Article Text

Download PDFPDF

No relation between coronary artery disease or electrocardiographic markers of disease in middle age and prenatal exposure to the Dutch famine of 1944–5
  1. L H Lumey1,
  2. Lauren H Martini1,
  3. Merle Myerson1,2,
  4. Aryeh D Stein3,
  5. Ronald J Prineas4
  1. 1Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York, USA
  2. 2Cardiovascular Disease Prevention program, St Luke's-Roosevelt Hospital Center, Columbia University, New York, New York, USA
  3. 3Hubert Department of Global Health, Rollins School of Public Health, Emory University, Atlanta, Georgia, USA
  4. 4Division of Public Health Sciences, Wake Forest University School of Medicine, Winston Salem, North Carolina, USA
  1. Correspondence to Dr L H Lumey, Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10032, USA; lumey{at}


Objective and Setting Associations between prenatal famine and coronary artery disease (CAD) have been examined before with inconsistent results. For further evaluation, we examined multiple cardiac risk markers in adult men and women with prenatal exposure to the Dutch famine of 1944–5.

Design Birth cohort study of 407 men and women with prenatal famine exposure, 344 born before or after the famine as time controls, and 324 unexposed siblings as family controls. Study subjects underwent standardised interviews and clinical examinations at age approximately 58 years.

Outcome Measures CAD events from medical history and medical and electrocardiographic (ECG) markers of CAD risk, including 10-year (Framingham) estimates for myocardial infarction and coronary heart disease death, major and minor ECG abnormalities, ECG estimates of left ventricular hypertrophy and left ventricular mass, cardiac autonomic neuropathy measures including the QT index, resting heart rate, heart rate variability from subsequent N–N intervals, and ECG markers of minor T-wave abnormalities, changes in QRS/T frontal plane angle and ST-segment.

Results No increase was seen in CAD risk (HR 1.17; 95% CI 0.73 to 1.88), Framingham risk (OR 1.14; 95% CI 0.90 to 1.44) or in ECG outcomes, adjusting for age and sex. Left ventricular mass estimated with body size was elevated by 3.78 g (95% CI 0.91 to 6.64) after prenatal famine, but showed a 0.65 g decrease (95% CI −2.63 to 1.34) when adjusted for body mass index.

Conclusions We see no relation between prenatal famine and adult CAD, Framingham risk, or any ECG predictors of increased cardiac disease risk.

  • Allied specialities
  • arrhythmias
  • atrial fibrillation
  • cardiac autonomic nervous system
  • cohort study
  • coronary artery disease
  • DOHaD
  • electrocardiography
  • epidemiology
  • Framingham CHD risk score
  • imaging and diagnostics
  • prenatal exposure delayed effects
  • prenatal nutrition
  • psychology/psychiatry
  • public health
  • quality of care and outcomes
  • risk factors
  • sick sinus syndrome
  • subclinical ECG abnormalities
  • sudden adult death syndrome

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Birth weight is inversely associated with coronary heart disease (CHD),1 blood pressure level2 and insulin resistance, type 2 diabetes mellitus and glucose intolerance.3–5 The nature of the relationship between low birth weight and clinical disease is not identifiable from those studies.6

The Dutch famine of 1944–5 at the end of World War II provides an opportunity to assess the impact of this exogenous stress during pregnancy on later outcomes. The famine was the result of a transport embargo imposed by the German occupying forces in early October 1944, which limited food supplies to the western part of the country. The severity and widespread nature of the famine have been documented elsewhere.7–9 Food rations, which had been stable at approximately 1800 kcal/day per person over the duration of the war, rapidly deteriorated and fell below 900 kcal/day by 26 November 1944, and were as low as 500 kcal/day by April 1945.10 The famine ended with the German surrender in May 1945 and rapid distribution of allied food supplies across the country.

Increases in coronary artery disease (CAD) have been reported after prenatal exposure to the Dutch famine11 ,12 but the results have been inconsistent.13 Our objective therefore was to test in a new study population whether or not there is a positive association between prenatal famine exposure and CAD, the Framingham risk score and selected electrocardiographic (ECG) markers of subclinical cardiac morbidity and mortality.

