Cancer incidence in persons with type 1 diabetes: a five-country study of 9,000 cancers in type 1 diabetic individuals

ARTICLE Cance incidence in pe sons wi h ype 1 diabe es: a i e-coun y s udy o 9,000 cance s in ype 1 diabe ic indi iduals Bendix Ca s ensen 1 &S ephanie H Read 2 &Sø en F iis 3 &Reijo Sund 4 &Ilmo Keskimäki 5 & Ann-Ma ie S ensson 6 &Ricka d Ljung 7 &Sa ah H Wild 2 &Joannes J Ke ssens 8 & Jessica L Ha ding 9 &Dianna J Magliano 9 &So ia Gudbjö nsdo i 6 & on behal o he Diabe es and Cance Resea ch Conso ium Recei ed: 4 No embe 2015 /Accep ed: 12 Janua y 2016 /Published online: 29 Feb ua y 2016 #The Au ho (s) 2016. This a icle is published wi h open access a Sp inge link.com Abs ac Aims/hypo hesis An excess cance incidence o 20–25% has been iden i ied among pe sons wi h diabe es, mos o whom ha e ype 2 diabe es. We aimed o desc ibe he associa ion be ween ype 1 diabe es and cance incidence. Me hods Pe sons wi h ype 1 diabe es we e iden i ied om i e na ionwide diabe es egis e s: Aus alia (2000–2008), Denma k (1995–2014), Finland (1972–2012), Sco land (1995–2012) and Sweden (1987–2012). Linkage o na ional cance egis ies p o ided he numbe s o inciden cance s in people wi h ype 1 diabe es and in he gene al popula ion. We used Poisson models wi h adjus men o age and da e o ollow up o es ima e haza d a ios o o al and si e-speci ic cance s. Resul s A o al o 9,149 cance s occu ed among pe sons wi h ype 1 diabe es in 3.9 million pe son-yea s. The median age a cance diagnosis was 51.1 yea s (in e qua ile ange 43.5– 59.5). The haza d a ios (HRs) (95% CIs) associa ed wi h ype 1 diabe es o all cance s combined we e 1.01 (0.98, 1.04) among men and 1.07 (1.04, 1.10) among women. HRs we e inc eased o cance o he s omach (men, HR 1.23 [1.04, 1.46]; women, HR 1.78 [1.49, 2.13]), li e (men, HR 2.00 [1.67, 2.40]; women, HR 1.55 [1.14, 2.10]), panc eas (men, HR 1.53 [1.30, 1.79]; women, HR 1.25 [1.02,1.53]), endome- ium (HR 1.42 [1.27, 1.58]) and kidney (men, HR 1.30 [1.12, 1.49]; women, HR 1.47 [1.23, 1.77]). Reduced HRs we e ound o cance o he p os a e (HR 0.56 [0.51, 0.61]) and b eas (HR 0.90 [0.85, 0.94]). HRs declined wi h inc easing diabe es du a ion. Conclusion Type 1 diabe es was associa ed wi h di e ences in he isk o se e al common cance s; he s eng h o hese associa ions a ied wi h he du a ion o diabe es. Keywo ds Cance incidence .Cance a e a io .Cance sub ypes .Diabe es du a ion Diabe ologia (2016) 59:980–988 DOI 10.1007/s00125-016-3884-9 Elec onic supplemen a y ma e ial The online e sion o his a icle (doi:10.1007/s00125-016-3884-9) con ains pee - e iewed bu unedi ed supplemen a y ma e ial, which is a ailable o au ho ised use s. *S ephanie H Read S ephanie. [email protected] on behal o he Diabe es and Cance Resea ch Conso ium 1 S eno Diabe es Cen e, Gen o e, Denma k 2 Ushe Ins i u e o Popula ion Heal h Sciences & In o ma ics, Uni e si y o Edinbu gh, Te io Place, Edinbu gh EH8 9AG, Sco land, UK 3 Ins i u e o Cance Epidemiology, Danish Cance Socie y, Copenhagen, Denma k 4 Cen e o Resea ch Me hods, Depa men o Social Resea ch, Uni e si y o Helsinki, Helsinki, Finland 5 Di ision o Heal h and Social Se ices, Na ional Ins i u e o Heal h and Wel a e, Helsinki, Finland 6 Depa men o Medicine, Sahlg enska Uni e si y Hospi al, Uni e si y o Go henbu g, Go henbu g, Sweden 7 Ins i u e o En i onmen al Medicine, Ka olinska Ins i u e , S ockholm, Sweden 8 In o ma ion Se ices, NHS Na ional Se ices Sco land, Edinbu gh, Sco land, UK 9 Depa men o Clinical Diabe es and Epidemiology, Bake IDI Hea and Diabe es Ins i u e, Melbou ne, Aus alia In oduc ion Pe sons wi h diabe es ha e an app oxima ely 20–25% highe cance incidence compa ed wi h pe sons wi hou diabe es, hough his a ies by cance si e [1]. In pa icula , pe sons wi h diabe es ha e been shown o ha e an ele a ed incidence o li e , panc ea ic, colo ec al, endome ial and kidney cance [1]. Con e sely, a dec eased isk o p os a e cance has been epo ed in men wi h diabe es [2–5]. The associa ion be ween ype 1 diabe es and cance is no well desc ibed. The majo i y o p e ious s udies assessing he link be ween diabe es and cance ha e no made a dis inc ion be ween he wo majo ypes o dia- be es. As ype 2 diabe es is a mo e p e alen han ype 1 diabe es, s udy popula ions ha e been p ima ily com- posed o pe sons wi h ype 2 diabe es, hus hinde ing he gene alisa ion o indings o pe sons wi h ype 1 diabe- es. I is possible ha he ela ionship be ween ype 1 diabe es and cance is di e en om ha obse ed be- ween ype 2 diabe es and cance as a esul o di e - ences in he unde lying disease cha ac e is ics, d ug he - apies and pa e ns o isk ac o s, such as obesi y. The obse ed excess isk o cance in pe sons wi h diabe es may, o some ex en , be a consequence o an i- diabe ic d ug he apies, such as exogenous insulin [6]. The e is some e idence