Improving pregnancy outcome in obese women: the UK Pregnancies Better Eating and Activity randomised controlled Trial

To assess appropriateness, acceptability and delivery of the intervention in target groups

To understand likely motivations for, and barriers to, changes in diet and activity in pregnancy by developing a quantitative measure of knowledge and attitudes about healthy eating and activity.

At the onset of this programme there were no pre-existing validated instruments to assess attitudes towards diet and PA in obese pregnant women. Initial work was undertaken to explore the relative clinical effectiveness of different behaviour change strategies used in previous studies of healthy lifestyle interventions for pregnant women. All relevant studies had focused on the reduction of gestational weight gain (GWG) as the primary end point. None was targeting insulin resistance, which was the focus of this programme. Ten controlled trials of interventions that aimed to reduce GWG through changes in diet or PA were reviewed. Meta-analysis showed that, overall, diet and PA change was effective in reducing GWG, but there was considerable heterogeneity in outcomes. 2 The analysis showed that sample characteristics and aspects of intervention design, content, delivery and evaluation differed between studies, and that these were likely to explain the variation between studies in effectiveness. A common issue was the failure to evaluate changes in behaviour or its psychological determinants, as well as inadequate detail of the intervention content. These were likely to have contributed to difficulty in identification of the processes by which weight change was effected. Because of this, it was difficult to discern active intervention ingredients. The study concluded that behaviour-based GWG reduction interventions should be more systematically designed, evaluated and reported to build on insights from behavioural science. This important conclusion reinforced our intention to systematically develop an evidenced-based intervention based on known theory, and our decision to pilot the intervention to evaluate all aspects of the intervention, particularly to evaluate change in behaviour, acceptability and its psychological determinants.

A questionnaire to measure attitudes towards diet and PA in pregnancy was then developed, based on psychological models of determinants of health behaviour, previous literature in pregnant women and interviews conducted with pregnant women (see Appendix 1). The questionnaire comprised the following constructs: attitudes to healthy behaviours, social norms, intention to change behaviour, knowledge regarding nutritional and PA recommendations and motivations for food choices. The outcomes of a study using this questionnaire were published as below by Gardner B, Croker H, Barr S, Briley A, Poston L, Wardle J; UPBEAT Trial. Psychological predictors of dietary intentions in pregnancy. J Hum Nutr Diet 2012;4:345–53. 3 Minor formatting edits have been made to the published text, which is reproduced with permission.

Abstract

Background: Consuming a healthy diet in pregnancy has the potential to improve obstetric outcome, including minimising the risk of macrosomia. Effective promotion of dietary change depends on identifying and targeting determinants of gestational diet. The present study aimed to model psychological predictors of intentions to reduce the intake of high-fat and high-sugar foods, and increase fruit and vegetable (F&V) consumption, among pregnant women.

Methods: One hundred and three pregnant women completed questionnaire measures of intentions to modify the consumption of the target foods, current intake, perceived vulnerability to, and severity of, adverse outcomes of non-healthful consumption of these foods (i.e. ‘threat’), benefits of dietary change to mother and baby, barriers to dietary changes and social approval for dietary change (‘subjective norms’). A cross-sectional design was used. Logistic regression analyses were undertaken to model dietary change intentions.

Results: Participants who reported excessive current intake of high-fat and high-sugar foods were more likely to commit to reducing the intake of these foods. The perceived benefits for mother and baby enhanced the mothers’ intentions to eat more F&V and fewer high-fat foods and marginally significantly increased their intentions to reduce the consumption of high-sugar foods. There were no effects of threat, barriers or subjective norms.

Conclusions: Lack of effects for barriers, threat and subjective norms may indicate that pregnant women discount barriers to health-promoting behaviour, understand the threat posed by unhealthy eating and perceive social approval from others. Dietary change interventions for pregnant women should emphasise likely positive outcomes for both mother and child.

Introduction

Research in gestational diet and nutrition has traditionally focused on preventing nutritional deficiencies in the maternal diet and ensuring adequate neonatal growth. 4 There is, however, growing interest in the potential for dietary changes among pregnant women with GDM or obesity to minimise the risk of macrosomia and associated adverse outcomes, including the risk of obesity in the child in later life. 4–6 Maternal glucose is the main substrate for the growing fetus and, consequently, maternal hyperglycaemia may result in fetal hyperinsulinaemia, thereby increasing growth rate and thus the risk of macrosomia. 7 Maternal insulin sensitivity reduces with advancing pregnancy. 8 Certain dietary interventions might, theoretically, reduce insulin resistance among pregnant women. 9 Reducing the consumption of foods rich in carbohydrates that release glucose rapidly into the bloodstream [i.e. high-glycaemic index (GI) foods] can promote glycaemic control and improve insulin sensitivity by manipulating the type of carbohydrate-rich foods and drinks consumed rather than necessarily restricting the quantity of carbohydrates per se. 7,10 A recent review found some evidence to support the use of low-GI diets, defined by the consumption of carbohydrates that release glucose gradually, in women with GDM, although it is less clear whether or not they could also benefit non-diabetic women. 7 Two intervention trials in non-diabetic women have yielded promising results, with reduced delivery rates of babies who are LGA (i.e. > 90th birthweight percentile) being observed among women who ate a low-GI diet relative to those consuming a usual diet. 11,12 Notably, both studies included women with a range of body weights, suggesting that dietary glycaemic control may benefit both normal and overweight women, although there are some concerns over ensuring adequate fetal growth, particularly in normal-weight women. 13 Another study investigating overweight and obese pregnant women found that a low-glycaemic load (GL) diet (i.e. foods with low values on an index that accounts for both GI value and carbohydrate content) improved cardiovascular risk factors, lengthened pregnancy duration and increased infant head circumference. 14 Further clinical trials are planned or are under way that aim to explore the impact of such diets on pregnancy outcome in obese women, these include the UK Pregnancies: Better Eating and Activity Trial [UPBEAT; International Standard Randomised Controlled Trial Number (ISRCTN) 89971375], described in this programme, and another in women who have had a previous macrosomic pregnancy. 15

Reducing the GI relies on changing dietary behaviour in pregnant women. Changing pregnant women’s food choices will be aided by identifying and targeting the determinants of diet in pregnancy. Behaviour change is often portrayed as a consequence of changes in psychological variables: modifying knowledge, attitudes and beliefs in turn influences intentions16 and ultimately alters behaviour. 17 However, interventions designed to improve diet in pregnancy, including those designed specifically to lower the GI or GL,12,14 have neglected psychological changes. 2,18 Identifying psychological determinants of dietary decisions in pregnancy would assist intervention development in two respects. First, psychological variables represent potential targets for intervention. Second, assessing changes on these variables can aid our understanding of why diet has or has not changed in response to an intervention. 17 However, little empirical evidence is available regarding the psychological determinants of dietary behaviour in pregnancy.

Pregnancy has been portrayed as a transitional life event that is psychologically characterised by heightened awareness of (and responsiveness to) threats to the health of either mother or child. 19 Risk perceptions may therefore underpin motivation to take health-protective action in pregnancy. The health belief model (HBM)20 proposes that responses to risk are underpinned by two psychological dimensions: threat perception and behavioural evaluation. 21 Threat perceptions are a function of perceived susceptibility to, and perceived severity of, a threat (e.g. GWG or its associated health complications), and behavioural evaluations refer to the expected benefits of, and barriers to, taking action (e.g. consuming a nutritionally balanced diet) to avert the threat. The HBM predicts that protective behaviour will be elicited where the individual perceives themselves as vulnerable to a serious threat, the action is deemed beneficial and there are few barriers to taking action. Although originally proposed as direct determinants of health behaviour, subsequent research has indicated that the HBM constructs largely influence action indirectly via the formation of intentions, which subsequently determine behavioural responses. 16

As determinants of intention, the threat perceptions and cost–benefit analyses proposed by the HBM compete with social influences in guiding health motivation in pregnancy. 22 For example, intentions to consume a healthy diet in pregnancy are likely to be a function not only of the subjective utility of a nutritionally balanced diet for averting health threats, but also expected (dis)approval from relevant others for consuming such a diet. 22 Expectations of friends, family and health-care professionals may therefore contribute to dietary decisions in pregnancy independently of a subjective threat and behaviour evaluations.

Dietary change focused on a reduced intake of saturated fat and high-sugar foods, as well as increased F&V consumption, has been proposed as an approach for reducing insulin resistance in pregnancy,9 and is consistent with the UK national guidance for healthy eating in pregnancy. 23 The present study assessed pregnant women’s attitudes and motivations concerning these food types, and examined which psychological variables predicted intentions to consume healthier quantities of these foods over the remainder of the pregnancy term. Our analysis draws on variables from the HBM,20 augmented by a measure of perceived social approval (i.e. ‘subjective norms’). 22 To isolate the unique influence of psychological variables on behaviour change intentions, we controlled for the current intake of each of the three foods, as well as the pregnancy-related personal characteristics that may influence dietary intentions or behaviour [pre-pregnancy body mass index (BMI), gestational age and parity]. 24,25

Materials and methods

Results

Discussion

A low-GI diet in pregnancy has the potential to improve obstetric outcome. 12,14 The design of diet-based interventions in pregnancy may be aided by identifying the psychological determinants of dietary choices because modifying these variables should translate into dietary change. 17 The present study examined appraisals of benefits and barriers, threat perceptions, evaluations of current behaviour and perceived social pressures (subjective norms) as potential influences on intentions to improve three aspects of diet linked to insulin resistance (increased F&V consumption, decreased high fat and high sugar consumption19) in a community sample of pregnant women. Perceived levels of current behaviour predicted intentions to eat fewer high-fat and high-sugar foods, with participants who felt that they ate too many high-sugar or high-fat foods being more likely to intend to reduce their consumption. Perceived benefits of action for mother and baby were associated with the intention to eat more F&V and less high-fat food, with participants who perceived greater benefits of adopting more healthy eating patterns being more likely to intend to do so. A similar association, which was only marginally significant, was observed between perceived benefits and intention to eat less high-sugar food. There were no associations between perceived barriers, threat perceptions and subjective norms and intentions.

Our analysis drew on variables derived from the HBM,20 which proposes that health behaviour arises from deliberation over the threat posed by health risks, as well as the benefits of, and barriers to, taking action recommended to minimise these threats. We observed consistently high levels of perceived threat among our sample and, perhaps as a result of minimal variation in threat levels, there was no association between threat and dietary intentions. As predicted by the HBM, perceived benefits for mother and baby were associated with healthy eating intentions, although, in contrast to theoretical predictions,16 perceived barriers had no impact. Although further evidence from larger samples is needed, these results suggest that, when making the cost–benefit analyses that underpin evaluations of health-related behaviours,28 pregnant women may weigh barriers to behaviour less heavily than do the general population. Perhaps this may be understood in the light of the immediacy with which health threats faced by pregnant women can be realised. Perceived health benefits of diet outside pregnancy may be typically distal and orientated towards the prevention of later morbidity or mortality, whereas barriers relate to immediate short-term obstacles (e.g. missing out on tasty foods, choosing less favoured options). By contrast, the benefits of healthy eating in pregnancy, such as the prevention of macrosomia and the reduced likelihood of delivery complications, are typically more proximal, becoming apparent at or shortly after birth. Consequently, these benefits may be easier to foresee or are perceived to be more real than the distal benefits associated with health behaviours outside pregnancy. To support this hypothesis, further exploration is needed of the health decision processes in pregnancy and other conditions when the health effects are more immediate than the decision processes in non-pregnant community samples. However, the stronger effects of perceived benefits than perceived barriers in our sample, coupled with raised threat perceptions, support the conceptualisation of pregnancy as a period of heightened responsiveness to potential health risks and a greater appreciation of the value of health-protective action. Pregnancy may therefore represent ‘an opportune time to initiate (behavior) change’.

Several behaviourally based pregnancy interventions have been described based on the provision of dietary ‘counselling’. 29–33 However, the term ‘counselling’ can refer to a variety of behaviour change techniques. 34 Although some interventions have focused primarily on identification of barriers to healthy eating,30,32 the results of the present study indicate that bolstering expectations of positive outcomes for both mother and child associated with adopting a healthier diet may have more impact. The under emphasis of the benefits of healthy diet in interventions tested to date may reflect an assumption that pregnant women are aware of the implications of gestational diet. 2 Changing dietary choices in pregnancy may require efforts to ensure that pregnant women consistently prioritise the benefits of healthy eating over beliefs that support consumption of unhealthy foods. No relationship was observed between perceived social approval from family or health-care professionals and healthy eating intentions. However, the high mean scores, coupled with the strong positive correlation between family and health-care professional expectations, suggest that our sample perceived strong and consistent approval for healthier dietary choices. The absence of an association between norms and intentions may therefore be the result of a lack of variation, although more work is needed to establish the generalisability of the normative beliefs of our sample.

The findings of the present study are limited in several respects. The survey was cross-sectional, and so we could not observe ‘prediction’ of intentions in a temporal sense or assess the effects of intentions on subsequent consumption of each of the three foods types. Although further longitudinal research is needed, there is considerable theoretical and empirical evidence to support the temporal relationships that we have inferred, as well as the assumption that modifying dietary intentions will change dietary behaviour. 35 The data were also based on self-report, which is susceptible to responses biased by participants’ motivations to portray themselves positively. 36 Retrospective self-reports may underestimate true pre-pregnancy weight especially in obese women,37 although objective weight data were not available to validate our BMI measure. Participants were not informed by us of weight gain or gestational diet recommendations or of the potential consequences of healthy and unhealthy eating in pregnancy, and the accuracy of their perceptions was not tested and cannot be estimated. Nonetheless, our findings suggest that modifying such perceptions may boost healthy eating intentions over the remainder of pregnancy. The representativeness of our sample is questionable; for example, participants were mostly white and well educated, and more than two-thirds were educated to university level. This may reflect a systematic participation bias within our pool of potential participants, although we cannot test this because no data were available from those who declined participation. However, our sample encompassed a wide distribution of participant BMI scores and gestational age. Most participants were in the second or third trimester of pregnancy, and more work is needed to explore beliefs and behaviour earlier in pregnancy because dietary changes achieved in early pregnancy and maintained over the remaining gestational period may be most beneficial. Our sample was also modest in size and, therefore, unable to identify small effects, neither could we reliably explore differences in the beliefs, intentions and behaviour of demographic subgroups of pregnant women. Further research, using larger samples, is needed into whether or not nutrition beliefs, intentions and behaviour differ systematically by, for example, parity, age, socioeconomic status or ethnicity. Such differences may have implications for the design and delivery of effective dietary interventions in pregnancy.

Notwithstanding these limitations, our results offer some insight into the health beliefs and dietary choices of pregnant women. Best practice for diet modification in pregnancy is likely to require the adoption of health promotion strategies to target the underlying psychological determinants of gestational diet.

Exploratory study: interviews with obese pregnant women to inform the development of a pilot randomised controlled trial intervention and protocol

This pre-trial exploratory interview study aimed to examine the sociocultural context within which women made decisions about pregnancy, diet and lifestyle, body image, health beliefs and their readiness to engage with the type of health interventions that the trial team was in the process of developing.

A total of 30 women were approached, and semistructured qualitative interviews were undertaken with 13 pregnant women with a BMI of ≥ 30 kg/m². Maximum variation sampling was used, selecting by maternal age, family composition (e.g. lone parent, first baby) and sociodemographic characteristics (employment, education level, ethnicity). The interviews aimed to explore pregnant women’s existing beliefs about diet and activity in pregnancy, their receptiveness to dietary change, perceived economic and time implications, increased activity levels and other views on acceptability. We explored obese women’s preferred approaches to intervention delivery (e.g. individual or group contacts, inclusion of family members, frequency of contacts, setting for intervention delivery, etc.) and barriers and constraints to change, including the impact of family circumstances and need for childcare. We investigated what types of activity obese pregnant women perceive to be feasible and what they believe to be the key features of a healthy diet in pregnancy. The interview schedule is provided in Appendix 2. Interviews were taped, transcribed and analysed using a framework approach geared to producing policy- and practice-relevant findings. The interviews were undertaken at either Guy’s and St Thomas’ Hospital, London, or the Royal Victoria Infirmary, Newcastle. The qualitative studies were given ethics approval [Integrated Research Application System (IRAS) reference number 09/H0802/05].

