DAUGHTREY, J., delivered the opinion of the court, in which, KEITH, J., joined.
MERRITT, J. (p.-), delivered a separate opinion concurring in the reversal of the judgment below but changing the instructions on remand.
OPINION
MARTHA CRAIG DAUGHTREY, Circuit Judge.The plaintiff, Wanda Glenn, filed this ERISA action against defendant Metropolitan Life Insurance Co., the plan administrator for Sears, Roebuck and Co., after MetLife terminated the long-term disability benefits that Glenn had been receiving as a former Sears employee, on the ground that her condition had improved to the point that she was no longer “totally disabled.” The district court upheld Met-Life’s determination, finding that the decision was not arbitrary and capricious, and granted summary judgment to the- defendants. For the reasons set out below, we conclude that the record does not support the conclusion that MetLife’s denial of Glenn’s claim was the result of a deliberative process or that it was based on substantial evidence. We therefore reverse the judgment of the district court and remand the case for further proceedings.
FACTUAL AND PROCEDURAL BACKGROUND
The plaintiff, Wanda Glenn, worked for Sears, Roebuck from 1986 until April 29, 2000, at which point she took a medical leave of absence and has not since returned to work. Glenn’s final position with Sears was as sales manager in the women’s department, which entailed a workweek of 40-50 hours per week and the training and direct supervision of 20-30 sales associates. The description of her job, prepared by Sears, indicated that the job required sitting up to 20 percent of the workday, standing for 20-60 percent of the workday, and some climbing, reaching, stooping, and lifting. Glenn was responsible for supervising other employees, ensuring that her department was properly stocked with merchandise, satisfying the needs and demands of customers, and identifying and solving various problems as they arose. She reported to the store’s general manager.
Glenn submitted a disability claim under Sears’s long-term disability plan on June 20, 2000, attaching a letter from her treating physician, Dr. Rajendera C. Patel, dated April 30, 2000, indicating that Glenn had been diagnosed with “severe dilated cardiomyopathy,” a disease of the heart muscle that causes the heart to become enlarged and, for that reason, to pump inadequately. Her symptoms included general fatigue and shortness of breath on exertion. Dr. Patel further stated, “From my standpoint, this patient cannot return to any kind of job that would require any significant physical or psychological stress.”
The Sears disability plan for which Glenn applied covered two distinct stages of “total disability.” The first provided that a participant was totally disabled when she was “completely and continuously unable to perform each of the material duties of [her] regular job.” The second category became relevant after the first 24 months of benefits and required that the participant be “completely and continuously unable to perform the duties of any gainful work or service for which [she is] reasonably qualified taking into consideration [her] training, education, experience, and past earning.” (Emphasis added.) Glenn’s claim was approved, and she began receiving long-term disability benefits fol*663lowing the completion of the 140-day elimination period mandated in the plan.
Two months later, in August 2000, Glenn filed for Social Security disability benefits at the direction of plan administrators, who steered Glenn to Kennedy & Associates, a law firm specializing in obtaining such benefits. The Social Security Administration initially denied Glenn’s claim, but an administrative law judge eventually issued a decision finding that Glenn was totally disabled as of April 30, 2000, and was entitled to benefits retroactive to October 2000. As a result, MetLife demanded reimbursement for its overpayment of benefits, based on the amount of Glenn’s newly awarded Social Security benefits, and collected $13,502.50 from her.
On May 20, 2003, MetLife sent a letter advising Glenn that if she wished to continue receiving long-term disability benefits, she was required to demonstrate that she met the second definition for total disability, ie., that she was “incapable of performing the material duties of any gainful occupation as defined by the plan.” The letter further advised that, in making a claim determination, MetLife would review “[Glennj’s vocational information, medical information and [Glennj’s specific restrictions and limitations that are supported by objective medical evidence.”
Glenn’s medical information was provided largely by Dr. Patel,- the cardiologist who oversaw Glenn’s care both before and during the period relevant to this appeal. The record indicates that Glenn had developed hypertension in the early 1980s and had been treated with anti-hypertension therapy since at least 1985. In 1989, she experienced “sudden cardiac death” but was resuscitated and implanted with a defibrillator system. Diagnosed with left ventricular dysfunction, Glenn was hospitalized twice again in the 1990s, once before and once after she began work at Sears. In 2000, she began experiencing prolonged chest tightness, shortness of breath, “increasing fatigue by the end of the day,” ■ and edema in her legs from “prolonged standing at work.” Because of her progressive ventricular dysfunction, at one point Dr. Patel referred Glenn to Dr. Carl Leier at Ohio State University Hospital to consider the possibility of a heart transplant. In March 2000, Dr. Patel ordered an angiogram, from which he diagnosed Glenn with severe dilated cardiom-yopathy. In a letter dated April 7, 2000, Dr. Patel indicated that his patient’s “main problem now is stress at work. She works in a Sears store and does have physical as well as psychological stress. Considering her low ejection fraction, I feel that she may not be able to continue to work in any kind of environment that would Cause any significant physical or psychological stress and demands.” Glenn took medical leave from Sears at the end of that month.