ha insulin- ea ed pa ien s ha e a highe cance incidence compa ed wi h non-insulin- ea ed pa ien s [7]. I exogenous insulin use is associ- a ed wi h inc eased cance isk, hen an excess isk o cance should be appa en in pe sons wi h ype 1 dia- be es, and his excess isk could po en ially be la ge han ha obse ed in pe sons wi h ype 2 diabe es. Al e na i ely, hype glycaemia has also been sugges ed as a possible mechanism h ough which diabe es is as- socia ed wi h cance [8,9]. I hype glycaemia does con- ibu e o he obse ed ele a ed isks o cance among pe sons wi h ype 2 diabe es hen simila associa ions would be expec ed o bo h ype 1 diabe es and ype 2 diabe es pa ien s. The small numbe o exis ing s udies in es iga ing cance occu ence among pe sons wi h ype 1 diabe es [10–13] ha e been unde powe ed o p o ide accu a e isk es ima es o si e-speci ic cance s and ha e subse- quen ly epo ed he e ogeneous indings [14]. Fu he limi a ions o p e ious s udies include un ep esen a i e samples, di icul ies in de ining ype 1 diabe es, sho ollow up and inadequa e e alua ion o he po en ial in luence o asce ainmen bias on diabe es du a ion. In his la ge mul ina ional s udy, we compa ed cance inci- dence among pe sons wi h ype 1 diabe es and he gene al popula ion using popula ion-based egis ies in i e coun ies. The e ec o diabe es du a ion on he associa ion wi h cance was also explo ed. Me hods Da a sou ces The da a sou ces o his p ojec we e compiled om he ol- lowing i e coun ies: Aus alia, Denma k, Finland, Sco land and Sweden. De ails o each coun y’s sou ce o diabe es, cance and mo ali y da a, as well as de ails o e hical app o - al, a e p o ided in he elec onic supplemen a y ma e ial (ESM). The s udy pe iods o each coun y we e: Aus alia, 2000–2008; Denma k, 1995–2012; Finland, 1972–2010; Sco land, 1995–2011; and Sweden, 1987–2012. ICD-10 (www.who.in /classi ica ions/icd/en/) and ICD-7 codes we e used o classi y indi idual cance diagnoses. In each coun y, pe sons wi h ype 1 diabe es we e de ined as pe sons who we e eco ded wi h a diagnosis o diabe es below he age o 40 yea s. Follow up ime and he numbe o cance s in pe sons wi h ype 1 diabe es we e classi ied by coun- y, sex, age and calenda ime (in 1 yea ca ego ies) and by Popula ion size by sex, age and calenda ime we e ob ain- ed om he espec i e na ional s a is ics and used o es ima e pe son-yea s a isk o each coun y. S a is ical me hods De ini ion o ou come We ollowed pe sons wi h ype 1 dia- be es om he s udy s a da e o he da e o diabe es diagno- sis, whiche e was la es , and excluded pa ien s wi h a p e- exis ing cance diagnosis. All indi iduals we e ollowed un il he i s cance occu ence, dea h o s udy end da e, whiche e came i s . The incidence o cance was de ined by he i s p ima y cance only and he da e o cance diagnosis was e ie ed om he espec i e cance egis ies. As he p e alence o ype 1 diabe es is low (<1%), we chose o compa e cance incidence in pe sons wi h ype 1 diabe es wi h ha in he o al backg ound popula ion a he han in a non-diabe ic popula ion. Fo con enience, we also chose o measu e ollow up among pe sons wi h ype 1 dia- be es up o he da e o dea h o he end o s udy, and igno ed he da e o cance diagnosis o a oid he ecalcula ion o pe son-yea s a isk o each cance ype. This app oach o e es ima ed he ollow up ime o pe sons wi h ype 1 dia- be es by <5% o all cance s combined and by much less han 5% o cance a speci ic si es (see h p://bendixca s ensen. com/DMCa/T1D/T1D-Ca.pd (accessed 8 Janua y 2016), sec ion 8.4.2, o sensi i i y analyses). We censo ed ollow up a cance diagnosis in he analysis o all cance s and in he g ouping o non-sex-speci ic cance s. S a is ical models Fo each cance si e and sex, da a we e classi ied by coun y, age and pe iod o ollow up, da e o Diabe ologia (2016) 59:980–988 981 diabe es du a ion (subdi ided a 0, 1, 2, 5, 10, 15 and 30 yea s). bi h (i.e. coho ), and gene al popula ion s diabe es pa ien s ( o disease du a ion). We i ed an age–pe iod–coho model o he cance incidence da a sepa a ely o each sex and o each s udied cance si e. The models we e i ed using cubic spline e ms o age, pe iod (da e o ollow up) and coho (da e o bi h), and using he e en s as ou come and he na u al log o pe son-yea s as he o se in a Poisson eg ession model [15,16]. Spline kno s we e chosen so he numbe o cases be ween consecu i e kno s was he same ac oss each o he ollowing a iables: age, pe iod and coho . We assumed a common e ec o ype 1 diabe es in he i s model and a se o common du a ion e ms in he second model. Thus, he model o he cance incidence a e (λ napcd ) was: log λnapcd  ¼ naðÞþgnpðÞþhncðÞþδd; whe e ndeno es coun y, adeno es age, pdeno es pe iod (calenda ime), cdeno es coho (da e o bi h) and ddeno es diabe es du a ion as e alua ed a he s a o each small in e al. In he simple model, donly ook