SEARCH

SEARCH BY CITATION

Keywords:

  • birth outcomes;
  • hyperemesis;
  • registry

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Objective

To study associations between hyperemesis gravidarum (HG) and birth outcomes.

Design

Population-based cohort study.

Setting

Norway.

Sample

Singleton births in the Norwegian Birth Registry, 1967–2009 (n = 2 270 363).

Methods

Multiple logistic regression was applied to study associations between HG and dichotomous outcomes; multiple linear regression to study associations between HG, birthweight and gestational length. Generalised estimating equations were applied to obtain valid standard errors. Sub-analysis on data with available information on smoking was conducted (1999–2009).

Main outcome measures

Small and large for gestational age (SGA/LGA), Apgar score after 5 minutes, very preterm and preterm birth (VPTB/PTB), perinatal death, stillbirth, neonatal death, birthweight and gestational length.

Results

No associations between HG and adverse pregnancy outcomes were observed in crude analyses, except for VPTB (odds ratio [OR] 0.79, 95% CI 0.67–0.93). In adjusted analysis, HG was associated with perinatal death (OR 1.27, 95% CI 1.08–1.48). Inverse associations were observed between HG and VPTB (OR 0.80, 95% CI 0.68–0.94) and LGA (OR 0.95, 95% CI 0.90–0.99). Sub-analyses showed no associations between HG and perinatal death (OR 1.29, 95% CI 0.91–1.83). The inverse associations between HG, VPTB and LGA were strengthened (OR 0.66, 95% CI, 0.48–0.91 and OR 0.86, 95% CI 0.79–0.93, respectively). Exposed babies had reduced birthweight and gestational length compared with unexposed, adjusted difference − 21.4 g and − 0.5 days, respectively. Adjustment for smoking slightly strengthened the impact of HG on birthweight.

Conclusions

Inverse associations for HG and VPTB and LGA were observed. HG was associated with slight reductions in birthweight and gestational age.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Hyperemesis gravidarum (HG) affects 0.8–3.2% of pregnant women depending on country of origin.[1] It is characterised by severe nausea and vomiting starting before 22 weeks of gestation, often leading to nutritional deficiencies and maternal weight loss.[2] Although rare, HG is the most common cause of hospitalisation during the first half of pregnancy.[3] The strongest risk factor for HG is having had it in a previous pregnancy.[4] The underlying aetiology remains largely unknown.

A recent systematic review of 24 studies reported that HG was associated with 42% increased risk of having a child with low birthweight (LBW), 28% increased risk of small for gestational age (SGA) and 32% increased risk of preterm birth (PTB).[5] However, heterogeneity in the definitions of HG, studied outcomes and adjustment for different confounders complicate the generalisation of these results.[6] A recently published article exploring the risk of placental dysfunction disorders among mothers with HG reported a 39% increased risk of SGA.[7] The largest epidemiological study on HG and pregnancy outcomes until now was based on data from the Swedish Birth Registry, including 3068 women with HG.[8] It reported positive associations between HG and LBW and PTB. The prevalence of HG in this study was 0.3%. The author suspected under-reporting as well as misclassification of HG because the HG diagnosis was not well-defined and prevalence varied extremely between hospitals, threatening the validity of these findings. Several studies, originating from the Middle East, Asia, America and Canada,[9-12] have shown no association between HG and birth outcomes. A study from Israel showed a lower frequency of spontaneous abortions among women with HG, and no increased risk of other adverse birth outcomes.[9] A study from the USA concluded that hyperemetic mothers had fewer stillbirths and spontaneous abortions compared with women without it.[13] Most studies included <200 women with HG and were case–control studies.

Hence, previous research on the association of HG and pregnancy outcomes suffers from heterogeneity of the definition of HG and yields conflicting results, warranting further research using large quality-assured data sets. The aim of this study was to explore associations between HG and selected birth outcomes using the Medical Birth Registry of Norway (MBRN), comprising more than 2.5 million births.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

This is a registry-based cohort study. MBRN was founded in 1967 as the first national birth registry in the world.[14] Notification of births after 12 weeks of gestation (from week 16 during 1967–98) is compulsory, using a textual form completed by the attending doctor or midwife shortly after birth. The form contains information on maternal health before and during pregnancy, the birth and the initial health of the child. Information entered is supplemented using the antenatal card and any hospital records. Registrations in MBRN are matched to the Central Persons Registry through unique identification numbers securing approximately 100% attendance.[15] The general validity of the MBRN for epidemiological research is considered to be high.[14]

Altogether, 2 583 651 births were registered in the MBRN from 1967 to 2009. We included singleton births after the 22 weeks of gestation. We excluded observations with missing information on identification number (n = 4712), parity (n = 34 767), age (n = 68), gestational length (n = 126 833) and birthweight of the child (n = 6968), as these observations may reflect a low validity of HG registration. We also excluded birthweights <500 g (n = 16 328) and > 6000 g (n = 7065). The final sample consisted of 2 266 345 observations. From 1999 to 2009, information on smoking was available (n = 545 748).

