E. GRADY JOLLY, Circuit Judge:

This case involves a denial of health benefits claimed by Jose Vega and his wife, Vilma Vega, under a health benefits plan they established for themselves and the employees of their business, the Corona Paint Body Shop, Inc. ("Corona"). The Vegas sued the insurance companies responsible for insuring and maintaining the plan, Pan-American Life Insurance Co. ("Pan-American") and National Life Insurance Services, Inc. ("National Life") — a subsidiary of Pan-American. The district court granted summary judgment for the insurance companies, relying in part on its holding that it could not consider additional evidence submitted by the Vegas to the district court when that evidence was not available to the plan administrator at the time it reached its decision. On appeal, a panel reversed the district court, holding that the district court erred in not considering the evidence presented by the Vegas.

We heard this case en banc to address three issues. First, the Vegas argue that we do not have jurisdiction under the Employee Retirement Income Security Act ("ERISA"), 29 U.S.C. § 1001 et seq., because the Vegas, as the sole owners of Corona, were not employees as that term is defined by the statute and related Department of Labor regulations. This issue has divided the Circuits and we recognize the need to clarify the scope of ERISA in this context. We hold today that where a husband and wife are sole owners of a corporation that has created an employee benefits plan covered by ERISA, and the husband and wife are also enrolled under the plan as employees of the corporation, they are employees for ERISA purposes and so our courts have jurisdiction under ERISA to review a denial of their claims.

The second issue we address is the panel's approach to reviewing a decision of an administrator operating under a conflict of interest, which in this case is that the corporation deciding the claim will have to pay the claim. Although in the past we have repeatedly stated that the district court may not engage in additional fact finding, the panel here sought to carve out an exception for conflicted administrators. The panel held that, when the administrator has a conflict of interest in denying a claim, it must meet a duty to conduct a good faith, reasonable investigation. In determining whether the administrator has met this duty, the panel elected to consider evidence that it believed such an investigation would have uncovered. We hold today that no such specific and uniform duty exists. We further hold that evidence may not be admitted in the district court that is not in the administrative record when that evidence is offered to allow the district court to resolve a disputed issue of material fact regarding the claim — i.e., a fact the administrator relies on to resolve the merits of the claim.

Finally, we turn to the merit of the summary judgment ruling by the district court. Even though the district court correctly refused to consider the additional evidence proffered by the Vegas, the district court nonetheless erred in upholding the administrator's denial of the claim. After reviewing the administrative record, we find no rational basis is contained therein for denying the Vegas' claim and therefore conclude that National Life abused its discretion.

I

The Vegas are the sole owners of Corona, a corporation structured as a Subchapter S corporation under the Internal Revenue Code. On March 20, 1995, Mr. Vega, on behalf of Corona, applied for an employer-sponsored group medical plan with Pan-American. Pan-American issued the policy, which covered Mr. Vega as an employee and his wife as a dependent. Under the plan, Pan-American was the insurer and National Life, a subsidiary of Pan-American, acted as the claims administrator of the plan. The plan granted National Life discretion in deciding claims.

In filling out the form for his wife, Mr. Vega denied that she had received any advice, consultation, or test for any medical condition (other than a recovered bladder infection) during the previous six months. Less than two months after Pan-American approved coverage for Mrs. Vega, she saw Dr. Bueso, who recommended surgery for posterior repair of the vagina. Mrs. Vega underwent the surgery and processed her claim for coverage under the plan.

In reviewing the medical records related to the claim, National Life discovered a notation by Mrs. Vega's gynecologist, Dr. Galvan, dated October 5, 1994, that stated "posterior repair." A representative of National Life then called Dr. Galvan's office and asked about the notation. National Life kept phone logs of two phone calls related to the inquiry. In the first phone call, the representative spoke to an assistant of Dr. Galvan's, Ramone, who told the representative that she would ask Dr. Galvan and call back. The second log states:

S/W Ramone

Last 2 entries were from phone conversations

Last pc was when patient called Dr. Galvan back and Ms. Vega had some questions regarding a surgical procedure. Dr. Galvan answered her questions about the procedure and wrote note in pt chart because they talked about it.

Was she anticipating surgery? He (Dr. Galvan) said she had questions and he answered them. Doesn't recall what prompted conversation.

On the basis of this information, National Life decided to deny Vega's claim.

National Life sent a letter to the Vegas explaining to them that it was denying the Vegas' claim. The letter stated:

During processing of your claims we learned that the information contained on your GEC regarding your health history was not accurate. Specifically, medical records received and reviewed from Dr. Pineda and Dr. Galvan indicate that your response to question number 3 was incorrect. Dr. Galvan's medical records indicate that on September 29, 1994 he consulted Ms. Vega for a check up and relaxation of tissue with breast tenderness. The records further state that on October 5, 1994, he recommended a posterior repair. Dr. Pineda's records indicate that on May 10, 1995 Ms. Vega obtained a consultation complaining of galactorrhea and a cytology was recommended. Had you advised us of Ms. Vega's medical history as you were obligated to do, coverage would have been denied at initial underwriting.

