Caffeine consumption and mortality in diabetes: An analysis of NHANES 1999-2010

ORIGINAL RESEARCH published: 20 Sep embe 2018 doi: 10.3389/ endo.2018.00547 F on ie s in Endoc inology | www. on ie sin.o g 1Sep embe 2018 | Volume 9 | A icle 547 Edi ed by: Cha uma hi Sabanayagam, Singapo e Eye Resea ch Ins i u e, Singapo e Re iewed by: Ka i a Venka a aman, Na ional Uni e si y o Singapo e, Singapo e Te su o Tsujimo o, Na ional Cen e Fo Global Heal h and Medicine, Japan *Co espondence: João Sé gio Ne es [email p o ec ed] Special y sec ion: This a icle was submi ed o Diabe es, a sec ion o he jou nal F on ie s in Endoc inology Recei ed: 10 June 2018 Accep ed: 29 Augus 2018 Published: 20 Sep embe 2018 Ci a ion: Ne es JS, Lei ão L, Mag iço R, Bigo e Viei a M, Viegas Dias C, Oli ei a A, Ca alho D and Clagge B (2018) Ca eine Consump ion and Mo ali y in Diabe es: An Analysis o NHANES 1999–2010. F on . Endoc inol. 9:547. doi: 10.3389/ endo.2018.00547 Ca eine Consump ion and Mo ali y in Diabe es: An Analysis o NHANES 1999–2010 João Sé gio Ne es1,2*, Lia Lei ão3, Ri a Mag iço4, Miguel Bigo e Viei a5, Ca a ina Viegas Dias6, Ana Oli ei a1, Da ide Ca alho 1,7 and B ian Clagge 8 1Depa men o Endoc inology, Diabe es and Me abolism, São João Hospi al Cen e , Po o, Po ugal, 2Depa men o Su ge y and Physiology, Facul y o Medicine, Ca dio ascula Resea ch Cen e , Uni e si y o Po o, Po o, Po ugal, 3Neu ology Depa men , Hospi al P o . Dou o Fe nando Fonseca, Amado a, Po ugal, 4Neph ology Depa men , Hospi al Cu y Cab al, Lisbon, Po ugal, 5Neph ology and Renal T ansplan a ion Depa men , Cen o Hospi ala Lisboa No e, Lisbon, Po ugal, 6NOVA Medical School, Lisbon, Po ugal, 7Facul y o Medicine, Ins i u o de In es igação e Ino ação em Saúde, Uni e si y o Po o, Po o, Po ugal, 8Ca dio ascula Di ision, B igham and Women’s Hospi al, Ha a d Medical School, Bos on, MA, Uni ed S a es Aim: An in e se ela ionship be ween co ee consump ion and mo ali y has been epo ed in he gene al popula ion. Howe e , he e ec o co ee consump ion in diabe es emains unclea . We aimed o e alua e he associa ion o ca eine consump ion and ca eine sou ce wi h mo ali y among pa ien s wi h diabe es. Me hods: We examined he associa ion o ca eine consump ion wi h mo ali y among 1974 women and 1974 men wi h diabe es, using he Na ional Heal h and Nu i ion Examina ion Su ey (NHANES) 1999–2010. Ca eine consump ion was assessed a baseline using 24 h die a y ecalls. Cox p opo ional haza d models we e i ed o es ima e haza d a ios (HR) o all-cause, ca dio ascula , and cance - ela ed mo ali y acco ding o ca eine consump ion and i s sou ce, adjus ing o po en ial con ounde s. Resul s: A dose-dependen in e se associa ion be ween ca eine and all-cause mo ali y was obse ed in women wi h diabe es. Adjus ed HR o dea h among women who consumed ca eine, as compa ed wi h non-consume s, we e: 0.57 (95% CI, 0.40–0.82) o <100 mg o ca eine/day, 0.50 (95% CI, 0.32–0.78) o 100 o <200 mg o ca eine/day, and 0.39 (95% CI, 0.23–0.64) o ≥200 mg o ca eine/day (p=0.005 o end). This associa ion was no obse ed in men. The e was a signi ican in e ac ion be ween sex and ca eine consump ion (p=0.015). No signi ican associa ion be ween o al ca eine consump ion and ca dio ascula o cance mo ali y was obse ed. Women who consumed mo e ca eine om co ee had educed isk o all-cause mo ali y (p=0.004 o end). Conclusion: Ou s udy showed a dose-dependen p o ec i e e ec o ca eine consump ion on mo ali y among women wi h diabe es. Keywo ds: ca eine, co ee, mo ali y, diabe es, na ional heal h and nu i ion examina ion su ey Ne es e al. Ca eine and Mo ali y in Diabe es INTRODUCTION Diabe es is a majo public heal h p oblem wi h an inc easing p e alence wo ldwide (1). Gi en i s signi ican bu den, i is impo an o iden i y li es yle ac o s o imp o emen o p ognosis. Ca eine is p o ided in di e en sou ces, mainly co ee, ea, and so d inks. Co ee con ains mo e ca eine han he majo i y o oods and cons i u es one o he mos commonly consumed be e ages wo ldwide. The de ec ion o a heal h- ela ed e ec associa ed wi h co ee consump ion may ha e a po en ial g ea impac in public heal h (2). Co ee has been e e ed as con aining se e al bioac i e compounds including an ioxidan s wi h po en ially bene icial p ope ies. An in e se associa ion be ween co ee consump ion and se um bioma ke s o in lamma ion and insulin esis ance has been desc ibed (3–5). A me a-analysis o p ospec i e s udies (6) and a sys ema ic e iew (7) showed ha co ee consump ion migh be associa ed wi h educ ion in he incidence o ype 2 diabe es. A ecen dose- esponse sys ema ic e iew concluded ha co ee consump ion was s ongly associa ed wi h a isk educ ion in all-cause and ca dio ascula disease (CVD) mo ali y. Howe e , his sys ema ic e iew excluded s udies which analyzed speci ic subpopula ions, such as hose including only people wi h diabe es (8). Al hough ca eine consump ion appea s o be associa ed wi h a dec eased isk o de eloping ype 2 diabe es, i is unclea i i s p o ec i e e ec pe sis s in people wi h es ablished diabe es. In a p ospec i e s udy including 4,365 pa ien s wi h a p io myoca dial in a c ion, d inking co ee was associa ed wi h lowe isk o ca dio ascula mo ali y and ischemic hea disease mo ali y (9). In his s udy, signi ican in e se associa ions we e epo ed in pa ien s wi hou diabe es, whe eas he associa ions we e weak and non-signi ican in he smalle g oup wi h diabe es. A e e iewing he li e a u e, we ound only h ee s udies speci ic o pa ien s wi h diabe es. One p ospec i e coho o 7,170 emale egis e ed nu ses wi h diabe es ound ha habi