Erin Smith brought this action for declaratory and injunctive relief against the Office of Civilian Health and Medical Program of the Uniformed Services (CHAMPUS) and William Perry, in his official capacity as the Secretary of Defense, challenging their refusal to pay for certain procedures recommended by Smith’s doctors as treatment for her breast cancer. The district court declared that the recommended procedures were covered by CHAMPUS, held that defendants had acted arbitrarily and capriciously in concluding otherwise, and entered a permanent injunction enjoining defendants from denying Smith coverage. We reverse the district court’s judgment and remand for the district court to enter an appropriate order affirming CHAMPUS’ initial determination.
I.
At the time of her complaint, Erin Smith was a forty-year-old woman diagnosed with advanced breast cancer. Smith’s doctors advised her that the cancer had spread to her local lymph nodes and that high-dosage chemotherapy (HDC) coupled with peripheral stem cell rescue would give her the best chance of survival. However, neither Smith’s doctors nor St. Vincent Hospital in Indianapolis where the treatment was to be administered would commence treatment until they received assurances that Smith had the means to cover the costs.
HDC involves administering the chemotherapeutic agents used in standard chemotherapy, but at higher dosages. While HDC is more effective at killing the cancer, it is also more likely to kill the patient’s stem cells, the cells which generate white blood cells, the primary component of the body’s immune system. To mitigate the potential damage to the patient’s immune system, doctors extract (the technical term is “harvest”) some of the patient’s stem cells prior to administering HDC. This harvesting process, referred to as autologous stem cell rescue (ASCR), is accomplished by one of two methods: autologous bone marrow transplant support (ABMT), in which the stem cells are harvested from the patient’s bone marrow, and peripheral stem cell rescue (PSCR), in which stem cells are harvested from the patient’s blood stream. Under either procedure, the harvested stem cells are frozen and stored until the HDC is complete, at which time the stem cells are reintroduced into the patient. Although Smith suggests otherwise, CHAMPUS maintains that HDC/PSCR and HDC/ABMT are essentially of equal medical value in treating breast cancer. Our review *952of the record supports CHAMPUS’s position.2
Married to an Air Force retiree, Smith receives her primary health care coverage through CHAMPUS. CHAMPUS was established by Congress pursuant to the Dependents’ Medical Care Act, 10 U.S.C. § 1071 et seq., to provide medical and dental benefits to dependents of present and former members of the military. CHAMPUS is not an insurance program where the insurer guarantees indemnification in return for a premium. CHAMPUS beneficiaries pay no premiums. Rather, CHAMPUS is funded by annual Congressional appropriations. Another unique feature of CHAMPUS is that it is an “at risk” program, meaning that unlike traditional health insurance programs, where beneficiaries usually know whether a treatment is covered beforehand, CHAMPUS beneficiaries typically receive medical care first and then submit a claim to CHAMPUS officials for an after-the-fact ruling on coverage. The beneficiary is “at risk” in the sense that the medical services received may not qualify for payment under CHAMPUS. Coverage determinations, as well as the other day-to-day administrative duties of CHAMPUS, are charged by statute to the Secretary of Defense who has delegated these responsibilities to the Director of the Office of CHAMPUS. 32 C.F.R. §§ 199.5, 199.7.
On Smith’s behalf, the oncology department at St. Vincent’s Hospital filed a claim with CHAMPUS outlining the particulars of Smith’s case and requesting a pre-treatment determination of whether HDC/PSCR would be covered. On August 2, 1994, Dr. David Bogner, CHAMPUS’s medical director, issued an initial determination that CHAM-PUS would not cover the prescribed HDC/ PSCR treatment. By letter he explained that under the terms of its Congressional mandate, CHAMPUS cannot cover treatments or procedures that are considered experimental or investigational. Based on his review, HDC/PSCR fell into that category:
After careful consideration of the documents you supplied, and following review of facts presented by our oncology consultants, technology assessment panels, and the current Phase III refereed medical literature, CHAMPUS finds that it is unable to offer benefits for high dose chemo*953therapy and autologous stem cell rescue in the treatment of breast carcinoma.
According to Dr. Bogner, the medical literature lacked sufficient evidence of the effectiveness of HDC/PSCR for the treatment of Smith’s condition:
CHAMPUS continuously reviews the current literature for outcomes of Phase III trials regarding high dose chemotherapy with autologous stem cell rescue. In the case of breast carcinoma, we have been unable to find sufficient evidence of this nature to date.
In light of this, Dr. Bogner concluded that “CHAMPUS must continue to consider this therapy as investigational for the treatment of breast carcinoma.”
Yet despite referencing specific reasons for rejection, in particular the lack of Phase III trials, Dr. Bogner remained open to any documented evidence of HDC/PSCR’s general acceptance as a treatment for breast cancer:
If you disagree with this CHAMPUS benefit determination, we invite you to submit pertinent documentation to support the position that high-dose chemotherapy with autologous stem cell treatment of breast carcinoma does, in fact, meet the generally accepted standards of usual professional medical practice in the general medical community.
However, he cautioned that “[wjhile personal opinions are valued, we must give the greatest weight to well-designed, Phase III, outcome based studies which have been published in refereed medical journals.”3 (Emphasis added.)