The letter went on to state:

The URB [Underwriting Review Board] remains willing to review and consider any additional information you may have which you feel would impact on our decision to rescind coverage. If you wish to appeal this decision, please do so in writing and send it to the attention of the Underwriting Review Board at the company address listed below.

The Vegas did not submit a request for review of the decision but instead hired an attorney who sent a letter on their behalf indicating that if Pan-American did not pay the claim, the Vegas would sue. Pan-American sent a reply indicating that it was prepared to go to court.

Shortly thereafter, the Vegas filed suit in state court alleging state law causes of action. Pan-American removed the action to federal court and each side sought summary judgment. In the pleadings filed in district court, the Vegas attempted to introduce testimony from Mrs. Vega's physicians (Dr. Galvan and Dr. Bueso) contradicting Pan-American's claim that, at the time the Vegas enrolled in the plan, Mrs. Vega contemplated posterior repair surgery. Pan-American then attempted to introduce expert testimony supporting its conclusion as a fair reading of the medical records.

The district court granted summary judgment to Pan-American after concluding that ERISA applied to the dispute and that Pan-American had not abused its discretion in denying the medical claim and rescinding coverage. The district court concluded that it could not consider the testimony introduced by the Vegas as it was not available to the plan administrator. On appeal, the panel held that such testimony could be considered, as there was evidence that Pan-American had violated its duty to conduct a "reasonable, good faith investigation of the claim." In reaching this conclusion, the panel relied heavily on the affidavits prepared by Dr. Galvan and Dr. Bueso. Vega v. National Life Ins. Services, Inc., 145 F.3d 673, 678-79 680-81 (5th Cir. 1998) (quoting full text of affidavits and treating testimony as relevant evidence for summary judgment purposes), reh'g en banc granted and vacated, 167 F.3d 197 (5th Cir. 1999).

II

The first issue we must address is whether the federal courts have jurisdiction under ERISA to hear this appeal. The Vegas contend that the trial court and the panel erred in concluding that ERISA covers this dispute. According to the Vegas, their status as sole shareholders of Corona renders Mrs. Vega neither a participant nor a beneficiary for ERISA purposes, so ERISA does not govern their claims. They urge that their suit belongs in state court.

ERISA preempts all state claims that "relate to any employee benefit plan." 29 U.S.C. § 1144(a). In order for ERISA to govern the Vegas' claims, two criteria must be met: (1) an employee benefit plan must exist, and (2) Mrs. Vega must have standing to sue as a participant or beneficiary of that employee benefit plan. See Madonia v. Blue Cross Blue Shield of Virginia, 11 F.3d 444, 446 (4th Cir. 1993); Apffel v. Blue Cross Blue Shield of Texas, 972 F. Supp. 396, 398 (S.D. Tex. 1997).

The ERISA statute defines "employee welfare benefit plan" as:

any plan, fund, or program . . . established or maintained by an employer . . . to the extent that such a plan, fund, or program was established or is maintained for the purpose of providing for its participants or their beneficiaries, through the purchase of insurance or otherwise, (A) medical, surgical, or hospital care or benefits . . . .

29 U.S.C. § 1002(1). The panel stopped at the initial step — finding that there was an ERISA plan — and determined that ERISA governed the Vegas' lawsuit. The first step is the easy part, however. In fact, the Vegas agree that an ERISA plan exists to the extent that the plan is established or maintained for the purpose of providing benefits to employees who are plan participants and their beneficiaries.

The Vegas' argument is that the plan as it regards Mrs. Vega is not an ERISA plan because she is not a participant or beneficiary. That is, to have standing to bring a suit under ERISA, a person must be either a "participant" or a "beneficiary" of an ERISA plan, see Weaver v. Employers Underwriters, Inc., 13 F.3d 172, 177 (5th Cir. 1994), and Mrs. Vega argues that her status as a co-owner precludes her from having standing to assert claims under ERISA. Thus, determining whether Mrs. Vega is a participant or beneficiary is key, and the panel did not reach the issue.

To be considered a "participant" of the plan, a claimant must be an "employee or former employee of an employer . . . who is or may become eligible to receive a benefit . . . ." 29 U.S.C. § 1002(7). Unhelpfully, ERISA defines "employee" as "any individual employed by an employer." Id. § 1002(6). A beneficiary is defined as "a person designated by a participant, or by the terms of an employee benefit plan, who is or may become entitled to a benefit thereunder." 29 U.S.C. § 1002(8).

Because Mrs. Vega is listed under the policy as her husband's dependent, her claims are governed by ERISA only if her husband qualifies as an employee under ERISA. Because Mr. Vega is a co-owner, we must decide whether a shareholder of a Texas corporation is precluded from qualifying as an employee for ERISA purposes.

