10ournal of EpidemiologSy and Community Health 1995;49:10-15

Childhood socioeconomic status and risk ofcardiovascular disease in middle aged USwomen: a prospective study

Michael D Gliksman, Ichiro Kawachi, David Hunter, Graham A Colditz,JoAnn E Manson, Meir J Stampfer, Frank E Speizer, Walter C Willett,Charles H Hennekens

Channing Laboratory,Department ofMedicine, HarvardMedical School andBrigham and Women'sHospital, Boston, USAM D GliksmanI KawachiD HunterG A ColditzJ E MansonM J StampferF E SpeizerW C Willett

Division of PreventiveMedicineJ E MansonC H Hennekens

and the Departmentsof Health and SocialBehaviorI Kawachi

NutritionW C WillettM J Stampfer

and EpidemiologyD HunterG A ColditzM J StampferW C Willett

Harvard School ofPublic Health, Boston,USA

Correspondence to:A/Professor I Kawachi,Channing Laboratory,Harvard Medical School,180 Longwood Avenue,Boston, MA 02115, USA.

Accepted for publicationOctober 1994

AbstractObjective - To examine prospectively therelationship of childhood socioeconomicstatus and risk of cardiovascular diseasein middle aged women.Design - A prospective cohort of womenwith 14 years follow up data (1976-90).Subjects - A total of 117 006 registered fe-male nurses aged 30 to 55 years in 1976 andfree of diagnosed coronary heart disease,stroke, and cancer at baseline.Main outcome measures - Incident fatalcoronary heart disease, non-fatal myo-cardial infarction, and stroke (fatal andnon-fatal).Results - Low socioeconomic status inchildhood was associated with a modestlyincreased risk of incident non-fatal myo-cardial infarction and total cardiovasculardisease in adulthood. Compared withmiddle aged women from white collarchildhood backgrounds, the age adjustedrisk of total cardiovascular disease forwomen from blue collar backgrounds was1-13 (95% CI 1.02,1.24) and that of non-fatal myocardial infarction was 123 (95%CI 1.06,142). No significant increase inrisk was observed for stroke or fatal cor-onary heart disease. Adjustment fordifferences in family and personal pastmedical history, medication use, exercise,alcohol intake, diet, birth weight, beingbreastfed in infancy, and adult socio-economic circumstance somewhat at-tenuated the increased risks observed forwomen from blue collar childhood socio-economic backgrounds. In multivariateanalysis, women whose fathers had beenmanual labourers had the highest relativerisk of total coronary heart disease (RR=1-53; 95% CI 1.09,2 16) and non-fatal myo-cardial infarction (RR= 1-67; 95% CI P1 1,2.53) when compared with women whosefathers had been employed in the pro-fessions.Conclusion - In this group lower childhoodsocioeconomic status was associated witha small but significant increase in the riskof total coronary heart disease as well asnon-fatal myocardial infarction. Forwomen from the most socioeconomicallydisadvantaged childhood backgrounds,the association is not explained by differ-ences in a large number of cardiovascularrisk factors, by differences in adult socio-economic status, or by differences in

indices of nutrition during gestation orinfancy.

(_J Epidemiol Community Health 1995;49: 10-15)

Low socioeconomic status is associated withincreased morbidity and mortality. Thesedifferences are most evident in the very youngas well as those over 65 years of age, especiallyafter the exclusion of occupationally relateddeaths.' 2 Socioeconomic status is also as-sociated with differences in cardiovascular dis-ease incidence3 and coronary heart diseasemortality.4 An inverse relationship betweensocioeconomic status and risk of stroke andcoronary heart disease has been reported inseveral prospective studies.5 Furthermore, thedecline in mortality from cardiovascular diseasein western societies over the past two decadeshas been most evident in higher socioeconomicgroups.6 The mechanisms underlying these as-sociations are not clear but have been the sub-ject of recent research.

Risk factors for cardiovascular disease aremore prevalent among children and adults fromlower socioeconomic backgrounds.'-9 Socialclass differences in diet and physical activity inchildren have been described.410 In addition,previous studies of childhood socioeconomicstatus and risk of cardiovascular disease havefailed to address adequately the problem ofconfounding by achieved adult socioeconomicstatus. For these reasons, it remains unclearwhether childhood status is independentlylinked to risk of cardiovascular disease inadulthood."'The nurses' health study offers a unique

opportunity to address these issues amongwomen. Baseline data allowed the assignmentof childhood socioeconomic status. Variabilityin achieved socioeconomic status of this cohortis less than in the general population. All femalenurses qualify for the same socioeconomic sta-tus rating as adults on the Duncan scale,'2which is independent of marital status andoccupation of spouse.

In this report we assess whether there isa relation between childhood socioeconomicstatus and risk of cardiovascular disease in alarge cohort of US female nurses after ad-justment for a large number of cardiovascularrisk factors, including sociodemographic vari-ables.

