OPINION

RONALD LEE GILMAN, Circuit Judge.

Richard Kalish, who suffers from a heart condition and depression, began receiving disability benefits under a group insurance plan (the Plan) issued to his employer by Liberty Mutual/Liberty Life Assurance Company of Boston. The benefits ended, however, when Liberty determined that Kalish was no longer disabled under the terms of the Plan. After exhausting his administrative appeals, Kalish brought this action to recover benefits from Liberty pursuant to 29 U.S.C. § 1132(a)(1)(B) of the Employee Retirement Income Security Act (ERISA). The district court granted judgment on the administrative record in favor of Liberty.

Kalish argues on appeal that the district court erred in holding that (1) the arbitrary and capricious standard of judicial review applies to this case despite Liberty’s conflict of interest as both insurer and plan administrator, (2) Liberty could deny benefits because Kalish was capable of sedentary work, (3) Liberty’s decision to deny benefits for Kalish’s heart condition was not arbitrary and capricious, and (4) Liberty was not required to consider evidence of Kalish’s alleged depression. For the reasons set forth below, we REVERSE the judgment of the district court affirming Liberty’s denial of benefits and REMAND this case for entry of an order requiring Liberty to award benefits plus interest from the date on which Kalish’s benefit payments ceased.

I. BACKGROUND

A. Factual background

Kalish worked for Deluxe Video Services, Inc., a company affiliated with Rank America, Inc., before his health problems began. As an employee, he was covered by Rank America’s long-term disability insurance policy. Liberty is both the insurer and administrator of the Plan. Under the terms of the Plan, Kalish was to receive monthly disability benefits equal to two-thirds of his monthly income if he became “unable to perform all of the material and substantial duties of his occupation on an Active Employment basis because of an Injury or Sickness.”

Kalish was the Director of National Transportation for Deluxe Video. He was responsible for ensuring that newly released video and DVD movies arrived at nationwide distribution centers on a same-day basis. According to the form that Kalish’s supervisor at Deluxe Video completed and returned to Liberty, Kalish’s job duties required him to “(1) direct all aspects [of] transportation operations, (2) handle all carrier negotiations, (3) travel to customer/earrier locations, [and] (4) direct [an] on-site staff of 50 employees.” His supervisor further described Kalish’s job as a “high stress position with many deadlines” and significant “vendor/customer contact.” In addition to four hours of sitting, Kalish’s position required him to spend one hour standing, one hour walking, and one hour reaching at shoulder level on a daily basis.

Kalish suffered two separate heart attacks in May and July of 2001. Immediately following the first heart attack, Kalish filed a claim with Liberty. He received short-term disability benefits for 180 days, the limit for such benefits under the Plan. Then, in December of 2001, Kalish filed another claim with Liberty, this time seeking long-term disability benefits.

In order to process this claim, Liberty requested additional documentation from *504 Kalish’s treating physicians. Among the physicians were Dr. Eisenberg, an internist, and Dr. Raminick, a gasteroenterologist, both of whom had been treating Kalish before his heart problems began and continued seeing him after his heart attacks. Drs. Eisenberg and Raminick provided Liberty with copies of their medical records, which document Kalish’s problems with hepatitis, diabetes, hypertension, peptic ulcers, and sleep apnea.

The bulk of the medical reports considered by Liberty, however, were generated by Dr. Rasak, the cardiologist who treated Kalish for his heart condition. In addition to providing Liberty with copies of Kalish’s medical records, Dr. Rasak sent Liberty a letter in January of 2002 that stated: “At this point, I do not see Mr. Kalish functioning well in a workplace and feel that he should be completely and totally disabled.” Dr. Rasak also opined that “never can [Kalish] work an 8-hour day” on account of his “severe coronary artery disease.” In support of this opinion, Dr. Rasak repeated that “Kalish is severely debilitated as a result of his extensive damage to his heart,” and that, “[i]f it were not for the patient’s hepatitis C, the patient may be a candidate for a cardiac transplant.” Dr. Rasak further observed that Kalish suffered from “hypertension, diabetes, hepatitis C, peptic ulcer disease, sleep apnea and hypercholesterolemia.”

