Pregnancy Nutrition Core Outcome Set (PRENCOS): A core outcome set development study

To develop a core outcome set for pregnancy nutrition.


| I N TRODUC TION
Antenatal diet interventions hold potential to promote population health and wellbeing across the life course, achieve the United Nations Sustainable Development Goals and break the cycle of malnutrition. [1][2][3][4][5][6] Nutrition, including maternal nutrition, is a high-priority issue for global health and requires urgent action. 7 Most women do not follow dietary guidelines for pregnancy, and inadequate intakes of macronutrients or micronutrients will predispose women to deficiencies. [8][9][10][11][12][13][14][15] The benefits of antenatal dietary interventions include reduced maternal blood pressure, reduced gestational weight gain, reduced gestational diabetes mellitus, reduced birthweight, reduced preterm birth and reduced perinatal death. [16][17][18][19][20][21][22][23][24] Antenatal dietary recommendations relate to outcomes for which evidence of sufficient quality exists. 7 More systematic reviews and meta-analyses are needed in maternal nutrition; however, variability in the reporting of outcomes limits the number of studies that can be compared. [25][26][27] In a review of outcomes in lifestyle interventions for weight management in pregnancy, over 50% were reported once. 28 This inconsistency leads to research waste, bias towards specific outcomes and a deficit of evidence for other nutrition-related outcomes, including outcomes that are important to patients. 29,30 In pregnancy research, the limited reporting of neonatal outcomes and measures of life impact has been identified. 16,28,31 This contrasts with the preferences of pregnant women, who value measures of life impact equally, if not more than, traditional clinical measures. 32,33 Core outcome sets (COS) can address the growing and critical problem of research waste and reduce inconsistency in outcome reporting across studies. 34 A COS is a list of outcomes that are considered critically important for a subject area. 29 In pregnancy and childbirth, 26 COS have been published but no COS exists for pregnancy nutrition research. 16 We aimed to address this gap through the development of the Pregnancy Nutrition Core Outcome Set (PRENCOS).

| M ET HODS
This study was prospectively registered with the Core Outcome Measures in Effectiveness Trials (COMET) initiative. 35 The protocol was published a priori. 36 All processes align with the Core Outcome Set-STAndards for Development and the Core Outcome Set-STAndards for Reporting (Appendix S1). 37,38 Methods included a systematic review, qualitative interviews, a modified Delphi survey and a consensus meeting (Figure 1). This study received institutional ethics approval from the clinical research site in Dublin, Ireland, in July 2019 (EC24.2019). Informed consent was obtained from all participants. This publication has emanated from research supported in part by a peer reviewed research grant from Science Foundation Ireland (SFI) under grants 12/RC/2273 and 16/SP/3827, and by a research grant from PrecisionBiotics Group Ltd. The funding source did not have a role in this study.

| Participants
Three main stakeholders were included in the COS development process: healthcare professionals (HCPs), researchers and women with personal experience of pregnancy who were not HCPs or researchers. The latter group included women who were currently or had previously been pregnant. The coronavirus disease 2019 (COVID-19) pandemic restrictions on in-person activities resulted in participant recruitment taking place almost exclusively online, via social media (e.g. Twitter), supplemented through snowballing referrals from personal contacts, professional networks and participants. Participants were explicitly selected from a variety of countries to ensure global representation.

| Patient involvement
This study also involved public and patient involvement. Prior to the interviews, study materials were presented F I G U R E 1 Pregnancy nutrition core outcome set methods.
to an advisory group of women with experience of pregnancy, and feedback was obtained on aspects such as clarity and relevance to the intended population for which this COS serves. Advisory group members did not take part in the study. Details on this advisory group have already been published. 39

