Ellis v. Liberty Life Assurance Co

958 F.3d 1271

Opinion

FILED

United States Court of Appeals PUBLISH Tenth Circuit

UNITED STATES COURT OF APPEALS May 13, 2020

Christopher M. Wolpert

FOR THE TENTH CIRCUIT Clerk of Court

_________________________________

MICHAEL D. ELLIS, Plaintiff - Appellee, v. No. 19-1074

LIBERTY LIFE ASSURANCE COMPANY OF BOSTON, a New Hampshire corporation,

Defendant - Appellant.

_________________________________

Appeal from the United States District Court for the District of Colorado (D.C. No. 1:15-CV-00090-LTB-KMT)

_________________________________

Byrne J. Decker, Ogletree, Deakins, Nash, Smoak & Stewart, P.C., Portland, Maine (Kristina N. Holmstrom (Ogletree, Deakins, Nash, Smoak & Stewart, P.C., Phoenix, AZ on the briefs) for Defendant-Appellant.

Shawn McDermott, McDermott Law, LLC, Denver, Colorado (Timothy Garvey, McDermott Law, LLC, Denver, Colorado on the briefs) for Plaintiff-Appellee.

_________________________________

Before HARTZ and EID, Circuit Judges. *

* The late Honorable Monroe G. McKay, United States Senior Circuit Judge, heard oral argument and participated in the panel’s conference of this appeal, but passed away before its final resolution. The practice of this court permits the remaining two panel judges, if in agreement, to act as a quorum in resolving the appeal. See United States v. Wiles, 106 F.3d 1516, 1516, n* (10th Cir. 1997); 28 U.S.C. § 46(d).

_________________________________

HARTZ, Circuit Judge.

_________________________________

In 2014, Liberty Life Assurance Company of Boston rejected the claim for long-

term disability benefits by Michael Ellis. As part of its employee-benefit plan, Comcast Corporation, for whom Ellis worked in Colorado from 1994 until 2012, had obtained from Liberty in 2005 a Group Disability Income Policy (the Policy). Ellis sought review of Liberty’s denial of benefits in the United States District Court for the District of Colorado under the Employee Retirement Income Security Act of 1974 (ERISA), 29 U.S.C. §§ 1001 et seq. The district court, reviewing the denial de novo, ruled that Liberty’s denial was not supported by a preponderance of the evidence. Liberty appeals. It contends that the court should have reviewed its decision under an abuse-of-discretion standard but that it should prevail even under a de novo standard. Ellis defends the district court’s choice of a de novo standard but argues he should prevail under either standard of review.

The central issue on appeal is what standard of review the district court should have applied. A plan administrator’s denial of benefits is ordinarily reviewed by the court de novo; but if the policy gives the administrator discretion to interpret the plan and award benefits, judicial review is for abuse of discretion. See Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 115 (1989). The Policy provided that it was governed by the law of Pennsylvania, which is where Comcast is incorporated and has its principal place of business. Among its terms was one that gave Liberty discretion in resolving claims for

3

benefits. A Colorado statute enacted in 2008, however, forbids such grants of discretion in insurance policies. The parties dispute both whether the statute applies to the Policy under Colorado law and whether Colorado law governs. We hold that in this dispute the law of Pennsylvania, rather than that of Colorado, is controlling. The uniformity and administrative-efficiency objectives of ERISA counsel us to adhere to the Policy’s choice of law. Liberty’s denial of benefits is therefore properly reviewed for abuse of discretion. Under that standard the denial must be upheld. Exercising jurisdiction under 28 U.S.C. § 1291, we reverse the decision of the district court.

I. BACKGROUND A. The Policy

Under the Policy, employees of Comcast are eligible for long-term disability benefits upon providing proof of disability due to injury or sickness and the expiration of an elimination period of at least six months, subject to proof of continuing disability and the need for regular attendance of a physician. 1 As relevant to the dispute before us, disability or disabled means that the employee “is unable to perform, with reasonable

1 The Policy language states, in relevant part:

When Liberty receives Proof that a Covered Person is Disabled due to Injury or Sickness and requires the Regular Attendance of a physician, Liberty will pay the Covered Person a Monthly Benefit after the end of the Elimination Period, subject to any other provisions of this policy. The benefit will be paid for the period of Disability if the Covered Person gives to Liberty Proof of continued:

1. Disability;

2. Regular Attendance of a Physician; and 3. Appropriate Available Treatment.

Aplt. App., Vol. II at 309.

4

continuity, the Material and Substantial Duties of Any Occupation.” Aplt. App., Vol. II at 296.

B. Ellis’s Medical History On February 1, 2012, while undergoing treatment for pneumonia, Ellis experienced severe chest pain as a result of a pulmonary embolism (blood clot in the lungs). He was administered nitroglycerin, but soon afterwards he had an abnormally slow heartbeat, followed by an approximately 24-second heart stoppage. He briefly returned to work after this incident, but his last day of employment with Comcast was February 29, 2012.

Ellis submitted a claim for short-term disability benefits, which Liberty approved in March 2012. He reported “poor concentration, dizziness, slowing of physical and mental skills” and was referred to a neurologist in June 2012. Aplt. App., Vol. I at 268. The neurologist who began treating Ellis, Dr. Alan Zacharias, recommended physical and cognitive therapy but also noted that Ellis had an “[u]nremarkable brain MRI,” had no evidence of a primary neuromuscular disease, and was alert and attentive. Id. at 269. Based on this report and documentation from two other providers, Liberty terminated short-term benefits in July 2012.

In October 2012, Ellis’s lawyer sent a letter to Liberty asking it to reinstate benefits without a formal appeal. Part of this submission was a neuropsychological evaluation by Dr. Dennis Helffenstein, whom the lawyer had asked to evaluate Ellis. He opined that Ellis’s testing “identified significant cognitive deficits suggesting bilateral frontal and bilateral temporal involvement. The pattern is consistent with cerebral

5

hypoxia[2]. There is absolutely no way Michael could do his job at this time from a cognitive standpoint.” Aplee. Supp. App., Vol. II at 578. Liberty reinstated short-term disability benefits through the maximum duration and advanced the claim for long-term- disability consideration.

To assess Ellis’s eligibility for long-term benefits, Liberty’s claim consultant asked Dr. John Crouch and Dr. Gilbert Wager (Liberty’s consulting neuropsychologist and internal-medicine specialist, respectively) to review Ellis’s records. The reports from both doctors expressed doubt that a 24-second heart stoppage could cause cerebral hypoxia or neurological injury. Dr. Wager explained that “[t]his scenario is unlikely, as permanent neurological injury is not a feature of an episode of cardiogenic syncope. In general, it takes about 4 minutes or longer of cerebral anoxia to cause neuronal cell death and permanent neurological damage upon loss of spontaneous circulation.” Aplt. App., Vol. I at 246.

Dr. Crouch requested Dr. Helffenstein’s raw data to assess the validity and reliability of Ellis’s claimed cognitive and psychiatric deficits. After receiving the raw data, Dr. Crouch stated in an addendum to his report that “multiple measures of response bias were administered and yield[ed] Normal findings, suggesting that [Ellis’s] impairments [were] valid/reliable.” Id. at 187. Liberty also placed Ellis under

2 According to the National Institute of Neurological Disorders and Stroke, “[c]erebral hypoxia refers to a condition in which there is a decrease of oxygen supply to the brain even though there is adequate blood flow.” National Institute of Health, Cerebral Hypoxia Information Page, https://www.ninds.nih.gov/disorders/all-disorders/cerebralhypoxia -information-page. Possible causes include “cardiac arrest.” Id.

6

surveillance in December 2012; the only video captured of Ellis revealed him “walking in a slow pace while utilizing a cane.” Aplee. Supp. App., Vol. II at 482. Liberty approved long-term benefits in April 2013 but noted that the cause of Ellis’s cognitive impairments was still unclear.

In May 2013 Liberty requested updated information from Dr. Dan Hadley, Ellis’s primary-care physician, and Dr. Zacharias, his neurologist. Dr. Hadley completed a restrictions form stating that Ellis could not work in a situation requiring more than 10-20 minutes of minimal concentration. Dr. Zacharias did not specify a work-related restriction and instead signed a restrictions form directing Liberty to “see neuropsych testing that supports his impairment.” Aplt. App., Vol. I at 239.

