Asadullah 2015.

Study characteristics
Methods	Study design: PCS How were missing data handled? disaggregated attrition analysis was done to identify significant differential attrition and found none. No mention made of ITT analysis, but the footnote to Table A3 indicated the intervention sample as 2098 and control sample as 1940 HHs, which suggests that only HHs that completed the entire follow‐up (2002–2011) were included in analysis as these numbers + total attrition numbers add up to 5626 participants. Randomisation ratio: N/A Recruitment method: NR Sample size justification and outcome used: NR Sampling method: panel data from 4‐round surveys conducted in the Rangpur, Kurigram and Nilphamari districts in Bangladesh used. Participants recruited from extremely poor, with intervention participants recruited through a first‐order community‐based participatory wealth ranking to identify the lowest 10% of the population for income distribution. Authors did not report how recruitment for the panel data survey was done. Study aim or objective: to determine long‐term effects of phase 1 of the CFPR‐TUP programme, predominantly a one‐off transfer of livestock assets, in terms of changes in food expenditure, HH assets, food security and microfinance participation of very poor women in Bangladesh. Study period: panel data obtained at baseline in 2002, interim follow‐up in 2005 and 2008, and final follow‐up in 2011. Unit of allocation or exposure: HHs
Participants	Baseline characteristics Intervention or exposure group (n = 2098) Age: HH head: years (mean): 43 Place of residence: NR Sex: female headed HHs, %: 41.3 Ethnicity and language: NR Occupation: primary occupation of working‐aged males, %: day labour 66.7; non‐farm self‐employed 16.1. Primary occupation of working‐aged females, %: day labour 24.2; servant 13.6; HH chores 48.7 Education: years of education of HHs head 0.32; Rabbani et al: HH cumulative schooling 1.62; literacy rate, % 7.45; HHs with ≥ 1 literate member, % 20.97; net enrolment of children aged 6–10 67.16; net enrolment of children aged 11–16 35.52 SES: per capita per annum income (mean): BDT 37.27; roof made of tin, %: 43.2; HH size (mean): 3.64 members; Rabbani et al 2006: mean HH size 3.62; % of HH loans intended for regular consumption 67.29 Social capital: NR Nutritional status: food security, %: always deficit 60.10; somewhat deficit 37.27; surplus 0.05. Always face food deficit, %: 66.8. Per capita food expenditure per day (mean): BDT 60.1; Rabbani et al 2006: % of HHs where people could not eat for 1 day 62.10 Morbidities: Rabbani et al 2006: % of people ill in the last 15 days: 15.21 Concomitant or previous care: NR Control group (n = 1940) Age: HH head: years (mean?) 43; Rabbani et al 2006: mean HH age 28.00 Place of residence: NR Sex: female headed HHs, %: 22.4 Ethnicity and language: NR Occupation: primary occupation of working‐aged males, %: day labour 59.6; non‐farm self‐employed 22.8. Primary occupation of working‐aged females, %: day labour 13.3; servant 8.0; HH chores 67.6 Education: years of education of HH head: 0.65: Rabbani et al: HH cumulative schooling: 2.20; literacy rate, %: 13.16; HHs with ≥ 1 literate member, %: 32.88; net enrolment of children aged 6–10 years: 71.66; net enrolment of children aged 11–16 years: 43.74 SES: per capita per annum income (mean): BDT(?) 49.23; roof made of tin, %: 54.4; HH size (mean) 3.99 members; Rabbani et al 2006: mean HH size 3.86; % of HH loans intended for regular consumption 50.3 Social capital: NR Nutritional status: food security, %: always deficit 41.91; somewhat deficit 49.23; surplus 1.34. Always face food deficit, %: 39.6. Per capita food expenditure per day (mean): BDT(?) 41.91; Rabbani et al 2006: % of HHs where people could not eat for 1 day 45.13 Morbidities: Rabbani et al 2006: % of people ill in the last 15 days 14.17 Concomitant or previous care: NR Overall group (n = 5626): NR Inclusion criteria: ultra‐poor woman who met ≥ 3 of the following 5 criteria: 1. HH dependent upon female domestic/seasonal work, e.g. begging, maid; 2. own < 10 decimals of land; 3. no active male adult member in HH; 4. no productive assets in HH; and 5. children of school age have to take paid work. Exclusion criteria: no women should have any of the 3 exclusion criteria: 1. no adult woman in the HH who is able to work; 2. participating in microfinance and 3. beneficiary of government/NGO development project. Baseline imbalance: HHs in treatment and control groups differed significantly in many baseline characteristics. Intervention group had fewer cash savings, poorer HH conditions, faced more food deficit, had a smaller number of assets, was more likely to be female headed and had fewer years of education of HH head. Male members from intervention HHs were also less likely be in non‐farm self‐employment, and more likely to be day labourers. Female members of intervention HHs were more likely to be day labourers and servants, and less likely to stay home for HH chores. Attrition per relevant group: 1588 (28.2%) HHs lost to attrition over the total survey period. 895/2993 (20%) were intervention HHs and 693/2633 (26%) were control HHs. Attrition over short‐run (2002–2005) and medium‐run (2002–2008) impact times were NR per group, but total attrition was 398 (7.1%) HHs in 2005 and 1067 (19.0%) HHs in 2008. Description of subgroups measured and reported: none reported Total number completed and analysed per relevant group: footnote to Table A3 indicates the intervention sample as 2098 and control sample as 1940 HHs. This suggests that only HHs that completed the entire follow‐up (2002–2011) were included in analysis. Total number enrolled per relevant group: intervention 2993; control 2633 Total number randomised per relevant group: N/A
