Fenn 2015.

Study characteristics
Methods	Study design: cRCT Study grouping: N/A How were missing data handled? WHZ data coded as missing if WHZ > +5 or WHZ < –5; HAZ data coded as missing if HAZ > +5 or HAZ < –6. A child's data were excluded from the analysis if the child was deemed to be a different child to the child enrolled at baseline. While checks were in place to ensure that the same child was measured every month, in some cases these were not followed. Used criteria for exclusion a decrease in height or length > 1 cm (measurement error) or an increase > 15 cm (considered the maximum height a child could grow in 6 months). All effect analyses were ITT. Randomisation ratio: 1:1:1:1 Recruitment method: Action Against Hunger provided the initial HH lists, and these were further verified and updated by the study research team. HHs defined as poor and very poor were selected. Field officers then visited identified HHs to share details of study and get informed consent before collecting baseline data. Study participants were enrolled by data collection team and were unaware which intervention they would be getting at enrolment. However, blinding of participants was not possible due to nature of intervention. Data collection team was different to cash and voucher disbursement team. Data collection team was responsible for collection of data and sensitisation of the study recipients to use of cash and vouchers. Data collection team was accompanied by local research mobilisers who facilitated the data collectors (e.g. in locating HHs), and were responsible for delivering key BCC messages. Sample size justification and outcome used: target sample size (about 632 HHs per group) was calculated to measure a detectable difference of prevalence of being wasted of 7% between the intervention and the control groups postintervention. Sample size was powered to detect a 0.19 WHZ difference between the intervention and the control groups. Sample size was reached for the standard cash transfer, food voucher and control groups. However, for the double cash transfer group, the sample size was 600 due to the different funding amounts given for this group, which did not allow for an equivalent number of HHs to be included compared to the other 3 groups. Target sample size was calculated using an estimated ICC of 0.02 for prevalence of being wasted from an Action Against Hunger nutrition survey in Dadu District. ICC for prevalence of being wasted was 0.01 (Fenn 2017). Sampling method: it was not possible to carry out a public randomisation, therefore, randomisation was done by the principal investigator using a random number table to generate the randomisation sequence and then drawing village names from a box. Block randomisation was done, allowing equal distribution of the villages to each group for small (< 40 HHs), medium (40 (SD 85) HHs) and large (> 85 HHs) villages. The investigator had no knowledge of the villages involved and was not involved in the intervention implementation or any data collection. HHs were selected from villages from 3 agricultural areas sharing similar livelihoods, geography and access to the same elements of the standard WINS programme. Action Against Hunger provided the initial HH lists, and these were further verified and updated by the study research team. HHs defined as poor or very poor using eligibility criteria decided by research team with village participation, and based on ownership of cultivated land and number of goats and with ≥ 1 children aged 6±48 months were selected. The study was a closed cohort and followed all children in the same eligible HHs regardless of their baseline anthropometric status. Study aim or objective: to evaluate 3 cash‐based transfer modalities on nutritional outcomes in children aged < 5 years from poor and very poor HHs in Dadu District, Sindh Province, Pakistan. Aimed to 1. compare the nutrition status of children receiving either a seasonal UCTs or a fresh food voucher with those with access to Action Against Hunger care only, after 6 months and 1 year; 2. assess the costs and cost‐effectiveness of the different interventions; 3. understand the factors that determined the ways in which HHs used the different transfers and 4. explore the role of the different processes involved in the study outcomes and how they interact with the context (Fenn 2015). Study period: data for the main impact analysis and findings reported here involved 3 periods: baseline (May–July 2015), 6 months after baseline (December 2015), and 1 year after baseline (June/July 2016). Unit of allocation or exposure: villages
