Obesity is prevalent in women of reproductive age in both high and low-to-middle income countries [1]. Pregnant obese women have a heightened risk of adverse pregnancy outcomes [2], but at present there is no evidence-based intervention that can be introduced into clinical practice to improve pregnancy outcome in obese women. The majority of attempts to develop interventions have hitherto focused on limiting gestational weight gain (GWG) according to the USA Institute of Medicine (IOM) recommendations [3]. Recent meta-analyses of relevant studies in obese women show modest restriction of GWG without robust evidence for improved clinical outcome [4, 5]. Limitations of the existing evidence include poor study design, small sample size, absence of a theoretical basis and, importantly, no a priori demonstration of the feasibility of the intervention in regard to changing the specific behaviours targeted [6]. We have developed a theoretically based behavioural group intervention (diet and physical activity) for obese pregnant women with the primary aim of improving maternal glucose homeostasis. As maternal insulin resistance is integral to many complications of obese pregnancy the dietary intervention focuses on lowering the dietary glycaemic index (GI), previously shown to improve pregnancy outcome in women with gestational diabetes (GDM) [7, 8]. Increased physical activity can also improve metabolic control and reduce GDM risk in pregnant women [9].

Prior to embarking on a large randomised controlled trial (RCT) and in accordance with UK Medical Research Council Guidance for development of a complex intervention [10], we first explored the theoretical basis for an intervention in obese pregnant women [5, 6, 11, 12], leading to development of a novel intervention (Phase 1). We now report on Phase 2, an exploratory trial to determine whether this intervention achieved the changes in dietary and physical activity behaviours anticipated, and to undertake a process evaluation of every aspect of fidelity of the intervention and the protocol.

The protocol for the exploratory trial is shown in Figure 1.

Verbal and printed information was provided to potential participants at a routine antenatal appointment in the first trimester and women were contacted >24 hrs later to ascertain willingness to participate. For those declining participation, consent to record basic demographic data and BMI were obtained. Those willing to participate were invited to return for their first study appointment in the early second trimester (>15⁺⁰ weeks to <17⁺⁶ weeks’ gestation). This window of recruitment allowed adequate time for arrangement of the one to one session with the health trainer followed by the eight week intervention programme prior to the oral glucose tolerance test, carried out between 27⁺⁰ and 28⁺⁶ weeks’ gestation. Research midwives received a study-specific manual, attended at least one training session with the trial manager and continued feedback and training sessions for the study duration.

Inclusion criteria: BMI ≥30 kg/m² and singleton pregnancy; gestational age >15⁺⁰ weeks and <17⁺⁶ weeks’ gestation.

Exclusion Criteria: Unable or unwilling to give written informed consent; gestation <15⁺⁰ weeks and >17⁺⁶ weeks; pre-existing diabetes; pre-existing essential hypertension (treated); pre-existing renal disease; multiple pregnancy; systemic lupus erythematosus (SLE); antiphospholipid syndrome; sickle cell disease; thalassemia; celiac disease; currently prescribed metformin; thyroid disease or current psychosis.

All data were entered onto a password protected secure database (MedSciNet Ltd). Randomisation was performed online. The randomised treatment was allocated automatically, balanced by minimisation for maternal age, centre, ethnicity, parity and BMI. Data were analysed using Stata (version 11.2, StataCorp, College Station, Texas). All women randomised between 29th March 2010 and 13th May 2011 were included. Postcodes were matched to two national indices of deprivation: the Index of Multiple Deprivation (IMD) for English addresses, or the Scottish Index of Multiple Deprivation (SIMD) for addresses in Scotland [13, 14].

This study describes a pragmatic and rigorously evaluated pilot study of a complex intervention for diet and activity behaviour change in obese pregnant women. The intervention was associated with a significant change in dietary behaviour. Process evaluation showed overall acceptability of the protocol but led to several refinements to improve acceptability and fidelity.

In any lifestyle intervention of diet and physical activity, it is important that the pilot study design includes methods to assess the potential of the intervention to change these behaviours in the anticipated direction of effect. Few assessments of dietary intake in similar investigations of overweight or obese women have been attempted [37–39]. We used a 24 hr recall method to assess dietary intake and while this may lead to under-reporting of energy intakes, the reported values are not dissimilar to those of non-pregnant women in general UK population [40]. The objectives of the dietary intervention, to bring about reductions in GL and the proportion of energy derived from saturated fatty acids, were both achieved. This suggests that obese pregnant women are amenable to changing their diet in response to an intervention based on established theory, and that dietary advice, frequently delivered by health professionals, is likely to be successful in achieving dietary change in obese pregnant women, as previously implied [38]. The reduction in dietary GL achieved was similar (33% v 45%) to that reported in obese type 2 diabetic non-pregnant subjects in which improved glycaemic control was achieved [41]. Recently a similar intervention in 759 pregnant women, showed a lower change in GL (13%), which was associated with a reduction in gestational weight gain in women who had previously delivered a large for gestational age infant [42].