he alues ‘gene al popula- ion’o ‘ ype 1 diabe ic pa ien s’. In he ex ended analysis, i ook he alues ‘gene al popula ion’o 0, 1, 2, 5, 10, 15 o 30 yea s, al hough he la e alues we e p esen in a sligh ly smalle da ase because he du a ion o ype 1 diabe es was no a ailable o p e alen cases a he s udy s a imes o Denma k and Aus alia. The a iable ‘du a ion o ype 1 dia- be es’was only inco po a ed in o si e-speci ic models in which he e we e a leas 200 cance cases among pe sons wi h ype 1 diabe es o each sex, wi h he addi ion o kidney cance cases because o i s es ablished associa ion wi h ype 2 diabe es. The model was hus a p opo ional haza ds model which assumed ha he HR o cance among pe sons wi h ype 1 diabe es ela i e o he gene al popula ion was cons an ac oss age and calenda ime ca ego ies. This is he same assump ion ha unde lies a adi ional s anda dised mo ali y a io analy- sis, bu we modelled he popula ion cance incidence a es using smoo h e ms ins ead o using empi ical a es in small in e als. This was done because a ai ly la ge ac ion o he ollow up occu ed in age g oups in which popula ion a es we e uns able, pa icula ly o a e cance ypes. Ou app oach s abilised popula ion a es by imposing he easonable as- sump ion ha a es o cance a ied smoo hly o e ime in bo h he o al popula ion and pe sons wi h ype 1 diabe es. We es ed whe he he speci ic age cu -o o 40 yea s was app op ia e by including an in e ac ion be ween age a diag- nosis (<30, 30–35, 35–40) and he HR o cance be ween hose wi h ype 1 diabe es and he gene al popula ion. We also i ed an ex ended model wi h sepa a e ype 1 diabe es HR o each coun y and es ed his agains he model wi h a common HR ac oss coun ies o assess he e ogenei y. All calcula ions and plo s we e ca ied ou in R, e sion 3.1.2 [17], and using he R Epi package [18]. A comple e accoun o all da a manipula ions and analyses is a ailable a h p://bendixca s ensen.com/DMCa/T1D/T1D-Ca.pd (accessed 8 Janua y 2016). Resul s Among pe sons wi h ype 1 diabe es, he e we e a o al o 9,149 i s inciden cance s in 3.9 million pe son-yea s o ol- low up. The dis ibu ion o cance by coun y, sex and si e a e p esen ed in Table 1(also included as ESM Table 1 o com- ple eness). His og ams o e en s and pe son-yea s by age, calenda ime and du a ion o each coun y (and he o al) a e p esen ed in ESM Fig. 1. Figu e 1shows he dis ibu ion o cance cases among pe - sons wi h ype 1 diabe es in he i e con ibu ing coun ies. In his s udy popula ion, he majo i y o cance cases occu ed be ween he ages o 40 and 60 yea s. The median age a cance diagnosis in his ela i ely young coho was 51.1 yea s (in e qua ile ange 43.5–59.5). Basic analyses We ound HRs (95% CIs) o o e all cance o 1.01 (0.98, 1.04) among men and 1.07 (1.04, 1.10) among women (Fig. 2and ESM Table 2) in compa ison wi h he gene al popula ion. When analyses we e es ic ed o non-sex- speci ic cance s (i.e. excluding p os a e, es is, b eas , ce ix, endome ium and o a y cance ), HRs (95% CIs) o 1.15 (1.11, 1.19) among men and 1.17 (1.13, 1.22) among women we e obse ed. The HRs associa ed wi h ype 1 diabe es o 23 cance subsi es a e shown in Fig. 2. Signi ican ly ele a ed HRs o bo h sexes we e obse ed o cance s o he li e , panc eas, kidney and s omach. Women wi h ype 1 diabe es had ele a - ed HRs o cance o he oesophagus, endome ium, o a y and hy oid. Among men wi h ype 1 diabe es, ele a ed HRs we e obse ed o colon cance and non-Hodgkin’slymphoma. Women wi h ype 1 diabe es exhibi ed signi ican ly educed HRs o melanoma, b eas cance and Hodgkin’slymphoma. Men wi h ype 1 diabe es had a 44% lowe incidence o p os- a e cance han ha obse ed in he gene al popula ion. In addi ion, we ound a bo de line signi ican 12% lowe isk o es is cance in men wi h ype 1 diabe es. The e was no signi ican in e ac ion be ween age a diag- nosis and he HR be ween ype 1 diabe ic pa ien s and he gene al popula ion o any o he i e coun ies. Signi ican he e ogenei y was obse ed be ween HR es i- ma es om he i e coun ies (p≤0.0001; see ESM Table 2 and ESM Figs 2–4). In pa icula , he e we e la ge di e ences by coun y in he HRs o cance a all si es in men, e.g. ype 1 diabe es was associa ed wi h an 18% inc eased cance inci- dence in Finland compa ed wi h a 10% educ ion in Sweden. 