Liberty completed a vocational report in July 2013 that identified several alternative occupations fitting Ellis’s training, education, experience, and physical capacities. The case manager who completed the report indicated that she was asked to “presume[] sedentary work capacity, and not to include any cognitive and/or mental restrictions and limitations.” Aplee. Supp. App., Vol. II at 455. Liberty also had a three- day surveillance conducted in August 2013, but no clear video of Ellis was obtained.

When Liberty asked Dr. Crouch for an updated clinical review of Ellis’s records, he reported in September 2013 that “based on available information, it is unlikely that the claimant could perform the job duties of alternate occupations comparable to his prior job.” Aplt. App., Vol. I at 189. But he said that an independent neuropsychological reevaluation was warranted if one had not been recently performed. Dr. Bob Gant, a neuropsychologist, was retained by an outside vendor at Liberty’s request and evaluated

7

Ellis in October 2013. He determined that Ellis’s neuropsychological test results were invalid because of “[c]lear evidence of symptom exaggeration and suboptimal effort.” Id. at 202. He said:

Mr. Ellis reported an unusual and elevated degree of neurological complaints which are likely to be vague and illogical. This was confirmed by other tests utilized during this examination which indicated that the degree of neurologic impairment reported by Mr. Ellis was highly atypical and illogical. Such a presentation includes symptoms that are illogical or inconsistent with symptoms of a bona fide neurologic disorder or they occur very rarely in neurologically impaired patients.

Id. at 203–04. Dr. Gant questioned whether Ellis even had cognitive impairment:

[W]ithin reasonable medical probability [Ellis] has not suffered cognitive impairment related to the asystole event which lasted 24 seconds on February 1, 2012. In fact, I am not certain that the patient suffers from cognitive impairment. It is likely that elements of secondary gain and/or impairment related to somatic exaggeration is responsible for [his]

presentation.

Id. at 196.

In November 2013, Dr. Crouch reviewed Dr. Gant’s report and stated that the results from Dr. Gant’s evaluation “are insufficient to support the presence of valid/reliable” cognitive impairment. Id. at 194. Dr. Crouch also agreed that it was “medically impossible for a 24 second asystole event to cause cerebral hypoxia.” Id. at 194. Liberty terminated Ellis’s disability benefits in December 2013.

Ellis appealed the denial in June 2014. He included as additional evidence in support of his appeal a March 2014 letter from his speech therapist, letters from the Social Security Administration from December 2013 declaring him eligible for disability benefits, and imaging from a Single Photon Emission Completed Tomography (SPECT)

8

scan together with an assessment report interpreting the images. The SPECT scan, which shows blood flow and oxygen perfusion to the brain, was interpreted by Dr. S. Gregory Hipskind, a nuclear neurologist to whom Ellis had been referred by Dr. Helffenstein. He read the scan as abnormal—consistent with “a diffuse, toxic/hypoxic encephalopathic process.” Id. at 220. Dr. Helffenstein had also conducted a second evaluation in May 2014 and his written report, completed in July 2014, later supplemented Ellis’s appeal. The report said that Ellis had demonstrated notable improvement in his results but had “reached maximum medical improvement from a neuropsychological standpoint” and was “totally and permanently disabled from competitive employment.” Id. at 144–45.

In September 2014, Liberty had Dr. Timothy Belliveau, another of its consulting neuropsychologists, review Ellis’s medical records and neuropsychological evaluations. Dr. Belliveau opined that the test data from Dr. Helffenstein’s 2012 exam and Dr. Gant’s 2013 exam probably indicated symptom over-reporting. Dr. Belliveau concluded that “[c]onsidered as a whole, and in the context of the claimant’s documented medical history, the neuropsychological test data provide insufficient support for the presence of cognitive or psychological impairment due to a presumed brain injury in February 2012.” Id. at 109. In light of Dr. Belliveau’s review, Liberty upheld its denial.

C. ERISA

The Policy is part of an employee-benefits plan governed by ERISA. Such plans may provide a variety of healthcare, retirement, life-insurance, disability, and other benefits. Congress enacted ERISA both to “ensure that employees would receive the benefits they had earned” and to encourage employers to offer these plans by “creat[ing]

9

a system that is not so complex that administrative costs, or litigation expenses, unduly discourage employers from offering ERISA plans in the first place,” Conkright v. Frommert, 559 U.S. 506, 516–17 (2010) (brackets and internal quotation marks omitted), and by providing tax incentives, see Ronald J. Cooke, 1 ERISA Practice and Procedure § 1:3 (2d ed. 2019).

ERISA requires every benefit plan to be fully described in written “plan documents” that govern the management of the plan by plan administrators. See 29 U.S.C. § 1102(a)(1). The documents must “specify the basis on which payments are made to and from the plan,” id. § 1102(b)(4), and the plan administrator must act “in accordance with the documents and instruments governing the plan,” id. § 1104(a)(1)(D); see Egelhoff v. Egelhoff ex rel. Breiner, 532 U.S. 141, 147 (2001). From those documents, employees can “learn their rights and obligations under the plan at any time.” Curtiss-Wright Corp. v. Schoonejongen, 514 U.S. 73, 83 (1995).

Plan administrators are subject to federal standards imposing fiduciary duties. See Miller v. Monumental Life Ins. Co., 502 F.3d 1245, 1249 (10th Cir. 2007). To avoid plan administrators having to “master the relevant laws of 50 states and to contend with litigation [that] would undermine the congressional goal of minimiz[ing] the administrative and financial burden[s] on plan administrators,” Egelhoff, 532 U.S. at 149–50 (internal quotation marks omitted), ERISA contains a broad preemption provision stating that it “shall supersede any and all State laws insofar as they may now or hereafter relate to any employee benefit plan” covered by the statutory scheme, 29 U.S.C. § 1144(a). See Aetna Health v. Davila, 542 U.S. 200, 208 (2004) (“The

10

purpose of ERISA is to provide a uniform regulatory regime over employee benefit plans”; and its preemption provisions “are intended to ensure that employee benefits regulation would be exclusively a federal concern.” (internal quotation marks omitted)); Miller, 502 F.3d at 1249. ERISA is one of the rare federal statutes recognized as “preempting the field.” See Nelson v. Great Lakes Educ. Loan Servs., Inc., 928 F.3d 639, 652 (7th Cir. 2019). Nevertheless, state laws that regulate insurance, banking, or securities are generally exempted from ERISA preemption. See 29 U.S.C. § 1144(b)(2)(A); Ky. Ass’n of Health Plans v. Miller, 538 U.S. 329, 341–42 (2003). But see Aetna Health, 542 U.S. at 217 (“[E]ven a state law that can be arguably characterized as ‘regulating insurance’ will be pre-empted if it provides a separate vehicle to assert a claim for benefits outside of, or in addition to, ERISA’s remedial scheme.”).

An employee covered by an ERISA-governed benefit plan who believes the plan administrator wrongfully denied benefits can bring suit in state or federal court. See 29 U.S.C. §§ 1132(a)(1)(B), (e). The plan administrator’s decision is reviewed by the court de novo unless the terms of the benefit plan give the administrator discretion to interpret the plan and award benefits. See Firestone Tire, 489 U.S. at 115. The Supreme Court has observed that granting plan administrators deference in interpreting plans promotes efficiency by encouraging resolution of disputes without litigation and promotes predictability “as an employer can rely on the expertise of the plan administrator rather than worry about unexpected and inaccurate plan interpretations that might result from de novo judicial review.” Conkright, 559 U.S. at 517.

11

II. ANALYSIS A. Choice of Law

The principal dispute in this appeal is whether Liberty, the administrator of Ellis’s ERISA plan, has discretion in determining whether to award or deny benefits. The Policy states: “Liberty shall possess the authority, in its sole discretion, to construe the terms of this policy and to determine benefit eligibility hereunder. Liberty’s decisions regarding construction of the terms of this policy and benefit eligibility shall be conclusive and binding.” Aplt. App., Vol. II at 329. Ellis does not dispute that the Policy grants Liberty the requisite discretion. But Colorado law provides:

An insurance policy, insurance contract, or plan that is issued in this state that offers health or disability benefits shall not contain a provision purporting to reserve discretion to the insurer, plan administrator, or claim administrator to interpret the terms of the policy, contract, or plan or to determine eligibility for benefits.

