I want to use Microdata with Schema.org to define the main content of my webpage, so I did something like this:
There are several solutions to this, besides the workaround unor posted. mainEntityOfPage and mainContentOfPage are optional and I'm not sure how much they will affect the appearance of the rich snippets. Using something like about with just the main phrase/page title may be as effective in ranking - I'm unsure on this however.
Personally I would add json-ld for breadcrumbs, which is always hidden, and need not be near the actual visible. It's easier to read and maintain than microdata, can be positioned anywhere and nothing in schema.org states you must stick to only one format. I've successful used JSON-LD for some tasks (for example: dynamic data) and microdata for the rest, which google validates and understands. The json-ld can be added right at the start of or at the end, etc. You can just embed the BreadcrumbList inside a WebPage element. Google gives its own json-ld breadcrumb example. Example from schema.org -
A second alternative solution but not exactly fantastic either is to create duplicate breadcrumb data where you want it, using either or to make the links invisible. Depending on your website's design and number of breadcrumbs you may be able to find a visible area containing the correct words and add the microdata there - for example within the menu structure of that page, or using the page heading. The example below has 2 hidden breadcrumbs, plus uses the page title from the tag.
New Products
Although hidden meta data isn't ideal you are duplicating only what is already visible on the page, and only links within your own site. Some machine-readable but invisible microdata is acceptable, eg dates, currencies, times. Link is used within this w3.org example
If you choose to make them visible you could duplicate the breadcrumbs in small text in the footer (for example).