Tolerance develops to benzodiazepines-- and every other antieplieptic, according to the new Epilepsia article.

 In general, efficacy to all AEDs decreases with long term exposure.  That's tolerance.  If being on an AED reduces seizures by 50%, then tolerance is defined as occuring when you return to less than 50% reduction of symptoms.  Thus defined, tolerance (of such severity that increased doses do not help) occurs in  10-50%.
 

Worse, there appears to be cross tolerance.  For example, and likely most significant for psychiatric patients, Depakote "lost >50% of its anticonvulsant efficacy in mice pretreated twice daily for only 3 days with [benzos]."

Why does tolerance occur?  On the one hand is the obvious metabolic concern-- autoinduction of hepatic enzymes-- but this is really only relevant with the first generation drugs (and especially CBZ and phenobarbital (these are such powerful inducers of cytochrome enzymes that they actually induce their own metabolism)-- while  Depakote is the opposite (inhibitor of cytochromes-- which is why you must reduce the initial doses of Lamictal when given with Depakote, so as not to "overdose" and increase the risk of rash)).  On the other hand are  pharmacodynamic effects, which are of three types: downregulation of binding sites; functional uncoupling (on the GABA-A receptor, benzo binding has less of a positive allosteric effect on GABA binding); downregulation of or decreased sensitivity of ion channels (for example Neurontin downregulates Ca+ channels, benzos reduce Cl- channels, etc.)  Activity on the ion channels (as opposed to receptors) would partially explain cross tolerance since these ion channels are the downstream target of many drugs.

No, wait, there's a fourth reason for "tolerance:" maybe the seizure disorder itself changes over time, so it looks like you became tolerant, but really you have a "new" seizure disorder.   This is analogous to bipolar disorder, which evolves over time-- how you present at 25 may be different than 35; your manias are different, etc.

So now we have a problem: is there any reason to think that tolerance to the antimanic/antidepressive effects of AEDs  wouldn't occur?  If seizures, why not mania?   If mania is a strictly biochemical dysfunction in the brain, shouldn't tolerance to its treatment occur?  Do we make patients worse by keeping them on the meds?  Or at least harder to treat?  And if mania isn't strictly biochemical-- if we're allowing that life happens-- do we really believe that a fixed dose of an anti-epileptic administered over years  is going to prevent a negative response to a life event?  And wait a second-- doesn't mania spontaneously remit even without medication?  Shouldn't we just, sort of, help nature along, or even get out of its way?

I'm not saying not to treat-- I'm saying not to overtreat,

A guy is on 1500mg Depakote today.  What do you do when the patient relapses?   Increase to 2000mg?  Then what?  When does it stop?  When does this practice not ultimately result in polypharmacy?

Any reason-- biochemical or epidemiological, I'll take any offer-- why we should not be treating symptomatically rather than prophylactically?   Antimanics when you're manic, then stop them when you're better? 

I know everyone thinks Osler helped write the DSM after finding the gene for psychiatry and Hippocrates is jealous because he's balding junior faculty , but perhaps we should go reread The Epidemics and rethink our principles.