Conclusions

The phase 1 exploratory study, undertaken with pregnant women with a BMI of ≥ 30 kg/m² who matched criteria for inclusion in the proposed UPBEAT pilot RCT other than gestational age at recruitment and concurrent underlying health problems (see Implications for intervention development and limitations) but were not involved in the subsequent pilot RCT, suggested that willingness to change lifestyle was diminished by pregnancy illness and symptoms and by the demands placed upon women’s time by their family and work responsibilities. A lack of concrete knowledge about the effect of changing diet or increasing PA on pregnancy or birth complications also appeared to reduce the perceived value of the intervention. Previous experience of being stigmatised by peers or health professionals affected some women’s willingness to participate. Although most of these findings were also identified during the pilot trial process evaluation, the influence of stigma was less apparent. It is unclear whether or not this is because women who felt stigmatised were less likely to participate in the pilot RCT or whether modifications to the recruitment approach reduced the experience of stigma.

Physical Activity Measurement Study

The measurement of PA in a large number of subjects within a trial setting presents practical and financial issues. Accurate objective assessment would be the ideal, and at the time of study validation studies had shown that certain accelerometers could provide reasonably accurate objective measures. However, expense and collection and download of the accelerometer data were preclusive for the purposes of the main trial. A study was therefore designed to determine whether or not pedometers could be used instead of accelerometers in the pilot trial and the main RCT. This has been published as Kinnunen TI, Tennant PW, McParlin C, Poston L, Robson SC, Bell R. Agreement between pedometer and accelerometer in measuring physical activity in overweight and obese pregnant women. BMC Public Health 2011;11:501. 38 © Kinnunen et al. ; licensee BioMed Central Ltd 2011. This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. For the purposes of this report minor formatting edits have been made to the original text.

Abstract

Background: Inexpensive, reliable objective methods are needed to measure PA in large-scale trials. This study compared the number of pedometer step counts with accelerometer data in pregnant women in free-living conditions to assess agreement between these measures.

Methods: Pregnant women (n = 58) with a BMI of ≥ 25 kg/m² at a median of 13 weeks’ gestation wore a GT1M (ActiGraph, Pensacola, FL, USA) accelerometer and a CW701 Digi-Walker™ Pedometer (Yamax, Bridgnorth, UK) for four consecutive days. The Spearman rank correlation coefficients were determined between pedometer step counts and various accelerometer measures of PA. Total agreement between accelerometer and pedometer step counts was evaluated by determining the 95% limits of agreement estimated using a regression-based method. Agreement between the monitors in categorising participants as active or inactive was assessed by determining kappa.

Results: Pedometer step counts correlated moderately (r = 0.36–0.54) with most of the accelerometer measures of PA. Overall step counts recorded by the pedometer and the accelerometer were not significantly different (medians 5961 vs. 5687 steps/day; p = 0.37). However, the 95% limits of agreement ranged from –2690 to 2656 steps/day for the mean step count value (6026 steps/day) and changed substantially over the range of values. Agreement between the monitors in categorising participants to active and inactive varied from moderate to good depending on the criteria adopted.

Conclusions: Despite statistically significant correlations and similar median step counts, the overall agreement between the pedometer and accelerometer step counts was poor and varied with activity level. Pedometer and accelerometer steps cannot be used interchangeably in overweight and obese pregnant women.

Background

Current recommendations emphasise that regular moderate-intensity leisure-time PA during an uncomplicated pregnancy may have benefits such as reducing fatigue, back pain, stress and depression and improving glycaemic control but has no known harmful effects on the health of the mother or the fetus. 39,40 However, the available evidence is limited, and larger and better-quality trials are needed to define the potential role for PA promotion in preventing pregnancy complications such as GDM and pre-eclampsia. 41 Most previous studies assessing PA in relation to pregnancy outcome have used questionnaires or other self-reported measurements of PA, as these are cheap to administer in large-scale studies. 42 Although some of the questionnaires have been validated in pregnant women either against accelerometer,43–45 a portable activity monitor46 or pedometer and a PA logbook,47 their validity has usually been low or moderate, especially with regard to the low-intensity activity that is common among pregnant women. 48 These limitations are also observed for PA questionnaires in other populations. 49,50 Therefore, inexpensive, objective PA measurement methods are needed for large-scale studies to obtain more accurate information on PA levels during pregnancy. Accelerometers and pedometers are the most commonly used objective methods of assessing PA in epidemiological studies, and have been used in a number of previous studies of pregnant women. 43,44,48,51–54 Although accelerometers provide more detailed information on PA than pedometers, pedometers are much less expensive and, therefore, more economically feasible for larger studies. 55 It is unclear whether or not pedometers and accelerometers provide comparable estimates of PA in pregnant women. This issue was recently explored in two small studies (n = 30 in both cases) examining pregnant women in free-living conditions56 and on a treadmill. 57 Similar comparisons have also been reported in healthy adults58,59 infected with human immunodeficiency virus60 and older people61 in free-living conditions. These studies suggest that pedometer and accelerometer step counts are highly correlated, but large individual differences in step counts exist. Nevertheless, pedometer step counts for assessing overall PA were advocated in most of these studies. 56,59–61

This study was designed as a preliminary investigation to determine appropriate PA measurement methods for a large RCT (UPBEAT) of a lifestyle intervention in obese pregnant women. Overweight and obese women have a higher risk of several pregnancy complications and may benefit from increasing their PA levels during pregnancy. 62,63 The aim of this study was to compare pedometer step counts with several accelerometer-derived measures of PA in overweight and obese pregnant women in free-living conditions.

Methods

Statistical analyses

All activity data were averaged over the valid days of recording. The majority of variables were not normally distributed and, therefore, non-parametric methods were employed for all analyses. Continuous variables were described using the median and interquartile range. Differences in background characteristics of included (n = 58) and excluded (n = 35) participants were tested using Mann–Whitney U-test for continuous variables and chi-squared test for categorised variables.

Agreement between the accelerometer and the pedometer was assessed in several ways. Absolute step count measurements were compared using the Wilcoxon signed-rank test. The relative agreement between pedometer-derived step counts and various accelerometer measures of PA was examined by determination of the Spearman rank correlation coefficient (r).

Total agreement between accelerometer-derived and pedometer-derived step counts was evaluated by determination of the 95% limits of agreement. The difference between both measures of step counts was plotted against the mean of both measures. As there was a statistically significant negative correlation between these variables, which was not resolved by transformation, the limits of agreement were estimated by a regression-based method. To test whether or not the limits of agreement varied by baseline BMI (25.0–29.9 kg/m² vs. ≥ 30.0 kg/m²) or gestational age (11–14 vs. ≥ 20 weeks’ gestation), interactions terms were added to regression models, and absolute residuals were compared by Student’s t-tests.

The classification of participants according to whether or not they recorded a daily mean of at least 8000 or 10,000 steps/day was compared between pedometer and accelerometer by calculating Cohen’s kappa over 2 × 2 contingency tables. Kappa was also determined to assess the agreement between those reaching 8000 pedometer steps/day and those achieving 30 minutes MVPA, as measured by accelerometer. Kappa values of 0.81–1.00 were regarded as indicating almost perfect agreement, while values of 0.61–0.80 indicated good agreement, 0.41–0.60 moderate agreement, 0.21–0.40 fair agreement and 0.0–0.20 slight agreement. 73

Confidence intervals (CIs) and p-values for r were estimated by bootstrapping over 5000 iterations. A p-value of < 0.05 was considered statistically significant. The majority of statistical analyses were performed using SPSS version 17.0 for Windows (SPSS Inc., Chicago, IL, USA); however, bootstrapping methods used Stata^® version 10.2 (StataCorp LP, College Station, TX, USA).

Results

Of the 93 women recruited, 32 (34%) had valid accelerometer data for fewer than three days, 17 (18%) had valid data for 3 days and 44 (47%) had valid data for 4 days. All valid days from women with valid data for at least 3 days (n = 61) were included in further analyses, excluding three women who did not have pedometer data. The final study sample consisted of 58 women (62% of those recruited). The excluded women (n = 35) were younger (median age 28 vs. 32 years; p = 0.018) and more often smokers during the previous year (46% vs. 13%; p = 0.002) than the included women, and fewer of them were highly educated (5% vs. 59%; p < 0.001), but gestational age, BMI, parity, ethnicity, marital status and employment status were similar to those of the included women. The background characteristics of the included women are described in Table 3. The median age was 32 years and the median BMI was 29.3 kg/m² (range 25.3–46.2 kg/m²).

Descriptive activity data

The median wear time of the accelerometer was 13 hours 40 minutes/day (Table 4). The women were sedentary for most of that time and total active time (median 4 hours 50 minutes/day) mainly comprised light-intensity activity. The median time spent in MVPA was 18 minutes/day.

TABLE 4 - Descriptive data on PA measures among participants with valid data for at least 3 days (n = 58)a

Minutes/day unless otherwise specified. The values represent the average of all valid days.

A normally distributed variable.

PA measure	Median (interquartile range)	Range
Accelerometer
Total included wear timeb	821.8 (754.0–869.3)	608.0–1111.0
Sedentary timeb	514.0 (464.3–583.1)	255.7–849.8
Total activity timeb	290.9 (245.8–340.1)	127.7–473.5
Light activityb	271.3 (218.8–315.4)	99.3–429.3
Moderate activity	18.0 (11.4–29.1)	5.3–70.0
Vigorous activity	0.0 (0.0–0.0)	0.0–4.3
Moderate or vigorous activity	18.0 (11.7–30.1)	5.3–70.0
Total counts/day	202,680 (166,951–248,348)	92,131–429,497
Average cpm	256.3 (209.5–323.7)	131.0–615.5
Steps, counts/day	5687 (4452–7086)	1545–11,453
Pedometer
Steps, counts/dayb	5961 (3727–8510)	267–12,833

Agreement between continuous pedometer and accelerometer measures of PA

There was no significant difference between the overall step counts recorded by the pedometer and the accelerometer (median 5961 vs. 5687 steps/day, respectively; p = 0.37; see Table 4). Pedometer step counts were significantly correlated with all accelerometer measures of PA except for sedentary time (Table 5). The correlation was good for accelerometer step counts (r = 0.78) and moderate (r = 0.36 to 0.54) for all other measures of PA. Pedometer step counts are plotted against accelerometer step counts in Figure 1. Despite these statistically significant correlations, the 95% limits of agreement were very broad, ranging between –2690 and 2656 steps/day for the mean value (mean of accelerometer and pedometer steps/day = 6026) (Figure 2). The limits of agreement also varied substantially over the range of values, indicating a differential bias. At the lowest recorded step count (mean of accelerometer and pedometer steps/day = 906), the limits were –927 to 4897 steps/day (a range of 5824), indicating that the accelerometer was on average recording more steps/day than the pedometer. In contrast, at the highest step count value (mean of accelerometer and pedometer steps/day = 12,018) the limits were –4753 to 33 steps/day (a range of 4786) indicating that, while the level of random disagreement had decreased, the direction of bias had reversed, with the accelerometer recording less steps/day than the pedometer on average. BMI and gestational age did not modify the limits of agreement as there were no statistically significant differences in the slope of the regression line (p = 0.28 for BMI; p = 0.68 for gestational age) or in the absolute spread from the regression line (p = 0.64 for BMI; p = 0.35 for gestational age).

TABLE 5 - Spearman correlation coefficients between pedometer step counts and accelerometer measures of PA (n = 58)

Variable measured	Correlation coefficient	95% CI	p-value
Sedentary time (minutes/day)	–0.30	0.51 to 0.05	0.023
Total activity time (minutes/day)	0.40	0.13 to 0.63	0.002
Light activity (minutes/day)	0.36	0.10 to 0.58	0.006
Moderate or vigorous (minutes/day)	0.47	0.18 to 0.69	< 0.001
Activity (minutes/day)	0.51	0.24 to 0.72	< 0.001
Total counts/day	0.54	0.28 to 0.74	< 0.001
Average	0.78	0.59 to 0.90	< 0.001

FIGURE 1.

Daily step counts as measured by accelerometers and pedometers. The figure also shows the line of equality and least-squares line.

FIGURE 2.

Differences in step counts/day between accelerometers and pedometers, plotted against the mean of both (n = 58). The limits of agreement were calculated using a regression method previously described by Bland and Altman. 74

Agreement between categorised pedometer and accelerometer measures of physical activity

Based on the accelerometer data, 15 (26%) of these women recorded ≥ 30 minutes MVPA/day, 12 (19%) recorded ≥ 8000 steps/day and three (5%) recorded ≥ 10,000 steps/day. The pedometer data showed that 18 (29%) of the women recorded ≥ 8000 pedometer steps/day and four (7%) recorded ≥ 10,000 steps/day.

There was moderate agreement between those achieving ≥ 8000 pedometer steps/day and those achieving ≥ 30 minutes MVPA/day (kappa 0.45, 95% CI 0.24 to 0.67) (Table 6). Agreement between the pedometer and the accelerometer in categorising women to < 8000 or ≥ 8000 steps/day was good (kappa 0.63, 95% CI 0.43 to 0.83). Very few women achieved ≥ 10,000 steps/day with either of the monitors (n = 3 for accelerometers and n = 4 for pedometers), thus the agreement between these was artificially high (results not shown).

TABLE 6 - Agreement between categorised pedometer and accelerometer measures of PA: kappa and 95% CI

15 (26%) of the participants recorded ≥ 30 minutes MVPA/day and 12 (19%) recorded ≥ 8000 steps/day.

18 (29%) of the participants recorded ≥ 8000 pedometer steps/day.

Accelerometer dataa	Kappa (accelerometer data vs. ≥ 8000 pedometer steps counts/day)b	95% CI
≥ 30 minutes MVPA/day	0.45	0.24 to 0.67
≥ 8000 steps/day	0.63	0.43 to 0.83

Discussion

Large-scale trials are needed to assess the potential impact of increased PA on reducing pregnancy complications and these trials should ideally use objective measurement methods to measure changes in PA. 40,41,75 Pedometers would be a cost-effective measurement tool for large studies, provided that the simple step count measure is broadly comparable to the more specific accelerometer data. Our results show that, although there was a significant correlation between pedometer step counts and most accelerometer measures of PA and no difference in median step counts between the two devices, the 95% limits of agreement were very broad, especially among those participants who were less active. In addition, the direction of difference between the monitors appeared to reverse across the range of activity levels, suggesting a complicated pattern of disagreement. Agreement between the monitors in categorising participants as active and inactive varied from fair to good depending on the criteria adopted, being good when achievement of ≥ 8000 steps/day was used as the criterion. Pedometer step counts have been compared with accelerometer data in a number of previous reports,56,58–61 all of which used one of the ActiGraph/CSA/Manufacturing Technology Inc. accelerometer models and one of the Yamax pedometer models, as in the present study. These accelerometer models are not entirely comparable to each other in measuring steps and activity counts. 76,77 The study by Harrison et al. 56 included 30 overweight or obese pregnant women at 26–28 weeks’ gestation in Australia. The participants wore an accelerometer (GT1M) and a pedometer for 5–7 days and the accelerometer data processing rules were very similar to those used in our study. Despite a statistically significant correlation (r = 0.69, p < 0.01) between the step counts of each monitor, the mean difference was 505 steps/day and the limits of agreement were large (from –2491 to 3501 steps/day).

