CHAPTER 1 Death by Bureaucrat How would you rather die? From a lethal reaction to a drug prescribed by your doctor? Or because your doctor failed to prescribe a drug that would have saved your life? If this choice sounds like one you would rather not make, consider this: Employees of the Food and Drug Admin- istration (FDA) make that decision on behalf of millions of Americans many times each year. More precisely, FDA bureaucrats decide whether or not new medicines (prescription drugs) should be allowed to go on sale in the United States. If the FDA rules against a drug, physicians in America may not legally prescribe it, even if thousands of lives are being saved by the drug each year in other countries. A BRIEF HISTORY OF THE FDA The FDA's authority to make such decisions dates back to the passage of the Food and Drug Safety Act of 1906. The law required that medi- cines be correctly labeled as to their contents and that they not contain any substances harmful to consumers' health. Due to this legislation, Dr. Hostatter's Stomach Bitters and Kickapoo Indian Sagwa, along with numerous rum-laden concoctions, cocaine-based potions, and supposed anticancer remedies disappeared from druggists' shelves. The law was expanded in 1938 with the passage of the Food, Drug, and Cosmetic Act, which forced manufacturers to demonstrate the safety of new drugs before being allowed to offer them for sale. (This law was prompted by the deaths of 107 people who had taken Elixir Sulfanilamide, an anti- biotic that had been errantly mixed with poisonous diethylene glycol, a chemical cousin of antifreeze.)CHAPTER ONE Death by Bureaucrat 5 The next step in U.S. drug regulation came after a spate of severe pressure) was delayed nearly a decade in America relative to Europe. birth defects among infants whose mothers during pregnancy had According to several researchers, the lag in the FDA approval of these taken a sleep aid known as thalidomide. When these birth defects first drugs cost the lives of at least 250,000 Americans, became apparent, the drug was already widely used in Europe and Can- ada, and the FDA was nearing approval for its use in America. In fact. TERRIBLE TRADE-OFF about 2.5 million thali mide tablets were already in the hands of U.S. physicians as samples. The FDA ordered all of the samples destroyed In effect, the law requires FDA bureaucrats to make what is truly a terri- and prohibited the sale of the drug in the United States. This incident ble trade-off. Lives are saved becaus safe or ineffective drugs are kept led to the 1962 Kefauver-Harris Amendments to the 1938 Food, Drug. off the market, but the regulatory process delays (or even prevents) the and Cosmetic Act, radically altering the drug-approval process in the introduction of some safe and efficacious drugs, thereby costing lives. United States. Let us now take a more systematic Every time a new drug is in there is a chance that it should not have been-either becau THE IMPACT OF THE 1962 AMENDMENTS has adverse side effects that outweigh the therapeutic benefits t is not safe) or because it really Before the 1962 amendments, the FDA was expected to approve a new does little to help the individuals ke it (it is not effective). When drug application within 180 days, unless the application failed to show such a drug is introduced, we say that a Type I error has been com- that the drug was safe. The 1962 amendments added a "proof of effi- mitted. Since 1962, the incidence -the thalidomide cacy" requirement and also removed the time constraint on the FDA. The possibility-has been reduced the added testing required by the FDA has free rein to determine how much and what type of evidence it FDA. But other people have been he victims of what is called Type II will demand before approving a drug for sale, and thus may take as long error. Their cost is the pain, suffering, and death that occur because as it pleases before either granting or refusing approval. the 1962 amendments have prevented or delayed the introduction of The 1962 amendments d increased the costs of introducing safe, efficacious drugs. Type II error-as with Septra or beta block- a new drug and markedly slowed the approval process. Before 1962, for ers-occurs when a drug should be introduced but is held back by example, the average time b filing and approval of a new drug FDA regulation. application was seven months. 