Accelerated biological aging six decades after prenatal famine exposure

To test the hyp othesi s that ea rly-lif e adversity a ccele rates th e pace o f biol ogical aging, we analyzed data from the Dutch Hunger Winter Families Study (DHWFS, N=95 1). DHW FS is a natural-experiment bir th-cohort s tud y of surv ivors o f in-utero exposure to famine conditi ons caused by the G erman occupation of the Western Neth erlands in Winter 1 944-5 , matched controls, and their s iblings. We condu cted DNA methylation ana lysis o f blo od samples collec ted when the survivo rs were aged 58 to q uantify bi ological aging using the Dun edinPACE , GrimA ge, and PhenoAge epigenetic clocks. Famine surviv ors had fas ter Duned inPACE , as compared with controls. This e ffec t was strongest a mong wo men. Results were similar fo r GrimAge, although effec t-sizes were smaller. We obse rv ed no diff erence s in PhenoAge between survivor s and controls. Famine ef fects wer e not ac c ounted fo r by blood-ce ll composit ion and were similar fo r individuals exposed ear ly and lat er in gestation. Findings suggest in-utero u ndernutri tion may accelerat e biol ogical aging in later life .

early-life adv ersi ty, biological aging, natural exper iment, prenatal exposu re , famine/undernutrit ion . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 3 Significance Statement Environmental condit ions dur ing gestation are hyp othesi zed t o shape heal t h across the life course. The Dutch Hunger Winter, a f amine caused by a German blockade of the Western Netherlands in late 1944 and ended by the all ied li berati on of the Neth erla nds in Spring 1945, has been studied as a “natural exp eri ment” in which the t iming of a child’s concepti on determined the ir exposur e to se vere under-nutriti on during gestation. We applied thi s natural- experiment design to t est e ffec ts of i n-utero adversi ty on midli fe bio logical aging, as measured by epigenetic clock s. We found that i ndividuals with in-utero famine expos ure had a faste r pace of biologica l aging six decades late r. The environmental c ondit ions surr oun ding pregnancy have potentia l to shape aging trajec torie s f or the next genera tion. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 4

Insults to ea rly-lif e devel opment are predicted by theo ry to impact trajec t ories of healthy aging (1-4). Consistent with t his hypothes is, longitudina l observa ti onal studies have identif ied assoc iation s between ea rly -life condi tions and later-li fe healt h o utcomes (5). But establishi ng the causali ty of su ch asso ciations is dif ficu lt due to poten tial co nfounding effe cts of family histor y and othe r fact ors tha t may affect both early-li fe devel opment and later aging outcomes (3, 6). Natural experiments , which seek t o over come this obstacl e to causal inf erence , are study des igns that tak e advantage of his torica l event s beyond the cont rol of individua ls or their famil ies that impact a subset o f otherwise comparable individua ls in a population. An established na tural-experiment des ign for inve stigating eff ects o f earl y-life adversity on late r- life hea lth is in-utero exposu re to famine (7). In studie s of th e Dutch Hunger Winter (1944-5), Siege of Leningrad (1941-4), Holodomor famine in Ukraine (1932-3), and Great Chinese Famine (1959-6 1), survivors of in-ute ro famin e exposure exhib it higher bu rdens of multiple aging- related d iseases and have sh orte r li fe spans as compared to unexposed ind ividuals bo rn befo re or after famine or in adjac ent, unaffe cted regions (8-13). Within the Fetal Origins and Developmental Or igins of Health and Disease li teratu res, these observat ion s are oft en interpre ted as r efle cting in-utero p ro gram ming of risk for car diovascu lar a nd metabolic diseas e later in lif e (4, 14, 15). However, an al ternativ e hypoth esis is that famine-in duced insul t in ear ly life impairs the deve lopment of more general robustness and resi lience ca pacities of the organism, resulting in accelera ted sys temic decline with aging. To explore this a lterna tive h ypothes is , we analyzed blood DNA methylat ion (DNAm) data colle cted in a natural-experimen t study o f in-utero famine exposure t o test di ffe rences in . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 5 the pace and progress of bio logical aging between famine-survivors and matched cont rols. The Dutch Hunger Winter Families Study ( DHW FS) enrolled a coho rt of survivo r s of in-utero exposure to the Dutch Famine (1944- 5), matched controls bo rn befo re or a fter th e famine in the same hospitals as the surv ivors, a nd their same-sex siblings (16). We compared famine survivors to unexposed contr ols on th ree DNAm measures of biological agi ng linked in pr ior studies with h isto ries o f early-li fe adv ersity, the DunedinPACE, Grim Age , a nd PhenoAge DNA m clocks (17-19). Our analysis furth er ex plored di ffer ences in the e ffec ts of fa mine between women and men and by gestational ti ming of exposure, and tested co nsist ency of findings in both between- and within-family co mparisons.

