It really depends on the size of the signatures which will be used to print the book: the more pages up, the greater the creep. Since the book will be perfect-bound, the inner margin should be measured on the innermost pages of the printed and folded signature. If your measurements work on these pages, you should be OK...

I’ve been design a perfect-bound magazine with similar measures and (I believe) 16 page signatures. Almost 12,5 mm is "lost" on each page in the spine. There are variables though. Thinner and softer paper might give other results, and the cover also effect this.

Make a dummy booklet using the actual paper that will go into the final books and measure how much will really get lost in the gutter.

James (Dunwich) is on the right track. But in practice, measuring the loss is actually quite difficult.

I suggest two possible approaches.

1) Print out proofs of your design-in-progress and trim to size. Then set the page(s) into the dummy, snug up to the gutter.

Measure the amount left hanging out over the fore-edge. Trim this amount from the inside margin, and set the page(s) back into the dummy. (I like to use a bit of double-sided tape to hold the page against a dummy page, aligned to the fore-edge.)

Now you’ll really be able to tell how your design works, in three dimensions.

2) Another approach is to take the dummy, open a center spread, and mark with a pencil where you think the inner margin should fall visually. Then measure in from the fore-edge and use that to determine the inner margin of your layouts.

Then go through the exercise of #1 above to evaluate your design.

A book printer can usually make up a dummy for you to the exact specs — trim, length, paper stock, and binding style. Might cost you a surcharge, depending on your account with the manufacturer, but worth it, imo.

Remember, with perfect binding, the signatures are trimmed on 4 sides. In the States, the gutter looses 1/8 of an inch. That's just gone. Then there is the "tightness" that perfect binding (or notch, for that matter) adds.

Even with the creased spine, the glue itself will add some stiffness. I haven't had to worry about this for a while, and understand there are now better glues. Still . . .

But perhaps I misunderstand -- usually, you can't perfect bind as few as 64 pages. I'd take 96 pages as the minimum. With that in mind, Kent's suggestion of having the printer do a dummy makes good sense to me.