Brenda EVANS, Plaintiff-Appellee,
v.
EATON CORPORATION LONG TERM DISABILITY PLAN, Defendant-Appellant.

*316 *317 ARGUED: Jeffrey David Zimon, Benesch, Friedlander, Caplan & Aronoff, L.L.P., Cleveland, OH, for Appellant. Robert Edward Hoskins, Foster Law Firm, L.L.P., Greenville, SC, for Appellee. Anna K. Raske, Benesch, Friedlander, Coplan & Aronoff, L.L.P., Cleveland, OH, for Appellant.

Before WILKINSON and SHEDD, Circuit Judges, and CLAUDE M. HILTON, Senior United States District Judge for the Eastern District of Virginia, sitting by designation.

*318 Reversed and remanded by published opinion. Judge WILKINSON wrote the opinion, in which Judge SHEDD and Senior Judge HILTON joined.

OPINION

WILKINSON, Circuit Judge:

Eaton Corporation, a multinational manufacturing company that funds and administers a long-term disability benefits plan for its employees, terminated Brenda Evans's benefits in 2004. She sued under ERISA to recover them, and the district court, faced with substantial conflicting medical evidence and a good case on both sides, concluded that Evans's position was the stronger one. But Eaton was entitled to an abuse of discretion standard of review, and the district court's judgment, though abuse of discretion in name, was de novo in fact. We therefore reverse the district court's award of benefits to Evans and remand with directions that judgment be granted to Eaton.

I.

Under, Eaton's disability plan, a claimant's illness or injury qualifies as a covered disability if "during the first 24 months of such disability . . . you are totally and continuously unable to perform the essential duties of your regular position" and thereafter "you are totally and continuously unable to engage in any occupation or perform any work for compensation or profit for which you are, or may become, reasonably well fitted by reason of education, training or experience—at Eaton Corporation or elsewhere."

Proving a claim requires "[o]bjective findings of a disability," such as "physical examination findings," "diagnostic test results/imaging studies," and "observation of anatomical, physiological or psychological abnormalities." As with so many ERISA plans today, Eaton reserves "discretionary authority to determine eligibility for benefits" and "the power and discretion to determine all questions of fact . . . arising in connection with the administration, interpretation and application of the Plan." A third-party, Broadspire Services, Inc., was Eaton's claims administrator.

In 1998, due to severe rheumatoid arthritis, Brenda Evans quit her job as an order processor at Eaton and filed for long-term disability benefits. Eaton paid without controversy through 2003. But that year, Evans's disability picture started to become cloudy. Her rheumatologist, Dr. Boyd, had prescribed a new arthritis medication, Enbrel, which did her some considerable good: "This is the best she has felt since I have been seeing her," Dr. Boyd's notes from the time report. "[The Enbrel] has helped her quite dramatically. She has very little joint pain now."

From that point forth, medical ambiguity would be the theme of Evans's story. To start with, arthritis was no longer her only problem: A car accident in 2002 had caused serious independent injuries to her back, and Dr. Boyd did not see improvement on that front. Evans's other treating physician, a general practitioner named Dr. Murphy, agreed both as to Evans's arthritic improvement and continuing back problems, and both doctors continued to certify that Evans was totally disabled. On the other hand, an MM and radiology exam from the period indicated that the injuries to Evans's back were not severe. And in a questionnaire, Evans stated that she could cook, shop, do laundry, wash dishes, and drive about seven miles a day.

In January 2004, two of Broadspire's in-house physicians reviewed Evans's 2003 medical records and concluded that she was no longer disabled—though neither did they regard her as fully able. One *319 wrote, for instance, that Evans "would appear" to have "some use of her hands, although perhaps not full use." The other, likewise noting Evans's limited movement and chronic back pain, found that her "rheumatoid arthritis has stabilized" and her back pain did not "preclude the claimant from performing sedentary job duties." Both indicated that more information might be helpful, but on balance recommended ending Evans's benefits. That winter, Evans also took a "Functional Capacity Evaluation," which concluded that she was "capable of performing any job in the light category of work in an 8 hour period with occasional pushing, pulling, standing, walking, climbing stairs and ladders, balancing, stooping, crouching, overhead reaching, desk level reaching, and floor level reaching," and added that "[h]er actual abilities to lift, sit, stand, walk and carry are greater than her perceived abilities." Thus, on April 26, Broadspire sent Evans a benefits cancellation letter.

