In this case, we consider Nancy Cornelius Martin's claim under the Employee Retirement Income Security Act (ERISA), 29 U.S.C.A. § 1001-1461 (West 1985 & Supp.1997), to obtain insurance benefits for the autologous bone marrow transplant procedure she underwent as treatment for her epithelial ovarian cancer. The magistrate judge1 concluded that the summary plan description (SPD) issued by Blue Cross & Blue Shield of Virginia (Blue Cross) did not exclude the high-dose chemotherapy and peripheral stem cell rescue elements of Mrs. Martin's procedure and issued a declaratory judgment that Blue Cross must therefore cover those elements of the procedure. Because we conclude that the magistrate judge erred in finding that the procedure was not experimental or investigative, we reverse. We also hold that ERISA permits recovery of attorneys' fees only by prevailing parties, and thus reverse the award of attorneys' fees in favor of Mrs. Martin.

Ronald W. Martin, Mrs. Martin's husband, was one of the owners of a business known as Ray Fisher and Ron Martin, Incorporated (the Company). During its twenty-eight years of operations, the Company maintained a welfare benefit plan for its employees. Mrs. Martin, as Mr. Martin's wife, was a beneficiary of the health insurance coverage established and maintained by the Company. On behalf of the Company, Mr. Martin often changed health insurance policies to minimize costs, although in recent years he changed policies less often. In October 1993, Mr. Martin contracted with Blue Cross to provide the Company's health insurance and to administer the plan. Mrs. Martin claims that although the Company received copies of the SPD, Blue Cross never provided the Company with a copy of the insurance contract itself (the Plan).

Around this same time, Mrs. Martin was diagnosed with epithelial ovarian cancer. In 1994, Mrs. Martin was referred to Dr. Steven Wolff, an oncologist at Vanderbilt University Medical Center (Vanderbilt). Dr. Wolff recommended that, to treat her ovarian cancer, Mrs. Martin undergo a regimen of high-dose chemotherapy coupled with stem cell rescue and autologous bone marrow transplant. The parties do not dispute that the procedure at issue consists of three phases. The first of these, stem cell rescue, involves the harvesting of the patient's bone marrow cells.2 Then, while the harvested cells are frozen and stored, the second stage--high-dose chemotherapy--is conducted. The dosage is so strong that the patient's remaining bone marrow cells are destroyed or damaged so severely that they grow back slowly. The third stage involves the reinfusion of the harvested cells into the patient. This reinfusion is necessary to save the patient's immune system, which would otherwise be crippled by the damage to the bone marrow cells during the high-dose chemotherapy stage. Dr. Wolff, Mrs. Martin's treating physician and expert witness, described the entire procedure as "the administration of cytotoxic therapy of such magnitude that the bone marrow, the organ of the body that makes all the blood cells, is substantially injured necessitating the administration or transfusion of additional bone marrow cells to aid in the recovery of bone marrow function induced by the damage of the chemotherapy or radiation therapy." (J.A. at 595.) There is evidence in the record that the term "autologous bone marrow transplant" is often used to describe not only the third stage of the process, but also the entire procedure, including stem cell rescue and high-dose chemotherapy.

Dr. Wolff testified that, at the time of Mrs. Martin's treatment, fewer than five women had ever undergone the same procedure for epithelial ovarian cancer. In fact, the procedure was classified as a Phase II clinical trial. Such clinical trials are regulated by the Food and Drug Administration and the Department of Health and Human Services and require a protocol document to standardize the procedure. As explained by Dr. Wolff, a protocol explains "what you're doing, why you're doing [it], the rationale, who is eligible for this type of therapy, what the therapy is, how to administer the therapy, how to care for the patients, what the complications of expected nature are and how you could analyze the information that you get." (J.A. at 605.) Such clinical trials also require that each participating patient be advised of the nature of the treatment and its attendant risks.

