Alcohol intake, wine consumption and thedevelopment of depression: the PREDIMED study

Alcohol in ake, wine consump ion and he de elopmen o dep ession: he PREDIMED s udy Gea e al. Gea e al. BMC Medicine 20132013, 11:192 h p://www.biomedcen al.com/1741-7015/x11/192 RESEARCH ARTICLE Open Access Alcohol in ake, wine consump ion and he de elopmen o dep ession: he PREDIMED s udy Al edo Gea 1 , Juan J Beunza 2 , Ramón Es uch 3,4 , Almudena Sánchez-Villegas 3,5 , Jo di Salas-Sal adó 3,6 , Pila Buil-Cosiales 7 , En ique Gómez-G acia 3,8 , Ma ía-Isabel Co as 3,9 , Dolo es Co ella 3,10 , Miquel Fiol 3,11 , Fe nando A ós 3,12 , José Lape a 3,13 , Rosa-Ma ía Lamuela-Ra en ós 3,14 , Julia Wä nbe g 3,8 , Xa ie Pin ó 3,15 , Lluis Se a-Majem 3,5 and Miguel A Ma ínez-González 1,3* , o he PREDIMED GROUP Abs ac Backg ound: Alcoholic be e ages a e widely consumed. Dep ession, he mos p e alen men al diso de wo ldwide, has been ela ed o alcohol in ake. We aimed o p ospec i ely assess he associa ion be ween alcohol in ake and inciden dep ession using epea ed measu emen s o alcohol in ake. Me hods: We ollowed-up 5,505 high- isk men and women (55 o 80 y) o he PREDIMED T ial o up o se en yea s. Pa icipan s we e ini ially ee o dep ession o a his o y o dep ession, and did no ha e any his o y o alcohol- ela ed p oblems. A 137-i em alida ed ood equency ques ionnai e adminis e ed by a die ician was epea ed annually o assess alcohol in ake. Pa icipan s we e classi ied as inciden cases o dep ession when hey epo ed a new clinical diagnosis o dep ession, and/o ini ia ed he use o an idep essan d ugs. Cox eg ession analyses we e i ed o e 23,655 pe son-yea s. Resul s: Mode a e alcohol in ake wi hin he ange o 5 o 15 g/day was signi ican ly associa ed wi h lowe isk o inciden dep ession (haza d a io (HR) and 95% con idence in e al (95% CI) = 0.72 (0.53 o 0.98) e sus abs aine s). Speci ically, wine consump ion in he ange o wo o se en d inks/week was signi ican ly associa ed wi h lowe a es o dep ession (HR (95% CI) = 0.68 (0.47 o 0.98)). Conclusions: Mode a e consump ion o wine may educe he incidence o dep ession, while hea y d inke s seem o be a highe isk. Keywo ds: Wine, Alcohol, Dep ession, Coho Backg ound Mos cul u es and coun ies include alcoholic be e ages as pa o hei usual die . Alcohol in ake is di e en o e wo ld egions ega ding he habi ual ype o be e age and he pa e n o consump ion ( equency and a e age in ake). In gene al e ms, he consump ion o alcoholic be e ages is inc easing wo ldwide [1]. Unipola dep ession is he mos p e alen men al diso de in he wo ld and i is inc easing s eadily [2]. The simul aneous p esence o alcohol- ela ed p oblems and dep ession is common [3]. Howe e , p oblema ic in ake needs o be dis inguished om mode a e in ake. P oblema ic alcohol in ake may be associa ed wi h dep es- sion no only because o inc eased in ake o e hanol bu also because o o he alcohol- ela ed unheal hy li es yles o because o he social en i onmen su ounding p oblema ic d inke s (job loss, amily p oblems, inancial p oblems o o he addic ions). Any o hese ci cums ances may be a po en ial igge o dep ession, e en in he absence o a speci ic de imen al ole o e hanol. Unipola dep ession and ca dio ascula disease a e likely o sha e some common pa hophysiological mechanisms. Mode a e alcohol in ake, especially alcohol om wine, has been epea edly epo ed o be in e sely associa ed wi h he incidence o ca dio ascula disease [4-6]. Some o he esponsible mechanisms o his in e se associa ion a e likely o be in ol ed also in a educed isk o dep ession * Co espondence: [email p o ec ed] 1 Depa men o P e en i e Medicine and Public Heal h, Medical School- Clinica Uni e sidad de Na a a, Pamplona, Spain 3 CIBER Fisiopa ología de la Obesidad y Nu ición (CIBERObn), Ins i u o de Salud Ca los III, Mad id, Spain Full lis o au ho in o ma ion is a ailable a he end o he a icle © 2013 Gea e al.; licensee BioMed Cen al L d. 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 (h p://c ea i ecommons.o g/licenses/by/2.0), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. Gea e al. BMC Medicine 2013, 11:192 h p://www.biomedcen al.com/1741-7015/11/192 [7,8]. In he exis ing li e a u e abou non-p oblema ic alcohol use and dep ession, longi udinal s udies a e incon- sis en [9-20]. Howe e , none o hem has in es iga ed he speci ic ole o each be e age. Nei he ha e hey used epea ed measu emen s on alcohol in ake du ing ollow-up o upda e he in o ma ion on exposu e o alcohol. We p ospec i ely e alua ed he incidence o dep ession among ligh o mode a e d inke s om an olde Medi e - anean coho a high ca dio ascula isk using epea ed measu emen s o in ake. An in e es ing cha ac e is ic o his popula ion is ha wine was he mos equen ly consumed alcoholic be e age. We also e alua ed speci ic- ally he associa ion o wine wi h inciden dep ession, using epea ed measu emen s o wine consump ion. Me hods Subjec s We assessed pa icipan s om he PREDIMED S udy (“P e ención con Die a Medi e ánea”(P e en ion wi h Medi e anean Die )), which is a la ge, pa allel-g oup, andomized, mul icen e , con olled, clinical ial conduc ed in Spain. The aim o he PREDIMED ial was o assess he e ec o he Medi e anean die on he p ima y p e en ion o ca dio ascula disease (www.p edimed.es; www.p edimed.o g) [21]. Pa icipan s we e men and