The other studies were not conducted in pregnant women and, generally, included subjects who were more active than our participants. However, the findings were essentially similar to those in the present study. Tudor-Locke et al. ,58 in a study of 60 adult volunteers in South Carolina, USA, observed a high correlation between accelerometer (CSA model 7164) and pedometer step counts (r = 0.86), but the accelerometer detected 1845 ± 2116 more steps/day, on average, than the pedometer and the limits of agreement were even broader (–2387 to 6077 steps/day) than reported in the present study. In a larger study of older adults in the UK (n = 121),61 Harris et al. reported that pedometer step counts were highly correlated to the accelerometer (GT1M) step counts (r = 0.86) and the mean step counts were similar. However, the limits of agreement were again large, around –3500 to 35,0081 steps/day. Ramirez-Marrero et al. 60 reported similar findings among 58 adults infected with human immunodeficiency virus in Puerto Rico. Although the limits of agreement were not calculated in that study, individual variation in differences in step counts seemed to be large. When comparing pedometer step counts with other accelerometer (ActiGraph model 7164) measures of PA, the correlations observed in the present study are generally similar, although weaker than in the previous studies. 58,60,61 Macfarlane et al. 59 observed, among 57 adult volunteers in Hong Kong, that the means for accelerometer (Manufacturing Technology Inc. model 7164) measures of PA increased with increasing pedometer step counts, but the CIs were broad.

Among the previous studies, Tudor-Locke et al. 58 were the only authors to conclude that agreement between pedometer step counts and accelerometer measures of PA was unacceptably low, despite others also reporting broad limits of agreement56,61 or CIs. 59 Future studies should pay more attention to correct interpretation of Bland–Altman plots and limits of agreement.

The present study confirms the findings of these studies in a sample of 58 overweight and obese pregnant women. Although there are currently no methods available to calculate 95% CI for the limits of agreement determined by a regression-based method, it is important to note that a larger sample size would not have affected the size of the limits of agreement. There are also no guidelines for acceptable 95% limits of agreement for step counts.

We propose that they should be no larger than ± 500 steps/day (i.e. a range of 1000 steps/day), which is likely to correspond to a maximum of a 10-minute difference in the duration of MVPA, such as brisk walking,71 and may therefore be of clinical and public health importance.

The difference between the accelerometer and the pedometer step counts was correlated to the mean of both measures in the present study, but not in the other studies. 56,58,61 In this study, the difference between the step counts was in the opposite direction for less active and more active women, that is the accelerometer detecting more steps among less active women and the pedometer detecting more steps among more active women. This discrepancy may be related to the general limitations of the monitors or differences in their sensitivity to detect PA. Pedometer accuracy is reported to be diminished at slow walking speeds, especially below 3 miles/hour,69,78 and both active and inactive participants undertook many episodes of low-intensity activity in the present study. This may also be the case with some accelerometers, although the GT1M model used in our study has been shown to have lower inter-monitor variability and lower sensitivity for low-intensity activity than the previous 7164 model. 76,77 On the other hand, the previous CSA model has also been reported to erroneously detect slightly more non-steps, for example when travelling by a motor vehicle. 79

The accuracy of the latest ActiGraph accelerometer model (GT3X, ActiGraph, Pensacola, FL, USA) and the Yamax Digi-Walker SW200 (Yamax, Bridgnorth, UK) was recently investigated in 30 pregnant women. 57 Both monitors underestimated the number of steps, especially at slow walking speeds, and even when the monitors were repositioned at a tilt angle, there was no correlation with the percentage of actual steps detected by either monitor. In contrast, Crouter et al. 69 suggested that the tilt angle reduced the accuracy of spring-levered pedometers in overweight or obese adults. The tilt angle may also reduce the accuracy of accelerometers in assessing vertical movement, which may happen more often among overweight and obese than normal weight people. 55 The tilt angle was not directly measured in the present study. However, BMI and gestational age did not significantly modify the results of the Bland–Altman plot, suggesting that the potential effect of the tilt angle on the results may have been the same regardless of BMI or gestational age.

We also assessed agreement between the monitors in categorising participants as active or inactive. Although agreement was relatively good (kappa 0.63) when using 8000 steps/day as the criterion for both monitors, agreement was lower (kappa 0.45) when comparing participants achieving 8000 pedometer steps/day with those achieving 30 minutes of MVPA/day measured by a accelerometer. Of the previous studies, Ramirez-Marrero et al. 60 reported fair agreement between 10,000 pedometer steps/day and 150 minutes of MVPA/week (kappa 0.25, p = 0.01). These discrepancies between pedometers and accelerometers in categorising participants into active and inactive may partly be because of the data processing rules, such as selection of the epoch length and intensity cut-off points to define MVPA for the accelerometer data.

Accelerometry should not be regarded as a gold standard to measure free-living PA nor necessarily as a more accurate method of measuring daily steps than with a pedometer. Although the previous version of the ActiGraph accelerometer was the only commercially available accelerometer66 that correlated reasonably with double-labelled water, most validation studies have been conducted in controlled environments. Validity is lower when applied to free-living settings. 55 Two armband accelerometers have also recently been shown to be highly correlated with double-labelled water in free-living conditions. 42 Both the accelerometer and the pedometer measure biomechanical body movement. Hence, validity of the monitors against energy expenditure should not be a major concern when assessing agreement between these devices.

This study had some limitations. First, although the participants were asked to record the times when they wore the monitors, we cannot be sure that both monitors were worn exactly for the same time. Some women had very low pedometer step counts but moderate accelerometer step counts, suggesting that the wearing time may have been different for each monitor or the pedometer may have been in a tilt angle. Therefore, studies in controlled conditions, such as that by Connolly et al. ,78 would be necessary to be certain that monitors were worn for exactly the same time. Second, almost all of our participants were in the first or second trimester of pregnancy and, therefore, we do not know whether or not the results can be generalised to the third trimester, when activity decreases and the abdominal circumference is much larger. On the other hand, these results may be generalised to non-pregnant overweight and obese women of similar age.

Third, participation was low (33%) and 34% of the participants were excluded because valid accelerometer data were available for fewer than 3 days. The activity levels of the participants were similar or slightly lower than those reported in pregnant women in other comparable studies. 42,44,54,80 The purpose of this study, however, was to compare methods of measurement, rather than to obtain representative estimates of PA levels in pregnancy.

Conclusions

Comparing median step counts or assessing correlation coefficient overestimates agreement between pedometer and accelerometer data. Examination of the 95% limits of agreement revealed a substantial lack of agreement between step counts measured by the two types of monitor. Pedometer step counts were not comparable to accelerometer data at an individual level in overweight and obese pregnant women. The choice of measurement method may depend on the target of the intervention. For example, accelerometers may be better at assessing changes in PA in trials that promote increases in moderate or vigorous PA or reduction in sedentary time, whereas spring-levered pedometers may be more appropriate for studies evaluating walking interventions in more active populations.

Dietary assessment and analysis

A decision was made to calculate the dietary GI in obese pregnant women using the commercially available WISP version 3 (Tinuviel software, Llanfechell, Anglesey, UK) dietary analysis software. In order to assess diet most accurately, it was decided that dietary data in the pilot study would be obtained from a triple-pass 24-hour dietary recall at three time points in duplicate, performed at baseline and post randomisation [28/29⁽⁺¹⁾ weeks’ gestation and 36/37⁽⁺¹⁾ weeks’ gestation] (for details see Chapter 4). This dietary assessment methodology was undertaken in preference to a 4-day food diary, owing to recall bias and data quality issues reported with this method in similar population groups. A shortened food frequency questionnaire (FFQ) was also employed for assessment of long-term dietary habits, with particular reference to the dietary GI and for validation in the pilot study (for details see Chapter 4).

Control arm

Following randomisation, women in the control arm (standard care) returned for data collection appointments with the study midwife at 27⁺⁰–28⁺⁶ and 34⁺⁰–36⁺⁶ weeks’ gestation, when possible coinciding with routine antenatal visits.

Intervention arm

Following randomisation, participants attended a one-to-one appointment with the HT and were invited to weekly group sessions for 8 consecutive weeks from approximately 19 weeks’ gestation. All women attended routine antenatal care appointments and received advice regarding diet and PA in accordance with local policies, which draw on the UK’s National Institute for Health and Care Excellence (NICE)’s guidelines. 93

Sample size

The primary outcome was change in dietary and PA behaviours at 28 weeks’ gestation (coinciding with the primary maternal outcome for the main RCT, GDM at 28 weeks’ gestation). No prior investigation in obese pregnant women was available to inform power at the planning stage. The sample size of 183 was determined by the predefined duration of phase 2, the exploratory phase. This number was adequate to enable power calculations for primary end points of the subsequent RCT, by providing estimates of the variance to within approximately 7% of the true value.

Ethics

Research Ethics Committee approval was obtained in all participating centres, UK IRAS reference number 09/H0802/5 (South East London Research Ethics Committee).

The intervention

The intervention was informed by psychological models of health behaviour including control theory94 and social cognitive theory. 95 Although no clear patterns between intervention characteristics and outcomes have been seen to date in lifestyle interventions in pregnancy, and few studies have described their theoretical basis,2,85 self-regulation techniques, drawn from control theory, suggest that behaviour change is facilitated by feedback about performance compared with prespecified goals. 94,96 This approach was utilised in this study by setting specific, measurable, achievable, relevant and time-specific (SMART) diet and activity goals, with behaviours recorded in a logbook. Identification of benefits of, and overcoming barriers to, behaviour change, and increasing self-efficacy were also included, and social support was facilitated through the group format. 95 As identified in social cognitive theory, the intervention aimed to build self-efficacy through mastery experiences (e.g. maximising the chance of success with SMART goals), vicarious experiences (encouraged through modelling in the group setting), and social persuasion (e.g. through enlisting social support). Following initial feedback from HTs regarding difficulties encountered by some women in attending sessions, for those women unable to attend, the session content was delivered by telephone or e-mail.

Dietary advice

Prespecified dietary outcomes were a change in the GI, GL [an indicator of carbohydrate quality (GI) and quantity consumed] and energy intake from saturated fatty acids (SFAs). The focus of the dietary advice to the intervention group was therefore on increased consumption of foods with a low dietary GI, including replacement of sugar-sweetened beverages with low-GI alternatives. Reduction in saturated fats and replacement with monounsaturated and polyunsaturated fat was also recommended. Exchange of foods was emphasised, for example a high-GI food for a low-GI food, rather than limiting energy intake.

Physical activity advice

Women in the intervention arm were encouraged to increase daily PA incrementally, setting goals of incremental step counts (monitored by pedometer) and maintaining the achieved PA level after the intervention period. Recommendations included an emphasis on walking at a moderate-intensity level. 39

Intervention delivery

The intervention was delivered by HTs. In the UK, HTs do not have prespecified health professional qualifications, but relevant experience [http://informationstrategy.dh.gov.uk/health-trainer-workforce (last accessed 10 March 2013), information now available at www.healthcareers.nhs.uk/explore-roles/public-health/health-trainer (last accessed 20 March 2017)]. All HTs received a comprehensive treatment manual, pre-study training (and within-study supervision) in behaviour modification and conducting group sessions (organised by Weight Concern; registered charity number 1059686). The sessions were held in a hospital setting in all but one centre, in which women attended a community children’s centre. At the initial one-to-one appointment women were provided with a participant handbook, reflecting the rationale and content of the HT sessions, a pedometer (Yamax SW200 Digi-Walker), a logbook for weekly SMART goals and related behaviours (steps, PA and diet) and a DVD of a specially devised pregnancy exercise regime. Potential benefits of attending group sessions were discussed. Each group session delivered a different element of the dietary and PA intervention. (Table 7). Goals from the previous week were reviewed and goals set for the following week. Discussion included barriers to behavioural change and ways these might be overcome.

The following information was obtained from all participants (at visits indicated in Figure 3).

Attitudinal assessment questionnaire

The attitudinal assessment included questions relating to perceived benefits and barriers and confidence to carry out the dietary and PA behaviours. 27,97 The target behaviours were to consume lower-GI carbohydrates, to reduce saturated fat intake and to increase PA.

Health status and mental health

The EQ-5D98 was used to assess health status and the Edinburgh Postnatal Depression Scale (EPDS) to assess mental health. 99

Dietary assessment

Repeated, triple-pass 24-hour recall data obtained at baseline (randomisation) and at 28 weeks’ gestation were evaluated twice, 1 week apart, in both the intervention and control groups. The 24-hour dietary recall is a standard retrospective, interviewer-led dietary assessment methodology used to capture information on all food and drinks consumed in the preceding 24 hours. This is carried out in three stages (the triple pass): (1) recording a ‘quick’ list of foods eaten or drunk, (2) collecting more detailed information of these foods and (3) reviewing all items once more in order to clarify any ambiguities or omissions. A short FFQ, for later validation, was also completed.

Physical activity assessment

At the first and third appointments participants were asked to wear an ActiGraph accelerometer (either GT1M or GT3X set to uniaxial mode) for 7 consecutive days, removing it for washing, bathing, swimming and at night. PA was also assessed by questionnaire [via the Recent Physical Activity Questionnaire (RPAQ)].

Process evaluation

A process evaluation, following the framework of Steckler and Linnan,100 was undertaken. This explored (1) context (environmental, socioeconomic or political factors), (2) reach (the proportion of the intended target audience that participates and which subgroups, if any, do not participate), (3) dose delivered and dose received (the proportion of intended intervention received), (4) fidelity (if each component of the complex intervention was provided as intended) and (5) acceptability (if the intervention materials and advice were well-received by providers and participants).

Qualitative semistructured interviews were conducted to capture women’s experiences and perceptions of the trial and intervention. Women were recruited from each of the participating study sites using a maximum diversity sampling approach, following an informed consent procedure. Interviews took place between November 2010 and February 2011, and were either face to face (n = 17), mostly in hospital settings, or by telephone (n = 4). Control (n = 12) and intervention (n = 9) interviewees were asked about their involvement in the research and their experiences of the trial appointments, measurements, blood tests and accelerometry recordings. Women in the intervention arm were additionally asked about their perceptions of the different components of the intervention, and how these impacted upon their lives. The interviews were conducted by one researcher and took place during pregnancy after the intervention had been provided. In addition, HTs completed audio diaries (130 recordings) in which they reflected on the fidelity and feasibility of the intervention delivery. Attendance at sessions was recorded on the study database.

Clinical outcome data

Outcome data also recorded (not reported)

These included maternal outcomes: diagnosis of GDM and pre-eclampsia, mode of delivery, blood loss at delivery, inpatient nights, detailed clinical and family history, health in current pregnancy, early pregnancy data (ultrasound scan, nuchal screening), blood pressure, routine blood results; neonatal outcomes: gestational age at delivery, birthweight, anthropometry and inpatient nights. Maternal urine and cord blood samples were also provided.

Health quality and attitudinal assessment questionnaires

The generic EQ-5D health-related quality-of-life instrument98 is reported as the proportion of women with problems on individual dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). It is given as a summary index score calculated from preference values of different combinations of the dimensions elicited using the time trade-off method in a sample representative of non-institutionalised adults in England, Scotland and Wales (range −0.59 to 100, where −0.59 is severe problems on all dimensions);101,102 and also the visual analogue scale of health-related quality of life (range 0 to 100, where 0 is worst imaginable health state). The change between baseline and 28 weeks’ gestation in the percentage of women with any problem was assessed using McNemar’s test of changes. Attitudes to target behaviours (attitudinal assessment questionnaire) are based on the average of multiple responses on five-point scales (three responses for diet, 13 for PA) with five indicating the greatest perceived barrier, perceived benefit or level of confidence.

Assessment of deprivation

These scales for estimation of deprivation in England and Scotland91,92 use different reference populations to determine the actual indices of deprivation, and are therefore not directly comparable. For the purposes of this study, the most deprived quintile is presented separately for women in each population and compared with the remainder of the population (quintiles 1–4).

Dietary analysis

The quality of dietary data was checked within 1 week of entry. Dietary coding utilised McCance and Widdowson’s Composition of Foods (6th edition)103 food codes and nutrient composition was evaluated using WISP version 3 for GI and GL values. 104 Estimates using previously published methodology were made when GI values were not available. 105 Twenty-four-hour recall data obtained at baseline (randomisation) and at 28 weeks’ gestation were evaluated twice, 1 week apart, and then averaged. The validity of the short FFQ was assessed against the dietary recall data. Prespecified dietary outcomes were a change in GI, GL and energy intake from SFA. Total energy intake and the proportion of energy derived from macronutrients were assessed. GWG at 28 and 36 weeks’ gestation, a relevant secondary outcome of interest, is also reported.