We analyzed data for N=951 cohort members with available DNAm data (N=487 famine survivors, N=159 time cont rols, N=30 5 sibling contro ls; Table 1 ). The characteris tics o f thi s analysis sample were similar to the D utch Hunger Winter Families Study in terview sample ( SI Appendix , Table S1 ). Our analysis p roceeded in thr ee step s . First, to tes t the hyp othesi s that in-utero famine exposure cont ribut ed to acc elera ted biological aging, we compared DNAm measures of pace of aging (DunedinPA CE) and biological age (GrimAge and PhenoAge) between famine-survivors and controls. Sec ond, we conducted dose-response analysi s to test i f part i cipants who were exposed to th e famine for more week s of gestation exhibit ed larger famine effec ts as compared with those expos ed for fewer weeks of gestation. Third, to explore sp ecif ic ity of famine eff ects . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 6 to exposure du ring specif ic per iods o f gestation, we classi fied famine sur viv ors accord ing to when in gestation th ey were expos ed , as described in the Meth ods sec tion , and computed effec t estimates for each wind ow of e xposure. In each step, we conduc ted analysis (a) in the full DHW FS; (b) using a between-families comparison of famine-survivo rs to un exposed time controls born immediately be fore or after th e famine; and (c) us ing a withi n-family comparison of famine-survivors to the ir unexpose d same-sex siblings. We also explored whether associations of famine exposu re with biological aging diffe red between me n and women. In-utero famine exposure was associated with faster biological aging as measured by DunedinPACE. Cohort members expo sed to famine in u tero had faster pac e of aging compared with unexposed coho rt members (Du nedinPACE β=0.1 5, 95 % CI [0.0 3, 0.2 8 ], p =.018). Differ ences be tween famine survi vor s and contro ls were of smaller magnitude for Gr imAge and PhenoAge and not statistica lly di ffe re nt from zero a t the a lpha=0.05 level ( βs.099).

were similar in ana lysis r estr ic ted to in clude on ly famine surv ivors a nd unrelated time controls. Results are shown in Fig. 1 A and reported in SI Appendix , Table S2 ; full resu lts fo r all biological aging measures are repor te d in SI Appendix , Table S3 ; a correlati on matrix is shown in SI Appendix , Fig. S1 . Longer prenatal famine exposure was associated with faster biological aging as measured by DunedinPACE. In dose- response analys is, cohor t members w ho were exposed t o the famine for more wee ks of gestat i on had faster b iological aging (per 10-weeks of exposu re DunedinPACE β=0.08 , 95 % CI [0.02 , 0. 14], p =.01 3). There were no dose res ponse eff ects for GrimAge and PhenoAge (βs=0. 04, p>. 160). Results are rep orted in SI Appendix , Table S4 ; full

for al l biol ogical aging measur es are repo rted in SI Appendix , Table S5 . . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 7 No timing-specificity for famine exposure in predicting biological aging as measured by DunedinPACE . The effects of in-utero famine exposure may vary depend in g on when in gestation famine exposur e occurs. W e estimated assoc iations of famine ex posure with biological aging at each of s ix time wi ndows from the precon cepti on peri o d through the end o f gestation. Because cohort members could be exposed in multiple time win dows, we included indicator variabl es for exposure in ea ch time window in the same regressi on. Effect-sizes for DunedinPACE ran ged from -0. 01 to 0. 18 and were somewhat larger for late r gestational exposure windows. Effect-si zes fo r Gr imA ge ranged from -0. 08 to 0.12. Effect-sizes fo r PhenoAge ranged from -0.1 9 to 0.28. For GrimAge and Pheno Age, there w ere no gestati onal timing patterns in e ffec t-sizes. Effe ct-sizes are repor ted in SI Appendix , Table S6. Sex differences in associations of in-utero famine exposure with biological aging. We conducted explora tory ana lysis o f sex diffe rences in famine ef fects u sin g sex-stratified regressions and analys is of effe ct-measure modificat ion. In strat ifi ed analy sis, famine effec ts were consis tentl y larger f or women and were near zero for men ( Fig. 1 B; SI Appendix , Table S2). Findings from sex-stratified d ose-resp onse analysis sh owed similar resu lts ( SI Appendix , Table S4). In sex-stratified analys is of gestat ional timing, results wer e dif ferent fo r women and men ( SI Appendix , Table S6 ). For wo men, DunedinPACE effect-sizes were simila r across gestat ional time windows (effec t-sizes ranged f ro m β=0. 07- 0.25). In contrast, for men, DunedinPACE effect- sizes were largest for late r-gestationa l exposures and smaller for exposure in early gestat ion (effec t-sizes ranged f rom β=-0. 16- 0.1 9). This pattern was similar f or GrimAge. There was no consisten t patte rn fo r PhenoAge. Results are shown in Fig. 2 . Formal tests of eff ect-measure . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 8 modification found sex di ffe rences w ere stat istica lly d iffe rent from zero at the p<0.05 level for DunedinPACE and Grim Age , but not PhenoAge. Sibling-comparison analysis. Finally, we repeated our ana lysis us ing a sibli ng- comparison design. This design holds constant al l fact ors tha t are shar ed b y siblings in a family. In the contex t of th e famine natura l e xperiment, the sibl ing comparison de sign aims to rule out the possib ilit y that d iff erences b etwe en exposed and unexposed individua l s refle ct di ffer ences between families in thei r pre ferenc es and/or abilit y to c onceive and carry t o term a child unde r famine conditi ons. In full-sample sibl i ng comparison analysis (n=227 pairs), famine survivors tended to b e aging faster than t heir u nexposed same-sex siblings; however, effect-size s were attenuated by roughly half as compared with the full-sample analysis and were not sta tist ically differ ent f rom zero. In sex-strati fied a nalysis, dif ference s between s isters di scordant f or famine exposure (n=129 pairs) we re nearl y id entical to famine-effec t est imates fro m our original models whereas, among brothers (n=98 pairs), effec t est imates were near zero or in the opposite d irec tion o f our or iginal anal ysis. Resul ts are r eported in SI Appendix , Table S2 and Table S3 . Sensitivity analysis . We repeated our analysis with add itiona l covar iates f o r estimated proporti ons of whi te blood cell types. In cell-count-adjusted analys is, effe ct -sizes were similar for DunedinPACE, GrimA ge, and Phen oAge ( SI Appendix , Table S7-9 ).