HG was defined according to International Classification of Diseases, 8th revision (ICD-8) diagnosis codes 638.0 and 638.9 during 1967–98, and ICD-10 diagnosis codes 021.0, 021.1 and 021.9 during 1999–2009.

Outcome variables were selected in line with previous research. Size at birth, small for gestational age (SGA) and large for gestational age (LGA), was calculated using Norwegian specific fetal growth tables, described in detail elsewhere.[16] Preterm and very preterm birth (VPTB) were defined according to completed pregnancy weeks; PTB <37 weeks, VPTB <32 weeks of gestation, respectively. Apgar score after 5 minutes was also explored as it can be regarded as an assessment tool for a newborn's health. Perinatal death was defined as stillborn with gestational length of ≥ 22 weeks, and deaths before the end of day 7 after birth. Stillbirth was defined as intrauterine death before birth, death during birth or at an unknown time. Neonatal death was defined as deaths during the first 28 days of life. Birthweight and gestational age were calculated as continuous variables (in grams and days, respectively).

Maternal age, parity, country of origin, education and smoking were considered possible confounders and adjusted for. Maternal age was categorised as <20, 20–24, 25–29, 30–34 and ≥ 35 years. Smoking status was categorised as nonsmoker, smoker or missing. By parity, the women were classified as primipara or para. Information on maternal education and ethnicity was obtained from Statistics Norway. Educational level was recorded according to the Norwegian Standard Classification of Education, divided into five groups; <13, 13, 14–17, > 17 years or missing. Maternal country of birth was categorised into immigrant groups that were geographically and culturally related, in line with previous research.[1] In addition, Apgar score was adjusted for by induction of labour and gestational length as this can influence the given score. Likewise, we adjusted perinatal death by gestational length. Gestational age and birthweight were additionally adjusted for by elective caesarean section as this could influence time point of birth. As different editions of ICD have been used, we also adjusted the analysis for time-period.

Associations between HG and dichotomous birth outcomes were assessed by multivariable logistic regression. Crude and adjusted odds ratios (OR) with 95% confidence intervals were calculated and used as a proxy for relative risks. Continuous outcomes were studied using multiple linear regressions. As smoking information became available from 1999, we subdivided the data set and performed additional adjustment for smoking in this population separately.

Given that many women were registered more than once in the data set, all analyses were performed using generalised estimating equations (GEE) to obtain valid standard errors. SPSS for Windows version 20.0 (SPSS Inc., Chicago, IL, USA) was used for all calculations.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Altogether 20 004 women (0.8%) were registered with HG. In the sample from 1999–2009 the prevalence was 1.4%. Women with HG were more likely to be nonsmokers compared with women without HG. According to age, modest differences were observed according to HG status. Women with HG were less likely to have educational level below 13 years compared with women without HG. Compared with women without HG, women with HG were more frequently of African or Asian descent (Table 1).

Table 1. Background characteristics of the sample
  n No hyperemesisHyperemesis
n % n %
  1. a

    The Middle East including Lebanon, Iraq, Palestine and Syria. North Africa including Algeria, Egypt, Libya, Morocco and Tunisia.

Maternal age, years
< 20121 880121 0285.48524.3
20–24573 142568 32725.3481524.1
25–29775 663768 87534.2678833.9
30–34543 556538 40324.0515325.8
≥ 35252 104249 70811.1239612.0
Parity
Primipara924 511916 21640.8829541.5
Para1 341 8341 330 12559.211 70958.5
Education, years
< 13955 958948 50242.2745637.3
13514 922510 26022.7466223.3
14–17614 282608 34727.1593529.7
> 17116 138115 1465.19925.0
Missing65 04564 0862.99594.8
Smoking habits
Nonsmoker451 745445 07819.8666733.3
Smoker63 73563 1912.85442.7
Missing1 750 8651 738 07277.412 79364.0
Maternal country of origin
Europe, USA, Canada2 133 0152 115 36394.217 65288.2
Turkey, Middle-East and North Africaa23 73623 3101.04262.1
Other Africa19 38218 8370.85452.7
Asia53 31352 3602.39534.8
Central and South-America28 26727 9631.23041.5
Other countries and missing863285080.41240.6
Total2 266 3452 246 34199.220 0040.8

Birth outcomes by hyperemesis status are summarised in Table 2. Bivariate analysis showed that women with HG were less likely to experience VPTB (P = 0.002). Associations between HG and other birth outcomes did not reach the level of statistical significance.