ActiGraph analysis

An epoch length (time sampling interval) of 15 seconds was specified. Data were processed using the MAHuffe software package. Sedentary behaviour was defined as < 100 cpm, light activity as 100–1951 cpm, moderate-intensity activity as 1952–5725 cpm and vigorous activity as > 5725 cpm. 68 As time spent in vigorous activity was very low, minutes of moderate and vigorous physical activity (MVPA) were combined. Runs of zero counts lasting > 60 minutes were excluded, as these indicated monitor removal. A valid recording was defined as a day in which > 500 minutes of monitored on-time was recorded in 24 hours. 106 Data from participants recording ≥ 3 days of valid accelerometry data on 3 or more days were included in the analysis. The specified PA outcome was an increase in minutes per day of MVPA recorded by accelerometry.

Recent Physical Activity Questionnaire

The RPAQ was modified for the assessment of PA in the preceding 7 days. Estimates of minutes per day spent in light, moderate and vigorous activity in each of the domains were calculated. Sedentary activities were defined as those with a metabolic equivalent (MET) of < 1.5. Light activities were those of 1.5–3 METs. Moderate activities were those of 3–6 METs. Vigorous activities were those of 6 METs or greater. 107 MVPAs were combined to give one summary variable.

Process evaluation

Interviews were recorded and transcribed verbatim. Transcripts were anonymised and a unique identifier (ID) number was used to maintain confidentiality. Data were imported into a qualitative software analysis package (NVivo version 8, QSR International, Warrington, UK) and subject to comparative thematic analysis. 108 To enhance study validity and reliability, themes arising from the research were discussed, the data supporting these were reviewed by co-researchers, and data were compared between sites and with existing literature. By these methods, assumptions were tested and observations of differences and their relationship to the theoretical models underpinning the study were explored.

Statistical analysis

Analyses followed the intention-to-treat principle. Following Consolidated Standards of Reporting Trials (CONSORT) guidelines, risk ratios and risk differences were estimated by binary regression for yes/no outcomes. When measures were repeated at baseline and 28 weeks’ gestation, results of mean [standard deviation (SD)] or n (%) are presented separately at each time point. Randomised comparisons at 28 weeks were made using linear regression with robust standard errors, adjusting for the baseline value. For PA data, dummy variables were used when the baseline values were missing. Correlations between PA as assessed objectively (accelerometry) and when self-reported (RPAQ) were explored.

Participants

The mean first visit BMI was 36.3 kg/m². More than half the women were white and the remainder were from black (38%) and minority ethnic communities. More than half (56%) already had at least one child. More than half of those from centres in England and over 40% in Scotland came from regions in the highest quintile of social deprivation (Table 8).

Diet and gestational weight gain

Table 9 shows the dietary intakes at baseline and at 28 weeks’ gestation. There were no differences between groups in energy intakes, GI, GL or other macronutrient at baseline. However, following the intervention, at 28 weeks’ gestation, total energy intake, dietary GL, GL (% energy), saturated fat (% energy) and total fat (% energy) were significantly lower, and fibre intake measured as non-starch polysaccharides was greater, in the intervention group than in the control arm. The proportion of energy derived from protein was higher in the intervention group, but absolute protein intake did not differ. There was a difference of 7 GI points between the intervention and control group, which achieved borderline statistical significance (p = 0.054).

There was a marginal difference in GWG of –0.9 kg between the intervention and control group at 28 weeks’ gestation (p = 0.065).

Physical activity

There were no differences between the intervention and the control arms in objectively measured PA variables at baseline or at 28 weeks’ gestation, after adjustment for baseline activity. Self-reported moderate to vigorous PA at 28 weeks’ gestation was increased in the intervention group (mean difference 34 minutes/day; 95% CI 9 to 59 minutes/day), but this was not supported by the objective data. Women in the intervention group self-reported walking for leisure for 14 minutes/day more than those in the control group at 28 weeks’ gestation (95% CI 5 to 23 minutes; p = 0.003). Agreement between the RPAQ questionnaire and accelerometry was very poor; for example, correlation between MVPA in the two formats at baseline was r = 0.275 (95% CI 0.107 to 0.428) and at 28 weeks’ gestation was r = −0.069 (95% CI –0.296 to 0.165) (Table 10).

Attitudinal assessment of target behaviours

Benefits, barriers and confidence in making the target PA and dietary changes were unchanged in either the control and intervention groups from baseline to 28 weeks’ gestation (Table 11).

Health Status, Mental Health Edinburgh Postnatal Depression Scale and EuroQol-5 Dimensions

There was no influence of the intervention on the numbers of women reporting problems in each of the EQ-5D domains, but, as a group, obese women experienced a significant increase in problems with mobility, self-care, usual activities and pain and discomfort from baseline to 28 weeks’ gestation. There was a 10% prevalence of probable depression at baseline and a 13% prevalence at 28 weeks (i.e. EPDS score of > 12), with no significant effect of the intervention on anxiety and depression at 28 weeks’ gestation (see Table 11).

Process evaluation

Maternal and neonatal outcomes

The primary maternal and neonatal outcomes for the subsequent RCT are shown in Table 13. There were no significant differences in GDM or LGA (≥ 90th customised centile) between control and intervention arms. There was also no significant difference in GWG between control and intervention arms (secondary outcome). The overall incidence of GDM, the primary outcome of the subsequent RCT (not powered for), in accordance with recent IADPSG criteria,6 was 30%, enabling calculation of the subsequent RCT sample size (1546 women) for the RCT, powered for a 25% reduction. As 38% of potentially eligible women took part in the pilot study, to achieve this sample size in the main RCT, approximately 4100 would need to be approached.

Implications for proceeding to phase 3, the randomised controlled trial

As detailed in the discussion and conclusion section of the above report, there were many benefits gained from undertaking the pilot trial which led to us to change the delivery of the intervention, but not any of the principal elements of the intervention. Feasibility and general acceptability were established and, most importantly, the intervention led to changed dietary behaviours. PA behaviours that, as we found in the process evaluation, were less amenable to the women proved difficult to change, but the intervention was continued as PA is recommended nationally for all pregnant women. The study also enabled us to better define the power calculations for the main trial, based on the incidence of GDM in the women in the pilot study (see Chapter 4). The process evaluation provided some similar outcomes as structured interviews in phase 1, but several different outcomes, probably because the women in the pilot were more representative of the population to be recruited for the intervention. Importantly, it was recognised that recruitment rates would be lower than expected when the original application for funding was submitted, and that full recruitment to the main UPBEAT intervention would not be feasible within the time frame allotted. Following the pilot trial, an application for an extension to the trial funding was therefore submitted to the National Institute for Health Research (NIHR), and additional funding was provided to allow completion of recruitment.

Abstract

Aim: The aim was to examine the prediction of GDM in obese women using routine clinical measures and measurement of biomarkers related to insulin resistance in the early second trimester.

Methods: A total of 117 obese pregnant women participating in a pilot trial of a complex intervention of dietary advice and PA were studied. Blood samples were obtained at recruitment (15⁺⁰–17⁺⁶ weeks’ gestation) and demographic data, clinical history and anthropometric measures recorded. The biomarkers analysed were plasma lipids [high-density lipoprotein (HDL) cholesterol, low-density lipoprotein (LDL) cholesterol, triglycerides], high-sensitivity C-reactive protein, alanine transaminase, aspartate transaminase, ferritin, fructosamine, insulin, adiponectin, tissue plasminogen activator, interleukin 6, visfatin and leptin. Univariate and logistic regression analyses were performed to determine independent predictors and the area under the receiver operating curve was calculated for the model.

Results: Of the 106 participants included in the analysis, 29 (27.4%) developed GDM. Participants with GDM were older (p = 0.002), more often of a parity ≥ 2, had higher systolic (p = 0.02) and diastolic blood pressure (p = 0.02), and were more likely to be black (p = 0.009). Among the blood biomarkers measured, plasma adiponectin alone remained independently associated with GDM in adjusted models (p = 0.002). The area under the receiver operating curve for clinical factors alone (0.760) increased significantly [area under the curve = 0.834, χ²(1) = 4.00; p = 0.046] with the addition of adiponectin.

Conclusions: A combination of routinely measured clinical factors and adiponectin measured in the early second trimester in obese pregnant women may provide a useful approach to the prediction of GDM. Validation in a large prospective study is required to determine the usefulness of this algorithm in clinical practice (clinical trial registry number ISRCTN89971375).

Introduction

The prevalence of obesity in adults and children continues to rise. Obesity remains the sixth most important determinant of adverse health and reduced adult life expectancy globally. 139 In the UK, the incidence of obesity in women of reproductive age has almost doubled in the past 20 years;140 the most recent World Health Organization (WHO) Global InfoBase of obesity (BMI of ≥ 30 kg/m²) in UK women aged > 15 years (2010) reports an age-adjusted prevalence of obesity of 26.3% across all ethnic groups. 38

Maternal obesity carries a significant risk of adverse pregnancy outcomes, particularly GDM. Short- and long-term metabolic complications follow a continuous linear relationship with BMI,141 with the risk of developing GDM rising from two- to eightfold with increasing BMI category. 142 Not all obese women develop GDM, but this heterogeneity poses a burden on limited resources, with all women with a BMI of ≥ 30 kg/m² currently managed as if at risk, often resulting in suboptimal management. Accurate and early identification of obese pregnant women who will subsequently develop GDM would enable early risk stratification, more appropriate use of health-care resources and the targeting of intervention strategies.

Currently, NICE, in the UK, recommends selected rather than universal GDM screening, according to risk factors that include obesity. Women who have previously delivered a macrosomic infant, who have had previous GDM or who have a first-degree relative with diabetes mellitus and high-risk ethnicity are also screened. A systematic review of screening for GDM undertaken for a health technology assessment reported low sensitivities (50–69%) and specificities (58–68%; eight studies) when traditional methods of risk factor screening were used. 143 Although there is at present no accepted early pregnancy intervention to improve clinical outcome in obese pregnant women,84 increased recognition of the problem has led to an international research effort to develop effective interventions. Several large-scale RCTs, including UPBEAT (registered as ISRCTN89971375), are investigating targeted dietary and PA interventions or metformin to improve glucose homeostasis and pregnancy outcome in overweight and obese women. 15,144

Research into the prediction of adverse outcomes in other pregnancy-related conditions, such as pre-eclampsia, has shown that a combination of clinical history and early pregnancy clinical measures, together with the addition of biomarkers measured in biological samples, may provide an effective strategy in early pregnancy risk assessment. 145 Several studies have adopted this approach to the prediction of GDM,146,147 but, to our knowledge, not previously in a population of obese women.

The aim of the present study was to undertake a preliminary investigation in obese pregnant women to determine whether or not the addition of biomarkers to routine clinical measurements further improves the prediction of GDM. For this purpose, we studied 117 women participating in a pilot trial for UPBEAT, and measured 16 biomarkers frequently implicated in the pathogenesis and prediction of GDM and/or type 2 diabetes mellitus and reflecting inflammatory pathways, markers of adipose tissue function, hepatic fat accumulation and vascular dysfunction. 146,148,149 As abnormal fatty acids (FAs) are associated with insulin resistance and fetal macrosomia and because significant differences were found in the dietary intake of SFAs and the ratio of polyunsaturated fatty acid to SFA, a detailed analysis of FAs was also undertaken. 150 Prediction models were then developed incorporating clinical and biomarker data.

Methods

UPBEAT is a multicentre RCT of a complex dietary and PA intervention aimed at improving glucose homeostasis in obese pregnant women. A pilot trial was undertaken in 183 women in four UK hospitals to evaluate changes in dietary and PA behaviours, trial all aspects of the protocol and undertake process evaluation. 150 Details of the intervention and protocol are available on the trial website (www.medscinet.net/upbeat/about.aspx).

NHS Research Ethics Committee approval was obtained by all contributing centres (UK IRAS reference number 09/H0802/5).

At recruitment (15⁺⁰–17⁺⁶ weeks’ gestation) and after informed consent had been obtained, information was collected on demographics, maternal history, maternal family and current pregnancy health. Randomisation to the intervention arm or the control arm (which consisted of standard antenatal care) was carried out at the second appointment, approximately 1 week later (7–10 days), between 16⁺⁰ and 18⁺⁶ weeks’ gestation, by a secure internet-based data management system (MedSciNet Ltd, www.medscinet.net/upbeat). The randomisation schedule was minimised according to ethnicity, parity (0 vs. ≥ 1), age and BMI (30–34.9 kg/m² vs. 35–39.9 kg/m² and > 40 kg/m²). Blood pressure was recorded using the Micro-life^® BP3BT0-A automated blood pressure monitor (Micro-life, Widnau, Switzerland), which is validated for use in pregnancy. Maternal skinfold thickness (triceps, biceps, subscapular and suprailiac) was measured in triplicate with Harpenden skinfold callipers (validated for values ≤ 80 mm; Holtain Ltd, Felin-y-Gigfran, Crosswell, UK), in addition to the following circumferences: waist, mid-arm, thigh and hip. Skinfold thicknesses at four sites (triceps, biceps, suprailiac and subscapular) were summed. Blood samples were obtained from 117 participants in the three centres that had facilities for sample handling and storage. Serum and plasma was stored at –80 °C for future analysis.

At 28 weeks’ gestation, an OGTT was performed on all participants. Diagnosis of GDM following a 75-g 2-hour OGTT at 27⁺⁰–28⁺⁶ weeks was defined in accordance with the IADPSG’s criteria (fasting blood glucose concentration of ≥ 5.1 mmol/l or 1-hour glucose of ≥ 10.0 mmol/l or 2-hour glucose of ≥ 8.5 mmol/l). 6 If a diagnosis of GDM was made, women were referred for routine GDM care in accordance with local criteria.

Biochemical analyses

Plasma total cholesterol, HDL cholesterol triglycerides, alanine transaminase, aspartate transaminase, high-sensitivity C-reactive protein, fructosamine (c311, Roche Diagnostics, Burgess Hill, UK) and ferritin (Elecsys 2010, Roche Diagnostics, Burgess Hill, UK) were measured on clinically validated automated platforms using the manufacturers’ quality controls (QCs) and calibration materials. The coefficient of variation (CV) was < 6%. Plasma insulin was measured with an enzyme-linked immunosorbent assay (Mercodia, Uppsala, Sweden) that does not cross-react with proinsulin and the interassay’s CV was < 7%. Baseline plasma adiponectin, interleukin 6, leptin (R&D Systems, Abingdon, UK), tissue plasminogen activator (Stago, Theale, UK) and visfatin (Phoenix Peptide, Karlsruhe, Germany) were measured by enzyme-linked immunosorbent assay. These methods had an interassay CV of < 10%.

Plasma non-esterified fatty acids (NEFAs) and FAs were measured on samples obtained after fasting for 12 hours at 27⁺⁰–28⁺⁶ weeks’ gestation in the Division of Nutritional Sciences, King’s College London, London, UK.

Non-esterified fatty acids were measured on a clinically validated automated platform (Clinical Analyser ILab 650, Instrumentation Laboratories, Warrington, UK) using the Randox (FA115, Randox Laboratories, Montpellier, France) kit. QC was performed after each 60-sample batch at the upper and lower range of the assay, with the following CVs: QC1 target of 1.24 mmol/l (%CV = 0.95) and QC2 target of 0.58 mmol/l (%CV = 0.97). Plasma FAs were measured by gas–liquid chromatography. 151 Esterified and non-esterified fatty acid methyl esters were analysed using the one-step transesterification direct method. 152 Main FA peaks [C16:0, C18:0, C18:1(n-9), C18:2(n-6), C18:3(n-3), and C20:4(n-6)] were recognised by referring to standard retention times (Sigma-Aldrich Co. Ltd, Gillingham, UK). C20:5(n-3) and C22:6(n-3) were determined by cod liver oil fatty acid methyl esters standards (Sigma-Aldrich Co. Ltd, Gillingham, UK). The remaining FAs were measured by gas–liquid chromatography mass spectrophotometry. Each plasma FA concentration was determined as the area under the peak matched with the known standard. 153

All analyses were performed on previously unthawed ethylenediaminetetraacetic acid and serum samples. Samples were processed by technicians blinded to the identity of the samples.

