Vitamin D and cognitive function: A Mendelian randomisation study

1 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 www.na u e.com/scien i ic epo s Vi amin D and cogni i e unc ion: A Mendelian andomisa ion s udy Jane Maddock1,2, Ang Zhou3, Alana Ca adino2,4, Elżbie a Kuźma5, Yanchun Bao6, Melissa C. Sma 6, Kai-Uwe Saum7, Ben Schö ke 7,8,9, Jo gen Engmann10, Ma ie Kjæ gaa d11,12, Ville Ka hunen13,14, Yiqiang Zhan15, Te ho Leh imäki16, Su i P. Ro io17, Liisa Bybe g18, Ja i Lah i19,20, Ped o Ma ques-Vidal 21, Abhiji Sen22, Lau a Pe na7, Hen ik Schi me 23,24, A chana Singh-Manoux10,25, Juha Au inen13,14, Nina Hu i-Kähönen26, Mika Kähönen27, Lena Kilande 28, Ka i Räikkönen19, Håkan Melhus29, E ik Ingelsson29,30, Id is Guessous31,32,33, Ka ja E Pe o ic34, Helena Schmid 35, Reinhold Schmid 36, Pe e Vollenweide 21, La s Lind29, Johan G. E iksson37,38, Ka l Michaëlsson18, Olli T. Rai aka i17,39, Sa a Hägg15, Nancy L. Pede sen15, Ka l-Heinz He zig40,41,42,43, Ma jo-Rii a Jä elin13,44, Juha Veijola42,45,46, Mika Ki imaki 10,47, Rol Jo de11,12, He mann B enne 7,8, Meena Kuma i6,10, Ch is Powe 2, Da id J. Llewellyn5 & Elina Hyppönen3,2,48 1MRC Li elong Heal h and Ageing a UCL, London, Uni ed Kingdom. 2Popula ion, Policy and P ac ice, UCL G ea O mond S ee Ins i u e o Child Heal h, London, Uni ed Kingdom. 3Cen e o Popula ion Heal h Resea ch, Sansom Ins i u e, Uni e si y o Sou h Aus alia, Adelaide, Aus alia. 4Wol son Ins i u e o P e en i e Medicine, Queen Ma y Uni e si y o London, London, Uni ed Kingdom. 5Uni e si y o Exe e Medical School, Exe e , Uni ed Kingdom. 6Ins i u e o Social and Economic Resea ch (ISER), Uni e si y o Essex, Colches e , Uni ed Kingdom. 7Di ision o Clinical Epidemiology and Aging Resea ch, Ge man Cance Resea ch Cen e (DKFZ), Heidelbe g, Ge many. 8Ne wo k Aging Resea ch (NAR), Uni e si y o Heidelbe g, Heidelbe g, Ge many. 9Ins i u e o Heal h Ca e and Social Sciences, FOM Uni e si y, Essen, Ge many. 10Depa men o Epidemiology and Public Heal h, Uni e si y College London, London, Uni ed Kingdom. 11Di ision o In e nal Medicine, Uni e si y Hospi al o No h No way, T omsø, No way. 12Endoc inology Resea ch G oup, Depa men o Clinical Medicine, Facul y o Heal h Sciences, UiT The A c ic Uni e si y o No way, T omsø, No way. 13Cen e o Li e Cou se Heal h Resea ch, Uni e si y o Oulu, Oulu, Finland. 14Oulu Uni e si y Hospi al, Oulu, Finland. 15Depa men o Medical Epidemiology and Bios a is ics, Ka olinska Ins i u e , S ockholm, Sweden. 16Depa men o Clinical Chemis y, Fimlab Labo a o ies and Facul y o Medicine and Li e Sciences, Uni e si y o Tampe e, Tampe e, Finland. 17Resea ch Cen e o Applied and P e en i e Ca dio ascula Medicine, Uni e si y o Tu ku, Tu ku, Finland. 18Depa men o Su gical Sciences, Uppsala Uni e si y, Uppsala, Sweden. 19Depa men o Psychology and Logopedics, Uni e si y o Helsinki, Helsinki, Finland. 20Helsinki Collegium o Ad anced S udies, Uni e si y o Helsinki, Helsinki, Finland. 21Depa men o Medicine, In e nal Medicine, Lausanne Uni e si y Hospi al, Lausanne, Swi ze land. 22Depa men o Public Heal h and Nu sing, Facul y o Medicine and Heal h Sciences, No wegian Uni e si y o Science and Technology (NTNU), T ondheim, No way. 23Depa men o Clinical Medicine, Facul y o Heal h Sciences, UiT The A c ic Uni e si y o No way, T omsø, No way. 24Depa men o Ca diology, Ake shus Uni e si y Hospi al, Lø enskog, No way. 25INSERM, U1018, Cen e o Resea ch in Epidemiology and Popula ion Heal h, Hôpi al Paul B ousse, Villejui , F ance. 26Depa men o Pedia ics, Tampe e Uni e si y Hospi al and Facul y o Medicine and Li e Sciences, Uni e si y o Tampe e, Tampe e, Finland. 27Depa men o Clinical Physiology, Tampe e Uni e si y Hospi al and Facul y o Medicine and Li e Sciences, Uni e si y o Tampe e, Tampe e, Finland. 28Depa men o Public Heal h and Ca ing Sciences, Uppsala Uni e si y, Uppsala, Sweden. 29Depa men o Medical Sciences, Uppsala Uni e si y, Uppsala, Sweden. 30Depa men o Medicine, Di ision o Ca dio ascula Medicine, S an o d Uni e si y School o Medicine, S an o d, Cali o nia, USA. 31Uni o Popula ion Epidemiology, Di ision o P ima y Ca e Medicine, Depa men o Communi y Medicine, P ima y Ca e and Eme gency Medicine, Gene a Uni e si y Hospi als, Gene a, Swi ze land. 32Depa men o Epidemiology, Rollins School o Public Heal h, Emo y Uni e si y, A lan a, Geo gia, USA. 33Depa men o Ambula o y Ca e and Communi y Medicine, Uni e si y o Lausanne, Lausanne, Swi ze land. 34Di ision o Gene al Neu ology, Depa men o Neu ology, Gene al Hospi al and Medical Uni e si y o G az, G az, Aus ia. 35Resea ch Uni o Gene ic Epidemiology, Ins i u e o Molecula Biology and Biochemis y, Cen e o Molecula Medicine, Medical Uni e si y o G az, G az, Aus ia. 36Depa men o Neu ology, Clinical Di ision o Neu oge ia ics, Medical Uni e si y G az, G az, Aus ia. 37Depa men o Gene al P ac ice and P ima y Heal h Ca e, Uni e si y o Helsinki and Helsinki Uni e si y Hospi al, Uni e si y o Helsinki, Helsinki, Finland. 