A month later, Smith’s attorney filed a request for reconsideration. Acting on Dr. Bogner’s invitation, counsel submitted affidavits from two of Smith’s oncologists along with one from a third doctor familiar with Smith’s case. Each expressed the opinion that HDC/PSCR was generally accepted in the medical community and not considered experimental for the treatment of breast cancer. Counsel also submitted two other bases for reconsideration: an article suggesting that most private insurance companies eventually approve coverage for HDC treatment and a reference to three recent district court decisions enjoining CHAMPUS from denying coverage. See Gripkey v. Mail Handlers Benefit Plan, No. 3:94-378-0, 1994 WL 276265 (D.S.C. Feb.14, 1994) (unpublished), Hawkins v. Mail Handlers Benefit Plan, No. 1:94CV6, 1994 WL 214262 (W.D.N.C. Jan.28, 1994) (unpublished), and Wheeler v. Dynamic Engineering, Inc., 850 F.Supp. 459 (E.D.Va.1994), aff'd, 62 F.3d 634 (4th Cir.1995).4 Two of these, Gripkey and Hawkins, involved similar determinations by Dr. Bogner denying coverage for HDC/PSCR for the treatment of breast cancer. Significantly, counsel did not supply references to clinical studies indicating that HDC/PSCR was generally accepted in the medical community for the treatment of breast cancer.
After reviewing the additional documentation, on September 14, 1994, Dr. Bogner denied the request for reconsideration. The denial noted counsel’s failure to produce evidence of favorable Phase III trials or any other “medical literature references upon *954which CHAMPUS could consider adoption of the benefit you have requested.” Without convincing evidence to the contrary, Dr. Bog-ner stated that it remained CHAMPUS’s position that HDC/PSCR for the treatment of breast cancer “has not become a national medical standard of practice, and is, therefore, not a CHAMPUS benefit.”
Smith responded by filing a complaint in the district court seeking a declaration that HDC/PSCR was covered under CHAMPUS and an injunction prohibiting CHAMPUS from denying coverage. The Secretary moved for summary judgment on the grounds that the decision to deny coverage was supported by the administrative record and was not arbitrary and capricious. See 5 U.S.C. § 706(2)(A) (arbitrary and capricious standard). Following expedited proceedings, the district court entered judgment for Smith. The principal basis for its ruling was that CHAMPUS had relied on outdated medical studies in making its determination. The court held these outdated studies could not defeat the affidavits submitted by Smith’s oncologists stating that HDC/PSCR was generally accepted for the treatment of breast cancer. The district court concluded that CHAMPUS’s actions were arbitrary and capricious and accordingly entered an order permanently enjoining the Secretary from denying coverage for Smith’s treatment.5 The Secretary appeals.6
II.
A. Standard of Review
Under the Administrative Procedures Act (“APA”), our review of an agency decision is to determine whether the decision was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A); II Kenneth C. Davis and Richard J. Pierce, Administrative Law Treatise § 11.4 at 200 (3d ed.1994). We note that had CHAMPUS denied Smith benefits based on the results of an optional adjudicatory hearing conducted pursuant to 32 C.F.R. § 199.10(d), judicial review would focus on whether the ruling was supported by “substantial evidence.” 5 U.S.C. § 706(2)(E); Camp v. Pitts, 411 U.S. 138, 141, 93 S.Ct. 1241, 1243, 36 L.Ed.2d 106 (1973) (per curiam); Citizens to Preserve Overton Park, Inc, v. Volpe, 401 U.S. 402, 414, 91 S.Ct. 814, 822-23, 28 L.Ed.2d 136 (1971). Smith did not request a formal adjudicatory hearing to determine the merits of her claim. Instead, she submitted her claim for an initial hearing and then reconsideration, both of which are nonadjudicatory and are not designed to produce a record developed by formal, trial-type proceedings. Therefore, as noted, our review is to determine whether CHAMPUS’s decision was “arbitrary, capricious, an abuse *955of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A).
