The trusted source for
healthcare information and
September 2001; Volume 3; 65-69
By Anthony R. Scialli, MD
Stress can be defined as a physiologic response to an aversive stimulus. Stressful life events produce increases in so-called stress hormones such as cortisol and catecholamines. The idea that physiologic changes in the mother might affect embryo-fetal development has been investigated for decades.
A recent Danish record-linkage study evaluated congenital malformation in a newborn and the occurrence of serious life events during gestation, which included death of or serious illness in a partner or an older child.1 Partners were identified as husbands, legal fathers of the index children, or men living with the women on January 1 of the birth year of the index children. Serious life events were considered exposures, and multiple exposures in the same pregnancy were counted as one exposure, using the date of the earliest exposure.
In a retrospective cohort design, the frequency of malformation among 3,560 exposed pregnancies was compared with 20,299 control pregnancies (a small proportion of women in each group contributed more than one pregnancy).1 Odds ratios (ORs) were calculated and logistic regression analysis was used to adjust for parental age and education, mother’s parity, previous stillbirth and miscarriage, place of birth, maternal epilepsy, maternal diabetes, maternal intoxication, substance abuse, cohabitation, place of residence, and sex of the index child.
Compared to unexposed women, exposed mothers were older, of higher parity, and more likely to be single, and had less formal education and a higher incidence of diabetes, epilepsy, and intoxication during pregnancy. Most exposures were due to death or serious illness in children; illness or death in a spouse or partner was infrequent. Considering all exposed women and all malformations, the adjusted OR was 1.23 (95% confidence interval [CI] 1.02, 1.49). For cranial neural crest-associated malformations, the adjusted OR was 1.54 (CI 1.05, 2.27). Exposure in two consecutive pregnancies gave an adjusted OR of 2.99 (CI 1.08, 8.43). Considering only exposures involving older children, the adjusted ORs for total and cranial neural crest malformations were 1.20 (CI 1.02, 1.58) and 1.81 (CI 1.19, 2.75), respectively. The strongest associations were seen when the exposure was due to the death of an older child, and when only cranial neural crest cell malformations were considered. The adjusted OR for exposure of this sort any time in pregnancy was 1.95 (CI 1.28, 2.97); if exposure occurred during the first trimester (when cranial neural crest cell migration occurs), the adjusted OR was 4.75 (CI 1.63, 13.8). If the child’s death was unexpected (defined as caused by accident, suicide, or murder), the adjusted OR was 8.36 (CI 2.41, 2.97).
Experimental animal studies have yielded conflicting reports on the effects of stress on resorption, litter size, birth weight, and offspring viability. Most efforts to induce birth defects by restraint stress during rat pregnancy2,3 and noise stress during pregnancy in rats and mice4-6 have given negative results.
However, immobilization of pregnant animals for 12 hours on a single day gave rise to offspring with an increase in fused or supernumerary ribs and exencephaly.2 A linear relationship was shown between the induction of supernumerary ribs and maternal weight loss during the experiment, suggesting that the skeletal anomaly may have been due to general toxicity. High-frequency noise stress for several days during pregnancy increased the incidence of exencephaly, cleft lip, open eyes, and fused sternebrae in mouse offspring.7
In some studies of mice, maternal restraint or food deprivation has produced an incidence of cleft palate as high as 69% compared to a 1% rate in controls,8 an effect possibly mediated by elevated corticosteroids.9 It is not clear whether stressors used in animal studies (noise, restraint, heat, or food deprivation) adequately model human psychologic stress.
Difficulties documenting the adverse effects of stress in human pregnancy include problems defining and measuring stress and the confounding interactions of measures used to combat stress, such as prescribed and recreational drugs. One study reported an elevated frequency of negative life events among 111 women who spontaneously aborted a normal fetus.10 The incidence of reported negative life events among 81 women who spontaneously aborted a chromosomally abnormal fetus was used as a standard.10 The small size of this study and the relatively small magnitude of difference between the study groups make the reported findings suggestive at best.