Participants	Baseline characteristics Standard cash transfer Age: child, months, mean: 25.6 (SD 12.3) Place of residence: NR Sex: girls, n/N (%): 433/905 (47.9) Ethnicity and language: ethnicity, n (%): Sindhi 587 (94.2), Balochi 36 (5.8), Punjabi 0 (0). Muslim religion, n (%): 622 (99.8) Occupation: NR Education: father primary education or more, n (%): 249 (40.0); mother primary education or more, n (%): 63 (10.1) SES: wealth category, n (%): most poor 112 (18.0); more poor 137 (22.0); poor 114 (18.3); less poor 134 (21.5); least poor 126 (20.2). Access to safe water, n (%): 49 (7.9) Social capital: NR Nutritional status: child dietary diversity, median: 7 (IQR 6–8); child wasted (WHZ < –2SD), n (%): 196 (22.0); child severe acute malnutrition (WHZ < –3SD), n (%): 69 (7.7); child stunted (HAZ < –2SD), n (%): 457 (50.9); child Hb, g/L, mean: 89 (SD 17). Morbidities: child diarrhoea, n (%): 228 (25.2); child ARI, n (%): 310 (34.3) Concomitant or previous care: child deworming, n (%): 125 (13.8); BISP (Benazir Income Support Programme) participation n (%): 46 (7.4) Double cash transfer Age: child, months, mean: 25.9 (SD 12.0) Place of residence: NR Sex: girls, n/N (%): 429/839 (51.1) Ethnicity and language: ethnicity, n (%): Sindhi 523 (87.8), Balochi 59 (9.9), Punjabi 14 (2.4) Occupation: NR Education: father primary education or more, n (%): 198 (33.2); mother primary education or more, n (%): 66 (11.1) SES: wealth category, n (%): most poor 129 (21.6); more poor 123 (20.6); poor 90 (15.1); less poor 128 (21.5); least poor 126 (21.1). Access to safe water, n (%): 92 (15.4) Social capital: NR Nutritional status: child dietary diversity, median: 7 (IQR 6–9); child wasted (WHZ < –2SD), n (%): 198 (24.0); child severe acute malnutrition (WHZ < –3SD), n (%): 74 (9.0); child stunted (HAZ < –2SD), n (%): 389 (46.5); child Hb, g/L, mean: 90 (SD 16). Morbidities: child diarrhoea, n (%): 229 (27.3); child ARI, n (%): 332 (39.6); child fever/malaria, n (%): 517 (61.7). Concomitant or previous care: child deworming, n (%): 93 (11.1); BISP participation n (%): 68 (11.5) Food voucher Age: child, months, mean: 26.2 (SD 11.9) Place of residence: NR Sex: girls, n/N (%): 417/866 (48.2) Ethnicity and language: ethnicity, n (%): Sindhi 612 (97.3), Balochi 17 (2.7), Punjabi 0 (0) Occupation: NR Education: father primary education or more, n (%): 241 (38.3); mother primary education or more, n (%): 80 (12.7) SES: wealth category, n (%): most poor 143 (22.7); more poor 145 (23.1); poor 91 (14.5); less poor 113 (18.0); least poor 137 (21.8). Access to safe water, n (%): 49 (7.8) Social capital: NR Nutritional status: child dietary diversity, median: 8 (IQR 6–8); child wasted (WHZ < –2SD), n (%): 165 (19.3); child severe acute malnutrition (WHZ < –3SD), n (%): 46 (5.4); child stunted (HAZ < –2SD), n (%): 473 (54.9) Morbidities: child diarrhoea, n (%): 236 (27.3); child ARI, n (%): 265 (30.6). Concomitant or previous care: child deworming, n (%): 111 (12.8); BISP participation n (%): 59 (9.4) Control Age: child, months, mean: 23.4 (SD 11.3) Place of residence: NR Sex: girls, n/N (%): 431/852 (50.6) Ethnicity and language: ethnicity, n (%): Sindhi 515 (82.9), Balochi 105 (16.9), Punjabi 1 (0.2) Occupation: NR Education: father primary education or more, n (%): 197 (31.7); mother primary education or more, n (%): 28 (4.5) SES: wealth category, n (%): most poor 154 (24.8); more poor 130 (20.9); poor 106 (17.1); less poor 132 (21.3); least poor 99 (15.9). Access to safe water, n (%): 57 (9.2) Social capital: NR Nutritional status: child dietary diversity, median: 8 (IQR 6–9); child wasted (WHZ < –2SD), n (%): 184 (21.9); child severe acute malnutrition (WHZ < –3SD), n (%): 62 (7.4); child stunted (HAZ < –2SD), n (%): 437 (51.7) Morbidities: child diarrhoea, n (%): 298 (35.0); child ARI, n (%): 273 (32.2) Concomitant or previous care: child deworming, n (%): 38 (4.5); BISP participation n (%): 104 (16.8) Overall: NR Inclusion criteria: HHs selected from villages from 3 agricultural areas sharing similar livelihoods, geography and access to the same elements of the standard WINS programme. HHs defined as poor or very poor using eligibility criteria decided upon by the research team with village participation, and based on ownership of cultivated land and number of goats and with ≥ 1 children aged 6 (SD 48) months were selected. Exclusion criteria: NR Pretreatment: baseline characteristics of clusters and participants between the different intervention and control groups were well balanced for mothers and their children, apart from the proportion of children who had received deworming treatment, which was lower in the control group. There were a few potential imbalances at the HH level and between villages due to the clustered nature of the study design. These included village size, ethnicity, access to safe water and distance to nearest health service. In the control group, there was a higher proportion of HHs of Balochi ethnicity. There also appeared to be differences in the SES and educational status of mothers and fathers (both lower) and a higher number of HHs participating in the BISP in control group (Fenn 2017). Attrition per relevant group: no evaluation clusters LTFU; response rates for HHs were 95.6% and children were 