Glenn underwent significant treatment during the next several months, taking as many as seven or eight prescription medications for her heart condition. Dr. Patel noted that, almost three months after ceasing to work at Sears, Glenn was “doing clinically well.” He observed that “some of her improvement is from the reduction of stress and strain of work. She still has good and bad days and in fact there are periods where she feels extremely tired and fatigued and runs out of steam.” In November 2000, however, Dr. Patel notified MetLife that Glenn was totally disabled from undertaking any occupation and that he did not expect her to be able to return to work.
In March 2001, Glenn underwent a further evaluation of her cardiomyopathy that found her ejection fraction found to be “mildly reduced” and reported “some improvement in her LV [left ventricular] function.” Through a walking regime, Glenn had been able to increase her “exer*664cise tolerance,” but Dr. Patel noted that she “still gets fatigued out and short of breath, particularly if she is under any kind of significant psychologic stress.” Dr. Patel also reported, “From overall evaluation, Wanda seems to be clinically stable and I have advised her to continue her current regimen including Lasix, potassium, Lanoxim, Norvasc, aspirin, calcium, Normodyne and Diovan.” In March 2002, a year later, when Dr. Patel filled out a MetLife long-term benefits evaluation form, he indicated that Glenn could lift and carry up to 20 pounds occasionally, had the ability to perform bilateral repetitive foot and hand movements, and could sit eight hours of the day, stand four, and walk two. He also checked “yes” when asked on the form whether she was “able to work in a sedentary physical exertion level occupation.” Although the evaluation form was intended to assess capacity for full-time work, Dr. Patel signed it without indicating whether he was releasing Glenn to return to work or what restrictions might apply if she did go back to work. In a later evaluation, Dr. Patel’s answers were consistent with his earlier responses, except that he added, “No emotional stress/No heavy exertion.” Dr. Patel filled out that form on June 12, 2002.
However, just six days later, Dr. Patel sent a progress report to Glenn’s internist indicating that Glenn “was back in our office earlier than expected,” complaining of fatigue, “shortness of breath on moderate exertion,” and “significant anxiety ... regarding the disability and having to return to work.” Dr. Patel concluded, “From my standpoint, again, considering her cardiomyopathy, I do not believe she will handle any kind of stress well at her work and she would be better off being on disability” rather than returning to her job. Dr. Patel reiterated his opinion that Glenn should not return to work in a letter dated July 22, 2002:
[Glenn] has dilated cardiomyopathy as well as a history of ventricular tachycardia requiring an implantable defibrillator device. The patient has been on multiple medications for her dilated cardiom-yopathy. The patient continues to have significant difficulty with exertional shortness of breath on any kind of moderate exertion. She also continues to have significant difficulty in returning to even any kind of sedentary job because any kind of psychologic stress at work causes significant problems with her cardiovascular condition and she decompen-sates fast. This has happened on multiple occasions in the past. The patient has tried to return to work in the past with exacerbation of her symptoms as well as exacerbation of her condition. At the present time, I do not believe Wanda should be forced to return to any kind of even sedentary work particularly because it is the psychologic stress of work that really exacerbates her cardiovascular condition and symp-tomology. The patient basically should be considered completely disabled from her dilated cardiomyopathy as well as history of ventricular tachycardia.
On August 28, 2002, MetLife notified Glenn that her long-term benefits would be terminated on September 16, 2002, based on the administrative decision that “[tjhere is no supportive medical documentation of the exacerbation of your cardiac condition and symptomology, due to subjective complaints of work-related stress.” The notice concluded that “records submitted for review do not support cardiovascular impairment that would prevent you from performing full time sedentary work” and pointed primarily to medical information showing that Dr. Patel had found her to be “clinically stable” with “a better exercise tolerance” on November 2, 2001, and — based on a check mark on MetLife’s evaluation form — able to return to work on *665March 13, 2002. MetLife’s letter did not address Dr. Patel's letter of July 22, 2002, in which he stated that Glenn should not return to work.
Glenn appealed the denial of disability benefits. While the appeal was pending, Dr. Patel submitted a letter dated February 12, 2003, in which he reiterated his opinion that Glenn could not return to work:
Previous reports filled out by me state that the patient was fit for sedentary work, however based on her clinical condition and her symptomology, there was never a time where I felt that this patient would be able to return to full-time employment. I strongly believe that employment would put significant stress on her overall system, and she may de-compensate. Ms. Glenn does not have any emotional condition that causes her cardiac problem. She has a cardiac problem that is exacerbated by any kind of stress.
My position is that she should be considered completely disabled, mainly from her cardiomyopathy and associated symptoms from her LV dysfunction.