In making this determination, the court reviews the administrative record as it stood when the agency acted, not extra-record material produced later in court. Florida Power & Light Co. v. Lorion, 470 U.S. 729, 743-44, 105 S.Ct. 1598, 1606-07, 84 L.Ed.2d 643 (1985); Camp v. Pitts, 411 U.S. at 142, 93 S.Ct. at 1244; Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. at 420, 91 S.Ct. at 825-26; National Med. Ent., Inc. v. Shalala, 43 F.3d 691, 695 (D.C.Cir.1995); Osaghae v. United States Immigration and Naturalization Serv., 942 F.2d 1160, 1162 (7th Cir.1991). “[W]e must consider whether the decision was based on a consideration of the relevant factors and whether there has been a clear error of judgment.” Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43, 103 S.Ct. 2856, 2866-67, 77 L.Ed.2d 443 (1983) (quotations and citations omitted). Where Congress has authorized an agency to promulgate criteria for dispersing benefits and to interpret those criteria, a reviewing court must guard against substituting its own judgment for that of the agency. State Farm, 463 U.S. at 43, 103 S.Ct. at 2866-67; Overton Park, 401 U.S. at 416, 91 S.Ct. at 823-24. “[An] agency’s interpretation [of its own regulations] must be given controlling weight unless it is plainly erroneous or inconsistent with the regulation.” Thomas Jefferson Univ. v. Shalala, 512 U.S. 504, -, 114 S.Ct. 2381, 2386, 129 L.Ed.2d 405 (1994) (internal quotation marks omitted); Hinsdale Hospital Corp. v. Shalala, 50 F.3d 1395, 1399 (7th Cir.1995) (deference particularly due when agency is interpreting regulations issued pursuant to complex scheme in which determinations necessarily require significant expertise and entail the exercise of judgment grounded in policy concerns). Whether an agency decision violates the legal standards of § 706 or whether an agency interpretation of its regulations is plainly erroneous are questions of law. Our review of the district court’s legal conclusions is therefore de novo. See 5 U.S.C. § 706; Director, O.W.C.P., United States Dept. of Labor v. Ball, 826 F.2d 603, 604 (7th Cir.1987) (interpreting statutory and regulatory language “a matter of law”; “For matters of law, the APA mandates de novo review.”); see Durasys, Inc. v. Leyba, 992 F.2d 1465, 1470 (7th Cir.1993) (conclusions of law underlying permanent injunction reviewed de novo).
B. Analysis
CHAMPUS’s denial of coverage can be characterized either as an interpretation of its regulations governing experimental treatments, in which case we review for a “plainly erroneous or inconsistent” interpretation, Thomas Jefferson Univ., 512 U.S. at -, 114 S.Ct. at 2386, or as a decision to adopt the view of one side of an ongoing debate in the medical community, in which case our review is under the arbitrary and capricious standard of 5 U.S.C. § 706(2)(A). Under either standard our review is highly deferential, monitoring only for glaring mistakes-obvious wrongness. Pozzie v. United States Dept. of Housing and Urban Development, 48 F.3d 1026, 1029 (7th Cir.1995) (“The ‘arbitrary or capricious’ standard of review is a deferential one which presumes that agency actions are valid as long as the decision is supported by a ‘rational basis.’ ”) (emphasis added) citing Western & Southern Life Ins. Co. v. Smith, 859 F.2d 407, 410 (6th Cir.1988); Kisser v. Cisneros, 14 F.3d 615, 618 (D.C.Cir.1994) (review is “highly deferential”); and Hussion v. Madigan, 950 F.2d 1546, 1550 (11th Cir.1992).
The general exclusion of coverage for experimental or investigational procedures under CHAMPUS is contained in 32 C.F.R. § 199.4(g)(15):
(g) Exclusions and limitations. In addition to any definitions, requirements, conditions, or limitations enumerated and described in other sections of this part, the following specifically are excluded from the Basic Program:
(15) [Treatments, etc.] [n]ot in accordance with accepted standards, experimental or investigational. Services and supplies not provided in accordance with accepted professional medical standards; or related to essentially experimental or *956investigational procedures or treatment regimens.
32 C.F.R. § 199.4(g)(15). The term “experimental” is defined as:
Medical care that essentially is investigatory or an unproven procedure or treatment regimen (usually performed under controlled medicolegal conditions) that does not meet the generally accepted standards of the usual professional medical practice in the general medical community.
32 C.F.R. § 199.2(b). The regulations do not set forth the criteria for determining whether a treatment or procedure meets “the generally accepted standards of the usual professional medical practice in the general medical community.” CHAMPUS’s Policy Manual defines “experimental” in essentially the same way as § 199.2; no additional standards are provided for ascertaining general acceptance. What qualifies, then, is a matter of agency interpretation subject to deferential court review.7
This is not the first time this court has addressed the ongoing dispute over the medical value of HDC/ASCR. In Harris v. Mutual of Omaha Cos., 992 F.2d 706, 712-13 (7th Cir.1993), we held that HDC/ABMT was “experimental” under the definition provided in the Rural Carrier Benefit Plan, a federal health insurance policy administered by the federal Office of Personnel Management. Our decision was at least partly based on a record containing eighteen articles and reports published between November 1986 and July 1992, each of which concluded that HDC/ABMT “is currently in the developmental stage and requires more clinical research before it can be considered standard treatment for breast cancer.” Id. at 709. In Fuja v. Benefit Trust Life Ins. Co., we held that, although HDC/ABMT had “proven effective in treating certain cancerous blood diseases such as leukemia and Hodgkin’s disease,” as regards “solid-type tumors including breast cancer,” 18 F.3d at 1407, “[t]he uncontradicted expert testimony and documentary evidence presented to the trial court establishes that HDC/ABMT is still of uncertain medical value and is presently being researched at a number of leading medical colleges and oncology research centers throughout the country.” Id. at 1410. In Bechtold v. Physicians Health Plan, 19 F.3d 322 (7th Cir.1994), we similarly held that HDC/ABMT was not covered under an ERISA-governed insurance plan excluding “experimental or unproven procedures.” Id. at 328. And most recently in Hooper v. Demco, Inc., 37 F.3d 287 (7th Cir.1994), we concluded that because payment for HDC/ ABMT was not required under a health plan excluding “experimental” or “investigative” procedures, a patient was not entitled to recover attorneys’ fees for the plan’s refusal to extend coverage. Id. at 294-95.