On April 27, 1994, Dr. Wolff requested pre-authorization from Blue Cross, specifically indicating that he was "requesting authorization for the treatment of this patient with high-dose chemotherapy with autologous stem cell and marrow rescue." (J.A. at 17.) Soon thereafter, on April 29, 1994, Blue Cross denied pre-authorization. The offered reason for the denial was that "[h]igh-dose chemo(radio)therapy is considered investigational in the treatment of epithelial ovarian cancer." (J.A. at 18.) On June 6, 1994, Dr. Wolff again wrote to Blue Cross, appealing its initial coverage decision. On June 9, 1994, Blue Cross reaffirmed its denial of coverage, stating that it "must deny proposed high-dose chemotherapy for diagnosis epithelial ovarian carcinoma from two viewpoints: a) policy denial--considered investigational in diagnosis of epithelial ovarian cancer [and] b) contractually excluded per amendment to group policy--not covered for ovarian cancer other than germ cell tumors." (J.A. at 20.) Mrs. Martin's final appeal to Blue Cross was by a letter from her lawyer on July 6, 1994, which included several medical articles on the effectiveness of the recommended treatment for ovarian cancer. Blue Cross responded on August 2, 1994, explaining that

[t]he Technology Evaluation and Coverage (TEC) program of the Medical Advisory panel of the Blue Cross and Blue Shield Association reviewed all of the scientific published peer reviewed literature on [autologous bone marrow transplant] for treating epithelial ovarian cancer. The treatment modality does not meet our coverage eligibility guidelines. I have enclosed a copy of these guidelines for you.

(J.A. at 1014.) Blue Cross appended to this letter its five-factor analysis for determining if a procedure was experimental or investigative.

In July of 1994, before receiving the final letter from Blue Cross but after receiving both denials of pre-authorization, Mrs. Martin decided to undergo the procedure. Mrs. Martin conceded that she "actually underwent the procedure ... with knowledge that Blue Cross was taking the position it wasn't covered." (J.A. at 297.) The procedure was successful, and she was discharged from the hospital in August of 1994.

On November 28, 1994, Mrs. Martin filed this action against Blue Cross in the United States District Court for the Western District of Virginia. She brought the suit under 29 U.S.C.A. § 1132(a)(1)(B) (West 1985), as a beneficiary of an employee welfare benefit plan. From the beginning of the lawsuit, Mrs. Martin maintained that the Company never received a copy of the Plan and therefore that the SPD controlled. The magistrate judge bifurcated the trial to determine, first, which document--the Plan or the SPD--controlled the analysis and, second, whether under the controlling document, Mrs. Martin's treatment was excluded.

On July 5, 1995, the magistrate judge held a hearing on the first issue and ruled from the bench that the Company never received a copy of the Plan. He therefore concluded that the SPD controlled. Later, on December 13, 1995, the magistrate judge held a second evidentiary hearing to determine if Mrs. Martin's treatment was covered. In a memorandum opinion issued on January 22, 1996, the magistrate judge reaffirmed his earlier ruling from the bench that the procedure was not experimental or investigative within the meaning of the SPD's exclusion. He further concluded that an exclusion for "autologous bone marrow transplants" in the SPD did not exclude the related services of high-dose chemotherapy and stem cell rescue. Therefore, the magistrate judge entered a declaratory judgment of coverage for those two phases of the procedure.

Thereafter, Mrs. Martin petitioned the court for attorneys' fees. The magistrate judge granted the motion and, in an order of final judgment entered on March 13, 1996, awarded attorneys' fees. Blue Cross appeals, claiming that the entire procedure was excluded under the Plan's and the SPD's exclusion for experimental or investigative procedures, as well as under the exclusions in the Plan and the SPD for autologous bone marrow transplants. Blue Cross also challenges the award of attorneys' fees in favor of Mrs. Martin. We conclude that under the terms of the Plan and the SPD, the procedure was experimental or investigative and that Mrs. Martin is not entitled to insurance coverage for any phase of the procedure or to attorneys' fees. Accordingly, we reverse.

Blue Cross claims that the magistrate judge erred in concluding that the procedure at issue was not experimental or investigative within the meaning of the Plan's and the SPD's exclusions. The magistrate judge assumed that, because "the SPD controls this case" (J.A. at 748), review of Blue Cross's denial of benefits was de novo. We disagree with the magistrate judge's assumption that the SPD controls the case and with the magistrate judge's application of a de novo standard of review. We will assume without deciding that the Plan itself was never received by the Company--a point the parties argue at considerable length--because, in our view, the receipt or nonreceipt of the Plan is irrelevant. In the Fourth Circuit, "if there [is] a conflict between the complexities of the plan's language and the simple language of the [SPD], the latter [will] control if the participant relied on the SPD or was prejudiced by it." Hendricks v. Central Reserve Life Ins. Co., 39 F.3d 507, 511 (4th Cir.1994) (quotation omitted). In other words, even if the Plan never arrived, it will nonetheless control unless Mrs. Martin can prove both conflict between the Plan and the SPD and that she relied on or was prejudiced by the SPD.