In response, MetLife referred the case to an Independent Physician’s Consultant Board. There, Dr. Chandrakant Pujara reviewed Glenn’s medical records and offered an opinion based' on a review of Glenn’s file. In a letter dated May 2, 2003, he noted that Glenn’s “ejection fraction indeed has improved with medical therapy and probably with biventricular pacing as well” but concluded that Glenn was unable to do any “exertional physical activity.” Dr. Pujara’s prognosis was, at best, ambiguous:
The actual impact of any form of real or perceived emotional stress on cardiac arrhythmias, or cardiomyopathy is difficult to gauge in an individual patient. However, based on Dr. Patel’s own physical capacity evaluation on June 12, 2002, the patient seems to be a reasonable candidate to try one of the sedentary job classes at least on a trial basis. If the job environment entails [a] significant degree of emotional stress, and the patient is not able to cope with that, then certainly permanent disability can be considered.
Despite the equivocal nature of this report, MetLife issued a final denial of disability benefits to Glenn on May 20, 2003. The notice of denial relied on the same information as in the prior termination notice, notably Dr. Patel’s June 2002 evaluation form, as well as Dr. Pujara’s assessment. Although MetLife acknowledged Dr. Patel’s letter of February 2003, in which he clarified his position regarding Glenn’s ability to return to work, MetLife concluded that “documentation on file does not support a disability that would prevent Ms. Glenn from performing any occupation, as defined in the plan.” As before, however, no mention was made of Dr. Patel’s July 2002 letter, in which he rendered essentially the same opinion as in his letter of February 12, 2003.
Having exhausted her administrative remedies under the plan, Glenn brought a civil action pursuant to Section 502(a) of the Employee Retirement Income Security Act of 1974, as amended, 29 U.S.C. § 1132 (ERISA). The district court denied her motion for judgment on the administrative record and, instead, granted MetLife’s cross-motion. This appeal followed.
DISCUSSION
A. Standard of Review
We review de novo the decision of a district court granting judgment in an ERISA disability benefit action based on an administrative record, see Wilkins v. Baptist Healthcare Sys., Inc., 150 F.3d 609, 613 (6th Cir.1998), and apply the same legal standard as did the district court. *666See id.; Whitaker v. Hartford Life & Accident Ins. Co., 404 F.3d 947, 949 (6th Cir.2005). In this case, the district court appropriately reviewed the record under the “arbitrary and capricious” standard, because the plan at issue granted the plan administrator discretionary authority to interpret the terms of the plan and to determine benefits. See Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 111-15, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989); McDonald v. Western-Southern Life Ins. Co., 347 F.3d 161, 168 (6th Cir.2003). Indeed, the plaintiff conceded that review for arbitrariness was the correct standard of review here.
Under that standard, we will uphold the administrator’s decision “if it is the result of a deliberate, principled reasoning process and if it is supported by substantial evidence.” Baker v. United Mine Workers of Am. Health & Ret. Funds, 929 F.2d 1140, 1144 (6th Cir.1991). Despite this deferential standard, however, our review is no mere formality. “The arbitrary-and-eaprieious standard ... does not require us merely to rubber stamp the administrator’s decision.” Jones v. Metro. Life Ins. Co., 385 F.3d 654, 661 (6th Cir.2004) (citation omitted). Instead, we are required to review “the quality and quantity of the medical evidence and the opinions on both sides of the issues.” McDonald, 347 F.3d at 172.
Moreover, as discussed below, we are entitled to take into account the existence of a conflict of interest that results when, as in this case, the plan administrator who decides whether an employee is eligible for benefits is also obligated to pay those benefits and to factor in the plan administrator’s failure to give consideration to the Social Security Administration’s determination that Glenn was totally disabled.
B. Conflict of Interest
In the instant case, MetLife is authorized both to decide whether an employee is eligible for benefits and to pay those benefits. This dual function creates an apparent conflict of interest. See Darland v. Fortis Benefits Ins. Co., 317 F.3d 516, 527 (6th Cir.2003), overruled on other grounds by Black & Decker Disability Plan v. Nord, 538 U.S. 822, 123 S.Ct. 1965, 155 L.Ed.2d 1034 (2003) (holding that plan administrators do not owe special deference to the opinion of a treating physician). In discussing the applicable standard of review, the district court identified this conflict of interest as a relevant factor in determining whether an abuse of discretion had taken place. See Glenn v. Metropolitan Life Ins. Co., 2005 WL 1364625 at *4 (S.D.Ohio, June 8, 2005) (citing Firestone Tire & Rubber, 489 U.S. at 115, 109 S.Ct. 948 (“[I]f a benefit plan gives discretion to an administrator or fiduciary who is operating under a conflict of interest, that conflict must be weighed as a ‘faeto[r] in determining whether there is an abuse of discretion.’ ”) (internal citation omitted)). However, the court’s analysis of the plan administrator’s basis for terminating benefits does not include any discussion of the role that MetLife’s conflict of interest may have played in its decision nor appear to give that conflict any weight. It appears to us, as a result, that this factor did not receive appropriate consideration by the district court.
