Word Count Text: 3597; Word Count Abstract: 250;

2 Word count text: 3597; Word count abstract: 250;

Number of references: 45;

1 Tables: 5; Figures: 0; Supplementary Figure: 1;

2 Supplementary Tables: 2.

Maternal educational level and blood pressure, aortic stiffness, cardiovascular structure and functioning in childhood: The Generation R Study

Running title: Maternal education and cardiovascular outcomes

Selma H. BOUTHOORN MDa,b, Frank J. VAN LENTHE PhDb, Layla L. DE JONGE MDa,c,d,

Albert HOFMAN MD PhDa,c , Lennie VAN OSCH-GEVERS, MD, PhDd, Vincent WV

JADDOE MD PhDa,c,d, Hein RAAT MD PhDb

aThe Generation R Study Group, Erasmus Medical Centre, Rotterdam, the Netherlands;

bDepartment of Public Health, Erasmus Medical Centre, the Netherlands;

cDepartment of Epidemiology, Erasmus Medical Centre, Rotterdam, the Netherlands;

dDepartment of Pediatrics, Erasmus Medical Centre, the Netherlands.

Corresponding author

Selma H. Bouthoorn MD. The Generation R Study Group, Erasmus MC, University Medical

Centre Rotterdam. P.O. Box 2040, 3000 CA Rotterdam, the Netherlands; Telephone: +31

(0)10-704 38496; fax: +31 (0)10 704 4645; e-mail: [email protected].

Conflict of interest: None

Abstract 3 2

4 3Background: In adults, low level of education was shown to be associated with higher blood

4pressure levels and alterations in cardiac structures and function. It is currently unknown

5whether socioeconomic inequalities in arterial and cardiac alterations originate in childhood.

6Therefore, we investigated the association of maternal education with blood pressure levels,

7arterial stiffness and cardiac structures and function at the age of 6 years, and potential

8underlying factors.

9Methods: The study included 5843 children participating in a prospective cohort study in the

10Netherlands. Maternal education was assessed at enrollment. Blood pressure, carotid-femoral

11pulse wave velocity, left atrial diameter, aortic root diameter, left ventricular mass and

12fractional shortening were measured at the age of 6 years.

13Results: Children with low educated (category 1) mothers had higher systolic (2.80 mm Hg,

1495% CI: 1.62-2.94) and diastolic (1.80 mm Hg, 95% CI: 1.25-2.35) blood pressure levels as

15compared to children with high educated (category 4) mothers. The main explanatory factors

16were the child’s BMI, maternal BMI and physical activity. Maternal education was negatively

17associated with fractional shortening (p for trend 0.008), to which blood pressure and child’s

18BMI contributed the most. No socioeconomic gradient was observed in other arterial and

19cardiac measurements.

20Conclusion: Socioeconomic inequalities in blood pressure are already present in childhood.

21Higher fractional shortening among children from low socioeconomic families might be a

22first cardiac adaptation to higher blood pressure and higher BMI. Interventions should be

23aimed at lowering child BMI and increasing physical activity among children from low

24socioeconomic families.

25Keywords: Maternal educational level, blood pressure, cardiac structures, cardiac function,

26arterial stiffness, childhood 5 3

44Introduction 7 4

8 45Cardiovascular disease (CVD) is the leading cause of death in most western parts of the

46world. Low socioeconomic position (SEP) has been recognized as a marked risk factor in the

47occurrence of CVD among adults . A better understanding of the origins and underlying

48mechanisms of socioeconomic inequalities in CVD is essential to develop effective strategies

49aimed at preventing or reducing these inequalities. Growing evidence suggest that CVD

50inequalities arise already in childhood, since several CVD risk factors including, low birth

51weight, short gestational age, physical inactivity, childhood obesity and familial hypertension,

52affect children from low socioeconomic families more frequently . These risk factors may

53cause inequalities in blood pressure levels in childhood, and may track into adulthood .

54Indeed, a recent study showed socioeconomic inequalities in blood pressure to appear at the

55age of 5-6 years old already, mediated by known CVD risk factors including birth weight,

56childhood BMI and maternal BMI .

