Nausea, with or without vomiting, occurs in approximately 70% of pregnancies. In our study of a final sample of 825 women coming to the University of California at Los Angeles Hospital and Clinics, there was an incidence rate of 70.9% for nausea (with or without vomiting) during the first 20 weeks of gestation.1 The vomiting rate was 47.3%. Various factors were associated with nausea and vomiting of pregnancy (NVP), e.g., white ethnicity was associated with decreased nausea and vomiting, housewife occupation was associated with increased vomiting, and a history of infertility and a history of nausea during previous pregnancies were associated with an increased risk of nausea and vomiting during the current pregnancy.
Numerous studies have shown NVP as a prognostic indicator of pregnancy outcomes, both favourable (e.g., decreased miscarriage, perinatal mortality, low birth weight, premature birth) and unfavourable (e.g., increased foetal anomalies, low birth weight), or with no association with pregnancy outcome regarding birth weight, foetal anomalies or length of gestation.
Differences among previous studies may be due to the following factors population differences; methods of subject selection; sample size differences (i.e., small sample sizes may not detect significant associations, and studies with a very large sample size may produce significant associations for small effects); different classification systems for NVP and foetal outcome; or failure to control confounding variables (i.e., variables associated with both NVP and pregnancy outcome that could produce a spurious association).
The objectives of our study were: (1) to investigate the association between NVP and pregnancy outcome in a large, diverse sample population, employing multivariate statistical methods in order to control confounding variables, and (2) to determine the generalizability of the identified associations of NVP with pregnancy outcome across numerous studies, using the statistical methods of meta-analysis.
The epidemiologic study of NVP was conducted in Los Angeles, California, on a final historical cohort of 903 women with a singleton pregnancy.2 Two groups of women were eligible for inclusion. Group 1 consisted of women who were delivered at UCLA Hospital and Clinics during a one-year period (April, 1983 to March, 1984). Group 2 consisted of women who were treated for miscarriage prior to 21 weeks of gestation at this institution during the same period. The criteria for inclusion in the study were: at least one prenatal visit at no later than week 20 of gestation and documentation of presence or absence of NVP in the medical records. Only the first pregnancy of an individual during the study period was included. The exclusion criteria included: uncertainty about the date of the last menstrual period, reporting nausea or vomiting or both in conjunction with diarrhoea or fever.
The demographic characteristics of the study sample were: 26.9 years of age (range 14-43); the ethnic mix included 57% Hispanic and 27% white women; and 51% were housewives by occupation. There were significant differences between Groups 1 and 2 in demographic characteristics.
NVP was graded in three levels: no nausea or vomiting, nausea without vomiting, and nausea with vomiting. The pregnancy outcome measures which were analyzed included:
Note that dichotomizing the continuous scales of gestation and birth weight is appropriate, because it reflects standard obstetric classification and identifies apparent cut-off points for increased risk. However, reducing an interval scale of measurement to categories results in loss of available information, and therefore provides less precise estimates of association.
Associations of nausea and vomiting with these pregnancy outcomes were analyzed using analysis of covariance for quantitative outcome measures and multiple logistic regression for the categorical outcome measures. In each case, maternal age, ethnicity, occupational status, and use of antiemetic medications were included as covariates in predicting pregnancy outcome.
Nausea and vomiting during the first 20 weeks' gestation occurred in 69% of patients: 23% had nausea without vomiting and 46% had both nausea and vomiting. Despite significantly less weight gain during the first 20 weeks among women experiencing vomiting (mean weight gains for vomiting: 3.36 kg, nausea only: 4.41 kg, and no symptoms: 4.73 kg, p < 0.001), there was no significant difference in maternal mean weight among all three groups at the time of delivery. Nor were there any significant differences in infant birth weight associated with the presence or absence of NVP for the following pregnancy outcomes: birth weight, placental weight, neonatal body length and head circumference, 1- and 5-minute Apgar scores, and presence of severe foetal anomalies.