| Systematic review
The systematic review was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) in April 2020 (CRD42020135052) and took place on 27 February 2020. 40 Results are reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Appendix S2; Figure 2). 41 The protocol was published previously. 36 In brief, studies were included if they had an intervention aiming to influence dietary intakes of foods/macronutrients in pregnancy or involved observations on dietary indices and outcomes in pregnant women. 36 Four databases were searched simultaneously and grey literature was not reviewed. The details of the search varied with each database but included variations of the terms 'pregnancy' and 'dietary intakes' (Table S1). No limits were added to the search, but articles were excluded if: (1) the intervention or the exposure was outside the scope of the research question (e.g. micronutrient intakes or descriptive studies); (2) the study design was ineligible (case reports, case series, case-control studies, commentaries, letters to the editor, narrative reviews or expert opinions); (3) it was not possible to identify the unique relationships of the dietary variables during pregnancy; or (4) the article was not written in English. 36 Search terms were reviewed by a librarian at University College Dublin and were approved by the researchers. After the removal of duplicates, the screening of titles, followed by abstracts, was completed independently by two reviewers (SLK and SC). Any disagreement was discussed between the reviewers and if not clarified, escalated to a third researcher. Only the full text of articles published in English were obtained. Data extraction was divided between SLK and SC, with each reviewing and manually extracting data independently from the assigned studies. 36 The reviewers cross-checked 10% of the studies for accuracy.

| Qualitative interviews
Details of the qualitative interviews were reported according to the Consolidated Criteria for Reporting Qualitative Research (COREQ) checklist for qualitative interview reporting (Appendix S3). 42 A pragmatic epistemological approach was employed to study the phenomenon of the lived experiences of women during pregnancy. 43 A semi-structured topic guide was developed by SLK and SOR, who are experienced qualitative researchers, and was piloted with members of the research team ( Table 1). The one-to-one interviews were conducted by SLK (a female registered dietitian, BSc, and PhD student) and no prior relationship existed with any participant. Participating women were provided with an information sheet, a short explanatory video summary and infographics, and informed written consent was received prior to interview. The interviews took place between December 2020 and January 2021 via Zoom and lasted 30-60 minutes. No repeat interviews were carried out. Summarisation and checking were employed to ensure that participants were understood appropriately. The understanding of the Delphi survey process and lay definitions for candidate outcomes were tested during the interviews. 36 Feedback was incorporated into the remaining study material activities. The interviews were digitally recorded and transcribed verbatim. Participant IDs were assigned and identifying information was removed from the transcripts. Field notes were taken by SLK and reviewed as part of the analysis. The demographic data collected were limited to the woman's experience of pregnancy and location.

| Outcome identification
All eligible outcomes from the interviews were brought forward to the Delphi survey. This was supplemented with outcomes that were identified from the systematic review. All outcomes were categorised using the COMET initiative taxonomy and outcomes were grouped under the appropriate outcome domains. 44 Clinically similar outcomes were grouped together into overarching outcomes, as done by several previously published COS in this field. [45][46][47][48][49][50][51] 'Bowel habits' for example, was included under 'gastrointestinal symptoms' and several cord blood outcomes were condensed under 'cord blood markers'. Conversely, given the importance of maternal wellbeing to pregnant women, it was felt that this outcome should remain separate to 'mental health', which specifically records mental health conditions. Pregnancy-specific outcomes such as birthweight have been defined in other pregnancy COS, so we did not seek to define these individual outcomes again. Instead, the focus of this COS was mostly on nutrition outcomes. Previously defined outcomes were grouped as appropriate, such as 'newborn anthropometry and body composition'. Full details of the change from 69 to 30 outcomes are found in Tables S3  and S4, and were included in the 'help text' of the Delphi survey.

| Delphi survey
We conducted a modified two-round Delphi survey using the web-based DelphiManager™ system (www.cometinitiative.org/delphimanager). 52 Participants were sent an information sheet and informative video prior to participation, informed by previous research. 53,54 A pragmatic target of 60 participants was set a priori. 36,55 The survey took place from May-August 2021. For round one, participants were given 6 weeks to rank each outcome using a nine-point Likert scale, as recommended by the Grading of Recommendations Assessment, Development and Evaluation (GRADE). 56 Participants could select 1-3 or 'not essential' for inclusion, 4-6 or 'important but not critical', 7-9 or 'critically important for inclusion' or 'unable to score'. 36 Participants received two reminder emails, spaced 2 weeks apart, to encourage completion. Participants were provided with their scores from round one and the median score per stakeholder group for each outcome. Participants were given 6 weeks to complete round 2, but because of the lower response rates, a deviation from the protocol was employed, including a final reminder and an additional 2 weeks to respond. If ≥70% of participants selected 'critical for inclusion' (score 7-9), the outcome was considered to have reached 'consensus in' and was included in PRENCOS. Alternatively, 'consensus out' was achieved if ≥70% of participants selected 'not essential' (score 1-3). All other outcomes were reviewed at the consensus meeting.