We analyzed blood DNA methylat ion data from partic ipants in a na tural-ex periment study of the Dutch Famine to test the hypothesis that in-ute ro famine expo sure would be . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 9 associated wit h accel erated b iologica l aging over six decades o f fol low-up. We found that survivors of in-uter o famine exposure had faster pac e of bi ological aging as measured by the DunedinPACE clock. However, differe nces in bio logical age measured by th e GrimAge and PhenoAge clocks were smaller and le ss consiste nt. We did not obs erve e vi dence for a timing specifi c eff ect o f famine exposur e on these measures. These find ings were robust to covaria te adjustment for ce ll coun ts and were s imilar in sibl ing-difference analysis, al though estimates were less pre cise. All three o f the DNAm clocks we anal yzed show eviden ce of as sociat ion wi th morbidity and mortality in other studies (20, 21). A prior quasi-experimental anal ysis of early-li fe economic adversi ty found evidenc e o f in-utero-exposure e ffec ts on la ter-lif e biological aging measured by both an earl ier ve rsion of the Duned inPACE clock and the Gri mA ge clock (19). However, only DunedinPACE showed consisten t eviden ce of as sociat ion wi th in-utero famine exposure in t he ful l DHWFS sa mple. It could be tha t DunedinPACE is somewhat more sensitive to precl inical health changes occurr in g in famine survivors as compared with GrimAge. DunedinPACE was developed from an alysis of the ra te of ph ysiol ogical decl ine in midlif e adults (22). It is designed to measure t he Pace of Aging phenotype (23), defin ed a s the rate of dec line in system integri ty. GrimAge, in contr ast, was developed f rom analysis of mortality r isk in mid- late li fe adul ts (24). It is des igned to measure biological age, or the cu rren t level of sys tem integrity. These des ign diffe rences may have consequenc es fo r sensi tivi ty i n the cont ext of midlife fo llow-up of in-utero famine e xposure. Alternat ively, th e eff ect-size differ ences between DunedinPACE and GrimA ge were small and could re flec t stat isti ca l noise. Follow-up in . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 10 other stud ies is n eeded t o clar ify t he significance o f the d iff erence in resu lt s for the two measures. Prior studies suggest that exposure d uring the early p hase of gesta tion may be more impactful (11, 25-27). Our analysis of gestational t iming of famine exposur e did not find evidence for ear lier gesta tion as a se nsitive p eriod. Ov erall, res ults suggest that any in-utero exposure is asso ciated wi th a fast er p ace of bio logical aging six decades la t er. Our analysis o f sex di ffer ences in famine effec ts found larger ef fects of in-u tero famine exposure on DNAm measures of biol ogical aging in women as compared with men. This was observed fo r all three epigenetic cloc ks, but was most pronounced f or the DunedinPACE and GrimAge clocks. In non-human ani ma ls, there is evidenc e that males are more vulnerab le to early-life insults (28). However, prior studies o f in-utero famine exposure o ften repo rted larger famine effec ts on card iovascula r and metabolic diseases among wo men as compared with men (29-32). There is some evidence that select ive fe rti lity and/or feta l loss res ult in f ewer male births dur ing periods o f famine (33). A result coul d be that the subse t of m ale babies born are especiall y robust. This would be cons i stent with o ur resu lts. Bu t a reduc tio n in male births is not documented in the case of the Dutch Famine (34, 3 5). There are not yet an alyses of s ex differ ences in morta lity among survivors of in-ute ro exposure to th e Dutch Famine; the most comprehensive stud y of mortal ity r eli ed on data fr om military consc ripts, who were all men (25). New models in population scien ce suggest that envir onmental condi ti ons, such as in-utero exposure to famine, can induce subs t antial varia tion in sex dif feren ces in s urvival (36). Replicat ion of the obs erved sex d iffe r ence in famine e ffec ts is ne eded. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 11 We acknowledge limitations. There i s no gold standard to measure b iologic al aging (37). We analyzed the DNAm measures of aging with the best availabl e eviden c e for re liabi lity and validity. As new measures are introdu ced, follow-up will be n eeded. However, agreement between dif feren t biologica l aging m easures build confid ence tha t our fin dings do capture aging processes. We were unable to c onduct dose-respons e analysis of fam ine-exposure severit y. Because o f the lack o f family -level nutri tion da ta and the c onsiste ncy of ra tions acr oss the affe cted areas of th e Nethe rlands , analysis of exposu re seve rity wi ll ne ed to be condu cted in diff erent s ett ings, such as where fa mine severity was graded acro ss geographic loca tions o r time (9). Survival bias c ould af fect the population of famine surv ivors a live or in suf fici ently good health to be su rveyed a t fo llow-up. However, characteris tics at birth of the DHWFS participan ts are s imilar to th ose of fa mine-affected bir ths iden tifi ed in hos pital re cords but not successful ly enr olled in the cohort, in cluding birth weight, length, placen ta l weight, maternal age, and birth order (16). Excess deat hs among survivors of in-uter o famine exposure by the time of our s tudy were <10% (13). Therefore, any b ias is l ike ly to b e modest. Moreover, i t is expected tha t health y-participan ts an d survival b iases would bias effe ct-est imates toward the null because n on-participa tion due to ill heal th and death wou ld remove th e most affect ed from the populat ion. Therefo re, our e stimates of famine e ffec ts are expe ct ed to be conservati ve. Finally, becaus e the c oh ort so fa r lacks fol low-up to determin e survival differ ences be tween famine-exposed and control parti cipants, the extent t o which diff erences observed in measures of bio logical aging will transla te int o dif ferenc es in h ealthspan and lifespan r emains to be de termined. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 12 Within the con text o f these limitati on s, our findings prov ide evid ence f or lo ng-term impacts of in-utero famine exposure that may extend to a wide range of a ging -related health outcomes. Now that surv ivors o f in-utero famine exposu re are appr oachin g their ninth d ecade of lif e, furthe r study of famine bi rths i n administrative record data are ne e ded to clar ify the scope of famine effec ts on hea lthspa n and lifespan.

