Objective
The objective of the study was to examine the association between physical exercise during pregnancy and fetal growth measures.
Study Design
Data on 79,692 liveborn singletons from the Danish National Birth Cohort were collected between 1996 and 2002. Mean differences in birthweight, length, ponderal index, head and abdominal circumference, and placental weight and hazard ratios of small- and large-for-gestational-age babies were calculated.
Results
Our data indicated smaller babies in exercising women compared with nonexercisers, but the differences were small, and only a few were statistically significant. Exercising women had a slightly decreased risk of having a child small for gestational age (hazard ratio, 0.87; 95% confidence interval, 0.83–0.92) and large for gestational age (hazard ratio, 0.93; 95% confidence interval, 0.89–0.98).
Conclusion
The findings do not indicate sizable effects on fetal growth measures related to exercise apart from a modest decreased risk of small- and large-for-gestational-age infants. These findings do not speak against advising pregnant women to be physically active during pregnancy.
In line with an increasing focus on physical activity as a preventive action against obesity and obesity-related diseases, national guidelines in many countries now recommend a substantial level of physical activity during pregnancy. Hence, in Denmark, a minimum of 30 minutes of moderate physical activity per day is recommended for healthy pregnant women. Exercise in pregnancy causes an acute reduction in oxygen and nutrient delivery to the placental site. On the other hand, maternal blood volume, cardiac output, and placental function have been found to be increased in exercising women, and the question remains whether these mechanisms adequately provide what the fetus needs of oxygen and nutrients.
Size at birth is a function of time spent in utero and the fetal growth rate. We already studied the association between physical exercise and preterm birth using data from the Danish National Birth Cohort and found a slightly decreased risk among exercising women compared with nonexercisers. The present study addresses fetal growth, which is an indicator of both fetal health and health later in life.
With a few exceptions existing studies on exercise and fetal growth have been small and carried out on highly selected groups of women. Two metaanalyses reported either no or only minimal differences in mean birthweight in offspring of exercising mothers compared with those of nonexercisers, and so did 3 observational studies of varying size (n = 148-7101), although lower birthweights were observed with vigorous exercise in late pregnancy. A Cochrane review on 11 randomized trials reported that existing data were not sufficient to “infer important risks or benefits for the mother or infant.”
Most studies have addressed birthweight or small for gestational age (SGA), and a few have included birth length or placental weight. It is recognized that fetal growth may be impaired, even if birthweight is within normal limits, and more sensitive markers of growth may therefore be needed. Small abdominal circumference is associated with liver size and possibly high serum cholesterol later in life, and small head circumference correlates with brain size and has been associated with lower IQ, high blood pressure, and higher risk of cardiovascular disease as well as impaired glucose tolerance.
The aim of this study was to examine the association between physical exercise during pregnancy and fetal growth measures: birthweight, length, ponderal index, abdominal and head circumference, and placental weight.
Materials and Methods
The Danish National Birth Cohort Study (DNBC) is a cohort of a little more than 100,000 pregnancies with written informed consents from the mothers. The initial data collection included telephone interviews, questionnaires, and blood samples and took place between 1996 and 2002.
For this study, we used data from 2 pregnancy interviews carried out at approximately 16 and 31 completed weeks of gestation. More details about the cohort are presented elsewhere. Some of the women provided data on more than 1 pregnancy, and these data were used to do a study among siblings using differences in physical exercise between the pregnancies as the exposure.
The Scientific Ethical Committee has approved the data collection in the Danish National Birth Cohort research database. Approval to use data from the birth cohort for the present study on preterm birth has been obtained. The Danish Data Protection Agency has approved storage, handling, and linkage of data. All of the participants signed an informed consent form before being included in the study.
Study population
A total of 100,418 pregnancies were recruited to the DNBC, and only pregnancies in which we had data from the first pregnancy interview were included (n = 90,165). Reasons for not having a first interview would be having an induced or spontaneous abortion before the interview was done or that we did not obtain contact with the woman for the first interview.
The following exclusions were made: molar or ectopic pregnancies (n = 18), miscarriages (n = 892), early induced abortions (n = 4), late induced abortions (n = 173), stillborn singletons (n = 285), multiple pregnancies (n = 1965), mother emigrated (n = 36), mother deceased in pregnancy (n = 1), and unknown outcome (n = 8), which resulted in 86,783 liveborn singletons. Furthermore, 503 were excluded because of missing or implausible birthweight data according to a method described by Alexander et al.
A total of 1004 pregnancies were excluded because of self-reported diabetes in or before pregnancy. Finally, only the first enrolled child of each mother was included to avoid nonindependent observations (5584 pregnancies excluded), except in the sibling study. Thus, the final dataset comprised 79,692 liveborn infants.