38Folkhälsan Resea ch Cen e , Helsinki, Finland. 39Depa men o Recei ed: 24 May 2017 Accep ed: 19 Sep embe 2017 Published: xx xx xxxx OPEN www.na u e.com/scien i ic epo s/ 2 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 The causal na u e o he associa ion be ween hypo i aminosis D and poo cogni i e unc ion in mid- o la e -li e is unce ain. Using a Mendelian andomisa ion(MR) app oach, we examined he causal ela ionship be ween 25(OH)D and cogni i e unc ion. Da a came om 172,349 pa icipan s om 17 coho s. DHCR7( s12785878), CYP2R1 s12794714) and hei combined syn hesis sco e we e chosen o p oxy 25(OH)D. Cogni i e es s we e s anda dised in o global and memo y sco es. Analyses we e s a i ied by 25(OH)D e iles, sex and age. Random e ec s me a-analyses assessed associa ions be ween 25(OH)D and cogni i e unc ion. Associa ions o se um 25(OH)D wi h global and memo y- ela ed cogni i e unc ion we e non-linea (lowe cogni i e sco es o bo h low and high 25(OH)D, pcu a u e ≤ 0.006), wi h much o he cu a u e a ibu ed o a single s udy. DHCR7, CYP2R1, and he syn hesis sco e we e associa ed wi h small educ ions in 25(OH)D pe i amin D-dec easing allele. Howe e , coe icien s o associa ions wi h global o memo y- ela ed cogni i e unc ion we e non- signi ican and in opposing di ec ions o DHCR7 and CYP2R1, wi h no o e all associa ion obse ed o he syn hesis sco e. Coe icien s o he syn hesis sco e and global and memo y cogni ion we e simila when s a i ied by 25(OH)D e iles, sex and age. We ound no e idence o se um 25(OH)D concen a ion as a causal ac o o cogni i e pe o mance in mid- o la e li e. The p o ound e ec o an ageing popula ion is e idenced by es ima es sugges ing 65·7 million people wo ldwide will be a ec ed by demen ia by 2030, inc easing o 115·4 million by 20501. The la ency pe iod om he onse o symp- oms o clinical diagnosis is ypically e y long and cogni i e changes can be obse ed decades be o e diagnosis2. So, o he de elopmen o e icien p ima y p e en ion s a egies, i is essen ial o iden i y isk ac o s ha ope a e a he ea ly p e-clinical s age. Hypo i aminosis D is hypo hesised o be one such isk ac o . Low 25-hyd oxy i amin D [25(OH)D] le els a e p e alen in olde indi iduals3, and adul s wi h cogni i e di icul ies ha e been shown o ha e hypo i aminosis D4–6. The e is a biologically plausible link be ween i amin D and cogni i e unc ion. The i amin D ecep o (VDR), i amin D me aboli es and enzymes equi ed o i amin D ac i a ion ha e been ound in he b ain and cen al ne ous sys em7. Addi ionally, expe imen al s udies ha edemons a ed ha ac i e i amin D may in luence b ain and neu on de elopmen 8,and ha e neu op o ec i e po en ial and an ioxidan e ec s7. S udies on VDR knockou mice ha e demons a ed ha hypo i aminosis D may play a ole in accele a ed ageing, beha iou al, social, mo o and senso y de ici s9–12, all o which can con ibu e o cogni i e decline. A numbe o obse a ional s udies ha e linked hypo i aminosis D wi h cogni i e impai men and/o demen ia5,6,13,14. In 2012, a me a-analysis including eigh c oss-sec ional s udies (n = 2,740) ound ha mean sco es on he Mini-Men al S a e Examina ion (MMSE) we e lowe among indi iduals wi h <50 nmol/l compa ed wi h ≥50 nmol/l 25(OH)D5. The au ho s also demons a ed ha 25(OH)D concen a ions we e on a e age 6·2 nmol/l lowe in Alzheime ’s disease pa ien s compa ed wi h con ols (n = 502)5. Ano he me a-analysis in 2012, includ- ing i e c oss-sec ional and wo longi udinal s udies (n = 7,688) sugges ed ha he isk o cogni i e impai men was doubled in pa icipan s wi h low i amin D s a us compa ed o hose wi h no mal le els6. A sys ema ic e iew in 2013 ound ha low i amin D s a us was associa ed wi h wo se cogni i e unc ion o a highe inci- dence o demen ia in 72% o he 25 c oss-sec ional s udies and 67% o he sixp ospec i e s udies included13.In 2017, a me a-analysis including i e longi udinal s udies suppo ed he hypo hesis ha 25(OH)D concen a ions <25 nmol/l may con ibu e o he de elopmen o demen ia14. Resul s om andomised con olled ials (RCTs) ha e no been as p omising15–18. Howe e , null indings om hese ials may be due o a numbe o easons including: sho ollow-up ime16,18, inclusion o younge pa icipan s16, inclusion o sup aphysiological doses18, i amin D being used in combina ion wi h o he subs ances15,17, in low doses17 o , high baseline 25(OH)D s a us15,16. In ligh o hese indings, in 2014 a g oup o in e na ional expe s came o he consensus ha hypo i aminosis D should be conside ed a isk ac o o cogni i e decline and demen ia as i may change he clinical p esen a- ion o demen ia due o accompanying como bidi ies, bu ha 25(OH)D should no be used as a diagnos ic o