As part of its review process, Liberty sent an investigator to Kalish’s home in January of 2002 to conduct an in-person interview and obtain a tape-recorded statement from Kalish. Liberty’s investigator observed that “the claimant appeared tired, but was very cooperative.” He also reported that Kalish “appears to have a limited lifestyle, due to tiring easily.” In reaching his final conclusion that Kalish was “highly credible,” the Liberty investigator stated that Kalish “appears to try to keep his spirits up, but came across as missing his job.”

A registered nurse employed by Liberty reviewed the medical reports provided by Kalish’s treating physicians. In her evaluation, she noted what she considered to be inconsistencies between the statements provided by Dr. Rasak regarding Kalish’s ability to work and the results of the medical tests performed on Kalish. She recommended that Kalish’s file receive peer review from an independent physician specializing in cardiology.

Liberty then retained Peer Review Analysis, an independent company, to evaluate Kalish’s file. The cardiologist conducting the review, Dr. Conrad, spoke with Dr. Rasak by telephone. During this discussion, Dr. Rasak expressed the opinion that “it would not be feasible for the patient to perform work involving any significant exertion,” but he conceded that “it might be possible for the patient to perform work of a sedentary nature.” On the other hand, Dr. Rasak noted that “depression in association with other medical problems might also impair the employee’s ability to function effectively in his job.”

Dr. Conrad, in performing his evaluation, examined Kalish’s medical records. He concluded that Kalish had “moderate to severe impairment of left ventricular function,” and that this “dysfunction is associated with an increased risk of recurrent congestive heart failure and increased mortality risk.” But Dr. Conrad found that “there [wa]s no clear evidence ... that work cessation would necessarily reduce the risk.” He further observed that Kalish “has shown significant clinical improvement, ... is able to exercise on a regular basis, ... [and] testing has shown good functional capacity.” Dr. Conrad concluded his review with a finding that the “records do not document a need for restrictions or limitations that would nec *505 essarily preclude the employee from performing the duties of his job.”

Liberty forwarded a copy of Dr. Conrad’s peer review to Dr. Rasak on February 27, 2002, along with a letter inviting him to respond to the findings of Dr. Conrad. Dr. Rasak was given one week to “detail the specific areas with which [he] disagree[d] and [to] provide medical documentation to support [his] position.” Liberty cautioned that, “in the absence of a response, [it would] conclude that [he] agree[d] with the peer review.” Dr. Rasak did not respond to this letter.

On March 15, 2002, Liberty denied Kalish’s claim for long-term disability benefits as of October 25, 2001, the date of a stress test taken by Kalish that Dr. Conrad had relied upon in concluding that Kalish was no longer disabled. Liberty indicated that it would keep his claim open for an additional two weeks, until March 28, 2002, to permit Kalish to respond to the denial. No documents were received from Kalish or his physicians before the March 28 deadline. On May 3, 2002, however, Dr. Rasak sent Liberty a copy of a report, dated April 1, noting that Kalish was suffering from depression.

Kalish wrote a letter to Liberty on May 15, 2002, to formally appeal the denial of his claim. He also advised Liberty that he would be submitting additional medical reports from Dr. Rasak and from a clinical psychologist, Dr. Last. On May 28, Liberty received a letter from Dr. Last, stating that he had been treating Kalish for depression and anxiety on a weekly basis since April 25. Dr. Last expressed his opinion that “[d]ue to the depression and anxiety, and medical problems, [Kalish’s] ability to work currently is significantly compromised.”