1967, it was thirty months; and by the Over the past thirty year cries over the harm caused by the late 1970s, it had risen to eigh years. The protracted approval pro- drug lag have sometimes induced the agency to shorten the testing cess involves costly testing by the drug companies-more than $1 billion period when the costs of Type I error are small relative to the dam- for each new drug-and delays the receipt of any potential revenue from ages due to Type II error-as with terminally ill patients. One famous new drugs. Both the delays and the higher costs reduced the expected example involved azidothymidine (AZT), which emerged as a possible profitability of new drugs, so fewer new drugs have been brought onto treatment for AIDS. Gay men, among whom AIDS was most prevalent the market. at the time, took the lead in pres ring the FDA to approve the drug Debate continues over how much FDA regulation is needed to quickly, and the FDA responded accordingly, giving it the OK after ensure that drugs are both safe and efficacious, but there is little doubt only eighteen months of testing. Similarly, Taxol, an important new that the 1962 amendments resulted in a U.S. "drug lag." On average, drug used to treat breast cancer, received expedited review by the FDA drugs take far longer to reach the arket in the United States than they because of pressure applied by women in whose families there was a do in Europe. Admittedly, it takes time to ensure that p enefit history of breast cancer. from, rather than are harmed by, new drugs, but regulation-induced drug The FDA now has a formal program in which it seeks to expe lag can itself be life-threatening. Dr. George Hitchings . a winner of the dite testing for drugs that seem to offer great promise for alleviating Nobel Prize in Medicine, estimated that the five-year lag in introduc- death or suffering. Nevertheless, although the average approval time ing Septra (an antibiotic) to the United States killed 80,000 people in for new drugs has shortened considerably, it still takes roughly ten this country. Similarly, the introduction of a class of drugs called beta times as long for a new drug to be approved as it did before the 1962 blockers (used to treat heart attack victims and people with high blood amendments. 6 CHAPTER ONE Death by Bureaucrat LESSONS FROM THE FDA STORY 5. Things aren't always as they seem. Many analyses of the effects of What can we learn from the FDA regulation of new drugs that will guide government policies take an approach that does not fully recognize us in thinking about other public issues of our time? There are several the actions that people would otherwise have taken. Thus, official key principles: pronouncements about the effects of policies routinely misrepre t-not because there is necessarily any attempt to 1. There is no free lunch. Every choice, and thus every policy, entails deceive but bec ause it is often difficult to know what would have a cost-something must be given up. In a world of scarcity, we can- happened otherwise. Pharmac manufacturers, for example, not have more of everything: so to get more of some things, we must have strong incentives cing drugs that are unsafe give up other things. Although FDA review of drugs saves lives by or ineffective bject to loss of reputa- preventing the introduction of unsafe or ineffective drugs, the cost tion and to lawsuits. For sim sons, physicians have strong is billions of dollars of added expenses, plus delayed availability of incentives to avoid prescribing such drugs for their patients. Even safe and efficacious drugs, resulting in the deaths of hundreds of without FDA regulat would thus be extensive testing of thousands of people. new drugs before their introduction. Hence, it is incorrect to ascribe 2. The cost of an action is the alternative that is sacrificed. Econo- the generally safe and effective nature of modern drugs entirely mists often express costs (and benefits) in dollars because this is a to FDA protection. The flip side, however, is that the drug devel- simple means of accounting for and measuring them. That does not opment process is inherently long, complicated, and costly. Even mean that costs have to be monetary, nor does it mean that ec without FDA oversight, some people would die waiting for new ics is incapable of analyzing costs and benefits that are quite human. drugs because self-interested manufacturers would insist on some The costs that led to the 1938 and 1962 amendments were the very testing, and cautious physicians would proceed slowly in prescribe visible deaths caused by sulfanilamide and the terrible birth defects ing new drugs. due to thalidomide. nt revisions to the FDA process for The people who work at the FDA (and members of Congress) reviewing drugs, as with AZT and Taxol, have been in response to are publicly castigated when they "allow" a Type I error to occur the deaths and other adverse health effects caused by the regulation- especially when it is a drug that kills people. Thus, FDA bureau- induced drug lag. crats have a strong incentive to avoid such errors. But when testing delays cause a Type II error, as with Septra, it is almost impossible 3. The relevant costs and benefits are the marginal (incremental) ones. to point to specific people who died because the drug was delayed. The relevant question is not whether safety is good or bad; it is As a result, officials at the FDA are rarely attacked directly for such instead how much safety we want-which can only be answered delays. Because the costs of Type II errors are much more difficult by looking at the added (marginal) benefits of more safety com- to discern than the costs of Type I errors, there is an inherent bias at pared to the added (marginal) costs. One possible response to the the FDA in favor of being "safe rather than sorry"-in other words, sulfanilamide poisonings or thalidomide was to have