C. Social Security Disability Determination
There is yet another factor that the district court appears to have given inadequate consideration. The plaintiff contends that the plan administrator was arbitrary in failing to consider the award of disability benefits that she secured from the Social Security Administration. The *667record reflects that Metlife notified her that payments for long-term disability were subject to a discount for amounts received from other sources and steered her to a law firm specializing in securing disability benefits from the Social Security Administration. After the firm secured an award of benefits for her based on a claim of total disability, MetLife deducted the amount of those government benefits from the disability payments that it was obliged to pay and demanded a refund from Glenn in the amount of $13,500. And, yet, in making the decision to terminate payments under the MetLife policy, the plan administrator gave no weight whatever to the Social Security Administration’s determination of total disability. Hence, Glenn posits that MetLife took blatantly inconsistent positions — relying on the finding by the Social Security administrative law judge that she was totally disabled but contending that she was capable of performing sedentary work in denying her ERISA benefits.
The district court recognized this inconsistency but declined to invoke the doctrine of estoppel. Although we conclude that this ruling was technically correct, the fact that MetLife and the Social Security Administration reached contrary conclusions regarding Glenn’s disability status has two ramifications for this appeal. The first stems from the fact that MetLife assisted Glenn in obtaining Social Security benefits and reaped a financial benefit of its own when that assistance was successful. The second issue relates to the fact that, in denying Glenn continuation of her long-term benefits, MetLife failed to address Social Security’s contrary determination of Glenn’s status. It is obvious that both factors are relevant in determining whether MetLife’s decision was arbitrary and capricious.
The courts have recognized that a disability determination by the Social Security Administration is relevant in an action to determine the arbitrariness of a decision to terminate benefits under an ERISA plan. In Ladd v. ITT Corp., 148 F.3d 753, 755-56 (7th Cir.1998), for example, the Seventh Circuit overturned a plan administrator’s denial of disability benefits after finding that the insurance company had encouraged and assisted the plaintiff in applying for Social Security benefits, which were granted after an administrative law judge found that the plaintiff was totally disabled. Writing for the court, Judge Posner noted that the concept of judicial estoppel was not technically applicable but, nevertheless, he concluded that the inconsistency in litigation positions had to be factored into a review of the plan administrator’s determination for arbitrariness:
The grant of social security disability benefits ... brings the case within the penumbra of the doctrine of judicial es-toppel — that if a party wins a suit on one ground, it can’t turn around and in further litigation with the same opponent repudiate the ground in order to win a further victory. The doctrine is technically not applicable here, because Met-Life and ITT, the defendants in this suit, were not parties to the proceeding before the Social Security Administration. Yet they “prevailed” there in a practical sense because the grant of social security benefits to Ladd reduced the amount of her claim against the employee welfare plan. If we reflect on the purpose of the doctrine, which is to reduce fraud in the legal process by forcing a modicum of consistency on a repeating litigant, we see that its spirit is applicable here. To lighten the cost to the employee welfare plan of Ladd’s disability, the defendants encouraged and supported her effort to demonstrate total disability to the Social Security Administration, going so far as to provide her with legal ■representation. To further lighten that *668cost, it then turned around and denied that Ladd was totally disabled, even though her condition had meanwhile deteriorated. In effect, having won once the defendants repudiated the basis of their first victory in order to win a second victory. This sequence casts additional doubt on the adequacy of their evaluation of Ladd’s claim, even if it does not provide an independent basis for rejecting that evaluation.
Id. at 756 (citations omitted).
We adopted the Seventh Circuit’s “penumbra” rationale in Darland v. Fortis Benefits Insurance Co., a case in which the ERISA plan administrator had, as here, “requested that Darland apply for Social Security disability benefits so as to reduce the amount of monthly disability payments that it paid Darland under the plan” and then demanded reimbursement for overpayment of insurance benefits at the same time it sought to terminate Darland’s benefits on the ground that he was no longer disabled. Darland, 317 F.3d at 530. We concluded that “a plan administrator’s decision denying disability benefits where the Social Security Administration has determined that the applicant was totally disabled” can be considered arbitrary and capricious, id. at 529, especially where “it is plainly evident that the Social Security standard for a disability determination is much more stringent than that required by [the defendant’s] insurance policy.” Id. at 530. The latter observation not only pertains in this case to the language of the policy1 but is all the more relevant in the wake of the Supreme Court’s subsequent ruling that the special deference extended to the opinion of a treating physician in a Social Security dispute is not applicable in ERISA cases. See Black & Decker Disability Plan, 538 U.S. at 833-34, 123 S.Ct. 1965.
The district court in this case recognized that MetLife had encouraged and assisted Wanda Glenn in obtaining Social Security disability benefits, just as MetLife had in Ladd, but found Glenn’s case to be distinguishable in two respects. First, the district court noted, correctly, that the record before the administrative law judge did not include the report, signed on the same day that the administrative judge’s decision was made, in which Dr. Patel answered “yes” to MetLife’s question, “Do you believe Ms. Glenn is able to work in a sedentary physical exertion level occupation?” However, the relative significance of this omission in the record is undercut by the existence of Dr. Patel’s subsequent clarification of July 22, 2002, in which he indicated that he had never considered his patient capable of resuming full-time work, and by the fact that there is no indication in the record that Glenn’s government benefits have ever been discontinued. Second, the district court found, incorrectly, that it was unclear whether or not MetLife had actually benefitted from the plaintiffs receipt of Social Security benefits. We find it abundantly clear on the record that Glenn reimbursed MetLife for overpayment of benefits, based on her receipt of Social Security benefits.