Statistical methods

The analysis was exploratory with the aim of identifying potentially useful combinations of clinical and biochemical predictors154 of maternal GDM; therefore, potentially useful biomarkers were not excluded and no adjustment was made for multiple testing. Standard distributional checks (Box–Cox regression and normal distribution plots) were carried out, and separate decisions made on the appropriate transformation. Based on these findings, log-transformation was carried out for all biochemical variables. Differences between patient groups are reported as geometric means and ratios of geometric means, with 95% CIs.

The association of clinical indicators with GDM was established using linear or logistic regression as appropriate, with robust standard errors. Biochemical indicators were assessed as predictors of GDM, adjusting for significant clinical indicators. After univariate analysis, those variables which were identified as independent predictors of GDM were included in the model. The overall performance of the markers as predictors of GDM was assessed by comparison of receiver operating curve areas. Areas under receiver operator characteristic curves are compared using a non-parametric approach suggested by DeLong et al. 154 and implemented in Stata as the commands roccomp and rocgold. When necessary, composite predictors were derived using multiple logistic regression.

All data analysis was carried out using the statistical package Stata version 11.2.

Results

A total of 11 women were omitted from the analysis because of inadequate OGTT data. Of the remaining 106 women (53 in the control group and 53 in the intervention group), 29 were diagnosed with GDM (27.4%). The demographic and clinical characteristics of women who developed GDM compared with those who did not are summarised in Table 14. In general, women with GDM were older, more often of higher parity (≥ 2), had higher systolic and diastolic blood pressure and were more likely to be black than those without. BMI was not significantly different between the two groups, although skinfold thicknesses were greater in women who developed GDM; women who developed GDM had greater triceps thickness (37 vs. 31 mm; p = 0.004) and total sum of skinfold thicknesses (93 vs. 86 mm; p = 0.03). There was no evidence of interaction in terms of prediction of GDM by treatment group (p = 0.85).

Table 15 shows the first trimester biomarkers for women who subsequently developed GDM and those who did not. Women with GDM had 34% lower plasma concentrations of adiponectin (95% CI –47% to –19%), adjusting for clinical predictors of age, parity ≥ 2, diastolic blood pressure and systolic blood pressure. There was a trend towards significance for fructosamine in the GDM group (p = 0.05), which attenuated to the null after adjustment (p = 0.82) (see Table 15).

In a combined logistic regression model, including the biomarkers and clinical risk factors, the only consistent predictive variables were adiponectin (OR for a halving in adiponectin concentration 4.04, 95% CI 1.69 to 9.64; p = 0.002) and maternal age (OR per additional year 1.18, 95% CI 1.04 to 1.34; p = 0.01; Table 16). An area under the receiver operating curve of 0.760 (95% CI 0.645 to 0.875) for prediction of GDM was achieved with clinical predictors (age, parity, ethnicity and blood pressure) alone. The area under the receiver operating curve increased significantly to 0.834 [95% CI 0.742 to 0.927, χ²(1) = 4.00; p = 0.046] with the addition of adiponectin (Figure 5).

FIGURE 5.

Receiver operating curve and summaries using the basic model (including age, parity, ethnicity, blood pressure), with the addition of adiponectin. AUC, area under the receiver operating curve.

Discussion

The present study highlights novel biochemical and clinical factors for the prediction of GDM in obese pregnant women and suggests that an algorithm based on simple clinical variables plus adiponectin concentration may provide a clinically useful method for the prediction of GDM in this population.

Four previous studies have identified a number of patient characteristics and biomarkers associated with the prediction of GDM. 146,149,155,156 These have been undertaken in populations of mixed risk, including non-white ethnicity,146,155,156 family history of diabetes mellitus,146,149,155,156 previous history of GDM,146,155,156 high pre-pregnancy BMI,146,149,156 older maternal age146,155,156 and differing parity. 149 Savvidou et al. 149 measured nine biomarkers in the first trimester and found that high concentrations of tissue plasminogen activator and low concentrations of HDLs increased the area under the receiver operating curve from 0.824 with clinical risk factors alone to 0.861 in a group of all comers, regardless of baseline BMI. The addition of adiponectin to prediction models for GDM has consistently increased the area under the receiver operating curve to values above those achieved with clinical measures alone. Further inclusion of adipokines and biomarkers has frequently demonstrated a modest, non-significant, increase in the area under the receiver operating curve. For example, in a case–control study of 400 women, those with GDM were reported to have higher levels of maternal serum visfatin and lower serum adiponectin concentrations at 11–13 weeks of gestation. The addition of adiponectin to the prediction model using clinical measures alone resulted in a significant change in the area under the receiver operating curve, whereas there was a non-significant increase after addition of visfatin [an area under the receiver operating curve of 0.828 (maternal characteristics alone), 0.854 (adiponectin) and 0.855 (adiponectin and visfatin)]. 146 Nanda et al. 156 measured three biomarkers and found that in the GDM group, compared with controls, adiponectin and sex hormone-binding globulin levels were lower. When screening for GDM by maternal characteristics alone, the detection rate was 61.6% (false-positive rate 20%), increasing to 74.1% with the addition of adiponectin and sex hormone-binding globulin. Alternative approaches to GDM risk assessment have included the measurement of biomarkers in the preconception period, with a recent report finding that maternal characteristics, fasting plasma glucose, glycosuria and preconception dyslipidaemia yielded an area under the receiver operating curve of 0.90 for the prediction of GDM;155 however, the varied diagnostic criteria for GDM used in previous studies limit comparisons with previous attempts to predict GDM. Importantly, no study has specifically addressed risk assessment in obese pregnant women, which has important implications for clinical practice given the recognition of obesity as the major risk factor for GDM, and the likelihood that the biomarker profile may be dissimilar to other risk groups in women with a high BMI.

The present results suggest that clinically useful prediction of GDM in obese pregnant women is achievable using a combination of clinical characteristics (older age, higher systolic and diastolic blood pressure, parity ≥ 2 and black ethnicity) combined with the plasma concentration of adiponectin. To reflect current clinical practice, routine clinical measurements recorded at antenatal visits were included. The inclusion of detailed maternal anthropometry (including skinfold thicknesses), which is undertaken in all women participating in UPBEAT, suggested a limited potential role for taking such measurements routinely as an aid to GDM prediction (see Table 18).

Adiponectin, an adipocyte-derived adipokine, is now recognised as being strongly associated with improved glucose metabolism and increasing insulin sensitivity, although the causality of this relationship remains debated. Irrespective of causal direction, adiponectin appears to provide a good ‘read-out’ of whole-body insulin sensitivity. In a recent meta-analysis of non-pregnant individuals, adiponectin was shown to be strongly predictive of type 2 diabetes mellitus, and inversely related to measures of insulin resistance and BMI. 157

The role of adiponectin in obese pregnant women may extend beyond usefulness as a biomarker. In the Hyperglycemia and Adverse Pregnancy Outcome study,158 serum concentrations of adiponectin decreased as glucose and maternal BMI increased and adiponectin was inversely associated with birthweight, neonatal skinfold thickness and total body fat (estimated using anthropometry), giving rise to the hypothesis that this cytokine may play a role in fetal growth regulation by modulation of placental nutrient transport in addition to maternal glucose homeostasis. Data in support of a placental origin of adiponectin remain equivocal, with evidence favouring maternal origin of adiponectin measured in the blood of pregnant women. 159 Maternal adiponectin has, therefore, the potential to be a ‘functional’ target for interventions in obese pregnant women whereby achievement of increased plasma concentrations could parallel a reduced risk of macrosomia. This may be a realistic target, as adiponectin has been shown to be modifiable by dietary intervention in non-pregnant populations. 160 Lifestyle interventions in pregnant women of differing pre-pregnancy BMI categories have been equivocal with regard to the effects on glucose metabolism and insulin resistance, although none has measured adiponectin concentration. 123,161 Following the completion of UPBEAT (1546 women), the influence of the intervention on plasma adiponectin concentration will therefore be explored.

To the best of our knowledge, there have been no previous studies of adiponectin concentration and GDM in an exclusively obese population, but the findings are consistent with other reports in women of all BMI categories with established disease or prior to the development of GDM. 156,162 A recent case–control study from Brazil of 79 and 129 women of mixed ethnicity with and without GDM, respectively, reported that GDM was associated with significantly lower serum concentrations of adiponectin in the third trimester (28–36 weeks) than in the control group (p = 0.0015). GDM and BMI both had an independent association with adiponectin concentration, with no significant interaction between the two factors (GDM, p = 0.04; BMI, p = 0.01; and interaction, p = 0.76; using a two-way analysis of variance test). 163 In contrast, although adiponectin concentration was significantly lower in women who developed GDM in our previous study in women of mixed risk, it did not contribute to the final model, which combined two factors (HDL cholesterol and tissue plasminogen activator antigen), both recognised to be related to adiponectin via linked hepatic/circulating triglyceride-mediated pathways. 148

Low serum adiponectin concentrations appear to be associated with ethnic groups known to have a higher risk of developing incident type 2 diabetes mellitus later in life. 164 In the present study, women of black ethnic origin had significantly lower plasma levels of adiponectin than non-black women, consistent with findings from previous work examining pregnant women of South Asian origin. 162 Although we acknowledge our diverse ethnic population is not representative of the UK population, having a large proportion of women of black ethnic origin (> 80%) has yielded further insights into the relationship between adiponectin concentration and ethnicity.

We also observed that adiponectin concentration was significantly related to current smoking status, a finding previously reported in a non-pregnant population in which the plasma adiponectin concentration increased in a stepwise fashion with never, past and current smokers. 165

The present study has some limitations. The sample size was small and the data obtained should be considered as a training data set for later validation in UPBEAT. Furthermore, fasting blood samples were not obtained at randomisation (15⁺⁰–17⁺⁶ weeks’ gestation), precluding the measurement of the fasting glucose or insulin concentration and assessment of HOMA2-IR (homeostatic model assessment of insulin resistance); however, as fasting is not mandatory for antenatal clinic visits, the present study was designed pragmatically, to be relevant to current clinical practice in obese women. In this high-risk group, it is possible that some women with GDM may have had undiagnosed type 2 diabetes mellitus; initial screening with fasting blood glucose or glycated haemoglobin (HbA_1c) concentrations was not undertaken as IADPSG recommendations for universal testing for overt diabetes mellitus at the first antenatal visit have not been implemented. It is possible, therefore, that low early pregnancy adiponectin concentrations may have been influenced by type 2 diabetes mellitus in a small minority of women. Nevertheless, others have shown that low adiponectin concentrations are strongly predictive of GDM in women in whom type 2 diabetes mellitus was excluded antenatally. 166 The algorithm developed in the present study was based on the diagnosis of GDM by IADPSG’s criteria; it follows that it is potentially valid only for this method of diagnosis. As IADPSG’s guidelines are increasingly being adopted, for example by WHO, it would be appropriate to evaluate in studies of larger sample size their predictive potential for GDM diagnosis by other commonly used criteria.

In summary, we have demonstrated that the risk of developing GDM in obese pregnant women may be predicted in the early second trimester of pregnancy by using an algorithm which incorporates routine clinical variables as well as the biochemical marker, adiponectin. Our findings extend the work of previous studies and, collectively, the findings suggest that by additionally measuring adiponectin concentrations in women at high risk before routine clinical diagnosis of GDM, a potential therapeutic window for intervention could be created. As GDM is associated with a greater risk of incident type 2 diabetes mellitus and 10-year cardiovascular risk in mothers,167 as well as with maternal and neonatal pregnancy complications, successful intervention has the potential to improve both short- and long-term outcomes. We conclude that further large-scale studies of GDM prediction in obese pregnant women are warranted.

Abstract

Background: Obesity in pregnancy is associated with fetal macrosomia, a raised neonatal fat mass and an increased risk of obesity and poor metabolic health in childhood that persists into adulthood. The offspring of obese women are more likely to be obese than the offspring of lean women when they become pregnant themselves, perpetuating a cycle of obesity and its associated negative metabolic consequences. Increasing PA during pregnancy could improve insulin sensitivity and reduce the risk of maternal and offspring adverse outcomes. UPBEAT is a study of a complex intervention designed to improve pregnancy outcomes through dietary changes and PA. Data from the pilot trial of 183 women were available for analysis. The relationship between the time spent at different PA levels and maternal and infant pregnancy outcomes was examined.

Key messages: Strong evidence exists that PA improves insulin sensitivity in non-pregnant populations, and lifestyle interventions of proven effectiveness in non-pregnant populations have been developed. Women who are active in pregnancy demonstrate better glucose control and favourable pregnancy outcomes. There is a lack of effective interventions to support obese pregnant women to be physically active.

Conclusions: No difference was detected in objectively measured PA between women randomised to the intervention and the control arms of the UPBEAT pilot trial. Low-intensity PA was lower in early pregnancy in women who delivered macrosomic infants. Maternal sedentary time at 35–36 weeks’ gestation was positively associated and moderate-intensity PA was inversely associated with neonatal abdominal circumference. Maternal PA is associated with infant birthweight and abdominal circumference and is an appropriate target for intervention to improve infant outcomes. The challenge remains to develop an effective intervention to support obese pregnant women to be physically active.

Background

The prevalence of obesity in pregnant women is increasing. In the UK, maternal obesity doubled from 7.6% to 15.6% between 1989 and 2007. 140 This is of concern because of the well-established association between maternal obesity (typically defined as a BMI of ≥ 30 kg/m² at the first antenatal appointment) and the risk of adverse pregnancy outcomes, including GDM, maternal hypertension, assisted delivery and macrosomia. 169 In addition to the adverse outcomes apparent at birth, it is becoming clear that offspring born to obese mothers are at an increased risk for obesity and its associated metabolic consequences in later life. A recent systematic review and meta-analysis of 20 published studies concluded that babies weighing > 4 kg at birth were more than twice as likely (OR 2.07, 95% CI 1.91 to 2.04) to be obese as adults as babies weighing ≤ 4 kg at birth. 170 In a birth cohort of 1400 individuals followed up at the age of 32 years, maternal pre-pregnancy BMI was associated with a worse cardiometabolic profile (higher BMI, waist circumference and blood pressure, and lower HDL cholesterol) in the adult offspring. 171

Interventions to reduce the impact of maternal obesity on adverse pregnancy outcomes have the potential to reap positive benefits not only for the offspring of the index pregnancy but also for subsequent generations born to these offspring. The relative importance of intrauterine exposure for an individual’s predisposition to obesity compared with the role of genetic factors and shared lifestyles during childhood is a subject of increasing interest. 172,173 It has been reported that the offspring of obese women already display metabolic disturbances at birth, supporting a role of the intrauterine environment in determining future metabolic health. An increasing maternal BMI is associated with higher levels of fat deposition in the abdomen and liver,174 insulin resistance and raised leptin levels172 at birth. An understanding of how the intrauterine environment exerts an impact on health that persists through childhood and into adulthood will help to identify appropriate targets for intervention.

Discussion

Maternal obesity is associated with an increased risk of obesity and an unfavourable cardiometabolic profile in the offspring, which persists into adulthood and, therefore, impacts future generations. Obese women are more insulin resistant than lean women during pregnancy and thus the offspring of obese women are exposed to higher levels of glucose and other nutrients in utero. Exposure to intrauterine overnutrition is associated with a higher birthweight, increased adiposity and indicators of poor metabolic health.

Physical activity during pregnancy has the potential to increase insulin sensitivity and improve the health of the offspring. Data presented here from the UPBEAT pilot trial and from other studies180,182,183 demonstrate that PA during pregnancy is associated with improved maternal glucose control and less macrosomia and adiposity in the offspring. 184 Specifically, in the UPBEAT pilot study, even LPA in obese women was associated with improved glucose tolerance and a lower infant abdominal circumference. It was of particular interest to note that more time spent on sedentary activities and less time spent on low- and moderate-intensity PA at 36 weeks’ gestation were related to an increasing infant abdominal circumference, whereas PA when measured at 28 weeks showed no such relationships. Should this observation be repeated in the larger cohort of the full RCT, in fully adjusted regression analyses, it could indicate that PA towards term may have specific benefits for reducing adiposity in the child.