Study setting: The Dutch Hunger Winter of 1944-1945 The Dutch famine was initia ted by a f ood supply embargo imposed by the German occupying forces in early Octobe r 1944. The severity and wide spread natur e of the fa mine are well documented (16, 38, 39). Prior to the embargo, nutrition in the Nethe rland s had generally been adequate. Off icial rations, which eve ntually con sisted of l itt le more than b read and potato es, fell be low 900 kcal/day in la te November 1944, and were as low as 500 kcal/day by April 1945. The macronutrient composit ion of th e ration remained relat ively stable ov er this pe riod, but the compositio n of non-rat ion foods changed, with a reduction in the inta ke of fa t. The famine ceased with l iberat ion in May 1945, after which Al lied f ood suppl ies were d istribu ted. Participants Famine-Exposed individuals were id e ntified from revi ew of arch ival obst et ric rec ords. We selected al l the 2,417 singleton b irths be tween 1 February 1945 and 31 March 1946 at . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 13 three ins titu tions in famine-exposed cities in the wes tern Ne therlands wh ose mothers were exposed to th e famine during or imm ediately p reced ing that pregnancy. Time Controls were sele cted f rom birt hs at the same hospita ls and in the sa me months of the yea r as the famine-exposed group during 1943 and 1947 (two years before and two years after th e famine). We sampled an equal number of bir ths in each month, a llocated a cross t he three ins titu tions ac cording to their s i ze, to obtain 890 singleton bir ths. Of the t otal famine-exposed and time-control bir ths, curren t addresse s wer e able to be traced fo r 70%. These individuals wer e invited by mail to jo in the s tudy and were additiona lly asked if a same-sex sibling born b efor e or afte r the famine would be availa ble to par ticipa te. A total o f N=547 of the famine-exposed group, N=176 of the time-control gro up, and N=308 same-sex unexposed siblings consent ed to part icipat e and underwent a co mputer-assisted structur ed inte rview by telephon e. Data Collect ion was conduct ed in 2003-200 5, approximately six decades af ter the famine. Of the N=1,031 participan ts who were interv iewed, N=971 also participa ted in a c lini c exam ( Fig. 3 ). Following the Helsinki g uidelines, we obta ined eth ical appr ov al both fr om the Instituti onal Rev iew Board o f Columbia Universi ty Medica l Center and fro m the Medical Ethical Com mittee of th e Leiden Un iversi ty Medical Center (LUMC). The study participan ts prov ided verbal consen t in a telephon e inte rvi ew, and in case of cl inical examinatio ns, a written in formed consent was obtain ed and addit ional METC approval for epigenetic s tudies was later con firmed by the METC of the LUMC. Measures . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 14 Famine exposure. We defined the pe riod of famine from archival reco rds of weekly ration distribu tions, as desc ribed pr evious ly (16). Briefl y, the sta rt of the famine-e xposure period was defined as Nov ember 26, 1944 , based on the thr eshold <900 kca l/day of d istribu ted foo d rations. The end of the famine-expos ure period was de fined as May 12, 19 45, one week following the Ge rman surrender. Part icipants’ exposur e during gestation w as determined fr om the date o f thei r mother’s last menstr ual period ( LMP) and their da te of b ir th. In cases where the LMP date was missing or implausible (12% of birt hs), LMP was estimated from birth-record data on birth we ight and date of birth using tables of gender, par ity, and bi rth weight speci fic gestational ages from the c ombined birth re cords o f the Amsterdam Midwives School (1948-57) and the Univers ity o f Amsterdam Wil helmina Gasthuis Hospita l (1931-65). Famine-exposed participants exper ie nced an average of 17 weeks o f gesta tion during which ration d istr ibutions were <900kcal/day. Following prior wor k with t he cohort (29, 40), participan ts were c lassif ied as famine exposed during each o f four 10-week gestational per iods on the basis o f rat ion dis tribut ions. F or each indi vidual, average rat ions w ere calcu lated f or each 10-week period of gestat ion and periods with av erage rations <900 kc al/day were classifi ed as famine-exposed. Among the N=547 partic ipants re cruit ed as f amine exposed, N=403 met criteria for exposu re in at least one 10-week gestational period. A further N=82 had LMP dates prior t o the end of th e famine, but fewer than 10 weeks of gesta tional exposu re to ration dis tribu tions <900kca l/day. A fi nal N=62 had LMP dates after the en d of the famine. Gestationa l peri ods for famine-exposed partic ipants and t ime-controls are shown in Fig. 4 . DNA methylation measures of biological aging. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 15 Biological aging is the p rogressive los s of syst em integrity with ad vancing c hronological age (41). Biological aging is thought t o arise f rom an accumulation of c ellu l ar-level changes that progressively und ermine the robustn ess and resil ience capacity of ce lls, tis sues, and organ systems (42-44). While there i s no gol d-standard measure of biologica l aging in humans (37), the curren t-state o f the ar t are algori thms that combine info rmation from dozens or hund reds of DNA methylation (D NAm) marks, c hemical tags on the DNA sequence th at regulate gene expression and ar e known to change with aging (45). These algorithms are often re ferr ed to as “epigenetic cl ocks” (46). We measured biological aging using epigenetic c lo cks computed fr om the exist ing DNA methylation databa se for th e DHWFS. Brief ly, DNA methylation (DNAm) was measured from blood col lect ed at t he clini c exam using the Illumina Infinium Human M ethylat ion 450k BeadChip and prep ro cessed as prev iously described (26). Furth er deta ils are p r ovided in the SI Appendix , Supplementary Methods . Our primary analysis focused on thr e e epigenetic clock s for whi ch valida ti on data across multiple studi es estab lish robu st asso ciations wi th health span and li fespan and sensitiv ity to exposures known t o hasten aging-related health d ecline: the Duned inPAC E clock, which measures pace of aging, and the Gri mA ge and PhenoA ge clocks, which measure biol ogical age. We calculated Dunedi nPACE usin g the R code avai lable on G itHub ( https://github.com/danbelsky/DunedinPAC E ). We calculated high-technic al-reliabili ty “PC” versions o f the G rimAge and PhenoA ge clocks deve loped by the Levi ne La b (47) using the code available f rom GitHub ( https://github .com/MorganLevineLab/P C- Clocks ). The clocks are described in detai l in th e SI Appendix , Supplementary Methods . . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 16 There are many other epigenet ic c loc ks, although none with comparable e vidence o f validity to Duned inPACE , GrimA ge, an d PhenoAge. Most other cloc ks were developed to predi ct differ ences be tween indiv iduals in th eir chron ological age (sometimes ref e rred to as “ firs t- generation clo cks”). For comparison purposes, we report resul ts in the SI Appendix , Table S3 and Table S5 for three o f the b est-kn own first-generation clock s, the Horv ath, Hannum , and Skin & Blood clo cks (46, 48, 49). We a lso repor t resu lts f or the original ve rsi ons of thr ee second- generation clo cks, the Zhang, GrimAg e, and PhenoAge clocks (24, 50, 5 1) . Original vers ions of the cloc ks were computed us ing the methylclock R package (52) and Pyth on code to calcula te GrimAge provided by Ake Lu. Analysis The analysis sample for this stud y was formed from partic ipants in the c lini c examination who provided a bl ood sample from which DNA was extracted and sto red at LU MC. For our analysis, DNA were available fo r N=960 indivi d uals. After quali ty con trols, DNA met hylation da tasets were available for N=951. These indiv iduals formed our ana lysis sample. We used regression analysi s to t est a ssociation s between in-utero famine exposure and DNAm measures of biological aging. F irst- and second-generation ep igeneti c clock values have high correlat ions with ch ronol ogical a ge. For analysis and interp retat ion, th e standard approach is to regress clock values on pa rtic ipa nts’ chronological age valu es and pre dict res idual valu es. These values, often refer red to as “age accelera tion r esiduals”, aim to quan tify the dif feren ce between how much aging a person has actually exp erienc ed rela tive to th e expectat ion based on their chronol ogical age. No residu alizati on was perfo rmed for Dunedin PAC E, which is a rate . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 17 measure and shows only moderate c orrelat ion with ch ronologica l age. To account fo r the non- independence o f measurements take n from siblings, we used generalize d estimating equation (GEE) regressions (53). Our models in cluded covar iates for par ticipan ts’ se x, age, and a ge- squared at the time of th e clin ic exa m. We explored sex di fferen ces in fa mine effects b y repeating analysis wi th inc lusion o f pr oduct te rm testing inte raction betwe en famine exposure and sex. We repeated our anal ysis wi th a contro l group res tric ted to the "ti me controls" born immediately before or aft er the famine. We tested consi stency of re sults i n within-family comparisons of sib lings using sibling-f ixed-effects (FE) regressions (54). We tested th e sensitiv ity o f associa tions b etween fa mine exposure and biol ogical aging to diffe rences between parti cipants in leuk ocyte co mposition of DNA samples by rep eati ng analysis with additional covaria tes fo r DNAm estimates of l eukocy te prop orti ons est imated using the Houseman equations (55). Funding sources This research was suppor ted by Nat io nal Institut e on Aging grants ( R01AG0 66887; R01AG0421 90). D WB is a fell ow of th e Canadian Institute for Advanced R e search Child Bra in Development Networ k. LHL was supported by a NIDI/NIAS/UMCG Fellowship, Royal Netherlands Academy of Sci ences KN AW at the Netherlands Inst itut e for A dvanced Study. MC was supported by th e Swiss National Centre of Competence in Researc h “LIVES - Overcoming vulnerabil ity: Li fe cou rse persp ectiv es ” financed by the Swiss Nationa l Scie nce Foundation (51NF40 -1859 01) and the European Union Horizon 2020 Research and Inn ovation Programme under the Mar ie Sklodowska-Curie Gr ant (801076). LLS was supported by National Inst itute on Aging grant (R00AG056 599). E W T wa s supported by a VENI grant from the Netherlands Organization f or Scien tifi c Resear ch ( 916171 28) and by the Joint Program ming Initiative (JPI HDHL proposal number 655) via ZonMw in The Netherlands (529051023). The funders had no role in s tudy des ign, data collec tion a nd analysis, decis ion to pub lish, or p r eparation o f the manuscript. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 18