p ognos ic bioma ke 19. The au ho s concluded ha i amin D supplemen a ion should be pa o he ca e managemen o olde adul s wi h cogni i e diso de s19. Howe e , whe he i amin D plays a causal ole in cogni- i e decline di ec ly o h ough i s impac on como bidi ies, o whe he i is a consequence o cogni i e decline emains unclea . De e mining he na u e o he ue ela ionship be ween i amin D and cogni i e unc ion is challenging due o s udy design issues. Fo ins ance, e en i well-conduc ed, obse a ional s udies may no cap u e all unmeasu ed con ounding and he e is a possibili y o e e se causali y20. RCTs a e he gold s anda d app oach o in e ing a causal associa ion, bu hey also ha e hei limi a ions20. Mendelian Randomisa ion (MR) is an app oach ha uses a gene ic a ian , which is associa ed wi h he exposu e o in e es , o es ima e he causal ela ionship be ween an Clinical Physiology and Nuclea Medicine, Tu ku Uni e si y Hospi al, Tu ku, Finland. 40Resea ch Uni o Biomedicine, Uni e si y o Oulu, Oulu, Finland. 41Biocen e Oulu, Uni e si y o Oulu, Oulu, Finland. 42Medical Resea ch Cen e (MRC) and Oulu Uni e si y Hospi al, Oulu, Finland. 43Depa men o Gas oen e ology and Me abolism, Poznan Uni e si y o Medical Sciences, Poznan, Poland. 44Depa men o Epidemiology and Bios a is ics, MRC–PHE Cen e o En i onmen & Heal h, School o Public Heal h, Impe ial College London, London, UK. 45Resea ch Uni o Clinical Neu oscience, Depa men o Psychia y, Uni e si y o Oulu, Oulu, Finland. 46Depa men o Psychia y, Uni e si y Hospi al o Oulu, Oulu, Finland. 47Clinicum, Uni e si y o Helsinki, Helsinki, Finland. 48Sou h Aus alian Heal h and Medical Resea ch Ins i u e, Adelaide, Aus alia. Co espondence and eques s o ma e ials should be add essed o J.M. (email: [email p o ec ed]) o E.H. (email: [email p o ec ed]) www.na u e.com/scien i ic epo s/ 3 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 exposu e and ou come20. This me hod can help o o e come some limi a ions o obse a ional s udies as i elies on he andom assignmen o gene ic a ian s om pa en s o o sp ing o educe he possibili y o con ounding20. Fu he mo e, since he gene ic a ian is es ablished a concep ion, he possibili y o e e se causali y is minimised20. I hypo i aminosis D is causally ela ed o wo se cogni i e unc ion, he gene ic a ian associa ed wi h i amin D s a us, should be associa ed wi h cogni i e unc ion. Using an MR app oach, a ecen s udy p o ided some suppo o a bene icial ole o 25(OH)D in educing he isk o demen ia21. We conduc ed a s udy o examine he causal na u e o he associa ion be ween i amin D s a us, as measu ed by 25(OH)D, and cogni i e unc ion in mid- o la e -li e using a MR app oach. Me hods Pa icipan s. In o ma ion came om 17 coho s: 1958 B i ish bi h coho (1958BC); Aus ia S oke P e en ion S udy (ASPS); The CoLaus S udy (CoLaus); English Longi udinal S udy o Ageing (ELSA); Epidemiologic s udy assessing p e en ion, ea ly de ec ion, and ea men o ch onic diseases among olde adul s (ESTHER); Helsinki Bi h Coho S udy (HBCS); Heal h and Re i emen S udy (HRS); No he n Finland Bi h Coho 1966 (NFBC1966); The P ospec i e In es iga ion o he Vascula u e in Uppsala Senio s (PIVUS); Swedish Twin Regis y (STR); The T omsø S udy (T omsø); TwinGene; UK Biobank; The UK Household Longi udinal S udy (Unde s anding Socie y, UKHLS); Uppsala Longi udinal S udy o Adul Men (ULSAM); Whi ehall II (WII); and Young Finns S udy (YFS). In all s udies, da a we e es ic ed o Whi e/Eu opean pa icipan s wi h gene ic and cogni i e da a (n = 172,349, Table1). In o ma ion on 25(OH)D concen a ions was a ailable in nine s udies (n = 26,856). All pa icipan s p o ided in o med consen and e hical app o al was g an ed by local esea ch e h- ics commi ees. An expanded desc ip ion is p o ided in supplemen a y ex . Gene ic a ian s. We used wo single nucleo ide polymo phisms (SNPs) based on hei demons a ed asso- cia ions wi h 25(OH)D concen a ions: s12785878 ( i amin D-dec easing allele, G), loca ed nea gene coding 7-dehyd ocholes e ol educ ase (DHCR7), and s12794714 ( i amin D-dec easing allele, A) nea 25-hyd oxylase (CYP2R1)22,23. When hese SNPs we e no accessible, p oxy SNPs in pe ec linkage disequilib ium we e used (Supplemen a y Table1). The assump ions o he use o hese SNPs o p oxy i amin D s a us in MR s udies ha e been assessed in a p e ious s udy using da a om 1958BC24. Da a on DHCR7 and CYP2R1 we e a ailable in all coho s. Fo analysis pu poses, bo h DHCR7 and CYP2R1 geno ypes we e coded as 0-2 depending on p esence o alleles associa ed wi h dec easing 25(OH)D concen a ions, whe e homozygous geno ypes we e coded as 2. A sco e was c ea ed by summing DHCR7 and CYP2R1 on he basis o hei e ec alleles24. The sco e is e e ed o as To al Males Age, y s Age ≥ 65y s 25(OH)D, nmol/l CYP2R1, MAF DHCR7, MAF N = 172,349 (%) (median, IQR) (%) (median, IQR) (%) (%) S udies wi h 25(OH)D (N = 28,070) 1958BC 5,633 49·1 50 (NA) 057·0 (33·9) 43·0 22·2 CoLaus 875 45·1 70 (6) 100 47·3 (33·6) 47·1 27·9 ESTHER 8,080 43·0*74 (4)*100*45·1 (27·1) 46·2 25·7 HBCS 1,059 59·1 67·6 (3·6) 91·8 61 (24) 38·5 38·0 NFBC66 3,488 43·5 46·5 (0·9) 050·2 (20·8) 40·8 38·9 PIVUS 891 50·3 70·1 (0·2) 100 56 (26·5) 39·9 35·0 T omsø 4,766 55·4 69 (11) 76·8 56·5 (25·9) 41·2 38·6 ULSAM 1,118 100 71 (0·9) 100 68·2 (24·7) 39·4 33·4 YFS 2,160 45·1 43 (9) 057 (25) 38·2 40·4 S udies wi hou 25(OH)D (N = 144,279) ASPS 780 43·3 64.7 (11) 49·2 NA 43·7 29·3 ELSA 5,382 45·6 65 (15) 50·6 NA 43·2 22·4 HRS 9,930 41·4 68 (15) 62 NA 43·8 27·0 STR 969 44·7 72.1 (10·1) 77·2 NA 38·8 30·0 TwinGene 2,362 51·1 69 (6) 100 NA 40·0 33·2 UKBiobank 111,936 47·5 58 (12) 19·7 NA 42·2 21·0 UKHLS 8,577 43·7 54 (24) 28·0 NA 42·4 21·6 WII 4,343 76·2 59·7 (10) 28·7 NA 42·5 22·7 Table 1. Pa icipan cha ac e is ics. IQR: In e qua ile ange; MAF: mino allele equency. 1958BC: 1958 B i ish bi h coho ; COLAUS: The Colaus s udy; ESTHER: Epidemiologische S udie zu Chancen de Ve hü ung, F ühe kennung und op imie en The apie ch onische E k ankungen in de äl e en Be ölke ung; HBCS: Helsinki Bi h Coho S udy; NFBC66: No he n Finland bi h coho 1966; PIVUS: The P ospec i e In es iga ion o he Vascula u e in Uppsala Senio s; TROMSO: The T omsø S udy; ULSAM: Uppsala Longi udinal S udy o Adul Men; YFS: Young Finns; ASPS: Aus ia S oke P e en ion S udy; ELSA: English Longi udinal S udy o Ageing; HRS: Heal h and Re i emen S udy; STR: Swedish Twin Regis y; TwinGene: Swedish Twin Regis y; UKBiobank: UK Biobank; UKHLS: The UK Household Longi udinal S udy (Unde s anding Socie y); WII: Whi ehall II * N’s based on pa icipan s wi h cogni i e da a. www.na u e.com/scien i ic epo s/ 4 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 he syn hesis sco e since i con ains he SNPs in ol ed in he syn hesis o 25(OH)D24. The ew pa icipan s wi h 3 o 4 25(OH)D-dec easing alleles ( anging om 9·6% in 1958BC o 17·4% in T omsø) we e g ouped. Geno yping echniques a e desc ibed in supplemen a y ma e ials. Quali y checks o each SNP including mino allele e- quencies and Ha dy Weinbe g equilib ium a e epo ed in Supplemen a y Table1. Mino allele equencies we e compa ed wi h HapMap da a and we e ound o be app oxima ely simila . 25-Hyd oxy i amin D. 25(OH)D concen a ions we e a ailable o nine s udies, wi h de ails o measu e- men me hods p o ided in he supplemen a y ex . In o de o examine analyses s a i ied by 25(OH)D concen- a ions, sex and s udy speci ic 25(OH)D e iles(T) we e c ea ed (Supplemen a y Table2). 25(OH)D was ound o be skewed, he e o e na u al log (ln) ans o ma ion was applied o app oxima e a no mal dis ibu ion when 25(OH)D was he ou come in analyses. Global and memo y cogni i e unc ion. De ails o cogni i e es s in each coho , g ouped o ep esen global and/o memo y cogni i e unc ion, can be ound in supplemen a y ma e ials. Each es was s anda dised o p oduce a mean o ze o and a s anda d de ia ion o one. To ob ain a summa y sco e o global/memo y cogni- i e unc ion, ele an es s we e summed and e-s anda dised. ULSAM and CoLaus had in o ma ion on global cogni i e unc ion only. Co a ia es. Resul s in all s udies we e adjus ed o sex, age (in yea s), mon h o blood collec ion o 25(OH) D, and, educa ion and dep essi e symp oms which we e conside ed po en ial con ounde s a p io i. A desc ip ion o how educa ion and dep essi e symp oms we e measu ed in each coho is ou lined in supplemen a y ma e ials. S a is ical analyses. Wi hin each s udy, linea eg ession models we e used o assess he ollowing: 1) pheno ypic associa ions i.e. se um 25(OH)D and cogni i e unc ion adjus ed o age, sex, mon h o 25(OH) D blood collec ion, educa ional a ainmen and dep essi e symp oms whe e possible; 2) associa ions be ween SNPs/syn hesis sco e and cogni i e unc ion adjus ed o age, sex, educa ion, dep essi e symp oms and p inciple componen s ( o accoun o popula ion s a i ica ion) whe e speci ied in supplemen a y ma e ials. The p esence o non-linea pheno ypic associa ions we e assessed by including a quad a ic e m o 25(OH)D in he eg ession model. In e ac ion by age and sex was also assessed. Resul s om wi hin-s udy analyses we e combined using andom e ec s me a-analyses. Analyses we e s a i- ied by 25(OH)D- e iles, sex and age (<65 yea s s. ≥65 yea s). I-squa e es s we e used o indica e he e ogenei y be ween coho s. Me a- eg ession was used o examine he e ogenei y be ween he coho s using esul s om he me a-analysed pheno ypic analyses. S udy cha ac e is ics ha we e hypo hesised a p io i o a ec he associa ion included sex, age (<65 yea s s. ≥65 yea s), i amin D assay (mass spec ome y o immunoassay) and coun y egion [ca e- go ised as UK, No dic (Finland, Sweden, No way), cen al Eu ope (Aus ia, Ge many, Swi ze land) and he US]. To examine he s eng h and sui abili y o he SNPs/syn hesis sco e as ins umen s o MR s udies, associa- ions be ween he SNPs/syn hesis sco e and ln25(OH)D (adjus ed o age, sex, mon h o 25(OH)D collec ion and s udy-speci ic componen s whe e speci ied) we e examined, and F-s a is ics we e calcula ed. The F-s a is ic was app oxima ed om he p opo ion o a ia ion in he model (R2) assessing 25(OH)D which is explained by he SNPs/syn hesis sco e [F-s a = (R2*(n − 2))/(1 − R2)]. The adjus ed R2 in each coho was weigh ed by he sample size o each coho . The F-s a is ic should ideally be g ea e han en in o de o an ins umen o be conside ed s ong enough o use25. Pa icipan s om UK Biobank we e used o in es iga e cogni i e domain-speci ic e ec s (using s anda dised luid in elligence, pai s ma ching and eac ion ime es s) o he SNPs/syn hesis sco e. Finally, a powe calcula ion was conduc ed o de e mine he smalles e ec size ha ou s udy able o de ec a a wo-sided alpha le el o 0·05 and a a powe o 0.80. Powe analysis was pe o med using Quan o 1.2 (Uni e si y o Sou he n Cali o nia, USA). Me a-analyses we e conduc ed a he Cen e o Popula ion Heal h Resea ch (Uni e si y o Sou h Aus alia) using STATA e sion 14. Resul s A o al o 172,349 pa icipan s had comple e da a o he SNPs and comple ed a leas one cogni i e es . Pa icipan s om eigh coho s had no in o ma ion on 25(OH)D concen a ions, while wo s udies had no in o - ma ion on memo y- ela ed cogni i e unc ion. Basic cha ac e is ics o all s udies a e p esen ed in Table1. Pheno ypic associa ion be ween se um 25(OH)D and cogni i e unc ion. In me a-analyses o he nine eligible s udies, he e was no e idence o in e ac ion by sex o age on pheno ypic associa ions (p ≥ 0·06, Supplemen a y Table3). The e was e idence o a non-linea ela ionship be ween se um 25(OH)D and cogni i e unc ion, p ≤ 0·006 (Supplemen a y Table3) a e adjus men o age, sex, mon h o blood collec ion, educa ional a ainmen and dep essi e symp oms. Howe e , his non-linea associa ion was d i en by a single s udy and weakened ollowing i s’ exclusion om he me a-analyses (Table2). When s a i ied by 25(OH)D e iles, pa ic- ipan s in T2 and T3 had highe sco es in global cogni ion compa ed wi h hose in T1 (0·05 SD, 95%CI 0·01, 0·09; p = 0·02 and 0·07 SD, −0·01, −0·15; p = 0·07, espec i ely), while no clea di e ences we e no seen o memo y cogni ion (P ≥ 0·11 o bo h compa isons). The e we e no o e all di e ences in pheno ypic associa ions by sex, coun y o i amin D assay, while associa ions appea ed somewha s onge among hose aged 65 yea s o abo e compa ed o younge pa icipan s (Supplemen a y Table4). www.na u e.com/scien i ic epo s/ 5 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 Mendelian andomisa ion: associa ion be ween SNPs/syn hesis sco e and cogni i e unc ion. The e we e no associa ions be ween 25(OH)D- ela ed SNPs/syn hesis sco e wi h global o memo y cogni ion (Fig.1, Supplemen a y Table5). Fu he mo e, he e was no e idence o an associa ion be ween SNPs/syn hesis sco e wi h ei he cogni i e measu e s a i ying by sex, age and 25(OH)D e iles (Supplemen a y Figu e1). Gene ic associa ions wi h cogni i e unc ion did no a y by age, sex o coun y (Supplemen a y Table4). In o ma ion o pai s ma ching (n = 110,545), eac ion ime (n = 109,911), easoning (n = 35,603), and p o- spec i e memo y (n = 36,311) was a ailable o he UK Biobank. We conduc ed domain speci ic analyses agains all he ou ou comes, bu obse ed no associa ions wi h he SNPs/syn hesis sco e (Supplemen a y Figu e2). Ins umen alida ion: associa ion be ween SNPs/syn hesis sco e and 25(OH)D. The SNPs/syn hesis sco e we e associa ed wi h 25(OH)D (Fig.2, Supplemen a y Table6). 25(OH)D concen a ions we e 2·7% (95% CI 0·7% o 4·1%), 3·3% (95% CI 0·5% o 4·3%), and 3·1% (95% CI 0·5% o 4·0%) lowe pe Global Cogni ion Memo y Cogni ion NBe a (95% C.I.) p I2(%) phe e o nBe a (95% C.I.) p I2(%) phe e o 25(OH)D T1 4,961 Re e ence 44,482 Re e ence 25(OH)D T2 5,269 0·05 (0·01, 0·09) 0·02 17·48 0·29 44,772 0·04 (−0·01, 0·09) 00·11 34·29 0·17 25(OH)D T3 5,293 0·07 (−0·01, 0·15) 0·07 74·05 <0·001 44,827 0·02 (−0·07, 0·11) 0·64 76·84 <0·001 p end = 0·15 p end = 0·89 p* cu a u e = 0·001 p┼cu a u e = 0·01 Table 2. Associa ion o sex-speci ic 25(OH)D e iles wi h cogni ion. *excluding 1958BC, pcu a u e = 0·04; ┼excluding 1958BC, pcu a u e = 0·16. Figu e 1. Associa ion o CYP2R1, DHCR7 and syn hesis sco e wi h global and memo y cogni ion. Figu e 2. Associa ion o CYP2R1, DHCR7 and syn hesis sco e wi h 25-hyd oxy i amin D (25(OH)D). www.na u e.com/scien i ic epo s/ 6 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 i amin D-dec easing allele o DHCR7, CYP2R1 and he syn hesis sco e espec i ely. The weigh ed F-s a is ic was 54·13 (R2 = 0.003), 71·69 (R2 = 0.004) and 113·50 (R2 = 