Marcus Hensley suffered a serious knee injury while deployed by the United States Army in Iraq combat in 2005. In September 2007, he underwent right knee surgery, and the Department of Veterans Affairs (“VA”) awarded him benefits for service-connected disability. He then applied for Social Security disability insurance benefits, claiming as severe impairments: posttraumatic stress disorder (“PTSD”), back pain, right knee pain, and facial twitching. The Commissioner denied the application, ruling that Hensley was not disabled as of May 19, 2011. Hensley sought judicial review; the denial was affirmed. Hensley v. Colvin, No. 4:12CV00352, Mem. & Order (E.D. Ark. Aug. 23, 2013).
Hensley filed this second application for disability insurance benefits in August 2012, while his appeal from the first denial was pending. He alleged the same severe impairments and a disability onset date of May 20, 2011. His earnings record limited him to insurance coverage through September 30, 2011. See 20 C.F.R. § 404.131. After an August 2013 hearing at which Hensley and a vocational expert (“VE”) testified, the ALJ denied the application, concluding that Hensley’s impairments were severe but he retained the residual functional capacity (“RFC”) to perform certain sedentary work during the relevant period, May 20 to September 30, 2011. The Appeals Council denied further review, the district court1 upheld the denial of benefits, and Hensley appealed. It was established in the prior proceeding that Hensley was not disabled prior to the alleged May 20 onset date. Thus, the question is whether he met his burden • to show that he became disabled during the four-month period at issue. Concluding that substantial evidenee on the administrative record as a whole supports the ALJ’s contrary determination, we affirm. See Welsh v. Colvin, 765 F.3d 926, 927 (8th Cir. 2014) (standard of review).
I. Background
A. The Medical Evidence Relating to Physical Impairments. In May 2011, Hensley saw his primary-care VA physician, Richard McKelvey, complaining of worsening low back pain and pain in his left knee, reporting that his surgically repaired right knee had improved. Dr. McKelvey observed that Hensley appeared well, was in “no acute distress,” and had a “normal” knee exam and gait. Dr. McKel-vey ordered x-rays of the lumbar spine and left knee, and an MRI of the lumbar spine. The spine x-ray showed “very mild spurring in the lower thoracic region” but was “otherwise unremarkable.” The left knee x-ray was “normal.” X-rays of Hensley’s right knee, taken in 2010, showed “small suprapateller joint fluid collection,” but were otherwise “normal” and “unremarkable.”
On July 13, 2011, Jose Escarda, M.D., reviewed the MRI of Hensley’s lumber spine. Dr. Escarda determined Hensley had “elements of strain from a pelvic ma-lignment” and recommended a home exercise program with physical therapy (“PT”). He also noted that Hensley’s knee assessment was “normal,” he had “full knee range, good lower limb and spinal flexibility,” and he was no longer using a cane. Hensley received PT in July and August of 2011 and reported his pain fell from seven or eight on a ten-point scale to four. He also used a “TENS” unit and heat pack to manage his pain. At his final PT session, Hensley said his pain that morning was *930“minimal” and he had no pain at times; the physical therapist concluded Hensley was “in alignment.” In September 2011, Dr. McKelvey noted Hensley was “doing very well,” had “full flexion” in his lower back, and his back pain was “stable” due to the home exercises and PT. Dr. McKelvey also reported that Hensley’s facial tics were “very well controlled” with medication.
At a May 2012 disability exam, Dr. McKelvey noted that Hensley was “well appearing,” in “no acute distress,” “with no restriction in mobility,” and “[fjully alert and oriented.” Dr. McKelvey summarized his exam findings in a “To Whom It May Concern” letter. The letter reported that Hensley’s back pain had “been stable for the past year” though severe at times; he was no longer using a cane; and he had “full forward flexion,” normal strength, and a normal gait. Dr. McKelvey reported that Hensley’s right knee pain was “stable” and the knee was less painful on palpation, had a full range of motion, flex-ion, and extension, and could bear weight without a brace or crutch. Treatment of Hensley’s facial tics was continuing, thus far with “limited results.” His symptoms of PTSD and depression were being treated with two medications and were “stable,” though he complained of nightmares.
