AIMS: To evaluate the efficacy of 50% autologous serum drops against conventional treatment in ocular surface disorders refractory to normal treatments in a prospective randomised crossover trial. METHOD: Patients fulfilling ophthalmological and haematological entry criteria were randomised to either 3 months of autologous serum 50% followed by 3 months of their conventional treatment, or 3 months of conventional treatment, followed by 3 months of autologous serum. Clinical assessments, including Schirmer's test, rose Bengal, and fluorescein staining, were carried out on entry and at monthly intervals. Impression cytology was performed at entry, 3 and 6 months. Grading was carried out on degrees of squamous metaplasia and goblet cell density. Subjective comfort was recorded daily using the "faces" scale. These categorical scores were converted to linear measurement using Rasch analysis. Statistical analysis was carried out using Wilcoxon's signed rank test and ANOVA. RESULTS: 16 patients were recruited with 31 eyes studied. The ocular surface diseases chiefly included Sjögren's syndrome (n = 6) and keratoconjunctivitis sicca (n = 5). Impression cytology available in 25 of 31 eyes showed significant improvement on serum treatment, p<0.02. Rasch weighted faces scores were statistically significantly better with serum, p<0.01. CONCLUSION: The results of this randomised study provide further evidence of the beneficial effects of autologous serum in severe ocular surface disorders. For most of these patients, autologous serum was superior to conventional treatment for improving ocular surface health and subjective comfort.