women aged 55 o 80 and 60 o 80 yea s, espec i ely. They we e ee o ca dio ascula diseasea baselineandme a leas oneo he wo ollowing c i e ia: ype 2 diabe es melli us o he p esence o h ee o mo e co ona y hea disease isk ac o s. P e ious his o y o ca dio ascula disease, any se e e ch onic illness, his o y o ood alle gy, in ole ance o oli e oil o nu s, d ug addic ion o ch onic alcoholism we e exclusion c i e ia in he PREDIMED ial. Subjec s we e sc eened o p oblema ic use o alcohol using he CAGE ques ionnai e (a ou -i em sc eening ool) [21] and hey we e no included in he ial i hey answe ed posi i ely o a leas wo ques ions on his ques ionnai e. Pa icipan s who me he p e ious inclusion c i e ia (n = 7,447) we e andomly assigned o h ee in e en ions: Medi e anean die supplemen ed wi h ex a i gin oli e oil, Medi e anean die supplemen ed wi h mixed nu s o a con ol g oup. They we e annually in e iewed by a die ician, ob aining in o ma ion abou li es yle, die and inciden diseases. Fu he aspec s o he me hods and design o he PREDIMED ial ha e been epo ed else- whe e in de ail [22,23]. The Resea ch and E hic Commi ee o he Hospi al Clínic (Ba celona, Spain), acc edi ed by he Depa men o Heal h and Human Se ices and egula ed by he Fede alwide Assu ance o he P o ec ion o Human Subjec s o In e na ional (Non-US) Ins i u ions # 00000738 app o ed he s udy p o ocol on 16 July 2002. This ial has been egis e ed in he London Cu en Con olled T ials wi h he numbe ISRCTN 35739639 [24]. All pa icipan s p o ided w i en in o med consen . Ou o 7,447 pa icipan s, we excluded 153 pa icipan s wi h o al ene gy in ake ou o p ede ined limi s (800 and 4,000 Kcal/day o men and 500 and 3,500 Kcal/day o women) [25]. Ano he 1,579 subjec s who epo ed a baseline p e alen dep ession, p e ious his o y o dep ession o use o an idep essan d ugs we e also excluded ( o ensu e he empo al sequence and, he e o e, a oid e e se causa ion bias). Addi ionally, 210 pa icipan s we e los o ollow-up o did no ha e alcohol in o ma ion and we e also excluded. Finally, a subsample o 5,505 pa - icipan s was included in he analyses o he p esen s udy (Addi ional ile 1). Exposu e assessmen Alcohol in ake was assessed a baseline wi h a alida ed 137-i em semi-quan i a i e ood- equency ques ionnai e (FFQ), which included nine ques ions on consump ion o di e en alcoholic be e ages (di e en ypes o wine, bee and spi i s). In he alida ion s udy o he Spanish e sion o his ques ionnai e, he in a-class co ela ion coe icien be ween alcohol in ake om he FFQ and epea ed ood eco ds was 0.82 [26]. This FFQ was epea edly adminis e ed each yea du ing ollow-up. In o de o upda e alcohol in o ma ion, we conside ed alcohol in ake using epea ed measu emen s o die om all a ailable FFQs. Pa icipan s we e di ided in o ou g oups acco ding o hei alcohol in ake: abs aine s, hose who epo ed d inking less han 5 g/day, hose wi h an in ake anging om >5 o 15 g/day, and he ou h g oup wi h an in ake highe han 15 g/day. To assess he speci ic ole o wine, pa icipan s we e di ided in o i e g oups acco ding o hei a e age numbe o weekly d inks o wine: abs aine s, less han one d ink/ week, one o less han wo d inks/week, wo o se en d inks/week, and mo e han se en d inks/week. Pa ici- pan s who did no d ink wine bu d ank o he ypes o alcoholic be e ages we e excluded om hese speci ic analyses. Ou come assessmen An inciden case o dep ession was de ined as a diagnosis o dep ession made by a physician and epo ed by pa ici- pan s in any o he ollow-up in e iews, o a posi i e epo o habi ual use o an idep essan d ugs. Con ounde assessmen We ob ained in o ma ion abou medical, socio-demo- g aphic, an h opome ic and li es yle a iables wi h s anda dized ques ionnai es (see www.p edimed.es) a he baseline in e iew. Gea e al. BMC Medicine 2013, 11:192 Page 2 o 10 h p://www.biomedcen al.com/1741-7015/11/192 S a is ical analysis Cox eg ession models we e used o assess he ela ionship be ween ca ego ies o baseline alcohol in ake and he subsequen incidence o dep ession du ing ollow-up. Haza d a ios (HR) and hei 95% con idence in e als (95% CI) we e calcula ed using he abs aine s g oup as he e e ence ca ego y. En y ime was de ined as he da e a ec ui men . Exi ime was de ined as he da e a diagnosis o dep ession o cases and, o pa icipan s who did no de elop dep ession, as he da e when comple ing he las in e iew, 1 Decembe 2010, o da e a dea h, whiche e came i s . We adjus ed he mul i a iable models o he ollowing po en ial con ounde s: age, sex, smoking s a us (non-smoke s, ex-smoke s and cu en smoke s), physical ac i i y (MET-min/d), o al ene gy in ake (Kcal/ day), baseline body mass index (kg/m 2 ), ma i al s a us (ma ied o no ), in e en ion g oup (Medi e anean die + i gin oli e oil, Medi e anean die + nu s, low a die ), ec ui ing cen e (14 cen e s), educa ional le el (in i e ca ego ies om illi e a e o uni e si y g adua e), and numbe o pe sons li ing a home (li ing alone o no ). We e alua ed he in e ac ion be ween sex and alcohol in ake on he de elopmen o dep ession by calcula ing he likelihood a io es be ween he ully adjus ed model and he same model adding he in e ac ion p oduc - e m. To ake ad an age o he yea ly epea ed measu emen s o die , we upda ed he calcula ion o alcohol in ake o pa icipan s e e y yea om baseline o he incidence o dep ession