0.006) o DHCR7, CYP2R1 and he syn hesis sco e espec- i ely. Since hese F-s a is ics a e >10, he SNPs/syn hesis sco e can be conside ed s ong p oxies o 25(OH)D in MR analyses25. Powe calcula ion. Wi h a sample size o 153,187, a powe o 0·80, and a wo-sided alpha le el o 0·05, we a e su icien ly powe ed ode ec a 0·0125 SD change in he cogni i e sco e pe i amin D-dec easing allele. Fo he smalles s a um wi h n = 4,957, ou s udy is powe ed o de ec an e ec size o 0·058 SD o la ge (Supplemen a y Table7). Con e ed o he e lec he associa ion be ween se um 25(OH)D and a s anda dised cogni i e ou come, his co esponds o a 0.09 SD and 0.51 SD di e ence by 1 SD change in log 25(OH)D o he ull and minimum sample, espec i ely. Discussion Using da a om 17 s udies wi h in o ma ion o up o 172,349 pa icipan s, we ailed o ind any e idence o a causal associa ion be ween i amin D and cogni i e unc ion, wi h sub-g oup analyses s a i ying by age, sex and 25(OH)D e iles p o iding a consis en lack o e idence o causali y. These indings sugges ha he non-linea pheno ypic associa ion be ween se um 25(OH)D and cogni i e unc ion (whe e cogni i e sco es we e lowe o bo h low and high 25(OH)D concen a ions), which was also con i med in ou da ase s, may be due o e e se causa ion o con ounding. The e a e a limi ed numbe o RCTs assessing he e ec o i amin D supplemen a ion on cogni i e unc ion15–18. In line wi h ou s udy, hese ha e p o ided li le e idence o causali y. In pa icula , ou MR indings we e sup- po i e o esul s om he bigges RCT o da e which was ca ied ou among olde women (≥65 yea s, n = 4,143) pa icipa ing in he Women’s Heal h Ini ia i e15. In ha RCT, o e a mean ollow-up o 7·8 yea s, he e was no e ec o supplemen a ion wi h 400 IU/day i amin D3 and 1,000 mg/day calcium on cogni i e impai men . Howe e , i has been a gued ha he inclusion o calcium, which can be ha m ul o he b ain, may ha e weak- ened he esul , and baseline concen a ions o 25(OH)D among pa icipan s may ha e been adequa e o mee hei cogni i e equi emen s26. Two p e-pos s udies ha e been conduc ed27,28. One ound no e ec o ou weeks o i amin D2 supplemen a ion on cogni i e unc ion among ins i u ionalised olde adul s27. In con as wi h ou indings, he o he s udy demons a ed an imp o emen in he cogni i e pe o mance among 20 olde adul s a e 16 mon hs o 800 IU/day i amin D3 supplemen a ion compa ed wi h con ols (n = 24)28. Howe e , he non- andom p e-pos design o hese s udies is a limi a ion as po en ial unmeasu ed con ounding canno be uled ou . One ea lie gene ic s udy sugges ed a bene icial e ec o 25(OH)D on educing he isk o Alzheime ’s disease21. Howe e , in his s udy he bene icial associa ion wi h Alzheime ’s disease was due o a signi ican associa ion wi h a a ian in he GC gene coding a ia ions in he i amin D binding p o ein, while in line wi h ou s udy, no associa ions we e seen o DHCR7, o CYP2R1. As we ha e desc ibed ea lie , while associa ed wi h 25(OH) D concen a ions, GC is no sui able o he use as i s’ p oxy ma ke in Mendelian andomisa ion analyses, gi en likely in luences on bioa ailabili y o 25(OH)D29. Indeed, he success o an MR s udy elies upon he abili y o he gene ic a ian o accu a ely p oxy he exposu e o in e es 20. In line wi h p e ious s udies22–24 we used wo a ian s which ha e been consis en ly associa ed wi h ci cula ing 25(OH)D concen a ions. Bo h a ian s a e loca ed ups eam o he 25(OH)D me aboli e, wi h DHCR7 in luencing subs a e a ailabili y and CYP2R1 coding he 25-OH-hyd oxyla e. Analyses s a i ying by 25(OH)D concen a ions sugges ed ha he associa ion be ween CYP2R1 and 25(OH)D is es ic ed o pa icipan s wi h he highes g oup. While his could sugges ha he CYP2R1 a ian has a a e-limi ing e ec on 25(OH)D syn hesis o indi iduals wi h he high 25(OH) D concen a ions, i is also possible ha he appa en di e ence is due o he wide ange o 25(OH)D concen a- ions o indi iduals in he highes e ile in his s udy. S a i ica ion by 25(OH)D e iles could also ha e led o collide bias, i.e. when he associa ion o wo a iables (i.e. gene ic a ian and cogni i e unc ion), changes upon condi ioning on a hi d a iable (i.e. 25(OH)D), when his hi d a iable is a ec ed by he o he wo. The e o e, cau ious in e p e a ion is equi ed when s a i ying MR analyses by 25(OH)D. In e p e ing he associa ion be ween i amin D and cogni i e unc ion is complica ed. Demen ia is o en accompanied by a ange o o he ch onic diseases/diso de s, whe e cogni i e decline may enhance ch onic dis- ease and ice- e sa19. Vi amin D supplemen a ion has been shown o educe mo ali y isk30. Hypo i aminosis D has been associa ed wi h a numbe o condi ions including os eopo osis, ascula disease and educed ol ac o y unc ion31, which can p ecipi a e he p og ession o demen ia. The e o e i is plausible ha he po en ial cogni i e bene i s o i amin D