The only pregnancy outcome associated with NVP was miscarriage. An increased level of NVP was associated with decreased risk of miscarriage (p < 0.01). Although there was no association between NVP and threatened miscarriage, a significant effect was found on the risk of subsequent miscarriage. Patients with threatened miscarriage without symptoms of NVP had a 54.1% risk of subsequent miscarriage; those with nausea only had a 32% risk; and patients with vomiting had a 10.3% risk (i.e., the rate of miscarriage decreased with increasing level of NVP, gamma = -0.66, p < 0.001).
Although the only pregnancy outcome associated with NVP was miscarriage, the issue arose whether these results are generalizable across studies and, therefore, different populations. Specifically, we addressed the question whether a decreased risk of foetal and neonatal mortality due to NVP was restricted to the first 20 weeks of gestation (i.e., miscarriage), or whether this decreased risk was conferred throughout the prenatal and neonatal period. To satisfy this second study objective, we conducted a meta-analysis, comparing the results of our study with other studies conducted between the 1950s and 1980s on the association of NVP with pregnancy outcome.3 The studies were English language publications identified through a MEDLINE search for January, 1966 to December, 1988. Earlier articles were identified from references cited in the original search articles. Studies dealing only with hyperemesis gravidarum were excluded. A study was included only if it was possible to obtain quantitative data, either from direct reporting or where frequencies could clearly be established from percentage data. These data had to be able to be classified in a 2x2 table (i.e., presence/absence of NVP x presence/absence of mortality) in one or more of the following intervals:
There were three studies from the same population; the largest of these was selected. Thus there were nine studies included in the meta-analysis, including our study.
Ideally, the association between NVP and mortality would be examined separately for each time interval. Unfortunately, except for the first 20 weeks of gestation, there were not more than two studies with data specific to each time interval. Therefore, we conducted our analysis separately for each of three time intervals: (1) the entire prenatal and neonatal period, (2) the entire prenatal period, and (3) only the first 20 weeks of gestation (i.e., focusing only on the risk of miscarriage), thereby narrowing the time interval with each subsequent analysis.
For each time interval examined, the results from all studies were combined to obtain an overall measure of association (a common odds ratio [OR]). A homogeneity chi-square test was also conducted to determine if associations between NVP and mortality were consistent across the studies that were pooled.
There were nine studies evaluated for the association between NVP and miscarriage and perinatal mortality (i.e., the entire prenatal and neonatal period). NVP was associated with decreased risk of mortality (common OR = 0.47, 95% CI: 0.43-0.53). However, as indicated by the significant homogeneity chi-square (p < 0.001), this association was variable among studies using different time intervals for analysis. For the eight studies evaluated for the entire prenatal period, similar results were obtained. However, for the seven studies evaluated for the first 20 weeks of gestation only, not only was there a significant decrease in the risk of miscarriage when NVP was present (common OR = 0.36, 95% CI: 0.32-0.42, p < 0.001), but the association was consistent across the studies examined (homogeneity chi-square: p > 0.1). These results indicate that the decreased risk of perinatal mortality associated with NVP is confined only to the first 20 weeks of gestation and, therefore, NVP is a prognostic indicator for decreased risk of miscarriage, but not for stillbirths and neonatal death.
In meta-analysis there remains an additional problem, that there may be unpublished studies with contradictory results. We then calculated how many such studies would be needed to nullify the generalizable results of our meta-analysis, the so-called "fail-safe N." The calculation told us that there would need to be 168 unpublished studies with conflicting results which would contradict our meta-analytical findings. It is doubtful that so many exist.
In summary, our epidemiological study of a large, diverse population showed that NVP was only associated with a decreased risk of miscarriage and in cases of threatened miscarriage, NVP was associated with a decrease in subsequent miscarriage. Our meta-analysis indicated that the only association of NVP with perinatal and neonatal mortality that was generalizable across populations was a decreased risk of mortality during the first 20 weeks of gestation (i.e., miscarriage) due to NVP.
Back to home