Evans filed an appeal with Broadspire, and the steady current of conflicting medical evidence and opinion continued. Three more Broadspire physicians, one a rheumatologist, reviewed her file during the appeal process; all three wrote substantial analyses of her condition and concluded that her infirmity was serious but not totally disabling. In July, Broadspire denied the appeal, and Evans filed a second, final appeal (the third review of her condition). This time she supplemented her file with a recent x-ray and x-ray report documenting serious spinal problems, as well as a letter from Dr. Boyd: "[T]here have been notes in my medical reports that have shown that she has been better at times. Saying that she is better is not to say that she is nearly normal or not to say that she is not disabled. One can have a severe arthritic problem and get some improvement with medications and still be incapable functionally of working in a job . . . I do not think her situation will ever change and I think she will be disabled long term."

Two new Broadspire physicians (one a rheumatologist) reviewed the file, including the new materials, and after careful study concluded that Evans could work: "Noteworthy is [the new] letter from Dr. Boyd," the Broadspire rheumatologist remarked. "However, his notes describe dramatic improvement and almost no synovitis. That is not to say that the claimant is simply better but that she has very low levels of disease. Thus there seems to be a contradiction between the information being relayed in Dr. Boyd's [letter] and what is being relayed in his notes." In addition, Broadspire sent Evans's file to an anonymous rheumatologist from the Medical Review Institute of America, who concurred with Broadspire's in-house reviewers that Evans could work. At the end of 2004, Eaton issued its final denial letter.

Evans filed a complaint in the U.S. District Court for South Carolina in September 2005. See Employee Retirement Income Security Act of 1974 (ERISA) § 502(a)(1)(B), 29 U.S.C. § 1132(a)(1)(B) (2000). But, presumably with a view to settlement, the parties agreed to a stay of litigation. As the consent order explains, "[t]he Plaintiff and the LTD Plan . . . have agreed to stay litigation to permit a reconsideration of the final appeal by the Plan. Administrator. . . . Plaintiff has indicated that she intends to undertake certain neurological/orthopedic evaluations which she plans to submit to the Plan Administrator for this final determination reconsideration." Thus Dr. Michael Bucci, a neurosurgeon and the third doctor to examine Evans in person, met with her in April 2006. He found "joint deformity . . . significantly painful to touch," "joint swelling," "diffuse [spinal] tenderness," "difficulty with toe and heel walking," and in sum, severe rheumatoid arthritis and severe *320 spinal stenosis: "I concur with her rheumatologist that the lady is permanently and totally disabled at this time."

Even at this stage, however, the medical opinion was not univocal. Eaton forwarded Evans's file to an independent medical reviewer, Dr. Trangle, who took issue with Dr. Bucci's report and concluded that Evans was not disabled. Dr. Trangle found that, as of "the relevant date of 5/31/04 . . . review of the records indicates that she had barely detectable elevated rheumatoid factor markers." And "[i]n regards to her lower back," the MRI scans and other records from 2003 and 2004 indicate "some spinal stenosis and some radiculopathy," but "it was mild in nature." In July 2006, Eaton rejected Evans's remanded appeal, stating that the records given to Dr. Bucci were incomplete; that Dr. Bucci incorrectly focused on Evans's current medical condition ("I concur . . . that the lady is permanently and totally disabled at this time.") rather than her condition when benefits were cut off in 2004; and that his view conflicted with Dr. Trangle's. Litigation resumed, and the parties agreed to have the court resolve the matter on the basis of memoranda and a joint stipulation, with relevant documents attached.