iden i ied in obse a ional s udies may be media ed by imp o emen s in accompanying ch onic diseases. Resul s should be in e p e ed wi h limi a ions in mind. I has been sugges ed ha he e a e sensi i e pe iods i.e. oe al de elopmen , g ow h and senescence du ing which i amin D is o pa icula signi icance o neu ocogni ion19. Ou s udy looked a e ec s on cogni i e unc ion in mid- o la e -li e, and while we ound no e idence s a i ying by age g oup (i.e. <65 e sus ≥65 yea s), we could no assess he associa ion in younge o e y old indi iduals. Ne e heless, he use o gene ic a ian s o p oxy 25(OH)D s a us assumes ha we ha e ep esen ed li e ime 25(OH)D s a us20. Gene alisabili y o he s udy esul s is es ic ed o caucasian popula ions. Cogni ion was assessed using di e en es s be ween he coho s he e o e a composi e measu e o global cogni i e unc ion was used o ob ain a mo e uni o m ep esen a ion o cogni i e unc ion; his may ha e masked some domain speci ic e ec s. Since ou p e ious s udy emphasised he ole o i amin D in cogni i e unc ion, a sepa a e memo y unc ion sco e was c ea ed o accoun o any disc imina ing e ec s o i amin D32. A ecen s udy sugges ed ha i amin D may be associa ed wi h speed o p ocessing and execu i e unc ioning33, we ound no e idence o domain-speci ic e ec s using da a om UK Biobank. MR s udies equi e la ge sample sizes pa ly due o he e y www.na u e.com/scien i ic epo s/ 7 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 small amoun o a ia ion in he exposu e explained by gene ic ins umen s20. Acco ding o he powe calcula- ion, ou s udy was su icien ly powe ed o de ec ela i ely small e ec s. Ne e heless, i should be acknowledged ha ou analyses may ha e been unde powe ed o de ec small causal e ec s ope a ing a he ex emes o 25(OH) D dis ibu ion. Conclusion We ound no e idence o a causal associa ion be ween 25(OH)D concen a ions and cogni i e pe o mance in mid- o la e -li e. Re e ences 1. P ince, M. e al. The global p e alence o demen ia: a sys ema ic e iew and me aanalysis. Alzheime ’s & Demen ia 9, 63–75. e62 (2013). 2. Smi h, A. D. Imaging he p og ession o Alzheime pa hology h ough he b ain. P oceedings o he Na ional Academy o Sciences o he Uni ed S a es o Ame ica 99, 4135–4137, h ps://doi.o g/10.1073/pnas.082107399 (2002). 3. Annweile , C. & Beauche , O. Vi amin D in olde adul s: he need o speci y s anda d alues wi h espec o cogni ion. F on ie s in aging neu oscience 6, 72 (2014). 4. Annweile , C. e al. Vi amin D and cogni i e pe o mance in adul s: a sys ema ic e iew. Eu opean Jou nal o Neu ology 16, 1083–1089 (2009). 5. Balion, C. e al. Vi amin D, cogni ion, and demen ia A sys ema ic e iew and me a-analysis. Neu ology 79, 1397–1405 (2012). 6. E gen, T., Sande , D., Bickel, H., Sande , K. & Fö s l, H. Vi amin D de iciency, cogni i e impai men and demen ia: a sys ema ic e iew and me a-analysis. Demen ia and ge ia ic cogni i e diso de s 33, 297–305 (2012). 7. Ga cion, E., Wion-Ba bo , N., Mon e o-Menei, C. N., Be ge , F. & Wion, D. New clues abou i amin D unc ions in he ne ous sys em. T ends in Endoc inology & Me abolism 13, 100–105 (2002). 8. B own, J., Bianco, J. I., McG a h, J. J. & Eyles, D. W. 1, 25-Dihyd oxy i amin D(3) induces ne e g ow h ac o , p omo es neu i e ou g ow h and inhibi s mi osis in emb yonic a hippocampal neu ons. Neu oscience Le e s 343, 139–143 (2003). 9. Bu ne, T. H., McG a h, J. J., Eyles, D. W. & Mackay-Sim, A. Beha iou al cha ac e iza ion o i amin D ecep o knockou mice. Beha iou al b ain esea ch 157, 299–308 (2005). 10. Kalue , A. V., Lou, Y.-R., Laaksi, I. & Tuohimaa, P. Impai ed mo o pe o mance in mice lacking neu os e oid i amin D ecep o s. B ain esea ch bulle in 64, 25–29 (2004). 11. Z ou, J. e al. P og essi e hea ing loss in mice wi h a mu a ed i amin D ecep o gene. Audiology and Neu o ology 13, 219–230 (2008). 12. Keisa la, T. e al. P ema u e aging in i amin D ecep o mu an mice. The Jou nal o s e oid biochemis y and molecula biology 115, 91–97 (2009). 13. Van de Scha , J. e al. The associa ion be ween i amin D and cogni ion: a sys ema ic e iew. Ageing esea ch e iews 12, 1013–1023 (2013). 14. Somme , I. e al. Vi amin D de iciency as a isk ac o o demen ia: a sys ema ic e iew and me a-analysis. BMC ge ia ics 17, 16 (2017). 15. Rossom, R. C. e al. Calcium and Vi amin D Supplemen a ion and Cogni i e Impai men in he Women’s Heal h Ini ia i e. Jou nal o he Ame ican Ge ia ics Socie y 60, 2197–2205 (2012). 16. Dean, A. J. e al. E ec s o i amin D supplemen a ion on cogni i e and emo ional unc ioning in young adul s–a andomised con olled ial. PloS one 6, e25966 (2011). 17. Mande s, M. e al. The e ec o a nu ien dense d ink on men al and physical unc ion in ins i u ionalized elde ly people. JNHA-The Jou nal o Nu i ion, Heal h and Aging 13, 760–767 (2009). 