LIPEZ, Circuit Judge.
This case requires us to determine whether the plan administrator of an employee benefits plan governed by the Employees Retirement Income Security Act (“ERISA”), 29 U.S.C. §§ 1001-1461, may reasonably conclude that the insured, who was killed in a one-car collision with a tree while driving with a blood alcohol content (“BAC”) of three times the legal limit, did not die as a result of an “accident” for purposes of his Accidental Death and Dismemberment (“AD & D”) life insurance policies. After careful review of the developing federal common law under ERISA, we uphold the plan administrator’s determination that, in this case, the insured was so highly intoxicated that his death was not an “accident.” In so doing, we affirm the judgment of the district court.
I.
The facts are undisputed. On August 2, 2002, Steven Stamp attended a meeting for employees of appellee Exxon Mobil Chemical Company (“Mobil”) at a resort in West-brook, Connecticut. The meeting consisted of presentations in the morning, a boat cruise in the afternoon, and dinner that evening. Mr. Stamp spoke with his wife, appellant Karen Stamp, at 5:20 p.m. and confirmed that after dinner he planned to drive to his parents’ home near Providence, Rhode Island, where she and their young daughter would join him to celebrate his brother’s 40th birthday the next day. Mrs. Stamp attests that he did not sound impaired during the phone call, that he “sounded in good spirits,” and that he had enjoyed the meeting and cruise. Mr. Stamp consumed several alcoholic beverages during the boat cruise and at dinner, but his co-workers reported that he did not appear to be impaired or unsteady when he left the resort between 8:30 and 9:00 that evening.
[86]*86At 9:20 p.m., while en route to the Providence area, Mr. Stamp placed a cell phone call to his friend Joe Kingsley, hoping to visit him on his way through Rhode Island. Kingsley declined, explaining that he needed to go to bed early. Kingsley described Mr. Stamp’s mood as positive and upbeat and reported that he did not sound intoxicated.
The drive from Westbrook to Mr. Stamp’s parents’ house should have taken about an hour and a half. Cell phone billing records of calls placed through directory assistance at 9:51 p.m. and 9:58 p.m. indicate that Mr. Stamp was already in the Providence area at that time, but Mr. Stamp evidently did not go directly to his parents’ house. MetLife and Mobil posit that he must have stopped somewhere and consumed additional alcohol instead. As further evidence of this stop, MetLife notes that Mr. Stamp’s hand had a stamp on it that read “copy.” In Met-Life’s view, this fact indicates that Mr. Stamp went to a bar where his hand was stamped at the door.
Just after midnight, Mr. Stamp died when his car went off the road in Johnston, Rhode Island and struck a tree. At the time of his death, his BAC was .265%, more than three times the legal limit in Rhode Island.1 The police report described the road conditions at the time of the collision as dry, the traffic as light, and there were no other adverse driving conditions.2 The autopsy report concluded that the cause of death was “[m]ultiple injuries due to blunt force trauma,” and noted “acute ethanol intoxication” as another significant finding. The report classified the manner of death as “accidental.”
Mrs. Stamp, the beneficiary of her husband’s life insurance policies, submitted a claim for benefits. Appellee Metropolitan Life Insurance Company (“MetLife”), acting as claim fiduciary, paid her claim for Basic Life Insurance benefits, but denied her claims for benefits from her husband’s Basic and Voluntary AD & D policies. MetLife left undetermined Mrs. Stamp’s claim for benefits under an additional Occupational AD & D policy.
Mrs. Stamp submitted a timely appeal to Mobil, which, in its capacity as plan administrator, had “full and exclusive authority to make final determinations as to all issues concerning plan administration,” including “discretionary and final authority to determine coverage and eligibility for benefits ... [and] to interpret and explain the terms of the Life Insurance Program.” Mobil denied Mrs. Stamp’s appeal of the Basic and Voluntary AD & D benefit decisions, and rejected her claim for Occupational AD & D benefits.