These cases alone do not establish that HDC/ASCR was experimental at the time CHAMPUS denied Smith coverage.8 The pace of medical science is ever quickening; yesterday’s esoteric experiment is today’s miraculous cure. But they do confirm that in the immediate past the medical efficacy of HDC/ASCR for the treatment of breast cancer was in dispute. That is noteworthy because at issue here is the point where a treatment which has been experimental in the past crosses the line into general acceptance — the point at which the medical value *957of a treatment is no longer generally disputed. Perhaps no such line exists; we are probably dealing more with a zone of perceived effectiveness than a precise dividing line. What is evident, though, and foremost in our minds as we consider this case, is the incompetence of courts to decide when exactly that line or zone has been traversed. Such decisions are judgment calls for medical scientists and health-care professionals, not judges. We are “not empowered to substitute [our] judgment for that of the agency.” Overton Park, 401 U.S. at 416, 91 S.Ct. at 824. Which is why our standard of review in these eases is highly deferential. To repeat, our narrow duty is to monitor those charged with knowing and deciding these matters for decisions that are patently wrong; decisions we may fairly denounce as “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706(2)(A). Where the interpretation of a regulation is at issue, we guard against readings that are obviously incorrect. Thomas Jefferson Univ. v. Shalala, 512 U.S. at -, 114 S.Ct. at 2386. Therefore, we cannot overturn CHAMPUS’s decision to treat HDC/PSCR as still experimental if at the time it denied coverage there was a significant, on-going dispute in the medical community over HDC/PSCR’s medical efficacy. Such a decision cannot be deemed arbitrary, for it entails the very judgment call Congress has assigned to the administrative agencies and not the courts. As the following reveals, the evidence before us clearly demonstrates such a dispute and therefore that CHAM-PUS’s determination must be upheld.
In an affidavit submitted to the district court, Dr. Bogner revealed the bases of CHAMPUS’s decision to deny coverage:
6. The initial CHAMPUS policy regarding the investigational nature of HDC/ ASCR in the treatment of breast carcinoma was based on four primary sources:
a. The American Medical Association Diagnostic and Therapeutic Technology Assessment (AMA DATTA) evaluation of January 1990 entitled “Autologous Bone Marrow Transplantation — Reassessment” by Elizabeth Brown, M.D_
b. The 1988 study entitled “Public Health Service Reassessment: Autolo-gous Bone- Marrow Transplantation” prepared by the Office of Health Technology Assessment, Agency for Health Care Policy and Research (OHTA/AHCPR) of the Public Health Service, and authored by Harry Han-delsman, D.O_
c. The June 1993 study entitled “Au-tologous Bone Marrow Transplant and Peripheral Blood Stem Cell Rescue for the Treatment of Breast Cancer” copyright by ECRI[9]....
d. News releases concerning the most recent ECRI assessment of “Auto-logous Bone Marrow Transplant and Peripheral Blood Stem Cell Rescue for the Treatment of Breast Cancer.” This summary information was published in Health and Technology Trends in June 1994. I also received a copy of essentially the same material directly from ECRI on June 7,1994....
Each of the studies Dr. Bogner referenced essentially treats HDC/ASCR as investiga-tional or experimental in the treatment of breast cancer. For instance, the 1988 Public Health Service study authored by Dr. Han-delsman states that “[t]he available evidence suggests that using ABMT for solid tumors [which include breast cancer], with the exception of neuroblastoma, has not shown meaningful increased survival times.” The study noted conflicting views on the topic. For example, at the time of the study the National Cancer Institute-Navy Medical Oncology Branch had taken the position that “ABMT *958cannot be considered the standard of care other than in programs that are investiga-tional,” while the M.D. Anderson Hospital and Tumor Institute held that “ABMT should no longer be considered investigational since favorable results have been achieved in the treatment of lymphomas, neuroblasto-ma, and stage IV breast cancer.” The study concluded that the effectiveness of ABMT was not certain and that “[i]n the absence of Phase III clinical trials directly comparing ABMT with conventional therapies in the treatment of most solid tumors and [given] the lack of other evidence convincing to the clinical community, its role continues to be undefined.”
Six years later, HDC/ASCR remained controversial. An overview of the draft report for a 1994 study by ECRI, which Dr. Bogner relied on, opens by characterizing HDC/ ASCR as an “experimental — and costly— treatment for advanced breast cancer [that] is diffusing into the health care system before there is any clear evidence that the treatment is better than conventional therapies.” It states that after reviewing “over 300 published studies and articles,” ECRI concluded that “results from the experimental procedure are not any better than published results for conventional therapy to treat advanced breast cancer.” (Emphasis added.) ECRI Senior Research Analyst Dr. Nelson Erlick, author of the report, advised that “[i]f a hospital wants to provide high-dose chemotherapy and stem cell rescue to breast cancer patients, it should be aware that the impetus for this is more political than scientific,” and that it is a “treatment that’s becoming mandated by popular opinion.” The summary also noted that “the design of some studies [of HDC/ASCR] is biased toward favorable results.”