982 Diabe ologia (2016) 59:980–988 Howe e , when analyses we e pe o med o cance subsi es by coun y, he e we e no consis en di e ences. S a is ically signi ican he e ogenei y in he es ima ed HRs o cance by coun y was ound o indi idual cance si es, including cance o he panc eas, kidney, p os a e, b ain/cen al ne ous sys em and leukaemia in men and colo ec al and ce ical can- ce in women. Analysis by diabe es du a ion A o al o 7,792 (85.2% o 9,149) pa ien s wi h ype 1 diabe es who had a e i iable da e o diabe es diagnosis we e subse- quen ly diagnosed wi h cance . High HRs (95% CIs) o o e all cance occu ence we e obse ed du ing he i s yea ollowing he diagnosis o dia- be es (men, HR 2.28 [1.87, 2.78]; women, HR 2.34 [2.00, 2.74]; Fig. 3and ESM Table 3). Following he i s yea a e diabe es diagnosis, he HRs declined o uni y (1.03 [0.83, 1.29]) o women and dec eased o 1.23 [0.95, 1.60] o men. A e a ound 5 yea s, he HR o cance occu ence in men wi h ype 1 diabe es inc eased o 1.34 [1.20, 1.49]. A e 15 yea s, he cance incidence among men wi h ype 1 diabe- es was simila o ha o he gene al popula ion. The 95% CIs we e simila ega dless o disease du a ion because highe incidence a es in people wi h a longe du a ion o diabe es esul ed in o al numbe s o e en s simila o hose obse ed in he la ge popula ion wi h a sho e du a ion o diabe es. Fo mos si e-speci ic cance s, he HRs dec eased wi h in- c easing diabe es du a ion (ESM Table 4and ESM Figs 5–8). Howe e , he e was e y li le a ia ion by du a ion o diabe- es o b eas cance incidence, while he HR o endome ial cance emained ele a ed o app oxima ely 18 yea s. The Table 1 Pe son-yea s and numbe o cance cases in pe sons wi h diabe es diagnosed a age unde 40 yea s by sex, coun y and cance Cance si e Men Women To al AU DK FI SC SE Sub o al AU DK FI SC SE Sub o al Pe son-yea s (×1,000) a 255.5 255.6 547.9 178.7 737.5 1,975.1 255.4 289.0 636.8 145.5 631.8 1,957.7 4,064.0 All si es 504 401 1,000 253 1,882 4,040 600 641 1,408 280 2,180 5,109 9,149 Non-sex-speci ic 443 352 832 222 1,480 3,329 364 351 680 146 1,122 2,663 5,992 Pe son-yea s (×1,000) b 255.5 258.9 553.8 179.5 746.0 1,993.7 255.4 293.9 648.0 146.8 645.2 1,989.3 3,983.0 Oesophagus 9 4 16 9 29 67 3 1 5 7 14 30 97 S omach 12 6 47 5 64 134 13 14 50 4 39 120 254 Colon NA 26 56 18 173 273 NA 16 66 9 119 210 483 Rec um NA 13 46 15 84 158 NA 10 41 6 57 114 272 Colo ec al 61 39 102 33 257 492 60 26 107 15 176 384 876 Li e 14 9 34 12 44 113 7 6 8 3 17 41 154 Panc eas 19 15 54 7 52 147 8 9 41 5 30 93 240 Lung 41 39 119 37 134 370 29 38 58 19 128 272 642 Melanoma 86 21 60 16 122 305 72 59 59 15 103 308 613 B eas –– –– – –184 184 546 99 710 1,723 1,723 Ce ix –– –– – –11 48 36 14 85 194 194 Endome ium –– –– – –25 40 97 12 149 323 323 O a y –– –– – –22 25 80 11 114 252 252 P os a e 41 12 148 11 341 553 –– –– – –553 Tes is 22 37 23 20 57 159 –– –– – –159 Kidney 21 23 67 14 62 187 15 15 47 4 37 118 305 Bladde 12 17 49 11 124 213 2 7 16 6 35 66 279 B ain/cen al ne ous sys em 17 31 35 18 79 180 13 42 32 28 98 213 393 Thy oid 15 6 18 4 15 58 45 29 104 12 51 241 299 Non-Hodgkin’s lymphoma 30 26 56 21 111 244 19 14 46 8 56 143 387 Hodgkin’s lymphoma 11 11 14 NA 20 56 2 4 9 NA 9 24 80 Mul iple myeloma 2 4 14 3 26 53 2 5 8 0 13 28 81 Leukaemia NA 14 39 13 38 104 NA 16 33 9 28 86 190 a Follow up only o he i s p ima y umou o any kind o end o s udy b Follow up un il dea h o end o s udy; used o analysis o speci ic si es AU, Aus alia; DK, Denma k; FI, Finland, SC, Sco land; SE, Sweden; NA, da a no a ailable Diabe ologia (2016) 59:980–988 983 lowe incidence o p os a e cance among men wi h ype 1 diabe es became mo e appa en wi h inc easing du a ion o diabe es. Discussion In his la ge mul i-coun y s udy, we compa ed he cance incidence in pe sons wi h ype 1 diabe es wi h ha obse ed in he gene al popula ion, and in es iga ed he in luence o diabe es du a ion on he isk es ima es. We ound ha women wi h ype 1 diabe es had a ma ginally ele a ed incidence o o e all cance compa ed wi h he gene al popula ion. Men wi h ype 1 diabe es did no ha e an ele a ed incidence o o e all cance , al hough he subs an ial in e se associa ion be ween ype 1 diabe es and p os a e cance inci- dence in men con ibu ed o he o e all esul and o hese sex di e ences. When analyses we e es ic ed o non-sex-speci ic cance s, he excess cance incidence was simila in bo h men and women wi h ype 1 diabe es: he e we e ele a ed HR es i- ma es o abou 15% in he i s 20 yea s a e diagnosis. We ound ha he cance incidence was subs an ially highe in bo h men and women wi h ype 1 diabe es du ing he i s yea o ollow up compa ed wi h a longe du a ion o diabe es. The cance incidence subsequen ly dec eased o ha o he gene al popula ion a e app oxima ely 20 yea s o ollow up o men and a e 5 yea s o ollow up o women. We ound ha ype 1 diabe es con e ed an excess isk o cance o he s omach, li e , panc eas, endome ium and kid- ney and a educed isk o p os a e cance . We also epo a lowe incidence o b eas cance in women wi h ype 1 diabe- es compa ed wi h he gene al popula ion. The e was e idence o he e ogenei y in HRs o some can- ce s by coun y, hough a lack o plausible biological mecha- nisms leads us o belie e hese di e ences a e likely o be an a e ac . P e ious s udies in es iga ing he associa ion be ween ype 1 diabe es s a us and cance incidence ha e epo ed mixed indings [14], wi h es ima es o he o e all cance