Data Source

This study was embedded within the Generation R study, a population-based prospective mother and child cohort study, designed to identify early environmental and genetic causes of normal and abnormal growth, development and health from fetal life until young adulthood. The Generation R study has been previously described in detail [26–29]. In total, 9778 mothers with a delivery date from April 2002 until January 2006 were enrolled in pregnancy (n = 8879) or at birth of their children (n = 899) in the entire Generation R Study. This includes 7069 women, who were enrolled in early pregnancy (<18 weeks of gestation, median: 13 weeks). The overall response rate of the study was 61% [29]. The assessments in prenatal phase were conducted using three questionnaires: Mother 1 Questionnaire in early pregnancy; Mother 3 Questionnaire in mid-pregnancy (18–25 weeks of gestation); Mother 4 Questionnaire in late pregnancy (gestational age ≥25 weeks) [27]. Overall, mothers received four postal questionnaires during the prenatal phase; the three questionnaires that were just mentioned above plus Mother 2 Questionnaire regarding diet. The 25-page Mother 1 Questionnaire was used for the present study and assessed at around 12 weeks of gestation. It includes topics of medical history, family history, previous and current pregnancies, quality of life, life style habits, housing conditions, ethnicity and educational level [27]. The study was conducted with the guideline proposed in the World Medical Association of Helsinki and has been approved by the Medical Ethical Committee of the Erasmus Medical Center, University Medical Center Rotterdam. Written consent was obtained from all of the participating women [30].

Study population

Seven thousand and sixty-nine women were enrolled before 18 weeks of their gestation [26]. The assessment by Mother 1 Questionnaire was planned at around 12 weeks of their pregnancy (median: 13 weeks). We excluded women who didn’t respond to the questionnaire (n = 497). Additionally, we excluded pregnancies with the following outcomes: twin pregnancies (n = 71), induced abortion (n = 23), fetal deaths before 20 weeks of gestation (n = 62), loss to follow up their pregnancy outcomes (n = 23). Further, we excluded women with missing data on the symptoms (nausea, vomiting and fatigue) (n = 158). Finally, we excluded women in case of lacking information on one or more items of the SF-12 (n = 1156). Thus, the population for analysis comprised 5079 pregnant women (see S1 Fig).

Measurement of symptoms

The questions posed to pregnant women regarding to nausea, vomiting and fatigue are ‘have you had nausea in the last three months’, ‘have you had vomiting in the last three months’ and ‘have you had tiredness in the last three months’. The possible responses were ‘daily, a few days a week, once per week, less than once per week and never’. ‘The last three months’ refers the latest three months before the subject completed the questionnaire. By using ‘never’ as the reference group, the other four categories were recoded as dummy variables for multiple regression analyses.

Health-related quality of life

Women’s HRQOL in the past month was measured by SF-12 in the questionnaire at around 12 weeks of gestation (median: 13 weeks). SF-12 yields two component summaries: the physical component summary (PCS) and the mental component summary (MCS) [24, 31]. The Cronbach’s alpha for SF-12 in our study is 0.83. SF-12 includes 12 items regarding 8 scales: physical functioning (two items), role limitations due to physical problems (two items), bodily pain (one item), general health (one item), vitality (one item), social functioning (one item), role limitation due to emotional problems (two items) and perceived mental health (two items). Recoding for some items was conducted, so that a high value indicated the same type of response for each item. Then the raw scores were transformed to provide scale scores that ranged from 0 (the worst) and 100 (the best). We then calculated the raw physical component summary score and the raw mental component summary score by summing up all the scale scores weighted based on US general population survey. Finally the raw PCS and MCS scores were transformed into the standard scores based on the normalized algorithms from the US general population with the mean value of 50 (add 50) and the standard deviation of 10 (multiply by 10) [31]. The standardization enables cross-cultural comparison [32].

