You really need to determine what is causing the variation in order to determine what is an appropriate way to compensate.
Transistor temperature variation can be compensated with appropriate bias compensation. Gm tends to be CTAT when the transistor is bias with a constant current, so a PTAT bias current can stablize Gm over temp (so called constant Gm biasing if you use an external resistor, an external resistor is not always needed to stablize gain).
http://www.designers-guide.org/Forum/YaBB.pl?num=1199713543Apparently you can read an excerpt from Tom Lee's RFIC book on Google to read more about constant Gm biasing.
Load tempertature and process variation can partially be compensated by making the bias current depend on the same type of resistor as the load. This works out fine for resistor loads, but not so well for inductor loads.