Liberty wrote Kalish a letter on July 17, 2002, informing him that a “consulting physician” would be reviewing his “recently received medical information” and that the process could require several weeks. But Kalish’s file was never reviewed by an independent physician. It was instead evaluated by a nurse employed by Liberty. She concluded that the “[n]ew medical records document no new objective or clinical change in his condition except for some symptoms of depression. However, the psychologist supplied no office notes and symptoms were not present or treated until 6 mos after the denial.” Based upon this review, Liberty upheld its denial of benefits to Kalish in a letter dated July 30, 2002.

B. Procedural background

Kalish filed suit against Liberty in a Michigan state court in December of 2002. Liberty removed the action the following month to the United States District Court for the Eastern District of Michigan, where, in accordance with a stipulated order, all of Kalish’s claims were dismissed with the exception of his claim for benefits under ERISA. The parties then filed cross-motions for judgment on the administrative record. On May 17, 2004, immediately following oral argument on the motions, the district court delivered an oral opinion granting Liberty’s motion for entry of judgment and denying Kalish’s motion. The district court’s order, which had a transcript of the oral opinion attached, was filed on June 25, 2004. This timely appeal followed.

II. ANALYSIS

A. Liberty’s alleged conflict of interest does not make application of the arbitrary and capricious standard of review inappropriate

“[A] denial of benefits challenged under § 1132(a)(1)(B) is to be reviewed under a de novo standard unless the bene *506 fit plan gives the administrator or fiduciary discretionary authority to determine eligibility for benefits or to construe the terms of the plan.” Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101, 115, 109 S.Ct. 948, 103 L.Ed.2d 80 (1989). “When such authority is granted, the highly deferential arbitrary and capricious standard of review is appropriate.” Borda v. Hardy, Lewis, Pollard, & Page, P.C., 138 F.3d 1062, 1066 (6th Cir.1998) (citation and quotation marks omitted). A plan administrator’s decision will not be deemed arbitrary and capricious so long as “it is possible to offer a reasoned explanation, based on the evidence, for a particular outcome.” Davis v. Ky. Fin. Cos. Ret. Plan, 887 F.2d 689, 693 (6th Cir.1989) (noting that “[t]he arbitrary and capricious standard is the least demanding form of judicial review”) (quotation marks omitted).

We “review a district court’s determination regarding the proper standard to apply in its review of a plan administrator’s decision de novo." Hoover v. Provident Life & Accident Ins. Co., 290 F.3d 801, 807 (6th Cir.2002). The district court in the present case applied the arbitrary and capricious standard of review. Kalish concedes that application of the de novo standard would be improper, but argues that “Liberty’s denial of benefits should be evaluated under a less deferential, modified-arbitrary-and-capricious standard because Liberty was operating under a conflict of interest, as it was both the insurer and an administrator of the Plan.”

We agree that Liberty’s operation of the Plan as both the insurer and the administrator creates a conflict of interest. See Killian v. Healthsource Provident Adm'rs, Inc., 152 F.3d 514, 521 (6th Cir.1998) (concluding that “there is an actual, readily apparent conflict” where “as administrator, it interprets the plan, deciding what expenses are covered, and as issuer of the policy, it ultimately pays those expenses”) (quotation marks omitted). The solution to this conflict, however, is not to alter the degree of deference granted under the arbitrary and capricious standard. See Borda, 138 F.3d at 1069 (holding that a conflict of interest “does not require [the court] to abandon the arbitrary and capricious standard of review”) (quotation marks omitted). Rather, the Supreme Court has determined that “the conflict must be weighed as a factor in determining whether this is an abuse of discretion.” Firestone Tire & Rubber Co., 489 U.S. at 115, 109 S.Ct. 948 (quotation marks omitted). The district court, therefore, did not err in evaluating Liberty’s denial of benefits under the arbitrary and capricious standard of review.

B. The district court erred in ruling that Liberty could deny benefits on the basis that Kalish was capable of sedentary work

In holding that Liberty’s denial of benefits to Kalish was not arbitrary and capricious, the district court reasoned that Kalish’s “treating physician ... said at one point that the Plaintiff may be capable of sedentary work.” This finding, Kalish argues, was an improper basis for the district court’s conclusion. We agree.

