No one may sell a new animal drug, or feed containing a new animal drug, without the approval of the Food and Drug Administration. 21 U.S.C. Sec. 360b. Obtaining approval takes a long time and costs a lot of money, for the FDA requires thorough experimentation to determine both the drug's effects on animals and whether its residues persist in the animals and enter the food chain. A drug that leaves potentially-dangerous residues will not be approved. Scientists' ability to determine the long-range effects of many chemical substances is limited; that lack of knowledge, coupled with the FDA's cautious approach, means that few new animal drugs have been approved. We must take it as given that for significant diseases there are no effective FDA-approved drugs. Acute mastitis, which may decimate a herd of dairy cattle in short order, cannot be treated with approved new animal drugs; neither can shipping fever and salmonella in feedlot cattle. The FDA has not approved any general anesthetic for field use in surgery on food animals. For the principal diseases of non-food animals, such as mink and fox, there are few, if any, approved remedies.

Many veterinarians find this state of affairs deplorable. Because they cannot buy in finished form the drugs they think they should be able to use, they have elected to make their own. They purchase the active ingredients, mix them in the proportions they think best, and administer their concoctions as professional judgment dictates. The veterinarians do not sell the drugs, so the FDA's usual methods of control do not come into play. The FDA looked the other way for decades. Cf. Heckler v. Chaney, 470 U.S. 821, 105 S. Ct. 1649, 84 L. Ed. 2d 714 (1985). Recently, however, the FDA became concerned about the end-run around its approval process--perhaps because the practice has become widespread--and decided to put a stop to it. The FDA might have prosecuted thousands of individual veterinarians for administering unsafe drugs (any new animal drug is deemed unsafe unless approved by the FDA, 21 U.S.C. Sec. 360b(a)), but the agency thought it easier to cut off the veterinarians' supply of active ingredients. This case is the result.

The FDA decided to intercept the drugs in the hands of the middleman: in this case Schuyler Laboratories, Inc., which purchased unblended ingredients from their manufacturer and resold them, still in bulk, to veterinarians. Such a middleman does not manufacture drugs and therefore ordinarily does not need to seek the FDA's approval via new animal drug applications. (Things may be different when the drug is an antibiotic listed in 21 U.S.C. Sec. 321(w)(3); we shall not have to decide.) In July 1986 the FDA swooped down on Schuyler and seized every container of drugs on the premises. It carted off 52 lots and filed a complaint seeking their forfeiture as "misbranded" drugs. See 21 U.S.C. Sec. 334(a)(1). (How the caption of this case came to be as it is baffles us.) The bulk drugs were "misbranded", according to the FDA, as a result of this requirement in 21 U.S.C. Sec. 352:

A drug or device shall be deemed to be misbranded--

* * *

(f) Unless its labeling bears (1) adequate directions for use; ... Provided, That where any requirement of clause (1) of this subsection, as applied to any drug or device, is not necessary for the protection of the public health, the Secretary [of Health and Human Services] shall promulgate regulations exempting such drug or device from such requirement.

The labels of the containers on Schuyler's premises gave the name of the drug and its manufacturer, but no directions for use. It would be difficult to create labels with such directions, because bulk drugs are to be mixed or otherwise compounded rather than used in their current form. The drugs are therefore "misbranded" unless Schuyler can take advantage of the proviso.

The regulation implementing the proviso states:

A drug in a bulk package, except tablets, capsules, or other dosage unit forms, intended for processing, repacking, or use in the manufacture of another drug shall be exempt from Sec. 502(f)(1) [21 U.S.C. Sec. 352(f)(1) ] of the act if its label bears the statement "Caution: For manufacturing, processing, or repacking"; .... But the exemption shall not apply to a substance intended for a use in manufacture, processing, or repacking which causes the finished article to be a new drug, unless:

(a) An approved new drug application or new animal drug application covers the production and delivery of the drug substance to the application holder by persons named in the application....

