br Methods br Results Recruitment

Methods
Results Recruitment ran between July 12, 2010 and June 30, 2011, with follow-up completed on June 30, 2013. Fig. 1 shows the reasons for ineligibility and the flow of participants through the trial. Of 287 patients assessed for eligibility, 86 were randomly assigned: 44 to the telephone group and 42 to the face-to-face group. Characteristics at enrolment were reasonably well balanced between the two groups (Table 1). At baseline and 12months, 98% and 78% of participants had returned a QOL questionnaire and were included in the primary analysis (98% and 88% for the face-to-face and 97% and 68% for the telephone consultation groups respectively). A total of 13 (15%) participants had withdrawn from the study by the completion of the trial, 8 (19%) from the face-to-face group and 5 (11%) from the telephone group. Any available follow-up data for withdrawn patients e.g. a returned questionnaire was used in analysis as per the ITT principle. One participant allocated to telephone consultation group reverted to face-to-face consultations. In the face-to-face consultation group 40 (95%) participants had at least one consultation, as allocated before the 12month follow-up. This figure was 36 (82%) for the telephone consultation group. For the apexbio participants remained in the study, the median (IQR) number of consultations scheduled by the hospital for each patient which were not then cancelled by the hospital was 5 (3 to 6) for the face-to-face group and 4.5 (3 to 5.3) for the telephone group. The median (IQR) number of consultation attended per patient was 3 (2 to 4) for the face-to-face group and 4 (3 to 4) for the telephone group. A summary of all observed outcomes and the treatment effect estimates is given in Table 2. The median (IQR) of the QOL score in the face-to-face group was 106 (95, 116) and 113 (105, 125) in the telephone group. Fig. 2 illustrates all observed QOL scores at baseline and 12month follow-up. The specified ANCOVA model estimated a mean (95% CI) QOL score of 102.5 (96.5, 108.4) in the face-to-face group and 108.2 (101.6, 114.7) in the telephone group. The treatment effect estimate (95% CI) was then 5.7 (−2.9, 14.3) units on the IMPACT QOL scale (P=0.19) in favour of the telephone consultation group. Sensitivity analysis confirmed that the conclusion of non-inferiority implied by this confidence interval was robust to a wide range of plausible reasons for missing outcome data. The complier-adjusted analysis of QOL scores showed similar conclusions when there was at least minimal consultation participation (as defined previously); the efficacy-based treatment effect estimate (95% CI) was 4.8 (−4.1 to 13.6) points; p=0.285. There was no evidence that the treatment effect differed between Crohn\'s disease and ulcerative colitis/indeterminate colitis subgroups (p=0.15). For the anthropometric scores, there was no evidence of a difference between telephone consultation and face-to-face consultation. Disease relapses defined by a single or multiple PUCAI or aPCDAI scores greater than 15 over the 24month follow-up period were so rare (4 [10%] in the face-to-face arm and 1 [2%] in the telephone arm) as to prevent a multi-factorial analysis, so Fisher\'s exact test was used. There were no statistically significant differences between the control and intervention arm of CSQ scores at the last available observation as completed by either the participating children or their parents. There was a statistically significant lower mean consultation time for telephone consultations across all available follow-up compared to face-to-face consultation. Mean consultation times were 9.8min (IQR 8 to 12.3) for telephone consultation, and 14.3min (11.6 to 17.0) for face-to-face consultation with an estimated reduction (95% CI) of 4.3 (2.8 to 5.7) min in consultation times (p<0.001). One participant in each consultation group had one or more hospital admissions due to IBD. There was no evidence for a difference in the proportion of out-patient appointments attended between trial arms by binomial logistic regression.