Although the fact that a claimant is able to engage in sedentary work is an appropriate consideration in some cases, the Plan language in the present case explicitly stated that a participant is disabled so long as “he is unable to perform all of the material and substantial duties of his occupation.” (Emphasis added.) Liberty has acknowledged that Kalish’s position as Director of National Transportation required him to stand for one hour a day, walk for one hour a day, and reach at shoulder level for one hour a day. Thus, while Dr. Rasak opined that Kalish might be able to engage *507 in sedentary work, he repeatedly insisted that Kalish was incapable of working an eight-hour day and “functioning well in a work environment because it would not be feasible for the patient to perform work involving any significant exertion.” Dr. Rasak made this determination relative to what he “understood [Kalish’s] job duties as Director of National Transportation” to be.

In McDonald v. Western-Southern Life Insurance Co., 347 F.3d 161, 170 (6th Cir.2003), the plan administrator based its decision to deny benefits on the opinion of a doctor who had “merely opined that [the claimant] might be able to return to work under certain limited circumstances.” The McDonald court concluded that “[t]he mere possibility that a participant might be able to return to some type of gainful employment ... is an insufficient basis upon which to support a plan administrator’s decision to deny that participant’s claim.” Id. at 171-72. Similarly, in the present case, the fact that Kalish might be capable of sedentary work cannot be a rational basis for finding that he was not disabled, given that his former occupation required him to walk, stand, and reach for several hours a day. We must therefore determine whether there are any other grounds that adequately support Liberty’s decision to deny benefits to Kalish.

C. The district court erred in sustaining Liberty’s denial of long-term benefits to Kalish on the basis of his heart condition

Kalish argues that Liberty’s denial of continuing benefits was arbitrary and capricious because Liberty relied solely on the opinion of Dr. Conrad, the cardiologist it retained to review Kalish’s file. The weight of Dr. Conrad’s evaluation should be discounted, Kalish maintains, for two reasons. First, Dr. Conrad allegedly could not provide a truly independent opinion because he was “employed by a firm that specializes in eliminating employee absences,” and “thus became part of the Liberty advocacy team.” Second, Kalish argues that Dr. Conrad’s opinion should be discounted because Dr. Conrad did not physically examine Kalish and instead reviewed only his medical files.

Kalish makes the legally valid point that, “when a plan administrator’s explanation is based on the work of a doctor in its employ, we must view the explanation with some skepticism.” Moon v. Unum Provident Corp., 405 F.3d 373, 381-82 (6th Cir.2005). But Dr. Conrad was not an employee of Liberty. Rather, he was an independent expert retained by Liberty to conduct a peer review of the work of Kalish’s treating cardiologist, Dr. Rasak, and to examine the other medical records in Kalish’s file. There is no evidence that Liberty attempted to tamper with or inappropriately influence Dr. Conrad’s evaluation, which could lead us to conclude that Liberty had acted arbitrarily and capriciously in relying upon his opinion. See McDonald, 347 F.3d at 168 (finding a denial of benefits to be arbitrary and capricious when premised on the report of an independent doctor whose initial opinion changed after “a telephone call from members of [the plan administrator]’s appeals committee”).

Even so, the Supreme Court has acknowledged “that physicians repeatedly retained by benefits plans may have an incentive to make a finding of ‘not disabled’ in order to save their employers^] money and preserve their own consulting arrangements.” Black & Decker Disability Plan v. Nord, 538 U.S. 822, 832, 123 S.Ct. 1965, 155 L.Ed.2d 1034 (2003) (citation and quotation marks omitted). This court has similarly observed that a plan administrator, in choosing the independent *508 experts who are paid to assess a claim, is operating under a conflict of interest that provides it with a “clear incentive to contract with individuals who were inclined to find in its favor that [a claimant] was not entitled to continued [disability] benefits.” Calvert v. Firstar Fin., Inc., 409 F.3d 286, 292 (6th Cir.2005) (noting that the “possible conflict of interest inherent in this situation should be taken into account as a factor in determining whether [a plan administrator’s] decision was arbitrary and capricious”) (quotation marks omitted). Thus, although “routine deference to the opinion of a claimant’s treating physician” is not warranted, we may consider whether “a consultant engaged by a plan may have an ‘incentive’ to make a finding of ‘not disabled’ ” as a factor in determining whether the plan administrator acted arbitrarily and capriciously in deciding to credit the opinion of its paid, consulting physician. See Nord, 538 U.S. at 832, 123 S.Ct. 1965.

