Introduction. The Journal of Nutrition Community and International Nutrition

Transcription

1 The Journal of Nutrition Community and International Nutrition A 2-Year Integrated Agriculture and Nutrition Program Targeted to Mothers of Young Children in Burkina Faso Reduces Underweight among Mothers and Increases Their Empowerment: A Cluster-Randomized Controlled Trial 1 3 Deanna K Olney, 4 * Lilia Bliznashka, 4 Abdoulaye Pedehombga, 5 Andrew Dillon, 6 Marie T Ruel, 4 and Jessica Heckert 4 4 Poverty Health and Nutrition Division, International Food Policy Research Institute, Washington, DC; 5 Helen Keller International, Ouagadougou, Burkina Faso; and 6 Michigan State University, East Lansing, MI Abstract Background: Recent evidence demonstrates the benefits of integrated agriculture and nutrition programs for childrenõs health and nutrition outcomes. These programs may also improve mothersõ nutrition and empowerment outcomes. However, evidence from rigorous evaluations is scarce. Objective: We examined impacts of Helen Keller InternationalÕs 2-y enhanced-homestead food production (E-HFP) program in Burkina Faso on the secondary impact measures of mothersõ nutrition and empowerment. Methods: We used a cluster-randomized controlled trial whereby 55 villages with 1767 mothers of young children were randomly assigned to 3 groups: 1) control, 2) E-HFP with the behavior change communication (BCC) strategy implemented by older women leaders, or 3) E-HFP with BCC implemented by health committee members. Data for the treatment groups were pooled for this analysis because no differences were found between the 2 groups in key mothersõ outcomes. We used difference-in-differences (DID) estimates to assess impacts on mothersõ dietary intake, diversity, body mass index (BMI; in kg/m 2 ), prevalence of underweight (BMI <18.5), and empowerment. Results: The E-HFP program significantly increased mothersõ intake of fruit (DID = 15.8 percentage points; P = 0.02) and marginally increased their intake of meat/poultry (DID = 7.5 percentage points; P = 0.08) and dietary diversity (DID = 0.3 points; P = 0.08). The prevalence of underweight was significantly reduced among mothers in treatment compared with control villages by 8.7 percentage points (P < 0.01). Although the changes in BMI did not differ between mothers in treatment and control villages, there was a marginally significant interaction (baseline underweight 3 change in BMI; P-interaction = 0.07), indicating that underweight mothers had a greater increase in BMI than did mothers who were not underweight. The E-HFP program also positively affected mothersõ overall empowerment score (DID = 3.13 points out of 37 possible points; P < 0.01) and 3 components of empowerment: meeting with women (DID = 1.21 points out of 5 possible points; P < 0.01), purchasing decisions (DID = 0.86 points out of 8 possible points; P = 0.01), and health care decisions (DID = 0.24 points out of 2 possible points; P = 0.05). Conclusions: Helen Keller InternationalÕs E-HFP program in Burkina Faso substantially improved mothersõ nutrition and empowerment outcomes. These positive impacts benefit the mothers themselves and may also improve their ability to care for their children. This trial was registered at clinicaltrials.gov as NCT J Nutr 2016;146: Keywords: empowerment womenõs underweight, dietary diversity, integrated agriculture-nutrition program, Africa, womenõs Introduction Improving the health and nutritional status of mothers and empowering them is critically important to ensure their own well-being and may also have positive impacts on their childrenõs health and nutritional status (1, 2). According to the 2010 Demographic and Health Survey of Burkina Faso, the prevalence of underweight [BMI (in kg/m 2 ) <18.5] among women is high at 16%, with the highest prevalence found among the poorest (24%) and those living in the east (31%) (3). WomenÕs underweight is also associated with illness and inadequate diets, in part due to seasonal cereal shortages (3, 4). ã 2016 American Society for Nutrition. Manuscript received September 17, Initial review completed December 9, Revision accepted February 22, First published online April 13, 2016; doi: /jn

2 Integrated agriculture and nutrition programs targeted to mothers of young children are one way to simultaneously improve the nutritional status of mothers and empower them (5). These positive impacts may be achieved through the provision of agriculture-related inputs and assets and training in optimal agriculture, health, nutrition, and hygiene practices and encouragement for their adoption. In addition, these types of programs often provide opportunities for mothers to interact, exchange knowledge, and support each other in the adoption of new practices. Program inputs and opportunities can therefore translate into increased access to and use of nutrient-rich foods, greater access to income, and increased participation in household decision making, facilitating the adoption of recommended agriculture, health, nutrition, and hygiene practices, all of which may positively empower mothers and improve their nutritional status. Despite the potential of integrated agriculture and nutrition programs to benefit women, supporting evidence from rigorous program evaluations is scarce. To date, evaluations of these programs have primarily assessed changes in dietary diversity (6, 7), whereas few have evaluated changes in womenõs nutritional status (8 11) or components of womenõs empowerment (11 13), and only 2 have examined both (9, 11, 12). These evaluations have consistently demonstrated positive impacts on household and womenõs dietary diversity (6, 7). However, only 2 have found a positive impact on reducing the prevalence of anemia (9, 11), and only 1 demonstrated a positive impact on increasing womenõs BMI (11). All 3 of the studies that measured components of womenõs empowerment demonstrated positive impacts in at least some components [income (13), food and household expenditure per adult equivalent (11), or participation in household decision making