21 C.F.R. Sec. 201.122. Schuyler's bulk drugs do not qualify under this regulation as written, because although they are "intended for use in manufacture" (etc.) Schuyler has not offered to show that an approved new animal drug application covers the substance that will ensue. Schuyler argues that the qualification in the regulation starting with "But the exemption shall not apply ..." is unlawful because not "necessary for the protection of the public health" within the meaning of Sec. 352(f). In other words, Schuyler wants the benefit of the first sentence of Sec. 201.122 without the burden of satisfying the remainder of the regulation. Everything after the first sentence, Schuyler believes, must be discarded; the bulk drugs would come within the remaining language.

The district court agreed with Schuyler and ordered the FDA to return the drugs. 674 F. Supp. 1344 (C.D.Ill.1987). Accord, United States v. Algon Chemical Inc., 689 F. Supp. 394 (D.N.J.1988). As it saw things, the limits on the permission to sell unlabeled bulk drugs are neither necessary nor prudent for three reasons: Congress wanted to leave medical professionals free to practice as their judgment required; veterinarians are not regulated in compounding drugs, implying liberty to obtain drugs to compound; and the middleman cannot know whether its customer has the necessary approval. We stayed the district court's order pending appeal.

The statute treats drugs without directions for use as presumptively misbranded. Schuyler bears the burden of establishing its entitlement to treatment under the exception. United States v. An Article of Device "Toftness Radiation Detector", 731 F.2d 1253, 1260-62 (7th Cir.1984). Whether there is a safe harbor for Schuyler to seek depends in the first instance on the FDA's judgment about the conditions necessary to protect the public health. The phrase "necessary for the protection of the public health" in Sec. 352(f), like the phrase "public interest, convenience, and necessity" in other regulatory statutes, delegates authority to an agency. See Chicago & North Western Transportation Co. v. Kalo Brick & Tile Co., 450 U.S. 311, 321, 101 S. Ct. 1124, 1132, 67 L. Ed. 2d 258 (1981); Colorado v. United States, 271 U.S. 153, 168-69, 46 S. Ct. 452, 455-56, 70 L. Ed. 878 (1926); Black v. ICC, 737 F.2d 643, 650 (7th Cir.1984). It requires the agency to make a judgment about where the public interest lies. Courts defer to the FDA when it construes its governing statutes. E.g., Young v. Community Nutrition Institute, 476 U.S. 974, 106 S. Ct. 2360, 90 L. Ed. 2d 959 (1986); cf. Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984); Continental Air Lines, Inc. v. Department of Transportation, 843 F.2d 1444, 1448-54 (D.C.Cir.1988). The FDA's task was not, however, to "construe" a statute. It was to make a judgment about a medical subject. Someone must draw on knowledge about risks and needs and decide how much risk is too much; that someone is the Executive Branch of the government. "[T]he first question in determining the deference appropriate to the agency[] ... is whether Congress has transferred discretion to the agency. If the legislation either calls for the agency's decision or contains no disposition of the subject, then the agency has been deputized to make a rule, and its decision should be respected." Homemakers North Shore, Inc. v. Bowen, 832 F.2d 408, 411 (7th Cir.1987). So Schuyler is fighting a doubly-uphill battle: it bears the burden of establishing an exception to the labeling requirement, which it can get only by showing that the FDA's views about the needs of public health are arbitrary and capricious.

The district court thought that they are because "it was not the purpose of the Act to involve the agency in the practice of the healing arts." 674 F. Supp. at 1348. If not that, however, what does the statute do? Congress gave the FDA comprehensive powers to license the manufacture of drugs and limit their sales. To regulate drugs is to be "involved" in the "practice of the healing arts." True, the Senate report that accompanied the original statute in 1935 said that the bill was "not intended as a medical practices act and [would] not interfere with the practice of the healing art[s]", S.Rep. No. 361, 74th Cong., 1st Sess. 3 (1935), but the statute has undergone much amendment since then, including a complete overhaul in 1962. Phrases such as the one in the 1935 Senate report--not repeated in 1962--never meant more than that medical licensure and discipline would continue to be the states' business; states, not the FDA, would decide whether (for example) physicians had selected wisely from among methods of treatment. The full quotation, redacted by Schuyler and the district court, makes the point: the statute is "not intended as a medical practices act and [would] not interfere with the practice of the healing art by chiropractors and others ". So states set medical qualifications, and practitioners who do not use drugs (i.e., chiropractors) may go on as before. Nothing in the history or structure of the Act permits drugs deemed ineffective or dangerous by the FDA to be available for use. United States v. Rutherford, 442 U.S. 544, 99 S. Ct. 2470, 61 L. Ed. 2d 68 (1979).