Despite the existence of lifestyle interventions of proven effectiveness to increase PA and improve insulin sensitivity in non-pregnant populations, there is a paucity of interventions with demonstrated effectiveness for pregnant women. Many intervention studies targeting PA in pregnant women, including the UPBEAT pilot trial, have failed to impact PA levels.

Despite the difficulty of designing effective interventions to support obese pregnant women to be physically active on the basis of this and other studies, it is clearly premature to abandon efforts to do so. 185 In the context of the increasing problem of obesity internationally, the potential benefits that could be achieved if obese pregnant women become sufficiently active are substantial. A better understanding of the barriers to obese women engaging in PA during pregnancy is needed to inform the development of interventions. Process evaluation of the UPBEAT pilot trial revealed that the intervention was considered acceptable and well received by most women, although some reported that the lifestyle information they received was too basic and that attending group sessions was too time-consuming. 150 Refinement of the protocol for the delivery of the intervention in the UPBEAT has been made in view of these findings, although the intervention remains the same. Previous work has also found that pregnant women report a lack of time and childcare, as well as physical discomfort, as barriers to PA. 150 The intervention design needs to address these barriers by considering flexibility in the mode and location of intervention delivery, and in terms of facilitating engagement in appropriate types of activity.

In summary, it is increasingly acknowledged that efforts to prevent obesity and its associated poor health should begin in utero174 and should aim to improve maternal glucose control as this reduces the risk of obesity among offspring. 176 PA is an appropriate target for intervention to achieve this. The identification of methods by which obese pregnant women can best be supported to be sufficiently active remains a challenge.

The role of insulin resistance in obese pregnancies

An alternative approach to restricting GWG is to focus on the adverse clinical outcomes associated with obesity, and to develop interventions which are directly associated with known underlying mechanisms. A pre-pregnancy BMI of ≥ 30 kg/m², irrespective of the amount of weight gained during pregnancy, is the most important independent determinant of the risk of caesarean section, delivery of a LGA infant and postpartum weight retention. 194 In addition, the evidence linking GWG with GDM, in contrast to the strong association with pre-pregnancy BMI, is relatively weak. 195 This is, at least in part, likely to be a reflection of the strong association between maternal fat mass and insulin resistance. 9 There is a physiological increase in insulin resistance during normal pregnancy and the obese pregnant woman is at greater risk of developing GDM. Maternal hyperglycaemia and, more recently, maternal hypertriglyceridaemia are strongly implicated in the development of fetal macrosomia. 196–199 Using the method of continuous blood glucose monitoring, Harmon et al. 198 have shown, as might be expected, that obese pregnant women have an exaggerated postprandial glucose response. As the magnitude of the postprandial response was directly implicated in increasing fetal adiposity and birthweight through fetal hyperinsulinaemia, a dietary intervention focusing on reducing postprandial hyperglycaemia by lowering the dietary GL could improve maternal glucose control, reduce the incidence of GDM and lower the incidence of delivery of LGA infants. Similarly, pre-eclampsia is associated with maternal insulin resistance, and improved glucose homeostasis might lower the risk of pre-eclampsia in obese women. 200

Improving glucose homeostasis in pregnancy

Specific dietary advice including intake of low-GI foods and reduction of dietary saturated fats, as well increased PA, could contribute to improved maternal glucose homeostasis. 87,201 In a study of 50 obese Danish women designed to limit GWG, Wolff et al. 202 found that an intense dietary regime (10 1-hour sessions with a dietitian) focusing on healthy eating resulted in a reduction in plasma insulin compared with women in the control arm of the study. Another study reported that a diet and exercise regime led to a reduction in GWG and a decrease in the incidence of GDM in 126 overweight and obese Australian women,110 but no difference in infant birthweight (3.5 vs. 3.4 kg). In non-obese women with mild GDM, in whom improved glucose homeostasis is achieved through a strict regime of dietary intervention and insulin treatment when required, a reduction in the risk of adverse pregnancy outcome is achievable, as shown in two RCTs. 4,203 Higher levels of PA in normoglycaemic pregnant women and those with GDM have also been shown to improve insulin sensitivity,9 but few data of adequate power are available for the obese pregnant population. A recent meta-analysis of eight prenatal PA intervention studies, however, showed that there is a lack of consistent evidence regarding the benefits of exercise combined or not combined with dietary advice for improving glucose tolerance in obese pregnant women, which was interpreted to reflect the limited power of current evidence and poor intervention compliance. 204

Systematic review of the literature

Louie et al. 7 conducted a systematic review of the influence of lowering dietary GI in pregnancies across all BMI categories. Of the eight studies included, two suggested that a low-GI diet can reduce the risk of LGA infants in healthy pregnancies, but one reported an increase in small-for-gestational-age infants. In the three studies in which pregnancies were complicated by GDM, the evidence supported the overall advantages of a low-GI diet. This review recommended that, until larger-scale intervention trials are completed, a low-GI diet should not replace the current dietary recommendations from government and health agencies, and that further research regarding the optimal time to start a low-GI diet for maximum protection against adverse pregnancy outcomes is warranted.

In a systematic review of nine randomised trials including 743 overweight and obese pregnant women, Dodd et al. 193 reported that there was no significant effect of interventions designed to limit GWG on weight gain or on delivery of a LGA infant. In a later systematic review of 13 randomised clinical trials of lifestyle interventions in overweight and obese pregnant women (n = 1228), we concluded that there was a modest influence on GWG (–2.21 kg; 95% CI –2.86 to –1.59 kg), but no significant effect on any relevant clinical outcome. 85 We have also reviewed dietary and PA interventions in normal BMI and obese pregnant women (n = 1656 women) for the purpose of limiting GWG; in a systematic review we assessed 12 trials. Overall, diet and PA change was effective in reducing GWG, but there was considerable heterogeneity in outcomes. 2 The analysis highlighted differences in sample characteristics and aspects of intervention design, content, delivery and evaluation which might explain variation in effectiveness. Furthermore, failure to evaluate changes in behaviour or its psychological determinants could have obscured identification of the processes by which weight change is effective, and limited the ability to discern active intervention ingredients. We concluded that interventions should be more systematically designed and built on insights from behavioural science.

More recently, Thangaratinam et al. ,84 in a meta-analysis of 44 clinical trials of lifestyle or dietary interventions or a combination of both during pregnancy across all BMI ranges, found a reduction in GWG (mean reduction 1.42 kg) with any intervention in comparison with the control. 84 PA alone was associated with a reduction in birthweight (mean difference –60 g, 95% CI –120 g to –10 g). Interventions based on diet were the most effective, being associated with reductions in maternal GWG (mean 3.84 kg, 95% CI 2.45 to 5.22 kg) and a modest improvement in obstetric outcomes. However, the combination of intervention methods did not result in a reduction in the incidence of LGA infants between the groups (relative risk 0.85, 95% CI 0.66 to 1.09). Among obese women, there was no evidence of an improvement in any clinical outcome. In an editorial to this review, we highlighted that there remains a paucity of information regarding intensity, duration and compliance of the interventions, all of which could account for the lack of efficacy, as well minimal evidence for any effect of the intervention on the targeted behaviours. If the intervention does not achieve a change in behaviour in the expected direction, it follows that there will be no influence on clinical outcomes. 185

The protocol presented here describes a complex behavioural intervention comprising dietary (low GI and reduced saturated fat intake9) and PA changes that we have developed with the aim of improving glycaemic control in obese pregnant women. The intervention is based on established control theory with elements of social cognitive theory. 94,95 The primary hypothesis being tested is that an antenatal intervention package of low-glycaemic dietary advice and reduced saturated fat intake combined with advice on increased PA will reduce the incidence of maternal GDM and delivery of LGA infants. A secondary hypothesis is that the intervention will reduce the risk of obesity in the child. Prior to undertaking a trial adequately powered to investigate clinical outcomes, we completed a pilot study (n = 183 women) to determine whether or not the intervention changed dietary and physical behaviours as expected. 150 This pilot study showed that diet but not PA (as objectively measured) changed with the intervention and that all aspects of the protocol were feasible. A process evaluation led to optimisation of intervention delivery. The trial steering committee recommended continuation with recruitment for the RCT, and it was decided that the PA aspect of the intervention should remain, as this follows standard guidelines for pregnant women. 39

Study design

Multicentre RCT. For participating centres see the UPBEAT website: www.medscinet.net/upbeat/.

Ethics approval

NHS Research Ethics Committee approval was obtained in all centres (UK IRAS, reference number 09/H0802/5).

Inclusion criteria

Women with a singleton pregnancy, 15⁺⁰–18⁺⁶ weeks’ gestation and a BMI of ≥ 30 kg/m² at their first antenatal appointment.

Exclusion criteria

Women unable or unwilling to give informed consent; at < 15⁺⁰ weeks or at > 18⁺⁶ weeks’ gestation; essential hypertension requiring treatment either pre-pregnancy or in index pregnancy; (see Erratum below for omission in error of ‘pre-existing diabetes, type 1 or type 2’), pre-existing renal disease; systemic lupus erythematosus; antiphospholipid syndrome; sickle cell disease; thalassaemia; coeliac disease; thyroid disease; current psychosis; multiple pregnancy; or currently prescribed metformin.

The protocol for the study is shown in Figure 6.

FIGURE 6.

The UPBEAT protocol summary. ANC, antenatal clinic.

Trial entry

Eligible women are identified in antenatal clinics and from GP and midwives referral letters. Verbal and written information is given. Research midwives contact potential recruits, obtain verbal consent and arrange the first appointment. Care is taken to search the clinical records for those women screened in early pregnancy for pre-existing diabetes mellitus (UK recommendations are that all women with pre-GDM should be screened in early pregnancy). Women with a diagnosis of diabetes mellitus are excluded from the study (see Exclusion criteria). For those who decline to participate, permission is sought to collect minimal pregnancy outcome data.

15⁺⁰–18⁺⁶ weeks’ appointment: baseline and randomisation

At the first appointment, written informed consent is obtained. Baseline demographic information, medical and family history and current pregnancy information are collected. A short validated FFQ205 is completed to evaluate dietary GL, dietary GI, saturated fat and total sugar intake and other dietary variables. Women are weighed, their pulse and blood pressure are checked, anthropometric measurements obtained and blood and urine samples taken. Behavioural and psychological measures include the EQ-5D,98 the EPDS,99 the IPAQ206 and a binge-eating screening questionnaire. 207 Randomisation occurs at this visit via a secure internet-based data management system (MedSciNet Ltd), which is the repository for all trial data. The randomisation schedule is minimised according to ethnicity, parity (0 vs. ≥ 1), age, BMI (BMI 30–34 vs. 35–39.9 and > 40 kg/m²) and centre. Randomised women are allocated sequential study numbers, regardless of allocation to the intervention or standard care group.

Intervention

Women randomised to the intervention group attend a one-to-one interview with the HT, which includes discussion of the potential benefits of attending the weekly sessions. In the UK, HTs help people to change their behaviour to achieve personal choices and goals and, generally, do not have prespecified health professional qualifications, but do have relevant experience. All HTs in this trial receive study-specific training in all aspects of the intervention and ongoing support throughout the trial. Women in the intervention group receive a participant handbook, a DVD of an exercise regime safe for pregnancy, a pedometer and a logbook for recording weekly SMART goals and steps (as assessed by a pedometer). They are invited to attend eight sessions with the HT on a weekly basis, each lasting 1–1.5 hours. Women are encouraged to attend all sessions, but are strongly recommended to attend a minimum of five. When the sessions are not attended in person, the HT covers the session material by telephone or e-mail. Attendance and coverage of session material are documented in the study database. Following a review of the dietary and physical intervention, each session is designed to focus on different approaches in achieving the goals set. These include SMART goals, self-monitoring, and provision of feedback regarding goal attainment, identification and problem-solving of barriers, enlisting social support and providing opportunities for social comparison. At each session, a review of the previous week’s goals is undertaken.

The dietary intervention aims to promote a healthier pattern of eating, similar to that used in diabetes mellitus prevention studies, but does not aim to restrict energy intake. In order to decrease the GL, dietary advice includes replacing starchy foods with a medium/high GI by foods with a lower dietary GI, and restricting the consumption of sugar-sweetened beverages (including fruit juice) but not fruit. Participants are also given dietary advice to reduce SFA intake. No advice was given regarding GWG.

Advice regarding PA focuses on increasing the daily step count incrementally, and being more active in daily life. Pedometers are used for monitoring and motivation. The emphasis is on walking at a moderate intensity with additional options included, especially for those who are already engaging in some PA. This degree of activity accords with that recommended by the UK’s Royal College of Obstetricians and Gynaecologists. 39

Standard care

Women randomised to the standard care group attend routine antenatal care in accordance with local health-care provision. The UK recommendations state that all pregnant women, and particularly those with a BMI of ≥ 30 kg/m², should be advised by a health professional at the earliest opportunity of the risks of obesity in pregnancy and be given advice about a healthy diet and safe levels of PA. Recommendations for referral to a registered dietitian are infrequently implemented. No specific advice is given about GWG and women are weighed only at their first antenatal visit. 93 Women should also be encouraged to lose weight after pregnancy.

27⁺⁰–28⁺⁶ weeks’ appointment

All women in both groups attend for an OGTT at 27⁺⁰–28⁺⁶ weeks’ gestation (fasting for a minimum of 10 hours, 75-g glucose load). At this visit weight and anthropometric measurements are taken, health in current pregnancy noted, additional blood and urine samples taken, and dietary FFQ, EQ-5D, EPDS, IPAQ and questionnaires about binge-eating completed. Early pregnancy data including blood pressure, blood chemistry and anomaly scan reports are entered from routine clinical records.

34⁺⁰–36⁺⁰ weeks’ appointment

Women in both arms of the study attend the research appointment at 34⁺⁰–36⁺⁰ weeks’ gestation. Current health in pregnancy is recorded, weight and anthropometric measurements taken, blood and urine samples collected and dietary FFQ, EQ-5D, EPDS, IPAQ and binge-eating questionnaires completed.

Unexpected adverse events are reported in accordance with good clinical practice guidance.

Pregnancy outcome data

Following delivery, information is collected from maternal medical records regarding health in late pregnancy, labour onset, mode of delivery, blood loss, antenatal and postnatal inpatient nights. When possible a cord blood sample is taken.

Neonatal and postnatal outcome data include Apgar scores, admission to special care baby unit and inpatient nights. To address the influence of the intervention on fetal growth and adiposity, neonatal anthropometry and length measurements are undertaken within 72 hours of birth.

Six months post partum

To determine whether or not the intervention has led to sustained change in maternal dietary and PA behaviours, diet is assessed by FFQ and PA by IPAQ. Maternal demographic data, health since pregnancy and smoking history are obtained. Maternal anthropometric measures are taken. EPDS, Three-Factor Eating Questionnaire-R18208 and binge-eating questionnaires are completed. To address safety and the influence of the intervention on the long-term health of the child, details regarding the child’s health from birth are obtained. If cord blood was not taken, and if the parents provide consent, a buccal cell sample is taken from the child’s mouth for deoxyribonucleic acid (DNA) extraction (Oragene, DNA Genotek, Ottawa, ON, Canada). To address the potential influence of the intervention on infant adiposity at 6 months and to obtain information on known determinants of childhood obesity, infant length and other anthropometric measures are taken. The mother provides information for an infant feeding and growth questionnaire209 and a validated questionnaire addressing appetite, the Baby Eating and Behaviour Questionnaire. 210 Information on activity using questions from the Infant Behaviour Questionnaire-Revised211 and sleep patterns are obtained212 and information on childcare (kindergarten, other carers) collected.

Primary maternal outcome

Gestational diabetes mellitus was diagnosed by OGTT at 27⁺⁰–28⁺⁶ weeks’ gestation in accordance with the criteria recommended by IADPSG (i.e. fasting blood glucose concentration of ≥ 5.1 mmol/l or 1-hour glucose of ≥ 10.0 mmol/l or 2-hour glucose of ≥ 8.5 mmol/l).

Primary neonatal outcome

A LGA infant is defined as an infant whose adjusted birthweight is > 90th centile for gestational age, adjusting for maternal height, corrected maternal weight, ethnicity, parity and sex of baby.