Interventions	Intervention: poverty‐reduction programme with direct one‐off transfer of livestock and livelihood training Food access intervention category: increase buying power Intervention type: income generation Description: multicomponent programme including orientation training on the programme, selection of chosen income‐generation microenterprise by female participants, transfer of productive assets worth BDT 10,000 to support the selected enterprise 1 month after orientation (90% of HHs chose livestock combination, other vegetable growing or non‐farm enterprises), community savings, monthly visit by health workers, weekly follow‐up sessions for technical advice, building of social capital (through village support networks and sponsorship of community leaders) and weekly stipends (BDT 70). Second round of support provided in case of loss of assets due to unforeseen shocks. Direct services include child health, immunisation, diarrhoeal disease control, vitamin A supplements for children aged < 5 years, tuberculosis control, and family services and pregnancy care. Duration of intervention period: one‐off asset transfer with follow‐up training from 2002 to 2011 Frequency: one‐off transfer of productive assets; weekly follow‐ups for technical advice; monthly visits from health workers Number of study contacts: baseline (2002) with 3 follow‐ups (2005, 2008 and 2011) Providers: NGO (BRAC) Delivery: NGO workers deliver training and assets. Training and weekly follow‐up with each HH. Co‐interventions: NR Resource requirements: NR Economic indicators: cost of intervention: BDT 10,000 (approximately USD 119) per HH Control: no intervention
Outcomes	Per capita food expenditure (per day) Food security: proportion experiencing food deficit always/somewhat of a food deficit/neither experiencing food deficit nor having surplus/with surplus food Dietary intake: per capita daily energy intake, with intake < 1805 kcal/day below what is required Anthropometry: HAZ; WHZ; BMI (women) Morbidity: HH members seriously ill
Identification	Sponsorship source: Research and Evaluation Division (RED) of BRAC Country: Bangladesh Setting: ultra‐poor HHs in poor communities Author's name: M Niaz Asadullah Email: m.niaz@um.edy.my Declarations of interest: yes; no potential conflicts of interest. Study or programme name and acronym: Challenging the Frontiers of Poverty Reduction – Targeting the Ultra Poor (CFPR‐TUP) Type of record: journal article Trial registration: N/A Protocol availability: yes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (Selection bias)	High risk	CBA, therefore, no randomisation performed.
Allocation concealment (Selection bias)	High risk	CBA, therefore, no allocation concealment.
Baseline characteristics similar (Selection bias)	High risk	There were significant differences in characteristics between the groups at baseline.
Baseline outcome measurements similar (Selection bias)	High risk	Serious baseline imbalance for ownership of natural and physical assets (all P < 0.01 with the exception of number of goat/sheep owned: P < 0.05); food security (all P < 0.01); per capita food expenditure and income (both P < 0.01); and financial market participation (all P < 0.01 with the exception of size of outstanding lending: P < 0.10; NS).
Blinding of participants and personnel (Performance bias)	Low risk	There was no blinding as this was a CBA but this is unlikely to have influenced the performance of the participants.
Blinding of outcome assessment (Detection bias)	High risk	There was no blinding. Outcomes were measured with surveys, based on self‐reports from participants. So knowledge of treatment allocation, and perceived benefit or not from this, could have influenced their reporting.
Protection against contamination (Performance bias)	Unclear risk	Nothing reported regarding how contamination was prevented. It is also unclear who the control groups were so it is difficult to assess potential contamination.
Incomplete outcome data (Attrition bias)	High risk	Attrition was relatively high for the total follow‐up period; with higher attrition in the intervention compared to the control group (31.4% with intervention vs 25.5% with control). Although reasons for attrition were unrelated to treatment, the characteristics of those who were LTFU and those who remained in the study differed significantly, and the characteristics of people who were LTFU in the intervention and control group were also different.
Selective outcome reporting (Reporting bias)	Unclear risk	We could access the protocol. In the report of Rabbani et al 2006, some outcomes that are NR in the article were listed such as schooling, purpose of cash borrowing, legal awareness, social inclusion and Eid spending. Note: minor differences in baseline values between Asadullah and Rabbani.
Other bias	Low risk	None identified.