Introduction

Low birthweight (LBW: <2500 g), of infants small-for-gestational-age (SGA) or preterm, increases the risks of neonatal mortality [1, 2] and infant mortality [3], poor cognitive development [4], stunting [5], and lifelong susceptibility to non-communicable diseases [6, 7]. LBW remains a major public health problem in low-income settings, particularly in South Asia [8] and the WHO has set a target to reduce LBW by 30% by 2025 [9]. The urgency of addressing intergenerational undernutrition by intervening prenatally or during pregnancy to decrease SGA and stunting at birth is acknowledged [5, 8, 10], but our understanding of effective means of doing so is limited. Renewed WHO interest in low birth weight as an indicator for tracking country health [9] means use of birthweight as an outcome (as opposed to preterm and SGA) is of particular importance.

We reviewed evidence on three types of pregnancy-focused intervention that might improve birthweight: food supplements, cash transfers, and behaviour change communications. To review the effect of food supplements during pregnancy on child nutritional status we drew upon recent meta-analyses of trials of food supplementation, which all demonstrated a significant increase in birthweight [11–14]. Gresham 2014 found a raw mean difference (RMD) of 125 g from food supplementation or 94 g in studies limited to low or middle income countries (LMICs) and undernourished women [11], Imdad and Bhutta 2012 73 g [12], Ota 2012 41 g, (47, 77) [14] and Stevens 2015 found a standardised mean difference of 0.2 (0.03–0.38) (weight difference in grams not given) [13]. While updating the search of Stevens 2015, which focused on LMICs, we identified a Cambodian study of supplementation with a corn-soy blended food in pregnancy, which showed no differences in anthropometry [15]. Recent trials of Lipid-based Nutrient Supplements (LNS) in pregnancy have shown inconsistent results. Increases in birthweight of 85 g were demonstrated in Ghana [16], and of 41 g in Bangladesh [17], but no differences in anthropometry were shown in Malawi (either at birth or 18 months) [18].

Girard’s 2012 review of the effects of nutrition education and counselling (NEC) on birth weight showed an increase of 105 g (18, 193) overall and 152 g (-81, 384) in LMICs, but when disaggregated by whether the NEC was combined with food or micronutrient supplements, only the trials of NEC combined with supplements significantly increased birthweight whilst NEC alone did not: mean difference -8 g (-87, 70) for NEC alone and 0.1 g (0.0, 0.3) with NEC plus additional messages (16). Similarly, Ota’s review showed a significant birthweight increase of 490 g, (428, 552) from 2 trials of nutritional education alone with low quality evidence, and no increase in one trial with adequately nourished women 15 g (-76, 106) [14]. Quality of evidence around NEC and birth outcomes was generally low [19].

To review evidence on cash transfers during pregnancy on birthweight, we conducted searches in PubMed, Embase, The Cochrane Library, African Index Medicus, Web of Science and the Reproductive Health library using the inception date and December 2016 as inclusion dates, with no language restriction. For the effect of cash transfers on birthweight or child nutrition we used the search terms “cash transfer”, pregnant OR pregnancy, AND birthweight OR “child nutrition” respectively, without search limitations. Studies were included if they were original peer-reviewed articles or systematic reviews describing studies in which cash transfers were targeted to pregnant women in LMICs. Studies were excluded if they did not include either birthweight or child nutrition as one of the outcomes. Limited evidence of the effect of cash transfers on birthweight comes from large-scale programmes of cash transfers and income supplements in Mexico and Uruguay, which show increases in birthweight of the order of 127 g [20] and 31 g [21] respectively. In Mexico, Rivera (2004) found that children at young ages (0–6 months) in Progresa incentive-based intervention areas were 1.1 cm longer [22]. Leroy’s review of conditional cash transfers on child nutrition found a 1.53 cm length gain (0.41 Z-scores) but little impact upon micronutrient status [23]. We found no publications comparing effects of food supplements versus cash transfers on birthweight.