57Alterations in arterial and cardiac structures such as aortic stiffness, aortic root size, left atrial

58and left ventricular enlargement, contribute to the risk of CVD among adults . In adults, it has

59been shown previously that a low level of education was associated with left ventricular

60hypertrophy, left ventricular dilatation and cardiac dysfunction . Factors recognized to

61influence these structures include elevated blood pressure levels and obesity . These factors

62are more common in children from low SEP families and were also found to cause a higher

63left ventricular mass in childhood already . However, no previous study examined the

64socioeconomic gradient in early structural changes of the heart among children. Therefore, to

65improve the understanding of the origins and underlying mechanisms of CVD inequalities, we

66investigated in a population-based, prospective cohort study, the association between maternal

67educational level and blood pressure levels, arterial stiffness and cardiac structures and

68function at the age of 6 years. Furthermore, we examined whether adiposity measures,

69lifestyle-related determinants and birth characteristics could explain these associations. 9 5

86Methods

87Study design 11 6

12 88This study is embedded within the Generation R Study, a population-based prospective cohort

89study from early pregnancy onwards described in detail elsewhere . Enrolment was aimed at

90early pregnancy, but was allowed until the birth of the child. All children were born between

91April 2002 and January 2006 and form a prenatally enrolled birth-cohort that is currently

92being followed-up until young adulthood. The study was conducted in accordance with the

93guidelines proposed in the World Medical Association of Helsinki and has been approved by

94the Medical Ethical Committee of the Erasmus MC, University Medical Centre Rotterdam.

95Written consent was obtained from all participating parents .

96Study group

97In total, 8305 children still participate in the study from the age of 5 years . The participating

98children and their mothers were invited to a well-equipped and dedicated research center in

99the Erasmus Medical Center - Sophia Children’s Hospital between March 2008 and January

1002012. Measurements were focused on several health outcomes including body composition,

101obesity, heart and vascular development and behaviour and cognition . In total, 6690 children

102visited the research center. Mothers of children who didn’t visit the research center had a

103lower educational level, more frequently smoked during pregnancy (p<0.001) and fewer of

104them gave breastfeeding (p=0.001) as compared to mothers of children who visited the

105research center. No differences were found in other characteristics between these groups

106(Supplementary Table S1). We excluded twins (n=167) and participants who lacked

107information on maternal educational level (n=594) and on cardiovascular measurements

108(n=50). Furthermore, children with echocardiographic evidence of congenital heart disease or

109kidney disease were excluded (n=36), leaving a study population of 5843 children

110(Supplementary Figure S1).

111Maternal educational level 13 7

14 112Our indicator of SEP was educational level of the mother. Level of maternal education was

113established using questionnaires at enrollment. The Dutch Standard Classification of

114Education was used to categorize 4 subsequent levels of education: 1. high (university

115degree), 2. mid-high (higher vocational training, Bachelor’s degree), 3. mid-low (>3 years

116general secondary school, intermediate vocational training) and 4. low (no education, primary

117school, lower vocational training, intermediate general school, or 3 years or less general

118secondary school) .

119Child cardiovascular structures and function

120We measured blood pressure with the child in supine position quietly awake. Systolic and

121diastolic blood pressure (SBP and DBP) was measured at the right brachial artery, four times

122with one minute intervals, using the validated automatic sphygmanometer Datascope Accutor

123Plus TM (Paramus, NJ, USA) . A cuff was selected with a cuff width approximately 40% of the

124arm circumference and long enough to cover 90% of the arm circumference. More than 90%

125of the children who visited the research center had four successful blood pressure

126measurements available. SBP and DBP were determined by excluding the first measurement

127and averaging the other measurements.

128Carotid-femoral pulse wave velocity, the reference method to assess aortic stiffness , was

129assessed using the automatic Complior device (Complior; Artech Medical, Pantin, France)

130with participants in supine position. The distance between the recording sites at the carotid

131(proximal) and femoral (distal) artery was measured over the surface of the body to the

132nearest centimeter. Through piezoelectric sensors placed on the skin, the device collected

133signals to assess the time delay between the pressure upstrokes in the carotid artery and the

134femoral artery. Carotid-femoral pulse wave velocity was calculated as the ratio of the distance

135travelled by the pulse wave and the time delay between the feet of the carotid and femoral 15 8

16 136pressure waveforms, as expressed in meters per second . To cover a complete respiratory

137cycle, the mean of at least 10 consecutive pressure waveforms was used in the analyses.