Exposure measures
In the first and second pregnancy interview, the women were asked the following questions: (1) “Now that you are pregnant, do you engage in any kind of exercise?” In the case of a positive answer, the following questions were posed: (2) “What kind of exercise do you engage in?”; (3) “How many times a week do you engage in … (answer in question 2)?”; (4) “How many minutes a time do you engage in … (answer in question 2)?”; and (5) “Do you engage in other types of exercise?” A positive answer to the last question released a loop with the aforementioned questions until a negative response was given. All questionnaires are available in English at www.bsmb.dk .
For each of the 2 pregnancy interviews, the total number of minutes spent on exercise per week was categorized into hours per week. For additional analyses, the active women were assigned to 1 of the following types of activities defined as the type of activity performed more than 50% of the total activity time: swimming, low-impact activities (aerobics/gymnastics for pregnant women, aerobics/gymnastics, dance, walking/hiking, yoga), high-impact activities (jogging, ball games, racket sports), workout/fitness training, bicycling, horseback riding, and a nonclassifiable category.
Data on gestational age at birth (days), mother’s age at conception (years), and sex of the offspring were taken from the hospital birth record obtained from the National Discharge Registry. The other covariates came from the first pregnancy interview and were prepregnant body mass index (kilograms per square meter), occupational status, parity, and smoking, coded as presented in Table 1 .
| Variable | % | Amount of exercise (h/wk) | Mean birthweight, g | SGA | LGA | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | >0 to ≤1 | >1 to ≤2 | >2 to ≤3 | >3 to ≤4 | >4 to ≤5 | >5 | |||||
| n | – | 49929 | 10704 | 8156 | 4447 | 2551 | 1494 | 2236 | – | 7797 | 8268 |
| Age at conception, y | |||||||||||
| <25 | 13 | 14 | 13 | 12 | 13 | 14 | 13 | 16 | 3508 | 16 | 11 |
| 25 to <35 | 75 | 74 | 78 | 78 | 76 | 76 | 76 | 73 | 3586 | 72 | 76 |
| 35 to <40 | 10 | 11 | 8.8 | 9.3 | 9.9 | 9.3 | 9.5 | 9.4 | 3585 | 9.9 | 12 |
| 40+ | 1 | 1.1 | 0.7 | 0.8 | 1.1 | 1.1 | 1.4 | 1.5 | 3552 | 1.3 | 1.3 |
| Parity | |||||||||||
| Nulliparous | 50 | 44 | 58 | 61 | 61 | 65 | 62 | 67 | 3485 | 65 | 25 |
| Pluriparous | 50 | 56 | 42 | 39 | 39 | 35 | 38 | 33 | 3668 | 35 | 65 |
| Sociooccupational status | |||||||||||
| Higher grade professionals | 8.9 | 8.1 | 10 | 10 | 10 | 10 | 10 | 9.1 | 3609 | 7.8 | 9.2 |
| Lower grade professionals | 27 | 25 | 30 | 30 | 32 | 32 | 32 | 27 | 3609 | 24 | 29 |
| Skilled workers | 19 | 19 | 20 | 18 | 17 | 15 | 15 | 13 | 3569 | 19 | 18 |
| Unskilled workers | 26 | 28 | 22 | 20 | 19 | 18 | 20 | 25 | 3543 | 29 | 25 |
| Students | 14 | 12 | 14 | 16 | 18 | 19 | 18 | 19 | 3567 | 14 | 13 |
| Out of work a | 5.1 | 5.8 | 3.5 | 3.8 | 3.8 | 4.2 | 4.1 | 5.5 | 3557 | 5.3 | 5.6 |
| Nonclassifiable | 1 | 1 | 0.9 | 0.9 | 0.8 | 1.2 | 0.7 | 0.8 | 3519 | 1.1 | 0.8 |
| Prepregnancy body mass index | |||||||||||
| <18.5 | 4.5 | 4.9 | 4 | 3.4 | 3.7 | 4.2 | 3.7 | 5.2 | 3350 | 8.4 | 1.6 |
| 18.5 to <25 | 68 | 67 | 70 | 71 | 72 | 74 | 75 | 73 | 3556 | 70 | 60 |
| 25 to <30 | 19 | 20 | 19 | 18 | 18 | 16 | 16 | 16 | 3651 | 16 | 25 |
| 30+ | 8 | 8.7 | 7.6 | 7 | 6.5 | 6 | 4.8 | 5.8 | 3678 | 6.5 | 14 |
| Smoking in early/midpregnancy, cig/d | |||||||||||
| 0 | 74 | 70 | 80 | 79 | 78 | 78 | 81 | 76 | 3617 | 60 | 81 |
| 1 to <10 | 15 | 15 | 13 | 13 | 14 | 13 | 13 | 15 | 3490 | 20 | 11 |
| 10+ | 12 | 14 | 7.7 | 7.6 | 8 | 8.6 | 6.6 | 8.8 | 3421 | 20 | 8 |
Outcome measures
From the National Discharge Registry, we had data on birthweight (grams) length (centimeters), ponderal index (weight in grams*100 divided by length in cubic centimeters), abdominal and head circumference (centimeters), and placental weight (grams). For practical reasons we use the overall term fetal growth measures for the endpoints in the study, acknowledging that these measures are no more than proxies or indicators of fetal growth.