To review the effect of behaviour change interventions on birthweight we used the databases and inclusion/ exclusion criteria listed above, with the search terms “Child OR infant AND weight-for-age OR length-for-age OR nutritional status OR weight OR length OR stunting OR wasting OR underweight OR z score and pregnan* combined with either [women's group* OR participatory learning* OR community mobilization OR community participation] OR [nutrition education OR nutrition counseling OR “nutrition education and counseling” OR nutrition information]”.

We found no publications evaluating the effects on birthweight or child nutrition of Participatory Learning and Action (PLA) or community mobilization, apart from a study by Nair et al in India which found no significant effects on child stunting or birthweight but lower odds of underweight at 18 months of age (adjusted OR: 0.81 (0.66, 0.99)) [24, 25]. However we found that nutrition education interventions using participatory community-based approaches focused on practical behaviour were more likely to be effective than knowledge transfer alone [26]. Moreover, PLA with women’s groups was effective in reducing other health outcomes, including neonatal and maternal mortality in rural communities [27], and was pro-poor in coverage [28] and impact [29], and may influence nutrition-related behaviours [30] and dietary diversity [31].

We hypothesised that an improved behaviour change strategy using Participatory Learning and Action (PLA) women’s groups combined with transfers of food or cash to pregnant women would lead to increases in birthweight and child weight-for-age z-score in the first 16 months. We did a cluster-randomised controlled trial to compare with usual government programmes the effects on birthweight and child nutritional status of three interventions—PLA alone, PLA plus food transfers, and PLA plus cash transfers. We also assessed improvements in maternal nutritional status, eating behaviour in pregnancy, feeding behaviours, and institutional delivery.

Materials and methods

Results

All dates of recruitment, follow-up and intervention implementation are provided in Fig 1 and the trial profile provides full details of participant flow (Fig 1). From 63,308 women who consented to menstrual monitoring, we recruited 25,092 pregnancies into interventions and surveillance, and after exclusions 10,936 permanently resident women who delivered between 4 June 2014 and 20 June 2015 were eligible for (WAZ) outcome analyses (control: n = 2426; PLA alone n = 2448; PLA plus cash n = 3065; PLA plus food n = 2997).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Participant flow.

https://doi.org/10.1371/journal.pone.0194064.g001

The study timeline describes the trial periods and activities (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Trial timeline.

https://doi.org/10.1371/journal.pone.0194064.g002

Table 1 shows baseline characteristics of women eligible for analysis in the four study arms. There were no substantial differences between arms, except in the proportion of women whose husbands were working overseas, which was higher in control (31%) than in intervention arms (22–26%). Many women were aged 20–24 (38.8%) or were adolescent (37.4% were 15–19 and 0.5% 12 to 14 years), few were between 30 and 46 years (5.5%) and around one-third were primigravid. Most had no schooling (64%) and only 15% had any secondary education. Disadvantaged Dalit castes and Muslims accounted for one-third and ‘middle’ Terai caste groups accounted for 43%. Yadav or Brahmin (least poor) castes were more prevalent in the PLA plus cash arm (28%) than in the control arm (19%). Most households (74%) had no (or shared) toilet and used biomass fuels for cooking (95%).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of trial participants eligible in weight-for-age z-score analysis.

https://doi.org/10.1371/journal.pone.0194064.t001

Outcome measures were available in all clusters. Capture rates for data and outcomes are described in S2 Table. Birthweight was analysed for 2087 children measured within 72 hours (control: n = 464; PLA alone n = 488; PLA plus cash n = 509; PLA plus food n = 626) and WAZ for 9242 children aged 0–16 months, average age 9 months (control: n = 2091; PLA alone n = 2095; PLA plus cash n = 2549; PLA plus food n = 2507). Capture rates for birthweight were low during the main surveillance period due to unexpectedly high numbers of women enrolling, especially in transfer arms, and conflict affecting field team performance. Capture rates were balanced across arms: 22% of potential birthweights (range 19–24% per arm) and 84% of potential WAZ (range 83–86%).