isk among pe sons wi h ype 1 diabe es anging be ween 5% lowe o 25% highe compa ed wi h he gene al popula ion, al hough epo ed 95% CIs include he es ima es epo ed he e [10–13, 19]. The disc epancies in hese indings may, in pa , be due o subs an ial di e ences in he c i e ia used o de ine ype 1 diabe es diagnoses. Fo example, a UK-based coho s udy which iden i ied 214 inciden cance s among 23,834 pe sons wi h diabe es diagnosed be ween 1972 and 1986 and epo ed a s anda dised incidence a io o 0.95 (95% CI 0.84, 1.08), used insulin ea men as a p oxy o ype 1 diabe es [12]. In con as , in he Swedish coho s udy ha iden i ied 258 can- ce diagnoses among 24,052 pe sons be ween 1964 and 2006 and epo ed an HR o 1.17 (95% CI 1.04, 1.33), ICD codes om hospi al admissions plus age we e used o asce ain ype 1 diabe es s a us [11]. The au ho s acknowledged he lack o speci ic ype 1 diabe es diagnos ic codes in ea lie ICD e - sions as a possible sou ce o misclassi ica ion bias because pe sons wi h ype 2 diabe es migh ha e been included in he ype 1 diabe es s udy popula ion. Ou inding o a signi ican ly ele a ed cance incidence among people wi h ype 1 diabe es in he i s yea a e diag- nosis o diabe es may be a ibu able o asce ainmen bias (i.e. ea lie de ec ion o p e-exis ing cance s) and o a lesse ex en o e e se causa ion (i.e. he cance i sel causing diabe es). P e ious s udies in es iga ing he ela ionship be ween ype 1 diabe es and cance ha e p o ided only spa se in o ma ion on he in luence o diabe es du a ion on his associa ion. These s udies ypically applied wide in e als o diabe es du a ion in he analyses because o hei limi ed sample sizes. As a con- sequence, hey we e unable o pe o m a de ailed analysis o he di e ences in cance occu ence acco ding o ime since diabe es diagnosis. In one Swedish coho s udy including da a om 29,187 pe sons who we e hospi alised wi h ype 1 diabe es be ween 1965 and 1999, HR es ima es we e s a i ied wi hin wo pos -diagnos ic pe iods o 1–14 o ≥15 yea s ol- lowing diabe es diagnosis [13]. These analyses yielded s anda dised incidence a ios o 1.1 (95% CI 1.0, 1.3) and 1970 1980 1990 2000 2010 0 20 40 60 80 Da e o cance dia g nosis Age a cance diagnosis Denma k 1,042 Finland 2,408 Sweden 4,062 Sco land 533 Aus alia 1,104 To al 9,149 984 Diabe ologia (2016) 59:980–988 Fig. 1 Dis ibu ion o cance s by coun y, age and da e o cance diagnosis 1.2 (95% CI 1.0, 1.3) o 1–14 and ≥15 yea s, espec i ely, which a e la gely consis en wi h ou esul s bu do no include a es in he i s yea a e diabe es diagnosis. HRs o si e-speci ic cance s in pe sons wi h ype 1 diabe- es we e simila o hose obse ed among pe sons wi h all ypes o diabe es and wi h ype 2 diabe es [1,20]. This inding sugges s a po en ial common mechanism among pe sons wi h ype 1 and ype 2 diabe es, o example obesi y, insulin ea - men o hype glycaemia. HR es ima es o si e-speci ic cance s we e highes o can- ce ypes in which obesi y is an es ablished isk ac o , namely s omach, colo ec al, kidney, endome ial and panc ea ic can- ce s [21]. While obesi y is less p e alen in pe sons wi h ype 1 diabe es han in pe sons wi h ype 2 diabe es, obesi y is inc easing among pe sons wi h ype 1 diabe es and may con- ibu e o an inc eased isk o ce ain cance s [22,23]. Fu he esea ch is he e o e equi ed o compa e obesi y p e alence in people wi h and wi hou ype 1 diabe es. The in luence o glucose-lowe ing medica ions on he ob- se ed associa ion be ween diabe es and cance has been ex- ensi ely in es iga ed [24,25]. Ou inding o a smalle excess incidence o cance among pe sons wi h ype 1 diabe es han was p e iously obse ed in pe sons wi h ype 2 diabe es does no suppo he no ion ha insulin he apies con ibu e o he obse ed ele a ed incidence. I exogenous insulin did con ib- u e o he obse ed associa ion, he s eng h o he ela ionship be ween ype 1 diabe es and cance incidence would be ex- pec ed o be s onge han ha obse ed among pe sons wi h HR o cance , T1D s p o p ula ion Leukaemia Mul iple myeloma Hodgkin’s lymphoma Non-Hodgkin’s lymphoma Thy oid B ain, CNS Bladde Kidney Tes is P os a e O a y Endome ium Ce ix u e i B eas Melanoma o skin Lung Panc eas Li e Colo ec al Rec um Colon S omach Oesophagus Non-sex-speci ic All si es 0.5 0.6 0.8 1.0 1.2 1.5 2.0 2.5 3.0 1.07 (1.04,1.10) 1.17 (1.13,1.22) 1.79 (1.25,2.56) 1.78 (1.49,2.13) 1.06 (0.93,1.22) 0.97 (0.81,1.17) 1.09 (0.99,1.21) 1.55 (1.14,2.10) 1.25 (1.02,1.53) 1.07 (0.95,1.21) 0.81 (0.73,0.90) 0.90 (0.85,0.94) 0.92 (0.80,1.06) 1.42 (1.27,1.58) 1.15 (1.02,1.30) 1.47 (1.23,1.77) 1.01 (0.79,1.28) 0.97 (0.85,1.11) 1.51 (1.34,1.72) 1.04 (0.88,1.21) 0.61 (0.41,0.91) 0.77 (0.53,1.12) 1.10 (0.89,1.36) 1.01 (0.98,1.04) 1.15 (1.11,1.19) 1.08 (0.85,1.37) 1.23 (1.04,1.46) 1.25 (1.11,1.41) 0.96 (0.82,1.12) 1.14 (1.04,1.24) 2.00 (1.67,2.40) 1.53 (1.30,1.79) 1.06 (0.96,1.17) 