98.3% within clusters at 6 months and for HHs were 95.0% and children were 96.8% at 1 year. 109 (1.6%) children missing (WHZ outliers, excluded children, missing data at either time point): control 31, double cash 32, food voucher 24 standard cash 22. From Figure 1: excluded/missing HH data: control 3.9%; double cash 5.3%; food voucher 4.7%; standard cash: 5.9%; children missing/excluded: control 4.3%; double cash 5.1%; food voucher 5.5%; standard cash 6.2%; mothers BMI outcomes: 1307 (26.9%) mothers with missing data: control 308; double cash 315; food voucher 334; single cash 350. For MUAC, Hb and anaemia, missing data were ≤ 5%. Description of subgroups measured and reported: NR Total number completed and analysed per relevant group: at 12 months: control: HH and carer 607; children 815; double cash: HH and carer 568; children 796; food voucher: HH and carer 602; children 818; standard cash: HH and carer 595; children 849. At 6 months: control: HH and carer 601; children 809; double cash: HH and carer 573; children 809; food voucher: HH and carer 603; children 834; standard cash: HH and carer 607; children 874. Total number enrolled per relevant group: cluster allocations: total 114 eligible clusters randomised to: control 28; double cash 24; standard cash 31; food voucher 31. HH allocations: HHs assessed for eligibility 5128. Total HHs eligible to participate 2494. Baseline allocation 2494. HHs to the study groups: control 632; double cash 600; standard cash 632; food voucher 632. Total HH number receiving intervention (after dropped out due to relocation at baseline): control: HH 621; children 852; double cash: HHs 596; children 839; standard cash: HHs 623; children 905; food vouchers: HHs 629; children 866. Total number randomised per relevant group: 114 clusters (villages) randomised to: control 28; double cash 24; food voucher 31; standard cash 31. HH allocations 2494: control 632; double cash 600; food voucher 632; standard cash 632.
Interventions	Intervention characteristics Standard cash transfer Food access intervention category: increase buying power Intervention type: UCT – standard cash Description: standard cash amount of PKR 1500 (approximately USD 14) disbursed at same time every month for 6 consecutive months (Fenn 2017) Duration of intervention period: 6 months (July 2015 to December 2015) Frequency: monthly Number of study contacts: baseline; 6 months; 12 months Providers: the EU. Delivery: cash disbursed at distribution points on a monthly basis either by mobile banks that travelled to a central location serving some of the participating villages or through central banks that served several villages. Delivered with verbal messaging from Action Against Hunger field staff, who were present at all distributions, that children should benefit from the transfers. Disbursement of cash and vouchers was done by different organisations, and the cash participants had further to travel to their distribution point, which may have added to the opportunity costs to HHs and reduced the actual transfer value. Co‐interventions: all villages had access to the WINS programme, which provided outpatient treatment for children aged 6 (SD 59) months with severe acute malnutrition, micronutrient supplementation (children and pregnant and lactating women) and BCC. Key BCC messages on the causes of undernutrition, the benefits of exclusive breastfeeding, improved complementary feeding practices, food and water hygiene, handwashing and sanitation were targeted at mothers. These messages were delivered monthly to all study participants in group sessions by the research mobilisers who also facilitated data collection activities, such as locating HHs and setting up times to be available, but were not involved in the data collection itself. Resource requirements: disbursement of cash and vouchers was done by different organisations, and cash participants had further to travel to their distribution point, which may have added to the opportunity costs to HHs and reduced the actual transfer value. Economic indicators: authors mentioned cost‐effectiveness analysis (to be published elsewhere). Double cash transfer Food access intervention category: increase buying power Intervention type: UCT – double cash Description: UCT of double cash amount of PKR 3000 (about USD 28) disbursed at same time every month for 6 consecutive months. Duration of intervention period: 6 months (July 2015 to December 2015) Frequency: monthly Number of study contacts: baseline; 6 months; 12 months Providers: funded by DG EU Humanitarian Aid and Civil Protection Action Against Hunger field staff Delivery: cash disbursed at distribution points on a monthly basis either by mobile banks that travelled to a central location serving some of the participating villages or through central banks that served several villages. Delivered with verbal messaging from Action Against Hunger field staff, who were present at all