and perceived economic contribution to the household (12)]. However, the study by Brun et al. (13) did not have a control group, and that by Kumar and Quisumbing (11) found that the positive impacts in the short term among early adopters of improved vegetables or group fish ponds compared with late adopters were not sustained over the long term. Early adopters in the individual fish pond group, on the other hand, did significantly increase their food and household consumption per adult equivalent over the long term (11). Recently, an evaluation of Helen Keller InternationalÕs(HKIÕs) 7 enhanced-homestead food production (E-HFP) program in Burkina Faso, with the use of a longitudinal cluster-randomized controlled trial, showed positive impacts on reducing childrenõs anemia, wasting, and diarrhea. These findings were consistent with the programõs impacts on increasing mothersõ production of nutrient-rich foods, their ownership of agriculture assets and 1 Supported by the United States Agency for International Development, Office of US Foreign Disaster Assistance through Helen Keller International; European Commission/International Fund for Agricultural Development; the Gender, Agriculture and Assets Project, supported by the Bill & Melinda Gates Foundation; the CGIAR Research Program on Agriculture for Nutrition and Health led by the International Food Policy Research Institute; and the USDA (AD). 2 Author disclosures: DK Olney, L Bliznashka, A Dillon, MT Ruel, and J Heckert, no conflicts of interest. A Pedehombga works for Helen Keller International. 3 Supplemental Figure 1 and Supplemental Table 1 are available from the Online Supporting Material link in the online posting of the article and from the same link in the online table of contents at 7 Abbreviations used: BCC, behavior change communication; DID, difference-indifferences; E-HFP, enhanced-homestead food production program; ES, effect size; HC, health committee; HKI, Helen Keller International; IYCF, infant and young child feeding; OWL, older women leader. *To whom correspondence should be addressed. d.olney@cgiar.org Olney et al. small animals, and their nutrition- and health-related knowledge (14, 15). In this article, we use data collected in that same trial to estimate the secondary program impacts on 3 important sets of outcomes for mothers: dietary intake of individual food groups and diversity, BMI and prevalence of underweight, and womenõs empowerment. To our knowledge, this is the first clusterrandomized evaluation of a homestead food production program that provides rigorous evidence of the impact on these indicators in the same cohort of women. Methods Program description. HKIÕs E-HFP program, implemented in the province of Gourma in eastern Burkina Faso, is an integrated agriculture and nutrition program that aimed to improve mothersõ nutritional outcomes and empowerment through a set of agricultural production and nutrition interventions targeted to mothers with children aged 3 12 mo at baseline. The agricultural production interventions centered on dedicating land to womenõs production during the secondary agricultural season and distributing inputs and training to increase production and consumption of nutrient-rich foods and to generate additional income (and control over that income) through the sale of surplus production. To increase fruit and vegetable production, program beneficiaries were provided with saplings, cuttings, and seeds of nutrient-rich fruits (e.g., mangoes and papayas) and vegetables (e.g., orange-flesh sweet potatoes, dark green leafy vegetables, and carrots) and some small gardening tools (e.g., hoes, shovels, and watering cans). Beneficiaries were also given chicks to increase production of animal source foods (i.e., eggs and meat from the chickens). In addition, program beneficiaries received training in optimal agriculture and poultry-raising practices to help them establish their homestead food production activities. Agriculture inputs and training were first provided to 4 female village farm leaders, who were program beneficiaries themselves, at community demonstration gardens (village model farms). These beneficiaries then provided similar training and inputs with the support of project staff to the other beneficiary mothers in their communities to enable them to establish their individual homestead food and poultry production activities. The nutrition intervention used a behavior change communication (BCC) strategy known as the Essential Nutrition Actions framework, which focuses on 7 primary health and nutrition behaviors. Twice a month, all beneficiary mothers were visited in groups or individually by 1 of 2 different types of community volunteers trained by the program, either older women leaders (OWLs) or health committee (HC) members. This component aimed to improve mothersõ own health and nutrition by enabling them to adopt optimal health and nutrition practices for themselves and their young children. For further details, see Olney et al. (15). Program impact pathways. The E-HFP program had the potential to improve mothersõ nutritional status through 3 primary program impact pathways: 1) increased production of micronutrient-rich foods to increase consumption of these foods, 2) increased production of micronutrientrich foods to increase income, and 3) increased knowledge related to maternal and child nutrition and health to increase adoption of optimal practices. The program also aimed to increase mothersõ ownership and control over assets (i.e., small agriculture assets, poultry, and garden produce) and their human capital (i.e., increased knowledge of optimal agriculture, health, and nutrition practices). The E-HFP program sought to directly increase mothersõ ownership of assets by direct transfers of seeds, saplings, small gardening tools, and chicks. An additional program approach to improving womenõs empowerment consisted of inviting beneficiary mothers to participate in mothersõ groups formed for program activities such as working at the village model farms and for the transfer of agricultural, health, nutrition, and hygiene-related knowledge through BCC. Taken together, the asset transfers and participation in program activities were expected to increase the ability of beneficiary mothers to contribute to their own well-being as well as that of their children and other household members through the homestead food