*669Having benefitted financially from the government’s determination that Glenn was totally disabled, MetLife obviously should have given appropriate weight to that determination. As we held in Calvert v. Firstar Finance, Inc., 409 F.3d 286, 295 (6th Cir.2005), a case with several parallels to this one, an ERISA plan administrator’s failure to address the Social Security Administration’s finding that the claimant was “totally disabled” is yet another factor that can render the denial of further long-term disability benefits arbitrary and capricious. In Calvert, as here, the plan administrator relied almost exclusively on the recommendation of a physician, hired by the administrator, who conducted a file review but not a physical examination of the claimant. See id. We held that such a review was clearly inadequate, given that in reaching the determination to deny benefits, the plan administrator not only overlooked surgical reports, X-rays, and CT scans that documented the claimant’s disability, but also failed to mention the Social Security Administration’s determination, “not ... even to discount or disagree with it, which indicates that he may not even have been aware of it.” Id. at 296.
Here, there is no possibility that the plan administrator failed to discuss the Social Security award for lack of awareness. And, “the S[ocial] S[ecurity] A[dministration] determination, though certainly not binding, is far from meaningless.” Id. at 294. That MetLife apparently failed to consider the Social Security Administration’s finding of disability in reaching its own determination of disability does not render the decision arbitrary per se, but it is obviously a significant factor to be considered upon review.
D. The Medical Evidence
Even more perplexing than the plan administrator’s failure to consider the award of Social Security benefits is the persistent failure to give any weight to Dr. Patel’s letters of July 22, 2002, and February 12, 2003, in which he clearly stated that he did not believe Glenn was capable of returning to work, sedentary or otherwise. This omission stands in stark contrast to the heavy reliance MetLife placed in its brief on the “physical capacity assessment” form that MetLife provided to Dr. Patel and that he filled in and signed on March 13, 2002. On the form, Dr. Patel indicated that his patient could sit for eight hours a day, stand for four hours, and walk for two hours. He also checked yes when asked, “Do you believe Ms. Glenn is able to work in a sedentary physical exertion level occupation?” That information is so inconsistent with other medical evidence and detailed reports supplied by Dr. Patel over a period of three years that it can best be described as aberrational.
Nevertheless, MetLife cannot be faulted for its initial reliance on the opinion that Glenn could “work at a sedentary physical exertion level occupation” in consulting a “transferable skills analyst,” who then compiled a list of clerical jobs (account information clerk, attendance clerk, classified ad clerk) that would be available to someone with the physical limitations listed on the physical assessment form. There is no indication in the record, however, that the transferable skills analyst considered the medical evidence in the record or Dr. Patel’s reasons for concluding that his patient should not return to full-time work of any kind. Nevertheless, the plan administrator, relying only on the vocational assessment, the information on the MetLife form supplied by Dr. Patel on March 13, 2002, and certain other technical information that seems to have been almost randomly selected from his reports (one paragraph reads in its entirety, without further explanation: ‘Tour June 18, 2002 examination showed an ejection fraction of 35 to 40%.”), declined to grant further long-term benefits in a letter dated *670August 28, 2002. Nowhere in the notice of denial does the plan administrator mention the decision of the Social Security Administration finding that Glenn was totally disabled.
Perhaps most curious is the plan administrator’s treatment of Dr. Patel’s then-most recent assessments of his patient’s condition. The plan administrator emphasized the content of yet another “physical capacities evaluation” dated June 12, 2002, that indicated by check-off information consistent with that on the March 13 form (“sit for eight hours, stand for four hours, walk for 2 to 4 hours,” etc.). But the June 12 form does not recommend that Glenn return to work, nor does it release her to return to work. Furthermore, the notice from the plan administrator made no mention whatever of the letter Dr. Patel sent on July 22, 2002, which, as described above, noted that Glenn “continues to have significant difficulty with exertional shortness of breath on any kind of moderate exertion” and “significant difficulty in returning to even any kind of sedentary job because any kind of psychologic stress at work causes significant problems with her cardiovascular condition and she decom-pensates fast.” Dr. Patel concluded:
At. the present time, I do not believe Wanda should be forced to return to any kind of even sedentary work particularly because it is the psychologic stress of work that really exacerbates her cardiovascular condition and symptomatololgy. The .patient basically should be considered completely disabled from her dilated cardiomyopathy as well as history of ventricular tachycardia.
This letter to MetLife is consistent with a follow-up report a month earlier that Dr. Patel sent to Glenn’s internist, Dr. Rhee, in which he concluded: “From my standpoint, again, considering her cardiomyopa-thy, I do not believe she will handle any kind of stress well at her work and she would be better off being on disability.”