Table 2. Pregnancy outcomes by hyperemesis status
  n No hyperemesisHyperemesisP-value
n (%)n (%)
Gestational age <32 weeks20 97820 830 (0.9)148 (0.7)0.002
Gestational age <37 weeks124 508123 349 (5.5)1159 (5.8)0.070
Small for gestational age238 069235 954 (10.5)2115 (10.6)0.757
Large for gestational age96557497 (11.0)2158 (10.8)0.360
Apgar score <7 after 5 min19 62019 421 (0.9)199 (1.0)0.715
Perinatal death20 35620 165 (0.9)191 (1.0)0.410
Neonatal death86388559 (0.4)79 (0.4)0.751
Stillbirth13 23913 119 (0.6)120 (0.6)0.772
Total2 266 3452 246 341 (99.2)20 004 (0.8)

Results from the multivariate regression analysis are presented in Tables 3 and 4, where Table 4 presents data from 1999 to 2009.

Table 3. Crude and adjusted odds ratios with 95% confidence intervals for adverse birth outcomes in women with and without HG from 1967 to 2009 (= 2 266 345)
  n Crude ORAdjusted OR
(95% CI)(95% CI)a
  1. a

    Adjusted for maternal age, parity, education, time period and maternal country of origin.

  2. b

    Also adjusted for gestational length.

  3. c

    Additional adjustment for induction and gestational length.

Gestational age <32 weeks2 266 3450.79 (0.67–0.93)0.80 (0.68–0.94)
Gestational age <37 weeks2 266 3451.05 (0.99–1.12)1.05 (0.99–1.11)
Small for gestational age (<10%)2 266 3451.00 (0.96–1.05)1.01 (0.97–1.06)
Large for gestational age (>90%)2 266 3450.95 (0.91–1.00)0.95 (0.90–0.99)
Apgar score <7 after 5 minutes1 634 0051.02 (0.89–1.18)0.99 (0.86–1.14)c
Perinatal death2 266 3451.06 (0.92–1.22)1.27 (1.08–1.48)b
Neonatal death2 266 3451.04 (0.83–1.29)1.16 (0.94–1.35)
Stillbirth2 266 3451.02 (0.85–1.23)1.13 (0.94–1.35)
Table 4. Crude and adjusted odds ratios (OR) with 95% confidence intervals (CI) for adverse birth outcomes in women with and without HG from 1999 to 2009 (= 545 748)
  n Crude ORAdjusted ORAdjusted OR
(95% CI)(95% CI)a(95% CI)b
  1. a

    Adjusted for maternal age, parity, education and maternal country of origin.

  2. b

    Also adjusted for smoking.

  3. c

    Also adjusted for gestational length.

  4. d

    Also adjusted for induction and gestational length.

Total 545 748   
Hyperemesis 7475 (1.4%)   
Gestational age
<32 weeks545 7480.66 (0.48–0.91)0.65 (0.47–0.89)0.66 (0.48–0.91)
<37 weeks545 7480.99 (0.89–1.10)0.99 (0.89–1.10)1.00 (0.91–1.11)
SGA545 7481.06 (0.97–1.15)1.01 (0.93–1.10)1.06 (0.97–1.15)
LGA545 7480.88 (0.81–0.95)0.88 (0.82–0.95)0.86 (0.79–0.93)
Apgar score <7544 2931.04 (0.86–1.27)1.03 (0.84–1.25)1.03 (0.85–1.25)d
After 5 min
Perinatal death545 7481.12 (0.80–1.58)1.09 (0.77–1.53)1.29 (0.91–1.83)c
Neonatal death545 7480.75 (0.39–1.35)0.70 (0.36–1.35)0.71 (0.37–1.37)
Stillbirth545 7481.23 (0.83–1.82)1.21 (0.82–1.79)1.25 (0.84–1.84)

The crude numbers in Table 3 showed that HG was inversely associated with VPTB, but was not associated with other outcomes. After adjustment for confounders, VPTB and LGA were inversely associated with HG. No association was seen for Apgar score after 5 minutes. HG was, however, associated with perinatal death reflected in an OR of 1.27 (95% CI 1.08–1.48). No association was observed for HG and stillbirth or neonatal death.