3 production activities and the adoption of optimal agricultural, health, nutrition, and hygiene practices. Study design and participants. Before the baseline evaluation, villages within 4 departments in the province of Gourma were identified for possible inclusion in the E-HFP program; participating villages needed to have water sources to support production during the dry season. Fifty-five (of 181) eligible villages were identified for randomization and were stratified by commune and village size before randomization. We used a cluster-randomized controlled trial design and randomly assigned these 55 villages to 1 of 3 groups: 1) control group, which received no interventions from HKI (25 control villages), 2) E-HFP program with BCC led by OWLs (15 OWL villages), or 3) E-HFP program with BCC led by HC members (15 HC villages) (Supplemental Figure 1). Within the selected villages, all households with a mother who had a child aged 3 12 mo at the time of the baseline survey (February May 2010) were invited to participate in the impact evaluation. Those who were in OWL or HC villages were also invited to participate in the program. The evaluation used a longitudinal design and followed the same households, mothers, and children over time. An endline survey was conducted 2 y after program implementation between February and June The study was designed to assess the impact on child nutritional status and health as primary impact measures, as well as on mothersõ nutritional status and empowerment as secondary impact measures. The comparison between the 2 treatment groups (HC and OWL villages) was expected to be particularly relevant for looking at impacts on maternal knowledge and adoption of practices related to child nutrition and health but was not expected to be of primary importance for impacts on mothersõ nutritional status and empowerment. For this reason, and because we observed no statistically significant differences between HC and OWL villages in any of the maternal indicators studied, we pooled the 2 treatment groups to increase power for the analyses presented in this article. Measures. Household surveys were used to collect data at baseline and endline on household demographic and socioeconomic characteristics, agriculture production, ownership and control over assets, mothersõ health, nutrition and hygiene-related knowledge, household and mothersõ dietary diversity, and mothersõ perceptions of intrahousehold processes related to womenõs empowerment. In addition to the household survey, mothersõ height was measured at baseline, and weight was measured at baseline and endline. Household consumption of individual food groups was examined by using detailed consumption data (7-d recall) on 57 commonly consumed foods. Dichotomous variables were constructed to capture household consumption of each of 11 individual food groups: cereals and grains; roots and tubers; legumes, nuts, and pulses; fruit; vegetables; meat and poultry; fish and seafood; milk and dairy products; oils and fats; sugar and sweets; and miscellaneous. Household dietary diversity scores were calculated from these data by using 11 of the 12 food groups included by Swindale and Bilinsky (16) for the construction of a household dietary diversity score. The egg food group was not included because of an oversight during survey design. MothersÕ dietary intake was assessed over a 24-h recall period, in which mothers were asked to report whether they had consumed foods from 18 different categories (rice; cereals; beans; dark green leafy vegetables; yams, sweet potatoes, carrots, and tomatoes; other starchy vegetables; mango and papaya; other fruits; meat; poultry; organ meats; fish; eggs; peanuts; milk; other dairy; fats; and sweets) (17). These foods were aggregated into the same food groups used for the household consumption module (for comparison) and into the 9 food groups recommended for constructing the womenõs dietary diversity indicator (using information available at the time of the study; i.e., starchy staples, dark green leafy vegetables, vitamin A rich fruit and vegetables, other fruit and vegetables, organ meat, meat and fish, eggs, legumes, nuts and seeds, and milk and dairy products) (18); the latter is reported in Table 1. Trained staff measured mothersõ height at baseline by using a wooden height board (Shorr Productions). MothersÕ weight was measured at baseline and endline with the use of an electronic scale. BMI was calculated, and underweight was defined as BMI <18.5. Outstanding cases of abnormally low (<13.0) or high (>40.0) BMI or low weight (<30 kg) were excluded from analyses, as were those who were pregnant at endline or who were missing the month of assessment at baseline, endline, or both (Supplemental Figure 1). To measure womenõs empowerment, we used a 30-question survey. Questions were developed based on womenõs empowerment questionnaire modules that had been pilot tested in Mali and were further pilot tested in Burkina Faso in the context of this study. To identify the underlying components of womenõs empowerment, the data were reduced with the use of exploratory factor analysis, which suggested 7 unique components: meeting with other women, spousal communication, social support, purchasing decisions, family planning decisions, health care decisions, and infant and young child feeding (IYCF) TABLE 1 Unadjusted mean proportion of mothers who consumed individual food groups in the past 7 d, mean dietary diversity score at baseline and after 2 y, and DID impact estimates for these indicators among beneficiary and nonbeneficiary mothers in the enhanced-homestead food production program 1 Variable Control (n = 506) Treatment (n = 766) DID 2 P value Cereals and grains Baseline Endline Roots and tubers Baseline 1 4 Endline Legumes, nuts, and seeds Baseline Endline Milk and dairy products Baseline 1 2 Endline Meat and poultry Baseline Endline Fish and seafood Baseline Endline Eggs Baseline 1 1 Endline Fruit Baseline 10 5 Endline Vegetables Baseline Endline Oil and fats Baseline Endline Sugars Baseline 3 3 Endline Miscellaneous Baseline 6 5 Endline WomenÕs dietary diversity score (0 9) Baseline Endline Values are means 6 SDs or percentages. DID, difference-in-differences. 2 DID impact estimates between treatment and control groups are coefficients 6 SEs or percentage points. Agriculture-nutrition program benefits women 1111

