COFFIN, Chief Judge.

Plaintiffs, 178 physicians who treat diabetes and one diabetes patient who use oral hypoglycemic agents to control the disease by lowering the blood sugar level, brought suit to enjoin the defendants Secretary of Health, Education and Welfare and the Commissioner of the Food and Drug Administration (FDA) from enforcing and the defendant drug companies from complying with the FDA’s proposal for altering the labeling of those drugs. The district court granted a preliminary injunction, being persuaded that there was a reasonable likelihood of success in showing that the FDA had failed to comply with the statutes and its own regulation requiring that under some circumstances labeling make reference to the existence of a serious medical controversy. We vacate the injunction for reasons important to the proper judicial role in reviewing administrative actions.

This controversy revolves around a long-term, federally funded study undertaken by the University Group Diabetes Program (hereafter the UGDP study) to determine the effects of oral hypoglycemic agents on vascular complications in patients with adult-onset diabetes. The study, involving twelve clinics and 1200 patients, consisted of four treatment groups: diet alone, diet plus regular insulin doses, diet plus varying insu *412 lin doses and diet plus fixed doses of either tolbutamide or phenformin (two hypoglycemic agents). After monitoring the patients for from five to eight years, the study concluded that the combination of diet and either tolbutamide or phenformin was no more effective than diet alone in prolonging life but that those oral agents might be more hazardous than diet or diet plus insulin insofar as cardiovascular mortality was concerned. The latter conclusion, which led the investigators to discontinue use of the agents in the study as an unethical risk, was based on findings that patients treated with the two agents used in the study suffered more than twice as many cardiovascular deaths than patients receiving the other treatments.

After the study received much publicity and criticism, the FDA convened an ad hoc committee of experts on May 21, 1970 to evaluate the study’s findings and the following day issued a press release agreeing with the UGDP study’s conclusions and indicating that the agency would require labeling changes to reflect those views. After more extensive evaluation, the FDA concluded that protection of the public required a strong warning to physicians recommending use of an oral agent only if other treatments were inadvisable and noting the UGDP’s findings regarding the apparently increased danger of cardiovascular mortality. This evaluation and proposed labeling change was first formally published in the FDA Drug Bulletin of June, 1971.

On October 7, 1971, the Committee on the Care of the Diabetic, consisting of eminent doctors and experts in the field including some of the plaintiff doctors, submitted through its counsel a petition to the FDA. It asked the FDA to rescind its labeling recommendation, insure that all future FDA comments on the UGDP study include references to its alleged deficiencies and controversial nature, provide petitioners with the complete raw data of the study, and, “in accord with its policy of fair balance”, disseminate with equal emphasis and frequency studies and individual expert opinions differing with the study. The petition was accompanied by a detailed scientific critique of the UGDP study and some 250 pages of scientific studies, papers and comments illustrating the nature and extent of the opposition viewpoint. The study was primarily criticized for inadequate patient selection controls and use of fixed, rather than variable, doses of the drugs, contrary to allegedly accepted medical practice. The FDA proposal was attached for extending the study’s findings to all oral agents and patients despite the study’s own warning that such extrapolation could not be made on a statistical basis. The petition also referred to two smaller studies which indicated no cardiovascular complications from oral agents. It was supplemented in January, 1972, by another 220 pages of scientific materials.

In the May, 1972, Drug Bulletin, the FDA published the “Final Labeling Approved For Oral Hypoglycemic Drugs”, which proposed changes in the “indications” section of the label and the addition of a “special warning” section. The proposal speaks of “the increased cardiovascular hazard which appears to be associated with oral hypoglycemic agents”, notes that the UGDP study was the basis for the change, recites its findings, states that these conclusions apply to all oral agents, not just those employed in the study, and ends with the comment that “Further studies are being undertaken to shed additional light on the role” of the oral agents. On June 5, 1972, the Commissioner formally replied to the Committee’s position with an eleven-page, single-space letter addressing generally the legal and medical issues and with a 100 page appendix dealing specifically with the scientific criticisms of the study, criticizing the two contrary studies referred to by the petition, and appending the comments of major medical groups and various scientific papers supportive of the FDA’s position. The Committee’s counsel responded on July 13 with a four-page let *413 ter suggesting that the FDA’s label might constitute misbranding in violation of two cited statutes, that this was one of the “rare cases” suggested by the Commissioner in which “substantial evidence” exists on both sides of an issue, making appropriate reflection of the controversy in the package insert. Counsel requested a formal evidentiary hearing, a stay of any further action pending final resolution, and the full patient records from the study. The Cpm-missioner’s response of August 3 stated that the petitioners were not entitled to a hearing and that only clinical studies were substantial evidence of drug effectiveness. He concluded by saying that, “we do not contend that you do not have standing either to prosecute your peition [sic] or to pursue an appeal to the courts” and that his two letters constituted final agency action reviewable by the courts pursuant to the Administrative Procedure Act.