Though somewhat more reserved in tone, the full ECRI report, contained in the record before us, is no more supportive of HDC/ ASCR. The summary of its clinical findings with respect to metastatic (stage IV) patients states that “[t]here is no evidence that HDC/ ABMT or HDC/ ASCR,[10] when taken as groups, provide longer survival times than conventional therapy.”11 To the contrary, “[m]ost HDC with stem cell rescue regimens are associated with shorter survival times compared to the conventional regimens.” (Emphasis in original.) The final paragraph of this section concludes:
Current data suggest that existing HDC with ASCR regimens are unlikely to produce significantly greater response durations and survival times than conventional therapy. These data also suggest that it is unlikely that controlled trials (in progress) will demonstrate any substantive improvement in the quality of life or survival times for metastatic patients. It is possible that investigators may identify effective regimens and/or determine prognostic factors to identify metastatic breast cancer patient subsets most likely to experience sustained, complete responses and increased survival times. Unfortunately, based on available studies, this seems improbable.12
Smith challenges the relevance of these conclusions because they apply to metastatic, i.e., stage IV, breast cancer and not to the milder stage III cancer that afflicts her. Smith has been described as having “advanced” breast cancer and as a “poor prognosis” patient with a “metastatic tumor” present in both her breast and lymph nodes and “an extremely high risk to develop recurrent disease.” CHAMPUS maintains that the ECRI report’s findings are therefore directly relevant. But whichever side is correct, the ECRI report also indicates that the effeetive*959ness of the procedure for treating stage III breast cancer remains uncertain: “There are insufficient data at this time for assessing the effectiveness of HDC with ASCR for inflammatory, stage III, or stage II ... breast cancer patients. Controlled studies are in progress.”
Another report, also referenced in Dr. Bogner’s affidavit, similarly supports CHAM-PUS’s decision. The January 1994 edition of the Journal of Clinical Oncology reported on an international conference on HDC with stem cell support. The report states that while “there is ample scientific background for vigorous clinical investigation in this important area,” there is “currently insufficient evidence to justify the use of HDC plus HSC [hematopoietic stem cell] transplantation outside the setting of a clinical trial for any stage of breast cancer.”
These studies are powerful evidence that HDC/ASCR remains experimental. As noted, when invited to submit countervailing evidence Smith supplied affidavits from three oncologists connected with her case, references to several district court decisions on the matter, and an article suggesting that insurance coverage is usually approved for HDC/PSCR. No refereed journal articles or other technical data reviewing the present state of medical opinion concerning HDC with stem cell recovery regimes was submitted.
We do not mean to disparage the opinions of Smith’s able physicians. Barely familiar with the vocabulary, it is fortunately not our duty to decide which medical view is correct. The function assigned to us by Congress in these matters is simply to guard against arbitrary decisions and plainly erroneous interpretations. Whatever our sympathies, beyond that we cannot go. Given the considerable evidence indicating that HDC, whether accompanied by ABMT or PSCR, remains controversial for the treatment for breast cancer, we cannot say CHAMPUS’s decision to deny Smith coverage for HDC/ PSCR was arbitrary and capricious or a plainly erroneous interpretation of its regulations. Widespread disagreement among qualified medical experts over a medical issue virtually precludes a reviewing court from concluding that an agency decision that agrees with one side is arbitrary or plainly wrong, even if the court finds other views more persuasive. See, e.g., Florida Manufactured Housing Ass’n, Inc. v. Cisneros, 53 F.3d 1565, 1572 (11th Cir.1995) (remarking that in the context of rulemaking, “[w]hen the agency is confronted with opposing views among specialists, it must be given the discretion to rely on the reasonable opinions of its own experts, even if a court finds other views more persuasive”); Franklin Sav. Ass’n v. Director, Office of Thrift Supervision, 934 F.2d 1127, 1144 (10th Cir.1991), cert. denied, 503 U.S. 937, 112 S.Ct. 1475, 117 L.Ed.2d 619 (1992) (observing in the context of reviewing agency action that “Conflicting expert opinion ... is not sufficient to allow a reviewing court to conclude the agency decision was arbitrary, capricious or an abuse of discretion”). We are therefore legally compelled to uphold CHAMPUS’s decision.
In so doing, we reject Smith’s allegation, accepted by the district court, that CHAM-PUS has arbitrarily equated general acceptance with completed Phase III trials. The Fourth Circuit in Wilson v. CHAMPUS, 65 F.3d 361 (4th Cir.1995), which was decided after the district court’s decision in this case, also concluded that CHAMPUS had imposed a rigid requirement — “an unwritten agency policy” — mandating Phase III trials before a treatment can be provided. Id. at 365-66. Whatever the evidence before the court in Wilson, on the record before us, Smith and the district court have misconstrued CHAM-PUS’s position. The record does not indicate that CHAMPUS has created a broad new requirement, much less “an unwritten agency policy,” that every new procedure successfully complete Phase III trials before being approved. Rather, CHAMPUS has determined that given the on-going controversy and absent convincing new evidence, Phase III trials will be necessary to determine whether HDC/ASCR can fairly be said to have moved beyond the experimental to the generally accepted.