MethodsThe nurses' health study cohort was established

10

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from

Childhood socioeconomic status and later heart disease

in 1976 when 121 700 registered female nurses30-55 years of age, residing in 11 US states,completed lifestyle and medical history ques-tionnaires sent by mail. Approximately 98%of the cohort is white.'3 Biennial follow upquestionnaires sent by mail update informationon life events, cardiovascular risk status, cardio-vascular disease, and other health relatedevents. At baseline all women who reportedcoronary heart disease, stroke, or cancer (otherthan non-melanoma skin cancer) were ex-cluded, leaving a cohort of 117006 womenavailable for follow up.

In 1980, a validated, semiquantitative foodfrequency questionnaire as well as questionsassessing physical activity were added to thefollow up and have been updated regularly.'415A total of88 366 nurses free ofprevalent cardio-vascular disease or cancer completed and re-turned the 1980 diet questionnaire.

In the 1992 questionnaire 88 968 par-ticipants completed information on their birthweight, whether they were breastfed as infantsand, if currently married or widowed, theirhusband's highest educational attainment.

EXPOSURE DATAIn 1976 each woman was asked to state theoccupation of her parents when she was aged16 years, in a form that allowed coding ofresponses consistent with the USA census bookcodes for occupations.'6 This classification,which has been used by the Census Bureausince 1940,17 assigns each subject to one ofeight separate categories (listed below). Thesecategories correlate strongly with socio-economic status ratings calculated by the val-idated method devised by Duncan'217 for thepopulation of the USA at a time contemporaryto the childhood and youth of the members ofthe nurses' health study cohort. This methodspecifically correlates with household incomeand occupational prestige, variables which in-dicate educational and other environmental op-portunities available to children.'7The only exception is for the category "farm-

er". The Duncan score is said to underestimatetheir true economic wellbeing which is en-hanced by an active "barter" economy. 12 16 17These categories (listed in descending order byDuncan's socioeconomic score) are as follows:

(1) Professional/technical;(2) Managers/officials/proprietors;(3) Clerical and kindred workers;(4) Skilled blue collar;(5) Semi-skilled blue collar;(6) Domestic and service workers;(7) Farmers/farm workers;(8) Labourers (except farm workers).To increase statistical power while retaining

the validity of the measure of childhood socio-economic status, these categories were con-solidated into three groups - white collar(categories 1, 2, 3), blue collar (categories 4,5, 6, 8), and farmer (category 7). Childhoodsocioeconomic status was assigned on the basisof the father's occupation, a method com-parable with previously published studies.'8 Of

the remainder, the father was either un-employed, retired, or his occupational statuscould not be determined. All women whosefathers were known to be dead by the timeeach woman was 16 years old were placedin the "father deceased" category. This wasdetermined by an algorithm which comparedthe father's date of death with the study par-ticipant's date of birth. Only 1861 women didnot supply sufficient information to enable theirclassification into one of these four categories.

In 1976 and 1984, participants were askedwhether their father or mother had suffered amyocardial infarction and if so, at what age. If aparticipant indicated that either or both parentshad suffered an infarction before age 60 years,the participant was considered to have a posi-tive parental history of early onset myocardialinfarction.

In 1992, participants who were currentlymarried or widowed were asked to indicatetheir husband's highest level of education. Atotal of 6649 women were divorced and werenot asked to complete this question. Educationlevels are highly correlated with socioeconomicstatus.'9 Responses were categorised as no highschool, some high school, high school graduate,college graduate, graduate school. Of the88 968 women who completed the 1992 ques-tionnaire, 74 668 indicated they were eithercurrently married or were widowed. Of these,all but 163 provided information of their hus-band's educational attainment.

In 1976, cigarette smoking was categorisedas never, former, and current, with currentsmokers categorised as 1-14 cigarettes/d,15-24/d, 25 + /d. Information on smoking sta-tus was updated biennially. Questions aboutphysician-diagnosed hypertension, hyper-cholesterolaemia, and diabetes were also firstasked in 1976 and updated biennially.

Also in 1976, weight was recorded and thenupdated in every biennial questionnaire. In avalidation study among 184 cohort membersliving in the Boston area, these self reportedweights correlated highly with those measuredby a technician.20 Height was taken from the1976 questionnaire. Body mass index was cal-culated as weight (in kg) divided by the height(in m2) and was categorised into quintiles.