PLA and food or cash transfer interventions were implemented via 539 groups in all 60 intervention clusters as planned. Coverage of PLA groups and group activities amongst women eligible for WAZ analyses is summarised in Table 2. Around 20 women attended each PLA group, with a higher proportion of pregnant women attending (usually accompanied by a female relative) in transfer arms. Attendance at PLA groups was high amongst enrolled women, especially in food and cash transfer areas where 99% of cash and 97% of food arm families had heard of groups and 98% of cash and 96% of food families had a member who attended. These proportions were significantly higher (p<0.0001) than those in the PLA only arm where 65% had heard of groups and 55% had attended or had a household member who attended, and 49% of pregnant women had themselves attended. Women in the PLA only arm attended only twice on average during pregnancy, compared with four times in PLA plus transfer arms. This may be associated with differential gestational age at enrolment. Average gestational age in months at enrolment was higher in PLA plus cash (mean 3.7, SD 1.4) and PLA plus food (3.7, 1.5) arms than in control (5.4, 2.1) or PLA only (5.2, 2.1) arms. This means that women in food or cash arms had on average 1.5 months longer to be exposed to groups than in the PLA only arm.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Intervention exposure across study arms in cases eligible for weight-for-age z-score analysis.

https://doi.org/10.1371/journal.pone.0194064.t002

Home visits reached 17% of women in the PLA arm and 36–37% in PLA plus transfer arms. Most home visits were to invite women to attend the group or deliver transfers to women unable to attend, but some were delivered as a group strategy. Coverage of strategies implemented by PLA groups to raise community awareness about nutrition and low birth weight was higher in PLA plus transfer arms (>91%) than in the PLA only arm (50%). Exposure to community strategies was highest for viewing picture cards and attending extra meetings (targeted especially for pregnant women or mothers-in-law) and lowest for rallies and video showing. Few husbands or other men attended the groups.

In total 28,527 cash transfers were delivered to 6,651 pregnant women and 25,679 food transfers were delivered to 6,836 pregnant women. For trial participants (Table 2), 32% of women in food and 30% in cash arms received 6 or more transfers over pregnancy, 46% in food and 47% in cash arms received 4–5 transfers, and 21% in food and 23% in cash received 0–3 transfers. In a qualitative study in the PLA plus cash arm, women reported spending their cash on milk and fruit in particular [40]. In the PLA plus food arm, two-thirds of women used the Super Cereal in a sweet snack called haluwa, prepared by adding oil or ghee, sugar, and water or milk. Around one-third prepared it as roti (flatbread), made with water and very little or no oil. In a subsample of 768 interviewed in late pregnancy, women ate Super Cereal 1·4 times per day and 5 days per week on average, and around 43% reported sharing it with other family members. The food was thought of as ‘medicine’ for the woman and her baby. Contamination of control areas was minimal. Only 5% of control group respondents had heard of women’s groups facilitated by the implementing NGO and 1% had attended groups or observed group strategies that involved community awareness-raising.

S3 Table provides comparisons of characteristics of women whose infants’ birthweights were and were not captured. Compared with those whose infants were not measured, mothers of infants whose weights were captured were significantly older, had more children, were more likely to be Hindu than Muslim, and were less likely to have primary or secondary education. Many younger, first time mothers, who were more likely than older women to have some education, were not at the home where they had enrolled in the study at the time of the birth. This may be explained by the custom of younger pregnant women and girls (especially Muslims) with no or few previous children returning to their parental home for delivery, while mothers with two or more children stay with their marital families.