This suit was filed on August 11, 1972 and a temporary restraining order issued that day. After a hearing and submission of affidavits of experts by both sides, the emergency district judge denied the preliminary injunction on August 30, finding that whatever irreparable injury might be suffered by the plaintiffs did not outweigh that suffered by the public represented by the defendants and that the plaintiffs had not demonstrated “a reasonable probability” of showing that the FDA’s decision to require the warning was arbitrary or capricious. A second motion for a preliminary injunction was denied on September 21 by the judge to whom the ease was permanently assigned because no new evidence or amendment of the complaint had been presented.

On October 17, 1972, the litigation entered an entirely new phase. On that date, plaintiffs filed a motion for leave to amend their complaint, supported by 13 affidavits by diabetes experts attesting to the controversy over the UGDP study, and new motions for a temporary restraining order and a preliminary injunction. The motions presented for the first time the argument that the FDA’s proposed label was itself misleading and thus rendered the drug misbranded in violation of the statute, because it failed to reveal the existence of a “material weight of contrary opinion” among “experts qualified by scientific training and experience” as allegedly required by the agency’s own regulation, 21 C.F.R. § 1.3. After oral argument, at which plaintiffs’ counsel admitted this was an. unprecedented case, being brought by the doctors and seeking to apply that regulation to the agency’s own labeling recommendation, the district court in a Memorandum and Order granted-on November 3, 1972, the motions to amend the complaint and for a preliminary injunction. It noted that “The application for preliminary injunction was heard on the affidavits filed by plaintiffs and defendants, and their arguments both oral and written” and stated that “the court is satisfied plaintiffs have made a showing that there is reasonable likelihood upon a full hearing on the merits they would be successful in establishing the defendants . . . have not in the order described in the May 1972 Bulletin complied with 21 C.F.R. § 1.3.; 21 U.S.C. § 321 (n) and 21 U.S.C. § 352(a).”, and that “there is [absent an injunction] a likelihood of irreparable injury to the plaintiffs” greater than would be visited üpon the defendants by such relief.

The district court had jurisdiction to review the administrative action under the Administrative Procedure Act (APA), 5 U.S.C. § 704, because it was “final agency action for which there is no other adequate remedy in a court.” See Abbott Laboratories v. Gardner, 387 U.S. 136, 87 S.Ct. 1507, 18 L.Ed.2d 681 (1967); Weinberger v. Hynson, Wescott & Dunning, Inc., 412 U.S. 609, 93 S.Ct. 2469, 37 L.Ed.2d 207 (U.S.1973). 1 There is no dispute over the scope of re *414 view — whether the agency action was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2) (A). 2 The action challenged was informal agency action, subject to judicial review under-that standard. 3 Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971); Camp v. Pitts, 411 U.S. 138, 93 S.Ct. 1241, 36 L.Ed.2d 106 (X1973).

. Significantly, there is also no controversy over the basis for judicial review — “the full administrative record that was before the [Commissioner] at the time he made his decision.” Overton Park, swpra, 401 U.S. at 420, 91 S.Ct. at 825. 4 Yet, as the district court’s brief memorandum indicates, the preliminary injunction was based not on a review of that record but on the affidavits presented by both sides to the court. As the Supreme Court has only recently re-itérated, “the focal point for judicial review should be the administrative record already in existence, not some new record made initially in the reviewing court.” Camp v. Pitts, supra, 411 U.S. at 142, 93 S.Ct. at 1244. There are strong policy reasons behind this requirement. *415 Litigation affidavits are often “merely ‘post hoc’ rationalizations which have traditionally been found to be an inadequate basis for review.” Overton Park, supra, 401 U.S. at 419, 91 S.Ct. at 825; see Trailways of New England, Inc. v. C.A.B., 412 F.2d 926, 931 (1st Cir.1969). Moreover, this rule is significant in limiting courts to their proper role. Courts are to determine whether an agency’s action was arbitrary or capricious in light of the information it confronted. It is a re-view, Ía second look at the same material, not a re-doing. And, of course, limiting review to the existing administrative record also saves judicial time.