| Consensus meeting
The consensus meeting was conducted on 15 September 2020 via Zoom and was moderated by an independent chair (SLK). Nine expert healthcare professionals or researchers with experience in pregnancy nutrition attended from medicine, midwifery, and dietetics and nutrition backgrounds. The geographical spread covered Australia, Ireland, Malaysia, Switzerland, Taiwan and the UK. All attendees had at least 1-3 years of experience with pregnancy nutrition research and six attendees had over 10 years of experience. The PRENCOS researchers (FMcA, SOR and F I G U R E 2 PRISMA flow chart. *Titles were not retrieved if the work was an abstract only (e.g. poster presentation) or was published in a language other than English. SC) also attended. Prior to the consensus meeting, the panel were sent their individual scoring for each round and outcome of the Delphi survey. Following the presentation of the round-2 Delphi results, the panel discussed the outcomes that did not reach 'consensus in'. Participants were asked to vote anonymously on whether the outcome was 'critical for inclusion' using Poll Everywhere (pollev.com). 57 Attendees (n = 12) had a single vote, and votes were treated equally. A lower threshold for inclusion of ≥50% among the panel was set where women with experience of pregnancy ranked an outcome as critical for inclusion. The panel considered this important, as women with experience of pregnancy were not included in the consensus meeting. This protocol deviation was endorsed at the meeting.

| Data analysis
EndNote (Clarivate, London, UK) was utilised for the systematic review data management and data were exported to Microsoft Excel (Redmond, WA, USA) files for the purpose of systematic review and data extraction. 36,58 Interview transcripts were analysed by SLK using inductive thematic analysis. 59 The coding was completed using NVivo12 (Lumivero, Denver, CO, USA). 60 Unique coding was applied to the data in response to all identified outcomes. The results of the Delphi survey were managed through the DelphiManager™ platform and analysed using SPSS 26.0 for Windows (IBM, Armonk, NY, USA).

| Outcome identification
The full text of 691 articles were reviewed and outcomes were extracted from 427 eligible studies (Appendix S4). Trials made up 35.4% (n = 151) of studies and 4.9% (n = 21) were secondary analyses of trials. Qualitative interview participants (n = 26) had between one and four previous pregnancies and lived in a variety of countries, including Australia, Denmark, India, Ireland, Pakistan and the USA (Table S2). The total number of unique outcome codes identified from the qualitative interviews was 113, but only 45 were relevant. Before refinement, 69 outcomes were identified: 45 were in both the qualitative interviews and systematic review and 24 came only from the published literature (Table S3). This list was consolidated after review by the researchers and 30 outcomes were included in the Delphi survey (Table S4).

| Consensus process
A total of 138 invitations were sent, 90 individuals consented to take part in the Delphi survey and 82 participants ranked all outcomes in round 1 (Table S5). The breakdown of stakeholders who completed all of the questions was 28% (n = 23) HCPs, 37.8% (n = 31) researchers and 34.2% (n = 28) women with experience of pregnancy. Nine additional outcomes were suggested in round 1: 'placental weight', 'previous pregnancytime since; outcome; weight before; weight after', 'breastfeeding', 'sitting time', 'knowledge about food safety in pregnancy, e.g. risk of bacterial infection in foods, what to avoid etc.', 'autism', 'long-term sequelae in mother (diabetes, obesity, metabolic syndrome) measured post pregnancy', 'development of the infant at 2 years and possibly beyond' and 'pre-pregnancy or early pregnancy weight and BMI (needed for GWG categories)'. After review against inclusion criteria and other candidate outcomes, 'placental weight' was added to round 2. The second round was completed by 73.2% (n = 60) of round-1 participants. Round-2 stakeholders included 23.3% (n = 14) HCPs, 46.7% (n = 28) researchers and 30% (n = 18) women with experience of pregnancy. HCPs had the greatest attrition rate (39.1%). Results for round 2 of the survey are reported for all participants in Table 2 and by stakeholder group in Table S6. After round 2, 11 outcomes reached 'consensus in' and none reached 'consensus out' ( Table 2). Out of the remaining 20 outcomes, two were voted as critical for inclusion in PRENCOS at the consensus meeting. The outcome 'delivery complications' was voted in by 100% of attendees. Although only 53% of consensus meeting attendees voted in 'maternal wellbeing', 99% of women with experience of pregnancy voted this as critical for inclusion in the Delphi survey. The outcome 'anaemia' was incorporated with 'maternal vitamin and mineral status' based on consensus discussion. The final PRENCOS resulted in 13 outcomes (Figure 3).