Mrs. Martin fails at the first step. The SPD provided "a brief list of exclusions that apply to [the] program." (J.A. at 837-38.) This "brief list" specifically stated that "Comprehensive Major Medical benefits will not be provided for the following: ... Experimental or investigative procedures...." (J.A. at 837-38.) The Plan itself excludes "[a]ny service determined to be Experimental/Investigative by the Company, in its sole discretion." (J.A. at 900.) The Plan further explains that the phrase

Experimental/Investigative ... describes any service or supply which is judged to be experimental or investigative by the Company in its sole discretion. The Company will use the following criteria to decide this:

1. any supply or drug used must have received final approval to market by the U.S. Food and Drug Administration;

2. there must be enough information in the peer reviewed medical and scientific literature to let the Company judge the safety and efficacy;

3. the available scientific evidence must show a good effect on health outcomes outside a research setting; and

4. the service or supply must be as safe and effective outside a research setting as current diagnostic or therapeutic options.

A service or supply will be experimental or investigative if the Company decides that any one of the four criteria is not met. (J.A. at 885.)3 Therefore, because the Plan gives more detail than the SPD but does not contradict its terms, there is no conflict between the Plan and the SPD.

Circuit precedent supports this conclusion. In Hendricks, we considered an argument that "the district court, in ruling on coverage for his treatment, improperly relied upon definitions of the terms 'experimental' and 'investigative' in the official plan document maintained in Central Reserve's files and did not rest its decision on the terms of the summary plan description circulated to employees." 39 F.3d at 511. There, as here, the summary plan description used only the words "experimental or investigative," while the plan defined the exclusion in greater detail. After carefully reviewing the common understanding of the terms "experimental or investigative," we upheld the view of "the district court that the definitions of 'experimental' and 'investigative' given in the official plan document are not so different from common understandings of the terms that they can be thought to conflict with those understandings." Id. at 512. We think that the reasoning of Hendricks is equally applicable here. Because the additional considerations outlined in the Plan "are not so different from common understandings of the terms," no conflict exists between the SPD and the Plan. As noted in Hendricks, when "the summary plan description and the plan itself do not conflict, our cases provide no prohibition against review of the official plan itself for a fuller understanding of the plan's terms. Indeed, in those circumstances the plan is the controlling document for determining the scope of benefits provided." Id.

We also find that no conflict arises from the Plan's grant of discretion to Blue Cross to determine whether a procedure is experimental or investigative. Although the SPD contains no such language, we find no conflict between the absence of discretionary language in the SPD and its presence in the Plan. Vesting the plan administrator with discretion in making coverage decisions simply does not conflict with the SPD's silence on the matter.4 See Atwood v. Newmont Gold Co., 45 F.3d 1317, 1321 (9th Cir.1995) (holding that even though the summary plan description did not include discretionary language, the grant of discretionary authority in the plan controlled); see also Jensen v. SIPCO, Inc., 38 F.3d 945, 952 (8th Cir.1994) (explaining that the plan will control "when the plan document is specific and the SPD is silent on a particular matter"). Therefore, the Plan controls, and it vests the administrator of the plan with discretion in making coverage decisions.

