One of the interesting things about the FDA as a government agency is that it largely seems to be trusted by Americans, and respected (if grudgingly) by the industry it regulates - possibly because its domain, medical science, is an area which requires non-fakeable technical expertise. Its "strange" respectability is highlighted in this book review.
J.S. Mill pointed out in On Liberty that medicine was a market ripe for failure, because most consumers of medicine are not in a position to evaluate its safety and efficacy. His observation presaged the problems that triggered the creation of the FDA (or its ancestor) in the early twentieth century: too many kids were dying from liver failure, brought on by the alcohol and ethylene glycol (antifreeze) that America's own witch doctors were using to sweeten their literal snake oil.
The drawbacks of having a central agency responsible for allowing the marketing of new drugs are several. First, like any agency, FDA is not immune to politics: for example, post-TGN-1412, scrutiny of any cytokine-interacting antibody increased to an almost paranoid degree (in corporatese this is known as a CYA). Perhaps more sinister on this count, it is often the case that a compound which receives marketing approval in its home market (where its developer is headquartered) does not receive it overseas (case in point, Paris-HQ'd Sanofi-Aventis and their weight-loss drug rimonabant, approved in Europe with a few side effects, and rejected in the U.S. for those same side effects. Coincidence?) Furthermore, all government agencies have limited resources, subject to the vagaries of central planning and economic swings. The decade began with an ominous trend in increasing approval times; this is the metric that individual biomedical companies care about. Much more alarming was the trend of decreasing numbers of new chemical or biological entities being submitted per year. This is the best index of overall biomedical innovation in a market and is affected not just by the agency but by the greater R&D environment of the industry - which of course, anticipates increasing conservatism in the agency that makes or breaks it, since you don't spend hundreds of millions chasing a compound that can't be legally marketed. Fortunately the NCE and NBE numbers are turning around; in 2009 there were 9 NCEs and 13 NBEs, compared to 1 NCE at one point in the mid-2000s.
It's important to point out that FDA approves drugs for a specific indication, not just for general sale, but that physicians are still allowed to prescribe them off-label. This creates strange pressures. First, the temptation for companies to actively market off-label is great. It's highly illegal, and highly lucrative (recent off-label marketing settlements in the last few years easily reach a half billion or more). More important is that this creates dangers for patients. Physicians have the authority to prescribe off-label at their discretion, and they often do, particularly with psychoactive drugs. While the companies that develop the drugs are obligated to produce general safety information, there are under no such obligations to produce safety information (much less efficacy) for off-label populations that physicians end up prescribing to, though they're stupid if they don't explore this as a potential new revenue source.
The point is this. By having an agency with such a structure - that approves drugs for specific indications - we create these pressures and these false senses of security (that off-label use is as well-characterized as the labeled indication). An agency that approved a drug only for safety, but not for efficacy, might be more beneficial to everyone. In other words, the FDA would require the company to label their drug as not harmful within certain dosing parameters, and contraindicated with certain other medicines. Since we're already trusting physicians to read NEJM and prescribe the drug on that basis, isn't this a far more efficient way to get drugs on the markets to patients who need it, without compromising safety? This discussion is a non-starter in the current administration but perhaps in a vigorous developing economy with a sense for social experimentation, we might see whether this would provide a greater good for a greater number; that is, increase innovation speed without increasing risk to patients. With ongoing drug research, overconservatism and the slowing of innovation is absolutely an ethical problem, because we can't forget the patients who haven't yet gotten sick.