In 1980, a history of regular aspirin use wasobtained and updated biennially. We calculatedthe average number of tablets or capsules takenper week and categorised aspirin use as <1aspirin per wk, 1-6/wk, >6/wk. In 1976, in-formation on postmenopausal hormone andoral contraceptive use was collected and up-dated biennially. Hormone replacement ther-apy and oral contraceptive use were categorisedas current, past, or never.The 1980 questionnaire inquired about the

frequency of vigorous exercise. Activity levelswere assessed by the number of "sweat in-ducing" exercise sessions per week. Thismethod has been validated in several stud-ies.21-23 We divided the cohort into those womenwho exercised vigorously at least once per week,and those who did not.Also in 1980, alcohol intake was assessed in

the semiquantitative food frequency ques-

I1I

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from

Gliksman, Kawachi, Hunter, Colditz, Manson, Stampfer, Speizer, Willett, Hennekens

tionnaire. Alcohol consumption was updatedin 1984 and then biennially. Daily alcohol con-sumption was categorised as nil, 1-4-9 g/d,5-14-9 g/d, >14-9 g/d. In this cohort the validityof self reported alcohol intake is high.2425

Dietary intake of trans fatty acids is as-sociated with the risk of incident coronary heartdisease.26 For the current analysis, the energyadjusted dietary intake of trans fatty acids wasassessed and divided into quintiles. Vitamin Eintake is also associated with the risk ofincidentcoronary heart disease.27 As much of the ob-served benefit was attributable to oral vitaminE supplementation, quantification of intake forthe current analysis included oral supplements.Vitamin E intake was categorised by quintiles.

In the 1992 questionnaire, participants re-ported whether they had been breastfed asinfants. Responses fell into one ofthe categories"yes", "no", or "don't know". Of the 87917women who answered this question, 14586indicated they "don't know". In the same ques-tionnaire, they were also asked to indicate iftheir birth weight (in lb) fell within the cat-egories: not sure; <5; 5 to 5-5; 5-5 + to 7; 7 +to 8-5; 8-5 + to 10; 10 +. Ofthe 87314 womenwho answered this question, 16 924 indicatedthey were "not sure".

ASCERTAINMENT OF CARDIOVASCULAROUTCOMESReports of non-fatal and fatal cardiovascularoutcomes were assessed by standard methods28which included physician review of medicalrecords. A diagnosis of myocardial infarctionwas confirmed if the relevant World HealthOrganization criteria were met.29These requirethe presence of symptoms and either typicalelectrocardiograph changes or elevation of car-diac enzymes. Medical records to validate re-ports of coronary artery bypass grafting wereobtained for events dating from 1978 on-wards.30A diagnosis of stroke was confirmed if the

event was characterised by a typical neuro-logical defect of rapid or sudden onset, lastingat least 24 hours and attributable to a cerebro-vascular event which was not secondary toinfection or neoplasia. Subdural haematomaswere excluded from analysis. Total non-fatal

Table 1 Age standardised relative risks (95% CI) of cardiovascular diseases bychildhood socioeconomic status 1976-90

Whzite collar Blue collar Farmer Father deceased*

No of women 41 225 54 396 9764 9760Person years exposure 563 577 742 515 133 051 131 024Cardiovascular diseaset:

Total RR(CI) 1-00 1-13(1-03,1-24) 0-89(0-75,1-06) 1-23(1-05,1-44)Non-fatal 1-00 1-17(1-04,1-31) 0-92(0-76,1-12) 1-28(1-08,1-54)Fatal 1-00 1-03(0-86,1-24) 0-81(0-59,1-11) 1-14(0-85,1-53)

Coronary heart disease:Total RR(CI) 1-00 1-17(1-03,1-33) 0-85(0-69,1-06) 1-29(1-06,1-57)Non-fatal MI 1-00 1-23(1-06,1-42) 0-93(0-72,1-19) 1-35(1-08,1-69)(Cases) (285) (481) (81) (102)Fatal 1-00 1-09(0-86,1-37) 0-69(0-46,1-05) 1-20(0-84,1-72)(Cases) (119) (179) (27) (39)

Stroke:Total RR(CI) 1-00 1-01(0-86,1-18) 0-96(0-74,1-23) 1-17(0-91,1-47)Non-fatal 1-00 1-02(0-86,1-22) 0-95(0-71,1-26) 1-22(0-93,1-60)(Cases) (210) (295) (61) (68)Fatal 1-00 0 95(0 69,1.30) 0-96(0-57,1-60) 1-07(0-64,1-78)(Cases) (68) (88) (19) (19)

* Father deceased when cohort member was aged 16 years.t Coronary heart disease and stroke.

and fatal cardiovascular disease comprised thesum of non-fatal myocardial infarction andnon-fatal stroke, and fatal coronary heart dis-ease and fatal stroke respectively.Most deaths were reported by relatives or

postal authorities. The National Death Indexwas searched biennially for non-responders. Insuspected cases of cardiovascular death, per-mission to review medical records was soughtfrom the next of kin.

STATISTICAL ANALYSISIncidence rates were computed for non-fatalmyocardial infarction and fatal coronary heartdisease, non-fatal and fatal stroke, and non-fatal and fatal cardiovascular disease. The co-hort at risk included only those who were freeof cardiovascular disease at the beginning ofeach follow up interval.