Participants and methods

Study population

We identified all available live-born singleton births at three institutions in famine-exposed cities in the western Netherlands.14 We selected all 2417 births between 1 February 1945 and 31 March 1946 (infants whose mothers were exposed to the famine during or immediately preceding that pregnancy). We sampled 890 births from 1943 to 1947 (infants whose mothers were not exposed to famine during this pregnancy) to serve as comparison groups. This figure was chosen to have adequate numbers of controls, with random selection of equal numbers in each birth month.

We provided the names and addresses at birth of all 3307 infants to the population register in the municipality of birth. Of these, 308 persons (9%) were reported to have died and 275 (8%) to have emigrated. For 294 persons (9%) a current address could not be located, and the population registry in Rotterdam declined to trace 130 individuals born out of wedlock. A current address was obtained for 2300 individuals (70% of the birth cohort).

All 2300 traced persons were invited by mail to participate. For optimal bias control, our study design originally called for the recruitment of same-sex sibling pairs, and the lack of an available sibling was a reason for ineligibility. We received a reply from 1767 persons, of whom 347 expressed willingness to participate together with a sibling. Of the 1420 who declined, 67% (951) reported not having a same-sex sibling available for study. Subsequently, we attempted also to enrol these 951 persons; 381 agreed to participate. This resulted in a total of 1075 individuals (751 from the hospital series and 324 of their siblings) (figure 1).

Figure 1

Flow chart of study participation.

All study protocols were approved by the research ethics committees of the participating institutions. Study participants provided verbal consent at the start of the telephone interview and written informed consent at the start of the clinical examination in 2003–5.

Telephone interviews

A 1-h telephone interview included questions on sociodemographic characteristics, health-related behaviours including smoking and drinking, the use of blood pressure and cholesterol medications, and information on CAD outcomes including the Rose/WHO angina questionnaire15 and a history of myocardial infarction (MI), coronary angioplasty and bypass surgery. Following the telephone interview, we scheduled a visit to the Leiden University Medical Center. Clinical examinations for the reported study outcomes were completed on 962 study participants (figure 1). Most clinical examinations were conducted within 6 weeks of the telephone interview.

Clinical examinations and measures

Participants were scheduled to arrive early in the morning and were instructed not to eat solid foods after 22:00 hours the previous evening.

Following registration and obtaining of informed consent, we collected fasting venous blood samples, including total serum cholesterol and high-density lipoprotein (HDL) cholesterol. These were measured promptly by standard enzymatic methods. We defined dyslipidemia as a total serum cholesterol/HDL-cholesterol ratio greater than 5.0 or the use of cholesterol-lowering medication.16

We measured height to the nearest 1 mm with a portable stadiometer (SECA, Hamburg, Germany), weight to the nearest 100 g with the participant standing in underclothing without shoes on a portable digital scale (SECA) and waist circumference to the nearest 1.0 cm at the level of the iliac crest and intersection with the midaxillary line. A single measurement was taken for height and weight. Two measurements were taken for waist circumference, with the mean value of the two taken for analysis unless these were more than 1.0 cm apart, in which case a third and fourth measure were taken and the three measures closest together from the available four were averaged.17 Body mass index (BMI) was calculated as weight(kg)/height(m)2.

Blood pressure was measured with an automated sphygmomanometer (Omron HEM 705-CP, Bannockburn, Illinois, USA). Three readings were obtained using the automatic setting from the non-dominant arm after several minutes of rest. In analyses, the mean of the two closest readings was used.18 Hypertension was defined as having a systolic blood pressure of 140 mm Hg or greater or a diastolic blood pressure of 90 mm Hg or greater or a history of having been prescribed antihypertensive medication.

We then recorded three immediately sequential resting ECG using the GE/Marquette MAC 1200 portable electrocardiograph (Depex b.v., Houten, The Netherlands). The recording technicians were trained to make a special effort to reduce chest electrode placement errors, thereby reducing interindividual variability and improving the consistency of serial ECG recordings.19 Each 10-s segment of simultaneous standard 12-lead ECG recording was sampled at a rate of 250 samples per second per lead. ECGs were recorded and stored electronically in the electrocardiograph. At the end of the examination session they were digitised and transmitted over analogue phone lines to the Central ECG Reading Center (EPICARE Center, Wake Forest University, Winston-Salem, North Carolina, USA) daily for visual inspection for technical errors and adequate quality. A copy of the ECG was printed and read promptly by Dr BJM Delemarre, Leyenburg Ziekenhuis, the Hague, The Netherlands, to detect significant clinical ECG abnormalities. The participant's family physician was notified when indicated.