4 decisions (Supplemental Table 1). Total scores for each component are the sum of responses, and all variables were coded so that higher scores indicate a greater level of empowerment. CronbachÕs a, which can range from 0 to 1, was calculated for each subscale to estimate internal consistency (>0.8, very good; , adequate; <0.5, not adequate) (19). Analysis. Statistical analyses were conducted in STATA 13.1 (StataCorp LP). Household demographic characteristics, mothersõ age, BMI, and prevalence of underweight at baseline were compared across the treatment and control villages for all households in the study. Variables were considered balanced at baseline if P > 0.05 for t tests comparing the 2 groups. Of 10 household- and 8 individual-level variables, only one (motherõs BMI) test was significantly different between the treatment and control villages (P = 0.03). Difference-in-differences (DID) impact estimates were used to measure the impact of the E-HFP program on household consumption of individual food groups and household dietary diversity, mothersõ intake of individual food groups and their dietary diversity, mothersõ BMI and prevalence of underweight, overall womenõs empowerment score, and the 7 components of womenõs empowerment identified through the exploratory factor analysis. The DID estimates describe the change in program indicators between the endline and baseline surveys for the treatment group relative to the control group. The specification for household consumption of individual food groups and dietary diversity controlled for whether the participating mother had ever attended formal schooling, whether the household head had ever attended formal schooling, household size, and a housing quality index factor score, calculated with the use of principal components analysis (20). In addition to these demographic and household controls, maternal age at baseline was also included for the specifications that assessed impact on mothersõ outcomes. Number of food shocks experienced in the 12 mo before the baseline assessment was also included for the specifications for mothersõ intake of individual food groups, mothersõ dietary diversity, mothersõ BMI, and prevalence of underweight. Specifications for mothersõ BMI and prevalence of underweight also controlled for the month the measurements were taken at baseline and endline. The standard errors for all specifications were corrected for clustering at the village level, the unit of randomization. The analytic sample was restricted to households or mothers with complete data at baseline and endline for a given indicator. To address possible attrition bias, attrition weights were calculated and applied to the sample descriptive statistics and impact estimates. Impact estimates were considered marginally statistically significant at P < 0.10 and statistically significant at P < For impacts that were significant, we also calculated effect size (ES) using the following formula: p t statistic 3 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1=n 1 þ 1=n2 ð1þ Ethical approval and consent. Trained fieldworkers provided information about the study to the household heads and mothers of the selected children, and oral informed consent was obtained from the household head and the mother of the selected child. The protocol was approved by the Ministry of Health of Burkina Faso, and the institutional review board of the International Food Policy Research Institute. The trial was registered with clinicaltrials.gov as NCT Results Trial profile and attrition. After informed consent, which began in February 2010, 1767 households agreed to participate in the baseline study (only 1 household refused to participate). Of the households that participated in the baseline study, 1481 participated in the endline study in 2012 (Supplemental Figure 1). At endline, one cluster (village) in the HC treatment group was lost due to internal community problems, resulting in the withdrawal of the village from the program (n = 61) Olney et al. Between baseline and endline, 16% of the original sample attritted. Analysis revealed a statistically significantly higher rate of attrition among households and mothers in control villages (19% for households and 29% for mothers) than in treatment villages (14% for households and 22% for mothers). The higher attrition rate in the control villages was possibly due to lack of program activities in these villages. As noted, all results include attrition weights. As a whole, attritted households were significantly smaller (both in terms of total household size and number of children aged <6 y), were less likely to be polygamous, were more likely to be female headed, and had a higher proportion of participating mothers with formal education than did nonattritted households. In most cases, these differences were similar for treatment and control villages. There were a few exceptions: in treatment villages, female-headed households, those with participating mothers with formal education, and those with a lower value of male assets were significantly more likely to attrit. In control villages, households in which the head had received any formal education were more likely to attrit. Mothers who attritted had a significantly higher BMI and were significantly less likely to be underweight at baseline than were mothers who did not attrit. Differences were similar across control and treatment villages but only