The answer (if there is one at all) will depend also on:

• what OS are we running under? - I am fortunate enough not to have to work with Windows, but have gathered that Strawberry Perl and previously Active State Perl provided some more GUI-ish facilities than are present in the typical *nix environment.
• what is the programmer's background? - long-time C hackers who are happy in the bowels of gdb will probably be happier with the traditional perl debugger then those coming from 'higher level' languages which possibly do have more 'integrated' features. Folks used to an Eclipse-like environment may miss those features when working with Perl.
• what structure does our application have? - the wishlist for IDE features will differ for linear utility scripts, simple CGI applications, forking/threading systems, heavily object-oriented/modularised applications, those that utilise existing frameworks...
• at what stage of the application development cycle are we? - initial modelling/design, in midst of iterative development, picking up a monolithic legacy application..

Emacs+ perl -d + $SHELL may well be enough for many users (even for forking applications, which are fun). There are threads here that say you don't even need the perl -d.

For "simple" (relatively) tasks plain old perl -d seems to 'work for me'. I used to find perl -d:Ptkdb was a bit too much overhead, especially over a ssh connection.

For more structured/architected applications I am watching Padre with interest.

For web applications Selenium looks interesting but I've not gotten into it yet.

So, "which is the best?" .. which would you like?