| Main findings
We identified 13 core outcomes that are critical for inclusion in pregnancy nutrition research studies. Most of these outcomes, including pregnancy, delivery or neonatal T A B L E 1 Semi-structured topic guide for PRENCOS interviews.

Main questions Prompts/related questions
What is your experience of pregnancy?
• How many pregnancies and children did you have?
• What were those experiences like?
How did diet affect your pregnancy?
• What effect did your diet have during pregnancy?
• Were you trying to achieve anything when making dietary changes/deciding what to eat or drink?
What outcomes do you think are related to diet in pregnancy?
• Which of these should we measure?
These are the outcomes we have found so far (show list) • What do you think of these?
• Is there anything missing?
complications, pregnancy loss or perinatal death, birth defects or congenital anomalies and neonatal anthropometry, are routinely but not universally measured, as highlighted by Cochrane. [61][62][63] This should limit the burden on clinical trialists. 16 The endorsement of these outcomes was expected given that they are clinically relevant to maternity HCPs and have significant impact for pregnant women. Their inclusion in PRENCOS will support implementation, increase the evidence base for the role of diet in addressing these critical outcomes and address associated gaps in the literature. 61

| Interpretation
We found women with experience of pregnancy favoured maternal wellbeing as a critical outcome. A lack of measurement of 'salutogenic' outcomes, defined as outcomes relating to positive maternal and neonatal health and wellbeing, has been identified in maternity research. 31,64,65 The World Health Organization (WHO) diet-related evidence base does not include measures of life impact, probably because of a lack of evidence for these outcomes. 16,44 Recent clinical practice guidelines in pregnancy recommend screening for mental health concerns. 66 Most COS in pregnancy and childbirth do not focus on mental health, although the recently published COS on postpartum haemorrhage identified maternal sense of wellbeing as a critical outcome for inclusion. 67,68 The inclusion of both maternal mental health and wellbeing in PRENCOS will help address this research gap and inform the role of antenatal dietary intakes in improving the maternal experience of pregnancy. Generating consistent data on dietary outcomes will help bring dietetics to the frontline of pregnancy research. [69][70][71] Already, evidence on protein or energy intakes and anaemia exists, allowing for systematic reviews. 16,20,62,72,73 Greater evidence on maternal diet quality will address the practical nutritional issues identified in clinical practice, including a lack of achievement of dietary standards. [8][9][10][11][12][13][14][15][74][75][76] Participants also ranked gestational weight change favourably, resulting in the inclusion of this outcome in the COS. The relationship between gestational weight gain and pregnancy outcomes has been illustrated many times, across all continents. [77][78][79] Recent evidence reiterates the relationship between interventions to promote healthy weight gain in pregnancy and Note: 'Consensus in' was achieved when ≥70% of participants scored the outcome as critical for inclusion and 15% or less scored the same outcome as not important (ranked 1-3 on Likert scale). All missing values are a result of participants selecting 'unable to score' for that outcome. All participants therefore completed all questions. The n reported is the number scoring the outcome from 1-9.
T A B L E 2 (Continued) optimal maternal and child outcomes. 80 Standardisation in reporting will further enhance the quality and certainty of evidence in this area. 81

