[refurb]

There have been a few examples where the FDA received complaints, looked at the data and asked the generic manufacturer to rest test/reformulate their drug since it wasn't therapeutically equivalent.

The best example is generic Wellbutrin[1]. Many patients complained of side-effects from generics that they didn't experience with the brand-name drug. The FDA did testing and realized that the generic Wellbutrin extended-release resulted in much high blood concentrations than the branded version. The generics in questions were withdrawn from the market.

The interesting thing is that to get approval for an "AB-rated" (therapeutically equivalent) generic drug, you don't have to be identical. The FDA only requires that the generic drug deliver 80-125% of the drug the branded drug does[2]. Normally that range is fine, but for some drugs (antiepileptics are a good example) the differences can produced a less than equivalent therapeutic outcome.

[1]http://www.fda.gov/Drugs/DrugSafety/PostmarketDrugSafetyInfo...

[2]http://www.nature.com/clpt/journal/v94/n4/full/clpt2013104a....