Infant anthropometric outcomes are presented in Table 3. Relative to the control arm (mean 2756 g, n = 464) birthweight measured within 72 hours (the birthweight primary outcome) appeared to be incrementally higher in PLA (28·9 g; 95% CI -37·7, 95·4; n = 488), PLA plus cash (50·5 g; 95% CI -15·0, 116·1; n = 509), and PLA plus food (78·0 g; 95% CI 13·9, 142·0; n = 629) arms, but the only significant difference was between control and PLA plus food. The difference of 49 g (95% CI -13·5, 111·8) between PLA only and PLA plus food was not significant. LBW prevalence was lower in the PLA plus food arm (19.1% OR 0.74, 95% CI 0.51,1.08) than in the control arm (22.5%), representing a non-significant 16% decrease associated with the intervention.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Effect of interventions on birth weight and infant anthropometry in first 42 days.

https://doi.org/10.1371/journal.pone.0194064.t003

When birthweights measured within 10 days were analysed (including those measured within 72 hours) as a secondary outcome, differences from control were not significant for PLA alone (36.1 g; 95% CI -30.8, 103.1) but were significant for PLA plus cash (68.8 g; 95% CI 3·2, 134·4) and PLA plus food (72.4 g; 95% CI 7·5, 137·2). Differences between these and the PLA only arm (32–36 g) were not significant. The effect of food arm on infant size was sustained when weight-for-age (WAZ) scores from 0 to 42 days were analysed. Infants in the food arm had mean WAZ scores 0.154 (95% CI 0.025, 0.283) greater than control but there was no difference between control and other arms PLA only (0.065 g; 95% CI -0.068, 0.198) and PLA plus cash (0.154g; 95% CI -0.025, 0.233). Amongst these 57% of weights were measured within 72 hours, 14% from 72 hours to 10 days, and 29% from 10 to 42 days.

No significant differences were seen in infant head circumference or infant length within 10 days. Length-for-age z-score (LAZ) measures within 10 days suggest that stunting (<-2 length for age z score) prevalence was lower than control in PLA plus cash and PLA plus food arms (18.9%, 14.8%, 15.6% respectively) but formal analyses were not undertaken following the analysis plan.

Table 4 shows maternal eating behaviour and maternal and newborn morbidity by study arm. Although response rates were lower than the 40% cut-off in our analysis plan, at the request of reviewers of this article, we tentatively analysed a small number outcomes only as shown. The findings suggest that higher proportions of pregnant women in intervention arms had an adequate number of food groups in the previous 24 hours compared with control, especially in the PLA plus cash arm (54% versus 39% in control), but this association did not reach significance. The possible difference may be attributed to increased consumption of dairy foods (64% versus 53% in control). Women in the cash arm consumed a significantly higher number of meals per day during pregnancy. Mother’s weight in late pregnancy was somewhat higher in all intervention arms compared with control.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Outcomes for maternal anthropometry, diet in pregnancy, maternal morbidity in pregnancy and post-partum and neonatal morbidity, for which formal analyses are not conducted due to low response.

https://doi.org/10.1371/journal.pone.0194064.t004

Differences in neonatal morbidity using one or more of seven danger signs for neonatal sepsis [41] were tested but showed no clear differences. Across all arms, mothers of 38–45% of newborns reported one or more of these danger signs and 12–17% recalled signs of acute respiratory infection with fever, which illustrates the high levels of morbidity in the population.

S4 Table shows birth outcomes and maternal deaths for the trial. Anthropometric outcomes collected at endpoint for child and mother are presented in Table 5. Table 6 shows breastfeeding, infant and young child feeding, morbidity, gestational age and institutional delivery outcomes. Analysis of the endpoint WAZ primary outcome indicated little difference between arms. Relative to the control arm (n = 2091) the mean adjusted differences in WAZ after imputation in the PLA alone, PLA plus cash and PLA plus food arms were -0·026 (95% CI -0·117, 0·065; n = 2095), -0·045 (95% CI -0·133, 0·044; n = 2549) and -0·033 (95% CI -0·121, 0·056; n = 2507), respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Analysis of anthropometric outcomes captured at endpoint by study arm.

