Clinical management of nausea and vomiting in pregnancy and hyperemesis gravidarum across primary and secondary care: a population‐based study

To assess how nausea and vomiting in pregnancy (NVP) and hyperemesis gravidarum (HG) are managed and treated across primary and secondary care.


Introduction
Although nausea and vomiting is a very common symptom in pregnancy (NVP), affecting up to 70% of women 1 who often do not require treatment, in some cases severity can reach critical levels requiring hospital admission and need for continuous monitoring. 2 This severe condition, referred to as hyperemesis gravidarum (HG), has a reported prevalence of around 1.1% worldwide 1 and is responsible for a range of complications due to malnutrition, dehydration and excessive weight loss. Maternal and child health are affected by possible adverse effects of HG [3][4][5] and pregnancy complications are more likely. 6 These sequelae have substantial financial impact on the health services and significant burden on professional health care provision. [7][8][9] In June 2016, the Royal College of Obstetricians and Gynaecologists (RCOG) published the first national guidelines on the clinical management of NVP and HG for the United Kingdom (UK) 10 providing an accurate handbook for controversial decision making processes in the management of this condition with prescribed medications. Although this is supported by international reviews of HG, [11][12][13] there is no published evidence of how this condition is actually managed by health professionals and what common clinical pathways are followed by affected women. In the UK, research based on surveys 14,15 have shown general dissatisfaction with current clinical care from women experiencing HG, with a claim that many hospital admissions could have been avoided had they received timely antiemetic prescribing to curtail worsening severity. It is unknown whether underprescribing or late recognition of NVP and HG is widespread as there are no population-based studies showing how this condition is managed between primary and secondary care.
Using the Clinical Practice Research Datalink database (CPRD-HES), we have assessed the spectrum of NVP and HG across primary and secondary care and described how severity and pathways of management vary by maternal and pregnancy characteristics and comorbidities. We assessed the extent to which women are prescribed antiemetics in primary care and how this relates to hospital admissions.

Data source
We used primary and secondary care health records linked at individual patient level from the Clinical Practice Research Datalink (CPRD GOLD), 16 which includes over 15 million patients from 684 general practices (GP) across the UK. 17 Recorded information consists of demographics, symptoms and diagnoses coded using the Read coding system, 18 clinical tests and drug prescriptions. Over half of CPRD GOLD patients are linked with the Hospital Episodes Statistics (HES) database 19 which consists of all admissions to English hospitals. 20 Information within HES includes diagnoses coded using the International Classification of Diseases system (ICD-10) 21 and hospital procedures coded using the Office of Population Censuses and Surveys Classification of Interventions and Procedures (OPCS-4). 22 Maternity data contained in HES includes extensive information on pregnancy, labour and delivery; it is the main data source for monitoring maternity statistics in England 23 and is used to for perinatal epidemiology research. 23,24 Our previous studies using the CPRD-HES linked population show maternities are representative of those across the English population. 25,26 Patients were not involved in the development of the research.

Study population
Women with pregnancies ending in live birth or stillbirth between October 1998 and April 2014 who had active primary care registration were selected by extracting information from GP and HES records (operational codes of delivery) to obtain the most complete and precise information on each pregnancy. For live births, a probabilistic matching algorithm was employed to link each mother's pregnancy records to the corresponding children by matching each delivery date to a child's estimated birth date or HES birth admission and ensuring they had matching a household code, a unique identifier indicating which individuals in each practice live together. Gestation and pregnancy details were extracted from HES maternity data and, when missing, from the child's HES birth record, the mother's or child's GP records in this order of priority.

Diagnoses and hospital admissions for NVP and HG
To capture the full spectrum of severity, as defined in the national guidelines, 10 we identified all primary care diagnoses and hospital admissions for NVP/HG during pregnancy, using specific Read codes for primary care (Table S1) and ICD-10 codes for secondary care (Table S2) that were approved by co-author CNP who is a consultant in obstetric medicine. Due to controversy over a true distinction between NVP and HG and the lack of a standard approach for the diagnosis and clinical management of these conditions, 4 the risk that these diagnoses could have been used interchangeably by health professionals to refer to the same condition was taken in to consideration by carrying out a comprehensive analysis considering both diagnoses.
To exclude presentations of nausea and vomiting for specific reasons, a restrictive criterion was applied excluding any NVP diagnoses with evidence of differential diagnoses (Tables S3 and S4) recorded in GP data from 1 week before up to 1 week after the NVP diagnosis date or in HES data as a secondary reason of admission. This resulted in 1.9% of GP consultations and 17.6% of hospital admissions for NVP being excluded.

