Ferguson 1998.

Methods	DESIGN: Cluster RCT LENGTH OF FOLLOW‐UP: 8 weeks and 3 months post treatment DATA ANALYSIS: Intervention received (only participants who completed follow up were included in analysis).  ATTRITION RATE: Overall attrition rate 11 (17%). Attrition at 8 weeks and 3 months respectively: Group 1: 0/33 (0%); 3/33 (9%); Group 2: 0/30 (0%); 8/30 (27%) UNIT OF DATA ANALYSIS: Individuals (not neighbourhoods). No intra‐class correlation coefficient reported. SAMPLE SIZE CALCULATION: Powered 0.8 with alpha=0.05 to detect an effect size of 0.5. However it is not stated to which outcome(s) this applies and the calculation does not appear to take into account the cluster design. Stated that an effect size of 0.5 with sample size of 63 is low and one or more hypothesis tests would be expected to yield non‐significant results. EQUIVALENT STUDY GROUPS AT BASELINE: Limited baseline data were provided and suggest that the experimental and comparison groups were similar in terms of their knowledge, age and college grade. Socio‐economic and sexual health data were not provided, though the author stated that the neighbourhoods were homogeneous in their average household income ranges. However, there were differences between groups at the study outset in the proportion who were sexually active (76% versus 60%). As only four communities were randomised, with only two per arm, other unreported chance imbalances may be likely. PROCESS EVALUATION: Not reported
Participants	NUMBER RANDOMISED: 63 AGE: mean 13; range 12 to 16 years GENDER: All female SOCIO‐ECONOMIC STATUS: Not reported specifically for participants but mentioned for the setting in general (see Setting below). ETHINCITY/RACE: African‐American (100%) LOCATION: USA; Charlottesville, Virginia; urban. PREVIOUS STI: Not reported SEXUAL RISK BEHAVIOUR: The majority of participants (76% in experimental group and 60% in comparison group) reported not ever having been sexually active at the start of the study. Of those who were sexually active, use of effective contraceptives for the most recent sexual intercourse at the start (pretest) was reported by 63% in the experimental group and 83% in the comparison group. OTHER: Inclusion of participants was contingent upon: having already successfully completed a pregnancy prevention programme (Camp Horizon); not being pregnant; and having never given birth.
Interventions	NAME OF STUDY: Not stated GROUP 1: Intervention: Culturally specific peer‐led education and skills based pregnancy prevention programme (n = 33) YEAR STARTED: Not stated PROVIDER(S): African‐American females aged 12 to 16 years who had been selected as peer counsellors and had received a 10‐week training programme devised by the author. Four were assigned to one experimental neighbourhood group and five to another. They led group discussions and facilitated role playing sessions. SETTING(S): Not explicitly stated but community based (urban public housing developments) in which average household income was 125% of federal poverty level, 80% of families were headed by adolescent mothers and 98% of residents were African‐American. TYPE: Information/education (contraception use; preventing pregnancy; delaying sexual activity); Practical skills (leadership skills; communication skills; sexual assertiveness skills). DURATION: 2 hours per week for 8 weeks THEORETICAL BASIS: Not reported STIs COVERED: STIs in general and HIV/AIDS  GROUP 2: Comparison group: Individual‐led pregnancy prevention programme (n = 30) Limited details provided. The comparison group differed primarily from the peer‐led experimental group in that the author alone taught the content, which was described as containing life management, family relations, academic and career modules and sexual and reproductive education.
Outcomes	PRIMARY/SECONDARY: Not stated which outcomes were primary. A statistical power calculation was provided, but it was not stated to which outcomes it applies (the power calculation might apply to one or both of two survey instruments that were used to assess most of the outcomes; if so, the outcomes would effectively all be co‐primary ‐ however, this is unclear). Behaviour (pregnancy prevention skills; frequency of sexual activity;  delayed first intercourse; effective contraceptive use) Knowledge (about reproduction, contraception and STIs) Health problem or state (pregnancy)
Notes	COST DATA: None reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Low risk	Neighbourhoods were randomly allocated to intervention by coin tossing. However, individuals (the unit of analysis) do not appear to have been randomly allocated. No explanation was given of how individuals were allocated within the cluster design.
Allocation concealment?	Unclear risk	No information provided.
Blinding? All outcomes	Unclear risk	No information provided. The author was involved in the conception, conduct and analysis of the study.
Incomplete outcome data addressed? All outcomes	High risk	Incomplete outcome data were not assessed in the analysis. However, the author noted that 8 females who dropped out of the comparison group had very low scores (not clear which scores) on the pre‐test and 8‐week post test and this may have possibly affected the overall 3‐month findings for the comparison group. The author also observed that although 11 females dropped out of the study by 3 months, the average age and college grade remained the same. Note also that there was missing data on effective use of contraceptives due to under‐reporting. A potential barrier to evaluation was that many sexually active participants did not answer the question on contraceptive use, leading to a small number of participants who reported using protection. Attrition rates were higher in Group 2. No reasons given for attrition.
Free of selective reporting?	Low risk	The outcomes were not clearly stated a priori and it is unclear which were primary or secondary. However, the reported outcomes each link to an hypothesis or question mentioned in the introduction, suggesting that outcome reporting was probably complete.
Free of other bias?	High risk	There were differences between trial arms at baseline in the proportion of young women sexually active. There were only two clusters per randomised trial group and the unit of analysis was individuals rather than clusters.