SGA and large for gestational age (LGA) were calculated as the 10th percentile or below and as the 90th percentile or above of the sex- and gestation-specific birthweight within the present study population. Analyses were repeated using an intrauterine fetal weight standard.
Statistical analysis
We examined the association between amount of exercise in pregnancy and birthweight, length, ponderal index, abdominal and head circumference, and placental weight, respectively, by fitting linear regression models. Both exercise data from the first and the second pregnancy interview were analyzed, but only data from the first interview are presented in the tables. Likelihood ratio tests for interaction between exercise and parity, sex of the offspring, and smoking were made.
In addition, a sibling analysis was made to examine whether changes in exercise between 2 pregnancies correlated with birthweight measures. In the case of more than 2 liveborn children in the cohort (n = 63), only the first 2 pregnancies were included. Hence, 5521 pairs of siblings were studied.
Five categories of change in exercise between pregnancies were generated: from much to nothing (–2), from much to little or from little to nothing (–1), no change (0), from nothing to little or from little to much (+1), and from nothing to much (+2). Little exercise was defined as less than 90 min/wk and much exercise as 90 minutes or more. Adjustment for time-varying variables was made. Those were the same as in the main analyses except for occupational status. Covariates were modeled as in main analyses except that each covariate was included in the model with values from both the first and the second pregnancy.
Finally, we estimated the hazard rate of SGA and LGA in the offspring as a function of gestational days by means of a Cox regression model. Using a model for the hazard rate (rather than, eg, logistic regression) has a number of advantages. First, gestational age is directly incorporated in the model; second, it enables us to take the different times of entry into follow-up into account; and lastly, it makes it possible to update exposure data in cases in which a second interview is available. Follow-up ended at birth, emigration, or maternal death. An event was defined as delivery of a liveborn SGA or LGA baby, respectively. If a second interview was available, exercise data were updated by the time of the second interview.
Analyses on fetal growth measures and SGA/LGA were further carried out on a population without possible clinical contraindications for exercise. In these analyses the following exclusions were made: not having a second pregnancy interview (because most pregnancy complications were reported here), metabolic disorders, hypertensive disorders (including gestational hypertension and preeclampsia) or other serious illnesses, vaginal bleeding, painful contractions, loss of amniotic fluid, and cervical incompetence or dilatation.
Gestational age at birth in days was modeled as quadratic splines in all linear regression models. To avoid too much weight on extreme values, maternal age and prepregnant body mass index were modeled as restricted quadratic splines.
An alternative to this way of adjustment for gestational age would be to analyze gestational age-specific z -scores for the fetal growth measures. However, we believe that the interpretation of regression coefficients is more direct when they are expressed in grams, centimeters, and so on rather than in SDs. The remaining adjustment variables were included as categorical variables and categorized as displayed in Table 1 . The decision on which factors to adjust for was made a priori. All analyses were carried out using SAS statistical software (version 9; SAS Institute, Cary, NC).
Results
Table 1 shows some characteristics of the women according to exercise, mean birthweight, SGA, and LGA. In 37% of 79,692 pregnancies, the mother engaged in leisure time physical activity at the time of the first pregnancy interview. This proportion decreased to 31% at the time of the second pregnancy interview. Compared with active women, nonexercisers had lower sociooccupational status; were more often parous, overweight, or obese; and were more likely to smoke 10 cigarettes per day or more. Compared with other physically active women, heavy exercisers (>5 h/wk) were more likely to be in the outer age groups, nullipara, unskilled workers, students or out of work, and underweight. Women who had given birth before, women with higher sociooccupational status or higher body mass index, and nonsmoking women tended to have offspring with larger birthweights. SGA in the offspring was more common among mothers who were young, nulliparous, smokers, or with low body mass.
Table 2 presents mean differences in fetal growth measures among women engaged in exercise during early/midpregnancy compared with nonexercisers. In the analysis only adjusted for gestational age, birthweight seemed to decrease with increasing amount of exercise, but this association almost disappeared after controlling for potential confounding factors.