Drug treatment for NVP and HG
Antiemetic prescriptions were extracted from primary care records using selected drugs codes (Table S5) according to national recommendations from the RCOG, 10 and grouped in the following drug classes: antihistamines, phenothiazine dopamine antagonists, serotonin antagonists and steroids. As some drugs have multiple indications, differential diagnostic indications were explored to ensure the drugs under analysis were prescribed for the purpose of treating NVP/HG, according to the British National Formulary (BNF). 27 Grouping women with NVP/HG by their clinical management pathway We categorised women into four mutually exclusive groups (Figure 1), broadly representing their level of NVP/HG burden, according to their presentation and treatment in primary care or occurrence of hospital admissions as follows: a primary care diagnosis only, treatment in primary care, early hospital admissions (≥20 weeks gestation), late hospital admissions. We compared these four groups to a control group of all remaining women, i.e. those with no evidence of NVP/HG diagnosis nor treatment.

Maternal characteristics
We assessed maternal characteristics and comorbidities based on their previous evidence as risk factors for NVP/ HG 28,29 and information currently available in the CPRD-HES source dataset. These were: maternal age at delivery, socioeconomic deprivation as measured by the Index of Multiple Deprivation (IMD 2010) in quintiles, 30 ethnicity, smoking status during pregnancy, parity, birth plurality, diabetes, hypertension, pre-eclampsia, parathyroid dysfunction, coronary heart diseases, anaemia, thyroid dysfunction, hypercholesterolemia and asthma.

Statistical analysis
Numbers and proportions of pregnancies for each NVP/ HG group (i.e. diagnosis only, antiemetic treatment, early hospital admission and late hospital admission) were presented overall and the change in prevalence of each group was shown over time. To assess whether maternal characteristics differed by NVP/HG group, we used multinomial logistic regression (mlogit 31 with rrr option) to estimate relative risk ratios (RRR) with 99% confidence intervals (CI) for the association of each level of NVP/HG burden with maternal characteristics/comorbidities, compared to the control group. RRRs were adjusted for all the maternal characteristics and pre-existing comorbidities available in the data except for the risk factor of interest (Table 1). In order to account for potential clustering effects from including mothers with more than one pregnancy, a cluster option was set in the analysis. Missing values, present only for three maternal characteristics, namely ethnicity (12.4% missing), smoking (27.8%) and deprivation status (0.2%), were imputed using the multinomial logistic regression imputation method available in Stata MPv15 (Stata Corp, College Station, TX, USA) statistical package, 32 applying the mi imp mlogit function, 31 setting 10 imputed datasets and using all maternal characteristics available as predictor variables. As a sensitivity analysis to assess whether clear clinical distinctions were actually being made between NVP and HG diagnoses, we conducted stratified analyses by (1)  ***Missing values for these variables are estimated in the regression models using multiple imputation, therefore RRR is not given. ****Not applicable, as the condition is used as confounder and therefore adjusted for. *****80% of pregnancies had this gestational condition diagnosed for the first time in the third trimester. ******Empty cells indicate there were fewer than five exposed cases, for which statistical analyses were not performed.
dividing the primary care diagnosis only group into (a) those with HG diagnoses and (b) those with only NVP diagnoses, and (2) dividing the primary care treatment group into those with (a) antiemetics prescribed for an HG indication and (b) antiemetics prescribed for an NVP indication only. We assessed the prevalence of these four groups over time and whether they varied by maternal characteristics.

Results
Within the study period, there were 417 028 deliveries ending in live births or stillbirths in 300 858 women. The prevalence of NVP/HG overall was 9.1% (37 856): 3.6% (14 815 pregnancies) with primary care diagnoses that did not obtain treatment, 3.4% (14 226) with primary care diagnoses that were administered antiemetic drug treatment, 1.5% (6390) with first hospital admission before 20 weeks and 0.6% (2425) with late hospital admissions from 20 weeks onwards. Between 1999 and 2013 ( Figure 2) there were statistically significant increases in early hospital admissions and antiemetic prescribing such that by 2013, early admissions occurred in 2.1% of pregnancies and antiemetics were prescribed in primary care in 5.2% of pregnancies compared with 2.5% of pregnancies with recognised NVP/HG that were left untreated (P < 0.001 for both).