The requirement that review be on the administrative record parallels and supports the exhaustion of administrative remedies doctrine which reflects similar policies. To the extent that the record reflects consideration of arguments made and evidence submitted, it also reflects the focus which the agency had in bringing to bear its expertise. The exhaustion requirement, as it applies to administrative agencies, is no mere technical rule to enable courts to avoid difficult decisions. It is grounded in substantial concerns not only of fairness and orderly procedure, N.L.R.B. v. Rexall Chemical Co., 370 F.2d 363, 365-366 (1st Cir.1967); United States v. Tucker Truck Lines, Inc., 344 U.S. 33, 36-37, 73 S.Ct. 67, 97 L.Ed. 54 (1952), but also of competence. Courts are not best equipped, as both sides here readily agree, to judge the merits of the scientific studies and the objections to them. Specialized agencies like the FDA are created to serve that function. In this case, the regulation which, in their motion to amend, plaintiffs contend specifically governs the content of a balanced label, 21 C.F.R. § 1.3, was never presented to the Commissioner nor referred to in the administrative record. It is the significance of this omission that governs our disposition.

Plaintiffs argue that while this regulation was never mentioned in the administrative proceedings, the concept of “fair balance” which it represents was fully presented and argued by them in their initial petition, was explicitly rejected by the Commissioner in his initial letter, and that the specific statutes under which this regulation was promulgated were mentioned in plaintiffs’ letter of response. While we recognize that the concept was put forward, are fully aware of the disavantages of further delay, and do not wish to render the exhaustion doctrine a rigid and technical barrier, several factors in this case lead us to insist that the specific argument now pressed be first thrashed out in the administrative arena.

Most significantly, this is an unprecedented argument. As plaintiffs’ counsel readily admitted in oral argument before the district court, there appears to be no prior case in which an FDA drug labeling decision was challenged not by the producer but by concerned medical practitioners, and no case in which the misbranding statutes and regulations were sought to be applied not to the manufacturer’s label but to the FDA’s proposal for alteration of the label in light of new information. It is thus not surprising that the dialogue that did occur regarding “fair balance” was on an entirely different plane. The Food, Drug and Cosmetic Act as amended in 1962 requires that applicants seeking approval for marketing of new drugs present “substantial evidence” of the drug’s effectiveness as well as evidence of its safety. The term “substantial evidence” is defined' in the statute to mean “adequate and well-controlled investigations, including clinical investigations, by experts qualified by scientific training and experience to evaluate the effectiveness of the drug involved”. 21 U.S.C. § 355(d). The FDA has promulgated a detailed regulation to further refine and clarify that definition. 21 C.F.R. § 130.12. See Hynson, supra. Yet the statute provides that approval of an application may be refused or withdrawn not only because “there is a lack of substantial evidence that the drug will have the effect'it purports or is repre *416 sented to have”, 21 U.S.C. § 355(d)(5), (e)(3), or there is “insufficient information to determine whether such drug is safe”, § 355(d)(4), see also § 355 (e)(2), but also if “based on a fair evaluation of all material facts”, the drug’s labeling is “false or misleading in any particular.” § 355(d)(6), (e)(4) (second sentence).

The Commissioner, in his ruling on plaintiffs’ petition, rejected the argument for “fair balance” because he said that Congress had determined in 1962 that “unsubstantiated expert opinion could no longer suffice to establish the effectiveness of drugs”, and that “Except perhaps in rare instances where there is substantial evidence on both sides of an issue, therefore, it is inappropriate to utilize the package insert to present all aspects of the evidence relating to safety and effectiveness.” Since he found that not the situation here, he saw “no basis” for a balanced label. Plaintiffs’ reply, although noting the possibility that the PDA’s proposal might constitute misbranding under the relevant statutes, see infra, primarily argued that “this is in fact one of those rare instances” in which “substantial evidence” exists on both sides of an issue and thus, as the Commissioner’s letter suggested, warranted balanced labeling treatment. The Commissioner’s reply was that the plaintiffs had not in fact presented “substantial evidence” as defined by Congress.