C.R.S. § 10-3-1116(2). On the other hand, the Policy has a choice-of-law provision stating that it is governed by Pennsylvania law, and Pennsylvania has no statute limiting discretion. Therefore, we must decide whether the Colorado statute applies to this dispute. We review the choice-of-law issue de novo. See Boone v. MVM, Inc., 572 F.3d 809, 811 (10th Cir. 2009). 3

3 Ellis claims that Liberty did not argue the choice-of-law issue below and therefore essentially conceded that Colorado law applies. But Ellis was untimely in not raising this concern until oral argument. See Lenox MacLaren Surgical Corp. v. Medtronic, Inc., 762 F.3d 1114, 1122–23 n.7 (10th Cir. 2014). And in any event Liberty did preserve the issue by arguing in its briefing to the district court that “the Policy expressly provides that [it] is governed by the laws of the state of Pennsylvania . . . . Plaintiff’s Opening Brief does not suggest, nor can it, that Pennsylvania has a statute prohibiting ‘discretionary clauses.’” Aplt. App., Vol. II at 444–45. The district court clearly thought that Liberty

12

The choice-of-law question could be avoided if ERISA preempts the Colorado statute. Liberty raised preemption in district court. But several circuits have held that similar statutes are saved from ERISA preemption because they come within the exception to preemption for laws that regulate insurance. See Orzechowski v. Boeing Co. Non-Union Long-Term Disability Plan, Plan No. 625, 856 F.3d 686, 692–95 (9th Cir. 2017); Fontaine v. Metro. Life Ins. Co., 800 F.3d 883, 886–89 (7th Cir. 2015); Am. Council of Life Insurers v. Ross, 558 F.3d 600, 604–07 (6th Cir. 2009). Perhaps for that reason, Liberty has not pursued the issue on appeal. In any event, there is no need to resolve that preemption issue here because our analysis leads to the conclusion that the Colorado statute does not apply for other reasons. 4 Our analysis will proceed as follows: (1) Because Ellis’s claim for benefits is a federal cause of action, federal law governs the elements of the claim. (2) But when federal law is silent on the specific question at issue (here, whether the Policy’s grant of discretion to Liberty is enforceable), the federal court may incorporate state law instead of constructing a uniform federal rule. In our view, the enforceability question should be answered by state law; that is, federal law should incorporate a state rule of decision to resolve the question. (3) When federal law incorporates a state rule of decision, the choice of which state’s law to incorporate is a matter of federal law. (4) As a matter of

had argued that Pennsylvania law governed, because it referred to Liberty’s argument in the paragraph of its opinion devoted to the choice-of-law issue. 4 Similarly, because we conclude that we must follow Pennsylvania law, we need not address whether the Colorado statute applies to the Policy—that is, whether the Policy was issued after enactment of the statute and whether the Policy was issued in Colorado.

13

federal law, to effectuate ERISA’s goals of uniformity and ease of administration, the law of the State selected by a choice-of-law provision in the plan documents should ordinarily provide the rule of decision for claims brought under the plan.

First, the Supreme Court has made clear that claims to enforce rights under an ERISA plan, even if styled as claims under state law, are federal claims. In Metropolitan Life Insurance Co. v. Taylor, the Court declared that ERISA so “completely pre- empt[ed]” claims within the scope of § 1132(a) that “any civil complaint raising this select group of claims is necessarily federal in character.” 481 U.S. 58, 63–64 (1987). Such actions “are to be regarded as arising under the laws of the United States . . . .” Id. at 65 (internal quotation marks omitted). The Court concluded that the plaintiff’s suit seeking only state-law contract and tort remedies for failure of his employer and the plan administrator to provide benefits in accordance with his ERISA plan was “necessarily federal in character by virtue of the clearly manifested intent of Congress.” Id. at 67; see Pilot Life Ins. Co. v. Dedeaux, 481 U.S. 41, 56 (1987) (stating, in case where plan participant brought state common-law claim for tortious breach of contract, that “Congress’ specific reference to § 301 of the [Labor Management Relations Act] to describe the civil enforcement scheme of ERISA makes clear its intention that all suits brought by beneficiaries or participants asserting improper processing of claims under ERISA-regulated plans be treated as federal questions governed by § [1132](a).”). The federal character of the ERISA suit is preserved even when (as will be further discussed below) a state-law rule of decision is incorporated for resolution of the claim. Thus, in Unum Life Insurance. Co. v. Ward, 526 U.S. 358, 376–77 (1999), the Court, holding that

14

the California notice-prejudice rule was not preempted and should be applied in resolving a claim under § 1132(a), said, “The notice-prejudice rule supplied the relevant rule of decision for th[e] § [1132](a) suit.” In sum, federal law governs the resolution of Ellis’s claim.

To say that federal law governs, however, is not to say that state law is irrelevant.

In resolving a federal claim, questions may arise that cannot be answered by statutory interpretation. The court then must either adopt a federal common-law rule of decision or incorporate state law. The Supreme Court in Kamen v. Kemper Financial Services, 500 U.S. 90, 92 (1991), addressed this matter when it had to decide in a shareholder- derivative action under the Investment Company Act (ICA) whether to require the representative shareholder “to make a demand on the board of directors even when such a demand would be excused as futile under state law.” First, the Court stated that federal law clearly governs “the contours of the demand requirement in a derivative action founded on the ICA,” explaining that “[b]ecause the ICA is a federal statute, any common law rule necessary to effectuate a private cause of action under that statute is necessarily federal in character.” Id. at 97. That did not mean, however, “that the content of such a rule must be wholly the product of a federal court’s own devising.” Id. at 98. On the contrary, absent a special reason the federal rule of decision should be state law that is incorporated into the federal remedial scheme:

Our cases indicate that a court should endeavor to fill the interstices of federal remedial schemes with uniform federal rules only when the scheme in question evidences a distinct need for nationwide legal standards, or when express provisions in analogous statutory schemes embody congressional policy choices readily applicable to the matter at hand.

15

Otherwise, we have indicated that federal courts should incorporate state law as the federal rule of decision, unless application of the particular state law in question would frustrate specific objectives of the federal programs.

Id. at 98 (citations, brackets, and internal quotation marks omitted). The Court held that a court considering a derivative action under the ICA “must apply the demand futility exception as it is defined by the law of the State of incorporation.” Id. at 108–09. Similarly, in United States v. Kimbell Foods, Inc., the Court wrote:

Controversies directly affecting the operations of federal programs, although governed by federal law, do not inevitably require resort to uniform federal rules. Whether to adopt state law or to fashion a nationwide federal rule is a matter of judicial policy dependent upon a variety of considerations always relevant to the nature of the specific governmental interests and to the effects upon them of applying state law.

440 U.S. 715, 727–28 (1979) (citation and internal quotation marks omitted); see Davilla v. Enable Midstream Partners L.P., 913 F.3d 959, 965–66 (10th Cir. 2017) (incorporating, “as a matter of so-called ‘federal common law,’” Oklahoma law as rule of decision in federal-law-governed trespass action).

Two Supreme Court opinions will illustrate the process of determining whether the courts should adopt a uniform federal common-law rule or incorporate a state rule of decision. Quite recently Rodriguez v. FDIC, 140 S. Ct. 713 (2020), rejected a uniform federal rule (the Bob Richards rule, named for the case that originated it, In re Bob Richards Chrysler-Plymouth Corp., 473 F.2d 262 (9th Cir. 1973)), which had been adopted by several circuit courts for determining how the members of an affiliated group of corporations that filed a consolidated tax return are to share a federal tax refund after it is delivered to the group’s designated agent. See id. at 716. The affiliated group in

16

Rodriguez was just a bank and its corporate parent. See id. Serious problems with the bank required the Federal Deposit Insurance Corporation (FDIC) to take it over in receivership. See id. Soon afterwards, the parent entered bankruptcy. See id. In the parent’s bankruptcy proceedings the FDIC and the trustee for the parent’s bankruptcy estate both claimed a large federal-income-tax refund that had been issued to the affiliated group. See id. The circuit court applied the Bob Richards rule. But the Supreme Court said that a uniform rule was hardly “necessary to protect uniquely federal interests.” Id. at 718 (internal quotation marks omitted). It recognized that “[t]he federal government may have an interest in regulating how it receives taxes from corporate groups,” or “in regulating the delivery of any tax refund due a corporate group,” or it “may wish to ensure that others in the group have no recourse against federal coffers once it pays the group’s designated agent.” Id. “But what unique interests could the federal government have in determining how a consolidated corporate tax refund, once paid to a designated agent, is distributed among group members?” Id. at 717–18. The Court noted that “corporations are generally creatures of state law and state law is well- equipped to handle disputes involving corporate property rights,” and it added that the fact that the controversy arose in the context of “federal bankruptcy and a tax dispute doesn’t change much.” Id. at 718 (citation and internal quotation marks omitted). The Court held that the state rule of decision should govern, although it left for the circuit court to determine what that rule was. See id.; see also United States v. Turley, 878 F.3d 953, 956–57 (10th Cir. 2017) (in dispute arising from a lease between a private lessor and the United States Postal Service, we recognized that “obligations to and rights of the

17

United States under its contracts are governed exclusively by federal law,” but rather than constructing a uniform federal rule, we incorporated Oklahoma law because “lease contracts for the postal service do not inherently implicate clear and substantial interests of the National Government, which cannot be served consistently with respect for state interests.” (brackets and internal quotation marks omitted)).