18. S ein, M. S., Sche e , S. C., Ladd, K. S. & Ha ison, L. C. A andomized con olled ial o high-dose i amin D2 ollowed by in anasal insulin in Alzheime ’s disease. Jou nal o Alzheime ’s Disease 26, 477–484 (2011). 19. Annweile , C. e al. ‘Vi amin D and cogni ion in olde adul s’: upda ed in e na ional ecommenda ions. Jou nal o in e nal medicine 277, 45–57 (2015). 20. Lawlo , D. A., Ha bo d, R. M., S e ne, J. A. C., Timpson, N. & Da ey Smi h, G. Mendelian andomiza ion: using genes as ins umen s o making causal in e ences in epidemiology. S a is ics in Medicine 27, 1133–1163 (2008). 21. Mok y, L. E. e al. Gene ically dec eased i amin D and isk o Alzheime disease. Neu ology 87, 2567–2574 (2016). 22. Wang, T. J. e al. Common gene ic de e minan s o i amin D insu iciency: a genome-wide associa ion s udy. The Lance 376, 180–188 (2010). 23. Ahn, J. e al. Genome-wide associa ion s udy o ci cula ing i amin D le els. Human molecula gene ics, ddq155 (2010). 24. Be y, D. J., Vimaleswa an, K. S., Whi ake , J. C., Hingo ani, A. D. & Hypponen, E. E alua ion o gene ic ma ke s as ins umen s o Mendelian andomiza ion s udies on i amin D. PloS one 7, e37465, h ps://doi.o g/10.1371/jou nal.pone.0037465 (2012). 25. S aige , D. O. & S ock, J. H. (Na ional Bu eau o Economic Resea ch Camb idge, Mass., USA, 1997). 26. Annweile , C. & Beauche , O. Vi amin D and cogni ion: ecommenda ions o u u e ials. Jou nal o he Ame ican Ge ia ics Socie y 61, 1049–1050 (2013). 27. P zyb elsk i, R. e al. Rapid co ec ion o low i amin D s a us in nu sing home esiden s. Os eopo osis in e na ional 19, 1621–1628 (2008). 28. Annweile , C. e al. Cogni i e e ec s o i amin D supplemen a ion in olde ou pa ien s isi ing a memo y clinic: a p e–pos s udy. Jou nal o he Ame ican Ge ia ics Socie y 60, 793–795 (2012). 29. Be y, D. & Hyppönen, E. De e minan s o i amin D s a us: ocus on gene ic a ia ions. Cu en opinion in neph ology and hype ension 20, 331–336 (2011). 30. Chowdhu y, R . e al. Vi amin D and isk o cause speci ic dea h: sys ema ic e iew and me a-analysis o obse a ional coho and andomised in e en ion s udies. (2014). 31. Holick, M. F. Vi amin D de iciency. New England Jou nal o Medicine 357, 266–281 (2007). 32. Maddock, J., Geo oy, M.-C., Powe , C. & Hyppönen, E. 25-Hyd oxy i amin D and cogni i e pe o mance in mid-li e. B i ish Jou nal o Nu i ion 111, 904–914 (2014). 33. Annweile , C. e al. Me a-analysis o memo y and execu i e dys unc ions in ela ion o i amin D. Jou nal o Alzheime ’s Disease 37, 147–171 (2013). Acknowledgemen s P ima y p ojec unding came om The Judi h Jane Mason & Ha old S anne Williams Memo ial Founda ion, o he esou ces a e lis ed in he supplemen a y ma e ial. Funde s played no ole in he design, analysis o epo ing o he s udy. www.na u e.com/scien i ic epo s/ 8 SCIENTIFIC RepoR S | 7: 13230 | DOI:10.1038/s41598-017-13189-3 Au ho Con ibu ions W o e he pape : J.M., E.H. S udy supe ision: E.H., C.P., D.L., H.B., M.Ki., M.Ku., M.R.J., N.L.P., O.T.R., S.H. Da a analysis: J.M., A.Z., A.C., A.S., B.S., E.K., J.E., J.L., K.S., L.B., M.Kj., M.S., O.T.R., P.M.V., S.R., T.L., V.K., Y.B., Y.Z. P o ision o adminis a i e echnical o ma e ial suppo : HSchi, M.Kä., M.Ki., N.H.K., N.L.P., O.T.R., R.J., S.H., S.R., T.L. Da a collec ion: E.H., C.P., A.S.M., E.I., H.B., H.Schi., H.Schm., J.G.E., J.V., K.R., L.L., M.Kä., M.ki., M.R.J., N.H.K., N.L.P., O.T.R., R.J., R.S., S.H., S.R., T.L. In e p e a ion, manusc ip e ision and app o al o submission: J.M., E.H., A.Z., A.C., C.P., D.L., E.K., Y.B. M.S., K.S., J.E., M.Kj., V.K., Y.Z., T.L., S.R., L.B., J.L., P.M.V., A.S., L.P., H.Schi., A.S.M., J.A., N.H.K., M.K., L.K., K.R., H.M., E.I., I.G., K.E.P., H.Schm., R.S., P.V., L.L., J.G.E., K.M., O.T.R., S.H., N.L.P., K.H.H., M.R.J., J.V., M.Ki., R.J., H.B., M.ku. Addi ional In o ma ion Supplemen a y in o ma ion accompanies his pape a h ps://doi.o g/10.1038/s41598-017-13189-3. Compe ing In e es s: EI is a scien i ic ad iso o P ecision Wellness, Cellink and Olink P o eomics o wo k un ela ed o he p esen p ojec . The e a e no decla a ions o in e es s o all o he au ho s. Publishe 's no e: Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps and ins i u ional a ilia ions. Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License, which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as 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 e- a i e Commons license, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his a icle a e included in he a icle’s C ea i e Commons license, unless indica ed o he wise in a c edi line o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons license and you in ended use is no pe - mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission di ec ly om he copy igh holde . To iew a copy o his license, isi h p://c ea i ecommons.o g/licenses/by/4.0/. © The Au ho (s) 2017