Now, aware of the stringency of the substantial evidence test, see Hynson, supra, the plaintiffs argue that the mis-branding statutes and regulation apply. Section 502 of the statute, 21 U.S.C. § 352, declares that “A drug . shall be deemed to be misbranded — (a) if its labeling is false or misleading in any particular.” The definitional statute (21 U.S.C. § 321 (n)) provides: “If an article is alleged to be misbranded because the labeling is misleading, then in determining whether the labeling is misleading there shall be taken into account (among other things') not only representations made or suggested . but also the extent to which the labeling fails to reveal facts material in light of such representations or material with respect to consequences which may result from the use of the article to which the labeling relates.” Implementing the latter definition is regulation 1.3:

“The existence of a difference of opinion, among experts qualified by scientific training and experience, as to the truth of a representation made or suggested in the labeling is a fact (among other facts) the failure to reveal which may render the labeling misleading, if there is a material weight of opinion contrary to such representation.”

One reading of this regulation would suggest that unsubstantiated individual clinical opinions of qualified experts, which are insuffcient under the “substantial evidence” test enacted in the effectiveness section, might be sufficient to create a fact omission of which might render the labeling misleading.

The Commissioner never considered the meaning of this regulation, its relationship to the substantial evidence test, the intersection of the safety, effectiveness, and misbranding requirements, or the applicability of the misbranding requirements, both statutory and regulatory, to an FDA proposal for re-labeling, for the simple reason that the issue was not presented to him. 5 Arguably these are simply issues of law which we are fully capable of resolving without administrative assistance. But as the Supreme Court has very recently noted in similarly resolving a closely analogous case, the interpretation of even defini *417 tional sections in the drug law will often involve expert knowledge and the ability to evaluate the scientific evidence that becomes relevant. Weinberger v. Ben-tex Pharmaceuticals, Inc., 412 U.S. 645, 650, 93 S.Ct. 2488, 2492-2494, 37 L.Ed. 2d 235 (1973). Moreover, we have here not only novel issues concerning the interpretation of the statute, which the specialized enforcement agency should first undertake, but also unprecedented inquiries as to the meaning of the agency’s own regulations. It is thus no answer to say that an agency need not be reminded of its own regulations. Finally, both the definitional statute and its implementing regulation on which the district court relied explicitly anticipate the exercise of administrative discretion, since they require only that the omission of expert differences of opinion be considered, along with all other relevant facts, in determining whether a label is misleading. As the Court recently reaffirmed: “uniformity and consistency in the regulation of business entrusted to a particular agency are secured, and the limited functions of review by the judiciary are more rationally exercised, by preliminary resort for ascertaining and interpreting the circumstances underlying legal issues to agencies that are better equipped than courts by specialization, by insight gained through experience, and by more flexible procedure.” Ben-tex, supra at 654, 93 S.Ct. at 2494.

An equally important reason for insisting on exhaustion here is that insofar as the record reveals the administrative process seems to have been working well in this instance. The initial ad hoe advisory committee convened by the agency included several eminent critics of the UGDP study, indeed some of the plaintiffs here. While the response to Dr. Bradley’s initial communications may not have been fully satisfactory, the response to the Committee’s petition was both lengthy, detailed, and technical. Even the rebuttal letter, essentially in the form of a petition for reconsideration, a useful procedure in insuring that objections to even the supposedly final agency decision are first brought to its attention, received a specific and complete response. Arguments, studies, and materials, were not ignored; difficult problems were not swept under the rug. The communications evidenced agency recognition of the plaintiffs’ expert status, and, as the concession of standing indicates, receptivity to criticism. Additionally, we were informed at oral argument that extensive negotiations between the parties to arrive at a mutually acceptable solution to the labeling problem had been carried on during much of this litigation. We thus have more than a pious hope that a remand to the agency will not only not be futile, but could well produce the most informed and responsible solution possible.