https://doi.org/10.1371/journal.pone.0194064.t005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Analysis of child feeding, morbidity and delivery-related outcomes captured at endpoint by study arm.

https://doi.org/10.1371/journal.pone.0194064.t006

There was no effect of pregnancy interventions on maternal anthropometry 0–16 months after birth (BMI, underweight, MUAC) or on child anthropometry (WAZ, LAZ, WLZ, head circumference), preterm birth or gestational age at delivery, infant feeding behaviours, including continued breastfeeding beyond 6 months, child dietary diversity score, consumption of green leafy vegetables and fruits, flesh foods, or child morbidity. However, significantly more institutional deliveries (OR 1·46; 95% CI 1·03, 2·06; n = 2651) and lower rates of colostrum discarding (OR 0·71, 95% CI 0·54, 0·93; n = 2548) were observed in the PLA plus food arm compared with control (institutional delivery n = 2251; colostrum n = 2087).

Comparing average WAZ between time periods after birth (-1·11 to -1·29 z-scores within 42 days and -1·60 to -1·68 z-scores at age 9 months on average), children became more underweight between early infant anthropometry and endpoint in all study arms. Prevalence of stunting increased between 42 days (17·4%-21·4% across study arms) and 9 months on average (29·1%-30·7% across study arms).

Dose-response analyses (reported with imputation) in Table 7 show that, in the PLA only arm, attending more groups was linked to higher WAZ at 9 months age on average, whereas in the PLA plus cash arm more transfers were associated with lower WAZ. More cash transfers were also associated with lower LAZ. There was no significant dose-response effect in the food arm. S5 Table shows no association between socioeconomic status and the effect of any of the three interventions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Dose-response analyses of the effect of pregnancy intervention intensity on weight-for-age and length-for-age z-scores measured at endpoint (0 to 16 months, average age 9 months), with and without multiple imputation of covariates.

https://doi.org/10.1371/journal.pone.0194064.t007

Discussion

Our trial of the effects of PLA women’s groups, with and without transfers of cash or food to pregnant women, suggests an increase in birthweight in the PLA plus food arm. However, better nutritional status soon after birth did not persist into early childhood, reflected by only small differences between arms in child anthropometry at 0–16 months. No intervention effects were seen for many secondary outcomes and the number of secondary outcomes tested prevents firm conclusions, but in the PLA plus food arm the prevalence of colostrum discarding was lower and the number of institutional deliveries was higher. We conclude that adding a fortified balanced energy protein supplement to a Participatory Learning and Action behaviour change intervention is likely to increase birthweight, and may be more effective than providing a modest amount of cash with PLA or conducting PLA alone. Sustaining the improvement in birthweight over infancy may require intervention beyond pregnancy, including nutritional support to lactating women and age-appropriate complementary feeding and consideration of other environmental challenges.

The study limitations relate largely to the compromised capture of birthweights during ethnic conflict in the data collection team, and consequent lack of robustness to our birthweight findings. Food and cash distribution were delayed during monsoons due to road access problems, and cash transfers could not be delivered in October 2014 because of security concerns.

As there are strong geographical and cultural similarities, our findings are likely to be generalizable within the plains of Nepal and to other South Asian populations living in plains areas with predominant rice production in India, Bangladesh, and Pakistan. Strong cultural differences between hill and plain Nepalese communities mean that they may be less generalizable to hilly areas of Nepal with greater food insecurity or where women’s freedom to move beyond their home environment is less restricted.

The effect of Super Cereal (fortified sweetened wheat-soy blended flour) supplement on birthweight in the context of a PLA intervention, and the lack of appreciable effect of PLA alone, is consistent with several reviews that showed increases from protein-energy supplements [11–14]. In contrast with our findings, pregnancy supplementation with a World Food Programme Super Cereal-like blend of corn rather than wheat in Cambodia showed no effect on birthweight or mother’s weight gain in pregnancy, although there was a trend for higher impact among women with a lower BMI as compared to women with a higher BMI in early pregnancy, and a decrease was found in preterm delivery and anaemia [15].