In contrast, Boyle v. United Technologies Corp., 487 U.S. 500, 512 (1988), held that federal common law, not state law, applied in a diversity case against a federal contractor for an alleged design defect when the contractor’s design conformed to government specifications. The estate of a Marine pilot sought to hold a government contractor liable for defective design of a military helicopter’s escape hatch that caused the pilot’s death. See id. at 503. Although no federal statute precluded government contractors from being held liable for design defects, see id. at 504, the Court adopted a federal rule that displaced liability under state law in certain limited circumstances: “when (1) the United States approved reasonably precise specifications; (2) the equipment conformed to those specifications; and (3) the supplier warned the United States about the dangers in the use of the equipment that were known to the supplier but not to the United States.” Id. at 512. The Court observed that “imposition of liability on Government contractors will directly affect the terms of Government contracts: either the contractor will decline to manufacture the design specified by the Government, or it will raise its price. Either way, the interests of the United States will be directly affected.” Id. at 507. To impose design-defect liability under state law would be “precisely contrary to the duty imposed by the Government contract (the duty to manufacture and deliver

18

helicopters with the sort of escape-hatch mechanism shown by the specifications).” Id. at 509.

The issue before us, therefore, is whether any federal policy or interest demands the creation of a uniform federal rule either requiring or prohibiting enforcement of discretion-granting provisions in ERISA plans. If not, we should leave to state law whether to permit or allow such provisions. In particular, is there a federal interest in requiring that decisions by administrators be subject to de novo judicial review—that is, depriving administrators of discretionary power? Or is there a federal interest in always allowing plans to grant discretion to administrators? In our view, decisions of the Supreme Court have pretty much answered those two questions. On the one hand, the Court has set forth with considerable sympathy how granting discretion to administrators advances certain ERISA objectives. See Conkright, 559 U.S. at 517–21. It would be hard to read the discussion in Conkright and conclude that an ERISA plan’s grant of deference is inconsistent with federal policies and objectives. On the other hand, the Court has established de novo judicial review as the default standard for reviewing administrator decisions. See Firestone Tire, 489 U.S. at 110–15. That holding would seem inconsistent with a determination that ERISA policy forbids discretionary bans like Colorado’s. Thus, we can assume that permitting grants of discretion and forbidding such grants are both consistent with ERISA. Accordingly, there would be no need for a uniform federal-common-law rule favoring one approach over the other. Adopting a state-law rule of decision is appropriate.

19

The next question is which state’s law to use. The general rule is that federal choice-of-law principles are used in resolving federal causes of action. 5 See, e.g., Berger v. AXA Network LLC, 459 F.3d 804, 809–10 & n.7 (7th Cir. 2006) (citing cases); Gluck v. Unisys Corp., 960 F.2d 1168, 1179 & n.8 (3d Cir. 1992); 17A James Wm. Moore et al., Moore’s Federal Practice § 120.31[1][b][ii], at 120–73 (3d ed. 2011) (“[I]n federal question cases, federal courts look to federal choice of law principles.”). Thus, in Kamen, after determining that no uniform federal rule was required and that the courts should apply state law regarding the demand-futility exception for derivative actions under the ICA, the Court did not look to state choice-of-law doctrine before declaring that federal courts must apply the “demand futility exception as it is defined by the law of the State of incorporation.” 500 U.S. at 109 (emphasis added). Similarly, after deciding that state contract law should be incorporated to resolve a dispute over a lease for a federal post office, this court in Turley, again without canvassing state choice-of-law doctrine, held that Oklahoma law should govern because that was where the property was located. 878 F.3d at 957. (This is not to say that federal law cannot incorporate state choice-of-law doctrine in resolving a federal claim. In Richards v. United States, 369 U.S. 1 (1962), the Supreme Court interpreted the Federal Tort Claims Act to apply not

5 Ellis cites our decision in Loveridge v. Dreagoux, 678 F.2d 870, 877 (10th Cir. 1982), for the proposition that when selecting the applicable state law we must “follow the conflict of laws rules of the forum state where jurisdiction is based on a federal question.” Aplee. Br. at 18. But Loveridge is readily distinguishable. Although the jurisdiction of the federal court was based on a federal-law claim (under the Securities Exchange Act of 1934), the choice-of-law issue arose with respect to a pendent state-law claim—a breachof -contract claim under Utah law. See id. at 872.

20

only the substantive law of the State where the negligence occurred but also that State’s choice-of-law doctrine, see id. at 10–11; see also In re Gaston & Snow, 243 F.3d 599, 604–07 (2d Cir. 2001) (applying forum State choice-of-law rule for contract dispute in bankruptcy court).

In particular, in ERISA cases the federal circuits have applied federal choice-of-

law principles to determine whether to give effect to a policy’s choice-of-law provision. Other circuits have identified three possible approaches, two of which have been adopted.

The Ninth Circuit has said that the choice-of-law provision in an ERISA plan should be followed if “not unreasonable or fundamentally unfair.” Wang Labs v. Kagan, 990 F.2d 1126, 1128–29 (9th Cir. 1993). The dispute before the court concerned an employee whose ERISA plan required him to reimburse medical expenses paid by the plan for injuries he received in a vehicle accident after he obtained a tort recovery for the accident. See id. at 1127. The employee argued that the plan’s reimbursement claim was barred by the applicable statute of limitations. See id. The plan contained a choice-of- law provision selecting Massachusetts law. See id. at 1128. The employer and the plan administrator were headquartered there, and most employees affected by the plan lived there; but the employee resided in California at all relevant times and the accident occurred in California. See id. at 1127–28.

The court relied on the Supreme Court’s decision in Carnival Cruise Lines v.

Shute, 499 U.S. 585 (1991). See Wang, 990 F.2d at 1128–29. Carnival held that forum- selection clauses, even in contracts of adhesion (such as a cruise ticket), should be enforced if not unreasonable or fundamentally unfair. See 499 U.S. at 592–95. Wang

21

reasoned that choice-of-law clauses would be less burdensome to plan beneficiaries than forum-selection clauses because beneficiaries could still litigate ERISA disputes in their home state. See 990 F.2d at 1129. It ruled that the choice-of-law clause in the ERISA contract was not unreasonable or fundamentally unfair since the employer was headquartered in Massachusetts, most covered employees resided in the state, and “[n]o sensible person would hesitate to join a health plan because claims would be subject to the limitations period of the employer’s headquarters state.” Id. The Eighth and Eleventh Circuits have followed Wang’s unreasonable-or-fundamentally-unfair test for choice-of-law provisions in ERISA contracts. See Brake v. Hutchinson Tech. Inc., 774 F.3d 1193, 1197 (8th Cir. 2014) (declining to apply law of South Dakota (plaintiff’s home state and the forum state) disallowing discretion clause in health-insurance policies); Buce v. Allianz Life Ins. Co., 247 F.3d 1133, 1149 (11th Cir. 2001) (adhering to plan’s choice of Georgia law (plaintiff’s home state and forum state)).