The ERISA plan at issue provides that Basic and Voluntary AD & D benefits will be paid if the insured is “physically injured as a result of an accident and die[s] within 90 days as a result of that injury or accident.” Occupational AD & D benefits will be paid if the insured’s death occurs “within one year as a result of an injury caused by an occupational accident ‘while at work.’ ” Mobil concluded that Mr. Stamp’s death was not the result of an “accident,” and therefore was not covered by his AD & D policies.3 The denial letter explained:
[87]*87In your appeal submission you argue that the weight of legal authority compels me to find that the collision in this case was an “accident” within the meaning of the plan. Counsel has reviewed these and other cases and advises that the weight of authority under applicable Federal law would not compel such a finding.
In as much as I am not bound by law, I look to the purpose of the plan. I believe that the purpose of the plan is to protect participants from risks that are outside of their control. The risks flowing from driving while intoxicated are completely within the control of the participant. While it is true that certain behavior that increases risk (such as skiing or horseback riding) would not result in loss of coverage, [driving while intoxicated] can be distinguished because it unreasonably increases the risk associated with a normally safe activity by interfering with an individual’s ability to perceive and respond to risk. To impose the costs of such unreasonable risk-taking on the plan would result in an unanticipated cost.
Mobil further explained that “[t]he fact that the coroner’s report and the police report use the term ‘accident’ does not govern the proper interpretation under the plan.”
Mrs. Stamp filed suit in district court, asserting claims for breach of contract and breach of fiduciary responsibility. The district court held that Mrs. Stamp’s common law claims are preempted by ERISA and treated her suit as an ERISA enforcement action pursuant to 29 U.S.C. § 1132(a)(1)(B). Mrs. Stamp has not appealed that determination. On cross-motions for summary judgment, the district court reviewed Mobil’s denial of benefits under an “arbitrary and capricious” standard and concluded that the “determination that Mr. Stamp’s death was not accidental is reasonable and supported by substantial evidence in the record.” The district court entered judgment in favor of MetLife and Mobil. Mrs. Stamp filed this timely appeal.
II.
We review the district court’s grant of summary judgment de novo. Wright v. R.R. Donnelley & Sons Co. Group Benefits Plan, 402 F.3d 67, 73-74 (1st Cir.2005). Where, as here, the ERISA plan grants the plan administrator discretionary authority to make benefit determinations, we must uphold the administrator’s determination unless it was “arbitrary, capricious, or an abuse of discretion.” Id. at 74 (quoting Doyle v. Paul Revere Life Ins. Co., 144 F.3d 181, 183 (1st Cir.1998)); cf. Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 115, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989). Under this standard, we will uphold the denial of benefits if the plan administrator’s decision was “ ‘reasoned and supported by substantial evidence.’ ” Id. (quoting Gannon v. Metro. Life Ins. Co., 360 F.3d 211, 213 (1st Cir.2004)). “Evidence is substantial if it is reasonably sufficient to support a conclusion.... ” Gannon, 360 F.3d at 213.
Mrs. Stamp argued before the district court that this deferential standard of review was inappropriate because Mobil’s [88]*88contract with MetLife is “experience-rated,” linking Mobil’s premium costs to the number and size of claims presented. She asserted that this link represented a “structural conflict of interest” that would mandate a less deferential standard of review. However, Mrs. Stamp has abandoned this argument on appeal. In her brief, she merely cites the legal proposition that a conflict of interest may reduce the degree of deference we owe the plan administrator; she does not argue that a conflict actually exists in this case. Thus, this argument is waived.4 See United States v. Zannino, 895 F.2d 1, 17 (1st Cir.1990) (“[I]ssues adverted to in a perfunctory manner, unaccompanied by some effort at developed argumentation, are deemed waived.”). Consequently, we review the denial of benefits under the “arbitrary and capricious” standard.
III.