outlawed new excessive testing. drugs altogether. That would guarantee that people would be harmed by new drugs. But surely this "solution" would not be 6. Policies always have unintended consequences, and, as a result, sensible, because the marginal cost (more Type II errors) would ex- their net benefits are almost always less than anticipated. In the ceed the marginal benefit (fewer Type I errors). case of government regulations, balancing incremental costs and benefits (see principle 3) fails to make good headlines. Instead, 4. People respond to incentives. This is true whether we are talking what gets politician d regulators promoted are abso- about consumers, suppliers, or government bureaucrats. Here the lute notions such as safety (and motherhood and apple pie). Thus, incentive to amend the law in 1938 and 1962 was the very visible if a little safety is good, more must be better, so why not simply h and disfigurement of individuals. The eventual FDA decision mandate that drug testing "guarantee" that everyone is free of risk ced up the review process was prompted by intense lobbying by from dangerous drugs? Eventually, the reality of principle 3 sinks riduals who believed (correctly, as it turned out) that they might in, but in this instance not before the drug lag had killed many fit from drugs not yet approved. people.20 CHAPTER THREE Flying the Friendly Skies? 21 Interestingly, evidence shows that consumers are indeed cognizant capacity), (ii) the experience levels of their pilots, (iii) the age of of the safety performance of airlines and that they "punish" airlines that their planes, and (iv) the length of the typical route they fly. Which perform in an unsafe manner. Researchers have found that when an air- airlines would be likely to have the fewest fatal accidents? Which line is at fault in a fatal plane crash, consumers appear to downgrade would be expected to have the most? their safety rating of the airline (i.e., they revise upward their estimates of the likelihood of future fatal crashes). As a result, the offending airline 3. Is safety likely to be a "normal" good (i.e., something people want suffers substantial adverse financial consequences over and above the to consume more of as they get richer)? Use your answer to this costs of losing the plane and being sued on behalf of the victims. These question to predict likely safety records of airlines based in North findings suggest a striking degree of safety awareness on the part of sup- America and Europe, compared to those based in South America posedly ignorant consumers. and Africa. Then go to www.airsafe.com to see if your prediction is confirmed or refuted by the facts. WHAT ABOUT TERRORISM? 4. Many automobile manufacturers routinely advertise the safety of their cars, yet airlines generally do not mention safety in their adver- Of course, this discussion leaves open the issue of how to handle safety tising. Can you suggest an explanation for this difference? threats posed by terrorists and other miscreants. For example, much of the information that goes into asse errorist threats is classified as 5. Many economists would argue that private companies are likely to secret, and its revelation to airline s might compromise key be more efficient than the government at operating airlines. Yet many sources of the data. Hence, there could be an advantage to having the economists would also argue that there is a valid reason for govern- government try to approximate the ient safety outcome by mandat- ment to regulate the safety of those same airlines. Can you explain ing certain screening provisions caling exactly why they are why the government might be good at ensuring safety, even though being chosen. Similarly, beca connected in networks (so it might not be good at operating the airlines? that people and baggage move from airline to another in the course 5. Professional football teams sometimes charter airplanes to take them of a trip), achieving the efficient set of screening rules for all airlines. Even so, this does not inform us to their away games. Would you feel safer on a United Airlines plane whether the government shot that had been chartered by the Washington Redskins than on a regu- pose those rules or the airlines should come to a voluntary joint larly scheduled United Airlines flight? We began this chapter with commonplace observation that air- lines are safer than cars. Yet many people still worry for their safety every time they get on an airplane. Are they being irrational? Well, the answer, it seems, is in the eye of the beholder. Measured in fatalities per mile traveled, airplanes are about 15 times safer than cars (and 176 times safer than walking, we might add). This number masks the fact, however, that 68 percent of aircraft accidents happen on takeoff and landing, and these operations occupy only 6 percent of flight time. It is presumably this fact that quite sensibly makes people nervous whenever they find themselves approaching an airport. DISCUSSION QUESTIONS 1. Is it possible to be too safe? Explain what you mean by "too safe." 2. Suppose it is possible to observe (or measure) four attributes of air- lines: (i) the size of their planes (measured in passenger-carrying CHAPTER THREE Flying the Friendly Skies? 