Pooling Data from Similar Randomized Clinical Trials Comparing Latanoprost with Timolol; Medical Results and Statistical Aspects

Abstract: Two different principles were studied. 1st - statistical analysis techniques were used to obtain medical results from a patient population. 2nd - the patient population was used to study the statistical analysis techniques. Medical conclusions: latanoprost and timolol treatment showed a statistically significant and clinically useful mean IOP-reduction in a typical worldwide clinical trial population. Latanoprost reduced the IOP 1.6 mm Hg more than timolol. The IOP-reduction was maintained with timolol and slightly enforced with latanoprost up to 6 months of treatment. The mean IOP-reduction was maintained during 2 years of latanoprost treatment. The overall risk of withdrawal due to insufficient IOP-reduction with latanoprost was 8%. The statistical methodological issues are of a general and reoccurring character in trial design of the IOP-reduction: should the statistical hypothesis testing be based on the mean intraocular pressure (IOP) or the proportion of patients who reach a specific IOP level, should the estimate of the IOP or IOP-reduction be based on single eyes, mean of bilaterally eligible and identically treated eyes or the difference between an eye with active treatment and a placebo treated contralateral eye, and is mean of replicated recordings useful? Statistical methodological conclusions: the most effective response variable varies with the selected patient population. Therefore, the trial design process should include a comparison of the variability, test power and required sample size for the possible response variables in a sample of the target population. At minimum a statistical consideration should be done.