0.94 (0.84,1.05) 0.56 (0.51,0.61) 0.88 (0.75,1.02) 1.30 (1.12,1.49) 0.97 (0.85,1.11) 0.92 (0.80,1.06) 1.25 (0.97,1.61) 1.24 (1.09,1.40) 1.11 (0.85,1.44) 0.99 (0.76,1.30) 0.92 (0.76,1.12) HR Fig. 2 HRs (wi h 95% CIs) o si e-speci ic cance s associa ed wi h ype 1 diabe es in men (blue) and women ( ed) ob ained om he analysis o combined coun y da a. HRs a e p esen ed on a loga i hmic scale. CNS, cen al ne ous sys em; T1D, ype 1 diabe es Diabe ologia (2016) 59:980–988 985 ype 2 diabe es because all people wi h ype 1 diabe es a e ea ed wi h insulin and a mino i y o people wi h ype 2 dia- be es ecei e insulin ea men . Fu he mo e, he absence o an associa ion be ween o e all o si e-speci ic cance isk and inc easing du a ion o diabe es in ou s udy does no suppo adose– esponse ela ionship be ween exogenous insulin use and cance incidence. We did no ha e su icien ly de ailed da a o e alua e associa ions be ween ype 1 diabe es and can- ce incidence acco ding o speci ic ypes o insulin. Diabe es-speci ic me abolic de iciencies such as hype - glycaemia may p o ide an al e na i e explana ion o he obse ed excess isk o some cance s among pe sons wi h diabe es, gi en ha hese de iciencies a e common in bo h ype 1 and ype 2 diabe es. Hype glycaemia may be a plausi- ble explana ion gi en he iden i ica ion o a dose– esponse ela ionship be ween glyca ed haemoglobin le els and he isk o ce ain cance s [8,26,27]. Howe e , u he wo k is equi ed o elucida e whe he obesi y o hype glycaemia in pe sons wi h ype 1 diabe es is esponsible o he obse ed associa ion wi h some cance ypes. We epo a 10% lowe incidence o b eas cance in women wi h ype 1 diabe es compa ed wi h he gene al pop- ula ion. P e ious s udies assessing he in luence o ype 2 o unspeci ied diabe es on b eas cance ha e displayed a simila o ele a ed isk compa ed wi h he gene al popula ion, al- hough hese indings a e la gely limi ed o pos -menopausal women [28]. Ou con adic o y inding may he e o e e lec ou younge coho (con aining ewe pos -menopausal wom- en) o may be con ounded by o he b eas cance isk ac o s such as pa i y. Ou inding o an inc eased incidence o panc ea ic and li e cance among pe sons wi h ype 1 diabe es is consis en wi h some ea lie s udies. Resul s om a me a-analysis based on nine s udies (bu wi h only a o al o 39 cases o panc ea ic cance ) indica ed ha pe sons wi h ype 1 diabe es we e a a wo old inc eased isk o panc ea ic cance compa ed wi h pe sons wi hou diabe es [29]. A Danish s udy (dis inc om he Danish da a in his s udy) obse ed an HR o 4.8 (95% CI 2.8, 7.7) o li e cance among diabe es pa ien s younge han 50 yea s a diagnosis. Al hough his de ini ion o ype 1 dia- be es is no s ic ly compa able wi h ou s, he esul is consis- en wi h ou indings [19]. Howe e , he associa ions seen o panc ea ic and li e cance may be a consequence o e e se causa ion. We ound an inc eased isk o hy oid cance in pe sons wi h ype 1 diabe es compa ed wi h he gene al popula ion, pa icula ly among women. The wo ldwide incidence a es o hy oid cance ha e inc eased s eeply in ecen decades [30, 31], pa ly due o enhanced de ec ion h ough he use o mo e sensi i e diagnos ic equipmen . Consequen ly, ou obse a- ion o an inc eased incidence o hy oid cance among pe - sons wi h ype 1 diabe es may be a diagnos ic a e ac because pe sons wi h ype 1 diabe es ecei e conside ably highe le els o medical a en ion han hose wi hou diabe es. Al hough a educed isk o p os a e cance in pe sons wi h ype 2 diabe es has been epo ed in nume ous s udies [2–5], only one p e ious s udy has iden i ied such an associa ion among pe sons wi h ype 1 diabe es (al hough some o he la e s udy da a we e included in he p esen s udy) [10]. The simila i y in p os a e cance incidence be ween ype 1 and ype 2 diabe ic pa ien s sugges s a simila unde lying mechanism in bo h ypes o diabe es. One p oposed mecha- nism in ol es he lowe es os e one le els p esen in men wi h diabe es, obesi y o bo h [32]. Highe es os e one le els 1.0 1.5 2.0 2.5 3.0 0 5 10 15 20 25 30 350 5 10 15 20 25 30 35 Time since dia g nosis o diabe es (yea s) HR o cance ; T1D s popula ion ab Fig. 3 HR (95% CI) o all cance s by du a ion o diabe es in men (a) and women (b) 986 Diabe ologia (2016) 59:980–988 we e p e iously shown o lead o an inc eased isk o p os a e cance [33], while hype glycaemia has also been shown o inhibi es os e one p oduc ion [34]. S eng hs and limi a ions This s udy used i e na ional coho s o pe sons wi h ype 1 diabe es, and is he e o e he la ges s udy o da e o in es i- ga e he ela ionship be ween ype 1 diabe es and cance inci- dence. By combining cases ac oss coun ies we assembled a su icien ly la ge sample size o in es iga e associa ions wi h si e-speci ic cance s. The la ges p e ious s udy included less han one- hi d o he numbe o cance s in people wi h ype 1 diabe es compa