| Strengths and limitations
This study has several strengths. We engaged multiple stakeholders across six continents with widespread expertise in our protocol and COS development. 36 We conducted a large systematic review with four databases, performed qualitative interviews and held a consensus meeting after the Delphi survey. 82 Shi et al. found that only 34.1% of COS in obstetrics and gynaecology used four or more databases, 36.4% included a consensus meeting and only three interviewed key stakeholders. 83 Another study found that 35% of COS in pregnancy and childbirth did not include or only partly included patients. 67 If COS studies do include lay experts, most do so in the Delphi survey. 84,85 We included women with experience of pregnancy in multiple aspects of the study and conducted patient and public involvement (PPI) with an advisory panel of women before finalising the protocol. The use of qualitative methodologies can increase confidence that all relevant outcomes have been identified and encourage appropriate language use in the Delphi survey. 86 It has been suggested that COS developers should ensure that outcomes important to patients are included in the COS, and we have done this in our study while balancing this against the challenges of participant burden. 87 The number of outcomes influences the usability, uptake, and impact of COS. 67 In pregnancy and childbirth, the number of outcomes in COS ranges from six to 51, and 35% include over 20 outcomes, which is considered large for COS in this area. 67 We incorporated processes to refine outcomes throughout the study, including providing stakeholderspecific results (Table S6). 88,89 The final number of outcomes was 13, which is the median value for all obstetric COS. We also employed methods to support study retention, including social media campaigns, keeping participants informed on progress and offering personal acknowledgement. 54,90 As a result, our attrition rate of 26.8% is in line with the literature, which reports attrition rates of up to 48%. 54,85,91 The restriction of our systematic review to studies reported in English may have impacted the identification of outcomes, and running the study in English only could have influenced participation by individuals from non-English speaking countries. 92 Nevertheless, we had representation from English-speaking participants from both high-and low-to middle-income countries. We had 18 women with experience of pregnancy completing both rounds of the Delphi survey. This is lower than our target, and is a challenge experienced by other COS developers in maternal and infant health. 93 Our numbers, however, are higher than several other obstetrics COS, where 'patient' participants have ranged from 0 to 25 completing round 2. 36,91 As a result of in-person restrictions associated with the COVID-19 pandemic, our recruitment took place mostly online. Future studies could consider care-based recruitment strategies to further increase patient participation. 94 We used a nine-point Likert scale in the Delphi survey, as this is the suggested scale for Delphi surveys and has been shown to produce smaller COS compared with a five-point Likert scale. 46,95 On the other hand, the use of a five-point scale may support patient participation as greater numbers of options in a rating scale may limit the respondent's ability to discriminate. 46,96 The methodology of COS development is an evolving field. 97 On this basis, COS developers need to tailor development to the context and circumstances of the particular COS. 97 We reduced our list of 69 outcomes and brought 30 outcomes to the Delphi survey. Although we followed a similar approach to many previously published COS in this area and the final outcomes resembled those of several other COS, the subjectivity of this is a limitation of the study. [47][48][49][50][51]93,98,99 Although it is accepted that COS do not necessarily need to F I G U R E 3 Final pregnancy nutrition core outcome set.
include measurement instruments, another limitation of this study is that we did not explore the preferred outcome measurement tools for the core outcomes, of which there is further heretrogenity. 51,100 Although identifying outcome measurement tools was beyond the scope of this study, there are guidelines that can inform this approach, ideally with one measurement tool for each outcome, ensuring it has content validity. 101,102 Measurement instruments vary with changes or advances in the field, and therefore identifying core outcomes is still of value.

| Research recommendations
With the uptake and use of PRENCOS, evidence will build for the outcomes that were considered critical for pregnancy nutrition research and enable comparison among a large volume of studies. The use and evaluation of PRENCOS in dietary studies in pregnancy will support the uptake of the COS. 103

| Clinical recommendations
This COS is designed for clinical research, but the findings will inform outcome measurement in antenatal practice. Highquality evidence generation will also inform clinical practice guidelines and HCPs will benefit from insight into the priorities of women with experience of pregnancy, including prioritising wellbeing among other routine measures.

| Future directions
The outcomes in PRENCOS are a guide for researchers, are based on consensus and offer flexibility, given their broad nature. Study-specific adjustments or deviations may need to be made based on the individual study, especially in the context of rare outcomes. The 13 outcomes relate to pregnancy or delivery only, but future work could explore expanding these outcomes into postpartum and other time points.

| CONCLUSION
Clinical trialists investigating dietary interventions in pregnancy should consider measuring the 13 outcomes in PRENCOS as a minimum. Future research can consider developing standards for the preferred measurement tool for each outcome.