Our colleague disagrees. He argues that Dr. Bogner’s letter of response to the “new information” submitted by Smith’s doctors “indicated that published Phase III trials are *960a sine qua non for coverage.” Post at 964. For support the dissent quotes the letter’s statement that “CHAMPUS bases its policies upon outcomes of Phase III clinical trials as published in the refereed medical literature.” The dissent treats this as nearly conclusive evidence that CHAMPUS has “created a requirement that every new procedure successfully complete Phase III trials before being approved.” Id. at 964 (emphasis added).
The record does not support such a sweeping interpretation. Like all documents, Dr. Bogner’s letter must be read in the context in which it was written: as an explanation for why CHAMPUS decided to maintain its initial determination that HDC/ASCR is an experimental treatment for breast cancer — not as a comprehensive statement of CHAM-PUS’s approach to all new procedures and treatments. The affidavit of Martha Maxey, CHAMPUS’s Health Benefits Program Analyst, (parts of which the dissent quotes, post at 962) confirms this reading: “In determining whether a procedure is investigational, the Office of CHAMPUS looks to national medical policy organizations [ie., not solely to completed Phase III trials] to ascertain that adequate evidence exists to establish a procedure or treatment as safe, effective and generally acceptable medical practice. The information provided by these national technology assessing agencies, such as the Office of Health Technology Assessment of the Agency for Health Care Policy and Research of the Public Health Service, generally provides direction as to the final structure of the CHAMPUS benefit policy.” “Each report” from these organizations, the affidavit continues, “represents a detailed analysis of the safety, clinical effectiveness and use of new or unestablished medical technologies.” As related above, in 1988 a study of HDC/ ABMT by OHTA concluded that the role for the procedure in treating solid tumors, like breast cancer, “continues to be undefined.” See ante at 957. Maxey’s affidavit states that in light of this and other more recent reports, which now include the 1993 and 1994 ECRI reports, “CHAMPUS has concluded that the clinical trials published to date have not provided definitive evidence of the benefit of ABMT for the treatment of solid tumors other than neuroblastoma.” These “clinical trials published to date” necessarily were non-Phase III trials, since thus far no Phase III trials have been completed. Dr. Bogner’s affidavit discussing the basis for CHAMPUS’s initial determination that HDC/ ASCR is experimental confirms this. See ante at 957-58. It is in this specific context — and not because of an arbitrarily imposed general requirement — that CHAM-PUS has determined that Phase III trials are necessary, as the next sentence of the affidavit shows: “Therefore, in the absence of Phase III clinical trials directly comparing ABMT with conventional therapies in the treatment of most solid tumors (including breast cancer) and the lack of other evidence convincing to the medical community, CHAMPUS considers its role to be undefined.” (Emphasis added.) Given this, and other similar evidence in the record,13 we cannot agree with the dissent that CHAM-PUS has adopted a mindless “Phase III or nothing” approach to all new treatments, nor that there is any reasonable factual dispute in this regard. Cf. post at 966-67.
To the contrary, Smith’s doctors were explicitly invited to “submit pertinent documentation to support” their view of the efficacy and general acceptance of HDC/PSCR. We assume that had they produced such evidence coverage would have been granted. Taken in context, Dr. Bogner’s cautionary statements about the importance of Phase III studies merely indicate that while the *961medical controversy over the efficacy of HDC/ASCR rages, CHAMPUS will sit tight until convincing evidence is forthcoming.14
In this regard, it is important to realize that we are not dealing with the introduction of, say, penicillin, where the treatment’s immediate and obvious success might obviate the need for further studies before it can be deemed generally accepted. Wilson noted the possibility of a treatment becoming generally accepted without Phase III trials and then, after referencing a letter from a number of physicians supporting HDC/PSCR, implied that HDC/PSCR must be such a treatment. Id. at 365-66. We assume from its ready acceptance of an informal testimonial that the Fourth Circuit did not have before it the technical studies we reviewed above. To summarize, these studies demonstrate, as does this entire controversy, that the medical value of HDC/ASCR — whether with ABMT or PSCR — is hotly disputed among medical professionals. In the context of such a dispute-which, we again emphasize, we are institutionally and professionally incompetent to resolve-CHAMPUS’s decision to await the results of Phase III trials or other persuasive proof before determining that the taxpayer must fund this procedure was not a “plainly erroneous” interpretation of its regulations or otherwise an abuse of its discretion.
Furthermore, even if CHAMPUS had set the hurdle too high by equating general acceptance with completed Phase III trials, that alone does not establish that HDC/ PSCR is generally accepted and thus covered by CHAMPUS. Striking a standard as too high does not make everything rejected under that standard automatically acceptable. We are still left with the fact that on the record before us there is no convincing evidence that HDC/PSCR has moved beyond the experimental stage to general medical acceptance — and most certainly none that would allow us to condemn CHAMPUS’s decision as plainly erroneous or arbitrary and capricious. Like everyone concerned with the fate of breast cancer patients, we sincerely hope HDC/ASCR proves the cure so many desperately seek. But we cannot substitute our hopes for CHAMPUS’s considered judgment.
III.
The judgment of the district court is REVERSED. We RemaND to the district court to enter an appropriate order affirming CHAM-PUS’s initial determination denying Smith coverage.