The Circuits are not in accord as to whether an individual partner, sole proprietor, or shareholder may be a "participant" in an ERISA plan established for the business's employees. CompareFugarino v. Hartford Life Accident Ins. Co., 969 F.2d 178, 185-86 (6th Cir. 1992) (holding that ERISA did not apply to insurance coverage of a sole proprietor and his family even though the group insurance policy purchased by the sole proprietor established an ERISA plan with respect to the sole proprietor's employees); Kwatcher v. Massachusetts Serv. Employees Pension Fund, 879 F.2d 957, 959-60 (1st Cir. 1989) (holding that the sole shareholder of a corporation was not an "employee" within the meaning of ERISA and, therefore, could not participate in an ERISA plan); Brech v. Prudential Ins. Co. of America, 845 F. Supp. 829, 832-33 (M.D. Ala. 1993) (holding that on account of his position as an employer, the sole proprietor of a company was not a participant of an employee benefit plan he had established with the purchase of a group insurance policy for himself and his employee sons, and ERISA therefore did not preempt his state law claims); and Kelly v. Blue Cross Blue Shield of Rhode Island, 814 F. Supp. 220, 226-29 (D.R.I. 1993) (holding that the sole shareholder and chairman of the board of the employer corporation was not a participant or beneficiary with respect to a health insurance policy purchased for the sole shareholder and the company's employees) with Madonia v. Blue Cross Blue Shield of Virginia, 11 F.3d 444, 449-50 (4th Cir. 1993) (holding that a physician's status as the corporation's sole shareholder did not bar him from being an "employee" under ERISA's definition of the term).

The cases holding that owners cannot also count as employees for purposes of ERISA base their conclusion on their interpretation of the ERISA statute, relevant regulations, and legislative history. First, pointing to the definitions given in 29 U.S.C. § 1002, the cases assert that Congress "meant to divorce owner-employees from plan participation." See Kwatcher, 879 F.2d at 959. An "employee" is "any individual employed by an employer." 29 U.S.C. § 1002(6). The statute defines the term "employer" as "any person acting directly as an employer, or indirectly in the interest of an employer . . . ." Id. § 1002(5). While these definitions are not particularly helpful substantively, they do reveal, according to these cases, that an employee and an employer are different beings — they cannot be the same person for ERISA purposes. These cases further reason that, as a matter of "economic reality," an owner should be considered an employer rather than an employee because he "dominates the actions of the corporate entity." Kwatcher, 879 F.2d at 960.

Furthermore, Department of Labor regulations related to defining when a plan is covered by ERISA provide that:

An individual and his or her spouse shall not be deemed to be employees with respect to a trade or business, whether incorporated or unincorporated, which is wholly owned by the individual or by the individual and his or her spouse . . . .

29 C.F.R. § 2510.3-3(c)(1) (1992). Various courts have concluded that this regulation, clarifying the definition, is reasonable and controlling. See Meredith v. Time Ins. Co., 980 F.2d 352, 357 (5th Cir. 1993); Kelly, 814 F. Supp. at 227. The Fifth Circuit has held that the owner of a business cannot, for purposes of ERISA, simultaneously be an employer and an employee. See Meredith, 980 F.2d at 358 (holding that an insurance plan covering only a sole proprietor and her spouse was not an ERISA employee welfare benefit plan because the plan did not benefit employees).

Finally, these cases posit that the purpose of ERISA suggests that an employer's policy is not part of an ERISA plan because Congress intended, in passing ERISA, to provide protection for employees, not employers. See Kwatcher, 879 F.2d at 960 (stating that "ERISA's framers were concerned that employers would exploit, misuse, or loot the huge reserves of funds collected for employee benefit plans"); Kelly, 814 F. Supp. at 228. These cases contend that Congress' intent to exclude owners from ERISA coverage is revealed in the statute's "anti-inurement" provision:

[T]he assets of a plan shall never inure to the benefit of any employer and shall be held for exclusive purposes of providing benefits to participants in the plan and their beneficiaries and defraying reasonable expenses of administering the plan.

29 U.S.C. § 1103(c)(1).

Madonia, a Fourth Circuit case, reached a different conclusion. Madonia considered the status of a physician who was the director, president, and sole shareholder of MNA, the corporation that he founded. MNA had an employee welfare benefit plan established for the corporation's employees. The court held that the doctor qualified as an "employee" of MNA under ERISA's definition of the term — "employee" is "any individual employed by an employer," 29 U.S.C. § 1002(6) — because he was an "individual" and he was "employed by" the corporation. See Madonia, 11 F.3d at 448.

The court also looked beyond the definition and followed the Supreme Court's mandate in Nationwide Mut. Ins. Co. v. Darden, 503 U.S. 318 (1992), that it should employ "`a common-law [agency] test for determining who qualifies as an `employee' under ERISA. . . .'" Madonia, 11 F.3d at 448-49 (quoting Darden, 503 U.S. at 323). Darden, in which the Court grappled with the question of whether an individual qualified as an employee or an independent contractor, directed that, because the statutory definition of "employee" was "circular and explain[ed] nothing," courts should use a common-law test for determining who qualifies as an employee under ERISA. 503 U.S. at 323. And, because the common-law test contains "no shorthand formula or magic phrase that can be applied to find the answer . . .[,] all of the incidents of the [employment] relationship must be assessed and weighed with no one factor being decisive." Id. at 324 (quotingNLRB v. United Ins. Co. of America, 390 U.S. 254, 258 (1968)).