The ECG data were processed using the 2001 version of the GE Marquette 12-SL program (Marquette 12SL ECG Physician Guide, available at ECGs were classified according to the Novacode and Minnesota codes, using variables derived from the median complex of the Marquette measurement matrix, and standard and derived indices of continuous measurements of duration and voltage were established.20–22

CAD was defined to include one or more of the following: a history of MI diagnosed at hospital admission, a history of angioplasty or bypass surgery, CAD-positive symptoms using the Rose questionnaire,15 or study ECG Q-wave abnormalities (Minnesota codes 1.1.1–1.2.5 and 1.2.7).21 Age at onset of CAD was calculated from the year of birth and the earliest of year at MI diagnosis, angioplasty, or bypass surgery, or age at interview.

Ten-year (Framingham) risk for ‘hard CHD’ (MI and CHD death) was estimated based on Framingham risk scoring as defined in tables 3.1–5 and 3.1–6 of the updated Adult Treatment Panel III criteria.23 ,24 Framingham risk points were calculated by summing sex-specific risk points for age, systolic blood pressure, total cholesterol, HDL-cholesterol and current smoking status.

Major and minor ECG abnormalities were classified hierarchically. We first identified participants with major ECG abnormalities and then among the remaining participants identified minor ECG abnormalities. Criteria for major ECG abnormalities included the following diagnoses: atrial fibrillation or flutter (Novacode 1.5); high-degree atrioventricular dissociation (Novacode 2.3.1 and 2.3.2); left bundle-branch block (Novacode 3.1.0 and 3.1.1); right bundle-branch block (Novacode 3.2.0); indeterminate conduction delay (Novacode 3.3.0 and 3.3.1); Q-wave MI (Novacode 5.1, 5.2, 5.3 and 5.4); isolated ischaemic abnormalities (Novacode 5.5 and 5.6); left ventricular hypertrophy (LVH) with ST-T abnormalities (Novacode 6.1.1); and other miscellaneous arrhythmias: supraventricular tachcycardia, ventricular preexcitation, ventricular tachycardia (Novacode 1.4, 1.7, 1.8, 1.9, 2.4). Minor ECG abnormalities included: first and second-degree atrioventricular block (Novacode 2.1 and 2.2.1); borderline prolonged ventricular excitation (Novacode 3.4.1 and 3.4.2); prolonged ventricular repolarization (Novacode 4.1.1 and 4.1.2); isolated minor Q and ST-T abnormalities (Novacode 5.7 and 5.8); LVH without ST-T abnormalities (Novacode 6.1.0); left atrial enlargement (Novacode 7.1); frequent atrial or ventricular premature beats (Minnesota code 8.1); and fascicular blocks (Novacode 10.1 and 10.2).

Measures of LVH and left ventricular mass (LVM) were calculated as continuous ECG-derived measures from gender-specific individual QRS complex measurements and body weights. Elevated LVH was defined as a Cornell voltage (CV, a function of the R-wave and S-wave amplitudes in defined ECG leads) of 2800 μV or greater in men and 2200 μV or greater in women. For LVM (g) estimates, we used a function of the individual's weight, together with the CV and Sokolow and Lyon (SL) voltage, and the ST-segment elevation or depression around the J-point in V5 (JDV5). For men, LVM was estimated as 0.023*CV + 0.01*SL + 1.32*weight (kg) + 10.6 and for women as 0.016*CV + 0.27*JDV5 + 1.16*weight (kg) + 35.3.25 ,26

Cardiac autonomic neuropathy (CAN) was assessed by measures of QT index (QTI),27 resting heart rate and heart rate variability (HRV). Two ECG time-domain measures of HRV were calculated: the root mean square of successive differences of N–N intervals (RMSD) and the SD of N–N intervals (SDNN). The calculation of these measures was verified by senior electrocardiographers and a biostatistician in a test data set of 264 ECGs. Other subclinical ECG markers of disease included minor T-wave abnormalities (Minnesota codes 5.3 and 5.4)28; QRS/T frontal plane angle; and continuous measurement of ST-segment elevation or depression, either at the J-point or 60 ms past the J-point of the ST segment. These markers are all independent predictors of future CAD and total mortality.29–35

Exposure to famine

We used the date of the last menstrual period (LMP) as noted in the hospital records to define the start of gestation unless it was missing or implausible (12%). In those cases we estimated the LMP date from relevant annotations on the birth record and approximate gestational age from birth weight and date of birth, using cut points from tables of gender, parity and birth weight-specific gestational ages from the combined birth records of the Amsterdam midwives school (1948–57) and the university obstetrics department (1931–65).14 ,36 We characterised exposure to famine during gestation by determining the gestational ages (in weeks after the LMP) during which the mother was exposed to an official ration of less than 900 kcal/day. We considered the mother exposed in gestational weeks 1–10, 11–20, 21–30, or 31 to delivery if these gestational time windows were entirely exposed. Therefore, pregnancies with LMP between 26 November 1944 and 4 March 1945 were considered exposed in weeks 1–10; between 18 September 1944 and 24 December 1944 in weeks 11–20; between 10 July 1944 and 15 October 1944 in weeks 21–30; and between 2 May 1944 and 24 August 1944 in weeks 31 to delivery. By these definitions, any participant could have been exposed to famine during at most two adjacent 10-week periods. Individuals exposed in at least one of the 10-week periods were considered to have had any prenatal famine exposure.

We also classified famine exposure based on the date of birth alone to facilitate comparisons with previous reports.11 ,12

Statistical methods

We computed means with SD, categorical distributions, or medians with 25th and 75th percentile values for selected exposure groups as appropriate for quantitative or categorical variables, and assessed differences among exposure groups with analysis of variance, χ2 tests, or Kruskal–Wallis rank ordered tests.

We used binary and ordinal logistic regressions to estimate the odds of prevalent CAD, ECG abnormalities or Framingham risk in relation to famine exposure and Cox proportional hazards models to incorporate information on age at CAD onset. We analysed all quantitative outcomes in their original units and also after natural log transformations to improve normality of the distributions of resting heart rate, LVM, RMSD, SDNN and QRS/T frontal plane angle. These transformations confirmed analyses in original units and are not reported separately. We carried out gender-specific analyses and evaluated possible significant interactions between exposure and stage of pregnancy and gender. In the absence of such interactions (all had p>0.10) we present pooled models with adjustment for gender. We included age at examination (with linear and quadratic terms) in all regressions. Additional adjustments for birth weight, education, smoking, drinking, prevalent hypertension, waist circumference and dyslipidaemia were carried out to assess potential confounding or intermediary variables. These adjustments did not alter the estimates and are not reported further. We used the combined population of unexposed births in the three hospitals and all siblings of the birth series as the reference.

Statistical analyses were conducted with SPSS V.10 and with STATA V.11, using the xtlogit, xtcox and ologit commands (with the vce cluster option) to account for within-family clustering when indicated. Associations of study outcomes with the four 10-week exposure periods taken as a set were evaluated with a 4 degree of freedom Wald test.


We obtained health histories by telephone from 1075 and complete medical examinations and ECG recordings from 962 study participants (figure 1). There were differences between famine-exposed and unexposed women in prevalent hypertension and in BMI and waist circumference (table 1). In addition, sibling controls were on average 1.5 years younger than other study participants.

Table 1

Selected demographic characteristics in 2003–5 by famine exposure category in 1944–5

The 70 individuals with prevalent CAD (7%) (table 2) were diagnosed on the basis of a history of bypass surgery (n=20), angioplasty (n=18), a hospital confirmed MI (n=6), angina-specific symptoms on the Rose/WHO questionnaire (n=11) or MI-specific ECG findings (n=25). Some individuals were diagnosed with more than one of the above conditions. The estimated 10-year Framingham risk for ‘hard’ CHD outcomes was 10–12% in men and 2–3% in women. Major ECG abnormalities were identified among 10% of examinees and minor ECG abnormalities among 29%, but the prevalence was unrelated to prenatal famine exposure. Other ECG outcome measures were also unrelated to the famine exposure category, except for LVM and HRV as expressed by SDNN.