Because the magistrate judge erred in finding a conflict between the SPD and the Plan and in concluding that the SPD controlled, he also erred in choosing the standard under which to review Blue Cross's denial of benefits. In actions under ERISA challenging the denial of benefits, we apply the standard of review described in Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 109 S. Ct. 948, 103 L. Ed. 2d 80 (1989). Under this standard, we must show deference to the denial of benefits by an administrator with discretionary powers to determine eligibility for benefits, and we will reverse the administrator's denial of benefits only upon a showing of abuse of discretion. See id. at 115, 109 S.Ct. at 956-57; Doe v. Group Hosp. & Med. Servs., 3 F.3d 80, 85 (4th Cir.1993). This standard is slightly modified, however, when the administrator labors under a conflict of interest. In our Circuit,

when a fiduciary exercises discretion in interpreting a disputed term of the contract where one interpretation will further the financial interests of the fiduciary, we will not act as deferentially as would otherwise be appropriate. Rather, we will review the merits of the interpretation to determine whether it is consistent with an exercise of discretion by a fiduciary acting free of the interests that conflict with those of the beneficiaries. In short, the fiduciary decision will be entitled to some deference, but this deference will be lessened to the degree necessary to neutralize any untoward influence resulting from the conflict.

Doe, 3 F.3d at 87; see also Bedrick v. Travelers Ins. Co., 93 F.3d 149, 152 (4th Cir.1996) ("Inasmuch as the law is highly suspect of 'fiduciaries' having a personal interest in the subject of their trust, the 'abuse of discretion' standard is not applied in as deferential a manner to" plans where "the insurer processes and pays claims and acts as plan administrator."). Therefore, the appropriate standard under which to review Blue Cross's denial of benefits to Mrs. Martin is not de novo, as used by the magistrate judge, but rather a modified abuse of discretion standard.

In light of the magistrate judge's application of the wrong standard in reviewing Blue Cross's denial of benefits, remand would ordinarily be appropriate. In our view, however, remand is unnecessary because we conclude, as a matter of law, that the procedure at issue was experimental or investigative within the meaning of the Plan.

We first note that we have twice considered the issue, in different contexts. In Hendricks, 39 F.3d at 509, we considered the identical procedure as administered to a patient with small cell lung cancer. The procedure at issue in Hendricks was a clinical trial, regulated by the Food and Drug Administration and the Department of Health and Human Services. Furthermore, Hendricks signed an informed consent form and was subject to a protocol. Finally, Hendricks was one of the first patients to undergo high-dose chemotherapy with stem cell rescue for his particular type of cancer. We concluded that although

the components of the treatment proposed are fairly well known and data demonstrate that high-dose chemotherapy generally has a tendency to improve a patient's response to cancer, the data failed to demonstrate that small cell lung cancer would be more effectively treated with high-dose chemotherapy than with standard chemotherapy.

Id. at 514. Later, however, in Wilson v. CHAMPUS, 65 F.3d 361 (4th Cir.1995), a case that did not cite Hendricks, we held that autologous bone marrow transplant with high-dose chemotherapy was not experimental or investigative when administered for breast cancer. We rejected the argument that "published, Phase III clinical trial results are required before a benefit can be provided." Id. at 365. Instead, we reasoned that "there is considerable evidence that Phase III clinical trials are not the critical aspect in determining whether a therapy has become 'generally accepted' within the medical community." Id. This reasoning, coupled with our review of the medical literature, led us to conclude that high-dose chemotherapy with stem cell rescue was not experimental or investigative when used to treat breast cancer.

Hendricks and Wilson make clear that an insurance policy may limit coverage depending on the type of cancer involved. Although both cases involved the same procedure, we reached different results for small cell lung cancer and for breast cancer. Here, we conclude that, at the time Mrs. Martin underwent the procedure, Blue Cross did not abuse its discretion in determining that autologous bone marrow transplant with high-dose chemotherapy for epithelial ovarian cancer was experimental or investigative.

