Instrumentation amplifiers are designed to work down to DC, which is effectively what your application is. You'll basically need to look at all the sources of error (offset voltage drift will be key) and assume they can drift over their full range over one period of your measurement.
Typically, your ability to control the operating temperature of your system will be key to obtaining good accuracy.
Luckily, this is exactly the sort of thing in-amps are designed for, and they will often be spec'ed for things like temperature coefficient of the voltage offset, etc.
This is just to point out one more thing you want to look out for.
Here's the typical spectral input noise for the AD8235 StevenVH recommended:
You notice that the noise is increasing at low frequencies below about 10 Hz. This increasing noise at low frequencies is a very common (universal?) feature of op-amps and in-amps, called 1/f noise. The name comes because this noise source has an rms amplitude roughly proportional to 1/f. So you can expect this curve to continue increasing in a well-defined way as the frequency drops to 1 Hz, 0.1 Hz, 0.01 Hz, etc where you say your application is operating.
However, depending on the manufacturing process, etc, different amplifier types will have different levels of 1/f noise. You will want to look at this behavior as a key parameter for selecting an amplifier for your design. If some data sheet doesn't report the noise at all at low enough frequencies to show 1/f noise, that is not a part you want to use here. That said, the AD8325 does appear to be very good in this regard, with its "knee" at only 10 Hz.