Based on Dr. Patel’s assessment of her condition, Glenn appealed MetLife’s decision to terminate her benefits, and the decision was reviewed internally by Met-Life. In a letter dated May 20, 2003, the plan administrator once again denied benefits, this time based on the evaluation form signed by Dr. Patel in June 2002, the transferable skills analysis from July 2002, and two items that had not been part of the earlier review: a letter from Dr. Patel dated February 12, 2003, and the report from the outside consultant, Dr. Pujara, dated May 2, 2003. Once again, there is no mention of Dr. Patel’s July 2002 letter stating unequivocally that his patient was not able to return to work.
The appeal decision does indicate that Dr. Patel’s letter “noted that his previous reports indicated that Ms Glenn was fit for sedentary work” but that “based on her clinical condition and symptomatology, Dr. Patel never felt that Ms. Glenn could return to full-time employment” and that she “should be considered completely disabled from her cardiomyopathy and associated symptoms from her LV dysfunction.” On the other hand, there is no explanation in the notice for the failure to give any weight to this report.
The summary of Dr. Pujara’s report reiterated the information that he found in the medical files submitted by MetLife— Dr. Pujara did not conduct an examination of Glenn but merely undertook a file review — and concluded: “The consultant noted that based on the medical information Ms. Glenn has achieved a relatively stable cardiac status.” What the summary omits is any indication that the consultant thought that she could return to work. In fact, Dr. Pujara noted that given the information in her file, “the patient certainly is not able to do any exertional physical ac*671tivity” and, with regard to the possibility of “sedentary activity,” concluded that “[i]f the job environment entails a significant degree of emotional stress, and the patient is not able to cope with that, then certainly permanent disability can be considered.” Although it appears that MetLife forwarded the June 2002 evaluation form . to Dr .Pujara, there is no indication that he was provided with copies of Dr. Patel’s letters of July 2002 and February 2003, in which Dr. Patel concluded that Glenn was not able to undertake sedentary work because any degree of stress caused her medical condition to “decompensate.”2
The Plan provides that “[wjhile a claim is pending, we, at our expense, have the right to have you examined by doctors of our choice when and as often as we reasonably choose.” MetLife instead opted to have a file review, a decision that does not render its denial of benefits arbitrary per se. However, it is yet another factor to be considered in the overall assessment of its decision-making process. See Calvert, 409 F.3d at 295. Moreover, as noted above, it appears that the physician charged with conducting the independent file review was not provided with Dr. Patel’s letters of July 2002 and February 2003.
As a result, Dr. Patel is the only physician to have personally treated and observed Glenn over the entire course of her dispute with MetLife. As noted above, the plan administrator need not accord special deference to the opinion of a treating physician. By the same token, it may not arbitrarily repudiate or refuse to consider the opinions of a treating physician. See Black & Decker Disability Plan, 538 U.S. at 834, 123 S.Ct. 1965. In its final rejection of Glenn’s claim, however, MetLife did not indicate that it had considered Dr. Patel’s letter of February 12, 2003, nor did it provide a reason for rejecting it.3 It also did not indicate whether or not it had considered Dr. Patel’s letter of July 22, 2002, which was consistent with the February 12 letter advising that Glenn could not return to work.
In the absence of an explanation from MetLife, the district court provided its own, noting that because Dr. Patel “expressly acknowledged that he had stated that plaintiff was fit for sedentary work ... MetLife did not abuse its discretion by viewing Dr. Patel’s recantation with skepticism.” In support of this conclusion, the district court invoked United States v. Willis, 257 F.3d 636, 645-46 (6th Cir.2001), for the proposition that affidavits recanting trial testimony are viewed with “extreme suspicion.” The court’s reliance on Willis is misplaced, however, because that case addresses the full recantation of a witness’s testimony in a criminal trial, a situation which is hardly equivalent to Dr. Patel’s follow-up letters, which were consistent with his earlier assessments of Glenn’s ability to work and which are more aptly described as clarifications, rather than recantations. Indeed, the concept of recantation has little application outside the unique context of the criminal case, and any presumption regarding the credibility of inconsistent testimony should be limited to that context.
We conclude that the only fair inference from the record would actually *672undercut Dr. Patel’s check-offs and his brief assessment on the functional capacity form dated March 13, 2002, because that information was in direct conflict both with his earlier assessments and with every detailed written explanation that he gave concerning Glenn’s disability.4 Clearly, the plan administrator rejected Dr. Patel’s more detailed reports while crediting the check-off forms. However, MetLife offered no explanation for its resolution of the conflict or, for that matter, whether it was given any consideration at all. The omission is critical, because the failure to consider evidence that is offered after an initial denial of benefits renders a final denial of benefits arbitrary and capricious. See Williams v. Int’l Paper Co., 227 F.3d 706, 712-14 (6th Cir.2000). As for the district court’s attempt to reconcile the conflicts in the record, we point out that the court’s role is to review the basis for the decision that was actually made by the plan administrator, not to provide an adequate basis where none was offered.
