DSCS have a better story (generally) than centralized systems for offline or slow networks. They tend to be faster, which is really noticable for developers (using TDD) who do lots of check-ins.
Centralized systems are somewhat easier to grasp initially and might be a better choice for less experienced developers. DVCSes allow you to create lots of mini-branches and isolate new features while still doing red-gree-refactor checkin on green style of coding. Again this is very powerful but only attractive to fairly savvy development teams.
Having a single central repository for support for exclusive locks makes sense if you deal with files that are not mergable like digital assets and non-text documents (PDFs and Word etc) as it prevents you getting yourself into a mess and manually merging.
I don't think the number of developers or codebase size plays into it that much, both systems have been show to support large source trees and numbers of committers. However for large code bases and projects DVCS gives a lot of flexibility in quickly creating decentralized remote branches. You can do this with centralized systems but you need to be more deliberate about it which is both good and bad.
In short there are some technical aspects to consider but you should also think about the maturity of your team and their current process around SCCS.