The Sixth Circuit has adopted a different approach, applying the test set out in Section 187 of the Restatement (Second) of Conflict of Laws for when a contractual choice-of-law provision should be enforced. 6 See DaimlerChrysler v. Durden, 448 F.3d 918, 922 (6th Cir. 2006) (“In the absence of any established body of federal choice of law

6 Ellis contends that § 188 of the Restatement should control this case, and that under § 188’s multi-factor test Colorado law should apply. But § 188 is titled (and unsurprisingly concerns): “Law Governing in Absence of Effective Choice by the Parties.” Restatement (Second) of Conflict of Laws § 188 (1971) (emphasis added). Here, the Policy contains a clear choice-of-law provision, and Ellis makes no argument why this is not an “effective choice” within the meaning of the Restatement.

22

rules, we begin with the Restatement (Second) of Conflicts of Law.” (internal quotation marks omitted)). Section 187 provides, in relevant part:

(1) The law of the state chosen by the parties to govern their contractual rights and duties will be applied if the particular issue is one which the parties could have resolved by an explicit provision in their agreement directed to that issue.

(2) The law of the state chosen by the parties to govern their contractual rights and duties will be applied, even if the particular issue is one which the parties could not have resolved by an explicit provision in their agreement directed to that issue, unless either (a) the chosen state has no substantial relationship to the parties or the transaction and there is no other reasonable basis for the parties’

choice, or

(b) application of the law of the chosen state would be contrary to a fundamental policy of a state which has a materially greater interest than the chosen state in the determination of the particular issue and which, under the rule of § 188, would be the state of the applicable law in the absence of an effective choice of law by the parties.

Restatement (Second) of Conflicts of Laws, § 187 (1971). Durden involved two women who each claimed to be the “surviving spouse” of an employee covered under his employer’s ERISA-governed pension plan, which had a choice-of-law provision selecting Michigan law. 448 F.3d at 921. After the employee passed away, his surviving spouse was entitled to benefits from the pension plan, including life-insurance proceeds. Id. The court ultimately decided that Ohio law governed because it had the most significant relationship to both marriages. See id. at 923–27.

When faced with resolving whether an ERISA plan’s choice-of-law provision governed in determining if the employee’s misconduct forfeited his benefits, the Fifth Circuit identified three possible approaches. See Jimenez v. Sun Life Assur. Co., 486 F. App’x 398, 407–08 (5th Cir. 2012). It noted the Wang and Durden tests, and said that in

23

international-disputes cases it had presumptively enforced a contractual choice-of-law provision unless the party hoping to avoid enforcement clearly showed “that the clause [was] unreasonable under the circumstances.” Id. at 408 (internal quotation marks omitted). But it declined to choose a standard because it held that the employee challenging the administrator’s denial of benefits failed to satisfy his burden of overcoming the contractual choice-of-law provision under all three approaches. See id.

In our view, the above three circuit approaches, all of which sound primarily in reasonableness, are inadequate because they overlook the uniformity and efficiency objectives central to ERISA. Over several decades the Supreme Court has repeatedly recognized and emphasized that ERISA policy is best effectuated if a plan administrator is subject to only one legal regime.

The choice-of-law issue obviously is most likely to arise for interstate employers.

And it is precisely in plans for interstate employers that the need for a single legal regime is most pressing. As stated in Shaw v. Delta Air Lines, Inc., 463 U.S. 85, 105 (1983), “By establishing benefit plan regulation as exclusively a federal concern, Congress minimized the need for interstate employers to administer their plans differently in each State in which they have employees.” (citation and internal quotation marks omitted). The Court noted that the imposition of patchwork regulation by a variety of states could be particularly burdensome to “[a]n employer with employees in many States [who] might find that the most efficient way to provide benefits to those employees is through a single employee benefit plan.” Id. at 105 n.25. The consequences would be harmful to both employers and employees:

24

The employer might choose to offer a number of plans, each tailored to the laws of particular States; the inefficiency of such a system presumably would be paid for by lowering benefit levels. Alternatively, assuming that the state laws were not in conflict, the employer could comply with the laws of all States in a uniform plan. To offset the additional expenses, the employer presumably would reduce wages or eliminate those benefits not required by any State. Another means by which the employer could retain its uniform nationwide plan would be by eliminating classes of benefits that are subject to state requirements with which the employer is unwilling to comply.

Id.

Likewise, in Fort Halifax Packing Co. v. Coyne, 482 U.S. 1, 11 (1987), the Court observed that “[a] patchwork scheme of regulation would introduce considerable inefficiencies in benefit program operation, which might lead those employers with existing plans to reduce benefits, and those without such plans to refrain from adopting them.” To avoid such outcomes, ERISA preemption “ensures that the administrative practices of a benefit plan will be governed only by a single set of regulations.” Id.

More recently the Court has reiterated that “[o]ne of the principal goals of ERISA is to enable employers to establish a uniform administrative scheme, which provides a set of standard procedures to guide processing of claims and disbursement of benefits.” Egelhoff, 532 U.S. at 148 (internal quotation marks omitted). “Uniformity is impossible, however, if plans are subject to different legal obligations in different States.” Id. “Requiring ERISA administrators to master the relevant laws of 50 States and to contend with litigation would undermine the congressional goal of minimizing the administrative and financial burden[s] on plan administrators—burdens ultimately borne by the beneficiaries.” Id. at 149–50 (brackets and internal quotation marks omitted). Such

25

“tailoring of plans and employer conduct to the peculiarities of the law of each jurisdiction is exactly the burden ERISA seeks to eliminate.” Id. at 151 (internal quotation marks omitted); see H.R. Rep. No. 93-533, 1973 WL 12549, at 4650 (1973) (“Finally, it is evident that the operations of employee benefit plans are increasingly interstate. The uniformity of decision which the Act is designed to foster will help administrators, fiduciaries and participants to predict the legality of proposed actions without the necessity of reference to varying state laws.”).

And the Court in Conkright, 559 U.S. at 517, reiterated that “ERISA induces employers to offer benefits by assuring a predictable set of liabilities, under uniform standards of primary conduct and a uniform regime of ultimate remedial orders and awards when a violation has occurred.” (brackets and internal quotation marks omitted). Of particular importance, the Court noted a potential problem with patchwork state regulation that could be fatal to interstate plans: “[A] group of prominent actuaries tells us that it is impossible even to determine whether an ERISA plan is solvent (a duty imposed on actuaries by federal law, see 29 U.S.C. §§ 1023(a)(4), (d)) if the plan is interpreted to mean different things in different places.” Id. at 517–18.

These concerns explain not only the preemption of most state law regarding ERISA plans but also the need for uniform interpretation and enforcement of plan provisions in those areas where state law is not preempted. The Supreme Court’s decision in Kennedy v. Plan Administrator for DuPont Savings and Investment Plan, 555 U.S. 285, 300 (2009), is instructive. In Kennedy an employee who participated in his company’s ERISA plan signed a form designating his then-spouse to receive benefits

26

under the plan upon his death. Id. at 288–89. When the couple divorced several years later, the divorce decree stated that it divested the woman of all rights she may have had in any of the employee’s benefit plans. See id. at 289. But the employee did not execute any documents removing her as a beneficiary of the pension plan. See id. After first holding that the divorce-decree divestment provision was not void under ERISA’s anti- alienation provision, see id. at 297, the Court considered whether the plan administrator had to honor the decree or could instead pay the former spouse benefits under the plan’s terms.

The Court unanimously held that the “plan administrator did its statutory ERISA duty by paying the benefits to [the former spouse] in conformity with the plan documents.” Id. at 299–300. ERISA not only “requires ‘[e]very employee benefit plan [to] be established and maintained pursuant to a written instrument,’” id. at 300 (quoting 29 U.S.C. § 1102(a)(1)), but further obliges the plan administrators to act “‘in accordance with the documents and instruments governing the plan insofar as [they] are consistent with [ERISA],’” id. (quoting 29 U.S.C. § 1104(a)(1)(D)). The Court also pointed out that ERISA allows a beneficiary to file suit “‘to recover benefits due to him under the terms of his plan,’” further reinforcing the command to abide by plan terms. Id. (quoting 29 U.S.C. § 1132(a)(1)(B) (emphasis added)). A claim under ERISA “therefore stands or falls by the terms of the plan, § 1132(a)(1)(B), a straightforward rule of hewing to the directives of the plan documents that lets employers establish a uniform administrative scheme, with a set of standard procedures to guide processing of claims and disbursement of benefits.” Id. (brackets and internal quotation marks omitted). Given the statutory

27

goals of uniformity and predictability in the administration of ERISA plans, “the cost of less certain rules would be too plain.” Id. at 301.