1 produce important safety benefits. But how should we achieve this? safety. Consumers would simply observe the safety offered by different Should carry-on bags be prohibited or just examined more carefully? airlines, the prices they charge, and then select the degrees of safety that How thoroughly should checked luggage be screened for bombs? Even best suited their preferences and budgets, just as with other goods. Infor- now, our answers to these questions are evolving as we learn more mation is not free, however. It is a se ree good, costly to obtain. As a about the extent of the threat and the costs of alternative responses to result, passengers may be unav fety record of various airlines it. Nevertheless, throughout the process, economic principles can help or the compet ance procedures of an us make the most sensible deci airline's mechan themselves may be uncer- In general, the efficient level of safety will not be perfect safety tain about the efficient level of rhaps because they have no because perfection is simply too costly to achieve. For example, to be way of estimating cks, for example. Such absolutely certain that no one is ever killed or injured in an airplane possibilities have been used to argue that the federal government should crash, we would have to prevent all travel in airplanes-an unrealistic and mandate certain minimum levels of safety, as it does today through the impracticable prospect. This means that if we wish to enjoy the advan- operation of the Federal Aviation Administration (FAA). Let's look at tages of flying, we must be willing to accept some risk-a conclusion this issue in some detail. that each of us implicitly accepts every time we step aboard an airplane. One argument in favor of government safety standards rests on the assumption that when left to their own devices, airlines would provide THE IMPORTANCE OF CIRCUMSTANCES less safety than passengers want. This might happen, for example, if customers could not tell (at a reasonable cost) whether the equipment, Changes in circumstances can alter the efficient level of safety. For training, and procedures employed by an airline are safe. If passengers example, if a technological change reduces the costs of bomb-scanning cannot cheaply gauge the level of safety, they will not be willing to equipment, the marginal costs of preventing terrorist bomb attacks will reward airlines for being safe or punish them for being unsafe. If safety be lower. It will be efficient to have more airports install the machines is costly to provide and consumers are unwilling to pay for it because and to have extra machines at la speed the screening pro- they cannot accurately measure it, airlines will provide too little of it. cess. Air travel will become technological change. The conclusion is that government experts, such as the FAA, should set Similarly, if the marginal benefits safety rise for some reason (perhaps safety standards for the industry. because the president of the United S s on board), it could be effi- cient to take more precautions, resulting in safer air travel. Given the fac- tors that determine the benefits and costs of safety, the result of a change DO CONSUMERS KNOW BEST? in circumstances will be some determinate level of safety that generally will be associated with some risk This argument seems plausible, but it ignores two key points. First, how of death or injury. Airplanes are complex systems, and an amazing number of com- is the government to know the efficient level of safety? Even if the FAA ponents can fail. Over the century or so that humans have been flying, knows the costs of all possible safety measures, it still does not have airplane manufacturers and airlines have studied every one of the mal- enough information to set efficient safety standards because it does not know the value that people place on safety. Without such information, functions thus far and have put into place design changes and operating procedures aimed at preventing recurring error. The efforts have paid the FAA has no way to assess the benefits of additional safety and hence off. Between 1950 and 2010, for example, the fatal accident rate on U.S. no means of knowing whether those benefits are greater or less than the commercial airlines was cut by almost 97 percent. added costs. Second, people want to reach their destinations safely. Even if they cannot observe whether an airline hires good pilots or bad pilots, they DOES THE GOVERNMENT KNOW BEST? can see whether that airline's planes land safely or crash. If it is safety that is important to consumers-and not the obscure, costly-to-measure Consumers have the greatest incentive to ensure that air travel is safe. reasons for that safety-the fact that consumers cannot easily measure and if information were free, we could assert with some confidence that metal fatigue in jet engines may be totally irrelevant to the process of the actual level of safety supplied by firms was the efficient level of achieving the efficient level of safety