om he epea ed FFQs. To upda e alcohol in ake wi h all a ailable longi udinal da a, and o analyze he associa ion be ween alcohol in ake and dep ession a all ime-poin s simul aneously, we used gene alized es ima ing equa ions (GEE) wi h S a a 12.0 (S a aCo p, CollegeS a ion,TX,USA).Weassumedabinomialdis- ibu ion wi h logi models and he uns uc u ed ma ix as he wo king co ela ion s uc u e [27]. We de ined he coho isk as pa icipan s who emained ee o dep ession a he beginning o each wo-yea ollow-up pe iod. Pa icipan s who had been classi ied as inciden cases we e excluded om subsequen ollow-up. Then, only he i s inciden case o dep ession was conside ed as ou come. In he analysis o epea ed measu emen s, o allow o a su icien ly long ollow-up pe iod, we modeled he incidence o dep ession in ela ion o he cumula i e a e age alcohol in ake om all a ailable die a y ques ionnai es up o he s a o each wo-yea ollow-up in e al, bu we excluded cases occu ing in he i s yea o ha pe iod and ook in o accoun only new cases occu ing a e one yea . Fo example, o a pa icipan ec ui ed in 2005 he incidence o dep ession om 2007 h ough 2008 was ela ed o he alcohol in ake epo ed on he 2005 and 2006 ques ionnai es. The e o e, we assumed, an induc ion pe iod longe han one yea bu no longe han wo yea s and alcohol in ake was conside ed as he cumula i e a e age o all a ailable FFQs, om baseline o he beginning o each wo-yea ollow-up pe iod. Se e al sensi i i y analyses we e conduc ed e i - ing Cox eg ession analyses a e a) es ablishing he 1 s and he 99 h o b) he 5 h and 95 h pe cen iles o ene gy in ake as allowable limi s, c) a e excluding diabe ics, d) a e excluding pa icipan s olde han 75 yea s, e) a e excluding pa icipan s younge han 65 yea s, ) a e depu a ing he abs aine s’g oup (excluding pa icipan s who epo ed any alcohol in ake h oughou hei li e bu no cu en ly a baseline), g) a e including pa icipan s wi h a p io his o y o dep ession a baseline and including his a iable in he mul iple adjus ed model, and h) a e es ablishing a wo- o h ee-yea induc ion pe iod (ins ead o one o wo yea s) be ween exposu e and ou come in he epea ed measu emen analyses. Mo eo e , o assess he in luence o wine consump ion on he de elopmen o dep ession, we conduc ed bo h Cox eg ession and GEE analyses, using abs aine s as he e e ence ca ego y. In hese analyses, we adjus ed o he same po en ial con ounde s as men ioned abo e, and also o alcohol in ake om o he sou ces. Addi ionally, we e- an he analyses o wine consump ion among only-wine d inke s o con ol o con ounding by o he alcoholic be e ages. Mo eo e , we e alua ed he po en ial non-linea associ- a ion be ween alcohol in ake and inciden dep ession wi h he use o es ic ed cubic splines, o pa icipan s d inking no mo e han 80 g ams o alcohol pe day (99.42% o pa icipan s) [28]. Finally, we examined he incidence and eco e y om dep ession applying ixed-e ec s eg ession as sugges ed by Fe gusson e al. [29] o con ol o obse ed and non- obse ed ac o s ha do no a y o a subjec o e ime. Resul s The main cha ac e is ics o he 5,505 pa icipan s (2,683 males and 2,822 emales) acco ding o hei o al alcohol in ake a e shown in Table 1. As men ioned be o e, his Medi e anean popula ion consis ed o olde men and wo- men, wi h a mean age o 67 yea s. Highe alcohol in ake was posi i ely associa ed wi h being male (88% o pa ici- pan s d inking mo e han 15 g/day we e male), p ac icing mo e leisu e- ime physical ac i i y, being a cu en o o - me smoke , ha ing highe o al ene gy in ake, being ma - ied, bu i was in e sely associa ed wi h educa ional le el. A o al o 443 inciden cases o dep ession we e iden i- ied du ing he ollow-up pe iod, which acc ued 23,655 pe son-yea s. The Cox eg ession analysis using baseline alcohol in ake as exposu e showed an in e se associa ion be ween o al alcohol in ake and he incidence o dep ession, wi h no sig- ni ican linea o quad a ic end in he mul iple-adjus ed analyses. In he mul iple-adjus ed model, only he associ- a ion o he ca ego y o low- o-mode a e in ake (>5 o 15 Gea e al. BMC Medicine 2013, 11:192 Page 3 o 10 h p://www.biomedcen al.com/1741-7015/11/192 g/day) emained s a is ically signi ican (HR (95% CI) = 0.72 (0.53 o 0.98)) (Table 2). Examining he associa ion o hose pa icipan s who d ank mo e han 40 g/day, we ound ha hey we e a highe isk bu his associa ion was no s a is ically signi ican , p obably due o he small num- be o hea y d inke s (HR (95% CI = 1.34 (0.69 o 2.59)). In he epea ed-measu emen analyses using as exposu e he yea ly upda ed in o ma ion on alcohol in ake, we obse ed a s onge in e se associa ion. We ound a s a is ically signi ican in e se associa ion o ligh d inke s (>0 o 5 g/day) wi h a ela i e isk (95% CI) o 0.73 (0.57 o 0.95) and also o low- o-mode a e d inke s (>5 o 15 g/day) wi h a ela i e isk o 0.69 (0.50 o 0.96). We ound no e ec modi ica ion o alcohol in ake by sex (P= 0.45). Al hough he in e ac ion was non-signi ican , we e- an he analyses sepa a ely o men and women o ake in o accoun biological di e ences be ween he sexes. The esul s a e a ailable in he supplemen a y ma e ial (Addi ional ile 2). To accoun o a non-linea associa ion, we used e- s ic ed cubic spline analysis. The U-shaped associa ion ound, ep esen ed in Figu e 1, showed a consis en p o- ec ion agains inciden dep ession o