Mr. Stamp’s AD & D policies stated that MetLife would pay accidental death benefits if Mr. Stamp died as a result of an “accident.” The term “accident” is not defined in the plan documents. The plan administrator had the task of applying this term to the circumstances surrounding Mr. Stamp’s death.
In reviewing the administrator’s denial of benefits, we are guided by the principles of the federal common law of ERISA. Pilot Life Ins. Co. v. Dedeaux, 481 U.S. 41, 56-57, 107 S.Ct. 1549, 95 L.Ed.2d 39 (1987) (concluding that Congress intended for a federal common law to develop in the ERISA context); Wickman v. Northwestern Nat’l Ins. Co., 908 F.2d 1077, 1084 (1st Cir.1990). In particular, we rely on Wick-man, a case in which we added to the ERISA common law by formulating an approach for interpreting the ambiguous term “accident” in AD & D insurance policies. 908 F.2d at 1087-88.
We began with the principle that “the reasonable expectations of the insured when the policy was purchased is the proper starting point for a determination of whether an injury was accidental under its terms.” Id. at 1088. But we noted that this was only a starting point. The operative inquiry into the insured’s “expectations” in Wickman actually concerned the insured’s state of mind at the time of the incident that caused his death, not at the time the policy was purchased. See id. at 1089 (“If he actually expected the result, even if he did not specifically intend it, then his actual expectations make his death not accidental.”). We see no necessary inconsistency between these two formulations of the relevant time frame. The conclusion that the insured expected, or reasonably should have expected, that death would result from the particular conduct at issue is tantamount to the conclusion that no reasonable person, when buying an insurance policy, could expect that AD & D benefits would be paid out when his death results from that sort of conduct. In any case, aside from the reference to the expectations at the time of purchase as a “starting point,” the analysis in Wick-man makes no further reference to those expectations and is instead concerned solely with the insured’s expectations related to the intentional conduct that caused his death. We adopt that approach as well.
The Wickman analysis thus began with an inquiry into the expectations of the [89]*89insured at the time of the incident that caused his death. However, this subjective inquiry was not determinative. We held that even if “the fact-finder determines that the insured did not expect an injury similar in type or kind to that suffered, the fact-finder must then examine whether the suppositions which underlay that expectation were reasonable.” Id. at 1088. We further observed that “ ‘the subjective state of mind of the insured cannot be generally known.’ ” Id. at 1087-88 (quoting Hoffman v. Life Ins. Co., 669 P.2d 410, 419 (Utah 1983)). Thus, in the usual case, where the fact-finder will find “the evidence insufficient to accurately determine the insured’s subjective expectation,” the fact-finder “should then engage in an objective analysis of the insured’s expectations.” Id. at 1088. We framed this objective analysis as an inquiry into “whether a reasonable person, with background and characteristics similar to the insured, would have viewed the injury as highly likely to occur as a result of the insured’s intentional conduct.” Id. This reasonable person analysis, “when the background and characteristics of the insured are taken into account, serves as a good proxy for actual expectation.” Id.
