Kandpal 2016.

Study characteristics
Methods	Study design: cRCT Study grouping: parallel group How were missing data handled? authors reported that ITT analysis was performed; however, what they defined as ITT analysis was NR Randomisation ratio: 1:1 Recruitment method: sample selected in 3 stages (Figure 1). First, provinces in which the programme had not been introduced as of October 2008 were enumerated. Of the 11 provinces available, 3 were excluded because of security concerns. From the remaining 8 provinces, 4 provinces were chosen to span all 3 macro areas of the country (North, Visayas and Mindanao). Next, in each of these 4 provinces, 2 municipalities were randomly chosen to represent the mean poverty level of areas covered by the programme. Within each selected municipality, 130 villages were randomly assigned to treatment and control groups of 65 villages each. Data for the HH assessment form to run the PMT for beneficiary selection were fielded in the 8 RCT municipalities between October 2008 and January 2009. This was followed by the implementation of Pantawid in treated villages, with the first payment of cash grants commencing in April 2009. Data used in this analysis were collected in a follow‐up survey from the 130 villages in October and November 2011, allowing for a programme exposure period of 30–31 months. Sample size justification and outcome used: because this evaluation was a cRCT with treatment assignment at the village level, a power analysis (Supplemental Table 1) was conducted using the 3 main outcomes of interest: monthly per capita HH consumption, school participation by children aged 6–14 years, and health facility visits by children aged 0–5 years. In keeping with the programme's stated objective of improving child health and nutrition, the central research question of the impact evaluation was to estimate the programme effect on child health and education. However, at the time of the power calculations, data on child anthropometric measurements were N/A for the Philippinesat, a decentralised level; as a result, these outcomes were omitted from the power calculations despite their being a central concern of the impact evaluation. The 2007 Family Income and Expenditure Survey and the 2003 National Demographic Health Survey data sets were used as proxies for outcome mean and variance in the comparison population. A modest hypothesised impact ensured an adequately powered study. The power analysis used a 10% increase in HH per capita expenditures, a 7 pp increase in school enrolments in children aged 6–14 years, and a 7 pp increase in health facility visit rate in children aged 0–5 years. Intracluster correlation coefficients were 0.12–0.25, depending on the outcome of interest. These factors combined to suggest an RCT size of 3900 HHs randomly selected from 134 enumeration clusters. Sampling method: eligible poor HHs were identified by the survey conducted by the National Household Targeting System for Poverty Reduction (NHTS‐PR) that used a PMT, which estimated per capita HH income on the basis of observable and easily provided information, including HH size and physical dwelling conditions. HHs with estimated per capita income below the poverty line were classified as poor. From this subset of poor HHs, Pantawid identified eligible HHs as being those with children aged 0–14 years or a pregnant woman at the time of the assessment, or both. Poor and eligible HHs received a combination of health grants and education grants every 2 months of PHP 500–1400 (USD 11–32), depending on number of eligible children in HH. Study aim or objective: to assess the impact of the Pantawid Pamilyang Pilipino Program (or Pantawid Pamilya) on HAZ and stunting of young children aged 6–36 months at the time of follow‐up survey. Study period: baseline data collected October 2008 to January 2009; follow‐up data collected October–November 2011 Unit of allocation or exposure: villages were allocated as clusters
Participants	Baseline characteristics Intervention or exposure Age: children aged ≤ 5 years: 1.1 per HH; children aged 6–14 years: 1.7 per HH Place of residence: owned a house and lot: 31.5% Sex: NR Ethnicity and language: NR Occupation: farming and livestock: 73.1% Education: no grade completed: 9.5% completed elementary school: 21.8%; high school graduate: 11.9% SES: house had no toilet: 41.4%; house had electricity: 42.4%; HH composition: 5.7 members Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Control Age: children aged ≤ 5 years: 1.1 per HH; children aged 6–14 years: 1.6 per HH Place of residence: owned a house and lot: 32.9% Sex: NR Ethnicity and language: NR Occupation: farming and livestock: 69.4% Education: no grade completed: 8.5%; completed elementary school: 21.8%; high school graduate: 10.4% SES: house had no toilet: 43.3%; house has electricity: 39.6%; HH composition: 5.7 members Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Overall Age: children aged ≤ 5 years: 1.1; children aged 6–14 years: 1.7 Place of residence: NR Sex: NR Ethnicity and language: NR Occupation: farming livestock: 71.3% Education: no grade completed: 9%; completed elementary school: 21.8%; high school graduate: 11.1% SES: had electricity in house: 41% Social capital: NR Nutritional status: NR Morbidities: NR Concomitant or previous care: NR Inclusion criteria: category 1 HH: 1418 poor HHs, i.e. HHs whose estimated per capita income fell below the poverty line, and that also had children aged 0–14 years or a pregnant mother (or both) at time of assessment. Exclusion criteria: 3 provinces were excluded because of security concerns Pretreatment: baseline data showed that HH characteristics were not significantly different between the 1418 category 1 treatment and control HHs. Attrition per relevant group: survey data included complete HAZ data on 194/241 treated children aged 36 months and 178/244 control children aged 36 months. Complete weight‐for‐age data collected for 204/241 treated children aged 36 months, and 189/244 control children aged 36 months. Anthropometric z‐scores were calculated on the basis of the WHO growth standard. Scores > 6 SDs above or below the reference mean were dropped from the sample. This trimming resulted in the dropping 10/194 treated children and 11/178 control children from the HAZ regressions, and the dropping of 2/204 treated children and 1/189 control children from the weight‐for‐age regressions. 15% of those eligible in the treatment villages reported that they did not participate; NR for control villages. Description of subgroups measured and reported: N/A Total number completed and analysed per relevant group: intervention group (children aged 6–36 months): 194 had height‐for‐age data, and 204 had weight‐for‐age data; control group (children aged 6–36 months): 178 had height‐for‐age data and 188 had weight‐for‐age data Total number enrolled per relevant group: intervention group: 241 children aged 6–36 months were part of HHs that underwent survey; control group: 244 children aged 6–36 months were part of HHs that underwent survey. Total number randomised per relevant group: sample of 1418 HHs was randomly assigned to 714 treated HHs and 704 control HHs for the impact evaluation. At time of the data collection in 2011, in these 714 treated HHs there were 241 children aged 3 years who could have been exposed to the programme in the first 1000 days of their lives, and 244 children in the same age range from poor HHs in control areas.