In discerning common-law principles in Madonia, the court looked to relevant Virginia law, which recognized that a "corporation is a legal entity separate and distinct from its shareholders" and that the corporate form may be disregarded only in extraordinary circumstances. Madonia, 11 F.3d at 449. In accordance with this principle, the court held that MNA's corporate identity was the "employer," and Dr. Madonia's separate identity was MNA's "employee." See id. The court specifically distinguished its case from cases dealing with sole proprietors of unincorporated businesses that, by definition, have no separate legal existence. See id. at n. 2 (noting that "[t]he question of a sole proprietorship is not one that is before us").

Madonia also addressed 29 C.F.R. § 2510.3-3(c)(1), the DOL regulation that provides that an owner of a business is not to be considered an employee. The court insisted that the introductory clause of the regulation, "[f]or purposes of this section," refers to 29 C.F.R. § 2510.3-3, which deals exclusively with the determination of the existence of an employee benefit plan. Therefore, according to the court, the regulation's exclusion of business owners from the definition of "employee" is "limited to its self-proclaimed purpose of clarifying when a plan is covered by ERISA and does not modify the statutory definition of employee for all purposes." Madonia, 11 F.3d at 449 (quoting Dodd v. John Hancock Mut. Life Ins. Co., 688 F. Supp. 564, 571 (E.D. Cal. 1988)). In other words, "[t]he regulation does not govern the issue of whether someone is a `participant' in an ERISA plan, once the existence of that plan has been established," id. at 449-50; instead, because the regulations provide that the plan must involve at least one employee, 29 C.F.R. § 2510.3-3(c)(1) simply insures that owners are not counted as employees to satisfy the "employee" requirement. The court thought that its result made "perfect sense" because "it would be anomalous to have those persons benefitting from [the employee benefit plan] governed by two disparate sets of legal obligations." Id. at 450. The court also believed that its result promoted Congress' objective of achieving uniformity through the enactment of ERISA because it fostered consistency in the law governing employee benefits. See id.

We are persuaded by the reasoning in Madonia and interpret the regulations to define employee only for purposes of determining the existence of an ERISA plan. Under this approach, Corona's employee benefit plan is an ERISA plan because it does not solely cover the Vegas, co-owners of the company; rather, it includes their employees, and Corona employs at least one other person besides the Vegas. Mr. Vega's status as a co-owner does not automatically foreclose ERISA coverage. Instead, whether Mrs. Vega has standing to bring ERISA claims under the plan depends on whether common-law principles define her husband as an "employee" or an "employer."

As in Madonia, the entity in this case is a corporation. Texas courts recognize the basic proposition that a corporation is a legal entity separate and distinct from its shareholders. See Western Horizontal Drilling, Inc. v. Jonnet Energy Corp., 11 F.3d 65, 67 (5th Cir. 1994). Also like Virginia law in Madonia, Texas law permits disregard of the corporate form only in very limited circumstances. See id. Thus, we hold that Corona's corporate form was the "employer," Mr. Vega was an "employee," and Mrs. Vega was her husband's beneficiary — in short, ERISA applies to this case, so this court has jurisdiction. As long as a Texas business corporation maintains a plan and at least one employee participant (other than a shareholder or a spouse of the shareholder), an employee shareholder and his beneficiaries may be participants in the plan with standing to bring claims under ERISA.

III

We turn now to the panel's assessment of the district court's summary judgment ruling. On appeal, the panel agreed that ERISA applied to the case, but it reversed and remanded, concluding that Pan-American abused its discretion in denying Mrs. Vega's claim and rescinding her coverage. The panel based this conclusion on Pan-American's failure to conduct a reasonable, good faith investigation of the facts surrounding the Vegas' claim. In doing so, the panel relied on the additional evidence presented by the Vegas in district court.

Although we have dealt with cases involving similar fact patterns, we have never before explicitly imposed a duty like the one imposed by the panel here. In reaching its decision, the panel relied on the existence of a conflict of interest between National Life's role as the administrator and Pan-American's role as the insurer. We therefore first address the role that a conflict of interest plays in our analysis. We then turn to the merits of imposing such a duty on the plan administrator, concluding that such an imposition is inappropriate. Finally, we reaffirm our long-standing rule that when performing the appellate duties of reviewing decisions of an ERISA plan administrator, the district court may not engage in fact-finding.

A

ERISA provides the federal courts with jurisdiction to review determinations made by employee benefit plans, including health care plans. 29 U.S.C. § 1132(a)(1)(B). Generally, there are two ways employee benefit plans may be created: (1) the employer funds the program and either contracts with a third party who administers the plan or provides for administration by a trustee, individual, committee, or the like; or (2) the employer contracts with a third party that both insures and administers the plan. In the latter situation, the administrator of the plan is self-interested, i.e., the administrator potentially benefits from every denied claim.

Section 1132(a)(1)(B) does not provide any guidance regarding the standard of review to be employed by the federal courts. InFirestone Tire Rubber Co. v. Bruch, 489 U.S. 101 (1989), the Supreme Court addressed this issue holding:

[A] denial of benefits challenged under § 1132(a)(1)(B) is to be reviewed under a de novo standard unless the benefit plan gives the administrator or fiduciary discretionary authority to determine eligibility for benefits or to construe the terms of the plan. . . . Of course, 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 "facto[r] in determining whether there is an abuse of discretion."