Although PLA alone has been effective in reducing neonatal mortality [27], and in changing nutrition-related behaviours [25, 30, 31, 42], effects of PLA alone on birthweight were not large enough to reach significance in our study. This is consistent with the lack of effect of PLA plus home visits on birthweight found by Nair et al [25]. Whilst PLA differs from conventional nutrition education (NE) due to strategy implementation mobilising communities, the lack of impact on birthweight of PLA alone is somewhat similar to Girard’s review, which showed substantial increases in birthweight after NE with nutrition support in the form of food or micronutrient supplements, but no increase after NE alone [19]. A review of fifteen studies of nutrition counselling alone also showed varying effects [11].

The lack of sustained difference in WAZ appears to contrast with the follow-up of an antenatal multiple micronutrient supplementation trial in the same districts of Nepal [43], which showed a 205 g difference between intervention and control children at two years, having demonstrated a 77 g difference at birth [44]. The population in the multiple micronutrient study was more urbanised and generally less poor than our cohort, and the difference had disappeared by nine years of age [45]. Our endpoint data collection occurred during monsoon when levels of diarrhoeal disease and child underweight were high.

Other studies have also found that the effects of pregnancy food supplementation were inconsistent in childhood. For example, the MINIMat study in Bangladesh compared the effects of early and late pregnancy supplementation with food and micronutrients on child growth in the first 54 months. Compared with late pregnancy supplementation, early supplementation was associated with reduced stunting in boys, but not girls. However, as in our study, no effect was seen on weight-for-age or weight-for-height z-scores [46]. In Malawi continued LNS supplementation to women during pregnancy and the first 6 months postpartum and to their children thereafter had no impact on child anthropometry at 18 months [47].

As feeding practices are similar to those found by Patel in northern India [48], we hypothesise that the disappearance of anthropometric differences detected at birth in our study may be associated with late introduction and insufficient quality or quantity of complementary foods. A high burden of infection, especially environmental enteropathy [49], associated with poor sanitation and low coverage of toilets in the area might also explain the increase in undernutrition we found between birth and 9 months on average [50]. Heady and Hoddinott estimate that reduction in stunting over time in younger children (0–12 months) in Nepal is 18% attributable to improvements in sanitation, 29% to health care, 25% to increased wealth and 19% to maternal education. They hypothesise that most of the improvements in nutritional status in Nepalese children between 2001 and 2011 were driven by increases in birth size and less by improved postnatal nutrition trajectories [51]. Whilst our findings support their finding of poor postnatal growth, the association between birth size and pregnancy interventions had disappeared nine months later.

Increases in birthweight within 72 hours in the cash arm of our study were not large enough to reach significance but evidence from South American settings of large-scale cash transfer programmes showed increases in birthweight of 31g in Uruguay through the PANES programme [21] and of 127g in Mexico through the Oportunidades programme [20] as well as improvements in child length in the first 6 months, though not necessarily later in childhood [22, 23].

Our ‘dose response’ analysis raises some concern that children born to mothers who received more cash transfers in pregnancy were slightly smaller at endpoint. We suggest that this result is likely to arise from residual confounding whereby women who received more transfers differed in some way not captured in our model from those who received fewer.

The likely effect of the PLA plus food arm on institutional delivery and breastfeeding is interesting as the changes were unlikely to be associated with the food supplement itself. Enhanced participation in women’s groups brought about by provision of food transfers may have stimulated improved uptake of institutional delivery. It might also be the case that certain food arm clusters had higher institutional delivery rates prior to the intervention, though exploratory analyses did not show any significant difference between arms in pre-intervention and run-in periods of the study. During the main trial period, three clusters in the food arm had institutional delivery rates above 73%, which was not found elsewhere. It could be that the provision of care was better in these areas than elsewhere, though we do not have data on this. Any explanation of this finding can only be speculative.

