Statistical analyses
Statistical analyses were conducted using SPSS version 16.0 for Windows (Chicago, IL, USA). Descriptive statistics were used to assess the patient characteristics in the three therapy groups. Pearson’s chi-square tests were used to detect differences in the proportions of patients reaching treatment objectives of energy and protein intake by ONS during admission and one-way ANOVA (Bonferroni corrected) to detect differences in the intake of ONS in ml. The absolute percentages of patients reaching their treatment objectives and the associated 95% confidence intervals (CI) were used to compare the effectiveness of the different ONS administration modes. The numbers needed to treat were computed as the inverse of the risk differences. The calculations were performed according to as treated. Deviations from randomized allocations were stated. Last measured mean intake and loss-to follow-up were analysed as randomized. Patients who had a follow-up of less than 2 days were excluded from analysis. The reproducibility of the total intake assessment by nurses and nutrition assistants was assessed by intraclass correlations (model 1, single measure).