Id. at 115 (citation omitted).

Under Bruch, therefore, when an administrator has discretionary authority with respect to the decision at issue, the standard of review should be one of abuse of discretion. However, the proceeding by which an administrator denies a claim tends not to be as well defined as, for example, adjudicatory hearings under the APA. The issues are often further complicated by the relative sophistication of the parties. An employer may have an incentive to choose a less expensive benefit plan for his employees, even though that plan grants a self-interested administrator discretion to resolve claims. When an employee files such a claim, the administrator has a financial incentive to deny the claim and often can find a reason to do so. The employee, on the other hand, is often not a sophisticated negotiator and therefore may not best present his case to the administrator.

In the interim, since Bruch, our Circuit has struggled with the appropriate standard of review for determinations by a self-interested administrator with discretionary authority. See,Salley v. E.I. DuPont de Nemours Co. 966 F.2d 1011 (5th Cir. 1992) (holding that a conflict of interest required the court to more closely examine the denial of health care benefits under the plan); Duhon v. Texaco, Inc., 15 F.3d 1302, 1306 (5th Cir. 1994) (holding that court must "weigh this possible conflict as a factor in our determination of whether the plan administrator abused his discretion, instead of . . . altering the applicable standard of review"); Sweatman v. Commercial Union Insurance Co., 39 F.3d 594, 599 (1994) (holding that conflict does not change the standard of review, but should be weighed in determining whether administrator abused his discretion); Wildbur v. ARCO Chem. Co., 974 F.2d 631, 638-42 (5th Cir. 1992) ("We note that the arbitrary and capricious standard may be a range, not a point. There may be in effect a sliding scale of judicial review of trustees' decisions — more penetrating the greater is the suspicion of partiality, less penetrating the smaller that suspicion is . . . .").

Other Circuits have also struggled with the role a conflict of interest should play in determining whether an administrator has abused its discretion. The Tenth Circuit, in Chambers v. Family Health Plan Corp., 100 F.3d 818, 824-27 (10th Cir. 1996), catalogs the various approaches under two categories: the "sliding scale" standard and the "presumptively void" standard.

Under the "sliding scale" standard, the court always applies the abuse of discretion standard, but gives less deference to the administrator in proportion to the administrator's apparent conflict. An example of this approach is the Fourth Circuit decision in Doe v. Group Hospitalization Medical Services, 3 F.3d 80 (4th Cir. 1993):

We hold that 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.

Id. at 86; see also Chambers, 100 F.3d at 826 (holding "that the sliding scale approach more closely adheres to the Supreme Court's instruction to treat a conflict of interest as a `facto[r] in determining whether there is an abuse of discretion'"); Sullivan v. LTV Aerospace Defense Co., 82 F.3d 1251, 1255 (2d Cir. 1996);Taft v. Equitable Life Assurance Soc'y, 9 F.3d 1469, 1474 (9th Cir. 1993); Van Boxel v. Journal Co. Employees' Pension Trust, 836 F.2d 1048, 1052-53 (7th Cir. 1987).

Under the presumptively void standard, once a claimant has demonstrated that the administrator acted out of self-interest, the administrator then has the burden of establishing that its action was nevertheless in the plan's interest. An example of the application of this approach is the Eleventh Circuit's decision inBrown v. Blue Cross Blue Shield of Alabama, Inc., 898 F.2d 1556 (11th Cir. 1990):

[W]hen a plan beneficiary demonstrates a substantial conflict of interest on the part of the fiduciary responsible for benefits determinations, the burden shifts to the fiduciary to prove that its interpretation of plan provisions committed to its discretion was not tainted by self-interest. That is, a wrong but apparently reasonable interpretation is arbitrary and capricious if it advances the conflicting interest of the fiduciary at the expense of the affected beneficiary or beneficiaries unless the fiduciary justifies the interpretation on the ground of its benefit to the class of all participants and beneficiaries.

Id. at 1566-67. Another example of this kind of approach is the Ninth Circuit's opinion in Atwood v. Newmont Gold Co., Inc., 45 F.3d 1317, 1323 (9th Cir. 1995):

The "less deferential" standard under which we review apparently conflicted fiduciaries has two steps. First, we must determine whether the affected beneficiary has provided material, probative evidence, beyond the mere fact of the apparent conflict, tending to show that the fiduciary's self-interest caused a breach of the administrator's fiduciary obligations to the beneficiary. If not, we apply our traditional abuse of discretion review. On the other hand, if the beneficiary has made the required showing, the principles of trust law require us to act very skeptically in deferring to the discretion of an administrator who appears to have committed a breach of fiduciary duty.

Id.; see also Armstrong v. Aetna Life Ins. Co., 128 F.3d 1263, 1265 (8th Cir. 1997) (finding conflict and reviewing decisions de novo); Kotrosits v. GATX Corp. Non-Contributory Pension Plan for Salaried Employees, 970 F.2d 1165, 1173 (3d Cir. 1991) (holding that the court will withhold deference when the administrator has been shown to be biased by a conflict of interest).