The district court's opinion of October 2006, after reciting the abuse of discretion standard of review, examined the evidence on both sides in great detail. Most important to the district court, "the only physicians who physically examined Evans . . . concluded that she was totally disabled," and "[t]hese doctors' conclusions," unlike those of the reviewing physicians, "are substantial objective evidence of disability under the Plan." Evans v. Eaton Corp., C.A. No. 8:05-2575-HMH, 2006 WL 2997153, at *9-10 (D.S.C. Oct. 18, 2006). By contrast, in the district court's view, the reviewing physicians' opinions rested on weak foundations, such as Dr. Boyd's notes recording Evans's improvement (which, Dr. Boyd explained, were never meant to indicate that Evans was well enough to work), the functional capacity evaluation (which proved no more than Evans's ability to exert herself for one day, not day in and day out), and. Evans's statements that she could cook, wash dishes, do crafts, and drive (which also did not prove an ability to work). Id. at *10-11.

In addition, the district court believed that the Medical Review Institute physician's report was confused on basic matters of fact (such as whether Evans had ever been considered disabled), and Dr. Trangle's report was contradicted by most of the objective evidence in the record. Id. at *12-14. Finally, Eaton could not in good faith set aside Dr. Bucci's conclusions after agreeing to a stay in order to get them—especially for no better a reason than the obvious and unavoidable fact, implicitly agreed to in the stay, that Dr. Bucci's examination in 2006 would focus on Evans's condition in 2006. Id. at *14-15. In sum, "Eaton abused its discretion in failing to find Evans' examining physicians' opinions more credible than the opinions of Dr. Trangle, Broadspire's in-house physicians, and the [Medical Review Institute] reviewer." Id. at *14. Eaton timely appealed.

II.

This case turns on a faithful application of the abuse of discretion standard of review, and so we' begin with what is most crucial: a clear understanding of what that standard is, and what such standards are for.

A.

The purpose of standards of review is to focus reviewing courts upon their proper role when passing on the conduct of other decision-makers. Standards *321 of review are thus an elemental expression of judicial restraint, which, in their deferential varieties, safeguard the superior vantage points of those entrusted with primary decisional responsibility.^[1] The clear error standard, for example, protects district courts' primacy as triers of fact. See Anderson v. Bessemer City, 470 U.S. 564, 574-75, 105 S. Ct. 1504, 84 L. Ed. 2d 518 (1985). AEDPA's reasonableness standards protect state courts' authority over state criminal convictions. See 28 U.S.C. § 2254(d) (2000). Chevron deference, like the Administrative Procedure Act's arbitrary-and-capricious and substantial evidence standards, protects agencies' authority in carrying out the missions for which they are created. See 5 U.S.C. § 706 (2000); Chevron U.S.A. Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 843, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984). Rational basis review protects the political choices of our government's elected branches. See FCC v. Beach Commc'ns, Inc., 508 U.S. 307, 314, 113 S. Ct. 2096, 124 L. Ed. 2d 211 (1993) (calling the standard "a paradigm of judicial restraint"). And trust law, to which ERISA is so intimately linked, Firestone Tire & Rubber Co. v. Bruck 489 U.S. 101, 110, 109 S. Ct. 948, 103 L. Ed. 2d 80 (1989), uses the abuse of discretion standard to protect a fiduciary's decisions concerning the trust funds in his care. See Restatement (Third) of Trusts § 87 (2007).

The precise definitions of these various standards, the nuances, separating them from one another, "cannot be imprisoned within any forms of words" for "we cannot escape, in relation to this problem, the use of undefined defining terms." Universal Camera Corp. v. NLRB, 340 U.S. 474, 489, 71 S. Ct. 456, 95 L. Ed. 456 (1951) (Frank-furter, J.). But what these and other such standards share is the designation of a primary decision-maker other than the reviewing court, and the instrument, deference, with which that primacy is to be maintained.

B.

The undisputed standard of review in this case is abuse of discretion. Although ERISA itself is silent on the standard for denials of benefits challenged under § 1132(a)(1)(B), Firestone establishes that a de novo standard applies "unless the benefit plan gives the administrator or fiduciary discretionary authority to determine eligibility for benefits or to construe the terms of the plan," in which case the exercise of assigned discretion is reviewed for abuse of discretion. 489 U.S. at 111, 115, 109 S. Ct. 948; see also Booth v. Wal-Mart Stores, Inc., 201 F.3d 335, 341-42 (4th Cir.2000). In the instant case, the Plan's language giving Eaton "discretionary authority to determine eligibility for benefits" and "the power and discretion to determine all questions of fact . . . arising in connection with the administration, interpretation and application of the Plan" is unambiguous, and Evans does not dispute the standard it requires. Thus the district court functions in this context as a deferential reviewing court with respect to the ERISA fiduciary's decision, and we review the district court's decision de novo, "employing the same standards applied by the district court in reviewing the fiduciary's decision." Stup v. UNUM Life Ins. Co., 390 F.3d 301, 306-07 (4th Cir.2004) (internal quotation omitted).