In the present case, however, Kalish has offered only conclusory allegations of bias with regard to Dr. Conrad. He failed to present any statistical evidence to suggest that, when retained by Liberty, Dr. Conrad has consistently opined that claimants are not disabled. See id. (stating that a determination of bias “might be aided by empirical investigation”); see also Calvert, 409 F.3d at 293 n. 2 (“The Court would have a better feel for the weight to accord this conflict of interest if [the claimant] had explored the issue through discovery. While ... discovery is ... [ordinarily not] permissible in an ERISA action premised on a review of the administrative record, an exception to that rule exists where a plaintiff seeks to pursue a decision-maker’s bias.”). In the absence of such evidence, we are unable to conclude on this basis that Liberty acted arbitrarily and capriciously in deciding to credit the opinion of Dr. Conrad over that of Dr. Rasak. See Nord, 538 U.S. at 832, 123 S.Ct. 1965 (noting that “a treating physician, in a close case, may favor a finding of ‘disabled’ ”).

But Kalish presents an alternative argument for why Dr. Conrad’s opinion should be discounted. He points out that, although Dr. Conrad reviewed Kalish’s medical records, he did not physically examine Kalish. Liberty insists that this argument was rejected by the Supreme Court in Nord. But the Court’s opinion in Nord did not address the argument advanced by Kalish. The Court instead focused on the “treating physician rule” in ERISA cases, which some courts had employed to “impose a heightened burden of explanation on administrators when they reject a treating physician’s opinion.” Id. at 831, 123 S.Ct. 1965 (“Nothing in the Act itself ... suggests that plan administrators must accord special deference to the opinions of treating physicians.”). Kahsh’s argument, in contrast, is that greater deference should be accorded to the opinion of Dr. Rasak, not because he was Kalish’s treating physician, but because Dr. Rasak actually conducted physical examinations of Kalish while Dr. Conrad reviewed only Kalish’s medical records.

Whether a doctor has physically examined the claimant is indeed one factor that we may consider in determining whether a plan administrator acted arbitrarily and capriciously in giving greater weight to the opinion of its consulting physician. See Calvert, 409 F.3d at 295 (“[W]e find that the failure to conduct a physical examination ... may, in some cases, raise questions about the thoroughness and accuracy of the benefits determination.”). But “reliance on a file review does not, standing alone, require the conclusion that [a plan administrator] acted improperly.” Id.

*509 In the present case, the eight-page report produced for Liberty by Dr. Conrad contains a six-page summary of Kalish’s medical records, but it dedicates only one page to analyzing those records. This limited commentary contains little more than Dr. Conrad’s conclusory assertions to the effect that “the available records do not document a need for restrictions or limitations that would necessarily preclude the employee from performing the duties of his job as described.” In summarizing the information in Kalish’s file, Dr. Conrad states that he reviewed the form submitted by Kalish’s employer, which describes the job duties of the Director of National Transportation. Dr. Conrad notes that Kalish’s position is “high stress with many deadlines” and “includes responsibility for directing all aspects of transportation operations, handling negotiations, travel to other sites, and direct supervision of employees.” At no point, however, does Dr. Conrad explain how someone with Kalish’s condition — which Dr. Conrad explained as “left ventricular dysfunction associated with an increased risk of recurrent congestive heart failure and increased mortality risk” — could be expected to function on a daily basis in the high-stress environment characteristic of Kalish’s former position.