In the interim since Chambers, the First Circuit has issued an opinion that defies the neat categories set forth in Chambers. In Doe v. Travelers Ins. Co., 167 F.3d 53, 57 (1st Cir. 1999), the court held:

[T]he requirement that [the administrator's] decision be "reasonable" is the basic touchstone in a case of this kind and that fine gradations in phrasing are as likely to complicate as to refine the standard. The essential requirement of reasonableness has substantial bite itself where, as here, we are concerned with a specific treatment decision based on medical criteria and not some broad issue of public policy. Any reviewing court is going to be aware that in the large, payment of claims costs [the administrator] money. At the same time, the policy amply warns beneficiaries that [the administrator] retains reasonable discretion based on medical considerations.

Id.

Having polled the other Circuits, we reaffirm today that our approach to this kind of case is the sliding scale standard articulated in Wildbur. The existence of a conflict is a factor to be considered in determining whether the administrator abused its discretion in denying a claim. The greater the evidence of conflict on the part of the administrator, the less deferential our abuse of discretion standard will be. Having said that, we note that we sympathize with the First Circuit's approach — our review of the administrator's decision need not be particularly complex or technical; it need only assure that the administrator's decision fall somewhere on a continuum of reasonableness — even if on the low end.

B

The panel opinion offers a new solution to the problem of how to evaluate decisions by self-interested administrators. According to the panel, when reviewing such a decision, the district court should determine whether the administrator has met its duty to conduct a good faith, reasonable investigation. If not, the court need not restrict its review to the facts before the administrator:

The court must pause, before limiting itself to the record before the administrator, to assure itself that the administrator conducted a reasonable, good faith investigation of the claim. That requirement must be cautiously and carefully imposed when the administrator has the inherent conflict of interest as exists in the case at bar. To hold otherwise would restrict the district court to reviewing only those materials before the administrator, even in cases where the administrator conducted an unreasonably lax, bad faith investigation of the facts.

Vega, 145 F.3d at 680 (5th Cir. 1998).

We have never before imposed a duty on the administrator like the one imposed by the panel here. The two cases that have come closest to imposing such a duty are Southern Farm Bureau Life Ins. Co. v. Moore, 993 F.2d 98, 104 (5th Cir. 1993), and Salley v. E.I. DuPont de Nemours Co., 966 F.2d 1011, 1015 (5th Cir. 1992). Neither of these cases, however, actually imposed such a duty.Moore simply stated that the administrator had conducted a reasonable investigation and, although Salley notes that the administrator did not conduct a reasonable investigation, Salley rested its decision on the sufficiency of the administrative record to support the denial.

Having considered the relative merit of placing such a burden on the administrator, we reject this rule and stand by our precedent: We will continue to apply a sliding scale standard to the review of administrator's decisions involving a conflict of interest. If we placed a duty on conflicted administrators to reasonably investigate, we would be adopting the presumptively void standard of the Eleventh, Eighth, and Third Circuits. In effect, we would shift the burden to the administrator to prove that it reasonably investigated the claim. A rule that permitted such a result would be at odds with the Supreme Court's instruction in Bruch to review such determinations under an abuse of discretion standard — a standard that demands some deference be given to the administrator's decision. Such a rule would also violate basic principles of judicial economy. There is no justifiable basis for placing the burden solely on the administrator to generate evidence relevant to deciding the claim, which may or may not be available to it, or which may be more readily available to the claimant. If the claimant has relevant information in his control, it is not only inappropriate but inefficient to require the administrator to obtain that information in the absence of the claimant's active cooperation.

Instead, we focus on whether the record adequately supports the administrator's decision. In many cases, this approach will reach the same result as one that focuses on whether the administrator has reasonably investigated the claim. The advantage to focusing on the adequacy of the record, however, is that it (1) prohibits the district court from engaging in additional fact-finding and (2) encourages both parties properly to assemble the evidence that best supports their case at the administrator's level. For instance, in this case, the administrator's decision does not seem to be adequately supported by the record. On the other hand, the additional information — what Mrs. Vega's personal physician meant by his notation — could have been more easily obtained by the Vegas. When National Life did call Dr. Galvan's office, it did not actually get through to the doctor. Although National Life should have done so before denying the claim, this controversy could have been resolved if the Vegas, who were represented by counsel, had presented this information to National Life when their claim was denied. Here, it is apparent that the Vegas' attorney dismissed the administrative process as a nuisance and placed all their eggs in the litigation basket.

We hold today that, when confronted with a denial of benefits by a conflicted administrator, the district court may not impose a duty to reasonably investigate on the administrator. Under our own precedent and the Supreme Court's ruling in Bruch, we must give deference to the administrator's decision. That the administrator decides a claim when conflicted, however, is a relevant factor. In a situation where the administrator is conflicted, we will give less deference to the administrator's decision. In such cases, we are less likely to make forgiving inferences when confronted with a record that arguably does not support the administrator's decision. Although the administrator has no duty to contemplate arguments that could be made by the claimant, we do expect the administrator's decision to be based on evidence, even if disputable, that clearly supports the basis for its denial.