It is notoriously difficult to venture a general definition of the term "abuse of *322 discretion," and none is canonical; indeed the term has different meanings in different legal contexts. See 1 Steven Alan Childress & Martha S. Davis, Federal Standards of Review §§ 1.02, 4.21, at 1-12, 4-131, 4-132 (3d ed.1999) (explaining that the term "fights simple definition" and has a "sliding contextual meaning" with "varying level[s] of deference"). In one sense, abuse of discretion is a term of exquisite balance. The word "abuse" recognizes that authority can be misused. The word "discretion" recognizes that the exercise of authority is often impossible without some leeway for judgment. But taken together, the two words convey an unmistakable message: that as a matter of priority as well as sequence, discretion is first, and review for abuse is only a posterior check on judgment which strays too far from the mark.

The language of "in-bounds" and "out-of-bounds" thus becomes all but irresistible in the abuse of discretion context, for the standard draws a line—or rather, demarcates a region—between the unsupportable and the merely mistaken, between the legal error, disorder of reason, severe lapse of judgment, and procedural failure that a reviewing court may always correct, and the simple disagreement that, on this standard, it may not. See Harry T. Edwards & Linda A. Elliott, Federal Standards of Review 68 (2007) (listing an "erroneous view of the law," "patently arbitrary application of the controlling law," "clearly erroneous assessment of the evidence," judgment call outside "the range of choices permitted," and error "in the weighing process by which [discretion is] exercised" as factors cabining the deference owed on abuse of discretion review (internal quotations omitted)). At its immovable core, the abuse of discretion standard requires a reviewing court to show enough deference to a primary decision-maker's judgment that the court does not reverse merely because it would have come to a different result in the first instance. See Henry J. Friendly, Indiscretion About Discretion, 31 Emory L.J. 747, 754 (1982) ("[T]he trial judge has discretion in those cases where his ruling will not be reversed simply because an appellate court disagrees."). The "deference that is the hallmark of abuse-of-discretion review," Gen. Elec. Co. v. Joiner, 522 U.S. 136, 143, 118 S. Ct. 512, 139 L. Ed. 2d 508 (1997), is deference enough to appreciate reasonable disagreement.

C.

The ERISA context permits a still more particularized conception of the abuse of discretion standard. First, in ERISA cases, the standard equates to reasonableness: We will not disturb an ERISA administrator's discretionary decision if it is reasonable, and will reverse or remand if it is not. Firestone, 489 U.S. at 111, 109 S. Ct. 948; Booth, 201 F.3d at 342. Second, the abuse of discretion standard is less deferential to administrators than an arbitrary and capricious standard would be; to be unreasonable is not so extreme as to be irrational. See Firestone, 489 U.S. at 109-10, 109 S. Ct. 948; Booth, 201 F.3d at 341. Third, an administrator's decision is reasonable "if it is the result of a deliberate, principled reasoning process and if it is supported by substantial evidence." Bernstein v. CapitalCare, Inc., 70 F.3d 783, 788 (4th Cir.1995) (internal quotation omitted). Fourth, the decision must reflect careful attention to "the language of the plan," as well as the requirements of ERISA itself. Booth, 201 F.3d at 342. One adds new assemblages of words to this legal landscape with caution, but it seems on the whole that we require ERISA administrators' decisions to adhere both to the text of ERISA and the plan to which they have contracted; to rest on *323 good evidence and sound reasoning; and to result from a fair and searching process.^[2]