Dr. Conrad bases his finding that Kalish could return to work primarily on an October 25, 2001 stress test. This test, which Dr. Conrad said demonstrated “significant clinical improvement,” was found by Dr. Conrad to both contradict and outweigh more recent echocardiology test results, which continued to show “moderate to severe impairment.” Although acknowledging “that the studies are not directly comparable,” Dr. Conrad determined that this lone stress test, conducted several months before, justified his ultimate conclusion that “[t]here is no clear evidence that the employee’s current cardiac status would necessarily preclude work involving light activity, as described.” (Emphasis added.) But the Plan language in the present case explicitly provides that Kalish will be deemed disabled if he is “unable to perform all of the material and substantial duties of his occupation on an Active Employment basis because of an Injury or Sickness.” Kalish’s position, which required him to actively supervise 50 employees and travel extensively, cannot reasonably be found to require only “light activity.”

Dr. Rasak, in contrast, offered the opinion that Kalish was incapable of working an eight-hour day and “functioning well in a work environment because it would not be feasible for [him] to perform work involving any significant exertion.” This determination was made by Dr. Rasak in light of what he “understood [Kalish’s] job duties as Director of National Transportation” to be. Dr. Conrad, in arriving at a contrary opinion, does not explain why he disagreed with Dr. Rasak’s conclusions. See Calvert, 409 F.3d at 296 (finding a reviewing physician’s report to be inadequate because, even though the reviewing physician “does mention [the claimant’s doctors] by name, he does not explain why their conclusions ... were rejected out-of-hand”).

The fact that Dr. Rasak had the opportunity to physically examine Kalish on numerous occasions, while Dr. Conrad relied exclusively on a file review, makes Dr. Conrad’s failure to discuss the findings of Dr. Rasak all the more troublesome. See id. at 295 (concluding that the plan administrator’s reliance on a “ ‘pure paper’ review” was “just one more factor” that supported the court’s ruling that a denial of benefits was arbitrary and capricious); see also McDonald, 347 F.3d at 170 (“The evidence presented in the administrative record did not support the denial of benefits when only [the adminis *510 trator]’s physicians, who had not examined [the claimant], disagreed with the treating physicians.”).

We further note that, although Dr. Conrad’s report mentions in passing that Kalish has been diagnosed with depression (see Part II.D. below), the report offers no conclusion with respect to whether Kalish’s depression would impact his ability to return to his former position. Both the district court and the parties have treated the diagnosis of depression as distinct from the diagnosis of the heart condition. But Dr. Rasak concluded that Kalish was unable to return to work after considering Kalish’s overall physical and mental health, which included an analysis of the interrelated effects of the heart condition and the depression. The fact that the report of Dr. Conrad does not discuss Kalish’s depression or acknowledge the connection between the two ailments further calls into question Liberty’s reliance upon the opinion of Dr. Conrad.

Finally, Dr. Conrad’s report omits any discussion of the findings by Liberty’s own field investigator, who personally interviewed Kalish at home in January of 2002. This investigator found that Kalish “appears to have a limited lifestyle, due to tiring easily.” Far from explaining how a person who fatigues easily could assume the responsibilities required by Kalish’s former position, Dr. Conrad’s report, prepared less than a month later, fails to even mention the investigator’s observations. Perhaps Dr. Conrad dismissed the investigator’s report without comment because he believed that Kalish’s performance on the October stress test suggested that Kalish had exaggerated his exhaustion levels to the investigator. But, if so, Dr. Conrad made this credibility determination without having physically examined Kalish and in contradiction of the Liberty investigator’s conclusion that plaintiff was “very credible.” We thus accord little weight to Dr. Conrad’s conclusions with respect to the fatigue level experienced by Kalish. See Calvert, 409 F.3d at 297 n. 6 (stating that, although “there is nothing inherently improper with relying on a file review, ... [if] the conclusions from that review include critical credibility determinations ..., reliance on such a review may be inadequate”).

