You really need to be a LOT more specific about exactly what you are doing and what considerations you've already made about it. Disclose, don't obfuscate. What are you doing and what are the primary considerations?
In a case where my problem was absolute accuracy from femptoamps to microamps (couldn't use a log-amp, for example), I used Coto relays at the first stage to change transimpedance gains. Nothing else would do the job properly. In another case, an integrator for the same purpose, charge injection was a serious problem for any discrete FET. A MOSFET generally does NOT remove and add equal amounts of charge, going on and off, for example. You can work hard to design them that way, but it takes custom design work and access to a brokered FAB process.
Anyway, you need to state exactly what you are doing.
EDIT: How low of a voltage? Bipolar input? What resolution and precision? What accuracy? What are you doing about offset voltage? (Reset phase? How?) And input bias currents (as I gather you are converting the low voltages to currents with a resistor?) What about noise? What opamp? What frequency of measurement? (Dual slope is generally slow, but must ask.) Why not consider a coto relay (better Coto relays tend to specify leakage resistances in the 10^12
ohms region, mercury wetted, and just plain bulky?)