It's the LF noise (0.1Hz to 10Hz band) that gets you. You might try one of the newer high-voltage (works at >5V) chopper amps like the LTC2057, or an OPA189 (and I think ADI has one too-- but can't remember the part number right now).
The Analog part is the ADA4522 series, available in singles, duals, and quads. The 0.1Hz to 10Hz noise is specified at 117nV p-p typical, which is about half of the LTC2057 at 200nV p-p. The voltage noise is also about half of the LTC2057 as well, so it can be a useful amplifier.
I'm not certain of this, as it's not directly specified, but it seems likely that all amplifiers in a dual or quad will run on the same chopper clock, making it easier to use two or more amplifiers in the same circuit without any chopper frequency beating artifacts. I'm using an ADA4522-2 in a DC servo design, and have found that when used as an integrator, it puts out no detectable switching noise whatsoever. Sure, my integrators have a corner of around 24Hz, and this should lowpass the 800kHz and 4.8MHz artifacts into oblivion, but still, no artifacts larger than around 200nV could be found up to 1.2MHz using an Audio Precision test set, and it's more likely that these were leakage from broadcast AM stations into the test circuit and cabling, since the spur frequencies were not related to those used in the ADA4522, and lined up with local AM broadcast stations.
So, it's a nice amplifier, and it's worth considering for any sorts of low noise, low offset and low drift uses. The specifics might favor the 2057 over the 4522, but it's worth evaluating IMHO, especially if you need to use more than one channel at a time.