Under no formulation, however, may a court, faced with discretionary language like that in the plan instrument in this case, forget its duty of deference and its secondary rather than primary role in determining a claimant's right to benefits. The abuse of discretion standard in ERISA cases protects important values: the plan administrator's greater experience and familiarity with plan terms and provisions; the enhanced prospects of achieving consistent application of those terms and provisions that results; the desire of those who establish ERISA plans to preserve at least some role in their administration; and the importance of ensuring that funds which are not unlimited go to those who, according to the terms of the plan, are truly deserving. Cf. Gilbertson v. Allied Signal, Inc., 328 F.3d 625, 632 (10th Cir.2003); Johannssen v. District No. 1-Pacific Coast Dist., 292 F.3d 159, 169 (4th Cir.2002); Commc'ns Workers of Am. v. AT & T, 40 F.3d 426, 432 (D.C.Cir. 1994). Thus, the language of discretion in an ERISA plan is a message to courts, counseling not judicial abdication, to be sure, but a healthy measure of judicial restraint.

It is with these principles in mind that we examine the facts at hand.

III.

The most striking feature of the medical evidence in this case is its cross-cutting ambiguity. On Eaton's side, the plan's definition of "disability" is demanding; a claimant must be "totally . . . unable to engage in any" paid work for which the claimant "may become" capable. Dr. Boyd, who was perhaps Evans's strongest advocate, insisted that Evans satisfied that definition, but to some degree his own remarks ("She has very little joint pain now.") undercut that insistence. And no fewer than nine physicians reviewed Evans's file in the course of an initial benefits decision, two appeals, and a remand; all agreed that she did not meet the definition. The quantity is both impressive and, taken by itself, insufficient: As Evans argues, an ERISA administrator cannot prevail merely by multiplying deficient medical opinions, nor by arraying an abundance of low-quality opinions against a few highquality ones. But the opinions in this case are not deficient; the nine reviewing physicians' reports, which are in the record, show every sign of reasoned judgment and good faith.

Broadspire rheumatologist Yvonne Sherrer's two reports are representative of the set. We note first their measured tone, which acknowledges Evans's serious medical problems without a hint of dismissiveness: *324 "[T]his claimant has had substantial disease," she writes, and further examination "is recommended," but Evans also shows "excellent response to therapy" and currently "mild disease limited to the upper extremities only." Second, the substance of Dr. Sherrer's reasoning is analytical rather than rote, and she addresses a thorough and balanced array of medical evidence: "The first x-rays," she writes, showed "some evolution of existent erosions" but no "substantial new change"; three years later, "ftlepeat x-rays of the same areas" also showed "no substantial new changes"; therefore, "these records document stability in x-ray findings. . . ." Third, Dr. Sherrer's conclusions are coherent in themselves and consistent with other reviewers, finding Evans in sum to have "seropositive rheumatoid arthritis" that has "responded to therapy," and "degenerative changes of the cervical and lumbar spine" that, while serious, are not severe enough to "render the claimant incapable of any occupation." Finally, although Dr. Sherrer's reports themselves are not what the Plan characterizes as "[o]bjective findings of a disability," they are based on such findings, including: notes from Evans's treating physicians; x-rays and x-ray reports; a radiology exam and MRI; and the Functional Capacity Evaluation. With eight other physicians submitting reports that display, on the whole, the same virtues of tone, reasoning, and evidence, no one could deny that Eaton's evidence was substantial and its process principled.

The evidence on Evans's side was substantial too. First, there is no dispute that she was disabled at one point—even on Eaton's strict definition of the term. Second, there is no dispute that she remains, and has never ceased to be, impaired. Third, the news that her arthritis shows real improvement with a new medication is by itself an inadequate' reason to cut off benefits when she started from such a low point; as Dr. Boyd argued, she might improve considerably and still remain disabled, nor can one forget that her back problems persisted even as her arthritis improved. Fourth, various diagnostic tests, including bloodwork and an x-ray and x-ray report, substantiated Evans's claims. Fifth and most significantly, every doctor who examined Evans in person concluded that she was disabled.