baseline mode a e alcohol in ake. Table 1 Baseline cha ac e is ics o pa icipan s acco ding o baseline alcohol in ake Alcohol in ake ca ego ies (g/day) 0 >0 o 5 >5 o 15 >15 N 1,818 1,356 1,279 1,052 Sex ( emale %) 78 59 38 12 Age (yea s) 68.5 (6.1) 67.0 (6.1) 66.4 (6.2) 65.6 (6.1) BMI (kg/m 2 ) 30.3 (4.1) 30.0 (3.9) 29.4 (3.5) 29.3 (3.3) Leisu e ime physical ac i i y (MET-min/d) 201 (210) 230 (232) 267 (276) 318 (270) Seconda y school o highe (%) 13 23 28 36 Cu en smoke s (%) 7 12 17 29 Fo me smoke s (%) 15 24 36 44 Ma i al s a us (% ma ied) 71 78 84 89 Li ing alone (%) 12 9 7 5 To al ene gy in ake (Kcal/day) 2,054 (511) 2,184 (509) 2,330 (511) 2,595 (533) Alcohol (g/day) - 2.0 (1.4) 9.8 (2.7) 35 (17) Wine (g alcohol/day) - 1.3 (1.4) 7.4 (3.7) 25 (16) Bee (g alcohol/day) - 0.5 (0.9) 1.7 (2.7) 4.8 (8.4) Spi i s (g alcohol/day) - 0.2 (0.5) 0.6 (1.5) 4.2 (8.8) Mean and S anda d De ia ions, o %. The PREDIMED S udy 2003 o 2010. BMI, body mass index; MET, me abolic equi alen ask. Table 2 Risks o dep ession acco ding o ca ego ies o daily alcohol in ake Alcohol in ake ca ego ies (g/day) 0 >0 o 5 >5 o 15 >15 P o line end Baseline alcohol in ake a Cases/Pe son-yea s 195/7,777 114/5,728 79/5,390 55/4,760 C ude model 1 (Re .) 0.79 (0.63 o 1.00) 0.59 (0.46 o 0.77) 0.44 (0.33 o 0.60) <0.001 Age and sex-adjus ed 1 (Re .) 0.91 (0.72 o 1.15) 0.81 (0.62 o 1.07) 0.81 (0.58 o 1.14) 0.347 Mul iple-adjus ed model c 1 (Re .) 0.97 (0.75 o 1.25) 0.72 (0.53 o 0.98) 0.79 (0.53 o 1.16) 0.522 Upda ed alcohol in ake b C ude model 1 (Re .) 0.63 (0.49 o 0.81) 0.49 (0.37 o 0.66) 0.37 (0.26 o 0.52) <0.001 Age and sex-adjus ed 1 (Re .) 0.73 (0.56 o 0.94) 0.71 (0.52 o 0.97) 0.71 (0.48 o 1.05) 0.727 Mul iple-adjus ed model c 1 (Re .) 0.73 (0.57 o 0.95) 0.69 (0.50 o 0.96) 0.69 (0.46 o 1.04) 0.773 a Haza d a ios (95% con idence in e als) o inciden dep ession acco ding o ca ego ies o baseline daily alcohol in ake. b Rela i e isks (95% con idence in e als) o ca ego ies o upda ed alcohol in ake, using epea ed measu emen s o die du ing ollow-up. To a oid e e se causali y bias, an induc ion pe iod o a leas one yea , bu no longe han wo yea s was assumed. We conside ed as inciden cases o dep ession hose occu ing only du ing he second yea o e e y wo-yea ollow-up in e al. c Adjus ed o age, sex, smoking, physical ac i i y (MET-min/d), o al ene gy in ake (Kcal/day), baseline body mass index (kg/m 2 ), ma i al s a us, in e en ion g oup, ec ui ing cen e , educa ional le el and he numbe o pe sons li ing a home. The PREDIMED S udy 2003 o 2010. Gea e al. BMC Medicine 2013, 11:192 Page 4 o 10 h p://www.biomedcen al.com/1741-7015/11/192 D inking wine ( wo o se en d inks/week) was in e sely associa ed wi h he incidence o dep ession (Table 3). In he mul iple-adjus ed model, we ound a HR (95% CI) o 0.68 (0.47 o 0.98) o he baseline consump ion o wo o se en d inks/week o wine. Using epea ed measu emen s o upda e wine consump ion, we also obse ed a s eng h- ening o his in e se associa ion wi h a ela i e isk o 0.57 (0.40 o 0.79) o hose who d ank wo o se en d inks/ week o wine. In ou sample, 1,350 pa icipan s we e only- wine d inke s. Among hem he esul s we e consis en wi h hose p e iously discussed (Table 3). Wine consump- ion accoun ed o 82% o alcohol in ake a iabili y in his sample. The numbe o only-bee d inke s was jus 308 and he numbe o only-spi i s d inke s was 61. Only 1.63% o pa icipan s d ank mo e han one d ink/day o bee , and only 0.09% o spi i s. Consequen ly, we could no do o bee o spi i s he ype o sepa a e analysis ha we did o wine. In o de o in es iga e possible sou ces o bias in he es ima ion o he ela ionship be ween alcohol in ake and dep ession, we conduc ed se e al sensi i i y analyses (Table 4). Resul s did no subs an ially change excep when excluding diabe ics and o o he analyses wi h a subs an ial numbe o exclusions, p obably due o a lack o s a is ical powe . Resul s we e also consis en when we in es iga ed he incidence and eco e y om dep ession using ixed-e ec s models (Addi ional ile 3). Discussion We ound ha o al low- o-mode a e alcohol d inking (5 o 15 g/day) was associa ed wi h lowe isk o dep ession. A s onge in e se associa ion was ound o low- o-mode - a e wine-d inke s ( wo o se en glasses/week). These esul s seem pa adoxical as p e ious s udies ound a di ec associa ion be ween alcohol in highe amoun s, especially Alcohol Use Diso de s (AUD), and dep essi e symp oms, [11,17-20,30]. Howe e , lowe amoun s o alco- hol in ake migh exe a p o ec ion as i has been obse ed o co ona y hea disease (CHD). In ac , i is belie ed ha dep ession and CHD sha e common pa hophysiological mechanisms [30,31]. Only one p e ious coho , including pa icipan s wi h a high educa ional le el, assessed low- o-mode a e le els o in ake and ound a simila in e se associa ion (HR (95% CI); 0.65 (0.49 o 0.86)) o he same angeo exposu e5 o15g/day ha weha eob- se ed [9]. The esul s o ou s udy and ha p e ious coho a e, he e o e, closely consis en . The p ac ical absence o hea y d inke s in ou sample and in ha co- ho , oge he wi h he la ge p opo ion o pa icipan s exposed o low- o-mode a e in akes a e likely explana- ions o he di e ging esul s wi h espec o coho s including hea y d inke s. The esul s o hose s udies ha e o be ca e ully in e p e ed. In se e al p e ious c oss-sec ional s udies, a