distributions, that children should benefit from transfers. Disbursement of cash and vouchers was done by different organisations, and the cash participants had further to travel to their distribution point, which may have added to the opportunity costs to HHs and reduced the actual transfer value. Co‐interventions: all villages had access to the WINS programme, which provided outpatient treatment for children aged 6 (SD 59) months with severe acute malnutrition, micronutrient supplementation (children and pregnant and lactating women) and BCC. Key BCC messages on the causes of undernutrition, the benefits of exclusive breastfeeding, improved complementary feeding practices, food and water hygiene, handwashing and sanitation were targeted at mothers. These messages were delivered monthly to all study participants in group sessions by the research mobilisers who also facilitated data collection activities, such as locating HHs and setting up times to be available, but were not involved in the data collection itself. Resource requirements: disbursement of cash and vouchers was done by different organisations, and the cash participants had further to travel to their distribution point, which may well have added to the opportunity costs to HHs and reduced the actual transfer value. Economic indicators: authors mentioned cost‐effectiveness analysis (to be published elsewhere). Food voucher Food access intervention category: food prices Intervention type: fresh food vouchers Description: food vouchers with a cash value of PKR 1500 (about USD 14), which could be exchanged for specified fresh foods (fruits, vegetables, milk and meat) in nominated shops. Action Against Hunger ensured that all food voucher villages had good access to these shops, by nominating shops in, or nearby, these villages. All villages were served by ≥ 1 nominated shop. Vouchers were disbursed at same time every month for 6 consecutive months. Duration of intervention period: 6 months (July 2015 to December 2015) Frequency: monthly Number of study contacts: baseline; 6 months; 12 months Providers: funded by the EU. Delivery: food vouchers disbursed to participating HHs at the village level. Disbursement of cash and vouchers was done by different organisations. The food voucher group had more direct contact with Action Against Hunger field staff during voucher disbursement, which could have affected the results through greater exposure to key messages. It is also possible that the vouchers themselves were too restricted. They were designed to purchase fresh fruit, vegetables, and fresh meat and were, therefore, dependent on what the vendors stocked, such as chicken being the only available meat. There were also many anecdotal reports regarding vendors overcharging for food items redeemed against the vouchers as a way to cover their own administration fees in recovering the voucher costs. In this respect, the actual transfer value given may have been lower than the face value. Co‐interventions: all villages had access to the WINS programme, which provided outpatient treatment for children aged 6 (SD 59) months with severe acute malnutrition, micronutrient supplementation (children and pregnant and lactating women) and BCC. Key BCC messages on the causes of undernutrition, the benefits of exclusive breastfeeding, improved complementary feeding practices, food and water hygiene, handwashing and sanitation were targeted at mothers. These messages were delivered monthly to all study participants in group sessions by the research mobilisers who also facilitated data collection activities, such as locating HHs and setting up times to be available, but were not involved in the data collection itself. Resource requirements: NR Economic indicators: authors mentioned cost‐effectiveness analysis (to be published elsewhere). Control Food access intervention category: no intervention (WINS programme) Intervention type: no intervention (WINS programme) Description: no additional intervention beyond the basic WINS (Women and Children/Infants Improved Nutrition in Sindh) programme activities that were provided to all groups. A pure control group was not feasible given WINS programme coverage across Dadu District (Fenn 2017). All villages had access to the WINS programme, which provided outpatient treatment for children aged 6 (SD 59) months with severe acute malnutrition, micronutrient supplementation (children and pregnant and lactating women) and BCC. Key BCC messages on the causes of undernutrition, the benefits of exclusive breastfeeding, improved complementary feeding practices, food and water hygiene, handwashing and sanitation were targeted at mothers (Fenn 2017). The key WINS programme messages were delivered to all study participants in group sessions by REFANI‐P research mobilisers each month. The key messages were targeted at the mother/carers of the eligible children, although other HH members are not excluded from access