Heimeshoff v. Hartford Life & Accident Insurance Co. adhered to the “plan-

documents rule” of Kennedy, recognizing that the “focus on the written terms of the plan is the linchpin of a system that is not so complex that administrative costs, or litigation expenses, unduly discourage employers from offering ERISA plans in the first place.” 571 U.S. 99, 108 (2013) (upholding plan provision that commenced three-year limitations period before ERISA cause of action accrues) (brackets and internal quotation marks omitted); see also U.S. Airways, Inc. v. McCutchen, 569 U.S. 88, 101 (2013) (“The plan, in short, is at the center of ERISA.”).

Our choice-of-law doctrine in the ERISA context must therefore account for the centrality of the plan in ERISA matters and the aims of uniformity and reduced administrative costs that are essential to ERISA’s purposes. See Durden, 448 F.3d at 928–29 (Merritt, J., dissenting) (“the overriding purpose and policy of uniformity behind the ERISA statute [and] behind the interpretation of ERISA benefits contracts” requires courts to enforce parties’ choice of law in ERISA plans); William Baude, Beyond DOMA: Choice of State Law in Federal Statutes, 64 Stan. L. Rev. 1371, 1420 (2012) (criticizing Sixth Circuit’s decision in Durden to follow Restatement instead of enforcing choice-of-law provision in plan documents: “In light of the plan-documents doctrine established by the Supreme Court, the better rule for ERISA cases is to follow a marital choice-of-law rule required by the plan documents.”).

28

These considerations apply with full force to the present context. To recognize the Policy’s grant of discretion to the administrator for plan participants in some states but not in others would create significant complications. Legislatures enact statutes forbidding discretionary provisions for the purpose of awarding more benefits to participants in insurance plans. But there are costs in doing so. In Conkright the Supreme Court pointed out that granting deference to the administrator promotes efficiency, predictability, and uniformity. See 559 U.S. at 517. The increase in costs from denying discretion can lead employers to reduce benefits or even to cancel plans or refrain from offering them altogether. Indeed, it is precisely because discretion-denying statutes “substantially affect the risk pooling arrangement between the insurer and the insured,” Ky. Ass’n of Health Plans, Inc. v. Miller, 538 U.S. at 342, that such statutes have been held to be laws regulating insurance that are exempted from ERISA preemption. See, e.g., Orzechowski, 856 F.3d at 694–95; Fontaine, 800 F.3d at 888–89; Standard Ins. Co. v. Morrison, 584 F.3d 837, 844–45 (9th Cir. 2009) (under a discretion- denying statute, “insureds may no longer agree to a discretionary clause in exchange for a more affordable premium”).

When the plan is a single-state plan, the pluses and minuses of denying discretion are relatively clear and manageable. Every employee is treated the same; each has a better opportunity to get benefits provided by the plan and each will bear his or her proportionate share of any employer costs that may affect what benefits are provided. But for multistate plans, employees in different states may be treated differently if the meaning (or enforceability) of the provisions of the plan differ depending on the state

29

where the employee lives or works. Those whose benefits are governed by discretion- denying statutes will have a better chance of receiving benefits than those governed by the law of states without such statutes. We have already noted that the Supreme Court in Shaw expressed how such a state of affairs would run contrary to ERISA policy to “minimize[] the need for interstate employers to administer their plans differently in each State in which they have employees,” and would adversely affect plan beneficiaries. 463 U.S. at 105. Moreover, the disparity in treatment of plan participants in different states may make it difficult, if not impossible, to determine the solvency of the ERISA plan. See Conkright at 517–18.

All this is not to say that discretion-denying statutes are good or bad. As stated above, ERISA itself is agnostic on the matter. But if the plan has a legitimate connection to the State whose law is chosen (since Pennsylvania is where Comcast is incorporated and has its principal place of business, there can be no question of the propriety of the Policy’s selecting the law of Pennsylvania), ERISA’s interest in efficiency and uniformity, as well as its recognition of the primacy of plan documents, compels the conclusion that the selected law should govern whether a discretion-granting provision is enforceable. A clear, uniform rule enforcing an ERISA plan’s choice of law is required to ensure plan administrators enjoy the predictable obligations and reduced administrative costs central to ERISA—particularly as the choice of law affects the validity of

30

discretionary clauses. We therefore decide that, as a matter of federal law, the choice of law in the Policy governs. Pennsylvania law applies to this dispute. 7 Ellis’s arguments to the contrary are unpersuasive. He first contends that choice-

of-law provisions incorporate only substantive law, and “[b]ecause Colorado law dictating the standard of review applicable to ERISA benefits decision[s] is procedural, it applies here despite [the] Policy’s choice-of-law provision.” Aplee. Br. at 18. But the very case Ellis cites for this proposition explains that in cases arising under federal law, federal rules govern procedural issues, meaning Colorado law would still be inapplicable. See FDIC v. Petersen, 770 F.2d 141, 142–143 (10th Cir. 1985) (in action brought by United States, limitations period from relevant federal statute applied because Illinois

7 In Dang v. Unum Life Insurance Co. of America, this court considered a claim for benefits under an ERISA plan without a choice-of-law provision. See 175 F.3d 1186, 1190 (10th Cir. 1999). To determine which state’s law regarding the notice-prejudice rule to incorporate as the rule of decision, Dang applied the forum state’s choice-of-law rule. See id. Its rationale for doing so, however, appears incorrect on its face. The court explained that “[a] federal court adjudicating state law claims must apply the forum state’s choice of law principles,” citing Klaxon Co. v Stentor Electric Manufacturing. Co., 313 U.S. 487, 496 (1941). Id. But Dang’s claim was brought under ERISA, see id. at 1188, and, as discussed at some length above, the Supreme Court has made clear that claims brought under ERISA’s civil-enforcement provision are federal claims. See Metropolitan Life, 481 U.S. at 67; Pilot Life, 481 U.S. at 56. Moreover, a few days before Dang was filed, the Supreme Court explained that even if a state-law rule of decision (there, a notice-prejudice rule, as in Dang) is incorporated into federal law to resolve an ERISA benefits claim, there is still no state-law claim. See Unum, 526 U.S. at 377 (1999). All that is involved is that “[t]he notice-prejudice rule supplied the relevant rule of decision for this § [1132](a) suit.” Id.

Dang, however, may have intended to declare merely that a federal court should follow a forum state’s choice-of-law rule when the applicable ERISA plan has no choice-of-law provision. But we need not decide whether Dang’s analysis or conclusions are correct in that context under current law because the Policy in this case contains a choice-of-law provision.

31

choice-of-law provision in guarantee contract presumably did not apply to such procedural issues). In any event, Ellis failed to make this argument to the district court. Since he does not argue for plain error on appeal, we consider the argument waived. See Richison v. Ernest Grp., Inc., 634 F.3d 1123, 1130–31 (10th Cir. 2011).

Ellis next mounts several arguments that applying Pennsylvania law would be unfair, but these would be unpersuasive even if we were not bound by the plan- documents rule. He argues that choice-of-law determinations should consider unfair surprise to litigants and that he had no contacts with Pennsylvania. But the Policy’s choice of Pennsylvania law was clear on its face, preventing any such surprise. Ellis also claims that unfairness is demonstrated by the fact that the district court ruled in his favor under de novo review but ruled in Liberty’s favor when it reviewed for abuse of discretion. This difference in result is a well-recognized possibility, which is the justification for state laws like Colorado’s that require de novo review. We have already observed, however, that the Supreme Court has endorsed choices by ERISA plans to provide abuse-of-discretion review, noting the potential benefits for both employees and employers. See Conkright, 559 U.S. at 517. We see no unfairness.

B. Review of Liberty’s Decision The district court initially ruled that it should review for abuse of discretion (it had originally decided that the Colorado discretion-denying statute did not apply because it postdated the Policy) and upheld the decision by Liberty. But on a motion for reconsideration by Ellis, it changed its mind regarding applicability of the Colorado statute, exercised de novo review, and ruled in favor of Ellis. Its first decision was

32

correct. We agree that under an abuse-of-discretion standard, Liberty’s denial of benefits must be affirmed.