s ong posi i e associa ion was ound. The e o e, he de ini ion o a clea empo al se- quence is necessa y o p eclude he possibili y o a e e se causa ion bias [3]. Some longi udinal s udies ha e assessed he e e se di ec ion o his associa ion, ha is, he conse- quences o dep ession on alcohol in ake. They ha e ound di ec associa ions; subjec s wi h dep ession end o in- c ease hei alcohol in ake as a consequence o hei mood diso de [2]. The e o e, in some c oss-sec ional analyses, he posi i e associa ion ound is likely o be he esul o he di ec e ec o dep ession on alcohol in ake, which may o e come he possible in e se associa ion be ween low- o -mode a e alcohol in ake and dep ession isk. They can also cap u e he de imen al e ec s o highe amoun s o alcohol in ake on dep ession isk. In con as , in a longi u- dinal analysis, he empo al sequence is p ese ed, and e- e se causa ion bias is less likely. In o de o minimize he isk o ha bias, we assumed an induc ion pe iod o a leas one yea and we conduc ed a sensi i i y analysis as- suming an al e na i e induc ion pe iod o wo o h ee yea s in he epea ed measu emen analyses. Ea ly cases may be subclinical cases o dep ession a he beginning o he in e al and may co espond o subjec s who changed hei alcohol in ake be o e de eloping o e clinical dep es- sion. The ac ha he esul s o hese sensi i i y analyses a e consis en wi h he main esul s gi es u he suppo o ou indings. Mo eo e , inciden cases o dep ession did no change hei alcohol in ake di e en ly om hose who we e no inciden cases (P>0.1; da a no shown). The di e ences be ween ou esul s and hose o p e ious s udies could also be explained by he di e en a e age alcohol in ake be ween popula ions, he di e en pa e n Figu e 1 Dose– esponse ela ionship be ween alcohol in ake (g/day) and dep ession isk (HR and 95% CI). Res ic ed cubic spline model. The PREDIMED S udy 2003 o 2010. The black line ep esen s he HR and he dashed lines ep esen he 95% con idence in e al. Adjus ed o age, smoking, physical ac i i y (MET-min/d), o al ene gy in ake (Kcal/day), baseline body mass index (kg/m 2 ), ma i al s a us, in e en ion g oup, ec ui ing cen e , educa ional le el and he numbe o pe sons li ing a home. Gea e al. BMC Medicine 2013, 11:192 Page 5 o 10 h p://www.biomedcen al.com/1741-7015/11/192 o consump ion and especially because he p edominan ype o be e age consumed is di e en , wine being he p e e en ial be e age in ou coho . The PREDIMED ial includes an olde , adi ional Spanish Medi e anean popula ion, ha consumed chie ly wine, and mainly in a con ex o socializa ion wi h amily o iends. Subjec s wi h p oblema ic use o alcohol (≥2 posi i e answe s in he CAGE ques ionnai e) [21] we e explici ly excluded om his ial because his was one o he exclusion c i e ia. Mo eo e , hey we e in a ial whe e an in e en ion wi h a Medi e anean die was done, including he ad ice o consume wine mode a ely only o hose who we e al eady wine d inke s a incep ion. Howe e , he in e en- ion designed o his ial did no a ain any signi ican Table 4 Sensi i i y analyses To al cases/pe son-yea s >5 o 5 g/day O e all 443/23,655 0.72 (0.53 o 0.98) Ene gy limi s: pe cen iles 5 o 95 403/21,681 0.78 (0.57 o 1.06) Ene gy limi s: pe cen iles 1 o 99 444/23,675 0.71 (0.53 o 0.96) Excluding diabe ics 227/11,653 0.96 (0.64 o 1.44) Excluding pa icipan s olde han 75 389/21,200 0.73 (0.56 o 1.25) Excluding pa icipan s younge han 65 307/14,785 0.75 (0.52 o 1.06) Depu a ing abs aine s’g oup a 422/22,882 0.75 (0.55 o 1.02) Including pa icipan s wi h p io dep ession a baseline 817/25,868 0.72 (0.57 o 0.90) Upda ed alcohol in ake assuming induc ion pe iod: 1 o 2 yea s b - 0.69 (0.50 o 0.96) Upda ed alcohol in ake assuming induc ion pe iod: 2 o 3 yea s b - 0.68 (0.47 o 1.00) a A e excluding hose pa icipan s who epo ed any alcohol in ake h oughou hei li e bu no cu en ly a baseline. b Repea ed-measu emen analysis. Rela i e isks (95% con idence in e al) o inciden dep ession acco ding o yea ly upda ed measu emen s o alcohol in ake. Conside ing inciden cases du ing he las yea o e e y wo-yea ollow-up in e al. The PREDIMED S udy 2003 o 2010. All adjus ed o age, sex, smoking, physical ac i i y (MET-min/d), o al ene gy in ake (Kcal/day), baseline body mass index (kg/m 2 ), ma i al s a us, in e en ion g oup, ec ui ing cen e , educa ional le el, he numbe o pe sons li ing a home and alcohol in ake a baseline. Table 3 Risks o dep ession acco ding o weekly wine consump ion Wine: d inks/week Abs aine s <1 1 o <2 2 o 7 >7 Baseline wine consump ion a Cases/Pe son-yea s 195/7,777 44/2,461 28/1,708 57/3,183 32/6,828 Age and sex-adjus ed model 1 (Re .) 1.05 (0.78 o 1.42) 0.82 (0.55 o 1.22) 0.75 (0.54 o 1.05) 0.79 (0.60 o 1.04) Mul iple-adjus ed model c 1 (Re .) 1.00 (0.72 o 1.39) 0.93 (0.61 o 1.43) 0.70 (0.48 o 1.00) 0.78 (0.57 o 1.06) Mul iple-adjus ed model d 1 (Re .) 0.99 (0.71 o 1.37) 0.92 (0.60 o 1.41) 0.68 (0.47 o 0.98) 0.76 (0.56 o 1.04) Upda ed wine consump ion b Age-adjus ed model 1 (Re .) 0.64 (0.46 o 0.89) 0.86 (0.58 o 1.27) 0.59 (0.43 o 0.82) 0.70 (0.48 o 1.02) Mul iple-adjus ed model c 1 (Re .) 0.64 (0.46 o 0.89) 0.87 (0.58 o 1.29) 0.57 (0.41 o 0.80) 0.68 (0.46 o 1.01) Mul iple-adjus ed model d 1 (Re .) 