You should collect statistics on your secondary indexes to enable the Optimizer to make the best use of them.
You cannot collect statistics on a UDT column. This includes UDT columns that are components of an index. The Optimizer uses dynamic AMP sampling information for equality predicates on UDT columns and default selectivity for other predicates on UDT indexes for costing. The dynamic AMP sampling provides limited statistics information about the index. For example, it cannot detect nulls or skew. If a UDT index access path does not show any improved performance, you should consider dropping the index to avoid the overhead involved in its storage and maintenance.