138Recently, it has been shown that pulse wave velocity can be measured reliably, with good

139reproducibility in a large pediatric population-based cohort .

140Two-dimensional M-mode echocardiographic measurements were performed using the ATL-

141Philips Model HDI 5000 (Seattle, WA, USA) or the Logiq E9 (GE Medical Systems,

142Wauwatosa, WI, USA) devices. The children were examined in a quiet room with the child

143awake in supine position. Missing echocardiograms were mainly due to restlessness of the

144child or unavailability of equipment or sonographer. Left atrial diameter, interventricular end-

145diastolic septal thickness (IVSTD), left ventricular end-diastolic diameter (LVEDD), left

146ventricular end-diastolic posterior wall thickness (LVPWTD), interventricular end-systolic

147septal thickness, left ventricular end-systolic diameter, left ventricular end-systolic posterior

148wall thickness, aortic root diameter, and fractional shortening were measured using methods

149recommended by the American Society of Echocardiography . Left ventricular mass (LV

150mass) was computed using the formula derived by Devereux et al :LV mass = 0.80 ×

1511.04((IVSTD + LVEDD + LVPWTD)3 − (LVEDD) 3)+ 0.6. To assess reproducibility of

152ultrasound measurements, the intraobserver and interobserver intraclass correlation

153coefficients were calculated previously for left atrial diameter, aortic root diameter, IVSTD,

154LVEDD and LVPWTD in 28 subjects (median age 7.5 years, interquartile range 3.0-11.0) and

155varied between 0.91 to 0.99 and 0.78 to 0.96, respectively .

156Explanatory variables

157The following factors were considered to be potential explanatory factors in the pathway

158between SEP and blood pressure, cardiovascular structures and function. These were chosen

159based on previous literature . 17 9

18 160Maternal characteristics

161Information on pre-pregnancy weight, pre-pregnancy hypertension (yes, no) and smoking

162during pregnancy (no, until confirmed pregnancy and continued during pregnancy) and

163financial difficulties, as indicator of family stress, was obtained by questionnaires. Also,

164another measurement of family stress was assessed with the family assessment device score at

165age 5. A higher score reflects more ‘stress’. Information on pregnancy-induced hypertension

166and preeclampsia was obtained from medical records. Women suspected of pregnancy

167hypertensive complications, based on these records, were crosschecked with the original

168hospital charts. Details of these procedures have been described elsewhere . Maternal height

169was measured during visits at our research center. On the basis of height and pre-pregnancy

170weight, we calculated pre-pregnancy body mass index (BMI) (weight/height2).

171Birth characteristics

172Birth weight and gestational age at birth were obtained from midwife and hospital registries.

173Information on breastfeeding during infancy (ever/never) was obtained by questionnaires.

174Child characteristics

175Weight of the child was measured while wearing lightweight clothes and without shoes by

176using a mechanical personal scale (SECA), and height was measured by a Harpenden

177stadiometer (Holtain Limited) in standing position, which were both calibrated on a regular

178basis. BMI was calculated using the formula; weight (kg) / height (m)2. Standard-deviation

179scores (SDS) adjusted for age and gender were constructed for these growth measurements .

180Watching television (< 2 hours/day, ≥ 2 hours/day), as indicator of sedentary behaviour, and

181playing sports (yes/no), as indicator of physical activity, were obtained from questionnaires at

182the age of 5 years. 19 10

20 183Confounding variables

184Gender and date of birth of the child were obtained from midwife and hospital registries.

185Ethnicity (Dutch, other western, non-western) was obtained from the first questionnaire at

186enrollment in the study. These factors were considered as potential confounding factors.

