenfeld D, Fraumeni JP Jr (eds).Cancer Epidemiology and Prevention. Philadelphia: WB Saunders, 1982,pp.83754.

22 Glattre E, Haldorsen T, Berg JP, Stensvold I, Solvoll K. Norwegiancase-control study testing the hypothesis that seafood increases therisk of thyroid cancer. Cancer Causes Control 1993;4:1116.

23 DAvanzo B, Ron E, La Vcchia C, Franceschi S, Negri E, Ziegler R.Selected micronutrient intake and thyroid carcinoma risk. Cancer1997;79:218692.

24 Franceschi S, Levi F, Negri E, Fassina A, La Vecchia C. Diet and thyroidcancer: a pooled analysis of four European case-control studies. Int JCancer 1991;48:39598.

25 Wingren G, Hatschek T, Axelson O. Determinants of papillary cancerof the thyroid. Am J Epidemiol 1993;138:48291.

26 Franceschi S, Fassina A, Tolamini R et al. Risk factors for thyroidcancer in northern Italy. Int J Epidemiol 1989;18:57884.

27 Dvanzo B, La Vecchia C, Katsouyanni K, Negri, E, Trichopoulos D. An assessment, and reproducibility of food frequency data providedby hospital controls. Eur J Cancer Prev 1997;6:28893.

28 Salvini S, Gnagnarella P, Parpinel MT et al. The food compositiondatabase for an Italian food frequency questionnaire. J Food ComposAnal 1996;9:5771.

29 Breslow NE, Day NE. Statistical Methods in Cancer Research Vol.1: TheAnalysis of Case-Control Studies. IARC Scientific Publications No.32.Lyon, France: International Agency for Research on Cancer, 1980.

30 Bruzzi P, Green SB, Byar DP, Brinton LA, Schairer C. Estimating thepopulation attributable risk for multiple risk factors using case-controldata. Am J Epidemiol 1985;122:90414.

31 Benichou J, Gail MH. Variance calculations and confidence intervalsfor estimates of the attributable risk based on logistic models.Biometrics 1990;46:9911003.

32 Mezzetti M, Ferraroni M, Decarli A, La Vecchia C, Benichou J.Software for attributable risk and confidence interval estimation incase-control studies. Comput Biomed Res 1996;29:6375.

33 Mantel N, Haenszel W. Statistical aspects of the analysis of data fromretrospective studies of disease. J Natl Cancer Inst 1959;22:71948.

34 Parkin DM, Whelan SL, Ferlay J et al. (eds). Cancer Incidence in FiveContinents, Vol. VII. IARC Scientific Publications No.143. Lyons,France: International Agency for Research on Cancer, 1997, pp.57073.

35 Decarli A, La Vecchia C, Cislaghi C, Negri E. Cancer mortality in Italy,1994, and an overview of trends from 1955 to 1994. Tumori 1998;84:31234.