The reduction in energy intake observed is consistent with other studies that have restricted the intake of fat from meat and dairy products which have not been replaced by other sources of food energy [43]. The reduced GL may also have contributed through effects on satiety [44]. To our knowledge this is the first study demonstrating that anticipated changes in diet occur following delivery of an intervention to lower GL and saturated fat in obese pregnant women without GDM. Importantly this occurred despite the focus being on reducing GL by lowering the intake of added sugars as well as advocating foods with a lower GI. Focusing on GI tends to modify the GL from starch whereas the GI from sugar sweetened beverages is less amenable to change. Consequently, dietary advice to decrease the intake of added sugar, particularly as sugar-sweetened beverages, is likely to have had an important impact on GL.

This study adds to the scant literature on the habitual diet of obese pregnant women. The macronutrient profile at randomisation was similar to that of women in the general population, with fibre (non-starch polysaccharide) intake below, and total sugars and saturated fat above recommended UK guidelines [45]. The overall energy intake and macronutrient profile accords with one previous report in obese pregnancy [38]. Because of the time required to rigorously assess diet using the 24 hr recall method which, according to the process evaluation is likely to have influenced recruitment and compliance, a short food frequency questionnaire (5–10 minutes) was evaluated for use in the subsequent RCT.

The few studies that have attempted to measure changes in PA in intervention trials in pregnancy have generally relied on self-report, and results have been equivocal [38, 46–49]. Accelerometry, the standard method of objective assessment used previously in observational studies in pregnancy [48], has, to our knowledge, only been employed in one relevant RCT, the FitFor2 study, a supervised exercise intervention in 121 overweight and obese women [50]. Consistent with Fitfor2, we found no effect of the intervention on PA using the Actigraph accelerometer, concurring with the reported absence of change in barriers to PA. The failure of accelerometry to mirror the increase in self-reported walking in the intervention group could reflect insufficient intensity of this activity, but also reporting bias [51] which is common in the reporting of low intensity activities, such as those frequently undertaken by pregnant women [52].

As reported elsewhere, compliance with accelerometry in pregnancy was an issue [53, 54]. Nonetheless, 60% of obese pregnant women providing baseline accelerometry data met the current guidelines for PA in pregnancy (i.e. > 30 minutes of MVPA per day). A similar level of activity has been observed in pregnant women (all BMIs) [54] and overweight and obese non-pregnant adults [55], but not previously amongst obese pregnant women. Levels of PA were similar to those we found previously among overweight and obese women [53], but substantially higher than those reported for non-pregnant women in the UK [56]. There is no consensus on change of MVPA over pregnancy [53, 57–59]. In this study of obese women both groups reduced the level of objectively measured MVPA as pregnancy progressed.

This assessment has highlighted a critical need to evaluate PA behaviour objectively. We may otherwise have erroneously concluded in the following RCT that increased PA does not affect clinically relevant outcomes. Despite showing no increase in PA, we have not recommended that the RCT focuses on diet only [4], but rather that women continue to be encouraged to adhere to PA recommended in clinical guidelines.

Although there were no changes in attitudinal outcomes, women were generally positive about the recommended dietary and physical activity behaviours despite perceived barriers to change. Attitudinal data relating to diet were comparable to a population sample of pregnant women [12]. The intervention did not achieve any reduction in perceived barriers, but despite this, important dietary changes were achieved which may infer low levels of self efficacy. However, barriers to increasing PA appeared too great to overcome, possibly reflecting increased physical discomfort with gestation, as indicated by the EQ-5D questionnaire.

The relationship between mental health, diet and PA in obese pregnancies warrants further investigation in the RCT in view of the high prevalence of depressive symptoms (EPDS score > 12). Another report has also found no effect of a complex behavioural intervention in obese women on these symptoms [60].

In terms of context, the process evaluation recruited women in urban hospitals serving regions with areas of high socio-economic deprivation. Obesity rates are higher amongst women with lower socio-economic status, fewer qualifications [61] and amongst particular ethnic groups, particularly black African and black Caribbean [62]. It was important therefore not only to explore if recruitment was feasible but also whether the intervention was acceptable to the women recruited. Prominent media coverage about obesity raised the possibility that women in the control group might proactively address diet and physical activity, and some interview data supported this, but the evaluation suggested that awareness through the media alone was not adequate to achieve sufficient behavioural change.

