BackgroundSocial isolation in older adults is associated with morbidity. Evaluating interventions to promote social engagement is a research priority.MethodsA parallel-group randomised controlled trial was planned to evaluate whether telephone friendship (TF) improves the well-being of independently living older people. An internal pilot aimed to recruit 68 participants by 30 September 2012, with 80% retained at 6 months. Randomisation was web based and only analysts were blind to allocation. A service provider was contracted to train 10 volunteer facilitators by 1 April 2012 and 10 more by 1 September 2012. Participants were aged > 74 years with good cognitive function and living independently in an urban community. The intervention arm of the trial consisted of manualised TF with standardised training: (1) one-to-one befriending (10- to 20-minute calls once per week for up to 6 weeks made by volunteer facilitators) followed by (2) TF groups of six participants (1-hour teleconferences once per week for 12 weeks facilitated by the same volunteer). Friendship groups aimed to enhance social support and increase opportunities for social interaction to maintain well-being. This was compared with usual health and social care provision. The primary clinical outcome was the Short Form questionnaire-36 items (SF-36) mental health dimension score at 6 months post randomisation. Qualitative research assessing intervention acceptability (participants) and implementation issues (facilitators) and an intervention fidelity assessment were also carried out. Intervention implementation was documented through e-mails, meeting minutes and field notes. Acceptability was assessed through framework analysis of semistructured interviews. Two researchers coded audio recordings of telephone discussions for fidelity using a specially designed checklist.ResultsIn total, 157 people were randomised to the TF group (n = 78) or the control group (n = 79). Pilot recruitment and retention targets were met. Ten volunteers were trained by 1 September 2012; after volunteer attrition, three out of the 10 volunteers delivered the group intervention. In total, 50 out of the 78 TF participants did not receive the intervention and the trial was closed early. A total of 56 people contributed primary outcome data from the TF (n = 26) and control (n = 30) arms. The mean difference in SF-36 mental health score was 9.5 (95% confidence interval 4.5 to 14.5) after adjusting for age, sex and baseline score. Participants who were interviewed (n = 19) generally declared that the intervention was acceptable. Participant dissatisfaction with closure of the groups was reported (n = 4). Dissatisfaction focused on lack of face-to-face contact and shared interests or attitudes. Larger groups experienced better cohesion. Interviewed volunteers (n = 3) expressed a lack of clarity about procedures, anxieties about managing group dynamics and a lack of confidence in the training and in their management and found scheduling calls challenging. Training was 91–95% adherent with the checklist (39 items; three groups). Intervention fidelity ranged from 30.2% to 52.1% (28–41 items; three groups, three time points), indicating that groups were not facilitated in line with training, namely with regard to the setting of ground rules, the maintenance of confidentiality and facilitating contact between participants.ConclusionsAlthough the trial was unsuccessful for a range of logistical reasons, the experience gained is of value for the design and conduct of future trials. Participant recruitment and retention were feasible. Small voluntary sector organisations may be unable to recruit, train and retain adequate numbers of volunteers to implement new services at scale over a short time scale. Such risks might be mitigated by multicentre trials using multiple providers and specialists to recruit and manage volunteers.Trial registrationCurrent Controlled Trials ISRCTN28645428.FundingThis project was funded by the NIHR Public Health Research programme and will be published in full inPublic Health Research; Vol. 2, No. 7. See the NIHR Journals Library website for further project information.