ual co ee consump ion was no associa ed wi h inc eased isk o ca dio ascula diseases o p ema u e mo ali y (10). A simila coho o 3,497 male heal h p o essionals wi h diabe es ound no associa ion be ween ca eine consump ion and CVD o all-cause mo ali y (11). Ano he s udy, es ic ed o a Finnish popula ion (3,837 pa ien s), ound ha in pa ien s wi h ype 2 diabe es, co ee d inking was associa ed wi h educed all-cause, CVD, and co ona y hea disease mo ali y (12). Howe e , in his s udy, no adjus men s we e made o diabe es du a ion, complica ions o diabe es, and insulin ea men . Conside ing ha he e is limi ed and con lic ing e idence ega ding he ela ionship be ween ca eine consump ion and mo ali y in people wi h diabe es, we examined he con inuous Na ional Heal h and Nu i ion Examina ion Su ey (NHANES) 1999–2010 da abase o e alua e he e ec o ca eine consump ion and ca eine sou ce on all-cause, ca dio ascula , and cance mo ali y among pa ien s wi h diabe es. MATERIALS AND METHODS S udy Design and Pa icipan s We pe o med an analysis o he con inuous NHANES da abase. The NHANES is a pe iodic su ey conduc ed by he Na ional Cen e o Heal h S a is ics (NCHS) o he Cen e s o Disease Con ol and P e en ion (CDC). NHANES is a s a i ied, mul i- s age su ey using a na ionally ep esen a i e sample o he non-ins i u ionalized ci ilian popula ion o he Uni ed S a es. Pa icipan s a e selec ed a andom h ough a complex s a is ical p ocess each yea , and hey comple e pe sonal s uc u ed in e iews a home and hen pe o m a physical examina ion a a mobile examina ion cen e ha includes heigh , weigh , and labo a o y measu emen s (13). We used da a om 1999 o 2010, ha includes 62,160 people. We es ic ed ou analysis o indi iduals wi h ≥18-yea -old (35,379 subjec s) and wi h diabe es (4,544 subjec s). Diabe es was de ined by a sel - epo ed p e ious diagnosis, a hemoglobin A1c le el o ≥6.5%, o a as ing plasma glucose le el o ≥126 mg/dL. Bo h pa ien s wi h ype 1 and ype 2 diabe es we e included. We excluded 596 subjec s due o implausible alimen a y epo s (as de ined in p e ious s udies: consump ion o <500 kcal/day o >3500 kcal/day) (14) o missing in o ma ion on ca eine consump ion and/o mo ali y. Finally, 3,948 subjec s we e included in ou p esen analysis. The NCHS Resea ch E hics Re iew Boa d e iewed and app o ed NHANES, and all pa icipan s p o ided w i en in o med consen . This s udy was egis e ed a www.clinical ials. go as NCT03367806. Assessmen o Exposu e In all cycles o NHANES 1999–2010, a 24-h die a y ecall was collec ed. Using an au oma ed mul iple-pass me hod, all ood i ems and quan i ies consumed in he 24 h p eceding he in e iew we e eco ded. Fo pa icipan s in he 1999– 2002 NHANES, only one in-pe son 24-h die a y ecall was adminis e ed. The cycles s a ing om 2003 onwa d included wo ecalls, he i s one in-pe son and he second one ia elephone collec ed 3 o 10 days ollowing he i s die a y in e iew bu no on he same day o he week. To calcula e he ca eine, ene gy, and nu ien in akes, o pa icipan s in he 1999–2002 NHANES, we used he nu i ional in o ma ion om oods and be e ages collec ed in he single 24-h die a y ecall. Fo pa icipan s in he 2003-2010 NHANES, he mean o he nu i ional in o ma ion om bo h ecalls was used (15). NHANES includes in o ma ion ega ding nu ien sou ce by ype o ood inges ed. This da a was used o asce ain he quan i y o inges ed ca eine o igina ing in co ee, ea, o so d ink o each pa ien . The impac o ca eine consump ion ob ained om each o he h ee ypes o d ink on di e en ou comes was e alua ed. Conside ing he mean ca eine con en pe uni o ca eina ed be e age (95 mg in 8 oz. o co ee, 48 mg in 8 oz. o ea, and 30 mg in 12 oz. o cola) (16), we di ided he daily in ake o ca eine om all sou ces and om co ee in o h ee ca ego ies (<100 mg, 100 o <200 mg, and 200 mg o mo e). Gi en he low numbe o pa ien s wi h ca eine in ake om ea and om so d inks, and he high a iabili y o ca eine in hese be e ages, hose pa ien s F on ie s in Endoc inology | www. on ie sin.o g 2Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es we e di ided in o e iles o consump ion and a e p esen ed as Supplemen a y Ma e ial. Con inuous a iables a e p esen ed as means wi h s anda d de ia ions excep o polyunsa u a ed o sa u a ed a y acids a io, ibe s pe day, and yea s o diabe es which we e summa ized using median (in e qua ile ange) due o hei igh -skewed dis ibu ions. Ca ego ical a iables a e p esen ed as pe cen wi h 95% con idence in e als. Ou comes The p ima y ou come was ime o dea h. As seconda y ou comes, we selec ed ime o ca dio ascula dea h and ime o dea h by TABLE 1 | Baseline cha ac e is ics o he s udy popula ion. Baseline cha ac e is ics acco ding o ca eine consump ion among women (n=1974) No consump ion (n=219) <100 mg/day (n=979) P- alue* 100 o <200 mg/day (n=438) P- alue* ≥200 mg/day (n=338) P- alue* P o end Age, yea s 58.3 (15.8) 61.6 (15.0) 0.022 59.5 (14.3) 0.469 58.6 (13.2) 0.829 0.041 Non-Hispanic Whi e, % 34.7% 53.0% 0.001 59.2% <0.001 79.2% <0.001 <0.001 Annual amily income <$25000, % 51.4% 48.8% 0.399 39.1% 0.018 38.6% 0.010 0.004 Educa ion le el -Less han 9 h g ade, % 22.0% 15.9% 0.119 10.1% 0.002 9.1% <0.001 <0.001 Cu en smoke s, % 6.3% 8.4% 0.312 12.7% 0.029 25.9% <0.001 <0.001 Fo me smoke s, % 24.0% 26.5% 0.548 34.3% 0.053 33.2% 0.079 0.023 Alcohol consump ion >20 g ams/day, % 2.3% 2.9% 0.651 5.2% 0.172 4.9% 0.219 0.126 Ca bohyd a es pe day, g am/100 kcal 12.9 (3.0) 12.7 (2.3) 0.448 12.1 (2.3) 0.020 11.8 (2.5) 0.001 <0.001 Polyunsa u a ed/sa u a ed a y acids a io 0.73 (0.52–1.11) 0.70 (0.50–0.94) 0.061 0.67 (0.49–0.92) 0.056 0.65 (0.49–0.84) 0.001 0.001 Fibe pe day, g am/100 kcal 0.95 (0.54–1.27) 0.91 (0.65–1.16) 0.957 0.85 (0.64–1.09) 0.107 0.78 (0.59–1.03) 0.045 0.002 Low physical ac i i y le el, % 44.9% 37.2% 0.109 37.4% 0.201 41.9% 0.607 0.469 BMI, kg/m2 34.7 (8.9) 33.2 (8.3) 0.083 33.0 (6.9) 0.034 33.8 (7.8) 0.342 0.775 Hype ension, % 67.3% 67.2% 0.978 72.5% 0.266 67.0% 0.957 0.889 Dyslipidemia, % 53.1% 60.3% 0.188 61.3% 0.128 58.7% 0.283 0.926 Time since diagnosis o diabe es, yea s 3 (0–13) 5 (0–12) 0.100 5 (1–13) 0.415 7 (1–15) 0.092 0.592 Diabe ic e inopa hy, % 24.0% 21.7% 0.591 22.0% 0.722 25.4% 0.661 0.219 Diabe ic kidney disease, % 23.2% 26.4% 0.511 27.8% 0.477 16.8% 0.100 0.005 Mac o ascula complica ions, % 19.3% 18.1% 0.696 19.2% 0.988 19.5% 0.908 0.629 Insulin ea men , % 21.5% 18.6% 0.465 21.0% 0.876 27.6% 0.274 0.021 Baseline cha ac e is ics acco ding o ca eine consump ion among men (n=1974) No consump ion (n=186) <100 mg/day (n=739) P- alue* 100 o <200 mg/day (n=470) P- alue* ≥200 mg/day (n=579) P- alue* P o end Age, yea s 57.3 (14.6) 60.0 (14.0) 0.080 58.0 (14.0) 0.660 58.1 (12.4) 0.601 0.245 Non-Hispanic Whi e, % 41.4% 54.3% 0.013 70.4% <0.001 80.3% <0.001 <0.001 Annual amily income <$25000, % 35.2% 34.7% 0.871 27.6% 0.110 27.9% 0.123 0.029 Educa ion le el - Less han 9 h g ade, % 18.2% 16.3% 0.616 12.6% 0.126 10.2% 0.020 0.001 Cu en smoke s, % 16.0% 11.8% 0.300 14.8% 0.826 25.7% 0.097 <0.001 Fo me smoke s, % 39.7% 45.4% 0.247 46.7% 0.199 45.1% 0.366 0.834 Alcohol consump ion >20 g ams/day, % 15.9% 9.8% 0.073 11.4% 0.302 11.6% 0.169 0.867 Ca bohyd a es pe day, g am/100 kcal 11.9 (3.1) 11.8 (2.5) 0.958 11.7 (2.4) 0.573 11.3 (2.6) 0.124 0.006 Polyunsa u a ed/sa u a ed a y acids a io 0.70 (0.47–0.98) 0.68 (0.48–0.91) 0.306 0.70 (0.50–0.91) 0.249 0.64 (0.47–0.88) 0.133 0.216 Fibe pe day, g am/100 kcal 0.83 (0.55–1.10) 0.83 (0.61–1.14) 0.624 0.76 (0.55–1.09) 0.444 0.78 (0.57–0.99) 0.134 0.010 Low physical ac i i y le el, % 40.9% 30.0% 0.030 32.8% 0.157 33.4% 0.120 0.773 BMI, kg/m231.9 (8.1) 31.2 (6.1) 0.481 31.4 (6.0) 0.601 32.4 (7.3) 0.661 0.102 Hype ension, % 66.7% 60.5% 0.239 60.1% 0.278 55.1% 0.046 0.062 Dyslipidemia, % 55.6% 58.2% 0.666 60.0% 0.486 63.0% 0.207 0.166 Time since diagnosis o diabe es, yea s 3 (0–11) 4 (0–11) 0.882 3 (0–10) 0.539 5 (0–12) 0.966 0.779 Diabe ic e inopa hy, % 26.4% 26.1% 0.932 18.3% 0.223 18.6% 0.432 0.198 Diabe ic kidney disease, % 15.1% 25.0% 0.010 16.9% 0.639 20.1% 0.197 0.408 Mac o ascula complica ions, % 23.2% 24.8% 0.662 20.5% 0.551 26.2% 0.448 0.378 Insulin ea men , % 20.9% 21.8% 0.814 21.6% 0.888 23.2% 0.614 0.563 Baseline popula ion cha ac e is ics acco ding o ca eine consump ion among women and men. * s no consump ion g oup. Kcal, kilocalo ie; BMI, Body mass index. Signi ican P o end alues a e shown in bold. F on ie s in Endoc inology | www. on ie sin.o g 3Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es cance . Mo ali y s a us and cause o dea h we e de e mined by NHANES linked Na ional Dea h Index public-access iles h ough Decembe 31, 2011. S a is ical Analysis All calcula ions ook in o accoun he complex su ey design o he NHANES da ase and we e analyzed acco ding o he CDC analy ic ecommenda ions (17). To assess he c ude associa ion be ween ca eine consump ion and ime o dea h, we pe o med a Kaplan-Meie cu e and log- ank es . We pe o med u he analysis using he Cox p opo ional haza ds models o adjus o po en ial con ounde s. We buil wo di e en Cox P opo ional Haza d models o analyze he p ima y ou come: one model including age a baseline, ace (Mexican Ame ican, o he Hispanic, non-Hispanic whi e, non-Hispanic black, o he ace), annual amily income (<$25000, $25000 o $75000, >$75000), smoking s a us (ne e smoke , cu en smoke , o o me smoke ), and diabe ic neph opa hy (glome ula il a ion a e <60 mL/min/1.73 m2o u ine albumin/c ea inine a io ≥300 mg/g) (model 1); Model 2 including all model 1 co a ia es plus body mass index (BMI) (<20.0, 20.0 o <25.0, 25.0 o <30.0, 30.0 o <35.0, 35.0 o <40.0, ≥40.0 kg/m2), educa ion le el [less han 9 h g ade, 9-11 h g ade, high-school g ade, some college o associa e’s (AA) deg ee, college g adua e and abo e], daily ca bohyd a e consump ion (g ams o ca bohyd a e pe 100 kcal), alcohol consump ion (no alcohol consump ion, <20 g ams/day, ≥20 g ams/day), yea s since diabe es diagnosis (undiagnosed, ≤5 yea s, 5 o ≤15 yea s, >15 yea s), diagnosis o hype ension, e inopa hy, mac o ascula complica ions (co ona y a e y disease, his o y o myoca dial in a c ion, o his o y o s oke), insulin ea men and su ey cycle (yea s 1999–2000, 2001–2002, 2003–2004, 2005–2006, 2007–2008, o 2009–2010). Rega ding cause-speci ic mo ali y, we only used he mo e es ic i e model (model 1), due o he low numbe o ou come e en s. We es ed o in e ac ions be ween ca eine consump ion and he o he 14 a iables in model 2 o all-cause mo ali y. We also conside ed physical ac i i y as an impo an po en ial con ounde . Physical ac i i y was measu ed di e en ly along he a ious NHANES cycles. The e o e, we chose o use a iables ha allowed ca ego iza ion o physical ac i i y le el in o h ee ca ego ies (low, in e media e, and high), o combine hem in o a single a iable. F om 1999 o 2006 he physical ac i i y le el was assessed wi h he ques ion “compa e ac i i y wi h o he s o he same age” (pa icipan s we e classi ied in o app oxima e e iles as low i “less ac i e,” as in e media e i “abou he same,” and as high i “mo e ac i e,” wi h 31, 28, and 41% o pa icipan s, espec i ely, alling in o hese ca ego ies). F om 2007 o 2010 he weekly me abolic equi