. Compare Mattive v. Healthsource of Savannah, 893 F.Supp. 1559, 1572 (S.D.Ga.1995) ("The Court recognizes that the ABMT and the PSCR accomplish the same thing when used in conjunction with HDC in treating cancer patients .... but the procedures for removing the stem cells are different.... Assuming that the HDC with ABMT would be excluded by [the policy language], the Court agrees with Defendant that the failure to exclude HDC with PSCR under this provision would indeed be a failure to exclude based on semantics.”), with Wheeler v. Dynamic Engineering, Inc., 850 F.Supp. 459, 468 (E.D.Va.1994) (“Although the two procedures are similar in that they both provide support for a patient receiving high dose chemotherapy, the two are distinct procedures.”), and Bishop v. CHAMPUS, 917 F.Supp. 1469, 1472 n. 1 (E.D.Wash.1996) (although PSCR and ABMT are both support treatments for HDC and not the principal components of HDC, "[t]he Court questions whether the CHAMPUS policy of equating PSCR and ABMT would survive judicial review, even under the deferential standard imposed by the APA.”).
At the time Smith submitted her request, CHAMPUS authorized ABMT to support HDC therapy for the treatment of non-Hodgkin’s lymphoma (intermediate or high grade stage III or stage IV); Hodgkin's disease (stage III or stage TV A or B) when standard chemotherapy has failed and the patient has “adequate marrow function and no evidence of marrow involvement or lymphoma;” neuroblastoma (stage II or IV) when further "conventional-dose therapy is not likely to achieve a durable remission;” and acute lymphocytic or non-lymphocytic leukemias "following the first or subsequent remission.” 1 CHAMPUS Policy Manual, Ch. 3, § 6.38245.1. CHAMPUS authorized PSCR when ABMT was no longer possible: "[Hjarvesting of the required stem cells by apheresis from peripheral blood [PSCR] rather than bone marrow [ABMT] can be allowed for those beneficiaries for whom it has been established that bone marrow harvesting cannot be accomplished due to documented bone marrow involvement” with cancer. Id. Whether that suggests ABMT is medically preferable to PSCR is unclear. Contra Wilson v. CHAMPUS, 65 F.3d 361, 364 (4th Cir.1995) ("PSCR is a procedure distinct from ABMT that some studies have shown to be more effective.”); but see id. at 362 ("A similar, alternative procedure [to PSCR] known as an autologous bone marrow transplant (“ABMT") retrieves [stem] cells from the patient’s bone marrow.”). We emphasize that nothing in the record before us indicates that the effectiveness of HDC is substantially altered by the method of stem cell rescue. If anything, the record suggests that HDC with PSCR may be slightly less effective than with ABMT. See, e.g., infra note 11.
. All clinical cancer trials occur in four phases. In Phase I the new medical procedure is tried out on human subjects for the first time, the aim being to determine the subject's maximum tolerance for a drug. In Phase II, therapies which have successfully passed Phase I are given to a larger group of individuals to determine if the procedure is efficacious for treatment of the disease. Phase III involves randomized clinical trials in which some patients receive the experimental treatment and others receive the conventional, nonexperimental treatment; the responses of the two groups are documented, analyzed and compared to assess the efficacy of the experimental treatment as compared with conventional treatments. Phase IV occurs after the drug is approved by the Food and Drug Administration and is to determine whether the drug is effective in other settings. Fuja v. Benefit Trust Life Ins. Co., 18 F.3d 1405, 1410 (7th Cir.1994).
. In Wheeler, plaintiff sought a declaration that her primary insurer, a self-funded plan sponsored by her employer, was obligated to cover her claims for HDC/PSCR breast cancer treatment. Plaintiff had secondary insurance coverage through CHAMPUS due to her husband's prior military service. 850 F.Supp. at 461. The district court's order required plaintiff's primary insurer to cover her claim and further held that CHAMPUS was secondarily liable. Id. at 469. CHAMPUS did not appeal this ruling, id. at 637 n. 2, so the issue of CHAMPUS’ secondary liability was not before the Fourth Circuit.
. Although Smith requested a preliminary injunction, the district court granted a permanent injunction instead because it believed it could not properly enter a preliminary injunction unless Smith posted a bond, which Smith claimed she was financially unable to do. However, the language in Fed.R.Civ.P. 65(C) is sufficiently flexible to permit a district court discretion in setting the amount of the bond, which could include even a nominal amount under appropriate circumstances. See Coyne-Delany Co., Inc. v. Capital Dev. Bd. of State of Ill., 717 F.2d 385, 391 (7th Cir.1983).
Since the district court’s decision, Smith has received the HDC/PSCR treatment. From what the parties told us at oral argument, we are pleased to report that she is alive and no longer undergoing treatment. The issue is not moot because CHAMPUS has not paid the hospital for the costs of this procedure. However, the future relevance of our decision is unclear. Since this case began, CHAMPUS has revised its policy manual to specifically exclude any kind of HDC treatment with stem cell rescue, whether ABMT or PSCR, for breast cancer: "CHAMPUS benefits will not be paid for ... HDC with or without ABMT, HDC with or without PSC[R] ... if not specifically listed as covered [in the policy manual], For example, since breast and ovarian cancers are not specifically listed as covered for HDC with ABMT or PSCfR], CHAMPUS benefits are excluded." 1 CHAMPUS Policy Manual, Ch. 3, § 6.38230.1, Exclusion para. A (Nov. 21 1994) (emphasis added). With this definitive statement of policy, it is questionable whether this holding will have any impact beyond this case.