Although we are mindful of the general rule that “when a plan administrator chooses to rely upon the medical opinion of one doctor over that of another in determining whether a claimant is entitled to ERISA benefits, the plan administrator’s decision cannot be said to have been arbitrary and capricious,” McDonald, 347 F.3d at 169, we nevertheless conclude that Liberty improperly denied the claim in the present case. Liberty relied exclusively on the independent file review of Dr. Conrad in making its determination, but Dr. Conrad’s report is inadequate in several crucial respects. He concludes that Kalish could return to a position requiring “light activity,” but Dr. Conrad does not state that Kalish can return to his “high-stress” position as Director of National Transportation. See id. at 172 (finding that reliance on the report of an independent physician was arbitrary and capricious where the reviewer “opined that [the claimant] could return to work, [but] he did not say what kind of work he could perform”). His report also fails to rebut the contrary medical conclusions reached by Dr. Rasak, the physician who conducted regular physical examinations of Kalish. See Nord, 538 U.S. at 834, 123 S.Ct. 1965 (“Plan administrators, of course, may not arbitrarily refuse to credit a claimant’s reliable- evidence, including the opinions of a treating physician.”). Finally, Dr. Conrad neglects to even mention the observations of Liberty’s own field investigator that undermine Dr. Conrad’s diagnosis. See Calvert, 409 *511 F.3d at 297 (finding that a report was inadequate where the record contained “conclusions which [the reviewing physician] never addresses head-on and simply seemed to ignore”). We therefore conclude that Liberty acted arbitrarily and capriciously in denying Kalish disability benefits on the basis of his cardiac condition.

D. The district court erred in holding that Liberty was not required to consider evidence of Kalish’s alleged depression before denying his claim

In concluding that Liberty did not act arbitrarily and capriciously in refusing to consider Kalish’s depression, the district court found that “the psychiatric difficulties that the Plaintiff has mentioned arose after the rejection and cannot be considered with the record that was before the Plan Administrator.” Because Liberty does not dispute that Kalish’s administrative appeal explicitly raised his depression as a basis for a finding that he was disabled, the “rejection” to which the district court refers must be the initial denial by Liberty of Kalish’s benefits on March 15, 2002. Kalish, for his part, acknowledges that the review of his case must be limited to the administrative record, but he contends that the district court construed the term too narrowly, and that the record properly includes all of the evidence that he submitted to Liberty before its final denial of his administrative appeal on July 30, 2002.

The administrative record in an ERISA case includes all documentation submitted during the administrative appeals process because this information was necessarily considered by the plan administrator in evaluating the merits of the claimant’s appeal. See Moon v. Unum Provident Corp., 405 F.3d 373, 378 (6th Cir.2005) (“Our review is confined to the administrative record as it existed on [the date] when [the administrator] issued its final decision upholding the termination of [the claimant]’s [long term disability] benefits.”); Tremain v. Bell Indus., Inc., 196 F.3d 970, 976 (9th Cir.1999) (excluding evidence only where failing to do so “would lead to the anomalous conclusion that a plan administrator abused its discretion by failing to consider evidence not before it”). The district court therefore erred when it failed to determine whether Liberty had given due consideration to Kalish’s evidence that he suffered from depression.

Turning to the merits of this issue, Liberty contends that Kalish’s depression does not entitle him to disability benefits because he did not list depression as a grounds for such benefits in his application, and because he did not begin treatment for his depression until six months after Liberty’s initial denial. But the only support that Liberty offers for the proposition that treatment for a disability must begin immediately upon a claimant’s filing for benefits is the unpublished case of Bishop v. Metropolitan Life Insurance Co., No. 01-2458, 2003 WL 21659439 (6th Cir. July 10, 2003). In Bishop, this court held that because there was no evidence that the claimant “was in fact disabled due to depression at the time of filing her claim for long-term disability benefits,” and because the claimant did not begin treatment until “a year after the filing of her claim,” the plan administrator had not acted arbitrarily and capriciously in denying benefits to the claimant. Id. at *6.