We uphold a plan administrator’s decision under the abuse-of-discretion standard “so long as it is predicated on a reasoned basis.” Adamson v. Unum Life Ins. Co. of Am., 455 F.3d 1209, 1212 (10th Cir. 2006). We ask only that the decision “reside[] somewhere on a continuum of reasonableness—even if on the low end.” Id. at 1212 (internal quotation marks omitted). “Plan administrators, of course, may not arbitrarily refuse to credit a claimant's reliable evidence, including the opinions of a treating physician,” Black & Decker Disability Plan v. Nord, 538 U.S. 822, 834 (2003), but a benefits decision can be reasonable even when the insurer receives evidence contrary to the evidence it relies on, see Holcomb v. Unum Life Ins. Co. of Am., 578 F.3d 1187, 1193–94 (10th Cir. 2009). “[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, 538 U.S. at 834.

Ellis contends that under the Supreme Court’s decision in Metropolitan Life Insurance Co. v. Glenn, 554 U.S. 105 (2008), we must use a less deferential standard when, as here, the administrator operates under a conflict of interest by both paying out benefits and adjudicating claims. But Glenn explicitly rejected this proposition, saying that “[t]rust law continues to apply a deferential standard of review to the discretionary decisionmaking of a conflicted trustee.” Id. at 115. It instead instructed that “the

33

reviewing judge . . . take account of the conflict when determining whether the trustee, substantively or procedurally, has abused his discretion.” Id.; see id. at 115–17. “[C]onflicts are but one factor among many that a reviewing judge must take into account.” Id. at 116. Indeed, we have relied on Glenn to explain that the effect of a conflict is case-specific and “‘prove[s] less important (perhaps to the vanishing point) where the administrator has taken active steps to reduce potential bias and to promote accuracy,’” including by utilizing independent physicians. Holcomb, 578 F.3d at 1193 (quoting Glenn, 554 U.S. at 117 (2008)).

Under the Policy a claimant cannot receive long-term disability benefits without proof of disability arising from an injury or sickness that renders the claimant unable to perform the duties of any occupation. We hold that Liberty did not abuse its discretion in deciding that Ellis had not established such a disability in light of the evidence presented to it.

Ellis obtained support for his claim of neurological impairment from several sources. First, in June 2012 he consulted with Dr. Zacharias, a neurologist, who noted Ellis’s complaints of “poor concentration, dizziness, slowing of physical and mental skills,” and recommended cognitive and physical therapy. Aplt. App., Vol. I at 268–69. He also wrote, though, that he “did not do detailed neuropsychological testing,” and that Ellis was “alert and attentive,” had no “evidence of a primary neuromuscular disease,” and had an “[u]nremarkable brain MRI.” Id. at 269. In August 2012, Dr. Zacharias wrote in a follow-up report: “I am still not sure what accounts for Michael’s condition. My best assessment would be something happened with hypoxic injury with either his

34

syncopal episode or his pulmonary embolism.” Id. at 265. He noted Ellis was progressing with therapy, “but still struggles significantly.” Id. at 265. Completing a restrictions form sent by Liberty in May 2013, Dr. Zacharias provided a diagnosis for Ellis of “hypoxic/ischemic encephalopathy” and in response to a question asking for a description of his “physical, mental and/or cognitive restrictions,” he directed Liberty “to see neuropsych testing that supports his impairment.” Id. at 239.

Also in May 2013, Ellis’s primary-care physician, Dr. Hadley, reported in response to a Liberty restrictions form that Ellis suffered from “cognitive impairment from hypoxic encephalopathy.” Id. at 192 (internal quotation marks omitted).

In April 2013 one of Liberty’s consulting neuropsychologists, Dr. Crouch, opined that Ellis “would likely be precluded from performing the usual duties of his job, regardless of accommodations provided.” Id. at 187. He also commented that Ellis’s test results from Dr. Helffenstein’s evaluation appeared “valid/reliable.” Id. at 187. In September 2013, Dr. Crouch reaffirmed that the available records “provide reasonable support for significant impairment,” although he suggested that Ellis be reevaluated. Id. at 189.

Ellis has relied most heavily on the conclusions of Dr. Helffenstein, a neuropsychologist, who first tested Ellis in August and September 2012. According to his report, “The testing identified significant cognitive deficits suggesting bilateral frontal and bilateral temporal involvement. The pattern is consistent with cerebral hypoxia. There is absolutely no way Michael could do his job at this time from a cognitive standpoint.” Aplee. Supp. App., Vol. II at 578 (internal quotation marks omitted). He

35

also opined “within reasonable neuropsychological probability that the cognitive deficits noted on testing related directly and solely to the medical event that occurred on February 1, 2012. It seems reasonable that an episode of cerebral hypoxia did occur during this event.” Aplt. App., Vol. I at 262. Dr. Helffenstein reevaluated Ellis in May 2014. He reported that Ellis “demonstrated . . . notable improvement on testing from my first evaluation with him to my re-evaluation,” id. at 144, but maintained that he was “totally and permanently disabled from competitive employment,” id. at 145.

Finally, Ellis obtained a SPECT scan in May 2014, and the neurologist interpreting the scan concluded: “The nature, location, and pattern of these abnormalities is most consistent with the scientific literature pertaining to a diffuse, toxic/hypoxic encephalopathic process and the patient’s clinical history which was received after the blind review.” Id. at 220.

But Liberty had sound reasons not to adopt the above views. Dr. Hadley, Ellis’s primary care physician, does not specialize in neurology or neuropsychology, and was likely just deferring to the views of specialists. And Dr. Crouch, a consulting neuropsychologist for Liberty, consistently expressed the view that a 24-second heart stoppage could not cause neurological injury, and he came to have doubts whether Ellis suffered cognitive impairments. In September 2013 he had suggested a reevaluation and after receiving Dr. Gant’s evaluation of Ellis in October, Dr. Crouch opined that the results “are insufficient to support the presence of valid/reliable impairment” and that “results from multiple measures of response bias were suboptimal, indicating that

36

observed abnormal test results were ‘related to the patient’s desire to obtain disability benefits.’” Id. at 194 (internal quotation marks omitted).

As for Dr. Zacharias, he admitted that he did not conduct detailed neuropsychological testing, and the brain MRI was “[u]nremarkable.” Id. at 269. The restrictions form he completed for Liberty included no new data and simply directed Liberty to see prior testing. Further, the persuasiveness of his conclusions in support of Ellis is diminished by his adoption of the theory that Ellis’s claimed deficits may have been caused by cerebral hypoxia stemming from his 24-second heart stoppage, as this was deemed medically implausible by essentially every other physician to review the case.

The SPECT scan is obviously objective data, but the relevance could reasonably be questioned by Liberty. Dr. Belliveau expressed doubts:

Scientific studies about the utility of SPECT procedures during evaluation of dementia or brain injury due to trauma may not necessarily be applicable to evaluation of brain injury due to hypoxic-ischemic events, and . . . the cognitive and psychological assessment methods used during neuropsychological examination represent a more direct process of determining the examinee’s functional status.

Id. at 110. Although Ellis submitted to Liberty a number of medical-journal articles and court documents discussing the utility of SPECT scans, these focused almost exclusively on evaluating traumatic brain injury—without any mention of their utility in assessing hypoxic injury. And Ellis has not alleged that his disability was caused by physical trauma to the brain.

37

There remains Dr. Helffenstein. Liberty could reasonably have questioned his objectivity. He was hired by Ellis’s attorney to evaluate Ellis; and his initial report in November 2012 appears to have been a bit overenthusiastic. Although he had been advised that the duration of Ellis’s cardiac standstill had been only 24 seconds, the report stated that Ellis’s “cognitive deficits noted on testing relate directly and solely to the medical event that occurred on February 1, 2012. It seems reasonable that an episode of cerebral hypoxia did occur during this event.” Id. at 262 (emphasis added). By August 2013 he had walked back this theory, stating that he would “totally concur” with the assessment that “it is highly unlikely that the reported 24-second period of asystole on February 1, 2012 would be the cause of [Ellis’s] cognitive complaints.” Id. at 228 (internal quotation marks omitted). He proposed instead that Ellis’s “cognitive dysfunction most likely relates to a more extended period of cerebral hypoxia,” but failed to identify how or when such an event might have occurred. Id. His final report in July 2014 then broadened his original hypothesis, suggesting that Ellis “experience[d] some type of neurological event (likely a hypoxic episode or episodes) during the early part of February of 2012 related to his various medical conditions.” Id. at 144. But he added: “I am not sure that any physician or neuropsychologist could point to a specific time or event that resulted in Mr. Ellis’[s] injury but, at this point, I am absolutely convinced that such an injury did occur.” Id.