B. The Medical Evidence Relating to Mental Impairments. The VA first diagnosed and treated Hensley for PTSD and depression between 2007 and 2009. In July 2011, he returned to the VA’s mental health clinic for the first time since April 2009 and saw Sandra Ellis, M.D., complaining of “intrusive thoughts, nightmares, hypervigilance and depression.” Dr. Ellis confirmed the prior diagnosis of PTSD and major depressive disorder. She assessed a Global Assessment of Functioning. (“GAF”) score of 51, increased the dosage of an antidepressant, and referred Hensley to outpatient PTSD group therapy. Hensley agreed with the referral and attended the first therapy orientation session, where he “was attentive and asked questions.” When he missed the last three sessions, the clinic notified him on September 20 that he was being discontinued from the program.
Hensley returned to the VA Mental Health Management unit in late August 2012, meeting with Advanced Practice Nurse Penelope Pollock. Hensley reported that he was sleeping five hours per night with fewer nightmares, was willing to decrease the dosage of his antidepressant, and was “interested in enrolling in treatment programming.” Nurse Pollock again scheduled him in the outpatient PTSD therapy program. He cancelled the initial session, failed to show for the next session, and was discontinued from the program for lack of attendance on October 2, 2012.
C. The Hearing Testimony. At the administrative hearings, Hensley testified that depression and PTSD were the primary reasons he felt disabled. He explained that those conditions made it difficult to “adapt to certain environments,” though he had no problem functioning with supervisors or coworkers. He testified that antidepressant medication provided stability, and he felt better and “more stable” after the increased dosage in July 2011. He outlined his typical activities, which included driving his children around town, helping with homework, running quick errands, attending church twice a month, and. assisting with family finances and chores such as mowing the lawn. The ALJ asked:
Q. Tell me why you can’t get a job where it’s mostly sitting all day and just do it.
A. My mental state. I used to be a very social guy. After the stuff I’ve been through and I’ve seen, it’s hard for me to socialize with people.
*931Q. What about a job where I set you in the corner and let you look at the wall? Your desk is up against the wall and you can do your work there. Tell me why you can’t do that?
A. I would be up and down out of my chair with pain. The leg starts hurting. My back kills me. I’ve got to have some type of support in my chair. This chair isn’t helping. I would be embarrassing myself due to my PTSD and anxiety and depression. I just can’t do it.
Prior to the hearing, the ALJ recognized that Hensley suffered from one or more severe impairments and therefore testimony by a VE would be needed to complete steps four and five of the well-established disability evaluation — determining whether Hensley had the RFC to perform his past relevant work or “other work [that] exists in significant numbers in the national economy.” See 20 C.F.R. §§ 404.1520, 404.1560. At the hearing, after Hensley testified, the ALJ asked VE Elizabeth Clem to assume that a person of Hensley’s age, education, and work experience was limited to jobs at the sedentary exertional level with the following limitations: never climb ladders, ropes, or scaffolds; occasionally climb ramps and stairs; occasionally balance, stoop, kneel, crouch, or crawl; and with a simple routine, repetitive tasks, only incidental interpersonal contact, and supervision that is simple, direct, and concrete. Clem testified that this person could not do Hensley’s past relevant work but could perform jobs that exist in significant numbers in the national or regional economies. Clem cited two examples from the Dictionary of Occupational Titles (“DOT”)— lamp-shade assembler, DOT 739.684-094, and stringing-machine tender, DOT 689.585-018.
D. The ALJ’s Decision. In a September 2013 decision, the ALJ found (i) that Hensley had severe impairments— degenerative disc disease of the lumbar spine, PTSD, anxiety disorder, and status-post right knee arthroscopy — that “could reasonably be expected to cause” his complained-of symptoms; (ii) that his “statements concerning the intensity, persistence and limiting effects of [his] symptoms are not entirely credible”; and (Hi) that he had the RFC described in the hypothetical to VE Clem, except that he was limited to no climbing. Based on this RFC and the testimony of Clem, the ALJ concluded that Hensley was not disabled because he retained the RFC to do certain unskilled sedentary jobs during the four-month time period at issue. The ALJ acknowledged that the VA had awarded Hensley disability payments but noted that “a finding of disability from another agency is not binding on the Social Security Administration,” which must make an independent determination of disability as defined by the Social Security Act.
II. Discussion
On appeal, Hensley argues that the ALJ erred in: (1) determining the RFC; (2) partially discounting his credibility; and (3) addressing the VA disability finding. “We review the district court’s decision upholding the denial of benefits de novo but, like the district court, we must uphold the ALJ’s decision if it is supported by substantial evidence on the administrative record as a whole.” Welsh, 765 F.3d at 927. “Substantial evidence is less than a preponderance, but enough that a reasonable mind might accept it as adequate to support a decision.” Cox v. Astrue, 495 F.3d 614, 617 (8th Cir. 2007) (quotation omitted).