Three different groups of investigators found that mothers of preterm babies (interviewed after delivery) had higher rates of psychologic abnormalities or reported more psychologically adverse life events than mothers of term infants.11-13 Because interviews were conducted after the pregnancy outcomes were known, recall bias may have affected data collection. Two studies did not control for factors such as family income, maternal smoking, ethanol use, nutrition, or prenatal care,11,12 which limit conclusions that can be drawn from them. One survey found that low socioeconomic status, previous preterm delivery, inadequate weight gain, a history of infertility, a history of induced abortion, bleeding from the vagina during the index pregnancy, abnormal placentation, a lack of leisure-time physical activity, ethanol ingestion, and an undesired pregnancy were associated (retrospectively) with preterm delivery.13
Another paper used two methods to avoid biases associated with interviewing women after childbirth.14 First, women were interviewed at 4-6 months’ gestation and prospectively followed for preterm labor. Second, interviews with women delivering preterm or delivering at term after successful treatment for preterm labor were conducted 3-9 months post partum, only when the baby was at home and healthy. Women who experienced premature labor had higher anxiety scores at both mid-pregnancy and several months post partum.14 The authors correctly recognized that this association did not show that maternal anxiety causes preterm labor, but they believed that "psychopathological tendencies," superimposed on other factors, might increase the risk of preterm labor. A subsequent study, which controlled for maternal demographic factors and the use of cigarettes, alcohol, and illicit drugs, also found measures of stress on a standardized instrument to be associated prospectively with a small (16%) but statistically significant increase in risk of premature birth.15 A case-control study of preterm deliveries in North Carolina was unable to show a significant association with job stress.16
Two studies dealing with severe anxiety in pregnant women are provocative. One administered an anxiety test to women in the third trimester and found that mean Apgar scores in the offspring of the severely anxious group were lower than in the normal group.17 This finding in itself is not meaningful, because Apgar is a nonparametric ranking system and does not permit comparisons of group means; however, in the normal anxiety group, all Apgar scores were 8 or higher; in the severely anxious group, Apgars scores ranged from 0 to 9. The authors concluded that extreme maternal anxiety might jeopardize fetal/neonatal well-being. An alternative explanation is that anxious women receive more analgesics during labor.18 Another study, however, found that of 200 women tested during pregnancy, 10 of 11 who demonstrated extreme degrees of anxiety had significant adverse outcomes (four miscarriages, five neonatal deaths, and one baby with congenital anomalies).19
Three other prospective studies deserve consideration. In one, data on psychologic state were collected and a limited group of abnormal obstetric outcomes was identified at the outset, rather than after delivery. This study found no relationship between maternal emotional state and adverse outcome.20 Another study, which set out to investigate a possible association between psychologic state and low birth weight, did not find a relationship with respect to maternal anxiety, but did show that major life events (such as death of a close relative or friend) occurred more often in women who then had low birth weight or premature babies.21 Importantly, the study also identified an association among major life events, smoking, and low socioeconomic class (possible factors in low birth weight and/or prematurity).
A more recent study that surveyed women in mid-pregnancy found an increased incidence of prematurity in women who had experienced severely stressful events (as categorized by the American Psychological Association).22 This report supports previous observations, but did not contain sufficient detail to permit full analysis. Women under stress are likely to alter their dietary patterns, and perhaps to use medications and recreational drugs in attempts to alleviate stress.
Given the difficulties with defining and testing for the reproductive effects of stress,23,24 it is difficult to compare available reports. It should be noted, however, that a prospective study of upper middle class women with a low self-reported incidence of tobacco and ethanol use showed that standard indices of stress during pregnancy correlate negatively with infant birth weight and gestational age.25 Although medical complications of pregnancy occurred in some subjects, stress indices were independently correlated with decreased birth weight and gestational length, making the association between stress and premature delivery more plausible than has been noted in other reports.
The latest record linkage study from Denmark1 represents the finest investigative effort possible, given the limitations of the technique. Scandinavian registries include records of birth, employment, residence, hospitalization, and death, and are virtually 100% complete for all citizens. These computerized, linked registries permit a large amount of data mining. Their major limitation, common to all retrospective record reviews, is that missing information cannot be considered. It is this missing information that ultimately prevents this study from supplying the answer to the question of whether stress during pregnancy causes malformations.
