Some variants of PDF fall under documented ISO standards (see also); these variants incur less format-obsolescence/unreadability risk. Whenever possible, it is obviously wise to produce PDFs that conform to an ISO standard!
Because Adobe can't resist cramming more bells and whistles into the format, though, it's eminently possible and quite common to produce non-standards-compliant PDFs. I offer a list of potential preservation risks in such PDFs, with the understanding that any given PDF may incur all, some, or none of them:
Embedding multimedia (audio, video) in PDFs is usually a bad idea. Should the embedded multimedia format become obsolescent, you'll have to can-open the PDF to yank it out and migrate it, afterwards trying to reconstitute the PDF. Complex objects are generally easier to preserve when different media types are kept in separate files (as with most HTML-based web pages).
If you don't embed fonts in your PDF, glyphs in those fonts may display wrongly or not at all on computers that don't have that font loaded. Over time, entire fonts can be lost or become unusable on modern equipment; I've heard from electronic thesis and dissertation managers that ETDs without fonts embedded have indeed suffered loss of information content.
Text non-computability risk
Scanning a print book page to PDF does not create computable text; it creates a picture of text. This is a common example of non-computable (non-indexable, non-searchable, non-text-minable, non-copy-and-pastable) digital "text," though hardly the only one. Some word-processing and page-layout programs produce PDFs in which (for example) the page layout is incorrectly hinted such that trying to copy-and-paste across two columns of text produces sentence salad! Hyphenation and page breaks also distort text for purposes of indexing.
Overzealous security risk
Many software packages that produce PDFs offer various kinds of "security measures" such as passworded file access or restricted printing. As with any technique that intentionally distorts or restricts access to files, such measures pose a serious risk to preservation. Do not use them on a PDF you hope to preserve.
Loss of text-structure risk
Computers are very good at producing nice typography, much less good at interpreting it. If the structure of a text (section divisions, headings, lists, etc) is important information, PDF is a poor choice to preserve that text, since (unless very carefully produced) it destroys structural information, retaining only appearance information.
(Conversely, if text appearance is vital, XML is usually a ridiculous preservation choice.)
Bad configuration risk
PDF is one of those million-option formats! Choosing software and PDF-production configuration poorly incurs any number of information-loss risks, from inappropriately low-resolution images to lost fonts to OS dependence to... you name it, really.