Rates were obtained by dividing incidentcases by person years in each exposure category.All relative risks were age adjusted and 95%confidence intervals (95% CI) were calculated.Proportional hazards models were used to eval-uate the effects of childhood socioeconomicstatus while controlling for known and sus-pected cardiovascular risk factors. When a diag-nosis of a cardiovascular end point was made,follow up was stopped.Where a significant association between

socioeconomic status and age adjusted rates ofcardiovascular diseases was found, we pro-gressively added known and suspected risk fac-tors for cardiovascular disease"212831-33 to themultivariate models. This was done to de-termine whether differences in any of thesevariables provided a plausible explanation forany differences in risk of incident cardio-vascular diseases between the childhood socio-economic status groups.Only those end points that occurred between

the return of the 1976 questionnaire and 1June 1990 were included in the age adjustedas well as the multivariate analyses. Since in-formation on aspirin, exercise, alcohol intake,and diet were first sought in 1980, multivariatemodelling including these variables was re-stricted to those end points that occurred afterthat date. In order to compare the effects ofthe addition ofthese variables with age adjustedfindings, we also calculated age adjusted rel-ative risks for outcomes occurring between1980 and 1 June 1990.Since information on spouses' education,

participant's birth weight, and breastfed statuswere collected in 1992, these variables couldonly be used in models where non-fatal diseasewas the outcome.

ResultsBetween 1976 and 1990 we documented 949cases of non-fatal myocardial infarction, 364cases of fatal coronary heart disease, 634 casesofnon-fatal stroke, and 194 cases offatal stroke.In age adjusted analyses (table 1), women fromblue collar and "father deceased" childhoodbackgrounds had a small but significantly in-creased risk of non-fatal myocardial infarction,

12

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from

Childhood socioeconomic status and later heart disease

Table 2 Age standardised relative risks (95% CI) of cardiovascular diseases bychidhood socioeconomic status 1980-90

White collar Blue collar Fanner Father deceased*

No of women 40 140 52 914 9499 9416Person years exposure 391 044 514 647 92 239 91 747Cardiovascular diseaset:

Total RR(CI) 1 00 1-13(1-01,1-26) 0-94(0-78,1-13) 1-13(0-94,1-35)Non-fatal 1 00 1-16(1-02,1-31) 0-98(0-79,1-21) 1-20(0-98,1-47)Fatal 1 00 1-04(0 85,1-29) 0-79(0-55,1-14) 0-93(0-65,1-33)

Coronary heart disease:Total RR(CI) 1 00 1-19(1 03,1-37) 0-92(0-71,1-17) 1-22(0-97,1-53)Non-fatal MI 1 00 1-25(1 06,1 47) 1-03(0-79,1-36) 1-34(1-03,1-73)(Cases) (223) (384) (70) (79)Fatal 1 00 1-07(0-82,1-39) 0-65(0-40,1-04) 0-93(0-59,1-45)(Cases) (94) (139) (20) (24)

Stroke:Total RR(CI) 1 00 1-00(0-84,1-19) 0-96(0 73,1 27) 1-03(0-78,1-37)Non-fatal 1 00 0-98(0-80,1-19) 0-93(0-68,1-27) 1-03(0-75,1-42)(Cases) (176) (236) (50) (48)Fatal 1 00 1-06(0-74,1-50) 1-05(0-60,1-84) 0-95(0-52,1-75)(Cases) (52) (75) (16) (13)

* Father deceased when cohort member was aged 16 years.t Coronary heart disease and stroke.

Table 3 Relative risks (95% CI) of coronary heart disease by childhood socioeconomicstatus in multivariate analysis

White collar Blue collar Farmer Father deceased*

Non-fatal myocardial infarctionModel 1 1 00 1-19(1-03,1-38) 1-07(0-83,1-37) 1-33(1-06,1-67)

2 1 00 1-19(1-03,1-38) 1-12(0-88,1-44) 1-29(1-03,1-62)3+ 1 00 1-20(1-02,1-41) 1-19(0-91,1-56) 1-25(0-98,1-61)4+ 1 00 1-18(1 01,1 39) 1-16(0-89,1-51) 1-24(0-97,1-60)5+ 1-00 1-18(1-01,1-39) 1-17(0-90,1-53) 1-20(0-93,1-54)6+ 1-00 1-17(0-99,1-37) 1-15(0-88,1-50) 1-18(0-92,1-52)

Fatal coronary heart diseaseModel 1 1-00 0-98(0-75,1-27) 0-76(0-48,1-22) 1-12(0-74,1-68)

2 1-00 0-97(0-75,1-26) 0-79(0-49,1-26) 1-09(0-73,1-64)3+ 1-00 0-97(0-73,1-28) 0-72(0-43,1-22) 0-95(0-60,1-51)4+ 1-00 0-94(0-72,1-25) 0-69(0-41,1-16) 0-93(0-59,1-48)

Total coronary heart diseaseModel 1 1-00 1-14(1-00,1-29) 0-99(0-79,1-23) 1-28(1-05,1-56)