alen s (MET) minu es o physical ac i i y (accoun ing o igo ous wo k- ela ed ac i i y, mode a e wo k- ela ed ac i i y, walking o bicycling o anspo a ion, igo ous leisu e- ime physical ac i i y, and mode a e leisu e- ime physical ac i i y) was di ided in o e iles (pa icipan s we e classi ied as low i included in he lowe MET-minu e e ile, as in e media e i in he middle MET-minu e e ile, and as high i in he highe MET-minu e e ile). The esul s o mo ali y associa ed wi h ca eine consump ion adjus ed o physical ac i i y a e p esen ed in he Supplemen a y Ma e ial. As sensi i i y analysis, we addi ionally pe o med he Cox p opo ional haza ds models using daily in ake o ca eine ( om all sou ces and om co ee, ea, o so d inks) as a con inuous a iable. Mul iple impu a ion by chained equa ions was used o dealing wi h missing da a ega ding co a ia es. Twen y impu a ions pe missing obse a ion we e pe o med and analyzed. A es o end o e inc easing ca eine consump ion ca ego ies was pe o med whe e each ca ego y median was modeled as a con inuous a iable in he eg ession. A wo-sided p- alue o <0.05 was conside ed s a is ically signi ican . Analyses we e pe o med wi h S a a ( e sion 14.2). RESULTS Associa ion o Ca eine Consump ion Wi h Die a y and Li es yle Fac o s Ca eine consump ion a baseline was associa ed wi h se e al o he die a y and li es yle ac o s, wi h some di e ences acco ding o sex (Table 1). In bo h men and women, compa ed wi h people who did no d ink ca eine-con aining be e ages, ca eine consume s we e mo e likely o be non-Hispanic whi e and cu en smoke s, o ha e a highe le el o educa ion, o ha e an annual amily income highe han $25000, and o consume less ca bohyd a es and less ibe s pe kilocalo ie inges ed. Women ha consumed ca eine we e mo e likely o be ea ed wi h insulin, had less diabe ic kidney disease, and a lowe a io o polyunsa u a ed/sa u a ed a y acids in ake. Ca eine Consump ion and Mo ali y Du ing a median 57 mon hs o ollow-up ( o al pe son-yea s, 21,606), 407 men and 351 women died. The e was a signi ican in e ac ion be ween sex and ca eine consump ion wi h espec o mo ali y (p=0.015 o in e ac ion in model 2). In he unadjus ed analysis (Figu e 1), and also a e mul i a ia e analysis, ca eine consump ion was associa ed wi h a dec ease in all-cause mo ali y in women (p=0.005 o end ac oss FIGURE 1 | Kaplan-Meie cu es o all-cause mo ali y by ca eine consump ion among women. F on ie s in Endoc inology | www. on ie sin.o g 4Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es ca ego ies, in model 2). A dose-dependen in e se associa ion be ween ca eine and all-cause mo ali y was obse ed. Haza d a ios (HR) o dea h among women who consumed ca eine, as compa ed wi h women who did no consume ca eine, we e as ollows: 0.57 (95% con idence in e al [CI], 0.40 o 0.82) o less han 100 mg o ca eine pe day, 0.50 (95% CI, 0.32 o 0.78) o 100 o less han 200 mg o ca eine, and 0.39 (95% CI, 0.23 o 0.64) o 200 mg o mo e o ca eine pe day (Table 2). In con as , his associa ion was no obse ed among men (Figu e 2), e en a e adjus men o po en ial con ounde s (Table 2). Speci ic causes o dea h we e also examined. The e we e 199 dea hs om CVD and 133 dea hs due o cance du ing he ollow-up. A e mul i a ia e adjus men , he e was no signi ican associa ion be ween ca eine consump ion and dea hs om CVD o cance , bo h in men and women (Table 2). Sou ce o Ca eine Consump ion and Mo ali y An analysis o ca eine consump ion acco ding o i s o igin on co ee, ea, o so d inks was also pe o med. In he unadjus ed analysis, and also a e mul i a ia e analysis, a educed isk in all-cause mo ali y was obse ed in women wi h diabe es who consumed ca eine om co ee (Table 3). The adjus ed haza d a ios we e as ollows: 0.74 (95% CI, 0.53 o 1.02) o less han 100 mg o ca eine pe day, 0.71 (95% CI, 0.46 o 1.09) o 100 mg TABLE 2 | Associa ion o ca eine consump ion wi h mo ali y. Associa ion o ca eine consump ion wi h mo ali y among women No consump ion (n=219) <100 mg/day (n=979) 100 o <200 mg/day (n=438) ≥200 mg/day (n=338) P o end Con inuous analysisa All-cause mo ali y No. o dea hs (%) 59 (26.9%) 170 (17.4%) 73 (16.7%) 49 (14.5%) 351 (17.8%) Unadjus ed HR – 0.71 (0.51–1.00) 0.63 (0.42–0.96) 0.46 (0.27–0.78) 0.010 0.86 (0.76–0.96) Model 1 HR – 0.60 (0.43–0.83) 0.52 (0.34–0.80) 0.42 (0.25–0.71) 0.020 0.88 (0.77–0.99) Model 2 HR – 0.57 (0.40–0.82) 0.50 (0.32–0.78) 0.39 (0.23–0.64) 0.005 0.86 (0.76–0.96) CVD mo ali y No. o dea hs (%) 16 (7.3%) 40 (4.1%) 15 (3.4%) 8 (2.4%) 79 (4.0%) Unadjus ed HR – 0.66 (0.33–1.30) 0.45 (0.22–0.93) 0.40 (0.15–1.08) 0.118 0.82 (0.68–0.99) Model 1 HR – 0.52 (0.26–1.02) 0.37 (0.18–0.76) 0.38 (0.14–1.00) 0.208 0.86 (0.70–1.06) Cance mo ali y No. o dea hs (%) 9 (4.1%) 26 (2.7%) 11 (2.5%) 8 (2.4%) 54 (2.7%) Unadjus ed HR – 0.87 (0.34–2.22) 0.70 (0.24–2.02) 0.49 (0.13–1.80) 0.204 0.88 (0.65–1.19) Model 1 HR – 0.71 (0.27–1.89) 0.52 (0.17–1.59) 0.35 (0.08–1.53) 0.126 0.84 (0.59–1.20) Associa ion o ca eine consump ion wi h mo ali y among men No consump ion (n=186) <100 mg/day (=739) 100 o <200 mg/day (n=470) ≥200 mg/day (n=579) P o end Con inuous analysis a All-cause mo ali y No. o dea hs (%) 50 (26.9%) 159 (21.5%) 82 (17.5%) 116 (20.0%) 407 (20.6%) Unadjus ed HR – 1.24 (0.81–1.91) 1.00 (0.64–1.57) 1.21 (0.81–1.82) 0.739 1.03 (0.97–1.09) Model 1 HR – 0.94 (0.62–1.41) 0.77 (0.49–1.23) 0.91 (0.60–1.39) 0.925 1.04 (0.98–1.10) Model 2 HR – 1.09 (0.70–1.69) 0.90 (0.55–1.47) 1.01 (0.65–1.56) 0.783 1.03 (0.97–1.10) CVD mo ali y No. o dea hs (%) 11 (5.9%) 50 (6.8%) 23 (4.9%) 36 (6.2%) 120 (6.1%) Unadjus ed HR – 2.09 (0.99–4.39) 1.59 (0.58–4.33) 1.92 (0.85–4.31) 0.749 1.07 (0.98–1.18) Model 1 HR – 1.47 (0.70–3.12) 1.13 (0.41–3.10) 1.18 (0.54–2.58) 0.517 1.07 (0.98–1.16) Cance mo ali y No. o dea hs (%) 12 (6.5%) 28 (3.8%) 18 (3.8%) 21 (3.6%) 79 (4.0%) Unadjus ed HR – 1.15 (0.51–2.63) 0.96 (0.37–2.47) 1.17 (0.52–2.62) 0.830 1.09 (0.97–1.22) Model 1 HR – 1.03 (0.42–2.54) 0.78 (0.28–2.13) 0.90 (0.37–2.18) 0.777 1.07 (0.95–1.22) Model 1: Adjus ed o age, ace, annual amily income, smoking s a us, and diabe ic kidney disease. Model 2: Adjus ed o co a ia es in Model 1 and body mass index, educa ion le el, daily ca bohyd a e consump ion, alcohol consump ion, yea s since diabe es diagnosis, diagnosis o hype ension, e inopa hy, mac o ascula complica ions, insulin ea men and su ey cycle. aHR o he con inuous analysis a e p esen ed o each 100 mg inc ease in ca eine consump ion. HR, Haza d Ra io; CVD, Ca dio ascula disease. Signi ican P o end alues and signi ican haza d a ios in he con inuous analysis a e shown in bold. F on ie s in Endoc inology | www. on ie sin.o g 5Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es FIGURE 2 | Kaplan-Meie cu es o all-cause mo ali y by ca eine consump ion among men. o less han 200 mg o ca eine, and 0.53 (95% CI, 0.35 o 0.80) o 200 mg o mo e o ca eine pe day (p=0.004 o end ac oss ca ego ies, in model 2). The e we e no signi ican associa ions o ca eine consump ion om co ee wi h ca dio ascula o cance mo ali y. Rega ding ca eine consump ion om ea and om so d inks, he e we e no signi ican associa ions wi h all-cause o cause-speci ic mo ali y (Supplemen a y Tables 1,2). Among men, he e we e no signi ican associa ions be ween sou ce o ca eine and mo ali y (Table 3, Supplemen a y Table 2). Sensi i i y Analysis The analysis o he associa ion be ween sou ce o ca eine and mo ali y using daily in ake o ca eine as a con inuous a iable showed simila esul s (Table 2,Supplemen a y Table 3). Among women, he HR o each 100 mg inc ease in ca eine consump ion o all-cause mo ali y was 0.86 (95% CI, 0.76–0.96; p=0.009) o unadjus ed analysis and 0.86 (95% CI, 0.76–0.96; p=0.011) o adjus ed analysis. The e ec o ca eine consump ion acco ding o ca eine sou ce on all-cause, ca dio ascula , and cance mo ali y among women and men we e also conco dan wi h he main analysis (Table 2,Supplemen a y Table 3). Fu he mo e, he associa ion o ca eine consump ion wi h mo ali y adjus ing o physical ac i i y was also consis en wi h ou main analysis (Supplemen a y Table 4). DISCUSSION Ou s udy showed a dose-dependen p o ec i e e ec o ca eine consump ion on all-cause mo ali y among women wi h diabe es. The e was no signi ican associa ion be ween ca eine consump ion and mo ali y among men wi h diabe es. Al hough ca eine consump ion in women was associa ed wi h lowe all-cause mo ali y, no associa ion was ound be ween ca eine consump ion and ca dio ascula o cance mo ali y. When compa ing ca eine consump ion acco ding o i s o igin on co ee, ea, o so d inks, women wi h diabe es who consumed mo e ca eine om co ee also had educed isk o all-cause dea h. No di e ences on mo ali y we e obse ed on he adjus ed analysis o consump ion o ca eine om ea o so d inks. Howe e , hese esul s should be in e p e ed cau iously as he numbe o e en s in each sou ce o ca eine ca ego y was low. P e ious s udies had al eady shown a p o ec i e e ec o co ee consump ion in he gene al popula ion. Lo ield e al., o example, showed a dec eased isk o all-cause mo ali y in people wi h highe co ee consump ion (18). Speci ically, an in e se associa ion was ound be ween co ee consump ion and diabe es- ela ed dea h. In he subg oup o pa icipan s wi h sel - epo ed diabe es, his associa ion seemed s onge . Rega ding s udies in people wi h diabe es, a s udy in he Finnish popula ion in 3,837 pa ien s, ound an in e se ela ionship be ween co ee d inking and all-cause and ca dio ascula -associa ed mo ali y (12). On he o he hand, wo p e ious s udies ha e ob ained neu al esul s in pa ien s o bo h sexes. In a s udy including only emale nu ses wi h diabe es, habi ual co ee consump ion was no associa ed wi h ca dio ascula diseases o p ema u e mo ali y (10). Fu he mo e, no signi ican associa ion be ween co ee consump ion and mo ali y was obse ed in a p ospec i e coho o male heal h p o essionals wi h diabe es (11). Al hough he Finnish popula ion s udy esul s a e no consis en wi h hese wo s udies, hei esul s also sugges a p o ec i e e ec o ca eine-con aining be e ages as obse ed in ou s udy. Possible explana ions o hese di e ences be ween s udies migh be he use o di e en endpoin s, di e en du a ion o ollow-up and di e ences in he popula ion’s cha ac e is ics. Ou esul s sugges ha biological di e ences may exis be ween men and women ega ding he e ec s o ca eine consump ion. Conside ing he pa hophysiology and complica ions o diabe es, se e al s udies ha e highligh ed di e ences be ween sexes (19). Gene ic and epigene ic mechanisms, nu i ional ac o s, and seden a y li es yle ha e been shown o di e en ly a ec diabe es complica ions acco ding o sex (19). Fu he mo e, ca eine may induce di e en hemodynamic e ec s in men and women. In a double-blind ial compa ing age-ma ched women and men, women showed an inc ease in ca diac ou pu , whe eas men showed inc eased ascula esis ance a e a die a y dose o ca eine (20). These di e ences may pa ially explain why ca eine in ake is associa ed wi h educed mo ali y in women wi h diabe es whe eas he e ec s in mo ali y a e neu al in men wi h diabe es. In he gene al popula ion, some s udies ha e also sugges ed di e ences in he esponse