When exploring observations from 1999 to 2009, adjustment for smoking was made in addition to adjustment for potential confounders available for the entire sample. Crude analyses showed inverse associations between HG and LGA and VPTB. The other pregnancy outcomes were not associated with HG. Adjustment for possible confounders, including smoking, did not influence these estimates. In contrast to the analysis based on the whole sample, there was no association between HG and perinatal death in the subset. Likewise, there was no association between HG and stillbirth or neonatal death rates.

Analysis of the continuous variables birthweight and gestational age is presented in Table 5 and 6, where Table 6 represents the data collected from 1999 to 2009. We observed a slight reduction in both gestational age and birthweight among women with HG compared with women without it both in the crude and adjusted analyses of the entire population. In the subsample, we found an additional reduction in birthweight when additional adjustment for smoking was applied.

Table 5. Differences in birthweight and gestational age according to hyperemesis status from 1967 to 2009 (= 2 266 345)
 MeanCrude differenceAdjusted difference β
(95% CI)(95% CI)a
  1. a

    Adjusted for maternal age, parity, education, time-period, maternal country of birth and elective caesarean section.

Birthweight, grams
HG−3525.9 g11
HG+3501.1 g−24.8 (−33.0 to −16.5)−21.4 (−29.4 to −13.3)
Gestational age, days
HG−279.7 days11
HG+278.9 days−0.8 (−1.01 to −0.63)−0.5 (−0.7 to −0.3)
Table 6. Differences in birthweight and gestational age according to hyperemesis status from 1999 to 2009 (= 545 748)
 MeanCrude differenceAdjusted difference βAdjusted difference β
(95% CI)a(95% CI)b
  1. a

    Adjusted for maternal age, parity, education, elective caesarean section and maternal country of birth.

  2. b

    Also adjusted for smoking.

Birthweight, grams
HG−3562.3111
HG+3531.6−34.2 (−47.1 to −21.2)−25.7 (−38.2 to −13.1)−34.4 (−46.9 to −21.9)
Gestational age, days
HG−278.71 1
HG +278.1−0.6 (−0.9 to −0.3)−0.4 (−0.7 to −0.1)−0.4 (−0.7 to −0.2)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Main finding

To the best of our knowledge, this is the largest study on HG and birth outcomes. HG was not positively associated with adverse birth outcomes, except for perinatal death when exploring data from 1967 to 2009. This association was not apparent in the subsample from 1999 to 2009. We observed inverse associations between HG and VPTB and LGA, all findings significant in the whole sample as well as the subsample.

Strengths and limitations

Large data sets reduce the risk of random error. The generalisability is high as notification to MBRN is compulsory and all observations are matched to the Central Person Registry. A recent validity study described the registration of HG in MBRN to be acceptable for large-scale epidemiological studies,[17] reporting sensitivity and specificity of 83.9% and 96.0%, respectively, for mild HG. When stricter diagnostic criteria were applied, the sensitivity fell to 64.3%. The reason for this may be that registration of HG relies on textual information and subsequent recoding into corresponding ICD-code by the staff at MBRN. According to the validity study, the MBRN coded HG primarily as 638.9 until 1999 and only as O21.9 after 1999. Therefore there is potential for misclassifications on several levels, which may result in fewer cases of severe HG being registered.[17] The observed exposure–outcome associations may therefore be stronger than reported.

A possible limitation to our study is the lack of information on time of debut and duration of HG. A recently published article reported that the risk of placental dysfunction disorders were highest among women with HG in the second trimester.[7] Information on HG is registered on the MBRN form shortly after birth, based on the antenatal card or hospital records if any. However, standard definitions of HG, including the ICD-10 definition, pinpoint the debut of HG before 22 weeks of gestation. This is also reflected in the high hospital admission rate for HG during the first half of pregnancy. Information on HG is documented before the outcomes are observed. The risk of recall bias is therefore reduced. The results should, however, be read with caution as the study design does not permit causal inference.