Table 2

Cardiac disease outcomes and risk factors in 2003–5 by famine exposure category in 1944–5

In regression analyses adjusting for age and gender, comparing famine-exposed study participants with hospital and sibling controls (table 3), no relation was seen between famine and the prevalence odds of CAD (OR 1.11; 95% CI 0.64 to 1.92), the risk of CAD (HR 1.17; 95% CI 0.73 to 1.88), or the odds of major or minor abnormalities or other ECG outcomes under consideration except one. Our estimate of LVM showed an increase of 3.78 g (95% CI 0.91 to 6.64) after prenatal famine exposure. The estimate was no different from zero, however, after adjustment for BMI, showing a decrease of 0.65 g (95% CI −2.63 to 1.34) (data not presented in table). The OR for minor T waves in individuals with any prenatal famine exposure relative to controls was 1.78 (95% CI 1.05 to 3.02). No ECG measures were statistically associated with any of the 10-week gestational periods alone or combined as a set (by Wald test).

Table 3

CAD, Framingham risk estimate, and selected ECG predictors of cardiac morbidity and mortality in 2003–5 among subjects with prenatal famine exposure in 1944–5 and unexposed controls

To validate the risk related to our selected subclinical ECG markers we related these to other CHD risk factors in our population. Serum cholesterol levels increased with heart rate, QTI and QRS/T angle while the reverse was seen for SDNN and RMSD. ECG abnormalities were also related to adverse levels of HDL-cholesterol (although less markedly than for total cholesterol) and systolic and diastolic blood pressure, BMI and waist circumference (data not shown).

For comparison with a previous study in the Dutch famine setting11 we present in table 4 our findings for the famine exposure categories in late, mid and early pregnancy. We see no significant association between exposure in any trimester of pregnancy and CAD prevalence. For CAD, the HR after prenatal famine exposure is 1.26 (95% CI 0.59 to 2.70), based on eight cases in this subgroup. In a combined analysis of the 19 cases from both studies with prenatal famine in early gestation and control subjects conceived before or after the famine, the OR was 1.38 (95% CI 0.79 to 2.38; p=0.29).

Table 4

CAD risk among subjects with famine exposure in late, mid and early gestation relative to controls


In this study of 1075 men and women born in Amsterdam, Rotterdam and Leiden towards the end of World War II and unexposed siblings, we found no evidence for an association between prenatal exposure to the 1944–5 Dutch famine and CAD, the Framingham CHD risk score, or multiple ECG measures of CAN. The strengths of the study include the unique cohort, examinations under well-defined conditions and a measure of famine exposure in specified periods of pregnancy.

We did not confirm earlier reports from The Netherlands of a threefold increase in CAD12 or an earlier onset of CAD11 after famine exposure during early gestation. These authors suggested that the reported increase may relate to changes in glucose and lipid mechanisms, but this explanation is not consistent with other findings in this population.13 We interpret the earlier positive reports as chance findings from a subgroup analysis of limited sample size in need of independent confirmation. Taken together, our study and earlier reports do not provide compelling evidence for changes in cardiovascular outcomes after prenatal famine exposure.

There are minor differences in CAD definitions between this study and earlier work. We defined abnormal Q waves by Minnesota codes 1.1.1–1.2.5 and 1.2.7, whereas earlier work used codes 1.1–1.2. We also included individuals with a history of MI as CAD cases. These small coding differences are not a likely explanation for the differences between the two studies.

Framingham risk estimates are a well-established CHD prediction tool.37 ,38 The similarity of Framingham risk estimates in famine-exposed individuals and unexposed controls provides further support to our overall findings that there are no significant associations between prenatal famine exposure and cardiovascular risk profile.

Major ECG abnormalities including pathological Q-waves have been associated with a higher risk of total mortality, cardiac disease and mortality, non-fatal cardiac events and stroke.14 ,39–41 This association is independent of other known cardiac disease risk factors. No associations were seen, however, between prenatal famine and major ECG abnormalities.