Blue Cross extensively reviewed the medical literature before denying Mrs. Martin's application for coverage. In April of 1994, the National Institutes of Health (NIH) produced a consensus statement on "Ovarian Cancer: Screening, Treatment, and Followup." (J.A. at 1035.) The statement specifically noted that "high-dose chemotherapy with hematopoietic growth factors or bone marrow transplantation is experimental, and its use should be limited to research settings." (J.A. at 1046.) Dr. John L. Colley, Blue Cross's expert witness, testified that when the coverage decision for Mrs. Martin was made, Blue Cross "had seen the NIH statement." (J.A. at 568.) Further, as explained by Dr. Colley, "[t]he independent, seventy (70) or so Blue [Cross/Blue Shield] plans across the country[formed an] association [that] has a technology assessment program[the Technology Evaluation Center] that looks at emerging technologies and ... they have a ... very elaborate evaluation process for literature review and attendance at scientific meetings...." (J.A. at 560.) In 1994, the Technology Evaluation Center produced an eighteen-page assessment that "review[ed] the available evidence to determine whether high-dose chemotherapy with autologous stem cell support (HDC/AuSCS) improves survival for patients with epithelial ovarian cell cancer." (J.A. at 1056.) The assessment "focuse[d] on evidence published since 1990, briefly review[ed] the earlier evidence, and synthesiz[ed] the prior conclusions with the new evidence." (J.A. at 1056.) After an extensive review of the available literature in the eighteen-page report, the Technology Evaluation Center concluded that "high-dose chemotherapy with autologous stem cell support for the treatment of epithelial ovarian cancer does not meet the Blue Cross and Blue Shield Association Technology Evaluation Center (TEC) criteria." (J.A. at 1074.) Dr. Colley testified that Blue Cross relied on this assessment in denying coverage for Mrs. Martin's procedure, and Blue Cross specifically mentioned the assessment in its final denial of coverage to Mrs. Martin. Dr. Colley also testified that, when Mrs. Martin's application for coverage was denied, nothing had "come to Blue Cross'[s] attention to cast doubt on the original understanding that [high-dose chemotherapy with autologous bone marrow transplant] was experimental [or] investigative for ovarian cancer." (J.A. at 568.) In fact, there is no testimony in the record that this treatment has yet become generally accepted or standard practice for treatment of epithelial ovarian cancer.

Our conclusion that Blue Cross conducted an adequate review of the medical evidence and did not abuse its discretion in denying benefits is sufficient to dispose of Mrs. Martin's claim. But because the magistrate judge considered evidence beyond that which was before the administrator,5 and because neither party objected to the consideration of such evidence, we note that the extrinsic evidence presented at the hearing does not undermine Blue Cross's denial of coverage.

The Plan requires that, to avoid exclusion as experimental or investigative, the procedure meet four criteria, one of which is that it "be as safe and effective outside a research setting as current diagnostic or therapeutic options." See supra p. 1205. Likewise, one of the five factors included in Blue Cross's final denial of coverage was that "[t]he available scientific evidence must demonstrate a net beneficial effect on health outcomes." (J.A. at 25.) Dr. Wolff, Mrs. Martin's own expert witness, agreed with the statement that "what's experimental is whether the therapy is better than standard alternatives for ovarian cancer." (J.A. at 489.) Therefore, although Dr. Wolff testified that application of Blue Cross's coverage factors otherwise indicated that the procedure was not experimental or investigative, and although his testimony was credited by the magistrate judge, he conceded that at least one of the factors precluded coverage because there was a lack of evidence to prove that this procedure was "as safe ... as current diagnostic or therapeutic options." The Plan makes clear that coverage will be denied if "any one of the four criteria" indicates that the procedure is experimental or investigative, and Dr. Wolff's concession cannot support Mrs. Martin's claim for benefits.

Moreover, Dr. Wolff answered affirmatively when asked whether "high-dose chemotherapy with autologous bone marrow transplant for ovarian cancer patients is experimental as you define the term experimental." (J.A. at 687.) Therefore, Dr. Wolff himself testified that, in the general sense of the term, Mrs. Martin's procedure was experimental or investigative. In other words, not only did Mrs. Martin's expert testify that at least one coverage factor indicated that the procedure was experimental or investigative, he also testified that he himself considered the procedure experimental or investigative.

Finally, Mrs. Martin's procedure, like that in Hendricks, was part of a clinical trial, more specifically a Phase II clinical trial.6 In this regard, the magistrate judge clearly erred in concluding that "what was considered a clinical trial in November 1993 on the evidence before the district judge in Hendricks, was not a clinical trial according to the preponderance of the evidence in this case." (J.A. at 771.) In fact, it is uncontradicted in the record that Mrs. Martin's procedure was part of a Phase II clinical trial. In his deposition, Dr. Wolff explained that "a Phase [II] study really tries to answer the question, in a specific disease situation how effective is the therapy intervention that you're trying. You have to also note that to go in Phase [II] studies means you're not taking standard therapy...." (J.A. at 496.) Although Wilson specifically rejected the argument that lack of a Phase III clinical trial renders a procedure experimental or investigative, Mrs. Martin's participation in a Phase II clinical trial strongly supports Blue Cross's conclusion that the procedure, when administered to Mrs. Martin, was experimental or investigative. Moreover, the required protocol for Mrs. Martin's procedure stated, "THIS STUDY IS ENTIRELY RESEARCH," and, "THIS STUDY IS AN INVESTIGATIONAL TREATMENT PROGRAM." (J.A. at 939.) Finally, Mrs. Martin signed an informed consent form before undergoing the procedure. In this consent form, it was noted that:

Both of these drugs are used in standard therapy for various tumors either separately or in combination. Recently, there is some evidence that each of these agents may be more effective if given at doses much greater than normal. The use of the higher doses and the combination of these drugs at the higher doses is the investigational (research) part of this study. The combination of these drugs in high-doses have been used in only a few patients. Preliminary observations suggest that this therapy may be effective.

It is the purpose of this study to determine the effectiveness and safety (side effects) of this high-dose therapy. It must be understood that compared to standard therapy the use of the therapy in this study will produce more serious side-effects and should be considered substantially more risky.

(J.A. at 951 (emphasis added).) Therefore, Mrs. Martin knew before undergoing the procedure that it was considered experimental or investigative.

In short, Blue Cross did not abuse its discretion in denying benefits to Mrs. Martin because its decision was based on an extensive review of the available literature. Moreover, the extrinsic evidence considered by the magistrate judge supports Blue Cross's denial of coverage. Dr. Wolff testified that at least one of Blue Cross's coverage factors indicated that the procedure was experimental or investigative and that he himself considered the procedure experimental or investigative. In addition, Mrs. Martin herself knew that the procedure was experimental or investigative because it was part of a Phase II clinical trial, was subject to a protocol, and required Mrs. Martin's informed consent. In light of this evidence, we cannot say that Blue Cross abused its discretion, even the more limited discretion afforded to a fiduciary acting under a possible conflict of interest, in concluding that the procedure here was experimental or investigative when used to treat epithelial ovarian cancer.

Finally, we note that we are not at liberty to "fragment" the procedure into its component parts under the exclusion for experimental or investigative procedures. In Hendricks, we concluded that "[t]o fragment the phases of treatment and consider each in light of the policy language produces an unrealistic and distorted analysis." 39 F.3d at 514. In doing so, we distinguished Doe v. Group Hosp. & Med. Servs., 3 F.3d 80 (4th Cir.1993), which held that fragmentation of the treatment was permissible under an exclusion for autologous bone marrow transplants. In Hendricks, we reasoned that because "the treatment which is experimental or investigative includes the high-dose chemotherapy as well as the preparational and recovery phases of the treatment, the scope of the exclusion here is broader" than the scope of the autologous bone marrow transplant exclusion considered in Doe. 39 F.3d at 515. Applying Hendricks, we hold that Mrs. Martin cannot recover the cost of any part of the procedure.7

The magistrate judge also awarded attorneys' fees to Mrs. Martin. We review the award of attorneys' fees under ERISA for abuse of discretion. See Denzler v. Questech, Inc.,, 80 F.3d 97, 103 (4th Cir.1996). Here, the magistrate judge abused his discretion by basing the award of attorneys' fees on a faulty legal standard. The magistrate judge stated that attorneys' fees "under ERISA[ ] should be recovered by the prevailing party absent special circumstances rendering the award unjust." (J.A. at 795.) We have specifically rejected such a standard, holding that "there [is] no presumption in favor of awarding attorney's fees to a prevailing insured or beneficiary." Denzler, 80 F.3d at 104. Therefore, the magistrate judge erred in applying a presumption that attorneys' fees should be recovered by a prevailing ERISA claimant.

Although we would ordinarily remand for an application of the correct legal principles, remand here is unnecessary because Mrs. Martin is not entitled to attorneys' fees as a matter of law. We have held that, in awarding attorneys' fees, we consider five factors:

(1) degree of opposing parties' culpability or bad faith;

(2) ability of opposing parties to satisfy an award of attorney's fees;

(3) whether an award of attorney's fees against the opposing parties would deter other persons acting under similar circumstances;

(4) whether the parties requesting attorney's fees sought to benefit all participants and beneficiaries of an ERISA plan or to resolve a significant legal question regarding ERISA itself; and

(5) the relative merits of the parties' positions.