We then applied this framework to the particular facts in Wickman, where the insured died after falling from a railroad bridge, having intentionally climbed over the guardrail on the side of the bridge. Although there was some evidence that the insured intended to kill himself, we determined that his actual expectations could not be ascertained. Id. at 1088. Therefore, we applied the objective prong of the Wickman analysis and concluded that “[objectively, he reasonably should have expected serious injury when he climbed over the guardrail and suspended himself high above the railroad tracks below by hanging on to the guardrail with only one hand.” Id. at 1089. Consequently, we upheld the determination that his death was nonaccidental.5
Our reasoning in Wickman has been widely accepted by our sister circuits.6 See Eckelberry v. Reliastar Life Ins. Co., 469 F.3d 340, 343-45 (4th Cir.2006). Applying Wickman, federal courts have, “with near universal accord,” upheld plan administrators’ determinations that “alcohol-related injuries and deaths are not ‘accidental’ under insurance contracts governed by ERISA.” Id. at 344; see also Lennon v. Metro. Life Ins. Co., 504 F.3d 617, 622-23 (6th Cir.2007); Cozzie v. Metro. Life Ins. Co., 140 F.3d 1104, 1110 (7th Cir.1998); Weatherall v. Reliastar Life Ins. Co., 398 F.Supp.2d 918, 924 (W.D.Wis.2005); Mullaney v. Aetna U.S. Health[90]*90care, 103 F.Supp.2d 486, 494 (D.R.I.2000); Walker v. Metro. Life Ins. Co., 24 F.Supp.2d 775, 782 (E.D.Mich.1997); Schultz v. Metro. Life Ins. Co., 994 F.Supp. 1419, 1422 (M.D.Fla.1997); Nelson v. Sun Life Assurance Co., 962 F.Supp. 1010, 1012 (W.D.Mich.1997); Miller v. Auto-Alliance Int’l, Inc., 953 F.Supp. 172, 176-77 (E.D.Mich.1997); Cates v. Metro. Life Ins. Co., 14 F.Supp.2d 1024, 1027 (E.D.Tenn.1996), aff'd, 149 F.3d 1182 (6th Cir.1998) (unpublished). But see West v. Aetna Life Ins. Co., 171 F.Supp.2d 856, 904 (N.D.Iowa 2001). Thus, the plan administrator was correct to conclude, in denying Mrs. Stamp’s benefit claims, that “the weight of authority under applicable Federal law would not compel” a finding that Mr. Stamp’s collision was accidental.7
In applying the Wiclcman analysis to drunk-driving deaths, courts have stated that “ ‘the hazards of drinking and driving are widely known and widely publicized’ ” and reasoned that, as a result, “the insured should have known that driving while intoxicated was highly likely to result in death or bodily harm.” Eckelberry, 469 F.3d at 345 (quoting Nelson, 962 F.Supp. at 1012). As the Supreme Court recently observed, albeit in a different context, “[d]runk driving is an extremely dangerous crime .” Begay v. United States, — U.S. -, 128 S.Ct. 1581, 1584, 170 L.Ed.2d 490 (2008). It is common knowledge that the danger grows even more extreme as the driver’s level of intoxication increases. Nat’l Hwy. Traffic Safety Admin., U.S. Dep’t of Transp., Setting Limits, Saving Lives: The Case for .08 BAC Laws, DOT HS 809 241, revised Apr. 2001. Accordingly,
any drunk driver who takes to the road should know he runs a risk of injuring another person [or himself]. The extent of the risk will of course vary from case to case, depending on how intoxicated the driver is, how far he drives, how fast he drives, and how many other drivers and pedestrians are sharing the road with him.
Lennon, 504 F.3d at 621 (quoting United States v. Rutherford, 54 F.3d 370, 376 (7th Cir.1995)). Following this logic, courts have emphasized the decedent’s level of intoxication when determining that a plan administrator’s denial of benefits was reasonable. See, e.g., id. at 623 (BAC of more than three times the legal limit); Eckelberry, 469 F.3d at 345 (BAC of one and a half times the legal limit); Cozzie, 140 F.3d at 1106 (BAC of more than two times the legal limit). We endorse this approach.8 [91]*91The Wickman analysis does not require a categorical determination that all alcohol-related deaths are per se accidental or nonaccidental. Rather, it leads us to consider the circumstances of the fatal event in question.9
In this case, the plan administrator had substantial evidence that Mr. Stamp was severely intoxicated when he died. The toxicology report indicated that Mr. Stamp’s BAC at the time of his death was in the range of .231% to .265%. MetLife’s medical department had advised that this level of intoxication “would cause delirium intoxication. In a sporadic drinker it would cause lethargy, stupor, combativeness, incoherency & vomiting. In a chronic drinker, there would be [illegible] emotional changes and mood changes.”10 An internal Mobil memorandum, prepared during the review of Mrs. Stamp’s claim, further elaborated:
Various on-line sources indicate that Mr. Stamp’s ability to control his vehicle would have been significantly impaired. For example, the National Highway Transportation Safety Administration states that all drivers, even experienced drivers, show impairment at a BAC of 0.08-one-third the level reported for Mr. Stamp. The “Police Handbook” published by the Oklahoma University Police has the following descriptions of the effects of various BACs:
0.20 Feeling dazed/confused or otherwise disoriented. May need help to stand/walk. Blackouts are likely.
0.25 All mental, physical and sensory functions impaired.
Our own medical department has confirmed that the assessments by MetLife and these internet sources are reasonable, and that serious impairment would be expected.
In its letter denying Mrs. Stamp’s claims, Mobil noted that “[t]here is no evidence to suggest that another vehicle or object, mechanical failure, or adverse weather or road collisions contributed to the collision,” and concluded that “Mr. Stamp’s BAC was at least a substantial contributing cause of the collision.” In light of this well-developed record of the severe impairment that would be expected from Mr. Stamp’s level of intoxication and the administrator’s reasonable inference that his drunk-driving was a “substantial contributing cause” of the crash, we cannot say that it was arbitrary and capricious for the administrator to determine that Mr. Stamp’s death was not “as a result of an accident.” In Wickman terms, it is not arbitrary and capricious to conclude that a reasonable person would view death or serious injury as a highly likely outcome of driving while so drunk that one may need help to stand or walk and is likely to black out.11
[92]*92IV.
Mrs. Stamp makes four arguments to support her claim that Wickman compels us to reach a contrary determination. First, she marshals evidence from phone calls earlier in the evening to make the case that her husband did not sound impaired, was “in good spirits,” and was looking forward to a family event the next day. She concludes that this evidence indicates “clearly” that Mr. Stamp “had no subjective expectation of death or injury.” We disagree. The conversations she cites shed little light on Mr. Stamp’s state of mind many hours later after he had consumed more alcohol. Moreover, Mr. Stamp’s actual subjective expectation that he would arrive safely at his parents’ house, even if it were clearly established in the record, is not dispositive. In Wick-man, we explicitly rejected the contention that “unless [the insured] actually expected to die, essentially that he specifically intended to commit suicide, his death must be considered an accident.”12 908 F.2d at 1087. Instead, the critical determination under Wickman lies in an objective analysis of what the insured reasonably should have expected when he decided to drive while highly intoxicated.
Second, Mrs. Stamp contends that the statistical evidence she presented to the district court “establishes that it cannot be reasonably concluded by either common sense or legal analysis that death resulting from driving under the influence of alcohol is ... ‘highly likely.’ ” Her statistics, gleaned from a comparison of the number of drunk driving arrests and the number of alcohol related driving fatalities, show that “the statistical probability of a fatal accident resulting while driving under the influence of alcohol is 1.19%.” However, as the district court properly noted, such statistics are “meaningless in this context.” They simply do not take into account the degree of intoxication of the driver in question. As we noted earlier, the risk of being involved in a fatal crash rises as blood alcohol levels rise. Accordingly, statistics that do not differentiate between various levels of intoxication are inappo-site. See Lennon, 504 F.3d at 623 (“[D]rivers with blood-alcohol levels above the legal limit as a group are far more likely to arrive home safely than drivers who are extremely drunk.”).