Interventions	Intervention characteristics Intervention or exposure Food access intervention category: increase buying power Intervention type: CCT Description: health grant: poor HHs with children aged 0–14 years or pregnant women (or both) received up to PHP 500 (USD 11) per HH per month, conditional on fulfilling the following requirements: 1. all children aged 5 years had to visit the health centre or rural health unit to receive age‐appropriate immunisation and vaccination, regular weight monitoring and monitoring for the management of childhood disease; 2. all pregnant women had to visit the health centre or rural health unit to undergo antenatal care, starting from the first trimester; 3. all school‐aged children (aged 6–14 years) had to receive deworming tablets twice per year; and 4. for HHs with children aged 0–14 years, the HH grantee (mother) or spouse (or both) had to attend family development sessions once per month. Education grant: grant of up to PHP 300 (USD 6.50) per child per month aimed to improve the school attendance of children aged 6–14 years living in poor HHs in selected areas. HHs only could receive the grant for ≤10 month/year to correspond with the duration of the school year, and for ≤ 3 children in the HH. Beneficiary HHs received the education transfer for each child as long as the child was enrolled in primary or secondary school and attended 85% of the school days every month. Duration of intervention period: 30–31 months: first payment of grants in April 2009 and a follow‐up survey in October and November 2011 Frequency: every 2 months Number of study contacts: 2 (baseline and end of study period) Providers: Pantawid program launched by the Philippine government Delivery: NR Co‐interventions: none Resource requirements: money; staff to manage the payments to the HHs; health workers to do the family development sessions; research staff, supervisors and a training manual for research staff collecting data Economic indicators: NR Control: no intervention
Outcomes	Diet diversity: child consumption of eggs/dairy/meat/fish in the past week Anthropometry: WAZ, underweight, severely underweight, HAZ, stunted, severely stunted Morbidity: fever, cough, or diarrhoeal disease in past 2 weeks
Identification	Sponsorship source: Consultative Group on International Agricultural Research (CGIAR) Research Program on Policies, Institutions, and Markets Country: Philippines Setting: poor HHs across all 3 macro areas of the country (North, Visayas and Mindanao) Author's name: Eeshani Kandpal Email: ekandpal@worldbank.org Declarations of interest: yes; no conflicts of interest. Study or programme name and acronym: Pantawid Pamilyang Pilipino Program (CCT programme) Type of record: journal article
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	Unclear risk	Method of randomisation NR.
Allocation concealment (Selection bias)	Unclear risk	Allocation concealment NR.
Baseline characteristics similar (Selection bias)	Low risk	Reported baseline characteristics were similar across intervention and control groups.
Baseline outcome measurements similar (Selection bias)	Unclear risk	Baseline data for nutritional outcomes NR.
Blinding of participants and personnel (Performance bias)	Low risk	Participants not blinded, but this was unlikely to have influenced participant behaviour.
Blinding of outcome assessment (Detection bias)	High risk	Unclear whether data collectors were blinded towards the group allocations, which may or may not have influenced the measurement of outcomes across groups, as some outcomes were self‐reported.
Protection against contamination (Performance bias)	High risk	7% of the control group participated in the intervention (ITT analysis performed), which may or may not have biased the estimated effects of the intervention towards 0.
Incomplete outcome data (Attrition bias)	High risk	Missing outcome data excluded from analyses and differed across groups, and reasons for LTFU were NR. For the outcome height‐for‐age, attrition was 47/241 (19.5%) in intervention group and 66/244 (27%) in control group. For the outcome weight‐for‐age, attrition was 37/241 (15.4%) in intervention group and 55/244 (22.5%) in control group.
Selective outcome reporting (Reporting bias)	Unclear risk	No protocol or trial registration number NR. While height‐for‐age and weight‐for‐age anthropometrical measurements in children aged 6–36 months were reported; weight‐for‐height (important to indicate wasting) was not. Results for the following outcomes were NR: monthly per capita HH consumption and health facility visits by children aged 0–5 years.
Other bias	Unclear risk	Misclassification bias: unlikely. Measurement bias: low risk. Quote: "Several rounds of training were conducted before data collection to ensure data quality, particularly of the anthropometric and dietary intake modules." However, the method for collecting dietary data (e.g. 24‐hour recall, or food frequency questionnaire) was NR. Incorrect analysis: low risk. Quote: "In order to take into consideration regional factors, including province‐specific eligibility cutoffs, and the clustered nature of the sample …, municipality fixed‐effects regressions were included. In addition, all SEs were clustered at the village level."