Id. at 104. Although we have never specifically held that only prevailing parties are entitled to attorneys' fees under ERISA, we have often indicated that reversal of a judgment under ERISA also requires reversal of any attendant award of attorneys' fees. See, e.g., Freeman v. Central States, Southeast & Southwest Areas Pension Fund, 32 F.3d 90, 94 (4th Cir.1994) (explaining that, because the district court clearly erred in its ruling, "[t]he judgment, as well as the derivative award of attorney's fees and costs, must therefore be reversed"); Elmore v. Cone Mills Corp., 23 F.3d 855, 863 (4th Cir.1994) (en banc) (explaining, in an action under ERISA, that "[o]ur reversal of the prior judgment also requires that we vacate the award of attorneys' fees to Plaintiffs"); Fuller v. FMC Corp., 4 F.3d 255, 264 (4th Cir.1993) (reversing a judgment under ERISA and holding that "[b]ecause of our rulings in favor of [the employer], we must also reverse the award in favor of [plaintiffs] for attorney's fees"). We have also suggested, without explicitly holding, that only prevailing parties are entitled to be considered for an award of attorneys' fees under ERISA. See Custer v. Pan American Life Ins. Co., 12 F.3d 410, 423 (4th Cir.1993) (stating that a plaintiff under ERISA "must demonstrate more than merely being the prevailing party on a single issue to demand entitlement to attorney's fees"). Moreover, many of our sister circuits have imposed a "prevailing party" limitation on the availability of attorneys' fees under ERISA. See Cottrill v. Sparrow, Johnson & Ursillo, Inc., 100 F.3d 220, 225 (1st Cir.1996) (considering "the degree of culpability or bad faith attributable to the losing party" and "the depth of the losing party's pocket" (emphasis added)); Boggs v. Boggs, 82 F.3d 90, 94 n. 1 (5th Cir.) (explaining that ERISA "allows the court to award ERISA beneficiaries, participants, and fiduciaries reasonable attorney's fees and costs when they are the prevailing party"), cert. granted, --- U.S. ----, 117 S. Ct. 379, 136 L. Ed. 2d 297 (1996); Little v. Cox's Supermarkets, 71 F.3d 637, 644 (7th Cir.1995) (asking the " 'bottom-line question' " of whether "the losing party's position [was] substantially justified and taken in good faith" (emphasis added)); Eddy v. Colonial Life Ins. Co., 59 F.3d 201, 206 (D.C.Cir.1995) (considering the "losing party's culpability or bad faith" and "the losing party's ability to satisfy a fee award" (emphasis added)); McPherson v. Employees' Pension Plan, 33 F.3d 253, 254 (3d Cir.1994) (noting that "[a]ttorneys' fees may be awarded to prevailing parties in actions brought under" ERISA); see also Flanagan v. Inland Empire Elec. Workers Pension Plan & Trust, 3 F.3d 1246, 1253 (9th Cir.1993) (admitting that it had "said in dictum that [the ERISA fee provision] permits an award of fees to a non-prevailing party," but holding that "plaintiffs cannot recover fees under section 1132(g)(1) until they succeed on any significant issue in litigation which achieves some of the benefit [they] sought in bringing suit" (quotations omitted and alterations in original)). But cf. Miller v. United Welfare Fund, 72 F.3d 1066, 1074 (2d Cir.1995) (stating that ERISA "contains no requirement that the party awarded attorneys' fees be the prevailing party"). Because of the strong suggestions in our own precedent, supported by the weight of authority from other circuits, we now make clear that in the Fourth Circuit, only a prevailing party is entitled to consideration for attorneys' fees in an ERISA action. We have denied Mrs. Martin's claim in full, and therefore she is not entitled to attorneys' fees. Accordingly, we reverse the magistrate judge's award of attorneys' fees.

We hold that the magistrate judge erred in concluding that Mrs. Martin's procedure was not experimental or investigative. Because circuit precedent precludes fragmentation of the autologous bone marrow transplant procedure for purposes of the experimental or investigative exclusion, we deny Mrs. Martin's claim in full, including her claim for attorneys' fees.

REVERSED.