o co ee be ween sexes. The in e se associa ion o co ee d inking wi h o al mo ali y has been shown o be educed in men compa ing o women (21–23). The di ec ion and s eng h o he associa ion be ween ca eine consump ion and mo ali y has a ied be ween s udies in he gene al popula ion. In a s udy including Japanese pa icipan s wi hou a his o y o cance , myoca dial in a c ion, o s oke a baseline, co ee consump ion was s ongly associa ed wi h educed all-cause and ca dio ascula mo ali y among women [HR o 0.48 (0.29–0.80) o 1–2 cups o co ee pe day and F on ie s in Endoc inology | www. on ie sin.o g 6Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es TABLE 3 | Associa ion o ca eine consump ion om co ee wi h mo ali y. Associa ion o ca eine consump ion om co ee wi h mo ali y among women No consump ion (n=734) <100 mg/day (n=707) 100 o <200 mg/day (n=316) ≥200 mg/day (n=217) P o end Con inuous analysisa All-cause mo ali y No. o dea hs (%) 127 (17.3%) 130 (18.4%) 58 (18.4%) 36 (16.6%) 351 (17.8%) Unadjus ed HR – 1.25 (0.94–1.66) 1.09 (0.75–1.60) 0.73 (0.50–1.09) 0.077 0.91 (0.83–0.99) Model 1 HR – 0.79 (0.60–1.04) 0.74 (0.47–1.15) 0.57 (0.38–0.85) 0.009 0.87 (0.78–0.98) Model 2 HR – 0.74 (0.53–1.02) 0.71 (0.46–1.09) 0.53 (0.35–0.80) 0.004 0.85 (0.76–0.95) CVD mo ali y No. o dea hs (%) 31 (4.2%) 32 (4.5%) 11 (3.5%) 5 (2.3%) 79 (4.0%) Unadjus ed HR – 1.51 (0.86–2.66) 0.98 (0.42–2.26) 0.50 (0.18–1.41) 0.123 0.83 (0.69–0.99) Model 1 HR – 0.96 (0.56–1.66) 0.72 (0.28–1.85) 0.42 (0.14–1.25) 0.110 0.81 (0.64–1.02) Cance mo ali y No. o dea hs (%) 19 (2.6%) 18 (2.6%) 12 (3.8%) 5 (2.3%) 54 (2.7%) Unadjus ed HR – 1.66 (0.80–3.42) 1.81 (0.74–4.38) 0.89 (0.23–3.41) 0.905 0.99 (0.76–1.29) Model 1 HR – 1.16 (0.54–2.48) 1.24 (0.47–3.23) 0.59 (0.15–2.39) 0.392 0.91 (0.64–1.29) Associa ion o ca eine consump ion om co ee wi h mo ali y among men No consump ion (n=656) <100 mg/day (n=572) 100 o <200 mg/day (n=358) ≥200 mg/day (n=388) P o end Con inuous analysisa All-cause mo ali y No. o dea hs (%) 132 (20.1%) 126 (22.0%) 70 (19.6%) 79 (20.4%) 407 (20.6%) Unadjus ed HR – 1.59 (1.19–2.13) 1.13 (0.77–1.66) 1.42 (0.93–2.08) 0.282 1.06 (1.00–1.12) Model 1 HR – 1.05 (0.76–1.45) 0.80 (0.57–1.11) 1.07 (0.74–1.55) 0.922 1.03 (0.96–1.11) Model 2 HR – 1.04 (0.74–1.47) 0.76 (0.53–1.08) 1.09 (0.76–1.56) 0.873 1.03 (0.96–1.11) CVD mo ali y No. o dea hs (%) 34 (5.2%) 45 (7.9%) 18 (5.0%) 23 (5.9%) 120 (6.1%) Unadjus ed HR – 2.13 (1.35–3.38) 1.15 (0.55–2.42) 1.48 (0.64–3.40) 0.791 1.09 (0.99–1.20) Model 1 HR – 1.33 (0.79–2.26) 0.70 (0.35–1.41) 0.97 (0.44–2.15) 0.563 1.06 (0.94–1.19) Cance mo ali y No. o dea hs (%) 17 (3.5%) 23 (3.5%) 19 (5.3%) 14 (3.6%) 79 (4.0%) Unadjus ed HR – 3.34 (1.58–7.05) 2.85 (1.39–5.84) 3.12 (1.79–5.42) 0.022 1.09 (1.00–1.19) Model 1 HR – 2.51 (1.10–5.71) 2.19 (1.00–4.81) 2.24 (1.10–4.56) 0.270 1.06 (0.93–1.20) Model 1: Adjus ed o age, ace, annual amily income, smoking s a us, and diabe ic kidney disease. Model 2: Adjus ed o co a ia es in Model 1 and body mass index, educa ion le el, daily ca bohyd a e consump ion, alcohol consump ion, yea s since diabe es diagnosis, diagnosis o hype ension, e inopa hy, mac o ascula complica ions, insulin ea men , and su ey cycle. aHR o he con inuous analysis a e p esen ed o each 100 mg inc ease in ca eine consump ion om co ee. HR, Haza d Ra io; CVD, Ca dio ascula disease. Signi ican P o end alues and signi ican haza d a ios in he con inuous analysis a e shown in bold. 0.45 (0.20–1.03) o 3 o mo e cups pe day compa ing wi h no consump ion) bu no in men (24). O he s udies ha e also shown in e se associa ions be ween consump ion o ca eine- con aining be e ages and mo ali y among women, albei wi h weake associa ions (22,25). Al hough some s udies sugges g ea e bene i s o consump ion o ca eine o co ee among women, hese indings a e no consis en ac oss s udies. Se e al s udies ha e shown simila in e se associa ions be ween co ee consump ion and mo ali y in women and men (26–28). The ype o ca eine-con aining be e age, he popula ion’s isk ac o s o he du a ion o ollow-up may explain he di e ences be ween s udies. The bene i s o co ee may be di ec ly ela ed o ca eine o o o he componen s p esen in co ee, including mine als, phy ochemicals, and an ioxidan s (25,29). The an ioxidan capaci y o hese d inks may con ibu e o he heal h-p o ec i e e ec desc ibed wi h deca eina ed co ee consump ion (7). The lack o signi ican di e ences on mo ali y ega ding ca eine consump ion om ea on ou s udy may be explained by insu icien powe . A ecen me a-analysis ound a signi ican associa ion be ween ea consump ion and educ ion o all-cause mo ali y; u he mo e, black ea was in e sely associa ed wi h cance mo ali y (30). Tsujimo o e al. also used NHANES da a o e alua e he e ec s o ca eine consump ion in he gene al popula ion (31). In hei main analysis, ca eine in ake was associa ed wi h a dec eased isk o all-cause mo ali y. Al hough i was no hei main objec i e, he au ho s also pe o med an addi ional analysis limi ed o he F on ie s in Endoc inology | www. on ie sin.o g 7Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es pa icipan s wi h diabe es. Con a y o ou esul s, hey epo ed a non-signi ican associa ion be ween ca eine consump ion and mo ali y among pa icipan s wi h diabe es. In he s udy by Tsujimo o e al., no adjus men s we e made o diabe es-speci ic pa ame e s, including diabe es du a ion, ype o ea men and complica ions o diabe es. Pa icipan s wi h diabe es we e no s a i ied acco ding o sex and no adjus men was pe o med o he p esence o kidney disease. Fu he mo e, in he s udy by