Dr. Murphy, for example, reviewing Evans's response to Enbrel in July 2003— well before there was any indication that her benefits might be terminated—wrote: "[fi]er arthritis symptoms have improved dramatically except for her back pain and hip pain. This has not improved at all." At the end of 2003 (still prior to the threat of a benefits termination), he completed a comprehensive medical evaluation documenting an array of medical problems: "She has seropositive, non erosive rheumatoid arthritis which has been very symptomatic . She has a chronic pain syndrome involving the right upper extremity after an injury and chronic low back pain which is a combination of post traumatic changes, lumbar facet arthritis and spinal stenosis. She has a history of hypertension, vasomotor instability, normocytic anemia, significant dependent edema which is getting worse over time and recurrent breast cysts. She has pain in the knees and hips, especially with walking. It's very painful to bend or to reach overhead. . . ." True, as Eaton points out, the Supreme Court has rejected the 4treating physician rule" in ERISA cases: "[C]ourts have no warrant to require administrators automatically to accord special weight to the opinions of a claimant's physician; nor may courts impose on plan administrators a discrete burden of explanation when they credit reliable evidence that conflicts with a treating physician's evaluation." Black & Decker Disability Plan v. Nord, 538 *325 U.S. 822, 834, 123 S. Ct. 1965, 155 L. Ed. 2d 1034 (2003). But that point may not be decisive when Eaton's own Plan elevates treating physicians' in-person examinations to the status of "[o]bjective findings of a disability," nor need it negate Dr. Bucci's view (since he was an examining but not a treating physician). Surely a reasonable person could find Evans disabled on such evidence.

Given this state of affairs, the par ties pick endlessly at each other's evidence. Evans argues, for example, that "although Dr. Sherrer did consider the October 2004 x-rays she did not consider the abnormal blood test of April 22, 2004"; Eaton counters that "Dr. Bucci's report failed to interpret the October 2004 x-ray and made no findings regarding the April 2004 lab results." Appellee's Brief 32 (emphasis in original); Appellant's Reply Brief 16. Such point/counterpoint shows what a close case this was, and thus how very important the abuse of discretion standard—which, like other such standards, bites mainly in close cases—should have been. The district court should have acknowledged the essential equipoise and stayed its hand. Elliott v. Sara Lee Corp., 190 F.3d 601, 606 (4th Cir.1999) ("[I]t is not an abuse of discretion for a plan fiduciary to deny disability pension benefits where conflicting medical reports were presented."). Instead, the court joined the fray, purportedly applying an abuse of discretion standard but actually re-weighing the evidence for itself.

Whole paragraphs of the court's opinion, for instance, go to attacking the Functional Capacity Evaluation and other physical evidence in Eaton's favor. Even more of the opinion goes to scrutinizing the work of the twelve doctors involved in some detail, ultimately concluding that Evans's three examining physicians were "substantially more persuasive" than Eaton's nine reviewing physicians. Evans, 2006 WL 2997153, at *8. This language of greater and lesser persuasiveness rings off-key in the abuse of discretion context. It is certainly possible that the evidence on each side of a disability dispute might be so lopsided that a decision for the less persuasive one is an abuse of discretion. But the evidence in this case was close, and the district court's revealing language is persistent: "Eaton abused its discretion in failing to afford greater weight to the physical examination findings of Evans' examining physicians." Id. at *10. "[T]he opinions of the non-examining physicians are not persuasive." Id. "Broadspire's inhouse physician reviewers are not as credible as Evans' examining physicians. . . ." Id. at *11. "[T]he [Medical Review Institute] reviewer is not as credible as Evans' examining physicians." Id. at *12. "Likewise, Dr. Trangle's opinion is not persuasive, and Eaton abused its discretion in crediting Dr. Trangle's opinion over the opinions of Evans' examining physicians." Id. at *13. Most vividly: "[T]he court finds that Eaton abused its discretion in failing to find Evans' examining physicians' opinions more credible than the opinions of Dr. Trangle, Broadspire's in-house physicians, and the [Medical Review Institute] reviewer." Id. at *14. These phrases belong in the voice of a primary factfinder; here they betray a misunderstanding of the district court's reviewing role.

It bears emphasis that the district court's view as to the superiority of the evidence in Evans's favor was not an unreasonable one. Were we reviewing the district court's own decision for abuse of discretion, we would sustain it. But we are reviewing the plan administrator's ruling under that standard, and its decision was not an abuse of discretion in any sense. Where an ERISA administrator rejects a claim to benefits on the strength *326 of substantial evidence, careful and coherent reasoning, faithful adherence to the letter of ERISA and the language in the plan, and a fair and searching process, there can be no abuse of discretion—even if another, and arguably a better, decisionmaker might have come to a different, and arguably a better, result.

Finally, Donovan v. Eaton Corp., 462 F.3d 321 (4th Cir.2006), featured the same defendant, the same plan, and the same general issue as this case. The panel there affirmed the district court's decision for the claimant, according to Evans, because it regarded Eaton as "picking and choosing" medical evidence in its, own favor. Appellee's Brief 28. Consistency, Evans argues, demands the same conclusion from us. But "picking and choosing" is just a pejorative label for "selecting," and what rightly offended the Donovan court was not Eaton's selectivity (which is part of a plan administrator's job), but its "wholesale disregard" of evidence in the claimant's favor. Donovan, 462 F.3d at 329. Here, by contrast, Eaton addressed the evidence in Evans's favor thoughtfully and at length. The benefits cancellation letter of April 2004, for example, gave due regard to the evidence in Evans's favor: "[Y]ou have a history of rheumatoid arthritis," Broadspire wrote, "but recent notes indicate that you had a decrease in your symptoms and are doing well on Enbrel." As to Evans's back, an MRI "revealed degenerative disk disease," but no "functional impairment that would preclude you from performing sedentary job duties." Eaton's assessment of the evidence in this case was fair, and it is entitled to the deference that is the result of careful work.

IV.

So standards of review do matter, for in every context they keep judges within the limits of their role and preserve other decision-makers' functions against judicial intrusion. But deference has a particular significance in the context of ERISA. ERISA's preamble refers to the "interests of employees and their beneficiaries" no fewer than four times in three paragraphs, 29 U.S.C. § 1001 (2000); no one doubts that the statute exists to protect employees' access to benefits, Firestone, 489 U.S. at 113, 109 S. Ct. 948. And yet a cavalier approach to the deference owed ERISA fiduciaries who contract for it would likely disserve that purpose, whatever the call on our compassion in a particular case, for the fact is that the "price [of greater coverage] would almost certainly [be] lower benefits levels and lower levels of plan formation." John H. Langbein, The Supreme Court Flunks Trusts, 1990 Sup.Ct. Rev. 207, 213. For more than thirty years, then, courts have balanced the need to ensure that individual claimants get the benefits to which they are entitled with the need to protect employees and their beneficiaries as a group from a contraction in the total pool of benefits available. At any point, Congress could have intervened. But the delicate balance persists. The district court lost sight of this balance. We therefore reverse the district court's award of benefits to Evans and remand with directions that judgment be granted to Eaton.

REVERSED AND REMANDED

NOTES

[1] De novo review, by contrast, signals no need to protect the primacy of another decisionmaker, because the reviewing court can perform the task as capably as the decisionmaker under review.

[2] In a vivid reminder of Judge Friendly's adage that "there is not just one standard of `abuse of discretion,'" Indiscretion About Discretion, 31 Emory L.J. at 783, Firestone remarks that "if a benefit plan gives discretion to an administrator or fiduciary who is operating under a conflict of interest, that conflict must be weighed as a factor in determining whether there is an abuse of discretion." 489 U.S. at 115, 109 S. Ct. 948 (internal quotation omitted). Thus this circuit "lessen[s] the deference normally given under [an abuse of discretion] standard . . . to the extent necessary to counteract any influence unduly resuiting" from an administrator's conflict of interest—but "in no case does the court deviate from the abuse of discretion standard." Ellis v. Metro. Life Ins. Co., 126 F.3d 228, 233 (4th Cir.1997). The decision as to whether to lessen deference is not always a simple one, compare Colucci v. Agfa Corp., 431 F.3d 170, 178-80 (4th Cir.2005), with Adams v. Louisiana-Pacific Corp., 177 Fed.Appx. 335, 343 n. 3 (4th Cir.2006), and is not fully briefed in this case. But the particular variation on the standard would not alter the result here, and we proceed on the assumption that either variation could apply.