A. The RFC Determination. RFC is defined as' the most a claimant can do despite his limitations, including both physical and mental limitations. 20 C.F.R. § 404.1545(a). “The Commissioner must *932determine a claimant’s RFC based on all of the relevant evidence, including the medical records, observations of treating physicians and others, and an individual’s own description of [his] limitations.” Myers v. Colvin, 721 F.3d 521, 527 (8th Cir. 2013) (quotation omitted). Where, as here, the claimant proves he cannot perform his past relevant work, the Commissioner has the burden of producing evidence that he has the RFC to perform other jobs. Goff v. Barnhart, 421 F.3d 785, 790 (8th Cir. 2005). However, “the burden of persuasion to prove disability and to demonstrate RFC remains on the claimant, even when the burden of production shifts to the Commissioner at step five.” Id. (alteration and quotation omitted).
Hensley argues that “reversal is warranted” because “no medical opinion supports the ALJ’s RFC” determination, and the ALJ failed to order a consultative examination (“CE”) to correct this lack of direct opinion evidence. We disagree. “Because a claimant’s RFC is a medical question, an ALJ’s assessment of it must be supported by some medical evidence of the claimant’s ability to function in the workplace.” Cox, 495 F.3d at 619. However, there is no requirement that an RFC finding be supported by a specific medical opinion. See Myers, 721 F.3d at 526-27 (affirming RFC without medical opinion evidence); Perks v. Astrue, 687 F.3d 1086, 1092-93 (8th Cir. 2012) (same).
In the typical Social Security disability case, the administrative record includes one or more opinions by the claimant’s treating physician(s) as to the impact of impairments on his RFC. In this case, after Hensley applied for Social Security disability benefits, he asked “if Dr. McKel-vey could complete an assessment of ability to do work related activities.” The VA advised “that primary care providers at the VA cannot complete those kinds of assessments.” Instead, Dr. McKelvey provided a “To Whom It May Concern” letter summarizing his May 2012 disability exam that addressed each of Hensley’s severe impairments. Together with the extensive VA treatment records, this provided adequate medical evidence of Hensley’s ability to function in the workplace. Accord Cox, 495 F.3d at 620 n. 6. In the absence of medical opinion evidence, “medical records prepared by the most relevant treating physicians [can] provide affirmative medical evidence supporting the ALJ’s residual functional capacity findings.” Johnson v. Astrue, 628 F.3d 991, 995 (8th Cir. 2011). With the medical record adequately developed, the ALJ was not requiréd to seek additional information from Dr. McKelvey or order a CE. See KKC ex rel. Stoner v. Colvin, 818 F.3d 364, 372-73 (8th Cir. 2016); 20 C.F.R. § 404.1519a(b).
Hensley’s remaining attacks on the ALJ’s RFC findings and determination fall in the category of deficiencies in opinion writing. “[A]n arguable deficiency in opinion writing that had no practical effect on the decision ... is not a sufficient reason to set aside the ALJ’s decision.” Welsh, 765 F.3d at 929.
Hensley first argues the ALJ “ignored” a November 2010 letter from Dr. McKelvey stating that Hensley “needs a residence on the ground level with no stairs.” This letter was written well prior to the time period at issue on a subject addressed by the climbing restriction included in the RFC. The ALJ properly focused on Dr. McKelvey’s later reports during the relevant period. “[A]n ALJ is not required to discuss every piece of evidence submitted.” Black v. Apfel, 143 F.3d 383, 386 (8th Cir. 1998). Relatedly, Hensley argues the ALJ erred in relying on the VE’s testimony because the ALJ’s hypothetical to the VE assumed a worker who can occasionally climb ramps and stairs, whereas the ALJ’s subsequent RFC finding limited him to no *933climbing. However, this discrepancy was irrelevant to the ALJ’s RFC determination. The two jobs the VE identified in response to the ALJ’s hypothetical — lampshade assembler and stringing-machine tender — do not require climbing. See DOT 739.684-094 (lamp-shade assembler), 689.585-018 (stringing-machine tender); cf. Van Vickle v. Astrue, 539 F.3d 825, 830 (8th Cir. 2008).
Hensley further argues the ALJ “essentially ignored” his alleged foot condition, mild bilateral halux valgus2 deformities, which allegedly prevented him from doing the standing and walking involved in sedentary work. Though Hensley did not list this impairment in his application for disability benefits, the ALJ inquired about it at the August 2013 hearing. Hensley testified he had the condition for as long as he .“could remember.” It caused some pain, but he had not yet agreed to surgery his VA medical providers offered in March 2011. The ALJ was not required to address this condition in his opinion. See Ostronski v. Chater, 94 F.3d 413, 419 (8th Cir. 1996) (refusing surgery suggests condition is not disabling).
Finally, Hensley argues the ALJ failed to consider whether he could sustain his ability to work over a full day in a real world work setting, as Social Security Ruling 96-8p required. However, Hensley “neither identifies evidence the ALJ failed to consider nor specifies how the ALJ’s assessment was unrealistic.” Juszczyk v. Astrue, 542 F.3d 626, 633 (8th Cir. 2008). The ALJ explicitly considered the persis-fence and limiting effects of Hensley’s symptoms and impairments, and the RFC took into account the effect particular work environments would have on Hensley’s ability to function by limiting the jobs he could perform to those with direct supervision and incidental interpersonal contact. Thus, the ALJ properly “engaged in a realistic assessment of [Hensley’s] abilities.” Id.
After careful review of the administrative record, like the district court we conclude that substantial evidence supports the ALJ’s RFC determination. Regarding Hensley’s mental impairments, the medical records reflect that Dr. Ellis assigned Hensley a GAF score of 51 in July 20113 and increased the dosage of an antidepressant medication. Dr. McKelvey observed that Hensley was “doing very well” and later reported that Hensley’s mental illness symptoms had stabilized. Hensley testified that, after the increased dosage, he got “better” and felt “more stable.” He also testified to having no problem functioning with coworkers or supervisors. In August 2012, Nurse Pollock noted that Hensley was continuing with the medications, agreed to a reduced antidepressant dosage, and was sleeping five hours a night and having fewer nightmares. After both sessions, Hensley expressed interest in attending a prescribed group therapy program but then failed to attend. The ALJ found that Hensley’s mental impairments were controlled, or at least controllable, during the relevant period. “If an impairment can be controlled by treatment *934or medication, it cannot be considered disabling.” Brace v. Astrue, 578 F.3d 882, 885 (8th Cir. 2009) (quotation omitted); see 20 C.F.R. § 404.1530(a) (“to get benefits, you must follow treatment prescribed by your physician if this treatment can restore your ability to work”). The ALJ’s RFC determination accounted for Hensley’s mental impairments by limiting him to jobs that involve simple, repetitive tasks learned by rote, with incidental interpersonal contact and simple, direct, concrete supervision.
Regarding Hensley’s physical impairments, the medical evidence during and immediately after the time period in question showed that, after completing PT, Hensley reported greatly reduced, sometimes nonexistent, back and knee pain. His physical therapist determined he “was in alignment,” Dr. McKelvey confirmed his pain was “stable” as a result of the therapy, and both Dr. Escarda and Dr. McKel-vey reported that Hensley’s knee range, flexion of knee and spine, gait, and strength were “normal.” Following the May 2012 disability exam, Dr. McKelvey confirmed Hensley’s pain had “been stable for the past year,” he had “no restriction in mobility,” was not using a cane, and could bear weight without a brace or crutch. Hensley testified he no longer used a knee brace because of effective pain medication. Dr. McKelvey also observed that Hensley’s facial tics were “very well controlled” with medication. Hensley’s testimony regarding his daily activities supported the ALJ’s finding that he had the RFC to perform a limited universe of sedentary jobs. The ALJ accounted for Hensley’s physical impairments by limiting him to sedentary jobs that do not require climbing of ladders, rails, or stairs and that require only occasional balancing, stooping, kneeling, crouching, and crawling.4
B. Subjective Complaints. Hensley claims the ALJ erred in discounting his complaints about the “intensity, persistence, and limiting effects” of his subjective complaints. The ALJ properly cited and considered the factors enumerated in Polaski v. Heckler, 739 F.2d 1320, 1322 (8th Cir. 1984), before discounting Hensley’s complaints of disabling pain and other subjective symptoms. The ALJ did not dispute that Hensley experienced symptoms, but found those symptoms not to be as limiting as Hensley claimed because, among other reasons, no treating physician had opined that he was disabled; he did not follow the recommended course of treatment for PTSD; his impairments were controlled by medication and treatment; and he performed “a wide range of daily activities.” “We will defer to an ALJ’s credibility finding as long as the ALJ explicitly discredits a claimant’s testimony and gives a good reason for doing so.” Schultz v. Astrue, 479 F.3d 979, 983 (8th Cir. 2007) (quotations omitted).