statistically significant among mothers in treatment villages. Given that the women who attritted had a higher BMI on average and were less likely to be underweight, any potential attrition bias would attenuate treatment effects on increasing BMI or decreasing underweight. Baseline characteristics. At baseline, there were no statistically significant differences in household characteristics between treatment and control groups, despite differential attrition rates by group. Households consisted of ;8 members, and almost all were headed by men (Table 2). The number and value of assets held by men were higher than those held by women, and formal education was rare among household heads and participating mothers in both treatment and control villages. Prevalence of maternal underweight was high at baseline and was significantly higher in treatment than in control villages. Likewise, mean BMI was significantly lower among women in treatment than in control villages. The mean age of mothers, however, did not significantly differ between groups at baseline (Table 2). Impacts on mothersõ BMI and prevalence of underweight. After 2 y of participating in the E-HFP program, beneficiary mothers living in treatment villages had a statistically significantly greater decrease in the prevalence of underweight than did those living in control villages, in which the prevalence of underweight remained the same (Table 3, Figure 1) [DID = 28.7 percentage points; P < 0.01]. The overall impact on mean BMI was not statistically significant, but there was a marginally statistically significant interaction between baseline underweight and treatment group on BMI (coefficient 6 SE: b = ; P-interaction = 0.07). This interaction shows that the effect of the program on BMI differed depending on whether women were underweight at baseline; among women who were underweight at baseline and participated in the program, BMI increased relative to the control group (coefficient 6 SE: b = ), whereas among those who were not underweight at baseline, there was virtually no difference in the change in BMI between the treatment and control villages (coefficient 6 SE: b = ) (Figure 2). Impacts on mothersõ and householdsõ dietary diversity. Consistent with the programõs positive impact on womenõs underweight, there were also greater increases in household and

5 TABLE 2 Unadjusted mean baseline key characteristics of beneficiary and nonbeneficiary households and mothers in the enhanced-homestead food production program 1 Variable Control Treatment P value Households, n Household size (residents present at least 6 mo), n Children (aged,6 y), n Polygamous households, % Female-headed households, % MenÕs asset value, XOF 72, ,694 74, , WomenÕs asset value, XOF 44, ,923 42, , Housing quality index factor score Food shocks experienced in the past 12 mo, n Household head had any formal education, % Women had any formal education, % Mothers, n Age, y BMI, kg/m Underweight (BMI,18.5 kg/m 2 ), % Values are means 6 SDs or percentages unless otherwise indicated. XOF, West African CFA (Financial Community of Africa) franc. maternal consumption of nutrient-rich foods between baseline and endline in treatment than in control villages. Specifically, there was a significantly greater increase in the proportion of households that had consumed fish or seafood and fruit, as well as a marginally statistically significant smaller decline in the proportion that consumed meat or poultry in the past 7 d among households in treatment villages compared with those in control villages (Table 4). In addition, there was a significantly greater decrease in the consumption of roots and tubers among households in treatment than in control villages. However, there were not significant impacts of the program on household dietary diversity scores. The changes in household consumption of individual food groups were due to changes in the proportion of households that bought and/or took these foods from storage in the past 7 d. For example, there was a greater change in the proportion of households in treatment than in control households that bought fish or seafood, fruit (particularly mangoes), meat, or poultry and a greater decrease in households that bought roots or tubers from baseline to endline in treatment than in control households (DID = 20.2 percentage points, P = 0.03; DID = 10.2 percentage points, P = 0.09; DID = 10.4 percentage points, P = 0.07; and DID = 24.6 percentage points, P =0.03, respectively). In the case of fruit, there was also a marginally significantly greater change from baseline to endline in the proportion of households in treatment compared with control households who consumed fruit (particularly mangoes) from their stores in the past 7 d (DID = 12.5, P = 0.06). Changes from baseline to endline in mothersõ intake of individual food groups as well as in their overall dietary diversity in treatment compared with control villages were similar to patterns seen in household dietary diversity. Mothers in treatment compared with control villages had a significantly greater increase in fruit intake in the past 24 h, a marginally significantly greater increase in meat and poultry intake (P = 0.08), and a significantly greater increase in cereal and grain intake (Table 1). In addition, there was a marginally significant positive (but small; ES = 0.10) effect of the program on increasing mothersõ dietary diversity score in treatment compared with control villages (P = 0.08). Impacts on womenõs empowerment. The program also had statistically significant positive impacts on 3 components of womenõs empowerment as well as on overall empowerment. Compared with the control group, improvements between baseline and endline among women from the treatment group were found in their overall empowerment score and specifically within the components of meeting with other women and contributing to purchasing and health care decisions (Table 5). ES for the overall score and these individual components TABLE 3 Unadjusted mean BMI and prevalence of underweight (BMI,18.5 kg/m 2 ) at baseline and after 2 y and DID impact estimates for these indicators among beneficiary and nonbeneficiary mothers in the enhanced-homestead food production program 1 Variable Control (n = 510) Treatment (n = 787) DID 2 P value BMI, kg/m 2 Baseline Endline Prevalence of underweight (BMI,18.5 kg/m 2 ) Baseline Endline Values are means 6 SDs or percentages. DID, difference-in-differences. 2 DID impact estimates between treatment and control groups are coefficients 6 SEs or percentage points. Agriculture-nutrition program benefits women 1113