1. L. A. Gavrilov, N. S. Gavrilova, Early-lif e programmin g of aging and longevity: The idea of high initial damage load (the HIDL hy pothesis). Ann. N. Y. Acad. Sci. 1019, 496–501 (2004). 2. D. J. Barker, Fetal o rigins of coronary heart disease. BMJ 311 , 171–174 (19 95). 3. P. M. Brakefield et al., What are the e ffects of maternal and p re-adult envi r onments on ageing in humans, and are there less ons from animal models? Mech. Ageing Dev. 126 , 431–438 (200 5). 4. P. D. Gluckman, M. A. Hanson, C. Coo per, K. L. Thornburg, Effect of In u ter o and early- life c onditi ons on adul t health and dis ease. N. Engl. J. Med. 359 , 61–73 (20 08). 5. Y. Ben-Shlomo , R. Cooper, D. Kuh, Th e last two decad es of life course epid emiology, and its rel evance f or res earch on ageing. Int. J. Epidemiol. 45, 973–988 (2016). 6. B. L. De Stavola, R. M. Daniel, Comme ntary: Incorpora ting concept s and methods from causal infe rence in to l ife c ourse ep id emiology. Int. J. Epidemiol. 45, 1006– 1010 (2016). 7. A. Vaiserman, Early-life origin o f adult disease: Evidence from natural expe r iments. Exp. Gerontol. 46, 189–192 (2011). 8. K. Grey et al., Severe malnutri tion o r famine exposure in ch ildhood and car diometabolic non-com municable disease later in li f e: A systematic revi ew. BMJ Glob. Health 6 , e003161 (202 1). 9. C. Li, L. H. Lu mey, Exposure to the Chinese famine of 1959-61 in early lif e a nd long-term health cond itions: A sys tematic rev ie w and meta-analysis. Int. J. Epidemiol. 46, 115 7– 1170 (2017). 10. C. Li, L. H. Lu mey, Early-life exposure to the Chinese famine of 1959-1961 a nd type 2 diabetes in adu lthood: A syst ematic r eview and meta-analysis. Nutrients 14 , 2855 (2022). 11. L. H. Lumey, A. D. Stein, E. Susser, Pre natal famine and adult h ealth. Annu. Rev. Public Health 32, 237–262 (20 11). 12. M. Cheng, N. Som met, M. Kerac, D. S. Jopp, D. Spini, Exposure to the 1959– 1961 Chinese famine and risk of non-communicable diseases in later lif e: A life course perspecti ve. PLOS Glob. Public Health 3 , e00021 61 (2023). 13. P. Eka mper, F. van Poppel, A. D. Stein, L. H. Lu mey, Independent and additi ve association of pre natal famine expos ure and inte rmediary li fe cond itions with adult mortality be tween age 18–63 years. Soc. Sci. Med. 119 , 23 2–239 (20 14). 14. P. D. Gluckman, M. A. Hanson, Living with the past: Evolu tion, deve lopment, and patterns o f disease. Science 305 , 173 3–1736 (2004). 15. R. C. Painter et al., Early onset of co ro nary artery disease a fter p renatal exposure to the Dutch famine. Am. J. Clin. Nutr. 84 , 322-32 7 (2006). 16. L. Lumey et al., Cohort profil e: The Dutch Hunger Winter Families Study. Int. J. Epidemiol. 36, 119 6–1204 (20 07). 17. G. H.-J. Graf et al., Social mobility and biological aging among older adults i n the United States. PNAS Nexus 1 , p gac029 (20 22) . 18. C. McCrory et al., Early li fe advers ity and age acceleration at mid-life and o lder ages indexed using the next-generati on Gr imA ge and Pace of Aging epigenetic clocks. Psychoneuroendocrinology 137 , 1056 43 (2022). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 19 19. L. L. Schmitz, V. Duque, In utero exposure to th e Great D epress ion is r efle c ted in la te-life epigenetic aging signatures. Proc. Natl. Acad. Sci. U.S.A. 119 , e220 85301 19 (2022). 20. E. Bernabeu et al., Ref ining epigeneti c predic tion o f chrono logical and bio l ogical age. Genome Med. 15, 12 (20 23). 21. R. F. Hillary et al., Epigenetic measure s of ageing predict the preva lence an d incidence of leading causes of dea th and diseas e b urden. Clin. Epigenetics 12, 115 (2020 ). 22. D. W. Belsky et al., DunedinPACE, a D NA methylation biomarker o f the pac e of aging. Elife 11, e73420 (20 22). 23. D. W. Belsky et al., Quanti ficat ion of biological aging in young adults. Proc. Natl. Acad. Sci. U.S.A. 112 , E 4104–41 10 (2015). 24. A. T. Lu et al., DNA methylation Gr imAge strongly predicts lifespan and he althspan. Aging 11, 303–327 (201 9). 25. P. Eka mper, F. van Poppel, A. D. Stein, G. E. Bijwaard, L. H . Lumey, Prenatal famine exposure and adult morta lity from ca ncer, cardiovascu lar dis ease, and oth er causes through age 63 years. Am. J. Epidemiol. 181 , 271–27 9 (2015). 26. E. W. Tobi et al., Early gestation as th e crit ical t ime-window for changes in the prenata l environment to a ffec t the adu lt human blood methylome. Int. J. Epidemiol. 44, 121 1– 1223 (2015). 27. E. W. Tobi et al., DNA methylation signatures lin k prena tal famine exposur e to growth and metabolism. Nat. Commun. 5 , 55 92 (2014). 28. J. C. K. Wells, Natural sele ction and se x differ ences in morbid ity and mortal i ty in ear ly life. J. Theor. Biol. 202 , 65–76 (200 0). 29. L. Lumey, A. D. Stein, H. S. Kahn, J. Ro mijn, Lipid profile s in middle-aged me n and women after famine exposure dur ing gestation: The Dutch Hunger Winter Families Study. Am. J. Clin. Nutr. 89, 1737–1 74 3 (2009). 