In relation to reach, just over one third of eligible women agreed to participate. Similarly low recruitment rates are consistent with other intervention studies, particularly in populations with lower uptake of health care. In one previous relevant study of lifestyle advice in non-obese pregnant women, recruitment was slower than expected and low attendance at group exercise sessions and participant concern about burdensome data collection contributed to dropout [63]. However, perceived advantages to participation such as extra clinical tests and continuity of care from research midwives supported study uptake and continuation. Given the continued rise in obesity in the adult population in England [61], approximately 1:5 pregnant women would be eligible for inclusion. Recruitment of the numbers needed to be approached (4100) for the full trial is therefore unlikely to be affected by a shortage of eligible women. Overall, the wide social and ethnic diversity amongst participants was similar in participants and those who declined, indicating that the intervention would be unlikely to increase health inequalities by attracting more educated and higher income participants [64]. Importantly, although obese pregnant women, once recruited, were generally willing to attend group sessions, practicalities often interfered with regular attendance, thereby influencing dose. However, the sessions did not appeal to all women. Some appreciated finding common issues with other group members, others preferred one to one contact. Evidence for health improvement interventions in group settings is varied [65] and this study adds to the recognition that a ‘one size fits all’ approach may not be effective [64], and that flexibility is key to retention. Fidelity was good with consistently high level provision of SMART goals by HTs, which were viewed as a positive achievement, particularly since poor adherence to goal setting has been associated with moderate attendance amongst pregnant women [66]. The high acceptability of the participant handbook and pedometer re-enforced the theoretical approach [16], and women also responded well to motivational techniques, but physical issues presented barriers to PA. The information provided was valued, including increasing awareness of safe PA in pregnancy, and seen to have important educational benefit. Several components of the intervention therefore appeared beneficial and were well received by women. The intervention is relatively intensive and presents costs for providers, and whilst a full assessment of cost and benefit was not conducted in this pilot, steps taken during the pre-clinical development phase (using HTs rather than clinicians to deliver the intervention, adopting local group-based approach) helped keep the overall costs of the intervention low, recognising that if beneficial, it should also be affordable to health providers and to women. There was also suggestion that the intervention may extend to peers and family, and some women aspired towards better fitness following birth. This study has reinforced earlier reports suggesting that rapport between study staff and participants, interviews requiring short time commitment, and participants' perception of the study as informative are all important recruitment and retention factors [67]. Formal evaluation of the reasons for the high refusal rate was not permissible due to ethical constraints, but the time commitment was frequently commented upon by the recruitment staff, as well as lack of appreciation of the health consequences of obesity in pregnancy.

In summary, this study has emphasised the value of a pilot trial to assess anticipated behaviour change. Although seldom attempted by others, we have also highlighted the importance of process evaluation in a complex intervention of diet and physical activity for pregnant women. The pilot trial demonstrated reductions in glycaemic load and in the proportion of energy derived from saturated fat are achievable in obese pregnant women without GDM. The process evaluation identified that dietary advice and education were well received, and confirmed that PA change is more problematic to achieve, although it remains important to consistently measure and support PA using technologies acceptable to women. The process evaluation also helped explain issues arising in relation to uptake, dose, fidelity and retention which informed the feasibility of the full trial.

As a consequence of this study, several modifications to improve compliance and fidelity have now been implemented in the protocol for the main trial. As well as process evaluation, HT feedback highlighted potential barriers to fidelity of the intervention and informed protocol modifications for the RCT. Flexibility has been increased regarding the timing and delivery of the sessions, and goal setting can be undertaken by telephone or email. It is recommended that women should receive at least 5/8 sessions. The two extra visits required for objective assessment of PA and accurate evaluation of diet have been omitted in all but two sites (as planned) and dietary assessment reduced to a validated FFQ. The RPAQ includes domestic and childcare activities, considered appropriate for pregnant women [68], but following feedback has been replaced by the shorter and more relevant IPAQ [69]. Maintenance of dietary and PA behaviour change in the participant and her family is being formally evaluated at 6 months and 3 years postpartum. To minimise loss to follow-up and attendance at these appointments, strategies which have been put in place include regular newsletters and sending greetings cards on special occasions such as the child’s birthday.

Assumptions should not be made that interventions in obese pregnant women necessarily change behaviour. We recommend that a pilot trial such as that described here, which has demonstrated evidence for anticipated change in behaviour, is a necessary prelude to any RCT of a complex intervention of diet and physical activity designed to improve pregnancy outcome in obese women. Without prior evidence for change in behaviour in the anticipated direction, pursuit of a large and costly trial would be futile. Similarly, we have demonstrated the value of early process evaluation, which can lead to important refinements in protocol to improve feasibility and compliance in the definitive trial.