187Statistical analyses

188First, univariate associations of maternal educational level to all covariates, blood pressure,

189carotid-femoral pulse wave velocity, cardiac structures and function were explored using Chi-

190square tests and ANOVAs. Second, we assessed the association of maternal educational level

191with blood pressure levels, carotid-femoral pulse wave velocity, cardiac structures and

192function using linear regression models adjusted for the potential confounders (model 1). To

193evaluate the mediating effects of all potential explanatory factors Baron and Kenny’s causal

194step approach was used . Only those factors that were significantly associated with the

195outcome (independent of maternal educational level; data available upon request) and

196unequally distributed across SEP groups (Table 1) were added separately to model 1 . To

197assess their mediating effects, the corresponding percentages of attenuation of effect estimates

198were calculated by comparing differences of model 1 with the adjusted ones (100 x (B model 1 –

199B model 1 with explanatory factor) / (B model 1 )). Finally, a full model containing maternal educational level

200and all the explanatory factors assessed the joint effects of the explanatory factors. The

201explanatory mechanisms were investigated only for outcomes (blood pressure, pulse wave

202velocity, cardiac structures and function) that differed by maternal educational level after

203adjustment for confounders (model 1). Interaction terms between maternal educational level

204and child’s sex on all the outcomes were not significant (p>0.1); therefore analyses were not

205stratified for sex. Multiple imputation was used to deal with the missing values in the

206covariates. Five imputed datasets were created and analysed together. A 95% confidence 21 11

22 207interval (CI) was calculated around the percentage attenuation using a bootstrap method with

2081000 re-samplings per imputed dataset in the statistical program R . All the other statistical

209analyses were performed using Statistical Package of Social Science (SPSS) version 20.0 for

210Windows (SPSS Inc, Chicago, IL, USA).

221Results

222 223Table 1 shows the characteristics of the study population. Of all participating children 22.4%

224(n= 1308) had mothers with a low educational level and 25.3% (n=1483) of the mothers were

225high educated. Compared with mothers with a high education, those with a low education had

226a higher pre-pregnancy BMI, suffered more frequently from pre-pregnancy hypertension, had

227more financial difficulties and a higher family assessment device score (p<0.001), indicating 23 12

24 228more stress within their families, fewer of them gave breastfeeding, and they more frequently

229smoked during pregnancy (p=0.002). Their children were on average lighter at birth and had

230a shorter gestational duration (p<0.001). Mothers with a high education were more frequently

231Dutch, while mothers with a low education had more frequently a non-western ethnic

232background (p<0.001). Children of low educated mothers less frequently played sports,

233watched more frequently ≥ 2 hours television per day and were heavier as compared with

234children of high educated mothers (p<0.001). Mean systolic (SBP) and diastolic blood

235pressure (DBP) levels (p<0.001) and carotid-femoral pulse wave velocity (p=0.02) were

236negatively associated with level of education. Also, left atrial diameter (p<0.001) and left

237ventricular mass (p=0.003) differed according to educational level (Table 1).

238Multivariable linear regression analyses adjusted for the potential confounders showed that

239the lower the education of the mother, the higher the systolic and diastolic blood pressure of

240their children (p for trend <0.001) (Table 2). No consistent associations of maternal education

241were observed with carotid-femoral pulse wave velocity, aortic root diameter, left atrial

242diameter, and left ventricular mass. Fractional shortening was negatively associated with

243maternal educational level (p for trend 0.008). There was no interaction between ethnicity and

244maternal education on these outcomes (p > 0.05).

245Maternal pre-pregnancy BMI, maternal pre-pregnancy hypertension, gestational age, and the

246child’s height and BMI were related both to maternal educational level and SBP. The selected

247explanatory factors for DBP included maternal pre-pregnancy hypertension, birth weight,

248watching television, playing sports, and the child’s height and BMI. Birth weight, child’s

249BMI, SBP and DBP were the selected explanatory factors in the association between maternal

250education and fractional shortening. For each selected factor, an interaction with maternal

251education was tested for significance, none of them were significant (p > 0.05). 25 13

26 252Regarding the proportion of explanation by each risk factor (Table 3 and Table 4), the child’s

253BMI was the most important contributor to the association of maternal education to SBP in

254the mid-low and low educational subgroup (15% attenuation (95% CI: -29% to -7%) and 29%

255(95% CI: -42% to -20%) ). Secondly, maternal pre-pregnancy BMI contributed to the

256association between maternal education and SBP accounting for 11% (95% CI: -22% to -6%)