Management variation by maternal characteristics
Maternal characteristics varied across the groups (Table 1).
In general, compared with control pregnancies, those among women with NVP/HG had higher proportions of younger women, with higher socioeconomic deprivation, or with Asian or Black ethnicity, and these proportions increased with level of NVP/HG burden with the highest among women with hospital admissions. The prevalence of comorbidities was generally higher in the affected groups compared with control pregnancies, particularly for preexisting diabetes, gestational hypertension, pre-eclampsia, gestational anaemia and thyroid dysfunctions, asthma, and hypercholesterolemia.
In the adjusted analysis (Table 1), results showed a clear increased risk of NVP/HG with younger maternal age across all levels of burden with the magnitude of risk highest for hospital admissions; whilst women under 25 years were 1.5 times as likely to be treated in primary care compared with women age 30-34 years, they were over twice as likely to be admitted to hospital. Women from more deprived socio-economic groups had a comparable prevalence of NVP/HG diagnoses; however, they were more likely to be treated with antiemetics in primary care, and to have early hospital admission compared with women from the least deprived group (test for trend P < 0.001 for all groups other than diagnosis only). Asian and Black women had considerable increased risks across all levels of NVP/HG burden, although there was no association between ethnicity and late hospital admission. Current smoking was associated with a decreased risk of NVP/HG across all levels other than late hospital admissions. There was no association with multiparity other than a decreased risk in primary care diagnosis of NVP/HG only. Multiple birth was associated with NVP/HG diagnosed and treated in primary care but the risk was highest in those with early hospital admissions. Diabetes and hypertension were not associated with NVP/HG diagnosed or treated in primary care, although they were associated with some increase in late NVP/HG hospital admissions and pre-existing diabetes increased the risk of early hospital admissions. Pre-eclampsia was associated with women treated for NVP/HG in primary care, while eclampsia was associated with NVP/HG diagnoses and late hospital admissions. Asthma and anaemia increased the risk of all levels of NVP/HG. There was an increased risk of primary care treatment and hospital admissions in women with thyroid dysfunction or hypercholesterolemia.

Antiemetic prescribing distribution
Distributions of primary care antiemetic prescribing for women with NVP/HG who were and were not admitted to hospital are shown in Table 2. Of those never admitted to hospital (29 041), antiemetics were prescribed in 49% of pregnancies; first, second and third-line treatment was prescribed for 42%, 11%, and 1% of pregnancies respectively. The most commonly prescribed antiemetic was prochlorperazine (21.1%), followed by the other first line drugs promethazine (15.4%) and cyclizine (13%). Ondansetron and steroids were very rarely prescribed to these women; metoclopramide was the most commonly prescribed NVP diagnoses were subject to exclusion criteria for differential diagnoses (see Materials and methods). *Total pregnancies consist of the entire population excluding the healthy women. **Antiemetics as defined in Materials and methods section. ***Percentage of pregnancies over the total shown in the main headers.
Of the 6390 pregnancies with early NVP/HG hospital admission (1.5% overall), only 38% had evidence of a primary care prescription of antiemetics before the admission and 50% had antiemetics prescribed following the admission. Overall, 34% received first-line treatment before admission, 9% second-line, and 1% third-line treatment. Individual drugs prescribed were similar to those for unadmitted women, with prochlorperazine being the most common first-line and metoclopramide the most common second-line treatment. Following admissions, cyclizine, metoclopramide and prednisolone prescription rates doubled compared to pre-admission rates and ondansetron increased from 0.4% pre-admission to 4.6% post-admission, reflecting the follow on from the higher level of treatment lines prescribed in secondary care.
Women with late admissions from 20 weeks' gestation onwards had even lower prescribing of antiemetics before their first admission (23% of pregnancies treated preadmission) and only 8% had antiemetics prescribed postadmission. First-line treatment was prescribed in 18% of pregnancies pre-admission, with second and third-line treatment prescribed in 6 and 2%, respectively. Prochlorperazine was still the most common drug prescribed preadmission, followed by cyclizine, metoclopramide, and promethazine.