The importance of a resolution of the apparent conflict between Dr. Patel’s responses on the check-off form and his later reports is even more pronounced given that Dr. Patel consistently deemed psychological stress to be a critical factor in Glenn’s treatment and the deciding factor in assessing her ability to return to work. On April 7, 2000, Dr. Patel wrote that Glenn’s “main problem now is stress at work.” On April 30, 2000, he concluded that Glenn “cannot return to any kind of job that would require any significant physical or psychological stress.” On July 29, 2000, after Glenn had been on leave from work for approximately three months, Dr. Patel observed that “some of her improvement is from the reduction of stress and strain of work.” On June 12, 2002, Dr. Patel also indicated that Glenn’s functional limits included “no emotional stress.” The law firm retained by MetLife to represent Glenn before the Social Security Administration certainly relied on Dr. Patel’s analysis of the impact of stress on Glenn’s condition, even if MetLife did not. Contending before the administrative law judge assigned to her claim that “[i]n addition to findings of impaired myocardial function, Dr. Patel indicated that the severity of Ms. Glenn’s symptoms and physical condition were aggravated by emotional stress,” the firm pointed out that “this opinion was rendered by a longtime treating source who had referred the claimant for implantation of a new cardiac device, referred the claimant for transplantation assessment, and provided consistent care while the claimant’s left ventricular functioning deteriorated.”
Despite the consistent and repeated references by Dr. Patel to stress as a factor in Glenn’s condition, MetLife concluded in its final denial of benefits to Glenn, “There is no supportive medical documentation of the exacerbation of your cardiac condition and symptomology, due to the subjective complaints of work-related stress.” But, this conclusion unjustifiably implies that Dr. Patel’s observations *673and notations regarding stress do not constitute “supportive medical documentation.” Although it is not clear from the record what sort of documentation Met-Life would have found sufficient to establish the negative effect of stress on Glenn’s medical condition, the plan itself does not restrict the type of evidence that may be used to demonstrate total disability. Rather, the plan provides that “[t]o qualify for benefits under the Plan, Total Disability must be supported by current medical documentation” and the claimant “must be under the regular care of a qualified physician under a course of treatment appropriate for the disability.” We conclude that the plan administrator’s rejection of Dr. Patel’s assessment, under the standard set out in the plan, was, in fact, arbitrary.
That conclusion is supported by our decisions in similar ERISA cases. In Evans v. Unumprovident Corp., 434 F.3d 866 (6th Cir.2006), for example, a claimant applied for long-term disability benefits on the basis that her frequent , seizures rendered her incapable of work. Her treating physician opined that the stress associated with her administrative duties was responsible for the severity and frequency of the episodes. See id. at 870. While on medical leave, the claimant saw an improvement in her condition. Her treating physician determined that it would be in the claimant’s best interest not to return to work. See id. at 871. The insurer denied the claim,, finding it unreasonable to speculate that a return to work would exacerbate the claimant’s condition. See id. In reaching this decision, the insurer relied heavily on an independent physician’s review, which declared that the impact of stress on the patient’s condition was entirely self-reported and had not been corroborated by medical studies. See id. at 875. We found that the insurer’s decision was arbitrary and capricious. “[Djespite the unwavering expert medical opinion of plaintiffs treating physician ... that stress is probably the most important seizure-provoking factor in all patients, and not just plaintiff, and that plaintiffs high-stress position would exacerbate her condition, defendant nonetheless unreasonably discounted stress as merely a ‘prophylactic’ factor that should be accorded minimal, if any, weight in its determination of disability.” Id. at 879.
In the present case, although Dr. Patel’s opinion of Glenn’s ability to perform sedentary work has not been consistently expressed, he likewise has been “unwavering” in his opinion that stress is an important factor in her condition. As in Evans, the plan in this case does not say that prophylactic determinations are not relevant to the decision, nor does it say that self-reported or “subjective” factors should be accorded less significance than other indicators. And, although it is not clear that stress is the most important factor in all cases of cardiomyopathy, as it apparently is with epilepsy, it is unreasonable for MetLife to have dismissed stress as an improperly documented, subjective, and irrelevant factor in its disability determination.
MetLife cites Eastover Mining Co. v. Williams, 338 F.3d 501 (6th Cir.2003), as justification for discrediting Dr. Patel’s opinion that Glenn was not capable even of sedentary employment. In • reviewing a benefits determination in a black lung case, the Eastover Mining court observed that “treating physicians may have strong pro-claimant biases and lack the expertise held by non-treating doctors.” Id. at 510 (citing Black & Decker Disability Plan, 538 U.S. at 832, 123 S.Ct. 1965). It is true that treating physicians, “in a close case, may favor a finding of ‘disabled’,” 538 U.S. at 832, 123 S.Ct. 1965, and that a presumption of deference to treating physicians makes little sense when “the relationship between the claimant and the treating physician has been of short duration, or. when *674a specialist engaged by the plan has expertise the treating physician lacks.” Id. Moreover, although Black & Decker Disability Plan makes clear that MetLife bears no heightened burden of explanation when making a disability determination that runs contrary to a treating physician’s opinion, see 538 U.S. at 831, 123 S.Ct. 1965, MetLife’s decision with respect to Glenn’s claim does not appear to be the result of a principled reasoning process, given that Dr. Patel had expertise in treating Glenn’s disability and had a long-term relationship with her.