2 1-00 1-13(1-00,1-39) 1-03(0-83,1-29) 1-24(1-02,1-51)3+ 1-00 1-14(0-99,1-31) 1-06(0-84,1-35) 1-18(0-94,1-46)4+ 1-00 1-12(0-98,1-29) 1-03(0-81,1-30) 1-16(0-93,1-45)5+ 1-00 1-12(0-97,1-29) 1-04(0-82,1-32) 1-11(0-89,1-38)6+ 1-00 1-11(0-96,1-27) 1-02(0-81,1-30) 1-10(0-88,1-37)

Model: 1 =Age, plus body mass index, smoking. 2 = 1, plus parental MI <60 y, past history ofhypertension, hypercholesterolaemia, diabetes. 3=2, plus aspirin (1-6 per week vs 0 or >6),hormone replacement therapy, past oral contraceptive use. 4=3, plus alcohol, exercise, transfatty acids (energy adjusted), quintiles vitamin E. 5 = 4, plus breastfed, birth weight, adult height.6= 5, plus spouses' education.* Father deceased when cohort member was aged 16 years.+ Denotes follow up for period 1980-90, when extra study variables used in models 3 and 4were measured.

total coronary heart, and cardiovascular diseasein the range of 15-30%, compared with womenfrom white collar childhood backgrounds. Inthe same model, women from "farmer" child-hood backgrounds had an apparent small butnon-significant decrease in the risk of fatalcoronary heart disease (RR=0-69; 95% CI0-46,1 05) and fatal cardiovascular disease(RR=0-81; 95% CI 0-59,1-11). We also ob-served an apparent small but non-significantelevation in risk of non-fatal stroke (RR= 1 22;95% CI0 93,1 60) and total stroke (RR= 117;95% CI 0-91,1-47) in the "father deceased"group.Between 1980 and 1990 we documented

756 cases of non-fatal myocardial infarction,277 cases of fatal coronary heart disease, 510cases of non-fatal stroke, and 156 cases of fatalstroke. When we limited analysis to this period(table 2), the increased risks observed amongthe "father deceased" group for non-fatal, fataland total stroke, non-fatal and total cardio-vascular disease, and total coronary heart dis-ease were no longer apparent.The small but significantly increased risk of

total cardiovascular and coronary heart diseasein the blue collar and "father deceased" groupswas largely due to an increased risk of non-fatal myocardial infarction. The relative riskof non-fatal myocardial infarction and totalcoronary heart disease fell slightly within theblue collar group when data on body massindex and cigarette smoking were added to themodel (table 3). For this group the addition ofvariables collected in 1980 had little effect onthe relative risks, but did widen the confidenceintervals because of the smaller numbers ofcases in the restricted analysis. The relativerisks for outcomes increased slightly amongthe farmer group but fell markedly among the"father deceased" group.When we refined the categories of white

collar and blue collar into the occupationalclassification scheme of Edwards'7 as shown intable 4 (apart from the "farmer group" forwhich these data are given in table 3), we

observed the largest increase in risks found inthe present analyses. The relative risk of totalcoronary heart disease was 1-53 (95% CI 1 09,2-16) and that of non-fatal myocardial in-farction was 1-67 (95% CI 1-11,2-53) in thelowest (labourer) compared with the highest(professional) childhood socioeconomic groupafter adjustment for all potential confoundingvariables. In multivariate analysis a significanttrend of increased risk of non-fatal myocardialinfarction (p=0 045) with decreased child-hood socioeconomic status was observed.

DiscussionIn this prospective cohort study we observed a

small but significantly increased risk of totalcardiovascular disease (defined as coronaryheart disease and stroke) among women fromblue collar childhood backgrounds comparedwith women from white collar childhood back-grounds. The apparent small increase in riskof cardiovascular disease among women fromlower socioeconomic backgrounds was duelargely to an excess of coronary heart disease.While the larger number of cases of heart dis-

Table 4 Relative risks (95% CI) of coronary heart disease (CHD) by disaggregated childhood socioeconomic status inmultivariate analysis

Professional Manager Clerical Skilled Semi-skilled Domestic Labourerblue collar blue collar

No of women 12377 17 171 11 677 27203 17432 6100 3661Person years 169 198 234 803 159 576 371 093 238 297 83 257 49 868exposureTotal CHD 1-00 1-13 0-87 1-12 1-02 1-26 1-53Model 6+ (0-87,1-47) (0-64,1-18) (0-88,1-42) (0-79,1-32) (0-92,1-73) (1-09,2-16)Fatal CHD 1-00 0 70 0-69 0-69 0-69 0-85 1 30Model 4+ (0-43,1-15) (0-40,1-20) (0-44,1-07) (0-43,1-12) (0-46,1 55) (0-70,2-41)Non-fatal MI 1 00 1-35 0-93 1-32 1 19 1-46 1-67Model 6+ (0 94,1 84) (0-65,1 34) (0-99,1 77) (0 87,1 63) (1-01,2 11) (1-11,2-53)

Model: see table 3. MI= myocardial infarction.