Because the plaintiffs failed to exhaust their administrative remedies regarding the issues they now present and, consequently, the district court reviewed the agency decision on something other than the administrative record, we must vacate the injunction. 6

Injunction vacated; case remanded for further proceedings consistent with this opinion.

Notes

. The right to petition the court of appeals for review under 21 U.S.C. § 355(h) is available only to a drug-marketing applicant after an order refusing or withdrawing approval of a drug application. Although the FDA has no direct statutory *414 power to compel labeling changes, it may refuse or withdraw approval of an application, under § 355(d) or (e), if new information demonstrates that the labeling is “false or misleading.” Since the drug companies here were willing to amend their labels in accordance with the PDA proposal, no withdrawal procedures were ever begun and thus the agency action was final in a meaningful sense. Abbott, supra.

Before this court, the government, for the first time, argued that plaintiffs did not have standing to sue, not being persons “adversely affected or aggrieved by agency action within the meaning of a relevant statute” as required by 5 U.S.C. § 702.’ Even assuming that the government may still raise this claim notwithstanding an explicit concession of standing by the specialized agency charged with interpretation and enforcement of the relevant statute, the case is properly before us. While there might be some doubt whether the plaintiff doctors, who clearly alleged “injury in fact”, Sierra Club v. Morton, 405 U.S. 727, 92 S.Ct. 1361, 31 L.Ed.2d 636 (1972) ; United States v. Students Challenging Regulatory Agency Procedures, 412 U.S. 669, 93 S.Ct. 2405, 37 L.Ed.2d 254 (U.S.1973), because of the impact of the label changes on their medical practice and on malpractice suits, have an interest “arguably within the zone of interests to be protected or regulated”, Data Processing Service v. Camp, 397 U.S. 150, 90 S.Ct. 827, 25 L.Ed.2d 184 (1969), there can be no question that the plaintiff patient’s interest is in the very center of that zone.

; While the plaintiffs now also contend that the agency’s failure to follow its own regulations is a violation of due process, thus apparently also invoking the review standard of § 706(2) (B), “contrary to constitutional right, power, privilege, or immunity”, that approach would change neither the nature of the review nor the result. If an agency action violates a regulation, it is “not in accordance with law” as well as violative of due process, United States v. Griglio, 467 F.2d 572 (1st Cir. 1972). Moreover, courts must review agency actions under both standards in all cases. Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 414, 91 S.Ct. 814, 28 L.Ed.2d 136 (1971). If, however, a regulation has been violated, there would seldom be occasion to decide the constitutional issue.

. Plaintiffs do not now claim that they are entitled to an evidentiary administrative hearing.

. The plaintiffs insist that the record must include not only their petition and the Commissioner’s response, which the record before us contains, but also the original patient records of the UGDP study, any intra-agency and other memoranda, factual reports and scientific studies before the Commissioner, and the minutes taken during the deliberations of the ad hoc advisory committee convened by the PDA the day before its first press release regarding the study. While in light of our discussion we need not resolve the propriety of each of these requests, we reiterate what we recently stated in an analogous situation: “We think the law requires production of the entire administrative record .... While there may be some instances in which the entire record need not be filed, where the correctness of factual findings are involved or where the complainants request the full record, we think the agency must produce it in court. Cf. 28 U.S.C. § 2112(b).” Silva v. Lynn, 482 F.2d 1282 at 1283 n. 1 (1st Cir. 1973).

. Similarly lie never considered the applicability or meaning of 21 C.F.R. § 1.106(b) (3) (i), which plaintiffs refer to as the “full disclosure” regulation, but which they failed to disclose hot only to the Commissioner but to the district court. Obviously, the relevance and impact of that regulation should also first be presented for administrative consideration.

. We therefore need not decide whether reversal would have been required, as argued by the defendants, because the district court, in granting the injunction, did not, as required by Fed.R.Civ.P. 52(a), “set forth the findings of fact and conclusions of law which constitute the grounds of its action”. AA’hile the context will often render obvious the grounds of decision, district courts should generally follow the rule’s mandate to obviate any possibility of misunderstanding, unnecessary reversal and/or delay.