6 FIGURE 1 Unadjusted mean prevalence of underweight (BMI,18.5 kg/m 2 ) at baseline and after 2 y among beneficiary and nonbeneficiary mothers in the enhanced-homestead food production program. **P, 0.05 for difference-in-differences impact estimate for change in prevalence of underweight from baseline to endline in treatment compared with control villages. C, control villages; T, treatment villages. indicated that the effects, although significant, were small (ES = 0.19, 0.26, 0.16, and 0.12 for total empowerment, meeting with other women, purchasing decisions, and health care decisions, respectively). No significant effects of the program were found in the components of spousal communication, social support, family planning decisions, or IYCF decisions. Discussion Our results confirm, as others have, that integrated agriculture and nutrition programs are successful in increasing household consumption and mothersõ dietary intake of nutrient-rich foods (6, 7). However, our results add to this literature by demonstrating, and for the first time, to our knowledge, through a randomized controlled trial, that these programs can also directly improve mothersõ nutrition and empowerment. Our study showed a significant decrease in the prevalence of underweight among mothers in our sample over a period of only 2 y and a significant increase in BMI among those mothers who were underweight at baseline. The program also had a positive impact on womenõs empowerment scores and on 3 dimensions of empowerment: social capital, represented by meeting with other women, and their participation in decision making regarding purchases and health care use. These benefits not only represent positive changes for the mothers themselves but can also potentially improve their ability to care for their children and, in turn, contribute to improving their childrenõs healthand nutrition. To our knowledge, this is the first evaluation of an integrated agriculture and nutrition program to demonstrate a positive impact on reducing underweight among mothers by a substantial 8.7 percentage points in treatment compared with control villages. This significant impact, coupled with the significant interaction of baseline underweight and treatment on change in mothersõ BMI, indicates that the E-HFP program benefited the most vulnerable mothers. Of the 3 studies that have looked at the impact of these types of programs on mothersõ weight, BMI, or prevalence of underweight (8, 10, 11), only one found a positive impact on BMI but no effect on the prevalence of underweight among women (11). In the study by Kumar and 1114 Olney et al. Quisumbing (11), early adopters compared with late adopters of improved home gardening technologies, but not fish farming, had higher BMI 10 y after the program started. The other 2 studies did not find a significant impact on either motherõs BMI (8) or adult underweight (10), although statistical power may have been an issue in both of these studies. Supporting the positive impact on decreasing the prevalence of underweight among mothers, we found statistically significant positive effects of the E-HFP program on household consumption of nutrient-rich foods, particularly of fruit, fish or seafood, and meat or poultry, which were consistent with changes in food purchases and use of food from stores between baseline and endline. Impacts on mothersõ dietary intake of individual food groups and diversity largely reflected impacts at the household level and were statistically significant for fruit and marginally significant for meat and poultry and dietary diversity scores. The parallel findings on householdsõ and mothersõ dietary intake and diversity are in line with qualitative research conducted in Burkina Faso that illustrates that women and children may be favored in the intrahousehold distribution of foods (21). This article demonstrates that contrary to research conducted in other countries, particularly in Asia (22), men are the first to cut back on food intake and that women have their own stores of foods that they can access when the household is facing cereal shortages (21). Another study conducted among pastoralists in Kenya describes a similar trend whereby men are the first to suffer the effects of food shortages, whereas women and children tend to be more protected (23). A program such as HKIÕs E-HFP program could be particularly important for safeguarding mothersõ access to food in this context by building or protecting their food stores through the additional food produced as a result of program inputs and activities as well as by increasing food availability. Stores of food are likely to be most important during the planting or lean season in rural Burkina Faso when cereal shortages are common, which can cause the prevalence of underweight to increase by ;6 percentage points (4). In addition, increased availability of food through increased production is likely to be important during the harvest season to make up for any deficits accrued during the preharvest FIGURE 2 Unadjusted mean BMI (in kg/m 2 ) at baseline and after 2 y among mothers who were underweight at baseline (BMI,18.5) and those who were not (BMI $18.5) among beneficiary and nonbeneficiary mothers in the enhanced-homestead food production program. *P = 0.07 for interaction of baseline underweight and treatment group for change in BMI from baseline to endline. NW/C, normal weight/control villages; NW/T, normal weight/treatment villages; UW/C, underweight/control villages; UW/T, underweight/treatment villages.