Bishop is distinguishable from the present case, however, because the plan administrator in Bishop fully considered the claimant’s evidence of depression before determining that the claimant was not disabled. When evidence of the depression emerged following the final denial of Bish *512 op’s appeal, the plan administrator reopened her case, gave her the opportunity to submit additional medical information, and had her entire ■ file reviewed by an independent physician. See id. at *4. The plan administrator ultimately denied benefits only after considering the review by the independent physician of Bishop’s evidence of depression.

In the present case, Dr. Rasak reported that “Kalish is severely debilitated as a result of his extensive damages to his heart. This has led to severe depression, which is marginally treated with medications.” Dr. Rasak made this diagnosis in January of 2002, two months before Liberty’s initial denial of benefits. Even before that, in November of 2001, Dr. Raminick observed that Kalish was “having significant depression” and “may need psychiatric referral.” Dr. Conrad, in conducting his file review in December of 2001, notes that Kalish was taking the antidepressant medication Remeron. Finally, during the appeal process, Kalish submitted two additional letters from his psychologist, Dr. Last, attesting to the severity of his depression.

Liberty informed Kalish, in a July 17, 2002 letter, that the evidence he submitted regarding his depression would be reviewed by a “consulting physician.” Instead, the only review of this evidence was made by a nurse employed by Liberty, who simply read the two letters submitted by Dr. Last. The nurse, in lieu of analyzing the issue on the merits, concluded that the opinions of Kalish’s three doctors — who had all diagnosed Kalish with depression— did not need to be considered by Liberty because Kalish had not sought treatment for his depression until six months after Liberty’s initial denial of benefits and because Dr. Last had not included his office notes along with his letters.

This assessment is factually inaccurate, however, because Kalish’s medical records reveal that he began taking antidepressants shortly after his two heart attacks and long before Liberty’s initial denial of benefits. And even without the office notes of Dr. Last, the repeated references to Kalish’s depression in the office notes of the three physicians treating Kalish for his heart condition were sufficient to require Liberty to investigate further before denying Kalish disability benefits. See Killian v. Healthsource Provident Adm’rs, Inc., 152 F.3d 514, 521 (6th Cir.1998) (“The key question, however, is what constitutes the universe of information that Healthsource should have considered.... Healthsource acted arbitrarily and capriciously in formulating the limitations it placed on the information it would take into account.”).

In defense of its failure to consider the opinions of Kalish’s doctors regarding the debilitating nature of his depression, Liberty argues that it was required to review only the evidence relating to Kalish’s heart condition because this was the sole medical condition listed on his application for benefits. But Liberty provides no support for this argument beyond its citation to the unpublished Bishop opinion. Moreover, such a contention strikes us as illogical in the context of long-term disability benefits. A person who is hospitalized for broken bones received in a fall, for example, may subsequently develop a severe infection. After the broken bones have healed, the person might still be unable to return to work because of the infection, even though the subsequent infection was not listed on the application for benefits. We see no reason that a mental illness triggered by a physical injury should be treated any differently.

Whether Kalish’s depression constitutes a disability within the meaning of the Plan must be considered in the first instance by *513 Liberty, and Liberty’s failure to do so before denying Kalish’s claim for benefits was arbitrary and capricious. See Moon, 405 F.3d at 379 (“[M]erely because our review must be deferential does not mean our review must also be inconsequential. While a benefits plan may vest discretion in the plan administrator, the federal courts do not sit in review of the administrator’s decisions only for the purpose of rubber stamping those decisions.”). Our need to remand this matter with instructions for Liberty to consider Kalish’s depression as a basis for finding him disabled, however, is obviated because of our conclusion that Kalish is entitled to disability benefits on account of his heart condition.

III. CONCLUSION

For all of the reasons set forth above, we REVERSE the judgment of the district court and REMAND this case for entry of an order requiring Liberty to award benefits plus interest from the date on which Kalish’s benefit payments ceased.