257and 12% (95% CI: -18% to -7%) respectively. Overall, 12% (95% CI: -26% to 0.3%) and

25824% (95% CI: -36% to -14%) of the association of maternal education and SBP was

259explained for mid-low and low education by including the explanatory factors (Table 3). For

260DBP, the child’s BMI and playing sports were the most contributing factors in the association

261between maternal education and DBP, explaining 8% (95% CI: -14% to -3%) and 7% (95%

262CI: -13% to -3%) in the lowest educational subgroup. The overall explanation was 12% (95%

263CI: -47% to -3%), 16% (95% CI: -30% to -8%) and 21% (95% CI: -35% to -12%) in the mid-

264high, mid-low and low educational subgroup respectively (Table 4). For both SBP and DBP

265the socioeconomic differences remained significant after including all the explanatory factors

266(Table 3 and Table 4). Complete elimination of the association of mid-low and low education

267with fractional shortening was observed after individual adjustment for BMI, SBP and DBP.

268SBP was the most contributing factor, explaining 27% (95% CI: -142% to 25%) and 42%

269(95% CI: -159% to -12%) respectively (Table 5).

274 27 14

28 275

287Discussion

288This study shows that socioeconomic inequalities in adult CVD may have their origins in

289childhood. We found socioeconomic inequalities in systolic and diastolic blood pressure to

290occur already at the age of 6 years. Furthermore, a positive association was found between

291maternal education and fractional shortening, which was explained by higher blood pressure

292levels and a higher BMI among children of lower educated mothers. We did not find evidence

293for socioeconomic inequalities in carotid- femoral pulse wave velocity, aortic root diameter,

294left atrial diameter, and left ventricular mass at the age of 6 years. 29 15

30 295Methodological considerations

296The strengths of this study are the prospective population-based design and the availability of

297many important covariates that may explain the association between maternal education,

298blood pressure levels, and cardiac structures and function. In addition, we had a large sample

299size of 5843 participants and the availability of several cardiovascular measurements in most

300of these children.

301To various extents, our results may have been influenced by the following limitations. We

302used maternal educational level as indicator of SEP. Maternal education may reflect general

303and health-related knowledge, health behavior, health literacy and problem-solving skills .

304Furthermore, it has been shown that education is the strongest and most consistent

305socioeconomic predictor of cardiovascular risk factors . It is less clear to what extent it

306captures the material and financial aspects of the household. We therefore repeated the

307analyses using household income level as determinant, and we found comparable results with

308the highest SBP and DBP and the highest fractional shortening in the lowest income group

309(Supplementary Table S2). Information on various covariates in this study was self-reported,

310which may have resulted in underreporting of certain adverse lifestyle related determinants. In

311our blood pressure measurements, we were not able to account for circadian rhythm. Yet,

312while this may have resulted in random error, it seems less likely that bias varied

313systematically across socioeconomic groups. Although the participation rate in The

314Generation R Study was relatively high (61%), there was some selection towards a relatively

315highly educated and more healthy study population . Non-participation would have led to

316selection bias if the associations of maternal educational level with the outcome differed

317between participants and non-participants. Previous research showed that this bias is

318minimal . However, it cannot be ruled out completely. Finally, unmeasured factors related to

319both SEP and blood pressure, such as salt intake and psychosocial factors, might explain some 31 16

32 320of the remaining effect of SEP on blood pressure. Sleep patterns may also explain some of the

321remaining effect, since this has been associated with cardiovascular health in previous

322studies .

323Maternal education, blood pressure and cardiac structures and function

324Previous literature concerning the association between socioeconomic circumstances and

325blood pressure in childhood is conflicting . This might be due to the use of different study

326designs and different study populations. In line with previous literature we found BMI to be

327the factor contributing most to the association between maternal education and blood pressure

328. Obesity may cause disturbances in autonomic function, insulin resistance and abnormalities

329in vascular structure and functions, which in turn may cause elevated blood pressure levels .