. This panel’s original opinion in this case was vacated when we granted a rehearing to reconsider the matter in light of the Fourth Circuit’s holding in Wilson v. CHAMPUS, 65 F.3d 361 (4th Cir.1995). Our dissenting colleague criticizes alleged inconsistencies between this opinion and the unanimous vacated opinion. Post at 966. We decline to discuss the details of an unpublished vacated opinion. See Circuit Rule 53(b)(2)(iv); cf. Kawitt v. United States, 842 F.2d 951, 954 (7th Cir.1988).
. The dissent seems to call for a much more vigorous judicial inquiry than our reading of the cases would allow. For instance, the dissent essentially argues at one point that CHAMPUS's inclusion of medical policy organizations and medical scientists in its interpretation of the term "general medical community” — as opposed to just oncologists — is a plainly erroneous interpretation of its regulations. See post at 965 & n. 2. It is clear to us, however, that judicial review must be very deferential, not de novo. 5 U.S.C. § 706(2)(A); Thomas Jefferson Univ., 512 U.S. at -, 114 S.Ct. at 2386; Pozzie, 48 F.3d at 1029. In reviewing a decision by CHAMPUS, this court cannot bypass the deferential standards Congress and the Supreme Court have established, especially in a highly technical area such as this where judges have very limited knowledge. Cf. post at 967-68, 968-69 (disputing explicit conclusions of ECRI report and related news release that HDC/ABMT is experimental for treatment of breast cancer).
. As we noted above, it is our understanding that HDC/ABMT and HDC/PSCR — which both fall under the broader label HDC/ASCR or "HDC with stem cell rescue” — are essentially the same in the treatment of breast cancer. See supra note 2; but see infra note 11.
. ECRI is a nonprofit agency established in 1955 and chartered by the Commonwealth of Pennsylvania that conducts technology assessments for government agencies, health care insurers, and managed-care providers, as well as for hospitals and clinical specialty societies. According to the policy statement accompanying its 1994 “Health Technology Assessment Report” on HDC/ASCR for the treatment of breast cancer, "ECRI reports are produced by a multi disciplinary staff of life scientists, epidemiologists, biostatisticians, engineers, information specialists, and other health professionals.... Neither ECRI nor its employees accept gifts or grants from or consult for medical device or pharmaceutical manufacturers, and each employee's IRS return is examined each year to ensure conformance with ECRI’s stringent conflict-of-interest rules.”
. Differing slightly from CHAMPUS's terminology, the ECRI report appears to use the "HDC/ASCR” label to refer solely to HDC/PSCR, instead of to HDC/PSCR and HDC/ABMT. The reason for the discrepancy' in terms is unclear, but it does not affect our analysis.
. The report continues: "Conventional chemo-therapies have an average median survival time of 15.98 ±n 0.82 months (73 studies, 5,498 patients). HDC/ABMT has an average median survival time of 12.66±n 1.88 months (13 studies; 257 patients), and HDC/ASCR regimens [including HDC/PSCR] yield average median survival times of 10.81 ±n 1.90 months (6 studies, 170 patients).”
.While we could attempt to argue with these conclusions, as the dissent does, post at 967-68, we simply would be taking sides in a medical dispute about which we have no independent expertise.
. For instance, in an affidavit given in September of 1994, Dr. Bogner gave the following as “the current hierarchy of the assessment sources used by the Office of CHAMPUS":
a. Congressional direction.
b. Outcome-based, Phase III trials published in refereed medical literature.
c. Formal technology assessments.
d. National medical policy organization positions.
e. National professional medical associations.
f. National expert opinion organizations.
g. Regional expert opinion organizations.
h. Individual and small group expert opinion. The multiplicity of sources to which CHAMPUS looks in making coverage determinations belies the suggestion that only completed Phase III trials are considered.
. We again emphasize that CHAMPUS's initial decision to treat HDC/ASCR as experimental was rooted in scientific studies questioning its medical efficacy in the treatment of breast cancer. See ante at 957. Only after CHAMPUS had reasonably determined that the treatment was experimental and controversial did it decide that Phase III trials would be necessary to resolve the dispute. We do not have the record in Wilson, but something in it (exactly what is not clear from the opinion) convinced the court that “CHAMPUS [had] relied on an unwritten agency policy mandating Phase III trials before a treatment is provided" thus creating a rigid "prerequisite” that treatments pass Phase III trials before being deemed nonexperimental. Wilson, 65 F.3d at 366. We have found no evidence of such a sweeping prerequisite in the record before us. Rather, the record indicates that in this particular instance CHAMPUS has determined that a Phase III trial is needed to resolve this hotly debated medical issue. Nothing in the record before us suggests that CHAMPUS would have required positive results from Phase III trials if convincing studies or other persuasive medical evidence had supported HDC/ASCR for the treatment of breast cancer.