HLA and ophthalmic disease associations: a critical review and improved method for statistical analysis.

A critical review of published ophthalmic histocompatibility antigen association studies found numerous conflicting or ambiguous reports. Methodologic errors that could account for these discrepancies were categorized into 3 groups (1) arbitrary application of statistical methods to determine significance, (2) poor sampling design, and (3) variable disease definition. Multivariate analysis, not previously applied in this area provides a more powerful tool for assessing statistical significance. A computer program using this methodology for large studies is now available.