Moreover, the focus of our objective analysis in Wickman was not on the statistical probability that death would occur from the decedent’s actions. Instead, we were concerned chiefly with what a reasonable person would perceive to be the likely outcome of the intentional conduct. Wickman, 908 F.2d at 1089. Russian roulette provides an archetypal example of this critical distinction. From a statistical standpoint, the likelihood of dying from a single round of Russian roulette is 16%-one in six. Such a death is not “highly [93]*93likely” if that phrase is taken to mean “more likely than not” or “substantially certain.” In fact, those who play Russian roulette have a decent chance, statistically speaking, of not being injured. Lennon, 504 F.3d at 623. Nonetheless, such a death “would not be publicly regarded as an accident” because the mortal risk associated with playing Russian roulette is patently obvious to any reasonable person. Wickman, 908 F.2d at 1087. Similarly, even if Mrs. Stamp had adduced evidence that those who drive while extremely drunk have a better than even chance of arriving home safely, that evidence would not have been dispositive. Statistical analysis is simply not at the core of the Wick-man analysis.13 Instead, as the Sixth Circuit has explained, Wickman’s framework reflects that “at some point the high likelihood of risk and the extensive degree of harm risked, weighed against the lack of social utility of the activity, become not marginally but so overwhelmingly disproportionate that the resultant injury may be outside a definition of ‘accidental’ that is not unreasonably narrow.” Lennon, 504 F.3d at 623. It was not arbitrary for the plan administrator here to conclude that Mr. Stamp’s decision to drive while grossly intoxicated qualifies as overwhelmingly and disproportionately risky conduct. •
Third, Mrs. Stamp challenges the conclusion that her husband’s death was causally related to his intoxication. She speculates that the accident could have been caused by an attempt to avoid a collision with another vehicle or with an animal. Because we are concerned with the definition of “accident” as a threshold of eligibility for benefits, the burden of proof is on Mrs. Stamp to show the existence of coverage. See 17A Lee Russ & Thomas F. Segalla, Couch on Insurance § 254:16 (3d ed.2007); see also Mario v. P & C Food Markets, Inc., 313 F.3d 758, 765 (2d Cir.2002) (asserting, in an ERISA case, that “the insured has the burden of proving that a benefit is covered, while the insurer has the burden of proving that an exclusion applies”); Jenkins v. Montgomery Indus., Inc., 77 F.3d 740, 743 (4th Cir.1996) (same); Blair v. Metropolitan Life Ins. Co., 974 F.2d 1219, 1221 (10th Cir.1992) (same). Mrs. Stamp’s speculation regarding alternative causes for the collision falls far short of the evidence needed to carry her burden. As a result, the plan administrator’s finding of fact as to causation is not “arbitrary and capricious.”
Fourth, Mrs. Stamp invokes the contra proferentem doctrine, which holds that the policy terms must be strictly construed against the insurer and in favor of the insured. However, this doctrine is only applied when courts undertake de novo review of plan interpretations:
When the administrators of a plan have discretionary authority to construe the plan, they have the discretion to determine the intended meaning of the plan’s terms. In making a deferential review of such determinations, courts have no occasion to employ the rule of contra proferentem. Deferential review does not involve a construction of the terms of the plan; it involves a more abstract [94]*94inquiry—the construction of someone else’s construction.
Morton v. Smith, 91 F.3d 867, 871 n. 1 (7th Cir.1996); see also Winters v. Costco Wholesale Corp., 49 F.3d 550, 554 (9th Cir.1995). We are engaged in that “more abstract inquiry” here, and thus the doctrine is inapplicable to our review.
V.
Under the “arbitrary and capricious” standard, “we need not decide what is the best reading of the words in the insurance policy,” Lennon, 504 F.3d at 624, nor how we would have applied those words de novo. Instead, we are called upon only to decide whether the plan administrator’s denial of benefits was “ ‘reasoned and supported by substantial evidence.’ ” Wright, 402 F.3d at 74 (quoting Gannon v. Metro. Life Ins. Co., 360 F.3d 211, 213 (1st Cir.2004)). As we have set forth above, the administrator’s determination that Mr. Stamp’s death was nonaccidental is compatible with Wickman. Objectively, he reasonably should have expected death or serious injury when he drove with a BAC of more than three times the legal limit. That conclusion does not lessen the tragedy of Mr. Stamp’s death. It simply means that the plan administrator’s denial of benefits was not arbitrary and capricious, and the judgment below is affirmed.
So ordered.