The effects on colostrum feeding and timely initiation of breastfeeding may have been mediated by greater attendance at women’s groups in this arm, or by the increase in institutional deliveries because these practices would have been encouraged in health facilities. However, a similar trial of women’s groups practising PLA in Bangladesh showed a shorter time to initiation of breastfeeding without any associated increase in facility deliveries [52], as well as increased exclusive breastfeeding [30]. It may be that breastfeeding behaviour was also susceptible to change amongst those who delivered at home in our study population but we have not tested this.

The increased effect on birthweight of adding transfers of food or cash to a PLA women’s group intervention suggests that providing women with extra resources during pregnancy increases their access to and intake of nutritious food. It may be that the PLA was more effective in the cash and food arms because the transfers incentivised group attendance and pregnant women and their mothers-in-law were thereby motivated to improve the diets of pregnant women.

Our study raises questions as to why supplementation with fortified wheat-soy blended food was apparently more effective in increasing birthweight than cash transfers. Perhaps women adhered to consumption of Super Cereal better than they adhered to spending their cash upon similarly energy-dense food? However a qualitative evaluation indicated that decision-makers in the household were often persuaded to spend cash transfers on pregnant women’s food intake [40]. In addition, some families saved their cash transfers to comply with PLA recommendations for women to attend antenatal care and seek institutional delivery in order to secure access to private providers [40]. Super Cereal did not have resale value on the local market and was not a substitute for regular meals: it differed from usual staples, was sweetened, and was not preferred. Cash could be spent as the pregnant woman chose or used for regular household expenses, though many reported spending their transfers on increasing milk intake [40]. The food ration was sufficient to allow for considerable household sharing, whilst the cash transfer was sufficient to buy 0.5 litres of milk daily, which, if shared, may not have increased intake as much as the food supplement.

A sub-study on pregnant women’s diets found that, compared with the control arm, women in the PLA plus cash arm consumed more micronutrient-rich foods, especially dairy, whereas women in the PLA plus food arm may have been allocated slightly more energy [53]. This suggests that the food and cash were affecting dietary patterns in different ways. This sub-study was conducted towards the end of the trial, on a sub-sample of pregnant women in their third trimester, when respondents were eating relatively small amounts of super cereal per day. It seems that regular consumption of fortified balanced protein-energy super cereal (during the earlier time of the study when birthweight was being measured) was more effective at improving birthweight than regular consumption of extra milk, as occurred in the cash arm. In this context, gender disparity in earned cash income is a stronger predictor of intra-household calorie allocation than household food security, whereas household food security is a stronger predictor of micronutrient allocation [54]. Yet, intra-household allocation of dairy was more equitable in the PLA plus cash arm than the control, and intra-household allocation of energy was more equitable in the PLA plus food arm, suggesting that the additional dairy purchased with the cash and the additional energy from Super Cereal in these transfer arms were channelled to the pregnant women (51). A sub-study on the impact of trial interventions upon women’s agency found that the impact on birthweight was not mediated by an increase in women’s empowerment. The PLA intervention increased agency in the study population by enabling newly married women to attend group meetings in the community, a privilege normally reserved for their women 10–20 years older than them. However, we found no impact on agency in domains other than group participation including health behaviours [55]. It could be that PLA increased awareness about maternal health and nutrition amongst decision-makers in households and that this led to increased antenatal care seeking and better hygiene. This in turn may have improved pregnant women’s health and eating behaviours, but our response rates were too low to draw conclusions on this.

Future studies should compare the effect of cash and nutritious food transfers on birthweight and on longer-term growth and survival, at least to two years. Ideally they should test the effect of nutritional interventions throughout the ‘thousand days’ period from conception, through pregnancy, exclusive breastfeeding and complementary feeding to 2 years of age, starting at the pre-conception stage with newly weds or couples planning a family. The apparent graded effect on birthweight through PLA alone, PLA plus cash, and PLA plus food also suggests that further studies using PLA for nutrition behaviour change are warranted.