7 TABLE 4 Unadjusted mean proportion of households that consumed individual food groups in the past 7 d, mean dietary diversity score at baseline and after 2 y, and DID impact estimates for these indicators among beneficiary and nonbeneficiary households in the enhanced-homestead food production program 1 Variable Control (n = 596) Treatment (n = 880) DID 2 P value Cereals and grains Baseline Endline Roots and tubers Baseline 3 7 Endline Legumes, nuts, and pulses Baseline Endline Milk and dairy Baseline 2 3 Endline Meat and poultry Baseline Endline Fish and seafood Baseline Endline Fruit Baseline Endline Vegetables Baseline Endline Oils and fats Baseline Endline Sugars Baseline Endline Miscellaneous Baseline Endline Household dietary diversity score (0 11) Baseline Endline Values are means 6 SDs or percentages. DID, difference-in-differences. 2 DID impact estimates between treatment and control groups are coefficients 6 SEs or percentage points. season. In our study, conducted during the postharvest season in both 2010 and 2012, the significant positive impact of the E-HFP program on reducing the prevalence of underweight among mothers by 8.7 percentage points is plausible given the demonstrated impact of the E-HFP program on increasing mothersõ agricultural production and ownership of agriculture assets and small animals during the previous year and control over the vegetables produced in their gardens (14), as well as due to the documented improvements to mothersõ dietary intake of nutrient-rich foods and diversity. The E-HFP program also improved empowerment among mothers of young children in terms of their total score and specifically within 3 domains: meeting with other women, participation in health care decisions, and participation in purchasing decisions. The positive impact onwomenmeetingwithother women in treatment compared with control villages likely reflects the program activities, which in part centered on beneficiaries working together at the village model farms as well as participating in meetings related to the nutrition and health BCC activities. It is also possible that because of these program activities, mothers were more likely to meet with other women outside of program activities to discuss the new agriculture, health, and nutrition practices they were being taught. Regardless, this positive increase in meeting with other women has the potential to build mothersõ social capital, which can increase their access to resources and social support (i.e., being able to access resources or safety when experiencing difficult situations) (24). The significant positive impacts of the E-HFP program on mothersõ participation in health care and purchasing decisions may in part be due to the positive impacts of the program on mothersõ health and nutrition-related knowledge (15), ownership of agricultural assets and small animals, agricultural production, and the perceptions of themselves, their husbands, and other community members that women are capable of successfully producing food from their own gardens (14). TABLE 5 Unadjusted mean scores of womenõs empowerment, scores within individual components at baseline and after 2 y, and DID impact estimates for these indicators among beneficiary and nonbeneficiary mothers in the enhanced-homestead food production program 1 Variable Control (n = 517) Treatment (n = 781) DID 2 P value Meeting with other women (0 5) Baseline Endline ,0.01 Spousal communication (0 14) 3 Baseline Endline Social support (0 4) Baseline Endline Purchasing decisions (0 8) Baseline Endline Family planning decisions (0 2) Baseline Endline Health care decisions (0 2) Baseline Endline Infant and young child feeding decisions (0 2) Baseline Endline Total womenõs empowerment score (0 37) Baseline Endline , Values are means 6 SDs. DID, difference-in-differences. 2 DID impact estimates between treatment and control groups are coefficients 6 SEs. 3 Sample for spousal communication is limited to women who live with a spouse (control: n = 464; treatment: n = 641). Agriculture-nutrition program benefits women 1115

8 Although we found significant positive impacts of the program on 3 components of womenõs empowerment, we saw no significant impacts of the program on changes in the components related to spousal communication, social support, and participation in family planning decisions or in IYCF decisions. The lack of impact on participation in IYCF decisions is most likely because almost all mothers already participated in IYCF decisions at baseline, and thus there was no scope for improvements in this aspect over the course of the program. The lack of impact on participation in family planning decisions and spousal communication is not surprising given that no particular program interventions addressed these components of womenõs empowerment and is likely related to current cultural beliefs and practices and would be unlikely to be changed by the types of activities implemented as part of the E-HFP program. The study has some limitations. First, although dietary diversity has been associated with womenõs BMI in rural Burkina Faso during the postharvest season (4), more comprehensive measures of dietary intake may have shed further light as to how changes in mothersõ dietary intake were related to the changes seen in the prevalence of underweight. However, collection of this type of data was not possible in the context of this study due to time and resource constraints. Second, although the reduction in the prevalence of underweight is plausible given the multiple positive documented impacts of the E-HFP program on relevant outcomes, it is possible that other unmeasured factors also contributed to reducing underweight such as changes in use of health care services, morbidity, or workload. Third, recently developed measures to assess dietary diversity [Minimum Dietary Diversity Women (25)] and empowerment for women [WomenÕs