0.63 (0.46 o 0.89) 0.86 (0.58 o 1.28) 0.57 (0.40 o 0.79) 0.66 (0.45 o 0.99) Baseline wine consump ion (Only wine-d inke s) a Cases/Pe son-yea s 195/7,777 40/1,454 13/752 25/1,363 26/2,236 Age-adjus ed model 1 (Re .) 1.13 (0.80 o 1.59) 0.79 (0.45 o 1.38) 0.86 (0.56 o 1.30) 0.59 (0.39 o 0.90) Mul iple-adjus ed model c 1 (Re .) 1.18 (0.81 o 1.72) 0.91 (0.49 o 1.69) 0.81 (0.50 o 1.30) 0.55 (0.35 o 0.87) Upda ed wine consump ion (Only wine-d inke s) b Age-adjus ed model 1 (Re .) 0.54 (0.34 o 0.87) 0.93 (0.52 o 1.66) 0.58 (0.36 o 0.92) 0.52 (0.24 o 1.14) Mul iple-adjus ed model c 1 (Re .) 0.55 (0.34 o 0.87) 0.93 (0.52 o 1.66) 0.57 (0.36 o 0.92) 0.52 (0.23 o 1.18) a Haza d a ios (95% con idence in e als) o inciden dep ession acco ding o ca ego ies o baseline daily alcohol in ake. b Rela i e isks (95% con idence in e als) o ca ego ies o upda ed alcohol in ake, using epea ed measu emen s o die du ing ollow-up. To a oid e e se causali y bias, an induc ion pe iod o a leas one yea , bu no longe han wo yea s was assumed. We conside ed as inciden cases o dep ession hose occu ing only du ing he second yea o e e y wo-yea ollow-up in e al. c Adjus ed o age, smoking, physical ac i i y (MET-min/d), o al ene gy in ake (Kcal/day), baseline body mass index (kg/m 2 ), ma i al s a us, in e en ion g oup, ec ui ing cen e , educa ional le el and he numbe o pe sons li ing a home. d Adding alcohol om sou ces o he han wine o he p e ious mul iple-adjus ed model. The PREDIMED S udy 2003 o 2010. Gea e al. BMC Medicine 2013, 11:192 Page 6 o 10 h p://www.biomedcen al.com/1741-7015/11/192 change in a e age le els o alcohol in ake (−1.4 g/day, −1.4 g/day, and −0.9 g/day in he Medi e anean die + i gin oli e oil, Medi e anean die + nu s and con ol g oups, e- spec i ely) wi hou any be ween-g oup signi ican di e - ence. In addi ion, he Medi e anean die has also been associa ed wi h lowe incidence o dep ession [32]. The e- o e, dummy a iables o he wo ac i e in e en ion g oups we e included in he mul iple-adjus ed model as po en ial con ounde s. The ela ionship be ween alcohol in ake and dep ession is biologically di e en among men and women [20,33,34]. Fo ins ance, alcohol bioa ailabili y is di e en be ween sexes [35], and women seem o be mo e likely o de elop dep ession [36]. The e o e, we p esen ed he analyses s a i- ied by sex despi e he e being no e idence o s a is ical in e ac ion by sex. Besides he sensi i i y analysis men ioned abo e, we conduc ed eigh o he sensi i i y analyses o explo e u he po en ial sou ces o bias. The e e ence ca ego y in he main analysis was he abs aine s’g oup. This g oup may include o me d inke s. The e o e, we excluded om he abs aine s’g oup all pa icipan s who epo ed d inking alcohol e e in he /his pas li e, and he esul s emained simila . Howe e , he esul s became non-signi ican a e he exclusion o p e alen diabe ics a baseline. This ac maybeexplainedbyalosso s a is icalpowe because almos 50% o ou pa icipan s we e diabe ics. Al e na i ely, i migh be a ibu ed o a be e disease classi ica ion among diabe ic han among non-diabe ic pa icipan s. Diabe ics usually a e mo e likely o ha e mo e equen medical isi s due o hei diabe es; he e o e, hey ha e mo e oppo uni ies o ecei e a medical diagnosis o dep ession han non-diabe ic pa ien s. To ou knowledge, wi h he excep ion o he ecen esul s o he Uni e si y o Na a a ollow-up S udy (“Seguimien o Uni e sidad de Na a a”(SUN coho )), no o he la ge p ospec i e coho s udy has p e iously epo ed conclusi e esul s abou he ela ionship be ween speci ic alcoholic be e ages and inciden dep ession. Non-alcoholic componen s o wine can accoun o he in e se associa ion ound in ou s udy. Fo ins ance, ans- es e a ol has been pos ula ed as a neu op o ec i e subs ance [37-40]. P e ious in es iga ions sugges ha he hippocampal complex may play a ole in he de elopmen o majo dep ession [41]. This neu op o ec ion applied o he hippocampus may p e en mode a e wine d inke s om de eloping dep ession. An al e na i e po en ial explana ion o his associa ion can be ha wine-d inke s o mode a e d inke s migh be heal hie in o he aspec s han non-wine d inke s o non-mode a e d inke s, as i has been sugges ed o o he ou comes [42]. In o de o accoun o con ounding by hese ac o s, se e al li es yle a iables, including quali y o he die and se e al indica o s o an o e all heal hy li es yle and heal h consciousness, had been included in he mul iple-adjus ed model. Some impo an s eng hs o he p esen s udy a e i s p ospec i e design, he use o epea ed measu emen s o alcohol in ake du ing ollow-up, he la ge sample size o a e y homogeneous popula ion and he good adjus men o po en ial con ounde s. In addi ion, he high p e alence o low- o-mode a e a e age alcohol in ake wi h almos null p e alence o excessi e d inking oge he wi h he p e e ence o wine consump ion o e ed a unique oppo - uni y o in es iga e he in luence o hese wo pa icula ac o s (ligh - o-mode a e d inking and wine consump ion), while i limi ed ou powe o de ec associa ions be ween hea y d inking and dep ession. Inhe en o nu i ional epidemiology me hods, we ha e o poin ou he possibili y o some deg ee o misclassi i- ca ion in he die a y assessmen me hods. Howe e , we used a FFQ ex ensi ely alida ed in Spain o he die a y assessmen [26,43,44]. In addi ion, he co ela ion coe i- cien in he alida ion s udy was highe o alcohol han o mos o he nu ien s. Mo eo e , i he e is some mis- classi