330Maternal pre-pregnancy BMI was also an important explanation of the socioeconomic

331gradient in SBP. Maternal BMI and a child’s BMI are correlated; the effect of maternal BMI

332may act for an important part through its effect on the child’s BMI. This is supported by our

333findings which showed that after adjustment for child’s BMI and maternal BMI, only child’s

334BMI remained significant (data not shown). A novel finding of this study is that indicators of

335sedentary behavior and physical activity, including watching television and playing sports,

336mediate the association between maternal education on DBP. Although these effects may pass

337through the effects of the child’s BMI on DBP, its significant association in the fully adjusted

338model suggests that there is probably also an independent effect of physical activity on DBP,

339which explains the socioeconomic inequalities in DBP. This is supported by a recent study of

340Knowles et al. who found higher physical activity to be associated with lower DBP levels

341among 5-7 year old children, independent of BMI and other markers of adiposity . These

342findings suggest that intervention should be aimed at increasing physical activity, especially

343among children with a lower SEP. 33 17

34 344Although children from low SEP families had higher BP levels and a higher BMI which is

345previously shown to be associated with larger left ventricular mass and atrial enlargement , we

346did not find a socioeconomic gradient in cardiac structures at the age of 6 years. We found,

347however, left ventricular mass to be lower in the mid-high and the mid-low educational

348subgroup as compared to the high educational subgroup. The direction of the association was

349not entirely equal to what was expected. Also, no socioeconomic gradient was observed and

350therefore commonly used cardiovascular risk factors explaining socioeconomic inequalities

351are not likely to be an explanation. Future research is necessary to replicate these findings.

352Another remarkable finding was the higher fractional shortening among children with mid-

353low and low educated mothers. One explanation might be that these children experience more

354stress during the echocardiographic assessment due to a lower lack of parental support

355because they, as well as their parents, are more impressed by the measurement setting,

356resulting in a higher fractional shortening. However, stress would also have increased their

357heart rate and adjustment for heart rate would then have explained, at least to some extent, the

358association between maternal education and fractional shortening. In our study, this

359explanation is less likely since heart rate, as indicator of stress, was not associated with

360fractional shortening. Since an increased fractional shortening was explained by higher blood

361pressure levels and a higher BMI, another explanation might be that it is the first indication of

362cardiac adaptation to higher blood pressure levels and higher BMI among low SEP children.

363However, in a later stage higher blood pressure levels and a higher BMI result in alterations

364of cardiac structures such as left ventricular and left atrial enlargement , which are associated

365with a lower fractional shortening. Thus our findings that lower education leads to a higher

366fractional shortening at the age of 6 are not fully understood and a change finding can also not

367be excluded. Confirmation in future studies is therefore highly recommended and replication

368would suggest that cardiac structural alterations can be prevented at this age by lowering 35 18

36 369blood pressure levels and body weight, especially among children from low socioeconomic

370families.

371Conclusion

372Our study adds to the small body of literature concerning socioeconomic differences in blood

373pressure and shows that inequalities arise in early childhood which were mainly explained by

374socio-economic inequalities in the child’s BMI and physical activity. No differences in

375arterial stiffness or cardiac structures were observed per educational subgroup. However, a

376socioeconomic gradient in fractional shortening was observed, with children from low SEP

377families having a higher fractional shortening as compared to children from high SEP

378families. This might be a first cardiac adaptation to a higher blood pressure and a higher BMI.

379The results require careful interpretation since the associations were relatively small and the

380clinical importance not yet fully understood. 37 19

38 Supplementary information is available at http://www.oup.com/ajh.

381Acknowledgements

382The Generation R Study is conducted by the Erasmus Medical Centre in close collaboration

383with the Erasmus University Rotterdam, School of Law and Faculty of Social Sciences, the

384Municipal Health Service Rotterdam area, Rotterdam, the Rotterdam Homecare Foundation,

385Rotterdam, and the Stichting Trombosedienst & Artsenlaboratorium Rijnmond (STAR),

386Rotterdam. We gratefully acknowledge the contribution of general practitioners, hospitals,

387midwives and pharmacies in Rotterdam. We also thank Wim Helbing for his valuable

388comments on an earlier draft of this manuscript.

Funding

The first phase of the Generation R Study is made possible by financial support from:

Erasmus Medical Centre, Rotterdam, Erasmus University Rotterdam and the Netherlands

Organization for Health Research and Development (ZonMw).

Conflict of interest

394 39 20

40 395References