Sensitivity analysis distinguishing NVP from HG diagnoses made in primary care
Among women with recognised NVP/HG who were never admitted to hospital, the proportions of pregnancies receiving an HG diagnosis (rather than an NVP diagnosis) were 21% of those without drug treatment and 41% of those with drug treatment. These proportions remained constant over time (Supporting Information Figure S1), indicating that NVP and HG diagnoses may have been used interchangeably in the medical records. Furthermore, the distribution of key maternal characteristics and comorbidities was very similar between those with HG diagnosed and those with only NVP diagnosed (Supporting Information Figure S2 and Table S6), again providing a rationale for NVP/HG being considered as the same clinical group.

Main findings
We found that 9.1% of pregnancies had NVP/HG that was clinically recognised in primary or secondary care; 7% did not result in hospital admission but were treated by GPs with antiemetics half of the time, and 2.1% resulted in hospital admissions. In all, 38% of women admitted to hospital had received previous antiemetics in primary care. The prevalence of affected women prescribed antiemetics in primary care has increased over time, with a turning point at 2008 after which affected women were more likely to be treated than not. Hospital admissions, however, also increased over time, showing an overall increase in the recognised clinical burden of NVP/HG. Moreover, NVP and HG diagnoses were used in a similar way for antiemetic prescribing and hospital admissions, likely because health professionals considered them a spectrum of illness, despite distinguishing clinical criteria for hyperemesis gravidarum diagnosis.

Strengths and limitations
The CPRD-HES is a well validated data source 33 widely used for epidemiological research, 34 broadly nationally representative. 17,35 and internationally recognised as an extremely meaningful source of clinical information for studying pregnancy complications and prescribed treatments in England.
To our knowledge, this is the first large epidemiological study evaluating the prevalence of clinically recognised and managed NVP/HG within primary and secondary care.
One of the major strengths of this study was the possibility to assess the antiemetic treatments offered to women with NVP/HG in primary care. However, secondary care prescribing was not available and although discharge prescriptions are usually in short supply, results on prescribing after admission need to be interpreted cautiously. It is possible that we overestimated treatments used for NVP/HG, considering antiemetics have multiple indications, but we think this is unlikely, as we assessed differential diagnoses for each consultation where antiemetics were prescribed and carefully excluded prescriptions for treating other conditions, including corticosteroids used for asthma, Crohn's disease, and other auto-immune conditions. Some antihistamines are available without prescription; however, we do not think this underestimated prescriptions, as no antiemetics were licensed for use in pregnancy in England during the study period, and pregnant women receive free prescriptions from their GP. We also applied rigorous exclusion criteria for NVP diagnoses that had differential diagnoses for these symptoms, such as gastrointestinal, metabolic, and genitourinary conditions, as indicated in RCOG national guidelines for NVP/HG. 10 We acknowledge that NVP is a common symptom in many diseases and is potentially also attributable to other conditions such as diabetes or pre-eclampsia; however, as these are not differential diagnoses, it is also possible that NVP/HG co-exists with other comorbidities.
We have included an extensive analysis of risk factors; however, results could have been affected by residual confounding, as we did not include certain factors such as body mass index or family support, which were not comprehensively recorded in the data. In addition, there was sub-optimal recording of certain demographics or lifestyle factors such as ethnicity or smoking status. However, quality of data has improved over time 35 and robust information on pre-existing comorbidities and pregnancy complications was available. 33 Moreover, imputation of missing values for the affected variables was used to minimise this limitation.
We have included women admitted initially for NVP/ HG after 20 weeks' gestation, and although the classic presentation of HG is a hospital admission prior to 20 weeks, 36 some women remain symptomatic throughout pregnancy 37 so it was important to capture this group, who may represent a severe and sustained burden of HG. However, we acknowledge that those women could have been admitted for excessive NVP due to other underlying conditions such as diabetes, gestational hypertension, eclampsia or hypercholesterolemia, which have been shown to be strong risk factors of late admissions.
It is important to acknowledge that our findings represent the clinical prevalence of NVP/HG in pregnancies ending in live and stillbirths only, as we did not include pregnancies ending in spontaneous or non-spontaneous abortion. Although some studies have indicated that HG can lead to pregnancy terminations, 38 more research is needed to assess how severe NVP and HG may relate to both early or late pregnancy losses.

