Assume that you have a table that has statistics on columns that correspond to a prefix of the key of the clustered or covering non-clustered index. There are some non-covering indexes for which this condition is not true. When the statistics of the table are updated automatically or manually with a specified sampling rate, a significantly smaller number of rows are sampledfrom the sampling rate than expected. This results in non-representative statistics, and inefficient query plans may be used for the queries that use the statistics.
The issue was first fixed in the following cumulative update of SQL Server. After you apply the fix, you have to enable trace flag 4199 to make the hotfix work.
Each new cumulative update for SQL Server contains all the hotfixes and all the security fixes that were included with the previous cumulative update. Check out the latest cumulative updates for SQL Server: