OPINION
I. INTRODUCTION
Jackie Cowen, claiming that her employment caused her left breast saline implant to deflate, sought workers’ compensation benefits from her employer, Wal-Mart. The Alaska Workers’ Compensation Board rejected her claim. We affirm the superior court’s affirmance of the board’s decision because we conclude that the opinions of two physicians were substantial evidence permitting the board to conclude that Cowen’s work activities did not cause her implant to deflate.
II. FACTS AND PROCEEDINGS
At the time of her injury, Jackie Cowen was employed as a Direct to Store Delivery Associate at Wal-Mart, where her job duties included loading and unloading freight, tracking merchandise, and emptying boxes. She has a history of fibrocystic breast disease for which she had undergone a bilateral mastectomy and multiple breast implant surgeries. On March 1, 2000, she felt lethargic and nauseated when she returned home at the end of her shift. She worked the next day and informed her store manager on March 3 that her left breast implant had collapsed. Although she stated in her Report of Occupational Injury that she had hit her left breast while lifting a box, she answered “maybe” on the attached questionnaire to the question of whether anything had struck her body.
Cowen informed her family physician, Derek Hagen, D.O., of her condition on March 2, 2000. Dr. Hagen referred her to plastic surgeon Sarah Troxel, M.D., who examined *423Cowen on March 6. At that time, Cowen denied any knowledge of trauma to her breast. Dr. Troxel confirmed leakage and deflation of Cowen’s left breast implant, but found no evidence of chest area bruising.
Dr. Troxel referred Cowen to George Siegfried, M.D. Dr. Siegfried has performed implant surgery since 1973 and performs approximately forty implant surgeries per year. Dr. Siegfried examined Cowen on March 14, 2000. His October 24, 2000 affidavit expressed his opinion that Cowen would remember the kind of trauma necessary to induce acute implant deflation. He also thought that a blow sufficient to cause an implant to deflate would probably cause the scar tissue around the implant to bruise, even after multiple breast surgeries. He stated that it was not possible that lifting boxes of merchandise and bumping them against the breast caused Cowen’s implant to deflate. He stated his opinion that the deflation was most likely due to natural wear caused by continuous abrasion. He stated that anything that would cause the pectoralis muscle to move would contribute to implant failure.
Cowen revisited Dr. Hagen on March 20, 2000. She complained of cough, congestion, body aches, and anxiety associated with the deflation. Dr. Hagen expressed his opinion in a letter dated May 5, 2000 that Cowen’s job duties directly caused her breast implant to deflate. He later stated in his deposition that he based his opinion on what Cowen told him had happened at work. Although he doubted that fluid from the implant directly caused her symptoms, he thought that the deflation’s effect on Cowen’s emotional state affected her pulmonary condition.
Cowen filed a worker’s compensation claim on April 12, 2000. She alleged that her left breast implant had deflated due to trauma that occurred while she was performing her regular job duties. Wal-Mart and its insurer controverted and answered her claim, asserting that the breast implant deflation was not work-related. Cowen, when deposed, testified that she could not recall any specific event at work that could have caused the implant to deflate.
At Wal-Mart’s request, Cowen was evaluated by another physician, Ajit Arora, M.D., who declined to offer his opinion regarding the cause of the injury. He noted that nausea was a symptom of fluid overload in the body, but found no association between the implant deflation and Cowen’s pulmonary condition. Rather, he concluded that her chest symptoms were related to her asthma and interstitial fibrosis.
The board ordered a second independent medical evaluation, by plastic surgeon Phil Haeck, M.D. He stated in a letter that he found no evidence of the type of trauma required to deflate Cowen’s implant. He opined that Cowen’s prior breast surgeries would not result in a diminished ability to sense pain except in the nipple areolar complex area of her breast and the adjacent skin. He also stated that her multiple surgeries would not alter the way bruising occurs after trauma. In his opinion, “shell-fold failure,” which occurs when the implant is under-filled, was the most likely cause of Cowen’s implant deflation. He noted that Cowen’s right breast implant was under-filled by approximately ten percent at the time of his examination. Although he stated that the relation between shell-fold failure and physical activity is unknown, he thought it unlikely that Cowen’s work activities accelerated the shell-fold failure and subsequent deflation of the implant. He also stated that he had never encountered a patient with a deflated saline implant who had any symptoms other than urinary frequency and decreased breast size.
At the board hearing, Cowen testified that she could not recall a specific blow that occurred the day before she noticed her implant had deflated. She stated that she lacks normal sensation or the ability to bruise in her chest area. She also testified that her only physical activity is at work and that she is physically inactive at home. Her husband corroborated her testimony.
The board denied and dismissed Cowen’s claim for medical and temporary total disability benefits. One member of the board dissented. Although the board found that Cowen had introduced sufficiént minimal evidence to raise the presumption that work *424was a substantial factor in her injury, it concluded that Wal-Mart had presented substantial evidence to overcome the presumption of compensability. Further, it concluded that Cowen had failed to prove her claim by a preponderance of the evidence. Superior Court Judge William F. Morse affirmed the board’s decision on September 24, 2002. Cowen appeals. She contends that Wal-Mart failed to present evidence sufficient to overcome the presumption of compensability.
III. DISCUSSION
A. Standard of Review
When the superior court acts as an intermediate court of appeals, we independently review the merits of the administrative decision.1 We review the administrative agency’s findings to determine whether they are supported by substantial evidence.2 Substantial evidence “is ‘such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.’ ”3 We should not “ ‘reweigh the evidence or choose between competing inferences,’ but [we should] simply determine whether such evidence exists.”4 The workers’ compensation board has exclusive authority to decide the credibility of witnesses.5
B. The Board Did Not Err in Concluding that Wal-Mart Produced Substantial Evidence To Rebut the Presumption that Cowen’s Claims Are Compensable Under the Workers’ Compensation Act.
The Alaska Workers’ Compensation Act creates a presumption that an employee’s claims are compensable.6 Applying the presumption requires a three-step analysis.7 First, the employee must establish the preliminary link between her employment and her alleged injury.8 At this step, the board should consider only evidence that tends to establish the link.9 Here the board found that Cowen introduced “sufficient ‘minimal’ evidence” to raise the presumption that her work was a substantial factor in her injury. It relied upon Cowen’s testimony describing her job duties and Dr. Hagen’s testimony that those duties directly caused her breast implant to deflate. The parties do not challenge the board’s finding that Cowen established the preliminary link between her job and the implant deflation.
Next, the employer has the burden of rebutting the presumption that the injury was work-related with substantial evidence.10 Benefits are awarded if the record establishes that a work-related injury is a substantial factor in the employee’s disability regardless of whether a non-work-related injury could independently have caused the disability.11 The employer’s substantial evidence must either (1) provide an alternative explanation for the injury that, if accepted, would exclude work-related factors as a substantial cause of the disability; or (2) directly eliminate any reasonable possibility that employment was a factor in causing the disability.12 The burden of production shifts to the employer, but not the burden of persuasion; only the employer’s evidence is examined.13 We have held that presentation of “a qualified expert who testifies that, in his or her opinion, the claimant’s work was probably *425not a substantial cause of the disability” rebuts the presumption of compensability.14 Medical testimony cannot constitute substantial evidence if it simply points to other possible causes of an employee’s condition, however.15
The board found that Wal-Mart had rebutted the presumption of compensability with substantial evidence. It relied upon the opinions of Dr. Siegfried and Dr. Haeck that Cowen’s work activities were not the cause of the implant failure. Both experts attributed the implant deflation to product failure. Dr. Siegfried stated his opinion that the implant failure was caused by product fatigue. He thought it probable that the implant wore out from everyday activity. Dr. Haeck attributed the implant collapse to “shell-fold failure” that occurred because the breast implant was under-filled. He explained that it was “unlikely ... that lifting activities at work accelerated the deflation of her breast implant.” Only Dr. Hagen linked the implant deflation to Cowen’s work activities.
Dr. Siegfried and Dr. Haeck also addressed the question whether Cowen’s employment was a factor in causing her injury. Dr. Siegfried stated that a blow sufficient to cause an implant to deflate would probably also cause scar tissue around the implant to bruise, even after multiple breast surgeries. Because Cowen experienced no bruising and could not recall a specific event in which she was hit in her left breast, he concluded that her work activities probably did not cause her implant to deflate. Dr. Haeck opined that Cowen’s multiple surgeries were unlikely to diminish her ability to sense breast pain or to alter the way bruises occur after trauma to the area. He “could find no evidence in the medical record or from [Cowen’s] history of any specific trauma which caused the deflation of [her] implant.” He thought it most likely that the implant was under-filled and that deflation was inevitable.
Cowen argues that Dr. Siegfried’s and Dr. Haeck’s opinions are “based on speculation and on factual assumptions unsupported [by] and contrary [to] the record.” Cowen points out that Dr. Siegfried cannot exclude work as a substantial factor in causing the implant failure. Dr. Siegfried initially stated that he could not rule out work as a substantial factor because he was unaware of the type of activities Cowen performed at work. After Cowen’s work activities were described to him, he stated that the pressure of boxes against Cowen’s chest was not likely to accelerate implant failure, but that any activity that caused the pecto-ralis muscle to move could contribute to it. Dr. Siegfried seemed to doubt, however, that Cowen’s work activities caused any more damage than would ordinary, everyday activities, such as “lifting a pencil or a glass of water” or “breathing.” Although he did not express his opinion in absolute terms, it was not entirely inconclusive.16 In stating that everyday activities were as likely to cause damage as work activities, he effectively discounted work-related aggravation as a substantial factor in Cowen’s disability. Although acceleration and aggravation are justifiable bases for finding a compensable injury,17 it was not clear that Cowen’s work-related movements contributed more to the abrasion than did her usual movements. A reasonable mind could therefore conclude that Dr. Siegfried’s opinions were adequate to rebut the presumption of compensability.
Cowen also argues that Dr. Haeck’s opinion was speculative and unsubstantiated. Dr. Haeck opined that the implant deflation was not a direct consequence of physical activity, whether undertaken at work or at home. He acknowledged, however, that “no one seems to know whether this phenomena is related to physical activity.”
*426Although we have held that medical certainty regarding the causes of a claimed injury is not necessary to satisfy the standard for substantial evidence,18 we have also stated that “merely reciting the proper words as an opinion is not necessarily enough to rebut the presumption of compensability.”19 Nevertheless, we have held that a doctor’s testimony that the employee’s job was not a substantial factor in causing her disability was substantial evidence to rebut the presumption of compensability even when the doctor admitted that the causes of the disability were unknown.20 We reasoned that it would otherwise create an irrebuttable presumption that the disability was work-related.21
A similar rationale applies here. We should not discount Dr. Haeck’s opinion, which reflects his considerable experience with and knowledge of breast implants, because of medical uncertainty. Dr. Haeck based his opinion on his best professional judgment. A reasonable mind could accept his opinions as adequate to support the conclusion that Cowen’s work activities were not a substantial factor in the shell-fold failure and subsequent deflation of the implant, despite his inability to state definitely whether the failure was related to physical activity.
After an employer rebuts the presumption that injuries are work-related, the employee must prove his or her claim by a preponderance of the evidence to prevail.22 The employee must induce a belief in the trier of fact that the asserted facts are probably true.23 The claimed injury is compensa-ble if the employee’s work is a substantial factor in causing it.24
The board, after weighing the evidence, including the opinions and the medical records, found the preponderance of the evidence demonstrated that Cowen’s breast implant deflation was not a work-related injury. Dr. Siegfried’s opinion persuaded the board that the implant spontaneously deflated due to abrasion when Cowen moved her pectoralis muscle. The board also gave “great weight” to Dr. Haeck’s opinion that the breast implant was going to fail regardless of Cowen’s physical activity at work or home and that her job duties did not accelerate the failure. The board gave less weight to Dr. Hagen’s opinion because he is a family practitioner who has less experience with breast implants. When medical experts provide contradictory testimony, the board determines credibility.25 “[I]f the Board is faced with two or more conflicting medical opinions — each of which constitutes substantial evidence — and elects to rely upon one opinion rather than the other, we will affirm the Board’s decision.”26 It was reasonable for the board to discount Dr. Hagen’s testimony because he had limited experience with breast implants. The evidence that was sufficient to rebut the presumption of compens-ability was also sufficient to support the board’s determination that Cowen failed to show by a preponderance of the evidence that her injury was work-related.
IY. CONCLUSION
For these reasons we AFFIRM the superi- or court order that affirmed the decision of the Alaska Workers’ Compensation Board.
BRYNER, Chief Justice, and MATTHEWS, Justice, not participating.