The adjusted ORs suggest a relationship between stressful events, particularly death of an older child, and the diagnosis of a malformation of presumed cranial neural crest cell origin. Most of these abnormalities were cleft lip (with or without cleft palate) and congenital heart defects. Although the cranial neural crest origin of these abnormalities has not been established definitively, particularly for defects involving areas other than the conotruncal regions of the heart, it is acceptable to give the authors the benefit of the doubt in this area. These abnormalities were selected for particular attention because of the presumed sensitivity of the cranial neural crest to the actions of some drugs (e.g., retinoids and ethanol) and because hyperglycemia and hypoxia experimentally interfere with neural crest cell migration. The authors postulate that increased cortisol and catecholamine could cause hyperglycemia and hypoxia.
Although arguments can be made against the authors’ interpretation of the developmental literature, these arguments would not invalidate the empiric results, i.e., the elevated OR is associated with stressful events and the presumed class of cranial neural crest anomalies. After all, a child with an orofacial cleft will not care whether the abnormality is classed as being of cranial neural crest origin and will not be concerned with whether hyperglycemia can be related experimentally to the anomaly.
Critics will object that the method of identifying women as exposed is flawed. A woman whose partner dies may experience less rather than more stress, de-pending on the nature of the relationship with the partner. But surely the death of a child will be experienced as stressful by all or nearly all women, and it is the death of an older child (especially if unexpected), that showed the strongest association with abnormalities of the index infant. To make the case even stronger, the authors excluded events in older children that were associated with malformations also seen in the index infant, thus removing the argument that the association was simply a marker of a genetic tendency toward malformation.
What about other unmeasured stressors, such as job loss or incarceration of a family member? Failure to include these stressors would not invalidate the results of the study because not measuring other stressors would have biased the study toward the null; in other words, some unexposed women actually may have been exposed to stress. If this misclassification bias operated differentially (more unmeasured stressors in exposed than in unexposed women), the general results would stand, although the attribution of death of a child as the most important stressor might be questioned. Pregnancy dating and exposure timing errors also would have biased the results toward the null, unless there was differential misdating of pregnancy between exposed and unexposed women, which is unlikely.
The case for the association between stress and malformation is increased, in fact, by the apparent gradient of response and by its biologic plausibility. Women exposed in two pregnancies gave an OR twice that of the OR for all pregnancies (2.99 vs. 1.23). Women who lost a child during the first trimester (when cranial neural crest cells are migrating) gave a higher OR than women with exposure during the second or third trimesters. Women who lost children in an accident, suicide, or homicide gave the highest ORs, which was consistent with the idea that the death of a child after an illness might produce less acute stress than the unexpected death of a healthy child.
So what could be the limitation of this large and carefully performed study, the results of which make so much sense? It is, in fact, the same limitation of other human studies on stress, which is the inability to separate the stress from the interventions used to treat the stress. People whose children die can be expected to begin or to increase self-medication (e.g., alcohol and cigarettes) and prescribed medication (e.g., barbiturates and benzodiazepines). The use of these agents would be expected to be proportional to the severity of the stress: A woman whose child dies in an accident may be more likely to be medicated or to medicate herself than a woman whose child dies after a long illness. It is true that this study controlled for intoxication, alcoholism, drug abuse, and smoking, but only to the extent that these exposures were recorded in one of the databases. Alcoholism, for example, was based on whether an ICD8 code of 30300 to 30399 appeared in the hospital discharge register during the index pregnancy; heavy alcohol use around the time of the stress would not have been recorded. Cigarette smoking data were available for only two years of the 12-year study.
Could exposure to prescribed and recreational drugs explain the association with neural crest cell-derived abnormalities? Most of these abnormalities were orofacial clefts and heart defects. Ethanol abuse during pregnancy has been associated with heart defects and, to a lesser extent, orofacial clefting. Maternal cigarette smoking has been associated with orofacial clefting in individuals with a variant TGF-alpha isoform. Congenital heart disease and orofacial clefting occur in pregnancies exposed to barbiturate—including anticonvulsant regimens (although it is not known whether the abnormalities are caused by the barbiturate, another anticonvulsant, or the combination). Benzodiazepine use during pregnancy has been associated, although inconsistently, with a small increase in the risk of orofacial clefting.
