Only three of the events were withdrawals, meaning that only 1.3% of approved drugs were recalled off the market. Consider that this is a really impressive error rate. Scientific studies are allowed to be wrong 5% of the time; the FDA is wrong 1%.
The others were “boxed warnings” and “safety communications”. These are often pretty minor. I can’t read the full study you linked, so I don’t have their examples, but some of the most recent FDA boxed warnings are:
Last year, the FDA added a new boxed warning to all opiates and benzodiazepines (drugs which have been around for the better part of a century) telling doctors that really, no, seriously, don’t prescribe these together. People have known not to prescribe them together for the better part of a century, but the FDA decided they should be louder about it so they added a new boxed warning. Also, I’ve prescribed opiates and benzodiazepines together several times when it’s been clinically necessary, and as long as you’re not an idiot about it it’s fine (but please don’t try mixing opiates and benzodiazepines at home without your doctor’s permission!).
There’s a boxed warning on Ritalin that it might cause heart problems in kids. Later research has shown this is probably false, but the FDA worried about it one time ten years ago and decided to add a boxed warning, which nobody ever got rid of. Ritalin remains as popular as ever.
In 2006, the FDA added a boxed warning to warfarin, a blood thinner which at that time had been out 52 years, saying that if you took too much of it, it could make your blood too thin.
This year, the FDA released a safety communication, discussing a very common antiseptic used by millions of people yearly, that, over the last fifty years of use, about one person per year has had a serious allergic reaction to it, and if you have a serious allergic reaction to it, you should call 911.
Maybe another way of making this point is to list the psychiatric drugs that have gotten boxed warnings sometime after they were released: every antidepressant, every typical antipsychotic, every atypical antipsychotic, every benzodiazepine, lithium, Depakote, Lamictal, etc. You may recognize this as every single psychiatric drug (except BuSpar, which doesn’t work).
What I’m saying is that an FDA boxed warning isn’t “Oh god, how did this poison ever make it through the approval process?!”. It’s an incremental update to existing safety regulations on a generally good drug as more evidence comes in and people’s priorities shift.
The study says on average these drugs were out for four years before the FDA added the warning. That’s probably because these are rare side effects that only appear after many people have been taking the drugs for several years. For example, if one in every million people who uses antiseptic soap has an allergic reaction, we can’t very well ban antiseptic soap until we try it on a million different people. That’s why post-marketing surveillance is a natural and important part of the regulatory process.
Another example: antipsychotics are not FDA-approved for dementia. I don’t even know if any company has ever tried to get antipsychotics FDA-approved for dementia. Giving demented people antipsychotics is a bad idea if there’s any other option. But people do it anyway, and then the demented people get heart problems and die. The FDA eventually figured that out and put a boxed warning on antipsychotics not to give them to demented people if possible. This didn’t show up before approval because nobody wanted them to be given to demented people so there was no study about whether it was a good idea.
My point is that boxed warnings are a natural part of pharmacology, that some new side effects being found post-approval means the system works, and that the FDA continues to have a 98.7% accuracy in approving drugs that don’t need to be withdrawn later.