I know the post is now marked as solved, but I like to read though them and try peoples answers out. The last one you provided wouldn't quite work for me:
=TEXT(TODAY()-E6,"yy ""Years"" mm ""Months"" dd ""Days""")
Against a date of "09/07/1983" (UK format) gave me "29 Years 12 Months 06 Days"
I can see what it is doing, 29 years since 83, 12 months since July and 6 days since the 9th.
The formula posted by Deb works in probably every case except a few , but one should be aware of it.
Today being June 14 , what is the result expected for the same date in earlier months ? For example what should we get for May 14 , April 14 , March 14 , February 14 and January 14 ?
For February 14 , we get 3 months and 28 days ; probably 4 months might be what we expect.
This difference of a day or two exists for some dates ; probably because of the use of fractional values such as 365.25 , 30.4375.
Sorry for the redundancy but each time I read "that" function name a red alert light turns on and triggers the auto response event. And regarding age calculations...
As far as I know , the DATEDIF function gave problems in one application ; however , if you google this topic , you will see that almost every link uses this function.
http://dmcritchie.mvps.org/excel/datetime.htm
There is Chip Pearson's VBA procedure , which appears to be sound ; again , once we go back in time , it becomes that much more difficult to verify any algorithm for all possible start and end dates.
Once proven that erratically and sporadically doesn't work, then it's use should be avoided from then on in the interest of safety, reliability and stability.
And I suppose that Redmond dumb&deaf guys responsible for that implementation neither acknowledged the non-working issues nor had the intention of solving it. It should be added too the exile from any Excel user documentation and help from version 2000, 13 years and 5 versions ago!, which leads to the conclusion that they still allow its use so as to don't make older and obsolete workbooks crash.
I wouldn't use anywhere any function whose behavior couldn't be predictable and I think that none of my users (either individuals or corporations) would approve me doing so. At least the serious ones.