13

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from

Gliksman, Kawachi, Hunter, Colditz, Manson, Stampfer, Speizer, Willett, Hennekens

ease than stroke may account for this, it shouldalso be noted that risk factors for myocardialinfarction and stroke differ.34The small increase in risk for women in the

blue collar group was not explained fully bydifferences in confounding or intermediatevariables. In particular, when we disaggregatedthe childhood socioeconomic categories andcompared the most disadvantaged with themost advantaged group in multivariate analysis,there was an excess risk of total coronary heartdisease and non-fatal myocardial infarction inthe range of 50-70%. This excess is similar tothat reported for middle aged men in the Ku-opio study,35 which used paternal occupationin association with other parental variables toassign childhood socioeconomic status withinthree categories. In that study, adjustment forcardiovascular risk factors and adult socio-economic status weakened but did not elim-inate the association between childhoodsocioeconomic status and prevalent ischaemicheart disease.When we adjusted associations for differ-

ences in cardiovascular risk factors and socio-demographic variables, the risk estimates wereattenuated for women whose fathers were de-ceased but were increased among women inthe farmer group. This raises the possibilitythat some of the increased incidence of cor-onary heart disease in the former and decreasedincidence in the latter were due to differences inrisk factors in adulthood which were associatedwith childhood socioeconomic status.

Recent reviews have examined longitudinaland case-control studies of the association be-tween early life environment and adult cardio-vascular disease,"1 36 and concluded that theobserved associations were compromised byinadequate handling of confounding variables,in particular of adult socioeconomic status.This may lead to an overestimation of the riskassociated with lower childhood socioeconomicstatus. Our analysis was performed among acohort of female registered nurses, a relativelyhomogeneous group for adult socioeconomicstatus. In addition, we adjusted for differencesin the socioeconomic status of spouses of thecohort members.

It is possible that the effect of differences inchildhood socioeconomic status among womenin this study was confounded by the socio-economic status of their spouses and henceby the economic wellbeing of the woman'sdomestic arrangements. While educational at-tainment is an important variable in assessingadult socioeconomic status,'9 some caveatsapply in assessing its validity in this analysis.We assessed spouses' educational status onlyin 1992. For some, this person may not be thesame spouse preceding the woman's cardio-vascular event. No information on spouses'educational attainment was available for casesof fatal cardiovascular disease. Finally, almost17% of the cohort were either unmarried in1992 or gave no reply to this question. How-ever, inclusion of this variable in multivariatemodelling did not materially affect the riskestimates.We used the father's and not the mother's

occupation to determine childhood socio-economic status.37 An analysis of the dis-tribution of mother's listed occupation in thiscohort showed that over 66% were employedin domestic duties and analyses ofmodels basedsolely on mother's occupation showed no in-creased risk for any group.A recent case-control study suggested

differences in coronary heart disease ratesamong adult men from different (adult) socio-economic groups could be explained by differ-ences in serum high density lipoproteincholesterol.'8 The possible role of adult bloodlipid subfractions in accounting for the differ-ences in the present study could not be de-termined, as haematological variables were notascertained in this cohort.

It has been argued that early life en-vironmental factors such as fetal nutrition orbeing breastfed in infancy might have long termeffects on the risk of cardiovascular disease.Babies who are small at birth or during infancyhave increased rates of cardiovascular diseaseand non-insulin dependent diabetes mellitus asadults.38 Adult height, which is positively re-lated to childhood nutritional status and in-versely related to the adult risk ofcardiovasculardisease,39 has been used in a number of studiesas a proxy for childhood socioeconomic ex-perience.40 42 However, inclusion of these vari-ables in multivariate modelling did not explainthe observed associations between childhoodsocioeconomic status and cardiovascular out-comes among women from blue collar child-hood backgrounds.The associations observed in this cohort may

have been due to unequal access to medicalintervention in some groups when cardio-vascular disease symptoms such as angina pec-toris first appeared. While this potential sourceof bias might be particularly important in anation without universal health care, when weanalysed the rate of coronary artery bypassgrafting, women from blue collar childhoodbackgrounds were somewhat more likely to havehad this procedure after adjustment for age(RR= 1-19; 95% CI 1-02,1 39). This may ex-plain, at least partly, the apparently strongerassociation between lower childhood socio-economic status and incidence of myocardialinfarction than mortality from coronary heartdisease. Although nurses from lower childhoodsocioeconomic backgrounds were more likelyto suffer a myocardial infarct, to the extent thatmedical interventions (such as bypass grafting)were effective, any association between child-hood socioeconomic status and mortality inthis cohort may have been underestimated.

All participants in the cohort were selectedfrom registers of the State Boards of Nursing.Thus, all cohort members were health pro-fessionals with recognised nursing credentials.If women from lower socioeconomic back-grounds were under-represented in this cohort,then our ability to detect a stronger effect ofpoverty or absolute deprivation during child-hood on risk of coronary heart disease in ad-ulthood would be limited.

In summary, our findings indicate that lowchildhood socioeconomic status is associated

14

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from

Childhood socioeconomic status and later heart disease

with a small but significant increase in risk ofcoronary heart disease in middle aged women.The increased risk among women from bluecollar backgrounds was not fully explained bydifferences in a large number of risk factors forcardiovascular disease. This was particularly sofor those women from the most disadvantagedchildhood socioeconomic background. Al-though the magnitude of the association ismodest, the public health impact could besubstantial.43 Increased preventive efforts tar-getting modifiable cardiovascular risk factorsamong children and adolescents before riskbehaviour becomes established has been ad-vocated.44 These data suggest that such effortshould be targetted towards the socio-economically disadvantaged.Our results suggest that neither differences

in fetal nutrition as evidenced by birth weight,nor differences in adult environmental cir-cumstances, could explain the association be-tween childhood socioeconomic circumstancesand risk of adult cardiovascular disease ob-served in our cohort. There is some evidencelinking nutritional status in early and middlechildhood years to the risk of subsequentcardiovascular disease.4546 If these findings areconfirmed, nutritionally based interventionprogrammes in childhood may offer a meansof addressing the imbalance in the later life riskof cardiovascular disease.A hypothesis worthy of investigation is

whether directly addressing socioeconomic

deprivation in childhood might be of addedeffectiveness in reducing the risk of coronaryheart disease in later life.

Supported by research grants (HL 24074, HL 34594, CA 40935and CA 40356) from the National Institutes of Health, USA.

1 Fox AJ, Goldblatt PO, Jones DR. Social class mortalitydifferentials: artefact, selection or life circumstances? In:RG Wilkinson ed. Class and health: research and longitudinaldata. London: Tavistock, 1986.

2 Carstairs V, Morris R. Deprivation: explaining differencesin mortality between Scotland, England and Wales. BM]1989;299:886-9.

3 Williams FLR, Lloyd 0. Cerebrovascular disease in Scotlandduring 1959 to 1983: its geographical distribution andassociations. Community Medicine 1989;11:306-15.

4 Marmot M. Socioeconomic determinants ofCHD mortality.Int Epidemiol 1989;18:3(suppl 1):S196-202.

5 Shaper AG, Phillips AN, Pocock SJ, Walker M, MacfarlanePW. Risk factors for stroke in middle aged British men.BM] 1991;302:1111-5.

6 Kawachi I, Marshall S, Pearce N. Social class inequalitiesin the decline of coronary heart disease among NewZealand men, 1975-1977 to 1985-1987. Int Epidemiol1991 ;20:393-8.

7 Baker IA, Sweetnam PM, Yarnell JWG, Bainton D, ElwoodPC. Haemostatic and other risk factors for ischaemic heartdisease and social class: evidence from the Caerphilly andSpeedwell studies. Int] Epidemiol 1988;17:759-65.

8 Stamler R, Shipley M, Elliott P, Dyer A, Sans S, StamlerJ. Higher blood pressure in adults with less education.Some explanations from INTERSALT. Hypertension 1992;19:237-41.

9 Gliksman MD, Dwyer T, Wlodarczyk J. Differences inmodifiable cardiovascular disease risk factors in Australianschoolchildren: the results of a nationwide survey. PrevMed 1990,19:291-304.

10 Gliksman MD, Lazarus R, Wilson A. Differences in serum

lipids in Australian children with age, sex, socioeconomicstatus and ethnic origin: is diet responsible? Int Epidemiol

1993;22:247-54.11 J, Whincup P, Shaper AG. Early life experience and

adult cardiovascular disease: longitudinal and case-controlstudies. Int ]fEpidemiol 1991;20:833-44.

Robinson JP, Athanasiou R, Head KB. Measures of oc-

cupational attitudes and occupational characteristics. Ann

Arbor, MI: Survey Research Centre;Institute for SocialResearch, 1969.

Manson JE, Rimm EB, Stampfer MJ, et al.Physical activity

and incidence of non-insulin-dependent diabetes mellitusin women. Lancet 1991;338:774-8.

14 Willett WC, Sampson L, Browne ML, et al. The use of a

self-administered questionnaire to assess diet four yearsin the past. AmjfEpidemiol 1988;127:188-99.

15 Willett WC. Nutritional epidemiology. New York: OxfordUniversity Press, 1989.

16 Edwards AM. US census of population, 1940: Comparativeoccupational statistics, 1870-1940. Washington, DC: USGovernment Printing Office, 1943.

17 Liberatos P, Link BG, Kelsey JL. The measurement of socialclass in epidemiology. Epidemtiol Rev 1988;10:87-121.

18 Hebert PR, BuringJE, O'Connor GT, Rosner B, HennekensCH. Occupation and risk of nonfatal myocardial in-farction. Arch Intern Med 1992;152:2253-7.

19 Winkleby MA, Jatulis DE, Frank E, Fortmann SP. Socio-economic status and health: how education, income, andoccupation contribute to risk factors for cardiovasculardisease. Am J Public Health 1992;82:816-20.