Tsujimo o e al., pa icipan s wi h missing in o ma ion on any o he po en ial con ounde s we e excluded, whe eas we included hese pa ien s using mul iple impu a ion. The di e ences in pa icipan s’ selec ion and in he s a is ical analyses echniques used p obably accoun o he di e ences be ween he s udies. Ou s udy has se e al s eng hs, including he e alua ion o a coho o pa icipan s om a la ge da abase ep esen a i e o he Ame ican popula ion. Da a was p ospec i ely collec ed and included ha d ou come measu es such as dea h and cause- speci ic mo ali y. The p esence o de ailed in o ma ion abou he pa icipan s allowed o adjus men o he main biologically plausible con ounde s. As o limi a ions, i should be no ed ha ca eine consump ion was e alua ed by 24-h die a y ecalls. I canno be excluded ha da a gene a ed using his me hod may no ep esen long- e m die a y habi s. We conside ha he inclusion o da a om non-consecu i e ecalls o es ima e usual die a y in ake dis ibu ions minimizes his isk. Al hough we p esen addi ional in o ma ion ega ding die in he s udied popula ion (such as consump ion o ca bohyd a e, sa u a ed a , o ibe ), no adjus men was pe o med o addi i es p esen in ca eine-con aining be e ages. None heless, o he s udies showed signi ican associa ion be ween co ee consump ion and dec eased isk o dea h e en a e adjus men o co ee addi i es, such as c eam, milk, suga , o honey (32). E en hough we ha e ound a signi ican associa ion be ween ca eine consump ion and mo ali y in women wi h diabe es, i is possible ha he di e ences ound a e due o chance, unmeasu ed con ounde s, o he possibili y ha ca eine consume s also pe o m o he p o ec i e beha io s, con ibu ing o a heal hy use e ec . To minimize his possibili y, we ha e conside ed die a y ac o s and physical ac i i y as po en ial con ounde s. As he numbe o dea hs in ou s udy was low, hese es ima es should be cau iously in e p e ed. In conclusion, his la ge obse a ional s udy showed a signi ican in e se associa ion be ween ca eine consump ion and dea h om all causes in women wi h diabe es. These esul s sugges ha ad ising women wi h diabe es o d ink mo e ca eine may educe hei mo ali y. This would ep esen a simple, clinically bene icial, and inexpensi e op ion in emale pa ien s. Fu he s udies, ideally andomized clinical ials, a e needed o con i m his bene i . New esea ch should also ocus on he di e en e ec s o ca eine consump ion in men and women and on he bene i s o o he compounds p esen in ca eine-con aining be e ages. DATA AVAILABILITY STATEMENT The da ase analyzed o his s udy can be ound in: h ps://www. cdc.go /nchs/nhanes/index.h m. AUTHOR CONTRIBUTIONS JSN, LL, RM, MBV, and CVD join ly designed and conduc ed esea ch, de eloped he analy ical s a egy and did he s a is ical analysis. BC e iewed he analy ic s a egy and s a is ical analysis. JSN, LL, RM, MBV, and CVD join ly con ibu ed o he i s d a . All au ho s con ibu ed o in e p e a ion o da a o he wo k, c i ically e ised he wo k, and app o ed he inal e sion o he manusc ip . ACKNOWLEDGMENTS We hank he pa icipan s and s a o NHANES. Pa s o his s udy we e p esen ed in abs ac o m a he 47 h Eu opean Associa ion o he S udy o Diabe es Annual Mee ing, Lisbon, Po ugal 11–15 Sep embe 2017. SUPPLEMENTARY MATERIAL The Supplemen a y Ma e ial o his a icle can be ound online a : h ps://www. on ie sin.o g/a icles/10.3389/ endo. 2018.00547/ ull#supplemen a y-ma e ial REFERENCES 1. Ogu so a K, Da Rocha Fe nandes JD, Huang Y, Linnenkamp U, Gua igua a L, Cho NH, e al. IDF diabe es a las: global es ima es o he p e alence o diabe es o 2015 and 2040. Diabe es Res Clin P ac . (2017) 128:40–50. doi: 10.1016/j.diab es.2017.03.024 2. Ding M, Sa ija A, Bhupa hi aju SN, Hu Y, Sun Q, Han J, e al. Associa ion o co ee consump ion wi h o al and cause-speci ic mo ali y in 3 la ge p ospec i e coho s. Ci cula ion (2015) 132:2305–15. doi: 10.1161/CIRCULATIONAHA.115.017341 3. A nlo J, Vessby B, Rise us U. Co ee consump ion and insulin sensi i i y. JAMA (2004) 291:1199–201. doi: 10.1001/jama.291.10.1199-b 4. Lopez-Ga cia E, Van Dam RM, Qi L, Hu FB. Co ee consump ion and ma ke s o in lamma ion and endo helial dys unc ion in heal hy and diabe ic women. Am J Clin Nu . (2006) 84:888–93. doi: 10.1093/ajcn/84. 4.888 5. Kolo e ou E, Panagio akos DB, Pi sa os C, Ch ysohoou C, Geo gousopoulou EN, Laska is A, e al. The e alua ion o in lamma o y and oxida i e s ess bioma ke s on co ee-diabe es associa ion: esul s om he 10-yea ollow- up o he ATTICA S udy (2002-2012). Eu J Clin Nu . (2015) 69:1220–5. doi: 10.1038/ejcn.2015.98 6. Jiang X, Zhang D, Jiang W. Co ee and ca eine in ake and incidence o ype 2 diabe es melli us: a me a-analysis o p ospec i e s udies. Eu J Nu . (2014) 53:25–38. doi: 10.1007/s00394-013-0603-x 7. Ding M, Bhupa hi aju SN, Chen M, Van Dam RM, Hu FB. Ca eina ed and deca eina ed co ee consump ion and isk o ype 2 diabe es: a sys ema ic e iew and a dose- esponse me a-analysis. Diabe es Ca e (2014) 37:569–86. doi: 10.2337/dc13-1203 F on ie s in Endoc inology | www. on ie sin.o g 8Sep embe 2018 | Volume 9 | A icle 547 Ne es e al. Ca eine and Mo ali y in Diabe es 8. C ippa A, Discaccia i A, La sson SC, Wolk A, O sini N. Co ee consump ion and mo ali y om all causes, ca dio ascula disease, and cance : a dose- esponse me a-analysis. Am J Epidemiol. (2014) 180:763–75. doi: 10.1093/aje/kwu194 9. Van Dongen LH, Molenbe g FJ, Soedamah-Mu hu SS, K omhou D, Geleijnse JM. Co ee consump ion a e myoca dial in a c ion and isk o ca dio ascula mo ali y: a p ospec i e analysis in he Alpha Omega Coho . Am J Clin Nu . (2017) 106:1113–20. doi: 