ed wi h ou s udy [14]. The popula ion-based egis y app oach also minimised selec ion bias. A u he s eng h o ou s udy was he inclusion o in o - ma ion on du a ion o diabe es, which enabled us o assess he in luence o asce ainmen bias and he po en ial impac o e e se causa ion. Fo example, i is well es ablished ha pan- c ea ic cance can induce diabe es [35]. Possible misclassi ica ion o ype 2 as ype 1 diabe es may ha e led o an o e es ima ion o he ela ionship be ween ype 1 diabe es and cance because o he highe le els o well- es ablished isk ac o s o cance (including obesi y and smoking) among pe sons wi h ype 2 diabe es. Howe e , he p e alence o ype 2 diabe es below 40 yea s o age is low and we he e o e belie e ha he in luence o misclassi ica ion bias is negligible, as shown by ou in e ac ion analysis (see h p://bendixca s ensen.com/DMCa/T1D/T1D-Ca.pd (accessed 8 Janua y 2016), sec ion 8.5). Mo eo e , we compa ed cance incidence a es among pe - sons wi h ype 1 diabe es wi h hose o he gene al popula ion, ins ead o only wi h hose o pe sons wi h diabe es. This may ha e dilu ed he impac o ype 1 diabe es on HR es ima es, al hough his e ec will be small because o he low p e a- lence o ype 1 diabe es (1.5–4.8 pe 1,000 people) in he s udy popula ions [36]. In addi ion, he la ge numbe o compa isons pe o med in oduced he po en ial o chance indings. Finally, he s udy popula ion was ela i ely young. Thus, we we e no able o meaning ully e alua e po en ial changes in cance incidence among olde pe sons wi h ype 1 diabe es. Implica ions Ou indings do no suppo changing he policy o cance sc eening in pe sons wi h ype 1 diabe es. Simila ecommen- da ions o li es yle app oaches o educe cance isk such as weigh managemen , inc easing physical ac i i y and a oiding smoking apply o pe sons wi h ype 1 diabe es as o he gen- e al popula ion. Fu u e wo k should be di ec ed a asce aining whe he he inc eased incidence o some cance s among pe sons wi h ype 1 diabe es leads o a aised isk o cance mo ali y among pe sons wi h ype 1 diabe es. Conclusions We ound ha , on a e age, ype 1 diabe es con e s an excess incidence o se e al cance s. In pa icula , pe sons wi h ype 1 diabe es had a highe incidence o cance o he li e , pan- c eas, kidney, endome ium and o a y and a lowe incidence o p os a e cance han hose in he gene al popula ion. Howe e , simila o he indings o ype 2 diabe es, he HRs o cance we e highes a ime o diabe es diagnosis and de- clined o e ime. Acknowledgemen s Some o he da a om his s udy we e p esen ed as a pos e a he 74 h Scien i ic Sessions o he ADA in 2014 and as an o al p esen a ion a he 2014 EASD Annual Mee ing. Funding This s udy was unded by he Eu opean Founda ion o he S udy o Diabe es. Duali y o in e es BC is a s ockholde o No o No disk and an em- ployee o he S eno Diabe es Cen e , a diabe es clinic and esea ch ins i- u ion owned by No o No disk. SHW ecei ed an hono a ium om Global MedEd/As a Zeneca in Sep embe 2014 o con ibu ing a lec u e o a se ies o educa ional ideos aimed a p ima y ca e p o essionals and specialis s in he Middle Eas . All o he au ho s decla e ha he e is no duali y o in e es associa ed wi h hei con ibu ion o his manusc ip . Con ibu ion s a emen The s udy was concei ed by SG, SHW, IK and BC; da a was p epa ed by BC, RL, RS, JJK, A-MS, SF, DJM and JLH; s a is ical analyses we e ca ied ou by BC; SHR w o e he i s d a o he pape and coo dina ed he w i ing p ocess. All au ho s con ibu ed o he in e p e a ion o he indings and he pape ’s c i ical e ision. All au ho s ha e app o ed he inal e sion o he manusc ip . BC is he gua an o o his wo k. Open Access This a icle is dis ibu ed unde he e ms o he C ea i e Commons A ibu ion 4.0 In e na ional License (h p:// c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons license, and indica e i changes we e made. Re e ences 1. Vigne i P, F asca F, Sciacca L, Pandini G, Vigne i R (2009) Diabe es and cance . Endoc Rela Cance 16:1103–1123 2. Bono as S, Filioussi K, Tsan es A (2004) Diabe es melli us and isk o p os a e cance : a me a-analysis. Diabe ologia 47:1071–1078 3. Weide pass E, Ye W, Vainio H, Kaaks R, Adami HO (2002) Reduced isk o p os a e cance among pa ien s wi h diabe es melli us. In J Cance 102:258–261 4. Fall K, Ga mo H, Gudbjo nsdo i S, S a in P, Ze helius B (2013) Diabe es melli us and p os a e cance isk; a na ionwide case- con ol s udy wi hin PCBaSe Sweden. Cance Epidemiol Bioma k P e : Publ Am Assoc Cance Res, cosponso ed by he Am Soc P e Oncol 22:1102–1109 Diabe ologia (2016) 59:980–988 987 5. Pie ce BL (2012) Why a e diabe ics a educed isk o p os a e cance ? A e iew o he epidemiologic e idence. U ol Oncol 30:735–743 6. Smi h U, Gale EAM (2009) Does diabe es he apy in luence he isk o cance ? Diabe ologia 52:1699–1708 7. Ca s ensen B, Wi e DR, F iis S (2012) Cance occu ence in Danish diabe