Most importantly, two neuropsychologists challenged Dr. Helffenstein’s methods and the validity of his results. Dr. Belliveau, Liberty’s consulting neuropsychologist, questioned the results of Dr. Helffenstein’s testing because of significant evidence of

38

symptom overreporting and other evidence of invalidity. He noted that tests differ in their ability to detect insufficient effort and that Ellis had passed the less sensitive tests but failed those that are more sensitive to insufficient effort. In light of the “multiple findings of invalid neuropsychological test data,” Dr. Belliveau concluded that the “available medical record documentation” in Ellis’s file “represents insufficient support for the conclusion that the claimant has permanent cognitive impairment due to hypoxic- ischemic encephalopathy.” Id. at 124.

Similarly, Dr. Gant, the independent neuropsychologist Liberty retained from an outside vendor, criticized Dr. Helffenstein for using outdated and inadequate tests. He conducted his own testing and evaluation but decided that many of the test scores were invalid. He reported: “It is unlikely that [Ellis] provided valid effort during this examination. Clear evidence of symptom exaggeration and suboptimal effort was identified.” Id. at 202. In particular, he observed:

Ellis reported an unusual and elevated degree of neurological complaints which are likely to be vague and illogical. . . . [Other tests] indicated that the degree of neurological impairment reported by . . . Ellis was highly atypical and illogical. Such a presentation includes symptoms that are illogical or inconsistent with symptoms of a bona fide neurologic disorder or they occur very rarely in neurologically impaired patients.

Id. at 203–04. He concluded that “within reasonable medical probability [Ellis] has not suffered cognitive impairment related to the asystole event which lasted 24 seconds on February 1, 2012,” and that “elements of secondary gain and/or impairment related to somatic exaggeration is responsible for [Ellis’s] presentation.” Id. at 196.

39

On this record we cannot say that Liberty’s denial of benefits was an abuse of discretion. Ellis criticizes several aspects of Liberty’s decision-making. Although some of the criticism has weight, a decision is not arbitrary and capricious just because some may be persuaded otherwise. Ellis first asserts that Liberty improperly relied on the conclusions of its hired reviewers despite flaws in their testing methods and reports. He argues that Liberty failed to credit Dr. Helffenstein’s claim that fatigue during testing could alone account for Ellis’s “sub optimal performance on symptom validity measures” during Dr. Gant’s testing. Id. at 135. But Dr. Helffenstein does not explain how fatigue could cause the apparently intentionally dishonest reporting observed by Dr. Gant. And even though Ellis was provided breaks during his 2012 evaluation with Dr. Helffenstein, Dr. Belliveau expressed doubts as to the validity of the scores obtained during that evaluation as well —contrary to the suggestion that fatigue fully accounted for the symptom exaggeration and other measures of invalidity that Dr. Gant observed.

Ellis also claims that Dr. Gant did not review Dr. Helffenstein’s raw data, and that Dr. Crouch, who did, opined that Dr. Helffenstein’s test findings were valid and reliable. But Dr. Gant was still able to criticize the testing on the ground that the tests were out of date and that “inadequate testing was done to evaluate patient effort and test validity”; and he suggested that the raw data be obtained. Id. at 202. In any event, Dr. Belliveau did review that data and, like Dr. Gant, criticized Dr. Helffenstein’s results and methods, not only stating that Dr. Helffenstein used outdated tests but also that his data indicated symptom overreporting. And Dr. Crouch agreed with Dr. Gant’s conclusion that Ellis had not been candid in Dr. Gant’s testing.

40

Ellis criticizes Liberty’s instructions for conducting a July 2013 vocational report.

The vocational case manager who submitted the report was asked “to base [the] report on a presumed sedentary work capacity, and not to include any cognitive and/or mental restrictions and limitations” in her assessment. Aplee. Supp. App., Vol. II at 455. Ellis argues that these instructions are “clear[] evidence that Liberty never intended to provide Ellis with a full and fair review of his claim, but instead, conducted a result-oriented investigation solely intended to terminate his benefits.” Aplee. Br. at 49. The argument is not totally off-the-wall, but it is a stretch. The record shows that the vocational report was for “an exploratory TSA [transferable skills analysis],” Aplee. Supp. App., Vol. I at 28, which would be necessary because in two months Mr. Ellis’s eligibility for disability would require inability to perform the “material and substantial duties of any occupation” rather than “of his own occupation,” Aplt. App. at 296 (emphasis added). That would not be a nefarious purpose for conducting the limited evaluation, particularly since Liberty was at the same time pursuing the medical basis of the alleged cognitive deficits and would thus later be able to assess Ellis’s ability to perform the alternate occupations identified in the vocational report in light of any mental limitations.

Ellis complains that Liberty instructed that Dr. Crouch, who had expressed some support for Ellis’s claim, should not be assigned to review the file on internal appeal. But the reason given for the instruction was that “he previously handled the file.” Aplee. Supp. App., Vol. I at 55. On its face, it seems reasonable, and apparently legally mandated, to have an appeal handled by persons other than those who handled the initial decision. See 29 C.F.R. § 2560.503–1(h)(3)(v).

41

Finally, Ellis claims that Liberty ignored other evidence demonstrating he was cognitively disabled, namely (1) Liberty’s surveillance of him, (2) his Social Security Disability Insurance (SSDI) award, and (3) a letter from his speech therapist and other clinical notes from various providers. But the record rebuts this assertion.

To begin with, Drs. Belliveau, Crouch, and Gant all reviewed the surveillance reports as part of their consideration of Ellis’s claim. In particular, Dr. Crouch remarked that there were “[n]o cognitive [symptoms] documented” in the first surveillance report, Aplt. App., Vol. I at 182, and “[n]o abnormalities noted in limited visual contact” in the second report, id. at 191, diminishing their relevance to Ellis’s claim of cognitive impairments.

Liberty also acknowledged the SSDI award. Its letter denying benefits stated that it was “aware [of] and fully considered” the December 2013 ruling of the Social Security Administration (SSA) granting the award. It explained, though, that its decision was “based upon updated medical records and testing, and different medical and vocational reviews that would not have been considered by the SSA in December 2013,” and that the SSA requirements are not the same as those in the Policy. Id. at 105. Ellis’s SSDI application was submitted before Dr. Gant’s evaluation and report and Dr. Belliveau’s review, and there is no indication that the SSA considered the later reports.

Similarly, the record indicates that all the clinical notes were reviewed by Liberty experts. The experts did not disregard them; it is just that they found that the record considered as a whole was inadequate to support Ellis’s claim. Dr. Belliveau explicitly stated his conclusion was based on “[t]he available medical record documentation,

42

including the scope, severity, and persistence of the claimant’s reported symptoms; observations of his treatment providers; [and] multiple findings of invalid neuropsychological test data.” Id. at 124 (emphasis added). Dr. Gant similarly arrived at his opinion “[a]fter reviewing the medical records, the report completed by Dr. Helffenstein, and completing [his] own evaluation.” Id. at 210. Again, the existence of evidence supporting Ellis’s claim does not render a denial of benefits unreasonable. See Holcomb, 578 F.3d at 1193–94 (upholding benefits denial even though insurer “had received a large volume of reports, letters, imaging studies, and exams that were not entirely consistent”).

In sum, Liberty relied on two expert neuropsychologists, Drs. Gant and Belliveau, who both concluded that there was insufficient evidence from Ellis’s medical records and test data to support his claim of cognitive deficits. Because the record shows Liberty and the experts it retained considered all the pertinent evidence submitted by Ellis and that Liberty reasonably gave less weight to much of Ellis’s evidence, we cannot say that Liberty abused its discretion in denying Ellis’s claim for benefits.

III. CONCLUSION We REVERSE the district court’s judgment in Ellis’s favor and REMAND for entry of judgment in Liberty’s favor.