Hensley attacks each of these findings. Each is supported by substantial evidence in the administrative record, and Hensley’s contentions that the ALJ should have weighed these facts differently or drawn different conclusions do not warrant relief under our deferential standard of review. “It is not the role of this court to reweigh the evidence presented to the ALJ.” Cox, 495 F.3d at 617 (quotation omitted). However, one contention deserves further discussion.
Hensley argues the ALJ erred in discounting his subjective complaints based on his failure to attend prescribed group *935therapy to treat his PTSD, without inquiring into the circumstances surrounding the alleged failure. Hensley relies on Pate-Fires v. Astrue, 564 F.3d 935, 937, 945-46 (8th Cir. 2009), where the evidence “overwhelmingly demonstrate^]” that the failure of a claimant to take her prescribed medication “was a medically-determinable symptom of her mental illness,” described as “bipolar disorder I severe, with psychotic features.”
The Social Security Administration has recognized that there are circumstances in which a claimant’s failure to follow prescribed treatment is justifiable and therefore does not preclude a finding of disability. See Soc. Sec. Ruling 82-59. Whether severe mental illness has resulted in justifiable noncompliance is a fact-intensive issue. Here, Hensley accepted and completed PT that lessened his physical impairments. In the relevant time period, he was prescribed and took medications that stabilized his mental impairments, PTSD and depression. In July 2011 and again in August 2012, he was prescribed and expressed interest in completing a group therapy program to treat his PTSD. But after attending one session in 2011, he twice failed to attend and was discontinued from that program.
When questioned at the hearing, Hensley testified that the VA mental health providers failed to contact him about the sessions. The medical records suggest otherwise, but in any event, there is no evidence that Hensley’s failure to attend “was a medically-determinable symptom of [his] mental illness.” Therefore, the ALJ could reasonably conclude that Hensley’s repeated failure • to attend a prescribed course of treatment was evidence that his mental impairment was less disabling than Hensley claimed. See Bradley v. Astrue, 528 F.3d 1113, 1115 (8th Cir. 2008).
C. VA Disability Finding. Hensley’s final argument is that the ALJ did not “properly evaluate and discuss” the VA finding that he is disabled. He relies on Morrison v. Apfel, 146 F.3d 625, 628 (8th Cir. 1998), but in that case the ALJ did not even mention a VA finding that the claimant was “permanently and totally disabled.” Here, the ALJ explicitly acknowledged the VA’s disability finding, and correctly noted that the disability finding of another agency like the VA was not binding on the Social Security Administration. See 20 C.F.R. § 404.1504. There was no error. See Pelkey v. Barnhart, 433 F.3d 575, 579-80 (8th Cir. 2006) (“the ALJ did not err because he fully considered the evidence underlying the VA’s final conclusion that Pelkey was ... disabled”).
The judgment of the district court is affirmed.
. The Honorable Brian S. Miller, United States District Court for the Eastern District of Arkansas, adopting the Recommended Disposition of the Honorable Joe J. Volpe, United States Magistrate Judge for the Eastern District of Arkansas.
. Hallux valgus is an "angulation of the great toe away from the middle of the body, or toward the other toes; the great toe may ride under or over the other toes.” Dorland's Illustrated Medical Dictionary 818 (32nd ed. 2012).
. This score indicates "moderate difficulty” in social and occupational functioning. Diagnostic and Statistical Manual of Mental Disorders 34 (4th ed. 2000). GAF scores are not determinative of RFC, see Nowling v. Colvin, 813 F.3d 1110, 1116 n. 3 (8th Cir. 2016), but they offer some evidence of a claimant’s ability to function. See Myers, 721 F.3d at 525. Hensley notes that his GAF score was lower in August 2012, but that assessment by Nurse Pollock occurred almost a year after the relevant time period.
. Notably, the two state-agency physicians who reviewed Hensley's medical record found he had the RFC to do certain light work. See 20 C.F.R. § 404.1567(a)-(b). The ALJ afforded Hensley a more restricted, sedentary-based RFC.