Eighteen decisions that change what VA has to do

A case is not an argument you make. It is a rule that already binds the agency, whether or not anyone in your file mentions it. These eighteen come up constantly in disability claims, and many of them exist because VA was doing something a court told it to stop doing.

Each page states what the court held, what it did not hold, and where the official opinion is. None of them tells you what your own record supports. For that, talk to an accredited representative.

Which court, and does it bind

Four bodies decide these questions, and they do not carry equal weight.

• The Board of Veterans' Appeals decides your appeal. Its decisions bind nobody else, including you in a later claim.
• The Court of Appeals for Veterans Claims (cited Vet. App.) reviews Board decisions. Its precedential opinions bind VA.
• The Court of Appeals for the Federal Circuit (cited F.3d or F.4th) reviews the Veterans Court. It outranks both, and it can strike down a VA regulation — which is what happened in Spicer.
• The Supreme Court is above all of them, and has decided one case on this list.

Not every Veterans Court decision counts the same. Most are single-judge memorandum decisions, and they are not precedent. The Federal Circuit issues nonprecedential decisions too. If a decision has no Vet. App. or F.3d/F.4th volume and page, assume it is not precedential until you have checked.

Not precedent does not mean unusable. The court's Rule 30(a) says parties "may cite nonprecedential dispositions by any court or agency," provided they identify clear precedent on point or state that none exists, explain why the disposition is relevant, and attach a copy if it is not in a publicly accessible database.

Watch the date on whatever you are reading. Older copies of these decisions — including the versions in commercial databases — carry a notice saying the decision "may not be cited as precedent," language from an earlier version of the rule. The current rule permits citation. What has not changed is that a nonprecedential decision does not bind anyone, so a brief that leans on one as though it does is making a mistake.

When VA owes you an exam, and what makes one count

Stegall v. West (1998) — a remand order is not a suggestion. It gives you a legal right to compliance, and the Board errs if it decides the case anyway.

McLendon v. Nicholson (2006) — the four elements that trigger VA's duty to provide an examination. The third is a low threshold: the evidence need only indicate that a nexus may exist.

Stefl v. Nicholson (2007) — an opinion has to support its conclusion with analysis. And "it isn't on the presumptive list" is not an answer to whether service caused it.

Barr v. Nicholson (2007) — an examination VA chooses to give you has to be adequate. Partly abrogated on a different point; read the page before citing it.

Nieves-Rodriguez v. Peake (2008) — VA cannot prefer its own examiner solely because that examiner read the claims file, and cannot discount your doctor solely because they did not. "Most of the probative value of a medical opinion comes from its reasoning."

How your story matters

Buchanan v. Nicholson (2006) — the Board cannot find your statements not credible merely because no record from the time says the same thing.

Jandreau v. Nicholson (2007) — the three situations in which what you say yourself is enough to establish a diagnosis, and why that is a question of fact for the Board.

Clemons v. Shinseki (2009) — what you claimed is not just the word you wrote on the form. You cannot diagnose yourself, so the claim covers whatever your symptoms and the evidence reasonably support.

Bonds v. McDonough (2022) — VA has to work out how many claims a filing raised before deciding how far any one of them reaches. A case about a 2013 filing, under rules that no longer exist; the page says so twice, because it is not a reason to leave a condition off today's form.

How a disability gets connected to service

Walker v. Shinseki (2013) — the continuity-of-symptomatology route applies only to the chronic diseases actually named in the regulation. This one narrowed veterans' options, and it is why most claims have to be built on the ordinary three elements.

Atencio v. O'Rourke (2018) — causation and aggravation are separate questions, and an examiner cannot answer both with one merged paragraph.

Ward v. Wilkie (2019) — VA cannot require that a secondary worsening be permanent. It had been instructing examiners to do exactly that, in its own manual.

Spicer v. McDonough (2023) — 38 C.F.R. §3.310(b) is unlawful to the extent it conflicts with the statute, and worsening caused by not being able to get treatment is compensable.

Adams v. Collins (2025) — VA's categorical bar on obesity as a secondary disability does not stand. The bar on direct service connection does. Easy to overread; the page says what it did not decide.

When the evidence is close

Lynch v. McDonough (2021) — "approximate balance" means nearly equal, not a dead tie. The Federal Circuit abandoned the preponderance language that had obscured that.

Bufkin v. Collins (2025) — but the Veterans Court reviews VA's approximate-balance call only for clear error, because it is predominantly factual. Which is the argument for building the record before the Board rather than hoping a court reweighs it.