Empowerment in Agriculture Index (26)] that would have been appropriate to use in this study had not yet been developed when the study was conducted. The use of different measures to assess womenõs dietary diversity and empowerment may limit the comparability of these findings to studies that use these newer assessment methods. Last, it should be noted that the findings of this study are relevant for villages with similar characteristics as those included in this study, which, most important in this context, is access to water for agricultural production. The implications of the positive impacts of the E-HFP program found on select aspects of womenõs nutritional status and empowerment may have positive impacts on their own wellbeing, food security, and their childrenõs health and nutrition, both concurrently and over the longer term (1, 27, 28). It will be important to examine these relations in future analysis and to also assess potential unintended negative effects of the E-HFP program on womenõs time use and ability to care for their children. The impacts documented in this study were achieved over a short 2-y period, but the changes in womenõs status and decision making are likely to affect womenõs lives and that of their children and families over much longer periods given their potentially transformational nature. Evaluations of short-term impacts such as this one are likely to underestimate not only the multiple benefits of integrated programs such as HKIÕs E-HFP program on targeted beneficiaries but also their potential to change community norms over the long term, especially if they can be scaled up. Acknowledgments DKO, AD, and MTR designed the evaluation with input from AP; DKO, AP, and AD led the data collection activities; DKO, LB, and JH led the data analyses; DKO, LB, AD, MTR, and JH drafted the manuscript; DKO had final responsibility for submitting this article for publication; and all authors contributed to interpreting the results from the evaluation presented 1116 Olney et al. here and had full access to the data, and also read and approved the final version of the paper. References 1. van den Bold M, Quisumbing A, Gillespie S. WomenÕs empowerment and nutrition: an evidence review. Washington (DC): IFPRI Publications; p Walker SP, Wachs TD, Grantham-McGregor S, Black MM, Nelson CA, Huffman SL, Baker-Henningham H, Chang SM, Hamadani JD, Lozoff B, et al. Inequality in early childhood: risk and protective factors for early child development. Lancet 2011;378: Institut National de la Statistique et de la Démographie (INSD) Ministère de lõéconomie et des Finances, Ouagadougou, Burkina Faso et ICF International, Calverton, Maryland U. The DHS program final report, Calverton (MD): DHS; 2012 (in French). 4. Savy M, Martin-Prével Y, Traissac P, Eymard-Duvernay S, Delpeuch F. Dietary diversity scores and nutritional status of women change during the seasonal food shortage in rural Burkina Faso. J Nutr 2006;136: Ruel MT, Alderman H. Nutrition-sensitive interventions and programmes: how can they help to accelerate progress in improving maternal and child nutrition? Lancet 2013;382: Girard AW, Self JL, McAuliffe C, Olude O. The effects of household food production strategies on the health and nutrition outcomes of women and young children: a systematic review. Paediatr Perinat Epidemiol 2012;26: Leroy JL, Frongillo EA. Can interventions to promote animal production ameliorate undernutrition? J Nutr 2007;137: Olney DK, Talukder A, Iannotti LL, Ruel MT, Quinn V. Assessing impact and impact pathways of a homestead food production program on household and child nutrition in Cambodia. Food Nutr Bull 2009;30: Talukder A, Haselow NJ, Osei AK, Villate E, Reario D, Kroeun H, SokHoing L, Uddin A, Dhungel S, Quinn V. Homestead food production model contributes to improved household food security and nutrition status of young children and women in poor populations: lessons learned from scaling-up programs in Asia (Bangladesh, Cambodia, Nepal and Philippines). F Actions Sci Rep 2010;(Special Issue 1): Habtemariam K, Ayalew W, Habte Gabriel Z, Gebre Meskel T. Enhancing the role of livestock production in improving nutritional status of farming families: lessons from a dairy goat development project in eastern Ethiopia. Livest Res Rural Dev 2003;15:Article Kumar N, Quisumbing AR. Access, adoption, and diffusion: understanding the long-term impacts of improved vegetable and fish technologies in Bangladesh. J Dev Eff. 2011;3: Bushamuka VN, de Pee S, Talukder A, Kiess L, Panagides D, Taher A, Bloem M. Impact of a homestead gardening program on household food security and empowerment of women in Bangladesh. Food Nutr Bull 2005;26: Brun T, Reynaud J, Chevassus-Agnes S. Food and nutritional impact of one home garden project in Senegal. Ecol Food Nutr 1989;23: van den Bold M, Dillon A, Olney D, Ouedraogo M, Pedehombga A, Quisumbing A. Can integrated agriculture-nutrition programmes change gender norms on land and asset ownership? Evidence from Burkina Faso. J Dev Stud 2015;51: Olney DK, Pedehombga A, Ruel MT, Dillon A. A 2-year integrated agriculture and nutrition and health behavior change communication program targeted to women in Burkina Faso reduces anemia, wasting, and diarrhea in children months of age at baseline: a clusterrandomized controlled trial. J Nutr 2015;145: Swindale A, Bilinsky P. Household dietary diversity score (HDDS) for measurement of household food access: indicator guide (v.2). Washington (DC): FHI 360/FANTA; Kennedy G, Ballard T, Dop M. Guidelines for measuring household and individual dietary diversity. Rome (Italy): Food and Agriculture Organization; Arimond M, Wiesmann D, Becquey E, Carriquiry A, Daniels MC, Deitchler M, Fanou-Fogny N, Joseph ML, Kennedy G, Martin-Prevel Y, et al. Simple food group diversity indicators predict micronutrient adequacy of womenõs diets in 5 diverse, resource-poor settings. J Nutr 2010;140:2059S 69S.