We do not question the good will of physicians, veterinarians, and other medical professionals. Still, no individual veterinarian will possess complete knowledge of the efficacy of a novel drug compound--let alone of its persistence in the food chain. The testing required to obtain a new animal drug approval is costly and extended precisely because information is so difficult to come by. Testing must isolate the effects of the drug in question from all other environmental influences, then follow the animals for years (even generations of animals) to identify the consequences. This requires data from large populations of animals and the application of powerful statistical techniques. No solitary medical professional can carry out this program of knowledge acquisition for even one drug, let alone for the bevy of drugs a veterinarian may choose to compound. All the good judgment in the world is useless when we lack essential knowledge.

Schuyler (and the veterinary groups supporting it as amici curiae ) responds that if veterinarians err in compounding and administering novel drugs, they will be subject to professional discipline, and the sellers of the afflicted animals to tort liability. All very well, if the sellers and veterinarians had the wealth to pay what could turn out to be staggering tort judgments, and if it should prove possible to trace cancers developing 20 years hence to residues in particular animals treated by particular veterinarians. But the sellers and veterinarians do not have the wealth needed to compensate all potential victims fully, and it would be difficult--more likely impossible--to attribute loss to a particular source many years after the fact. One need only think of the complexities of DES litigation to see a shadow of what would happen if courts had to trace drugs back through the food chain to find their sources. Even though DES was administered directly to patients, it is often impossible to identify the seller of the particular drug, and identified sellers may be out of business.

Some of the bulk drugs seized from Schuyler pose risks. Schuyler was holding for sale dimatridazole and nitrofurans, both of which have been found to cause cancer in animals and inferentially to pose a risk to humans. The FDA has withdrawn existing approvals to sell finished drugs containing dimatridazole, 52 Fed.Reg. 25312 (1987), and has proposed to withdraw approvals for finished nitrofuran products, 49 Fed.Reg. 34965 (1984). Schuyler nonetheless believes that veterinarians (and other medical professionals) should be free to compound drugs containing these ingredients, in their sole judgment. Reasonable persons could think it appropriate to rely on tort law and professional discipline, despite the difficulties we have outlined, to assure the safety of drugs used in food animals (and for other purposes), but Congress and the FDA have reached a different conclusion.

The district court believed, however, that Congress made an earnest of its desire to steer clear of regulating medical judgment by excluding veterinarians from several features of the regulatory system. For example, although the manufacture of drugs is tightly controlled, Congress exempted from these controls practitioners who "prescribe or administer drugs ... and who manufacture, prepare, propagate, compound, or process drugs ... solely for use in the course of their professional practice." 21 U.S.C. Sec. 360(g)(2). The district court reasoned: "It makes no sense for Congress to grant this exemption ... if it then takes away the very drugs that would be compounded." 674 F. Supp. at 1349.

This conclusion depends on a suppressed assumption: that Sec. 360(g)(2) covers all drugs, rather than only drugs made from components that lawfully may be obtained and compounded. There isn't any support for the assumption. The straightforward reading of this language is the one the FDA gives it: "[V]eterinarians may use in the compounding of prescriptions for their private practice whatever bulk drugs or other pharmaceuticals they may lawfully purchase." FDA Compliance Policy Guide Sec. 7125.10 (July 13, 1976). It does not guarantee that any particular drug will be available lawfully. Rutherford shows as much in holding that physicians may not obtain laetrile for administration to dying patients, even on the assumption that the drug is harmless and the prescription (or compounding) private. The FDA treats Sec. 360(g)(2) as allowing veterinarians to "prepare, propagate, compound, or process drugs from ingredients they lawfully acquire ", and the added words are no more than those implied in every statute. Schuyler does not say that Sec. 360(g)(2) allows veterinarians to steal drugs in order to compound them, or to import drugs that are otherwise banned; no more does it allow them to obtain drugs that are otherwise mislabeled. To the extent there is any doubt, we accept the FDA's reading of its own statute. Young v. Community Nutrition Institute. Medical professionals may use their privilege under Sec. 360(g)(2) to mix drugs available outside the new drug application system--"grandfathered" drugs available in 1935 and recognized then as safe and effective. They also may further compound drugs that have been approved by the FDA, to at least a limited extent. This "extra-label use" is governed by a non-enforcement policy related to the one sustained in Heckler v. Chaney. If the need is great and the risk small, the FDA will not contend that re-mixing approved drugs, or using them in a way other than per the directions on the label, causes their "adulteration" under the act. FDA Compliance Policy Guide Sec. 7125.06 (Nov. 1, 1986).