to key messages (Fenn 2015) Duration of intervention period: 6 consecutive months (July–December 2015). Frequency: WINS programme – monthly BCC messages Number of study contacts: baseline; 6 months; 12 months Providers: no intervention except for WINS programme Delivery: no intervention except for WINS programme Co‐interventions: all villages had access to the WINS programme, which provided outpatient treatment for children aged 6 (SD 59) months with severe acute malnutrition, micronutrient supplementation (children and pregnant and lactating women) and BCC. Key BCC messages on the causes of undernutrition, benefits of exclusive breastfeeding, improved complementary feeding practices, food and water hygiene, handwashing and sanitation were targeted at mothers. Messages delivered monthly to all study participants in group sessions by the research mobilisers who also facilitated data collection activities, such as locating HHs and setting up times to be available, but were not involved in the data collection itself. Resource requirements: NR Economic indicators: NR
Outcomes	Anthropometry: WHZ; wasting (WHZ ≤ –2SD); severe wasting (WHZ ≤ –3SD); BMI; MUAC; HAZ; stunting (HAZ ≤ –2SD); severe stunting (HAZ ≤ –3SD) Biochemical: Hb Morbidity: prevalence of anaemia; incidence of diarrhoea, ARI and fever/malaria
Identification	Sponsorship source: 6 study authors received funding from the DfID (DFiD PO 6433). 2 study authors received funding from the Directorate‐General for European Civil Protection and Humanitarian Aid Operations of the European Union (ECHO/ERC/BUD/2015/91001). The funders had no role in study design, data collection and analysis, decision to publish or preparation of manuscript. Standard cash and food voucher groups were funded by the EU. Double cash group funded by EU Humanitarian Aid and Civil Protection. Country: Pakistan Setting: poor and very poor HHs in Agrarian district Comments: ISRCTN registry ISRCTN10761532 Author's name: Bridget Fenn Email: bridget@ennonline.net Declarations of interest: yes; no competing interests (Fenn 2017). Study or programme name and acronym: REFANI Pakistan Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Low risk	Randomisation by principal investigator using a random number table to generate randomisation sequence and then drawing village names from a box. Block randomisation was done, allowing equal distribution of the villages to each group for small, medium and large villages.
Allocation concealment (Selection bias)	Low risk	Allocation at village level.
Baseline characteristics similar (Selection bias)	Low risk	Village‐level characteristics were balanced at baseline but not all HH or child‐level characteristics. However, these were adjusted for in the analyses (child age at baseline, child sex; and for mother's analyses adjusted also for SES and baseline values of the outcome variables).
Baseline outcome measurements similar (Selection bias)	Low risk	There were some imbalances but these were adjusted for in the analyses.
Blinding of participants and personnel (Performance bias)	Low risk	Blinding in this type of study is not possible. It is unlikely that lack of blinding would influence participant or personnel behaviour or experience, beyond changes expected due to the intervention.
Blinding of outcome assessment (Detection bias)	High risk	Quote: "masking of the interventions to both participants and data collectors was not possible in this setting and for this type of study. Precautions were taken at the start of the study to try to mask the different interventions to participants, e.g., through incorporating buffer zones and training data collectors to keep the information to themselves, but it soon became clear that participants were aware of the other interventions." Comment: some outcomes were self‐reported and could have been susceptible to lack of blinding.
Protection against contamination (Performance bias)	Low risk	Cash transfers/food vouchers were only distributed to specific HHs, and all HHs in the same village were allocated to the same intervention. Therefore, contamination was unlikely.
Incomplete outcome data (Attrition bias)	Low risk	Children were excluded from the analyses if they had outlying data. In general, attrition was low (response rate for HHs was 95.6% and children was 98.3% within clusters at 6 months and for HHs was 95.0% and children was 96.8% at 1 year. For child outcomes, missing data were low (mostly < 5%). For maternal outcomes, only BMI had high missing data (26.9%) but ≤ 5% for other outcomes. No clusters were LTFU.
Selective outcome reporting (Reporting bias)	Low risk	All outcomes reported as specified in the protocol.
Other bias	Low risk	Misclassification bias of exposure: low risk; allocation by investigators. Measurement bias: low risk; appropriate instruments used and field workers trained in data collection. Incorrect analyses: low risk; clustering was taken into account in analyses.