30. J. Wang et al., Exposure to the Chines e famine in chil dhood inc reases t ype 2 diabetes risk in adu lts. J. Nutr. 146 , 2289 -22 95 (2016). 31. C. Yu et al., Victims of Chinese famine in early life have inc reased ri sk of me tabolic syndrome in adulthood. Nutrition 53, 20-25 (2 018). 32. H. Chen, W. N. Ne mbhard, H. G. Stockwell, Sex-specific e ffec ts of fetal exp osure to the 1959- 1961 Chinese famine on risk of adult hyper tension. Matern. Child Health J. 18, 527-5 33 (2014). 33. S. Song , Does famine influence sex ra tio at bi rth? Evidenc e from the 1959-1961 Great Leap Forward Famine in China. Proc. Biol. Sci. 279 , 2883–2890 (2 012). 34. A. D. Stein, P. A. Zybert, L. H. Lumey, Acute undernutr itio n is no t associa te d with excess of females at bi rth in humans: The Dutch Hunger Winter. Proc. Biol. Sci. 271 , S13 8–141 (2004). 35. J. S. Cramer, L. H. Lu mey, Maternal pr econcept ion die t and the sex ratio. Hum. Biol. 82, 103–107 (201 0). 36. V. Iannuzzi, M. G. Bacalini, C. Frances chi, C. Giuliani, The role o f genetics a nd epigenetics in sex dif ferenc es in human survival. Genus 79, 1 (2023). 37. L. Ferrucci et al. , Measuring biological aging in humans: A quest. Aging Cell 19, e1308 0 (2020). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 20 38. Z. Stein, M. Susser, G. Saenger, F. Ma rolla, Famine and Human Development: The Dutch Hunger Winter of 1944-1945 (Oxford University Press, 1975). 39. I. de Zwarte, The Hunger Winter: Fighting Famine in the Occupied Netherlands, 1944– 1945 (Ca mbridge University Press, 2020). 40. A. D. Stein et al., Anthropometric me asures in middle age afte r exposure t o famine during gestation: Evidence from the Dutch famine. Am. J. Clin. Nutr. 85, 869–876 (2007). 41. T. B. L. Kirkwood, Understanding the odd science of aging. Cell 120 , 437–4 47 (2005). 42. C. López-Otín, M. A. Blasco, L. Partridge, M. Serrano, G. Kroemer, The hallmarks of aging. Cell 153 , 1194– 1217 (20 13). 43. C. López-Otín, M. A. Blasco, L. Partridge, M. Serrano, G. Kroemer, Hallmark s of aging: An expanding universe. Cell 186 , 243–27 8 (2023). 44. J. Campisi et al., From discoveries in a geing research to the rapeuti cs for he althy ageing. Nature 571 , 183–192 (2 019). 45. J. Rutledge, H. Oh, T. Wyss-Coray, Me asuring biological age using omics dat a. Nat. Rev. Genet. 23, 715–72 7 (2022). 46. S. Horvath, K. Raj, DN A methylation-based biomarkers and th e epigeneti c c lock th eory of ageing. Nat. Rev. Genet. 19, 371–3 84 (2018). 47. A. T. Hig gins-Chen et al., A computati onal solut ion fo r bols tering rel iabil ity of epigeneti c clocks: Implicati ons fo r clin ical trials a nd longitudinal track ing. Nat. Aging 2 , 644–661 (2022). 48. G. Hannum et al., Genome -wide met hylation p rofi les rev eal quant itati ve v iews of human agin g rates. Mol. Cell 49, 359–367 (2013). 49. S. Horvath, DNA methylation age of h uman tissues and cell types. Genome Biol. 14, R115 (2013). 50. M. E. Levine et al., An epigenetic b io marker of aging for li fespan and heal t hspan. Aging 10, 573–5 91 (2018). 51. Y. Zhang et al., DNA methylation sign atures in pe riphera l blood s trongly pr edict al l-cause mortality. Nat. Commun. 8 , 14617 (2017). 52. D. Pelegí-Sisó, P. de Prado, J. Ronkainen, M. Bustamante, J. R. González, M ethylcl ock: A bioconduct or package to estimate D NA methylation age. Bioinformatics 37 , 1759–1 760 (2020). 53. J. A. Hanley, A. Ne gassa, M. D. d. Edw ardes, J. E. Forrester, Statis tical anal y sis of correla ted data using generalized est i mating equations: An orien tation. Am. J. Epidemiol. 157 , 36 4–375 (200 3). 54. A. H. Petersen, T. Lange , What Is the causal inter pretat ion of sibling comparison designs ? Epidemiology 31, 75–81 (2020). 55. E. A. Houseman et al., DNA methylati on arrays as sur rogate measures of c ell mixture distribu tion. BMC Bioinformatics 13, 86 (2012). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 21 Table 1. Characteristics of the Dutch Hunger Winter Families Study DNA methylation sample. The table shows charact erist ics o f the analysis sample overa ll (l eft column) and the famine- exposed and contr ol groups (middle a nd right columns). D H W FS Fa m i n e - e xp o s e d Tim e co n t r o l s S ib lin g c o n tr o ls ( N = 9 51 ) ( N= 48 7) ( N = 1 59 ) ( N =305 ) M e an / % ( SD ) M e an / % (SD ) M e an / % (S D ) M e a n / % (S D ) A g e ( y e a rs ) 58 ( 4) 59 ( 1 ) 5 9 ( 2 ) 57 ( 6 ) Me n ( % ) 45 % 47 % 4 5% 42 % Du r a t i on of e x pos ur e ( w e e ks ) 1 7 ( 7 ) D u n ed i n P A C E 0 .97 ( 0.11) 0 . 9 7 ( 0 .11) 0.95 ( 0. 11 ) 0. 9 7 ( 0.1 1) PC G r i m A g e 6 9.92 ( 5.03) 7 0.42 ( 4 .20) 70 . 1 8 ( 4. 60 ) 68 . 98 ( 6.2 0) PC P h en oA g e 5 0.18 ( 5.87) 5 0.61 ( 5 .18) 50 . 6 1 ( 5. 10 ) 49 . 27 ( 7.0 7) Con t ro l s P a ne l I : D H W F S s a m pl e ( N = 95 1 ) . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 22 Fig. 1. Differences in biological aging between survivors of in-utero famine exposure and unexposed control participants in the Dutch Hunger Winter Families Study. The figure shows effec t-sizes of in-utero famine exposu re associat ions with three DNA meth ylation (D NAm) measures of biologica l aging, DunedinPAC E, PC Grim Age , and PC PhenoA g e (N=951). Panel A shows effec t-sizes es timated in the fu ll cohor t. Panel B shows ef fect-si zes e stimated for women and men separately. Effec t-sizes wer e estimated fr om generalized es timating equation regressions and are d enominated in s tandard-deviation un its o f the aging measures, interpre table as Cohen`s d values. Err or bars show 95% confidenc e inte rval s. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 23 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 24 Fig. 2. Differences in biological aging between survivors of in-utero famine exposure and unexposed control participants in the Dutch Hunger Winter Families Study by gestational timing of famine exposure. The figure shows effec t-sizes es timated fo r fa mine exposure during six gestational time windows. Famine -exposed participan ts were exposed during up to two periods. The devel opmental period s are ordered in the x-axis in chronol ogical orde r rela tive to the famine. The left-most tic k shows effec t-sizes fo r late-gestati onal expos ure (def ined as exposure fo r the f inal 10 weeks o f gestation; N=139 exposed). The sec ond tick to the left shows effec t-sizes fo r exposure during the p enultimate 10 weeks of gesta tion (N = 146 exposed). The third t ick shows e ffec t-sizes f or expos ure during the se cond 10 weeks of ge station (N =125 exposed). The fourth tic k shows ef fec t-sizes for exposure du ring the f irst 1 0 weeks of gestation (N=74 exposed). The fif th ti ck shows effec t-sizes fo r earl y gestational expo sure with durat ion <10 weeks (N=94 exposed). The right-most tick shows e ffec t-sizes f or prec o nceptual exposu re, i.e., for exposure du ring the per iod pr eceding concept ion (N=52 exposed ). Numbers exposed do not add up to the to tal exposed sample because many parti cipants wer e ex posed in two adjacent period s (N=143). Effect-si zes are repor ted fo r DunedinPACE, PC Gr imA ge, and P C PhenoAge. Effect-sizes were estimate d from a multivariate regression in which indica tor variables f or each expo sure window were included as p redic tor var iables along with covariat es for sex, age, and age-squared. Effect- sizes are d enominated in s tandard-deviation un its of the aging measures, interpretab le as Coh en`s d values. Full resul ts are report e d in SI Appendix , Table S6 . Effect-sizes ar e plot ted sep arately f or women (circ les) and men ( triangles). The figure shows a consisten t sex-specif ic patt er n in DunedinPACE and P C GrimA ge effect-sizes. Women who survived in-uter o famine exposu re, whether at ear ly or later gesta tion , tended to have faster pace of bi ological aging, as me asured by DunedinPACE, and older biological age, as measured by PC GrimA ge DNA m clock. In contrast, men who survived later -gestational famine exposure tended to have faster pace of biologica l aging and older biol ogical age; whereas men who survived ear ly-gestational famin e exposure te nded to hav e slower pa ce of bio logical aging and younger biological age, as measured by DunedinPACE and P C GrimA ge DNAm clock, respecti vely. There was no c onsiste nt pattern in PC PhenoA ge effect-sizes. . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 25 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated biological aging six decades after prenatal famine exposure 26 Fig. 3. Flow diagram of the Dutch Hunger Winter Families Study. The figure shows how the analysis sample size was arrived at in each step for survivors of in-utero famine exposure, time controls, and same-sex sibling controls. N=1,031 participants completed telephone interviews. Of this group, 971 participated in the clinic exam. DNA extracted from blood samples was analyzed to determine DNA methylation and data passed quality controls for N=951 individuals. The figure illustrates the number of individuals in each exposure and control group included in the telephone interview, clinic examination, and analysis sample. 26 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint Accelerated b iological aging six decad es after p renatal famine exposure 27 Fig. 4. Gestational timing of exposure to famine in the Dutch Hunger Winter Families Study. The figure shows individua l gestation s of N=547 famine-exposed partic ipa nts (colo red l ines) and N=176 time controls (gray lines). Each gestation is plo tted as a s ingle h orizon tal lin e. The start of the l ine is the date of th e mother’s last menstrual pe riod (LMP). Th e end of th e line is the parti cipant’s date of bi rth. Individ ual gestations are plot ted fr om the to p of the graph to the bottom, ordered by LMP date. For th e famine-exposed participan ts, the se gment of each line showing the firs t 10 weeks of gestat i on is colo red gold. The segment showing the second 10 weeks is col ored o range. The segmen t showing the thi rd 10 weeks is co lor ed red. The segment showing the last 10 weeks is color ed purple. For the time contr ols, 10-week gestational p eriods are color ed in gray, with lighte r shad es for th e earl ier gestat ional per iods. The x-axis shows the date. The vertica l dashed l ines show t he start and end of th e famine expos ure period (November 26, 1944 - May 12, 1 945). . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted November 4, 2023. ; https://doi.org/10.1101/2023.11.03.23298046doi: medRxiv preprint