Secondary outcomes

Subgroup analysis

Women who are treated for GDM; differences in diagnostic thresholds between centres will be accommodated by minimisation by centre. Other subgroups likely to be of interest include demographic and socioeconomic status (as assessed by the Index of Multiple Deprivation), ethnic groups, BMI categories, groups of different parity and smokers.

Interaction tests will be used to determine whether or not treatment is particularly effective in individual subgroups. Performance of subgroup analysis will be dependent on sufficient data. Because of the well-known risk of false positives, both main effects and interaction tests will be performed before considering results for subgroups.

Sample size

In the pilot RCT,150 30% of women in the standard care arm developed GDM in accordance with the IADPSG’s criteria. 208 A total of 1546 women (including allowance for 20% dropout) (773/arm) will be recruited to provide 80% power to detect a 25% reduction in the incidence of GDM. Considering LGA deliveries, for a 30% relative risk reduction from an estimated 17.2% of LGA to 12.0% in the intervention arm, 1546 women would give 80% power. 215,216

Analysis

To determine whether the trial participants are representative of the general population, relevant parameters available from electronic summary patient records will be compared between eligible women agreeing and declining to take part. Analyses will follow the intention-to-treat principle.

Following CONSORT guidelines, risk ratios and risk differences will be estimated by binary regression for yes/no outcomes. When measurements are repeated over time, results [mean (SD) or n (%)] will be presented separately at each time point. Randomised comparisons with 95% CIs will be made using linear regression with robust standard errors, adjusting for the baseline value when appropriate.

Multiple regression models will be used to address the influence of maternal exposures on neonatal and infant (6 months) body composition and the role of paternal factors.

Exclusion criteria

Women unable or unwilling to give informed consent; < 15^+ 0 weeks or > 18^+ 6 weeks’ gestation; essential hypertension requiring treatment either pre pregnancy or in index pregnancy; pre-existing diabetes mellitus (type 1 or type 2); pre-existing renal disease; systemic lupus erythematosus; antiphospholipid syndrome; sickle cell disease; thalassaemia; coeliac disease; thyroid disease; current psychosis; multiple pregnancy; currently prescribed metformin.

Study design

We did this multicentre, randomised controlled trial at antenatal clinics in eight inner-city NHS Trust Hospitals in the UK – London (three centres), Bradford, Glasgow, Manchester, Newcastle, and Sunderland. The detailed study design and protocol have been published elsewhere. 186 A flow chart of the protocol is shown in the appendix (Figure 7). We did the study according to the UK’s National Institute for Health and Care Excellence (NICE) guidelines for diabetes in pregnancy, in which early pregnancy biochemistry screening for glucose intolerance and risk of gestational diabetes is not recommended. 221 The NHS research ethics committee approved the study protocol for all centres (UK integrated research application system, reference 09/H0802/5). The trial steering committee approved the protocol and the analysis plan and provided oversight of all aspects of the trial, including safety.

FIGURE 7.

Trial protocol summary. ANC; antenatal care.

Participants

We recruited women older than 16 years with a BMI of 30 kg/m² or higher and a singleton pregnancy between 15 weeks and 18 weeks plus 6 days of gestation. We excluded individuals if they were unwilling or unable to give informed consent; if they had underlying disorders, including a pre-pregnancy diagnosis of essential hypertension, diabetes, renal disease, systemic lupus erythematosus, antiphospholipid syndrome, sickle-cell disease, thalassaemia, coeliac disease, thyroid disease, and current psychosis; or if they were currently being prescribed metformin. All participants provided written informed consent. For women who declined to participate, we recorded age, BMI, ethnic origin, socioeconomic status, and outcome data if permission was granted.

Randomisation and masking

We randomly allocated participants to either standard antenatal care or the behavioural intervention plus standard antenatal care. We used a computer-generated randomisation procedure via a password-protected website. Allocation to study groups was done by the centre’s UPBEAT trial midwife. We used minimisation, according to ethnic origin (black, white, Asian, other), parity (primiparous, multiparous), age (≤ 24, 25–29, 30–34, ≥ 35 years), BMI (30.0–34.9, 35.0–39.9, ≥ 40 kg/m²), and centre. In view of the nature of the intervention, participants and staff were aware of allocations.

Procedures

Within 1 week of randomisation, women in the intervention group attended an individual interview at their trial centre with a health trainer (a person with skills in assisting behavioural change, but not necessarily a health professional) who received coaching in all aspects of the intervention and ongoing support throughout the study period. 186 The intervention, which was informed by control theory and elements of social cognitive theory, consisted of eight further health trainer-led group or individual sessions of 1 h duration once a week for 8 weeks. 186 If a participant could not attend a session in person, the material was covered by telephone or e-mail, providing flexibility in intervention delivery. Every session addressed approaches to achieving SMART goals (i.e. specific, measurable, achievable, relevant, time-specific) and reviewed the previous week’s goals. Women assigned to the intervention received advice on self-monitoring, identification and problem-solving of barriers to behaviour change; enlisting social support; and providing opportunities for social comparison. We encouraged participants to attend all sessions and provided them with a handbook in which information was included about the intervention and the theory behind it, with recommended foods and recipes, and suggestions for physical activity. We also gave the women a DVD of an exercise regimen that was safe for pregnancy, a pedometer, and a log book for recording their weekly SMART goals. The intention of the intervention was to improve glucose tolerance through dietary and physical activity behaviour change. With the dietary intervention we aimed to promote a healthy pattern of eating but not necessarily to restrict energy intake. We tailored recommendations to the woman’s habitual diet and cultural preference, and suggested exchanging carbohydrate-rich foods with a medium-to-high glycaemic index for those with a lower glycaemic index to reduce the glycaemic load, and restricting dietary intake of saturated fat. With respect to advice on physical activity, we focused on incremental increases in walking from a pedometer- assessed baseline, tailored to pre-existing activities. The emphasis of the exercise intervention was on walking at a moderate intensity, with additional options included, particularly for women already engaging in some physical activity. Further details are available in the protocol. 186 Women in the intervention group continued with their routine antenatal care appointments.

Women who were allocated to the standard antenatal care group continued to attend routine antenatal appointments at their trial centre, according to local practice. Typically, women would attend nine appointments. Recommendations of UK guidelines are for women with obesity to be advised, at first contact with a health professional, and at no other time, about a healthy diet and the benefits of physical activity. 93,221 We did not provide any additional information, including any details of the nature of the intervention.

For diagnosis of gestational diabetes, we gave all participants an oral glucose tolerance test (75-g load) between 27 weeks and 28 weeks plus 6 days of gestation. We used diagnostic criteria recommended by the International Association of Diabetes and Pregnancy Study Groups (IADPSG) (i.e., fasting venous glucose of 5.1 mmol/l or higher, 1 h venous glucose of 10.0 mmol/l or higher, 2 h venous glucose of 8.5 mmol/l or higher, or a combination of these). 6 We used these criteria not only because of their increasing adoption globally (and by WHO) but also because of differences in routine diagnostic criteria used by trial centres. We referred women who were diagnosed with gestational diabetes for antenatal diabetic services, according to local practice at every centre.

To assess the efficacy of the behavioural intervention, we gathered maternal dietary data and physical activity scores, calculated gestational weight gain and took maternal anthropometric measurements. We used standard laboratory methods to measure biochemical outcomes between 27 weeks and 28 weeks plus 6 days of gestation.

We used a food frequency questionnaire186,222 to assess the diet of participants for the month before randomisation and for the month before the study visit at between 27 weeks and 28 weeks plus 6 days of gestation. We adapted this questionnaire from one used in the UK arm of the European Prospective Investigation into Cancer Study. 222 We used WISP 3.0 (Tinuviel Software, Llanfechell, Anglesey, UK) to calculate nutritional composition and glycaemic load per 100 g of food and beverage items. We excluded from the analysis data for participants who we estimated were under-reporting (≤ 4.5 MJ/day) and over-reporting. 223

We measured physical activity at randomisation and at the study visit between 27 weeks and 28 weeks plus 6 days of gestation. We used the International Physical Activity Questionnaire (IPAQ) and summarised data according to established methods. 186 We calculated physical activity (minutes/week) as metabolic equivalents (METs) – i.e., the ratio of energy expenditure for an activity to energy expenditure at rest – with the formula 8.0 × vigorous activity + 4.0 × moderate activity + 3.3 × light activity (walking).

At delivery of the infant, we measured and weighed the baby. We calculated customised birthweight centiles with Gestation Related Optimal Weight (GROW) software, version 6.7.5.1 (Gestation Network, Perinatal Institute, Birmingham, UK).

Outcomes

The primary maternal outcome was gestational diabetes. Pre-specified secondary outcomes included dietary measures, physical activity scores, gestational weight gain, maternal anthropometric measurements (mid-arm and thigh circumference and subscapular, triceps, biceps, and suprailiac skinfold thicknesses), and biochemical outcomes (maternal fasting plasma glucose, fasting plasma insulin, insulin resistance [calculated by homoeostatic model assessment, HOMA2-IR]213, fasting triglycerides, LDL cholesterol, HDL cholesterol, and VLDL cholesterol). We pre-specified several other secondary clinical maternal outcomes: pre-eclampsia (defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or both, on at least two occasions 4 h apart, with proteinuria ≥ 300 mg/24 h or spot urine protein to creatinine ratio ≥ 30 mg/mmol creatinine, or urine dipstick protein ≥ 2+); severe pre-eclampsia (defined as systolic blood pressure ≥ 170 mmHg, diastolic blood pressure ≥ 110 mmHg, or both, with proteinuria ≥ 500 mg/24 h or spot urine protein to creatinine ratio ≥ 50 mg/mmol creatinine, or urine dipstick protein ≥ 3+); mode of delivery (elective or emergency caesarean section, vaginal delivery, or operative vaginal delivery); induction of labour; blood loss at delivery (> 1000 ml or > 2000 ml); inpatient nights (antenatal and postnatal); referral to diabetic antenatal service after oral glucose tolerance test; and a requirement for insulin or metformin during pregnancy. Prespecified maternal secondary outcomes not reported here are listed in the Supplementary text below.

The primary neonatal outcome was delivery of a large-for-gestational-age infant, which we defined as the 90th or higher customised birthweight centile for gestational age, adjusting for maternal height and weight, ethnic origin, parity, and sex of the baby. We pre-specified several secondary neonatal outcomes: gestational age at delivery; delivery at less than 37 weeks and less than 34 weeks; birthweight; birthweight 4.0 kg or heavier, 2.5 kg or lighter, or 1.5 kg or lighter; customised birthweight centile (≥ 95th, ≤ 10th, and ≤ 5th); neonatal death; days in special care baby unit; total inpatient days; discharge home on oxygen; confirmed infection; retinopathy of prematurity; neonatal hypoglycaemia; intraventricular haemorrhage; need for mechanical ventilation and duration; necrotising enterocolitis; pulmonary haemorrhage, skinfold thicknesses and circumferences; and birthweight centiles as population centiles (≥ 90th, ≥ 95th, ≤ 10th, and ≤ 5th). Pre-specified neonatal secondary outcomes not reported here are listed in the Supplementary text.

Adverse events other than those prespecified as secondary outcomes included miscarriage, late termination of pregnancy, maternal accident, placental abruption, antenatal and postnatal sepsis, iatrogenic premature birth, intrauterine complications (fetal cardiac, renal, respiratory, and neurological), fetal death in utero, unspecified neonatal complications at birth, and confirmed neonatal sepsis.

Statistical analysis

We calculated that a sample size of 1546 women (allowing for 20% dropout) would provide at least 80% power to detect a clinically important 25% reduction in the incidence of gestational diabetes, from 30% (observed in the pilot study of 183 women)150 to 23%. From a review of published population birthweight centiles in obese UK women,16 1546 infants provided 80% power to detect a 30% relative risk reduction for large-for-gestational-age infants (17.2% to 12.0%). 215

Our analysis was by intention to treat. We expressed treatment effects for binary endpoints as risk ratios (relative risk) with 95% CIs, using binomial regression and adjusting for maternal BMI, ethnic origin, and parity (i.e. minimisation variables for intervention allocation). We calculated risk differences and did significance tests for both primary endpoints. For continuous measurements, we used linear regression with robust SEs, adjusting for baseline data or the variables used for minimisation. For physical activity data, we did median regression. For biochemical data, we did log transformations for normality, as appropriate. To check for the potential of a variable response to the intervention, we did subgroup analyses with interaction tests for BMI, ethnic origin, socioeconomic status, parity, and smoking. Moreover, to ascertain whether attendance at intervention sessions affected outcome, we did further interaction tests.

For the main analysis, we followed the missing-at-random assumption. Predictors of missingness, which we included to ensure an unbiased measure of treatment effect, were maternal BMI, ethnic origin and parity. To test the possibility of undetectable bias attributable to missing data, we did a series of analyses under different missing-not-at-random assumptions for the primary maternal and neonatal endpoints, with the Stata command rctmiss. We tested the assumptions that the odds of disease in participants with missing data were variously half or double that for women with complete data, in both study groups or in one group only. We did all analyses with Stata version 13.1.

This trial is registered with Current Controlled Trials, ISCRTN89971375.

Role of the funding source

The funders had no role in study design, data collection, data analysis, data interpretation or writing of the report. The corresponding author had full access to all data in the study and had final responsibility for the decision to submit for publication.

Pre-specified outcomes; as in published protocol

Prespecified secondary outcomes not included in the manuscript include:

Resource utilisation and costs

The trial protocol was based on the delivery of eight sessions of health training to participants. Figure 9 presents patterns of engagement with sessions and how successful trainers were at setting goals. The most common outcome is that participants received all eight sessions and, therefore, worked towards a series of eight training goals. The second most common pattern represents early withdrawal. Overall, the pattern suggests that the more face-to-face sessions received, the more likely that the participant would be to receive eight training goals via either face to face or follow-up contacts.

FIGURE 9.

Total number of face-to-face health training sessions attended by participants and the total number of goals set during the training programme.

Table 31 presents resource utilisation (by study group) in four main categories. The first category presents average resources of the intervention (health training) and indicates the total number of patient contacts in which at least one goal was set. The second component examines antenatal admissions and type of admission (day case admission or overnight) and the related reason for admission. Third, resources related to pregnancy cessation present either mode of birth or rate of miscarriage or termination. Finally, postnatal admissions provide information on length of maternal admissions, rates of neonatal admission and neonatal length of stay. All resource use data are presented either for the full sample (all available information) or for complete cases (which informs the subsequent cost-effectiveness analysis).

Health training aimed to provide weekly 1.5-hour sessions for 8 weeks during the intervention period of the study. Participants received an average of 5.37 contacts, in which at least one goal was set, of which 4.10 contacts occurred face to face and 1.27 were follow-ups (telephone or e-mail).

The most commonly reported reason for an overnight prenatal hospital admission was for ‘hyperemesis/vaginal bleeding’. The most common day case was reported as related to trauma. Overall, the mean number of inpatient nights was 2.754 (SD 2.398) in the control group and 2.824 (SD 3.462) in the intervention group.

Pregnancy cessation was used in Healthcare Resource Groups (HRGs) categories; HRGs are commonly used to inform reimbursement of hospital trusts and as such form the subsequent basis for cost-effectiveness analysis. The most commonly recorded HRG mode of birth was normal delivery [comorbidities or complications (CC) score = 0] (intervention, 24.3%; control, 23.9%), followed by emergency caesarean section (CC score 0–1) (intervention, 16.1%; control, 19.4%). Rates of miscarriage (between 15 and 24 weeks) were low (intervention, 1.3%; control, 0.9%) and likewise observed terminations (between 15 and 24 weeks) were rare (intervention, 0.1%; control, 0.3%). Responses to EQ-5D questionnaires (required for complete case analysis) are missing in all cases of miscarriage and termination; therefore, the figure was drawn from the ‘full sample’. No data about the end of pregnancy were missing in the intervention group (complete case analysis); however, 0.5% of data were missing in the control group.