20 Willett WC, Stampfer MJ, Bain C, Rosner B, Speizer FE.Cigarette smoking, relative weight, and menopause. AmJ7Epidemiol 1983;117:651-8.

21 Paffenbarger RS Jr, Blair SN, Lee IM, Hyde RT. Measure-ment of physical activity to assess health effects in free-living populations. Med Sci Sports Exer 1993;25:60-70.

22 Washburn RA, Adams LL, Haile GT. Physical activityassessment for epidemiologic research: the utility of two

simplified approaches. Prev Med 1987;16:636-46.23 Washburn RA, Goldfield SRW, Smith KW, McKinlay JB.

The validity of self-reported exercise-induced sweating asa measure of physical activity. An _7 Epidemiol 1990;132:107-13.

24 Colditz GA, Willett WC, Stampfer MJ, et al. The influenceof age, relative weight, smoking, and alcohol intake on thereproducibility of a dietary questionnaire. Int I Epidemiol1987;16:392-8.

25 Giovannucci E, Colditz GA, Stampfer MJ, et al. The as-sessment of alcohol consumption by a simple self-ad-ministered questionnaire. Am T Epidemiol 1991;133:810-17.

26 Willett WC, Stampfer MJ, Manson JE, et al. Intake of transfatty acids and risk of coronary heart disease amongwomen. Lancet 1993;341:581-5.

27 Stampfer MJ, Hennekens CH, Manson JE, et al. A pro-spective study of vitamin E consumption and risk ofcoronary disease in women. N Engl 7 Med 1993;328:1444-9.

28 Stampfer AG, Colditz GA, Willett WC, Speizer FE, Hen-nekens CH. A prospective study of moderate alcoholconsumption and the risk of coronary disease and strokein women. N Engl _

Med 1988;319:267-73.29 World Health Organisation. IHD registers: report of the Fifth

Working Group. Copenhagen: WHO, 1971.30 Willett WC, Green A, Stampfer MJ, et al. Relative and

absolute excess risks of coronary heart disease amongwomen who smoke cigarettes. N Engl ] Med 1987;317:1303-9.

31 Colditz GA, Bonita R, Stampfer MJ, et al. Cigarette smokingand risk of stroke in middle-aged women. N Engl_7 Med1988;318:937-41.

32 Manson JE, Stampfer MJ, Colditz GA, et al. A prospectivestudy of aspirin use and primary prevention of cardio-vascular disease in women. _JAAIA 1991,266:521-7.

33 Stampfer MJ, Colditz GA, Willett WC, et al. Post-menopausal estrogen therapy and cardiovascular disease:ten-year follow-up from the Nurses' health study. N EnglJMed 1991;325:756-62.

34 Reed DM. The paradox of high risk of stroke in populationswith low risk of coronary heart disease. Amn 7 Epidemiol1990;131:579-88.

35 Kaplan GA, Salonen JT. Socioeconomic conditions in child-hood and ischaemic heart disease during middle age. BM]1990;301:1121-3.

36 Hill AB. The environment and disease: association or caus-ation? Proceedings of the Royal Society of Medicine 1965;58:295-300.

37 Haug MR. Measurement in social stratification. AnnualRevues in Sociology 1977;3:51-77.

38 Barker DJP, Gluckman PD, Godfrey KM, Harding JE,Owens JA, Robinson JS. Fetal nutrition and cardiovasculardisease in adult life. Lancet 1993;341:938-41.

39 Yarnell JW, Limb ES, Layzell JM, Baker IA. Height: a riskmarker for ischaemic heart disease: prospective resultsfrom the Caerphilly and Speedwell heart disease studies.Eur Heart]7 1992;13:1602-5.

40 Walker M, Shaper AG, Phillips AN, Cook DG. Shortstature, lung function and risk of a heart attack. Int ]Epidemiol 1989;18:602-6.

41 Nystrom Peck AM, Vagero DH. Adult body height, self-perceived health and mortality in the Swedish population._7 Epidemiol Community Health 1989;43:380-4.

42 Marmot MG, Shipley MJ, Rose G. Inequalities in death -specific explanations of a general pattern. Lancet 1984-i:1003-6.

43 Rose G. Mass prevention: population-directed approach.Acta Cardiol 1988;29(suppl): 131-9.

44 Report of the expert panel on blood cholesterol in children andadolescents, Bethesda, MA: National Institutes of HealthPublication 91-2732, HHS, USA, 1991.

45 Ariouat JF, Barker DJ. The diet of girls and young womenat the beginning of the century. Nutr Health 1993;9: 15-23.

46 Fall CH, Barker DJ, Osmond C, et al. Relation of infantfeeding to adult serum cholesterol concentration and deathfrom ischaemic heart disease. BM] 1992;304:801-5.

15

on Decem

ber 14, 2020 by guest. Protected by copyright.

http://jech.bmj.com

/J E

pidemiol C

omm

unity Health: first published as 10.1136/jech.49.1.10 on 1 F

ebruary 1995. Dow

nloaded from