Interpretation
We have shown that the actual prevalence of clinically recognised NVP/HG is higher than previously reported, in agreement with a recently published study based on eight English primary care settings. 39 Using the linked CPRD-HES data source, we have been able to provide this important missing information to complete the picture of NVP and HG management. Most of the current literature that describes the prevalence of HG or NVP is based on either medical records of hospital admissions 28,[40][41][42] or questionnaires filled in by the affected women to assess the severity of symptoms in an attempt to detect the actual occurrence of HG or NVP, for which women may not always consult a healthcare professional. 9,[43][44][45] We found that overall HG was diagnosed and managed in primary care alone in 2.5% of pregnancies, of which 75% were treated with antiemetics, showing an higher HG burden than previously reported figures. In our study, the level of NVP/HG burden varied by maternal characteristics and comorbidities, consolidating the current knowledge 29 that young mothers, women of Black and Asian ethnic origin, those from more deprived socio-economic groups, and women with multiple pregnancies are generally more likely to be affected across the whole severity spectrum. In particular, women from more deprived backgrounds were much more likely to be admitted to hospital and slightly more likely to be treated with antiemetics, but had similar risk of diagnosis-only to women from less deprived backgrounds, indicating that earlier treatment may prevent later hospital admissions.
An Australian review 46 revealed that the suboptimal management offered to women affected by NVP was due to the lack of national standard guidelines, concerns about drug teratogenicity, and underestimation of the impact of NVP on women's lives. A pregnancy Sickness Support survey recently published in the UK also reported significant problems accessing treatment and high levels of dissatisfaction with care. 14 General failure of an appropriate HG treatment provision was reported nationally 15,47,48 and internationally, 49 with a consequent feeling of isolation and dissatisfaction among the affected women, exacerbated by further evidence of lack of high-quality studies to support any particular intervention. 50 Despite a general consensus that some women are denied access to antiemetics that could help relieve the severe symptoms of these conditions, 47 we found that use of antiemetics in pregnancy has increased over time. This could be a sign of rising awareness of the impact of these conditions on the quality of life together with a growing confidence in GPs' prescribing, supported by growing evidence for the safety of antiemetics in pregnancy. 10 However, we also found that women with early hospital admissions were much less likely to be treated in primary care before their admission compared with women who never experienced hospital admissions (38 versus 49%), potentially supporting the hypothesis that some hospital admissions may be preventable with timely treatment.

Conclusions
The actual burden of clinically recognised NVP/HG is larger than the reported figures, currently affecting almost 10% of pregnancies due to a proportion of women reporting clinically relevant symptoms that are managed at primary care level, half of which are treated with antiemetics. Higher NVP/HG severity levels generally confirm the consolidated knowledge of which women are more at risk of developing this condition, with no relevant differences between NVP and HG diagnosis. Doctors' confidence in prescribing antiemetic drugs to pregnant women is increasing, although 62% of women with hospital admissions were not prescribed an antiemetic, raising urgent calls to clarify whether optimal and timely treatments could help prevent hospital admissions.

Disclosure of interests
All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author). CN-P reports personal fees from Alliance Pharma during the conduct of the study; personal fees from Sanofi Aventis, grants from Leo Pharma, personal fees from Warner Chilcott, personal fees from UCB outside the submitted work. All other authors report no potential conflicts of interest. Completed disclosure of interests forms are available to view online as supporting information.

Contribution to authorship
LF conducted data management and the analysis. CN-P, SD, RK, and LJT contributed to the design and analysis and interpretation of the data, and preparation, critical review, and approval of the manuscript. The corresponding author attests that LF, CN-P, SD, RK and LJT meet authorship criteria and that no others meeting the criteria have been omitted.

Details of ethics approval
The study was approved by ISAC (Independent Scientific Advisory Committee) for MHRA Database Research (protocol number 14_165R) on 23 September 2014.

Funding
The work was funded by the Rosetrees Trust and the Stoneygate Trust through external peer review for scientific quality. The funders had no role in study design, data collection, analysis, interpretation of the results, and decision to publish, or preparation of the manuscript. All the authors are independent from the funders. Rosetrees Trust, Grant number RB48AG.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. Change in proportion of pregnancies with clinically recognised NVP/HG by level of burden, distinguishing NVP diagnosis from HG diagnosis Figure S2. Distribution of maternal characteristics across different HG and NVP level of burden groups Table S1. Read codes for NVP and HG diagnoses Table S2. ICD10 codes for NVP and HG diagnoses Table S3. Read codes for NVP differential diagnoses Table S4. ICD10 codes for NVP differential diagnoses. Table S5. Antiemetics codes used for extracting antiemetic prescriptions  Table S6. Distribution of hyperemesis gravidarum level of burden according to maternal comorbidities for women with NVP diagnosis (with or without treatment) and HG diagnosis (with and without treatment) &