After the Board: two cases, one name

Andrews v. McDonough — 34 Vet. App. 151, decided May 28, 2021. If you elected Direct Review, winning at the Veterans Court does not reopen the record. The Board may not consider new evidence on remand, and the window to change dockets closes long before a court sees the case. Check your own docket and your own window before you send VA anything.

Andrews v. McDonough — 34 Vet. App. 216, decided June 22, 2021. A different veteran and an unrelated question: when a court grants a claim instead of sending it back. Reversal is rare, and this decision rejected the broad rule the appellant wanted — but a remand whose predominant purpose is to let VA look for a reason to deny is not allowed. Read alongside Bufkin.

These are two different cases. Same name, same Secretary, same year, same volume of the same reporter, sixty-five pages apart, nothing in common. The court tells its own files apart by the veterans' initials. This is not a hypothetical hazard: the two decisions arrived here in the same batch, one described and the other supplied, and it took reading both to notice.

Decisions that illustrate but do not bind

These are nonprecedential. They are not authority and no page here treats them as any. They are worth reading because they show the rules above meeting an actual record:

• Nisler v. Gibson (2014) — a private opinion saying a connection is "certainly plausible" held too speculative to carry weight. Equivocal language costs an opinion its value.
• Hazelett v. McDonald (2015) — a Board that found the veteran competent to describe his symptoms and then treated his statements as not competent evidence at all, skipping Jandreau's third situation.
• Hoskins v. Snyder (2017) — McLendon's third element failing: the veteran's own physician's questionnaire said he did not have the presumptive condition, so nothing indicated even a possible link.
• Condit v. McDonough (2022) — Andrews (34 Vet. App. 216) distinguished, on the three things its own record lacked: the Board never mentioned the manual, it found the veteran's statements not competent rather than not probative, and only one medical opinion existed with its adequacy still disputed. A clear picture of how narrow that gate is. It also carries the stale Rule 30(a) notice described above.
• Monroe v. McDonough (2023) — the Atencio problem again, in another record.
• Howard v. McDonough (2023) — Lynch did not displace "at least as likely as not." The two standards answer different questions.
• Johnson v. McDonough (2023) — a reminder of jurisdiction: the adequacy of an examination is a question of fact, and the Federal Circuit cannot review it.
• Gaines v. McDonough (2024) — an examination held adequate where causation and aggravation were answered in separate sections, which is the Atencio problem avoided.
• Long v. Hunter (2025) — McLendon's low threshold applied.
• McCarthy v. McDonough (2024) and McCarthy v. Collins (2025) — the same case twice, and the reason to check a decision's history is in the pair. See below.

Case law moves

Three of the pages here carry that history already. Barr was abrogated in part by Walker six years after it was decided. Lynch has a withdrawn first opinion at 999 F.3d 1391, so a citation to that volume points at something that no longer exists. And the Spicer that governs is a Federal Circuit decision that vacated an earlier Spicer of the same name, in the same case, holding the opposite.

The clearest illustration is not one of the eighteen. In September 2024 the Veterans Court decided McCarthy v. McDonough and affirmed the Board on every argument the veteran raised. Seven months later the same judge issued McCarthy v. Collins on the same docket, which says the Court withdrew the 2024 decision, granted reconsideration, found a threshold error, and remanded. The second decision was issued "in lieu of" the first. Anyone who pulled the 2024 decision and built on it was building on something that no longer exists, and nothing on the face of that document said so.

Adams is the instructive opposite: both sides appealed it to the Federal Circuit, which dismissed by agreement in 2026 without reaching the merits — later history that exists and changes nothing. You only know which kind you have by looking.

A citation can also point at the wrong case rather than at a dead one. Checking the history of Andrews v. McDonough tells you nothing if you have the wrong Andrews, and there are two — described above. Two other name collisions on this list are worth the same care: Adams v. Principi (Fed. Cir. 2001), which supplies the rule on when a remand is inappropriate, is nothing to do with the Adams v. Collins on obesity; and Spicer exists as both a Veterans Court decision and the Federal Circuit decision that vacated it. Match the volume, the page and the year, not the name.

That is normal, and it is the single most common way a confident citation turns out to be wrong. Before relying on any decision, check its later history. A law librarian at any public law school can show you how in ten minutes; so can the research service at an accredited representative's office.

Everything here was read against the official opinions, not a summary of them, on the date at the foot of this page.

Reviewed October 6, 2026