Minor ECG abnormalities may be an important early sign of cardiac risk. In the West of Scotland Coronary Primary Prevention Trial, the presence of multiple risk factors including minor ECG abnormalities doubled the risk of CHD events.42 No association was seen, however, between prenatal famine and minor ECG abnormalities.

In the Cardiovascular Health Study, LVM was predictive of cardiac heart mortality in women but not in men.43 We found an association between prenatal famine exposure and LVM defined by ECG measures and an individual's weight26 but not for CV alone. After adjustment for body size, however, LVM was no longer associated with prenatal famine. This is a reflection of the finding that BMI, weight and waist circumference in this population are elevated among women with prenatal famine exposure.17 We interpret the increase in minor T-wave abnormalities after prenatal famine exposure as chance variation, given the lack of association between famine exposure and other relevant cardiac parameters.

Lower HRV and higher heart rate and QTI indicate poorer autonomic function. These measures are reliable estimates of autonomic function and are recommended as risk predictors for use in large population studies.29 Abnormal HRV is a sensitive marker of CAN that predicts overall mortality and incident CVD events.29–35 We observed no changes in HRV in relation to prenatal famine exposure.

Three additional studies have focused on energy intake in wartime. In England in 1942–4, pregnant women were examined to determine if the wartime rations were sufficient to prevent nutritional deficiencies. Their offspring were later traced and examined.44 In that study, pregnancy nutrition did not reach starvation levels. In the second study, men and women born during the siege of Leningrad (1941–4) were examined.45 No long-term effect on cardiovascular outcomes was seen in either study. In the third study, CHD and stroke were increased in women aged 49–70 years who reported to be severely exposed to the Dutch famine at the age of 10–17 years.46 This study, however, does not speak to prenatal famine exposure.

We assigned famine exposure based on the date of mothers' LMP from the birth record. In contrast, previous studies of birth cohorts from the Dutch famine have mostly used date of birth, assuming a gestation period of 40 weeks for all participants across exposure categories. This might lead to some misclassification, especially of early pregnancy exposures.14 When reclassified according to the date of birth as in previous work on CAD11 our data still show no association with famine exposure. This again suggests that earlier positive reports from the Dutch famine may have been based on chance findings in selected subgroups.

Our negative conclusions are based on finding small insignificant differences in discrete and continuous outcomes, comparing famine-exposed individuals and unexposed controls. Such negative findings are not due to insufficient study power. For continuous outcomes in our study, we had 80% power to detect between-group differences as small as 1/3 SD. We interpret reported positive associations between CAD and early famine as chance findings from underpowered studies as the statistical power to detect the increase in risk was only 40%.

Our study adds unique information on the role of nutrition in pregnancy under acute maternal famine conditions, but does not speak to the possible long-term effects of lesser degrees of chronic undernutrition in pregnancy. Our use of sibling controls to reduce the potential bias related to family-level factors in this area of research is novel. Although extensive recorded information is available from the time of birth and from the time of examination at the age of approximately 59 years, we are limited in that there is no source for intermediary data points except for the individual interview. Not all eligible subjects from the birth series participated in the follow-up examinations. There is no suggestion, however, that the study sample was biased by selective losses to follow-up.14

In conclusion, we see no association between prenatal famine exposure and any of a wide range of clinical cardiovascular outcomes or ECG predictors of future CAD morbidity or mortality risk in middle age.


The authors would like to thank the Vroedvrouwenscholen of Amsterdam and Rotterdam and the obstetrics department of the Leiden University Medical Center in Leiden for their help in accessing their archives and the study participants for their cooperation. The clinical examinations were carried out at the study center of gerontology and geriatrics, Leiden University Medical Center, under the supervision of L de Man. They also thank Dr B J M Delemarre, Leyenburg Ziekenhuis, the Hague, The Netherlands, for reading ECG and providing clinical alerts when indicated.



  • Funding Funding was provided by the US National Institutes of Health (R01 HL-067914).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Ethics approval was provided by Columbia University (USA) and Leiden University (Netherlands) review boards.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Enquiries are welcome for collaborations on future analyses.