ica ion i will be mo e likely non-di e en ial and, he e o e, he associa ion would p obably be d i en owa ds he null alue. Ano he limi a ion in ou s udy is ha we a e no exclusi ely using a clinical diagnosis o dep ession. P obably we a e achie ing a high speci ici y a he expense o losing sensi i i y. Mo eo e , he e is a possibili y ha pa e ns o alcohol consump ion may be associa ed wi h decisions o seek ca e. I hea y d inke s we e less likely o seek medical ca e, his could esul in he a es o dep ession being unde -es ima ed among hea y d inke s. Conclusions In conclusion, low- o-mode a e o al alcohol in ake and speci ically wine consump ion may educe inciden de- p ession, while hea y d inke s seem o be a highe isk. Fu he coho s udies a e needed o con i m hese esul s. Addi ional iles Addi ional ile 1: Flow cha o pa icipan s: he PREDIMED S udy. Addi ional ile 2: Main analyses s a i ied by sex. Haza d a ios (95% con idence in e als) o inciden dep ession acco ding o ca ego ies o baseline daily alcohol in ake, and Rela i e isks (95% con idence in e als) o inciden dep ession acco ding o ca ego ies o upda ed alcohol in ake, using epea ed measu emen s o die du ing ollow-up, s a i ied by sex. The PREDIMED S udy 2003 o 2010. Addi ional ile 3: Fixed-e ec s models. Odds a ios (95% con idence in e als) o a new inciden episode o dep ession, o eco e y om an episode o dep ession, acco ding o ca ego ies o baseline daily alcohol in ake, using ixed-e ec s models. The PREDIMED S udy 2003 o 2010. Gea e al. BMC Medicine 2013, 11:192 Page 7 o 10 h p://www.biomedcen al.com/1741-7015/11/192 Abb e ia ions AUD: Alcohol use diso de s; CHD: Co ona y hea disease; CI: Con idence in e al; FFQ: Food F equency Ques ionnai e; GEE: Gene alized es ima ing equa ions; HR: Haza d a io; MET: Me abolic equi alen ask; PREDIMED: (P e ención con Die a Medi e ánea) P e en ion wi h Medi e anean Die ; SUN coho : Uni e si y o Na a a ollow-up S udy (“Seguimien o Uni e sidad de Na a a”). Compe ing in e es s D . Es uch epo s se ing on he boa d o and ecei ing lec u e ees om he Resea ch Founda ion on Wine and Nu i ion (FIVIN); se ing on he boa ds o he Bee and Heal h Founda ion and he Eu opean Founda ion o Alcohol Resea ch (ERAB); ecei ing lec u e ees om Ce ece os de España and Sano i-A en is; and ecei ing g an suppo h ough his ins i u ion om No a is. D . Salas-Sal adó epo s se ing on he boa d o and ecei ing g an suppo h ough his ins i u ion om he In e na ional Nu and D ied F ui Council; ecei ing consul ing ees om Danone; and ecei ing g an suppo h ough his ins i u ion om E oski and Nes lé. D . A ós epo s ecei ing paymen o he de elopmen o educa ional p esen a ions om Mena ini and As aZeneca. D . Lamuela-Ra en ós epo s se ing on he boa d o and ecei ing lec u e ees om FIVIN; ecei ing lec u e ees om Ce ece os de España; and ecei ing lec u e ees and a el suppo om PepsiCo. D . Se a-Majem epo s se ing on he boa ds o he Medi e anean Die Founda ion and he Bee and Heal h Founda ion. D . Pin ó epo s se ing on he boa d o and ecei ing g an suppo h ough his ins i u ion om he Residual Risk Reduc ion Ini ia i e (R3i) Founda ion; se ing on he boa d o Omega o ; se ing on he boa d o and ecei ing paymen o he de elopmen o educa ional p esen a ions, as well as g an suppo h ough his ins i u ion, om Fe e ; ecei ing consul ing ees om Abbo Labo a o ies; ecei ing lec u e ees, as well as g an suppo h ough his ins i u ion, om Me ck and Roche; ecei ing lec u e ees om Danone and Es e e; ecei ing paymen o he de elopmen o educa ional p esen a ions om Mena ini; and ecei ing g an suppo h ough his ins i u ion om Sano i-A en is, Kowa, Unile e , Boeh inge Ingelheim, and Ka o Bio. No o he po en ial con lic o in e es ele an o his a icle was epo ed. Au ho s’con ibu ions AG conduc ed he li e a u e e iew, pa icipa ed in he design o he p esen s udy, cleaned he da a, conduc ed he main s a is ical analyses and p epa ed he i s d a o he manusc ip . J-JB pa icipa ed in he s a is ical analyses plan and in he design o he p esen s udy, and e ised he manusc ip . RE ini ia ed he collabo a i e p ojec , designed da a collec ion ools, moni o ed da a collec ion o he whole ial, e ised he manusc ip and con ibu ed o he in e p e a ion o indings. AS-V conduc ed pa o he li e a u e e iew, cleaned and analyzed da a, pa icipa ed in he design o he p esen s udy, e ised he manusc ip and con ibu ed o he in e p e a ion o indings. JS-S implemen ed he ial in Reus, moni o ed he da a collec ion and e ised he manusc ip . PB-C pa icipa ed in he implemen a ion o he ial in Pamplona, pa icipa ed in he design o he da a collec ion ools and e ised he manusc ip . EG-G implemen ed he ial in Málaga, moni o ed he da a collec ion and e ised he manusc ip . M-IC implemen ed he ial in Ba celona, moni o ed he da a collec ion and e ised he manusc ip . DC implemen ed he ial in Valencia, moni o ed he da a collec ion, and e ised he manusc ip . MF implemen ed he ial in Palma de Mallo ca, moni o ed he da a collec ion and e ised he manusc ip . FA implemen ed he ial in Vi o ia, moni o ed he da a collec ion and e ised he manusc ip . JL implemen ed he ial in Se illa, moni o ed he da a collec ion and e ised he manusc ip . R-ML-R implemen ed he ial in Ba celona, moni o ed he da a collec ion and e ised he manusc ip . JW pa icipa ed in he design o he p esen s udy, moni