Covariates

Based on previous studies of determinants of pregnant women’s HRQOL, we considered the demographic characteristics, life-style related factors, and indicators of health status as potential confounders [9, 17, 18, 33, 34]. Data on these variables were collected in self-reported questionnaires at enrollment. The demographic characteristics included maternal age, gestational age, ethnic background (native Dutch people, other Western immigrant and non-Western immigrant), educational level (low, mid-low, mid-high, high), parity, marital status. Maternal ethnic background and education level were defined according to the classification of Statistics Netherlands [35]. Education was categorized into four subsequent levels based on the Dutch Standard Classification of Education: high (university degree), mid-high (higher vocational training, Bachelor’s degree), mid-low (>3 years general secondary school, intermediate vocational training) and low (no education, primary school, lower vocational training, intermediate general school, or 3 years or less general secondary school) [36].

Lifestyle-related factors included body mass index (BMI), tobacco and alcohol use; indicators of health status included chronic non-infectious conditions, infectious/inflammatory conditions, uro-genital symptoms, sleep quality, headache, anxiety, and depression. Tobacco/alcohol use was measured by asking ‘have you smoked in the past three months” and “have you drunk any alcohol in the past three months”, respectively. The amount of alcohol use was also measured.

Women were asked whether in the past 12 months they had one or more of 14 chronic non-infectious conditions on the standard list of chronic conditions according to Statistics Netherlands [37], i.e. diabetes, high blood pressure, a heart condition, migraine, epilepsy, chronic eczema, intestinal disorder, a severe back disorder, arthritis, multiple sclerosis, a thyroid disorder, chronic bronchitis, asthma, nose allergy (such as hay fever). Women were asked whether in the past three months they had one or more infectious/inflammatory conditions, i.e. fever, flu, sore throat or throat infection, runny nose or cold, sinusitis, ear infection, pneumonia, eye infection, cold sore, mouth infection, rash, dermatitis, fungus infection of skin or feet, warts, shingles, diarrhea or enteritis, cystitis or pyelitis and jaundice. An open question followed by asking about other infectious or inflammatory condition not mentioned. Women were asked whether they had one or more of the 10 uro-genital conditions/symptoms in the past three months, i.e. urination/urethra: frequent need to urinate, pain, burning feeling, itching; vagina: discharge, burning feeling, itching; bleeding after sexual intercourse; non-painful ulceration of urethra or vagina; enlarged lymph glands in groin. We summed up the presence of chronic non-infectious conditions, infectious/inflammatory conditions or uro-genital conditions/symptoms respectively and categorized the results into three categories: none condition/symptom, one condition/symptom, two or more conditions/symptoms. Frequency of ‘sleep badly’ and ‘headache’ were measured in the same way with the measurement of nausea, vomiting and fatigue, and were dichotomized as ‘yes’ or ‘no’. Anxiety and depression were measured with two questions: “Have you ever had a period in which you were anxious or worried (for at least two consecutive weeks) and “Have you ever had a period in which you felt very down or depressed (for at least two consecutive weeks).

Statistical analysis

Descriptive analysis was applied to characterize the study population. Differences of mean scores in physical component summary and mental component summary among subgroups were compared using one-way ANOVA. Correlations between symptoms were assessed. The Spearman correlation coefficient between nausea and vomiting was 0.50 (p<0.01); the coefficient between nausea and fatigue was 0.32 (p<0.01); the coefficient between vomiting and fatigue was 0.14 (p<0.01). Cohen’s effect sizes (d) were calculated by dividing the difference in mean scores among subgroups by largest SD and interpreted as: 0.2≤d<0.5 small difference, 0.5≤d<0.8 moderate difference, d≥0.8 large difference (see S2 Table) [38]. Multivariate linear regression was applied to assess the independent associations between nausea, vomiting, fatigue and scores of physical component summary and mental component summary by establishing a set of models. All models included the variable gestational age at enrollment. The crude model included three variables: frequency of nausea, vomiting and fatigue. In model 1, effect estimates were additionally adjusted by demographic characteristics. In model 2, effect estimates were additionally adjusted by the lifestyle-related factors. In model 3 (full model), we additionally adjusted by indicators of health status. Multicollinearity was checked and not serious. Multiple imputations were employed to account for the missing data in covariates. The imputed covariates were ethnic background, educational level, marital status, parity, smoking, alcohol use, headache, sleep badly, anxious or worried, feeling down and depressed, chronic non-infectious conditions, infectious/inflammatory conditions and uro-genital conditions/symptoms. Five imputed datasets were generated, based on which the pooled estimates were used to report betas and their 95% confidence intervals (CIs). Imputations were based on the relationships between all variables included in this study [39]. We also applied the multivariate linear regression analyses using the non-imputed data. Differences between women who were included in the present study (n = 5079) and women who were excluded (n = 1990) were assessed using Chi-square tests, and independent-sample t tests. Sensitivity analysis was performed by splitting the population into two subgroups: less than 14 weeks of gestation and over 14 weeks of gestation, and then comparing their outcomes (see S4 Table).