The district court's final objection to Sec. 201.122 was that "it shifts to the claimant an unreasonable burden to show that the exemption applies.... Because of the nature of bulk drugs ..., it is impossible to tell, at this stage, whether the finished product will be a new drug since there is no finished product yet. Thus, the FDA would force the claimant to produce proof which is non-existent in order to invoke the exemption. This result is unreasonable." 674 F. Supp. at 1349. It is not, however, "impossible" to tell what will become of bulk drugs. The seller need only ask the buyer what use lies in store. If the answer discloses impending compounding that would produce a "new animal drug", the seller must ask the buyer to display a filed or granted new drug application. That inquiry would not detain Schuyler, because it is stipulated that (a) all of its sales are to veterinarians, and (b) not a single veterinarian in the United States holds an approved new animal drug application for any drug, and none has such an application on file. More: The Federal Register contains a list of all filed and approved new animal drug applications, so Schuyler could check through that source as well. The task of inquiry set by the regulation, far from being "impossible", turns out to be easy. The easy answer reveals that Schuyler can never satisfy this aspect of the regulation when individual veterinarians are the buyers, but this is different from showing that it is impossible to acquire essential information.

The effect of Sec. 352(f) and Sec. 201.122 is that ingredients that can be used to produce "new" drugs may be sold only to firms that hold approved (or have filed) new animal drug applications. The FDA carefully scrutinizes the activities of these manufacturers. This is a sensible result in light of the structure of the statute.

The statute itself is not universally praised. As Schuyler points out, the FDA has not approved drugs adequate to deal with significant animal diseases. Careful scholars have concluded that the FDA is overly cautious, deprecating benefits and overestimating risks in a way that keeps beneficial drugs off the market, while unduly deferring the introduction of other drugs and raising the cost (and, ironically, the riskiness) of drugs finally made available. E.g., David Leo Weimer, Safe--And Available--Drugs, in Instead of Regulation: Alternatives to Federal Regulatory Agencies 239 (Robert W. Poole, Jr., ed. 1982); Sam Peltzman, The Health Effects of Mandatory Prescriptions, 30 J.L. & Econ. 207 (1987). The certification process hampers small firms and may produce a concentrated industry with higher prices still. Cf. Peter Temin,The Role of Regulation and Technology in the Creation of the Modern Pharmaceutical Firm (1978). It may be no accident that the trade associations of the large drug houses have appeared in this case as amici curiae in support of the FDA. Perhaps the individual judgment of medical professionals (including the pharmaceutical houses that do not recklessly expose their treasuries to depletion in tort litigation) does better than bureaucratic judgment despite the difficulties we have mentioned. Subjects such as these are for Congress and the FDA to consider. Judges' role is to decipher and enforce the existing scheme, whatever they think of its wisdom.

The statute essentially forbids the sale, in any form, of drugs formulated or put to new uses after 1935, without the approval of the FDA. Schuyler, which bears the burden of proof, has not offered to show that the FDA has approved the formulations and uses its customers would make of its bulk drugs. The statutory "hook" for implementing the prohibition is the labeling rule. The 52 lots of bulk drugs seized from Schuyler are mislabeled under Sec. 352(f), and Schuyler does not come within the regulatory exemption. The drugs are forfeit.

REVERSED.