For the analysis of further surgery for macular holes, performed or planned, we will use logistic regression (
xtlogit). Best-corrected visual acuity (BCVA), measured using Snellen charts at a standard distance of 6 m, will be transformed to a LogMAR scale with two decimal places [
9], and then analysed using linear regression (
xtreg). Measurements of BCVA corresponding to count fingers (CF), hand movements (HM), perception of light (PL), and no perception of light (NPL) will be replaced with values of 2.10, 2.40, 2.70, and 3.00, respectively. The number of patients in each BCVA category and their corresponding numerical values will be tabulated by treatment arm. For the patient-reported experience of positioning at 3 months, which is on the scale 0 (very difficult) to 10 (very easy) we will remain masked to treatment arm and consider an applicable cut-off value to dichotomise this variable and then use logistic regression (
xtlogit). For the patient-reported outcome ‘Given what you know now, would you still have elected to have the operation?’ we will pool the responses ‘Don’t know’ and ‘No’ into one category, and use logistic regression (
xtlogit). If the proportion of ‘Don’t know’ responses exceeds 10% (19 patients) we will undertake a sensitivity analysis by pooling the ‘Don’t know’ with ‘Yes’ responses and compare how this alternative pooling affects the odds ratio estimate. The Complications of Age-related Macular Degeneration Prevention Trial Research Group [
10] observed a skewed distribution of the VFQ-25 score; therefore, we will assume that our results will be similarly skewed, and perform a logistic transformation (log(x/(100 − x))) of the VFQ-25 scores, and with this transformed outcome use linear regression (
xtreg). When adjusting analyses of BCVA and VFQ-25 for their respective baselines we will use the same normalising transformation for the baseline measurement as for the follow-up measurement.