Interpretation

The systematic review previously mentioned, found no association between HG and perinatal death.[5] The only cohort study reporting a positive association was the largest study included, comprising 2270 cases of HG, yielding an OR of 1.64 for perinatal death.[6] This study used hospital data only, which may explain a 0.4% prevalence of HG. There was also a lack of confounder control. In contrast, MBRN also comprises the milder form of HG (ICD-10 code O21.0) and provides information on more confounders. In contrast to these studies, Roseboom et al. found an OR 0.46 for perinatal death among women with HG, suggesting an inverse assocation.[18] This study had an overall prevalence of HG of 0.2%, which might indicate a weakness in registration of HG and subsequent underestimation of effects. We found an inverse association between HG and the risk of VPTB, which is a known risk factor for perinatal death. Therefore, we adjusted this analysis by gestational length (results not shown). This did not change the estimate, suggesting that the observed association between HG and perinatal death is realised through other pathways.

Previous research has shown that smoking influences the risk for both HG and adverse pregnancy outcomes, so a subanalysis was performed for observations with this information available. In the subsample, there was no association between HG and perinatal death in the crude or adjusted analyses. Stepwise regression showed that smoking was not responsible for changing the risk estimate. The difference between the total sample and the subsample is primarily the use of different disease classification systems, reflecting different time periods, where ICD-8 was used from 1967 to 1998 and ICD 10 from 1999 to 2009. One might suspect a better registration of HG as well as potential confounders in the subsample, given the difference in prevalence. There might be increased awareness of the necessity to register health information in general more accurately.[17] Differences in data quality might therefore partly explain the differences regarding HG and perinatal death observed in the two data sets. It is highly unlikely that the subsample is too small to reveal a possible association between HG and perinatal death.

There were minor differences in birthweight and gestational length for babies born to mothers with and without HG. The clinical significance of these findings seems limited. Several studies have found similar and slightly higher reductions in birthweight.[6, 8, 18-20] Other studies have not found a similar association.[9, 11] Depue et al. reported higher birthweight and fewer stillbirths among children born of hyperemetic mothers than among controls,[13] but the differences were not statistically significant. A potential reason for the slight differences observed in this study, might be that women with HG receive satisfactory treatment to secure fetal growth. The frequency of hospital admissions due to HG is high[3] and while admitted she receives fetal monitoring and symptomatic treatment in line with national guidelines. Unfortunately, information on hospital admissions or treatment received was not available in the MBRN and could not be included in the analysis. The timing of exposure to adverse stimuli in utero may be of greater importance than the direct effect on birthweight. The theory of fetal programming states that presence of negative stimuli during crucial periods of development can increase the risk of disease in adulthood.[21, 22] Starvation has been identified as an important negative stimulus. Metabolic disturbances due to HG may resemble those seen during starvation, suggesting that HG might have long-term health consequences for the child.[22]

In line with previous research, HG was found to be associated with a 22% decreased risk for LGA.[6, 18, 20] In both the total sample and subsample, HG was associated with a substantially decreased risk for VPTB, but there was no association for all PTB. The reason for this finding remains a puzzle. VPTB has a multifactorial aetiology such as uterine infection or cervical insufficiency.[23] Previous research has shown an inverse association between HG and adverse outcomes such as spontaneous abortion,[24] but the mechanisms for such an association are not clear. Further studies are warranted to determine if an association between HG and VPTB exists.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Our findings suggest that HG may be associated with perinatal death. This result should be handled with caution as the association was not observed in the subpopulation. HG was inversely associated with LGA and VPTB. Further research is warranted, to study the association between HG and perinatal death in particular.

Acknowledgements

We want to acknowledge the Norwegian Resource Centre for Women's Health, Rikshospitalet, Helse Sør-Øst, for providing the funding for this study. Furthermore, we wish to thank the Medical Birth Registry of Norway for providing us with data for this study.

Contribution to authorship

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

KFV, ÅV, PM, SV and AG had the initial idea for the project and applied for the data. NS, KFV, ÅV and AG analysed the data, and KFV, AG and ÅV drafted the manuscript. AG, AV, SV and PM all contributed to critical revision of the manuscript before submission. All authors have approved the final version of the article.”

Details of ethics approval

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

The regional committee for medical research ethics approved the study (REK sør-øst 2010/2618-1).

Funding

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References

Financial support for this study was obtained from Norwegian Resource Centre for Women's Healh, Rikshospitalet, Helse Sør-Øst, The National Institute of Public Health and the Norwegian Research Council.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of interests
  9. Contribution to authorship
  10. Details of ethics approval
  11. Funding
  12. References