10.3945/ajcn.117.153338 10. Zhang WL, Lopez-Ga cia E, Li TY, Hu FB, Van Dam RM. Co ee consump ion and isk o ca dio ascula e en s and all-cause mo ali y among women wi h ype 2 diabe es. Diabe ologia (2009) 52:810–7. doi: 10.1007/s00125-009-1311-1 11. Zhang W, Lopez-Ga cia E, Li TY, Hu FB, Van Dam RM. Co ee consump ion and isk o ca dio ascula diseases and all-cause mo ali y among men wi h ype 2 diabe es. Diabe es Ca e (2009) 32:1043–5. doi: 10.2337/dc08-2251 12. Bidel S, Hu G, Qiao Q, Jousilah i P, An ikainen R, Tuomileh o J. Co ee consump ion and isk o o al and ca dio ascula mo ali y among pa ien s wi h ype 2 diabe es. Diabe ologia (2006) 49:2618–26. doi: 10.1007/s00125-006-0435-9 13. CDC (2017). Na ional Cen e o Heal h S a is ics and Cen e s o Disease Con ol and P e en ion. Abou he Na ional Heal h and Nu i ion Examina ion Su ey [Online]. A ailable online a : h ps://www.cdc.go /nchs/ nhanes/abou _nhanes.h m [Accessed May 4, 2018]. 14. Rhee JJ, Sampson L, Cho E, Hughes MD, Hu FB, Wille WC. Compa ison o me hods o accoun o implausible epo ing o ene gy in ake in epidemiologic s udies. Am J Epidemiol. (2015) 181:225–33. doi: 10.1093/aje/kwu308 15. Zip G, Chiappa M, Po e KS, Os chega Y, Lewis BG, Dos al J. Na ional heal h and nu i ion examina ion su ey: plan and ope a ions, 1999-2010. Vi al Heal h S a . (2013)1:1–37. 16. Somogyi LP. Ca eine In ake by he US Popula ion. Food and D ug Adminis a ion and Oak idge Na ional Labo a o y (2010). 17. Johnson C, Paulose-Ram R, Ogden CE, Ca oll M, K uszon-Mo an D, Doh mann S, e al. Na ional heal h and nu i ion examina ion su ey: analy ic guidelines, 1999-2010. Vi al Heal h S a . (2013) 1–24. 18. Lo ield E, F eedman ND, G auba d BI, Gue in KA, Black A, Huang WY, e al. Associa ion o co ee consump ion wi h o e all and cause-speci ic mo ali y in a la ge US p ospec i e coho s udy. Am J Epidemiol. (2015) 182:1010–22. doi: 10.1093/aje/kw 146 19. Kau zky-Wille A, Ha ei e J, Pacini G. Sex and gende di e ences in isk, pa hophysiology and complica ions o ype 2 diabe es melli us. Endoc Re . (2016) 37:278–316. doi: 10.1210/e .2015-1137 20. Ha ley TR, Lo allo WR, Whi se TL. Ca dio ascula e ec s o ca eine in men and women. Am J Ca diol. (2004) 93:1022–6. doi: 10.1016/j.amjca d.2003.12.057 21. Kleemola P, Jousilah i P, Pie inen P, Va iainen E, Tuomileh o J. Co ee consump ion and he isk o co ona y hea disease and dea h. A ch In e n Med. (2000) 160:3393–400. doi: 10.1001/a chin e.160. 22.3393 22. Lopez-Ga cia E, Van Dam RM, Li TY, Rod iguez-A alejo F, Hu FB. The ela ionship o co ee consump ion wi h mo ali y. Ann In e n Med. (2008) 148:904–14. doi: 10.7326/0003-4819-148-12-200806170-00003 23. Wu JN, Ho SC, Zhou C, Ling WH, Chen WQ, Wang CL, e al. Co ee consump ion and isk o co ona y hea diseases: a me a-analysis o 21 p ospec i e coho s udies. In J Ca diol. (2009) 137:216–25. doi: 10.1016/j.ijca d.2008.06.051 24. Sugiyama K, Ku iyama S, Akh e M, Kakizaki M, Nakaya N, Ohmo i-Ma suda K, e al. Co ee consump ion and mo ali y due o all causes, ca dio ascula disease, and cance in Japanese women. J Nu . (2010) 140:1007–13. doi: 10.3945/jn.109.1 09314 25. Ande sen LF, Jacobs DRJ , Ca lsen MH, Blomho R. Consump ion o co ee is associa ed wi h educed isk o dea h a ibu ed o in lamma o y and ca dio ascula diseases in he Iowa Women’s Heal h S udy. Am J Clin Nu . (2006) 83:1039–46. doi: 10.1093/ajcn/83.5.1039 26. Tamakoshi A, Lin Y, Kawado M, Yagyu K, Kikuchi S, Iso H. E ec o co ee consump ion on all-cause and o al cance mo ali y: indings om he JACC s udy. Eu J Epidemiol. (2011) 26:285–93. doi: 10.1007/s10654-011-9548-7 27. F eedman ND, Pa k Y, Abne CC, Hollenbeck AR, Sinha R. Associa ion o co ee d inking wi h o al and cause-speci ic mo ali y. N Engl J Med. (2012) 366:1891–904. doi: 10.1056/NEJMoa1112010 28. Gun e MJ, Mu phy N, C oss AJ, Dossus L, Da ois L, Faghe azzi G, e al. Co ee d inking and mo ali y in 10 eu opean coun ies: a mul ina ional coho s udy. Ann In e n Med. (2017) 167:236–47. doi: 10.7326/M16-2945 29. Van Dam RM, Feskens EJ. Co ee consump ion and isk o ype 2 diabe es melli us. Lance (2002) 360:1477–8. doi: 10.1016/S0140-6736(02)11436-X 30. Tang J, Zheng JS, Fang L, Jin Y, Cai W, Li D. Tea consump ion and mo ali y o all cance s, CVD and all causes: a me a-analysis o eigh een p ospec i e coho s udies. B J Nu . (2015) 114:673–83. doi: 10.1017/S0007114515002329 31. Tsujimo o T, Kajio H, Sugiyama T. Associa ion be ween ca eine in ake and all-cause and cause-speci ic mo ali y: a popula ion- based p ospec i e coho s udy. Mayo Clin P oc. (2017) 92:1190–202. doi: 10.1016/j.mayocp.2017.03.010 32. Ga dene H, Rundek T, W igh CB, Elkind MS, Sacco RL. Co ee and ea consump ion a e in e sely associa ed wi h mo ali y in a mul ie hnic u ban popula ion. J Nu . (2013) 143:1299–308. doi: 10.3945/jn.112.173807 Con lic o In e es S a emen : The au ho s decla e ha he esea ch was conduc ed in he absence o any comme cial o inancial ela ionships ha could be cons ued as a po en ial con lic o in e es . Copy igh © 2018 Ne es, Lei ão, Mag iço, Bigo e Viei a, Viegas Dias, Oli ei a, Ca alho and Clagge . This is an open-access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (CC BY). The use, dis ibu ion o ep oduc ion in o he o ums is pe mi ed, p o ided he o iginal au ho (s) and he copy igh owne (s) a e c edi ed and ha he o iginal publica ion in his jou nal is ci ed, in acco dance wi h accep ed academic p ac ice. No use, dis ibu ion o ep oduc ion is pe mi ed which does no comply wi h hese e ms. F on ie s in Endoc inology | www. on ie sin.o g 9Sep embe 2018 | Volume 9 | A icle 547