ic pa ien s: du a ion and insulin e ec s. Diabe ologia 55:948–958 8. Yang X, Ko GTC, So WY e al (2010) Associa ions o hype glyce- mia and insulin usage wi h he isk o cance in ype 2 diabe es: he Hong Kong Diabe es Regis y. Diabe es 59:1254–1260 9. Wild SH (2011) Diabe es, ea men s o diabe es and hei e ec on cance incidence and mo ali y: a emp s o disen angle he web o associa ions. Diabe ologia 54:1589–1592 10. Ha ding JL, Shaw JE, Pee e s A, Ca ensen B, Magliano DJ (2015) Cance isk among people wi h ype 1 and ype 2 diabe es: disen angling ue associa ions, de ec ion bias, and e e se causa- ion. Diabe es Ca e 38:264–270 11. Shu X, Ji J, Li X, Sundquis J, Sundquis K, Hemminki K (2010) Cance isk among pa ien s hospi alized o ype 1 diabe es melli us: a popula ion-based coho s udy in Sweden. Diabe Med: J B Diabe Assoc 27:791–797 12. Swe dlow AJ, Laing SP, Qiao Z e al (2005) Cance incidence and mo ali y in pa ien s wi h insulin- ea ed diabe es: a UK coho s udy. B J Cance 92:2070–2075 13. Zendehdel K, Ny en O, Os enson CG, Adami HO, Ekbom A, Ye W (2003) Cance incidence in pa ien s wi h ype 1 diabe es melli us: a popula ion-based coho s udy in Sweden. J Na l Cance Ins 95:1797–1800 14. Go don-Dseagu VL, Shel on N, Mindell JS (2013) Epidemiological e idence o a ela ionship be ween ype-1 diabe es melli us and can- ce : a e iew o he exis ing li e a u e. In J Cance 132:501–508 15. Ca s ensen B (2007) Age-pe iod-coho models o he Lexis dia- g am. S a Med 26:3018–3045 16. Ha ell JFE (2006) Reg ession modeling s a egies. Sp inge , New Yo k 17. R Co e Team (2014) R: a language and en i onmen o s a is ical compu ing, Vienna, Aus ia 18. Ca s ensen B, Plumme M, Laa a E, Hills M (2014) Epi: a package o s a is ical analysis in epidemiology. R package e sion 1.1.67. A ailable om h p://CRAN.R-p ojec .o g/package=Epi, accessed 16 Decembe 2015 19. Wide o L, G idley G, Mellemkjae L e al (1997) Cance inci- dence in a popula ion-based coho o pa ien s hospi alized wi h diabe es melli us in Denma k. J Na l Cance Ins 89:1360–1365 20. Tsilidis KK, Kasimis JC, Lopez DS, N zani EE, Ioannidis JPA (2015) Type 2 diabe es and cance : umb ella e iew o me a- analyses o obse a ional s udies. BMJ 350:g7607 21. Renehan AG, Tyson M, Egge M, Helle RF, Zwahlen M (2008) Body-mass index and incidence o cance : a sys ema ic e iew and me a-analysis o p ospec i e obse a ional s udies. Lance 371:569–578 22. Sco ish Diabe es Su ey Moni o ing G oup (2014) Sco ish Diabe es Su ey 2014. NHS Sco land. A ailable om www. diabe esinsco land.o g.uk/Publica ions.aspx?ca Id=3, accessed 19 Janua y 2016 23. Swedish Na ional Diabe es Regis e (2013) Annual epo . Cen e o Regis e s, Region Väs a Gö aland. In: www.nd .nu/pd s/ Annual_Repo _NDR_2013.pd , accessed 19 Janua y 2016 24. Ka ls ad O, S a up-Linde J, Ves e gaa d P e al (2013) Use o insulin and insulin analogs and isk o cance —sys ema ic e iew and me a-analysis o obse a ional s udies. Cu D ug Sa 8:333–348 25. Tang X, Yang L, He Z, Liu J (2012) Insulin gla gine and cance isk in pa ien s wi h diabe es: a me a-analysis. PLoS One 7:e51814 26. de Bee JC, Liebenbe g L (2014) Does cance isk inc ease wi h HbA 1c , independen o diabe es? B J Cance 110:2361–2368 27. Liao W-C, Tu Y-K, Wu M-S, Lin J-T, Wang H-P, Chien K-L (2015) Blood glucose concen a ion and isk o panc ea ic cance : sys em- a ic e iew and dose- esponse me a-analysis. BMJ:g7371 28. La sson SC, Man zo os CS, Wolk A (2007) Diabe es melli us and isk o b eas cance : a me a-analysis. In J Cance 121:856–862 29. S e ens RJ, Roddam AW, Be al V (2007) Panc ea ic cance in ype 1 and young-onse diabe es: sys ema ic e iew and me a-analysis. B J Cance 96:507–509 30. Pelleg i i G, F asca F, Regalbu o C, Squa i o S, Vigne i R (2013) Wo ldwide inc easing incidence o hy oid cance : upda e on epi- demiology and isk ac o s. J Cance Epidemiol 2013:965212 31. Vigne i R, Maland ino P, Vigne i P (2015) The changing epidemi- ology o hy oid cance : why is incidence inc easing? Cu Opin Oncol 27:1–7 32. Tsilidis KK, Allen NE, Appleby PN e al (2015) Diabe es melli us and isk o p os a e cance in he Eu opean P ospec i e In es iga ion in o Cance and Nu i ion. In J Cance 136:372–381 33. Gann PH, Hennekens CH, Ma J, Longcope C, S amp e MJ (1996) P ospec i e s udy o sex ho mone le els and isk o p os a e cance . JNa lCance Ins 88:1118–1126 34. Ba e -Conno E, Khaw KT, Yen SS (1990) Endogenous sex ho - mone le els in olde adul men wi h diabe es melli us. Am J Epidemiol 132:895–901 35. Huxley R, Ansa y-Moghaddam A, Be ing on de Gonzalez A, Ba zi F, Woodwa d M (2005) Type-II diabe es and panc ea ic can- ce : a me a-analysis o 36 s udies. B J Cance 92:2076–2083 36. F ese T, Sandholze H (2013) The epidemiology o ype 1 diabe es melli us. In: Esche AP, Li A (eds) Type 1 diabe es. InTech, Rijeka, C oa ia 988 Diabe ologia (2016) 59:980–988