C

We turn next to the panel's use of the doctors' affidavits in reaching its decision. A long line of Fifth Circuit cases stands for the proposition that, when assessing factual questions, the district court is constrained to the evidence before the plan administrator. Meditrust Financial Services Corp. v. Sterling Chemicals, Inc., 168 F.3d 211, 215 (5th Cir. 1999); Schadler v. Anthem Life Insurance Company, 147 F.3d 388, 394-95 (5th Cir. 1998); Thibodeaux v. Continental Casualty Insurance, 138 F.3d 593, 595 (5th Cir. 1998); Barhan v. Ry-Ron Inc., 121 F.3d 198 (5th Cir. 1997); Bellaire General Hosp. v. Blue Cross Blue Shield of Michigan, 97 F.3d 822, 828-29 (5th Cir. 1996); Sweatman v. Commercial Union Insurance Co., 39 F.3d 594, 597-98 (1994); Duhon v. Texaco Inc., 15 F.3d 1302, 1306-07 (5th Cir. 1994); Southern Farm Bureau Life Ins. Co. v. Moore, 993 F.2d 98, 101-02 (5th Cir. 1993); Wildbur v. ARCO Chem. Co., 974 F.2d 631, 639 (5th Cir. 1992).

Our case law also makes clear that the plan administrator has the obligation to identify the evidence in the administrative record and that the claimant may then contest whether that record is complete. See, e.g., Barhan, 121 F.3d at 201-02. Once the administrative record has been determined, the district court may not stray from it except for certain limited exceptions. To date, those exceptions have been related to either interpreting the plan or explaining medical terms and procedures relating to the claim. Thus, evidence related to how an administrator has interpreted terms of the plan in other instances is admissible. See Wildbur v. ARCO Chemical Co., 974 F.2d 631, 639 n. 15 (5th Cir. 1992) (compiling cases). Likewise, evidence, including expert opinion, that assists the district court in understanding the medical terminology or practice related to a claim would be equally admissible. However, the district court is precluded from receiving evidence to resolve disputed material facts — i.e., a fact the administrator relied on to resolve the merits of the claim itself.

In this case, the record amounted to a number of exhibits attached to Pan-American's motion for summary judgment. The exhibits contained the relevant plan documents, Mrs. Vega's medical record, and the phone logs documenting Pan-American's contact with Mrs. Vega's doctors. The dispute here was essentially a factual one that would resolve the merits of the claim: Did Mrs. Vega receive notice that she would need posterior repair surgery prior to applying for membership in the plan? The testimony that the Vegas sought to introduce is evidence related to this factual dispute, which easily could have been presented to the administrator by the Vegas' counsel. The district court therefore correctly held that it could not admit new evidence for the purpose of resolving this dispute on the merits of the claim.

Our motivating concern here is that our procedural rules encourage the parties to resolve their dispute at the administrator's level. If a claimant believes that the district court is a better forum to present his evidence and we permit the claimant to do so, the administrator's review of claims will be circumvented. This result is plainly contrary to Bruch, which requires us to apply an abuse of discretion standard of review. Although we recognize that there is a concern that a self-interested administrator can manipulate this process unfairly (e.g., by permitting the administrator to exclude from the record information that would weigh in favor of granting the claim), we think that this concern is largely unwarranted in the light of adequate safeguards that can be put in place.

Before filing suit, the claimant's lawyer can add additional evidence to the administrative record simply by submitting it to the administrator in a manner that gives the administrator a fair opportunity to consider it. In Moore, we said that "we may consider only the evidence that was available to the plan administrator in evaluating whether he abused his discretion in making the factual determination." Moore, 993 F.2d at 102. If the claimant submits additional information to the administrator, however, and requests the administrator to reconsider his decision, that additional information should be treated as part of the administrative record. See, e.g., Wildbur, 974 F.2d at 634-35. Thus, we have not in the past, nor do we now, set a particularly high bar to a party's seeking to introduce evidence into the administrative record.

We hold today that the administrative record consists of relevant information made available to the administrator prior to the complainant's filing of a lawsuit and in a manner that gives the administrator a fair opportunity to consider it. Thus, if the information in the doctors' affidavits had been presented to National Life before filing this lawsuit in time for their fair consideration, they could be treated as part of the record. Furthermore, in restricting the district court's review to evidence in the record, we are merely encouraging attorneys for claimants to make a good faith effort to resolve the claim with the administrator before filing suit in district court; we are not establishing a rule that will adversely affect the rights of claimants.

In the light of our precedent and the abuse of discretion standard set forth in Bruch, we will not permit the district court or our own panels to consider evidence introduced to resolve factual disputes with respect to the merits of the claim when that evidence was not in the administrative record. We therefore stand by our precedent and reaffirm that, with respect to material factual determinations — those that resolve factual controversies related to the merits of the claim — the court may not consider evidence that is not part of the administrative record.