Compliance with our interventions in the study area was high, especially in the PLA + food and PLA + cash arms. Despite the limitations of our birthweight data, adding food supplements to a PLA intervention in pregnancy increased birthweight significantly. Increases in weight or length at birth were not sustained through the first 16 months of life, which suggests that pregnancy interventions alone are unlikely to have lasting impacts on child nutritional status in these settings. Investment in fortified food supplementation combined with PLA groups for pregnant women may increase birthweight. Although continued support for breastfeeding, complementary feeding, improved water, sanitation and hygiene and infection control is also essential to break the intergenerational cycle of undernutrition in South Asia, an improvement in birthweight is an important achievement in its own right, since birthweight is an indicator of fetal development and a predictor of chronic disease in later life (2, 4, 5).

Conclusions

This study is the first to use women’s groups practicing Participatory Learning and Action (PLA) to influence birthweight as the primary outcome in a cluster-randomised controlled trial, and to compare the impact on birthweight of nutritious food and cash transfers combined with PLA. It is also the first to assess the effect of these pregnancy interventions on child nutritional status after birth and the first to assess impact of cash transfer using an RCT design. We confirm others’ findings that provision of food, a fortified balanced energy protein supplement in our case, in pregnancy can increase birthweight, but evidence for the impact of cash transfers (7.5 USD/mo) is less clear. PLA alone did not increase birthweight substantially and did not attain significance.

Provision of nutritious food supplements to pregnant women can increase the weight of babies at birth and may be more effective than provision of a modest amount of cash to buy additional food in pregnancy. However, in environments with very poor sanitation and hygiene, where infant and young child feeding behaviours are inadequate, gains in weight resulting from pregnancy interventions are not sustained, which suggests that nutrition interventions in pregnancy may need to continue beyond pregnancy until at least 2 years of age. Further studies comparing the effects of food and cash and assessing the impact of PLA on birthweight and child nutrition are needed, as are studies evaluating the long-term impacts on health of increases in birthweight.

Supporting information

Acknowledgments

Vikas Paudel passed away before the submission of the final version of this manuscript. Naomi Saville, corresponding author, accepts responsibility for the integrity and validity of the data collected and analyzed.

Thank you to the District Public Health Offices of Dhanusha and Mahottari and their health workers at all levels who supported the trial on the ground. We thank: the Trial Steering Committee: Prof Keith West (chair), Prof Simon Cousens (statistical adviser), Dr Elaine Ferguson, Prof Sonia Bhalotra, and Dr Shobha Rao; the Data Monitoring Committee: Dr Rajiv Bahl, Dr Andrea Rehman, Dr Lynette Neufeld, Dr Amod Poudyal and Dr Rita Thapa; and advisors from the Donor and Government Advisory Group in Nepal, including the Family Health Division, Child Health Division and Nutrition Section of the Nepal Ministry of Health.

We thank MIRA team members for their efforts in implementation of the study: Shiva Shanker Chaube, Neha Sharma, Ritesh Shrestha and all the field team. Lalbabu Kushuwa and Dhruba Devkota of Save the Children helped us to implement the food and cash transfers. Pramila Ghimire, Srijana Nakermi, Marco Cavalcante, and Allison Prather of World Food Programme Nepal provided Super Cereal for the study and advice on managing its distribution. We are grateful to administrators in UCL Institute for Global Health, Save the Children UK, and MIRA for administrative and managerial support.

We thank all the committed staff of MIRA who made the study possible, including more than 66 data collectors and their 16 supervisors, 720 volunteer enumerators, 40 intervention implementers, 539 volunteer mobilisers and 540 Female Community Health Volunteers responsible for facilitating women’s group and social transfer interventions, and the thousands of women who agreed to participate.