This study, then, is consistent with the proposition that maternal stress during pregnancy is associated with an increased risk of certain congenital malformations in the offspring, but the association has not been proven to be causal. Are there public health implications of these results? Perhaps so: Pregnant women who are subjected to stress might benefit from nondrug stress-reduction measures, including exercise, yoga, relaxation exercises, meditation, and massage. Although the benefits of these types of interventions were not evaluated in the current study, it is reasonable to infer that stress reduction may do some good and is unlikely to do any harm.
Dr. Scialli, a member of the Alternative Therapies in Women's Health editorial board, is Professor, Department of Obstetrics and Gynecology, Georgetown University Medical Center, Washington, DC.
1. Hansen D, et al. Serious life events and congenital malformations: A national study with complete follow-up. Lancet 2000;356:875-880.
2. Beyer PE, Chernoff N. The induction of supernumerary ribs in rodents: Role of maternal stress. Teratog Carcinog Mutagen 1986;6:419-429.
3. Barlow SM, et al. Delay in postnatal growth and development of offspring produced by maternal restraint stress during pregnancy in the rat. Teratology 1978;18: 211-218.
4. Nawrot PS, et al. Embryotoxicity of various noise stimuli in the mouse. Teratology 1980;22:279-289.
5. Kimmel CA, et al. Teratogenic potential of noise in mice and rats. Toxicol Appl Pharmacol 1976;36: 239-245.
6. Nawrot PS, et al. Embryotoxicity of broadband high-frequency noise in the CD-1 mouse. J Toxicol Environ Health 1981;8:151-157.
7. Cook RO, et al. Effects of high-frequency noise on prenatal development and maternal plasma and uterine catecholamine concentrations in the CD-1 mouse. Toxicol Appl Pharmacol 1982;66:338-348.
8. Rosenzweig S, Blaustein FM. Cleft palate in A/J mice resulting from restraint and deprivation of food and water. Teratology 1970;3:47-52.
9. Montenegro MA, et al. The influence of earthquake-induced stress on human facial clefting and its simulation in mice. Arch Oral Biol 1995;40:33-37.
10. Neugebauer R, et al. Association of stressful life events with chromosomally normal spontaneous abortion. Am J Epidemiol 1996;143:588-596.
11. Gunter LM. Psychopathology and stress in the life experience of mothers of premature infants. A comparative study. Am J Obstet Gynecol 1963;86:333-340.
12. Newton RW, et al. Psychological stress in pregnancy and its relation to the onset of premature labour. Br Med J 1979;2:411-413.
13. Berkowitz GS. An epidemiologic study of preterm delivery. Am J Epidemiol 1981;113:81-92.
14. Omer H, et al. Psychological variables and premature labour: A possible solution for some methodological problems. J Psychosom Res 1986;30:559-565.
15. Copper RL, et al. The preterm prediction study: Maternal stress is associated with spontaneous preterm birth at less than thirty-five weeks’ gestation. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol 1996;175:1286-1292.
16. Brett KM, et al. Employment, job strain, and preterm delivery among women in North Carolina. Am J Public Health 1997;87:199-204.
17. Crandon AJ. Maternal anxiety and obstetric complications. J Psychosom Res 1979;23:109-111.
18. Standley K, et al. Dimensions of prenatal anxiety and their influence on pregnancy outcome. Am J Obstet Gynecol 1979;135:22-26.
19. Grimm ER. Psychological tension in pregnancy. Psychosom Med 1961;23:520-527.
20. Grimm ER, Venet WR. The relationship of emotional adjustment and attitudes to the course and outcome of pregnancy. Psychosom Med 1966;28:34-49.
21. Newton RW, Hunt LP. Psychological stress in pregnancy and its relations to low birth weight. Br Med J 1984;288:1191-1194.
22. Lou HC, et al. Psychosocial stress and severe prematurity. Lancet 1992;340:54.
23. Barnea ER, Tal J. Stress-related reproductive failure. J In Vitro Fert Embryo Transf 1991;8:15-23.
24. McAnarney ER, Stevens-Simon C. Maternal psychological stress/depression and low birth weight. Is there a relationship? Am J Dis Child 1990;144:789-792.
25. Wadhwa PD, et al. The association between prenatal stress and infant birth weight and gestational age at birth: A prospective investigation. Am J Obstet Gynecol 1993;169:858-865.