All analyses were conducted with Statistical Package for Social Sciences (SPSS) version 21.0 for Windows (IBM Corp., Armonk, NY, USA). Significance differences were indicated at the level of p<0.05.

Non-response analyses

Compared with the participating women in the study (n = 5079), the excluded women (n = 1990) were more often with low education, non-Dutch, single and in their first pregnancy (p<0.05) and reported lower prevalence of infectious/inflammatory conditions and uro-genital conditions/symptoms (p<0.05) (see S3 Table). Given the amount of missing data on covariates, we could not conclude that the study included healthier women, or the contrary, compared with the excluded women.

Strengths and limitations

A strength of this study is the large sample size compared to earlier studies [15, 17–21, 45]. Information regarding a comprehensive set of potential confounders was available. Some limitations should be taken into account. Causation could not be evaluated with the current cross-sectional analyses. We recommend that future studies evaluate time trajectories of nausea, vomiting, fatigue and HRQOL during pregnancy. Women who were included in the present study were younger, higher educated, more often of Dutch origin and more frequently had infectious/inflammatory conditions and uro-genital conditions/symptoms than women excluded from the sample for analysis. Given the amount of missing data on the covariates in the excluded population, we could not conclude that the excluded population was healthier or more morbid. The selection bias may have occurred; for example, if the excluded women with nausea, vomiting and fatigue provided higher (or lower) HRQOL scores than the included women with these symptoms. Furthermore, the women in this study may not fully represent the general population in the Netherlands, as all of them resided in Rotterdam. We asked women to think about the frequency of their symptoms in the previous three months, while for the most SF12 items we only asked them to recall within past month. Although we included many potential confounders in the models, remaining unmeasured confounders, such as work status [17], therapeutic approaches to relieve nausea, vomiting and fatigue, could also explain associations between nausea, vomiting, fatigue and HRQOL. Regarding the measurement of covariates in the present study, we acknowledge that the anxiety and depression were not measured by either a psychometric instrument or a diagnostic interview. The questions were unspecific, which did not capture the information of severity. Misclassification may not be ruled out.

In our study, generic HRQOL was measured. For future studies, we recommend to include both generic measures of HRQOL and specific measures such as the ‘health-related quality of life for nausea and vomiting during pregnancy’ (NVPQOL) [16]. Munch et al. showed that the NVPQOL was more sensitive to measure the impact of pregnancy-related symptoms on HRQOL compared to the SF-36 [18]. Previous studies indicated that the degree of the negative impacts of nausea and vomiting may be associated with the severity of these symptoms [17, 19]. In the present study, we measured the frequency rather than the severity of the symptoms. Women’s interpretation of the question and the framing of frequencies may have influenced the results. It is controversial that women never presented with fatigue in early pregnancy. We recommend to measure the severity by symptom-specific instruments such as the Motherisk-PUQE (pregnancy-unique quantification of emesis and nausea) scoring system [46] or the Multidimensional Assessment of Fatigue (MAF) scale [47].