On average, maternal length of stay was similar to the national averages (intervention group, 2.123 days; control group, 1.996 days). Mean rates of neonatal admissions demonstrate some differences between groups (intervention 7.6% vs. control 4.1%); however, when compared in the full sample the difference diminishes (intervention 8.5% vs. control 7.5%), suggesting that this difference may be an artefact of sample attrition. Mean neonatal length of stay displays similarities between groups (intervention 2.235 nights vs. control 2.096 nights), although inspection of conditional means (i.e. mean given a neonatal admission) demonstrates interesting differences [intervention 12.951 nights (n = 61) vs. control 18.058 nights (n = 52)].

Table 32 presents unit cost data for each item of resource use. All unit costs are based on the NHS price year 2012–13 to represent the halfway time point of randomisation into UPBEAT. The HTs providing intervention were valued at the same rate as ‘clinical support worker (hospital)’. Specific costs were applied for all antenatal admissions. The numbers of CCs used to inform specific to modes of birth were based on either the presence of systemic comorbidities [pre-eclampsia, pre-existing hypertension, gestational hypertension, cardiovascular disease, pre-existing diabetes mellitus, GDM (previous history or currently diagnosed), genitourinary or general infection or existing mental health disorder] or labour complications [preterm labour (< 37 weeks’ gestation), long labour (i.e. second stage > 4 hours) or postpartum haemorrhage (blood loss of > 500 ml)]. Average bed stay post partum generally depends on mode of delivery; however, the cost for ‘non-elective inpatient – long-stay excess bed-days’ is applied when the maternal length of stay exceeds 2 days. Neonatal stay is incorporated into the cost of the mode of delivery; however, neonatal admissions are treated separately. In the absence of further information from the trial, all neonatal nights are assumed to be neonatal critical care, normal care.

Table 33 presents the total costs by study group, and subtotal costs (by item groups) and costs by specific item. Again, to aid consideration of the implications of missing data, the results are further subdivided into those drawn from the full sample and those available to inform the complete case cost-effectiveness analysis.

The mean cost of health training was £485 based on complete cases (n = 510). This is lower than would have been expected based on the protocol description of the intended intervention (i.e. ‘eight group sessions with the HT on a weekly basis, each lasting 1.5 hours’ equating to £624). However, when compared with the intervention cost based on the full sample (£390) the complete case estimate is higher and may be attributed to several potential factors (e.g. consumer preferences, attrition during intervention due to the loss of the pregnancy, etc.).

Overall, aggregated costs of all antenatal admissions would seem significantly higher in the intervention group (intervention £251 vs. control £177), although again, compared with results from the full sample (intervention £189 vs. control £168), any difference may be related to underlying patterns of missing data.

As complete case data are contingent on the completion of the EQ-5D, the cost of pregnancy cessation (for complete case analysis) excludes costs related to pregnancy loss and is solely based on the HRG mode of birth. Over all the varying modes of birth, the averaged cost is found to be identical (intervention £2177 vs. control £2177).

Costs associated with longer stays (i.e. more than two nights) are slightly higher in the intervention group than in the control group (intervention £536 vs. control £432).

The cost related to neonatal stay was higher in the intervention group than in the control group (intervention £187 vs. control £128). This may initially seem counterintuitive when considering that the conditional mean length of stay was higher in the control group; however, the rate of neonatal admissions in the intervention group was almost double that in the control group (7.6% vs. 4.1%, respectively), meaning that over the entire observed sample costs were higher.

Outcomes

The EuroQol-5 Dimensions, three-level (EQ-5D-3L), scores at baseline, 20 weeks’ and 36 weeks’ gestation are presented in Table 34. Comparing the two groups, scores are displayed as similarities across time points and as an aggregated QALY.

Cost-effectiveness analysis

Table 35 presents the results of a seemingly unrelated regression. The model tests the joint distribution of the effect of treatment (health training) on the incremental total costs and QALY, whether or not the effects are significantly different from those in the control group and whether or not the cost and QALYs are significantly correlated.

The results suggest that health training results in a non-significant mean increase in QALYs of 0.002 (95% –0.004 to 0.009) and a significant increase in mean total costs of £722 (95% CI £473 to £970). The ICER is £331,630/QALY.

To further explore uncertainty surrounding the ICER, Figure 10 presents results of the non-parametric bootstrap (10,000 replications) on a cost-effectiveness plane.

FIGURE 10.

Cost-effectiveness plane (10,000 replications).

To illustrate implications of uncertainty within the context of health care decision-making, the results of the bootstrap analyis are further expressed on a CEAC (Figure 11). This suggests that, based on the available information from UPBEAT, the probability that health training would be below NICE’s stated upper threshold of willingness to pay (£30,000/QALY) is only 1%.

FIGURE 11.

Cost-effectiveness acceptability curve illustrating the probability that the intervention is cost-effective at increasing values of willingness to pay for an additional QALY.

Health economic analysis

Provision of the intervention was estimated to cost, per participant, an average of £485 more than routine care. Allocation to the intervention was associated with a non-significant increase in QALYs (0.002, 95% CI –0.004 to 0.009; p = 0.52). Base case cost-effectiveness analysis found that the ICER was £331,630 per QALY. Based on the total cost to the NHS provider and health gains from randomisation to 36 weeks’ gestation, UPBEAT provided no supporting evidence to suggest the intervention represented value for money based on NICE’s benchmarks for cost-effectiveness.

Centres

Others

Peer-reviewed papers arising from programme

Gardner B, Wardle J, Poston L, Croker H. Changing diet and physical activity to reduce gestational weight gain: a meta-analysis. Obes Rev 2011;12:e602–20. http://dx.doi.org/10.1111/j.1467-789X.2011.00884.x

Kinnunen TI, Tennant PW, McParlin C, Poston L, Robson SC, Bell R. Agreement between pedometer and accelerometer in measuring physical activity in overweight and obese pregnant women. BMC Public Health 2011;11:501. http://dx.doi.org/10.1186/1471-2458-11-501

Gardner B, Croker H, Barr S, Briley A, Poston L, Wardle J. Psychological predictors of dietary intentions in pregnancy. J Hum Nutr Diet 2012;25:345–53.

Oteng-Ntim E, Varma R, Croker H, Poston L, Doyle P. Lifestyle interventions for overweight and obese pregnant women to improve pregnancy outcome: systematic review and meta-analysis. BMC Med 2012;10:47. http://dx.doi.org/10.1186/1741-7015-10-47

Poston L, Briley AL, Barr S, Bell R, Croker H, Coxon K, et al. Developing a complex intervention for diet and activity behaviour change in obese pregnant women (the UPBEAT trial); assessment of behavioural change and process evaluation in a pilot randomised controlled trial. BMC Pregnancy Childbirth 2013;13:148.

Poston L. Gestational weight gain: influences on the long-term health of the child. Curr Opin Clin Nutr Metab Care 2012;15:252–7. http://dx.doi.org/10.1097/MCO.0b013e3283527cf2

Briley AL, Barr S, Badger S, Bell R, Croker H, Godfrey KM, et al. A complex intervention to improve pregnancy outcome in obese women; the UPBEAT randomised controlled trial. BMC Pregnancy Childbirth 2014;14:74.

Maitland RA, Seed PT, Briley AL, Homsy M, Thomas S, Pasupathy D, et al. Prediction of gestational diabetes in obese pregnant women from the UK Pregnancies Better Eating and Activity (UPBEAT) pilot trial. Diabet Med 2014;31:963–70. http://dx.doi.org/10.1111/dme.12482

Hayes L, Bell R, Robson S, Poston L, UPBEAT Consortium. Association between physical activity in obese pregnant women and pregnancy outcomes: the UPBEAT pilot study. Ann Nutr Metab 2014;64:239–46. http://dx.doi.org/10.1159/000365027

Hayes L, Mcparlin C, Kinnunen TI, Poston L, Robson SC, Bell R, on behalf of the UPBEAT Consortium. Change in level of physical activity during pregnancy in obese women: findings from the UPBEAT pilot trial. BMC Pregnancy Childbirth 2015;15:52.

Poston L, Bell R, Croker H, Flynn AC, Godfrey KM, Goff L, et al. Effect of a behavioural intervention in obese pregnant women (the UPBEAT study): a multicentre, randomised controlled trial. Lancet Diabetes Endocrinol 2015;3:767–77. http://dx.doi.org/10.1016/S2213-8587(15)00227-2

White SL, Lawlor DA, Briley AL, Godfrey KM, Nelson SM, Oteng-Ntim E, et al. Early antenatal prediction of gestational diabetes in obese women: development of prediction tools for targeted intervention. PLOS ONE 2016;11:e0167846.

Flynn AC, Seed PT, Patel N, Barr S, Bell R, Briley AL, et al. Dietary patterns in obese pregnant women; influence of a behavioral intervention of diet and physical activity in the UPBEAT randomized controlled trial. Int J Behav Nutr Phys Act 2016;13:124.

Conference proceedings

Sandall J, Hunt C, on behalf of the UPBEAT team. ‘Fat Groups, and All That Sort of Business, It Does Nothing for Me’: Gaining Insights into UPBEAT: the UK Better Eating and Activity Trial (UPBEAT). Annual Academic Meeting in Obstetrics and Gynaecology, Royal College of Obstetricians and Gynaecologists, London, UK, 18–19 November 2010.

Sandall J, Hunt C, Poston L, Briley A, Oteng-Ntim E, Khasaezadeh N, et al. Inside the ‘black box’: process evaluation of the upbeat pilot trial. Arch Dis Child Fetal Neonatal Ed 2011;96:Fa134.

Hunt C, Sandall J, Briley A, Robson S, Poston L, on behalf of the UPBEAT team. Inside the ‘Black Box’: Process Evaluation of the Upbeat Pilot Trial. Poster, Perinatal Medicine 2011, Harrogate, UK, 15–17 June 2011.

Osbourne K. Follow-up Study of Women Participating in UPBEAT. Master’s in Public Health. King’s College London, London, UK, 2013.

Hilton C. Factors Associated with Physical Activity in UPBEAT. Master’s in Research. King’s College London, London, UK, 2016.

Sandall J. Research Masterclass: Evaluation of Complex Interventions Drawing on Recent Process and Implementation Evaluation of the UPBEAT RCT. University of Technology, Sydney, NSW, Australia, 1 October 2014.

Sandall J, on behalf of UPBEAT trial group. A Multicentre Randomised Controlled Trial of a Behavioural Intervention in Obese Pregnant Women; the UPBEAT Study. 10th International Normal Birth Conference, Grange-Over-Sands, UK, 15–17 June 2015.

McParlin C, Kinnunen TI, Tennant PWG, Poston L, Bell R, Robson SC. Objective Measurement of Physical Activity in Overweight and Obese Pregnant Women. British Maternal & Fetal Medicine Society, Newcastle, UK, 10 June 2010.

Kinnunen TI, McParlin C, Tennant P, Robson S, Poston L, Bell R. Physical Activity in Overweight and Obese Pregnant Women Assessed by Two Objective Methods. 11th International Conference on Obesity, Stockholm, Sweden, 11–15 July 2010.

Barr S, Poston L, Briley A, Oteng-Ntim E, Holmes B, Kinnunen T, et al. Habitual Dietary Intake of Obese Pregnant Women in the UK. Nutrition Society, London, UK, 6 June 2011.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. British Maternal & Fetal Medicine Society, Harrogate, UK, 16 June 2011.

Sandall J, Hunt C, Poston L, Briley A, Oteng-Ntim E, Khasaezadeh N, et al. Inside the ‘Black Box’; Process Evaluation of the UPBEAT Pilot Trial. British Maternal & Fetal Medicine Society, Harrogate, UK, 16 June 2011.

Briley A, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. Development of a Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Protocol. 58th Annual Meeting, Society of Gynecologic Investigation, Miami, FL, USA, 16–19 March 2011.

Sandall J. What Happens When the Trial Finishes? Using Normalization Process Theory to Understand Implementation; Translating Research into Policy and Practice: Strategies and Challenges for Making Change Happen! Invited presentation, International Workshop on Trials of Complex Interventions, University of Technology Sydney, NSW, Australia, 17 February 2011.

Molyneaux E, Micali N, Poston L, Howard L. Depression, Obesity and Dietary Patterns During Pregnancy. Marcé Society, Paris, France, 3 October 2012.

Hayes L, Kinnunen T, Robson S, McParlin C, Poston L, Bell R. Factors Associated with Level of Physical Activity in Obese Pregnant Women Participating in the UPBEAT Pilot Trial. British Maternal & Fetal Medicine Society, Glasgow, UK, 19 April 2012.

Hayes L, Kinnunen TI, Robson S, Poston L, Bell R. Maternal Blood Glucose and Objectively Measured Physical Activity. Poster, European Diabetes Epidemiology Group Annual Meeting, Swansea, UK, 12–15 May 2012.

Hayes L, Kinnunen TI, Robson S, Poston L, Bell R. Relationship Between Objectively Measured Physical Activity and Maternal Blood Glucose. International Society for Behavioural Nutrition and Physical Activity Annual Meeting, Ghent, Belgium, 22–25 May 2013.

Maitland R, Sattar N, Seed PT, Briley AL, Thomas S, Pasupathy D, et al. Prediction of Gestational Diabetes in Obese Pregnant Women. Society for Gynecologic Investigation, Orlando, FL, USA, 20–23 March 2013.

Maitland R, Sattar N, Seed PT, Briley AL, Thomas S, Pasupathy D, et al. Prediction of Gestational Diabetes in Obese Pregnant Women. Diabetes UK, Manchester, UK, 13–15 March 2013.

Flynn A, Kader S, Poston L, Goff L. Improvements in Diet Quality in the UPBEAT study: A Lifestyle Intervention to Improve Outcomes in Obese Pregnancy. Physiology Society Meeting, Obesity: A Physiological Perspective, Newcastle, UK, 10–12 September 2014.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. Physiology Society Meeting, Obesity: A Physiological Perspective, Newcastle, UK, 10–12 September 2014.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. 17th World Congress of Basic and Clinical Pharmacology, Cape Town, South Africa, 13–18 July 2014.

Hayes L, Bell R, Robson S, Poston L, on behalf of the UPBEAT Consortium. Association Between Physical Activity in Obese Pregnant Women and Health of the Offspring. EU Early Nutrition Academy, The Power of Programming: International Conference on Developmental Origins of Adiposity and Long Term Health, Munich, Germany, 13–15 March 2014.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. EU Early Nutrition Academy. Power of Programming Meeting, Munich, Germany, 13–15 March 2014.

Hayes L. Association Between Physical Activity in Obese Pregnant Women and Health of the Offspring. Association for the Study of Obesity, Conference on Maternal Obesity, London, UK, 12 March 2014.

Hayes L, Robson S, Poston L, Bell R, on behalf of the UPBEAT Consortium. Changes in Physical Activity During Pregnancy and Relationship with Pregnancy Outcomes. European Diabetes Epidemiology Group Annual Meeting, Quartu Sant’Elena, Sardinia, 29 March–1 April 2014.

Schneeberger C, Flynn A, Barr S, Seed P, Inskip H, Poston L. Maternal Diet Patterns and Glycaemic Load in Obese Pregnant Women Taking Part in a Pilot Trial of a Lifestyle Intervention (the UPBEAT Trial). American Diabetes Association’s 74th Scientific Sessions, San Francisco, CA, USA, 13–17 June 2014.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. Keynote Speaker, 3rd National Diabetes in Pregnancy Conference, Gateshead, UK, 9 October 2014.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. A Complex Intervention to Improve Outcome in Obese Pregnancies; the UPBEAT Study. American Diabetes Association’s 75th Scientific Sessions, Boston, MA, USA, 5–9 June 2015.

Ireland VF, Flynn A, Croker H, Goff LM. Observational Study of Eating Behaviours, BMI and Diet in Obese Pregnant Women. British Dietetic Association Research Symposium, Birmingham, UK, 2 December 2015.

Poston L. Obesity and Pregnancy. Hot Topic Conference World Obesity Confederation, invited speaker, London, UK, 30 October 2015.

Poston L, Holmes B, Kinnunen T, Croker H, Bell R, Sanders T, et al. The UPBEAT trial. 9th World Congress of Developmental Origins of Disease, Cape Town, South Africa, 8–11 November 2015.