IV

We turn finally to the merits of the district court's summary judgment ruling. Although we find that the district court did not err in refusing to consider the additional testimony of Dr. Bueso and Dr. Galvan, we cannot agree with the district court that National Life did not abuse its discretion in denying the claim. In the case at hand, the employer has contracted with both National Life and Pan-American. The record does not adequately address the relationship between these two companies. It is clear that Pan-American is a subsidiary of National Life, and we therefore must regard Pan-American as owned and controlled by National Life. What is not clear from the record is whether National Life exercises control over the day-to-day decisions of Pan-American.

Although our cases have addressed conflicts of interest in evaluating whether there has been an abuse of discretion, none of those cases involved the arrangement presented in this case in which the administrator is a separate but wholly-owned subsidiary of the insurer. Instead, in these previous cases, the insurer and the administrator have operated within the same entity. In this case, given the ownership and control of Pan-American by National Life, we must regard their relationship as something more than purely contractual and therefore conclude that Pan-American's decision was, to some degree, self-interested. Although the Vegas have demonstrated the minimal basis for a conflict, they have presented no evidence with respect to the degree of the conflict. On our sliding scale, therefore, we conclude that it is appropriate to review the administrator's decision with only a modicum less deference than we otherwise would.

A review of the evidence available to National Life at the time it denied the claim illustrates that its decision was not reasonable. National Life concluded that the Vegas made a material misrepresentation in response to the question, "Have you or your dependents had any consultation, advice, tests, treatment or medication for any medical condition(s) during the past 6 months?" Pan-American argues that the notation in Mrs. Vega's medical record clearly indicates that she received advice or consultation about a medical condition. To be material, however, the advice or consultation must be related to a medical condition that Mrs. Vega had at the time. In this case, there is simply no competent evidence that, at the time the notation was made in her medical record, Mrs. Vega suffered from a condition that required posterior repair surgery.

Shortly after enrolling in the plan, Mrs. Vega underwent surgery for posterior repair. The occurrence of the operation shortly after enrollment and the handwritten note by Dr. Galvan in Mrs. Vega's medical records certainly create a doubt regarding whether the procedure had been recommended to her prior to her enrollment in the plan.

The explanation of this notation provided by Dr. Galvan's assistant does not necessarily dispel this concern. Neither, however, does it provide evidence to support a conclusion that Mrs. Vega suffered from a medical condition for which she required posterior repair surgery; the notation is simply ambiguous. The evidence makes clear that Dr. Galvan made the notation during a telephone conversation with Mrs. Vega. If Mrs. Vega had called with questions about an actual ailment that required surgery, one would expect Dr. Galvan to set up an appointment with her, which he did not do. It is therefore far from a foregone conclusion that Dr. Galvan's notation was related to a medical condition that Mrs. Vega was experiencing at that time. This is the only information available in the record that supports the denial of the claim.

Plainly put, we will not countenance a denial of a claim solely because an administrator suspects something may be awry. Although we owe deference to an administrator's reasoned decision, we owe no deference to the administrator's unsupported suspicions. Without some concrete evidence in the administrative record that supports the denial of the claim, we must find the administrator abused its discretion.

If an administrator has made a decision denying benefits when the record does not support such a denial, the court may, upon finding an abuse of discretion on the part of the administrator, award the amount due on the claim and attorneys' fees. See, e.g.,Salley, 966 F.2d at 1014. We find such an abuse of discretion here, and we will remand to the district court for a determination of damages and reasonable attorney's fees and for entry of judgment.

V

In this case, we were first confronted with a jurisdictional issue — whether Mr. Vega was an employee and therefore a participant of the plan for purposes of ERISA. We hold today that, under Texas law, a sole shareholder of a corporation who is also an employee of that corporation is an employee for purposes of determining whether he is a participant of an ERISA plan.

This case also involves a complex issue with respect to how we deal procedurally with ERISA claims. Given the Supreme Court's language in Bruch, we must review this sort of claim under an abuse of discretion standard. Recognizing that a rule that unduly permits litigation in the district court may result in claimants' being less than forthcoming in the initial claim procedure, we reject the panel's formulation of an administrator's duty to conduct a good faith, reasonable investigation. Instead, we reaffirm that decisions like this one will be reviewed under an abuse of discretion standard — i.e., we will give deference to the administrator's decision. The amount of deference we accord to the administrator will decrease the more the administrator labors under an apparent conflict of interest. We nevertheless always give some deference to the administrator's decision. Finally, because we are bound by an abuse of discretion standard, we will not permit additional evidence to be admitted with respect to materially factual issues.

In this case, even applying a standard under which we accord deference to the plan administrator, we cannot affirm the district court's summary judgment ruling in its favor. The administrative record contained no evidence that would support denying the claim. Although the record contains innuendoes and hints that Mrs. Vega may have made a material misrepresentation on her enrollment form, there is no concrete evidence to support this finding. For the foregoing reasons, the ruling of the district court is REVERSED. We RENDER on the question of liability and REMAND to the district court for a determination as to the amount of damages and attorney's fees.

REVERSED, RENDERED and REMANDED for entry of judgment for the plaintiffs and award of attorney fees.