Finally, we note that a claimant’s potential ability to “return to work under certain limited circumstances,” McDonald, 347 F.3d at 170, should not be taken to satisfy the requirements of the Sears ERISA plan when those circumstances are, as here, very limited and would not permit Wanda Glenn, pursuant to the language of the plan, to perform the duties of gainful work for which she is reasonably qualified, “taking into consideration her training, education, experience, and past earning.” As we observed in McDonald:
The mere possibility that a participant in an ERISA plan might be able to return to some type of gainful employment, in light of overwhelming evidence to the contrary, is an insufficient basis upon which to support a plan administrator’s decision to deny that participant’s claim for [long-term disability] benefits.
Id. 170-71.
CONCLUSION
Even when it is reviewed under the highly deferential standard applicable in this case, we conclude that the plan administrator’s determination to deny benefits to Glenn cannot be sustained.
[The] obligation under ERISA to review the administrative record in order to determine whether the plan administrator acted arbitrarily and capriciously in making ERISA benefits determinations ... inherently includes some review of the quality and quantity of the medical evidence and the opinions on both sides of the issues. Otherwise, courts would be rendered to nothing more than rubber stamps for any plan administrator’s decision as long as the plan [administrator] was able to find a single piece of evidence — no matter how obscure or untrustworthy — to support a denial of a claim for ERISA benefits.
McDonald, 347 F.3d at 172.
For the reasons set out above, we conclude that MetLife’s decision to deny long-term benefits in this case was not the product of a principled and deliberative reasoning process. MetLife acted under a conflict of interest and also in unacknowledged conflict with the determination of disability by the Social Security Administration. In denying benefits, it offered no explanation for crediting a brief form filled out by Dr. Patel while overlooking his detailed reports. This inappropriately selective consideration of Glenn’s medical record was compounded by the fact that the occupational skills analyst and the independent medical consultant were apparently not provided with full information from Dr. Patel on which to base their conclusions. Moreover, there was no adequate basis for the plan administrator’s decision not to factor in one of the major considerations in Glenn’s pathology, that of the role that stress played in aggravating her condition and, in the language of the MetLife policy, in preventing her return to “gainful work or service for which [she is] reasonably qualified taking into consideration [her] training, education, experience, and past earning.” Taken together, these factors reflect a decision by MetLife that can only be described as arbitrary and capricious.
*675We therefore REVERSE the judgment of the district court and REMAND the case with directions to reinstate Glenn’s long-term disability benefits, retroactive to the date on which they were terminated,5 and for such other relief as the district court finds appropriate in view of our ruling in this case.
. After 24 months, the plan awards disability benefits to an individual who is "completely and continuously unable to perform the duties of any gainful work or service for which he is reasonably qualified, taking into consideration his training, education, experience, and past earning.” Social Security disability insurance benefits are awarded only if the individual's impairment is of such severity that he "cannot, considering his age, education, and work experience, engage in any other kind of substantial gainful work which exists in the national economy, regardless of whether such work exists in the immediate area in which he lives.” 42 U.S.C. § 423(d)(2)(A). The Social Security determination does not take into account past earning or location of the applicant.
. At oral argument of this case, the defendants conceded that Glenn had not been offered a part-time position at Sears and that she would have no chance of receiving benefits under the Sears ERISA plan if she went to work at for another employer on the trial basis that Dr. Pujara suggested.
. MetLife maintains in its brief on appeal that the denial notice “specifically discussed” Dr. Patel's letter of February 12, 2003, we find that the word "discussed” is somewhat misleading; "mentioned” would be a more accurate choice.
. This case is not the first in which we have been called upon to review MetLife’s over-reliance on an aberrational report from a physician. In Spangler v. Lockheed Martin Energy Systems, Inc., a case with several interesting parallels to this one, the plan administrator denied benefits based on the responses of the plaintiff's treating physician on the same form that Dr. Patel filled out in this case, a physical capacities evaluation that we described as "somewhat aberrant,” given that "all of [the physician's] prior reports and statements clearly indicate[d] that Spangler [wa]s not able to perform any work.” 313 F.3d 356, 362 (6th Cir.2002). We concluded that "MetLife, as Spangler contends, 'cherry-picked' her file in hopes of obtaining a favorable report from the vocational consultant as to Spangler’s ability to work.” Id.
. See Kalish v. Liberty Mutual / Liberty Life Assurance Co., 419 F.3d 501 (6th Cir.2005) (remanding for retroactive reinstatement of benefits where plan administrator arbitrarily and capriciously refused to consider evidence of claimant's depression as factor in disability); Williams v. Int'l Paper Co., 227 F.3d 706 (6th Cir.2000) (remanding for retroactive reinstatement of benefits where plan administrator had opportunity to review all evidence and did so in a manner that was arbitrary and capricious).