o ed he da a collec ion and e ised he manusc ip . XP implemen ed he ial in Ba celona, moni o ed he da a collec ion and e ised he manusc ip . LS-M implemen ed he ial in Las Palmas de G an Cana ia, moni o ed he da a collec ion and e ised he manusc ip . M-AM-G ini ia ed he collabo a i e p ojec , designed da a collec ion ools, moni o ed da a collec ion o he whole ial, implemen ed he ial in Pamplona, supe ised all he s eps in he s a is ical analyses and p epa a ion o he manusc ip , and con ibu ed o he in e p e a ion o indings. He is he gua an o . All au ho s c i ically e ised he manusc ip o impo an in ellec ual con en and app o ed he inal e sion o be submi ed o publica ion. Acknowledgemen s We hank he o he membe s o he PREDIMED G oup: Uni e si y o Na a a, Depa men o P e en i e Medicine and Public Heal h, Pamplona, Spain: E. Toledo, M. Bes-Ras ollo, A. Sánchez-Tain a, B. Sanjulián, E. Goñi, M. Ma ques, A. Ga cía-A ellano, I. Zazpe, J. Bas e a-Go a i, E. H. Ma ínez-Lapiscina. Uni e si y o Na a a, P ima y Ca e Cen es, Pamplona, Spain: M. Se ano- Ma ínez, J. Díez-Espino, N. O uño, N. Be ade, V. Ex eme a-U abayen, C. A oyo- Azpa, L Ga cía-Pé ez, J. Villanue a Telle ía, F. Co és Ugalde, T. Sag edo A ce, Mª D. Ga cía de la Noceda Mon oy, Mª D. Viga a López, Mª T. A ceiz Campo, A.U asunSampe ,MªV.Gue oRubioandB.Chu ioBe aza. Uni e si y o Na a a, School o Pha macy, Pamplona, Spain: J. A. Ma inez, A. Ma í. Hospi al Clinic, Ins i u d’In es igacions Biomèdiques Augus Pi i Sunye , Ba celona, Spain: M. Se a, A. Pé ez-He as, C. Viñas, R. Casas, L. de San ama ía, S. Rome o, J. M. Baena, M. Ga cía, M. Olle , J. Ama , I. Duaso, Y. Ga cía, C. Iglesias, C. Simón, Ll. Quinza os, Ll. Pa a, M. Li oz, J. Bena en , J. Clos, I. Pla, M. Amo ós, M. T. Bone , M. T. Ma in, M. S. Sánchez, J. Al i uba, E. Manzano, A. Al és, M. Co án, C. Valls, A. Sala-Vila and M. Doménech. Uni e si y Ro i a i Vi gili, Reus, Spain: M. Bulló, R. González, C. Molina, F. Má quez, N. Babio, M. So li, J. Ga cía Roselló, F. Ma in, R. To , A. Isach, B. Cos a, J. J. Cab é and J. Fe nández-Balla . Ins i u e de Rece ca Hospi al del Ma , Ba celona, Spain: S. Tello, J. Vila, M. Fi ó, H. Sch öde , R. De la To e, D. Muñoz-Aguayo, R. Elosúa, J. Ma uga and M. Fe e . Uni e si y o Valencia, Valencia, Spain: P. Ca asco, R. Osma, M. Guillén, P. Guillem-Saiz, O. Po olés, V. Pascual, C. Rie a, J. Valde ama, A. Se ano, E. Láza o, A. Sanma ín, A. Gi bés, V. San ama ía, C. Sánchez, Z. Plá, E. Sánchez, C. O ega-Azo ín, J. I. González and C Saiz. Uni e si y Hospi al o Ala a, Vi o ia, Spain: I. Sala e ía, T. del Hie o, J. Algo a, S. F ancisco, A. Alonso, J. San Vicen e, E. Sanz, I. Felipe, A. Alonso Gómez and A. Loma-Oso io. Uni e si y o Málaga, Málaga, Spain: R. Bení ez Pon , M. Bianchi Alba, J. Fe nández-C ehue Na ajas, R. Gómez-Huelgas, J. Ma ínez-González, V. Velasco Ga cía, J. de Diego Salas, A. Baca Oso io, J. Gil Za zosa, J. J. Sánchez Luque and E. Va gas López. Ins i u o de la G asa, Consejo Supe io de In es igaciones Cien í icas, Se illa, Spain: J. Sánchez Pe ona, E. Mon e o Rome o, A. Ma ín Rod íguez and V. Ma ín A anz. Ins i u e o Heal h Sciences IUNICS, Uni e si y o Balea ic Islands, and Hospi al Son Espases, Palma de Mallo ca, Spain: M. Ga cía-Valdueza, M. Moñino, A. P oenza, R. P ie o, G. F on e a, M. Gina d, F. Fiol, A. Jo e and J. Ga cía. Depa men o Family Medicine, P ima y Ca e Di ision o Se illa, Se illa, Spain: M. Leal, E. Ma ínez, J. M. San os, M. O ega-Cal o, P. Román, F. José Ga cía, P. Iglesias, Y. Co chado, E. Mayo al and C. Lama. School o Pha macy, Uni e si y o Ba celona, Ba celona, Spain: M. C. López- Saba e , A. I. Cas ello e-Ba gallo, A. Medina-Remón and A. T esse a-Rimbau. Uni e si y o Las Palmas de G an Cana ia, Las Palmas, Spain: J. Ál a ez-Pé ez, E. Díez Bení ez, I. Bau is a Cas año, I. Maldonado Díaz, F. Sa miendo de la Fe, C. Ruano, M. J. He nández, B. Macías Gu ié ez, P. Hen íquez and I. Cas o. Hospi al Uni e si a io de Bell i ge, Hospi ale de Llob ega , Ba celona, Spain: E. de la C uz, A. Gale a, Y. Sole , F. T ías, I. Sa asa, E. Pad es and E. Co bella. P ima y Ca e Di ision, Ca alan Ins i u e o Heal h, Ba celona, Spain: C. Cabezas, E. Vinyoles, M. A. Ro i a, L. Ga cía, G. Flo es, J. M. Ve dú, P. Baby, A. Ramos, L. Mengual, P. Rou a, M. C. Yus e, A. Gua ne , A. Ro i a, M.I. San ama ía, M. Ma a, C. de Juan and A. B au. O he in es iga o s o he PREDIMED ne wo k: M. T. Mi ja ila (Uni e si y o Ba celona), M. P. Po illo (Uni e si y o Basque Coun y), G. Sáez (Uni e si y o Valencia) and J. Tu (Uni e si y o Balea ic Islands). AG hanks he Spanish Go e nmen o he FPU ellowship ecei ed. Au ho de ails 1 Depa men o P e en i e Medicine and Public Heal h, Medical School- Clinica Uni e sidad de Na a a, Pamplona, Spain. 2 School o Medicine, Uni e sidad Eu opea de Mad id, Mad id, Spain. 3 CIBER Fisiopa ología de la Obesidad y Nu ición (CIBERObn), Ins i u o de Salud Ca los III, Mad id, Spain. 4 Depa men o In e nal Medicine, Hospi al Clinic, Uni e si y o Ba celona, Ba celona, Spain. 5 Depa men o Clinical Sciences, Uni e si y o Las Palmas de G an Cana ia, Las Palmas de G an Cana ia, Spain. 6 Human Nu i ion Uni , IISPV, Uni e si a Ro i a i Vi gili, Reus, Spain. 7 Spain P ima y Ca e, Se icio Gea e al. BMC Medicine 2013, 11:192 Page 8 o 10 h p://www.biomedcen al.com/1741-7015/11/192