Alzheimer's Disease & Dementia Diagnosis
Alzheimer's Disease & Dementia Diagnosis
A dementia assessment is not a single test that returns a result. It is a sequence in which the common, cheap and reversible answers are dealt with first, and the confirmatory and expensive ones are reached only if they are still needed. This page sets out what each part of the assessment is actually for, how cognition is tested and what the scores mean, what imaging and biomarkers add, and why the exclusion list is the part that matters most to the person in front of you.
The question the assessment actually answers
An assessment is often imagined as a test for dementia. It is not, and the difference is practical: a positive or negative result is less useful than a cause. The assessment answers four questions, in this order, and the order is what keeps it efficient.
| Question | What answers it | Why it comes in this position |
|---|---|---|
| Is there decline? | A history from the person and from someone who knows them, plus documented cognitive testing | Everything else depends on it. A complaint without documented impairment is not the same thing |
| Is it progressive, and how fast? | A second assessment separated by months, and the trajectory from the history | Progression is the defining feature. It cannot be established in a single visit |
| What is the cause? | The phenotype, the imaging pattern, and where indicated the biomarkers | This is where Alzheimer's is distinguished from the dementias around it, and where treatment is decided |
| What else is contributing? | The medication review, and assessment of hearing, sleep, mood, pain and vascular risk | Usually the largest reversible component, and usually the last thing considered |
A number of diagnoses are made badly because the third question is answered before the fourth. The label is reached, the contributing causes are never looked for, and the person is left with a prognosis when what they needed was a hearing aid, a sleep study and three medicines stopped.
The history, and who has to be in the room
The single most informative part of a dementia assessment is not a scan. It is a history taken from two sources: the person, and someone who sees them regularly. Where only one of those is present, the assessment is poorer, and the reason is worth explaining.
The person's own account establishes what they have noticed, what they are worried about, and what they can still do. The informant's account establishes something the person often cannot: when it started, how fast it moved, and what has already been quietly taken over. That last item is the most reliable measure of functional decline available, and it is almost never volunteered because the family does not think of it as information.
The specific things worth establishing are these, and they are worth writing down before an appointment:
| Element | Why it changes the answer |
|---|---|
| The first domain affected | Memory, language, visuospatial function and behaviour each point to a different underlying pathology and to different tests |
| Onset and tempo | Years suggests neurodegeneration; weeks to months suggests an autoimmune, infectious, neoplastic or prion process, and those are urgent to exclude |
| Fluctuation | Marked within-day variation points towards Lewy body disease or delirium rather than uncomplicated Alzheimer's |
| What the family has taken over, and when | The most reliable functional measure available, and the one most often omitted |
| Every medicine, including over-the-counter | The anticholinergic and sedative burden is a common contributor and it is reversible by subtraction |
| Falls, however trivial they seemed | A subdural collection can present weeks later as cognitive change alone |
| Family history and age of onset in relatives | Relevant to genetic counselling in early-onset disease, and to risk discussion in everyone |
| Mood, sleep, hearing and pain | Each is treatable, each can present as cognitive decline, and each is commonly present alongside it |
One practical note: the examination includes more than cognition. Gait, parkinsonism, eye movements, reflexes and the peripheral nervous system are all examined, because the findings there are often what separates one dementia from another.
Cognitive testing: what the scores are for
Cognitive testing does two things that a conversation cannot. It puts a number on the decline so that it can be compared later, and it maps which domains are affected. What it does not do is diagnose a cause, and it is worth saying so because both errors are common: a score treated as a diagnosis, and a normal score treated as reassurance.
A brief bedside screen — the MMSE, the MoCA, or an equivalent — is done in most clinics. It is quick, it is repeatable, and its main value is trend: the same test repeated at intervals tells you far more than any single number. It is also imperfect. Education, language, hearing, vision and culture all move the score, and a highly educated person can fall substantially and still score within the normal range.
Formal neuropsychological testing is worth doing when the picture is early, or atypical, or when the answer will change what happens next. It takes longer and it produces something a screen cannot: a profile. Which domains are impaired, and which are intact, is what separates a memory disorder from an attention disorder, a language disorder from an executive one, and genuine decline from depression or from a person who is simply anxious in a testing room.
Alongside cognition, function is assessed deliberately — managing money and medicines, using transport and the telephone, cooking and shopping, and later bathing, dressing and toileting. That functional line is what separates mild cognitive impairment from dementia, and it is what the care plan is built on.
Imaging: what each modality adds
Structural imaging is part of every new assessment, and it earns its place by answering three separate questions at once. It is not there to produce a picture for its own sake.
| Modality | What it shows | What it does not do |
|---|---|---|
| Structural MRI | The pattern and degree of atrophy, including whether the medial temporal regions are disproportionately affected; the burden of small vessel disease; and any structural or reversible cause | It does not demonstrate Alzheimer's biology. Atrophy can be present without it, and the pattern can be ambiguous |
| CT | A rapid exclusion of haemorrhage, a large lesion or a subdural collection, and ventricular size | It is considerably less informative than MRI about atrophy patterns and small vessel disease, and it is a fallback rather than an equivalent |
| Amyloid PET | Whether amyloid plaques are present in the brain, in living people | It does not stage the disease, and a positive result in an unimpaired person is not a dementia diagnosis |
| FDG or perfusion PET | The regional pattern of reduced metabolism, which separates one dementia from another when the phenotype is atypical | It shows function rather than the specific protein pathology |
| Tau PET | The distribution and burden of tau, which tracks more closely with symptoms than amyloid does | Largely a research and selected-clinical tool; availability is limited |
| Dopamine transporter imaging | Whether there is loss of the dopaminergic nerve terminals, which separates Lewy body disease from Alzheimer's when that is the question | It answers one specific question and is ordered when that question is live |
Two imaging findings are worth naming because they change management directly. Marked ventriculomegaly out of proportion to the atrophy, in someone with the gait and bladder triad, raises normal pressure hydrocephalus — the item on the differential that can improve after treatment. And a heavy burden of small vessel disease moves the plan towards vascular prevention, which is a treatment rather than an observation.
Biomarkers: amyloid, tau and what a positive result means
This is the largest change in the field in a generation. Alzheimer's used to be a diagnosis of exclusion, confirmed only at autopsy. It can now be identified in life, and increasingly from a blood sample. That change is real and it is useful, and it is also routinely over-read in both directions.
What is being measured falls into three groups. Amyloid — measured on PET or in cerebrospinal fluid — indicates that the amyloid arm of the pathology is present. Tau, and specifically phosphorylated tau, tracks more closely with symptoms and with stage. Neurodegeneration is measured on MRI as atrophy, and in blood as neurofilament light chain. The framework used in research combines them; in clinic, what matters is that amyloid and tau answer different questions.
Blood phosphorylated tau — p-tau217 in particular — has performed well enough in research settings to change who needs a lumbar puncture or a PET scan, and it is becoming part of clinical practice. It is a step change in accessibility, and it is also the reason the interpretation rule matters more than it used to.
Biomarkers confirm biology. They do not by themselves diagnose dementia. A positive amyloid or tau result in someone with normal cognition is not a diagnosis and does not mean decline is inevitable; it means the biology is present, which is a different statement. Conversely, a biomarker result is read alongside the phenotype and the examination, never as a standalone answer. And the test is worth doing when the result would change management — an atypical presentation, an unexpectedly young onset, or eligibility for antibody treatment — rather than as a screening test in someone with no symptoms.
The practical consequence of a confirmed result is narrow but real: it is the entry criterion for anti-amyloid antibody treatment, which is discussed on the treatment page. The consequence of a negative result is also real, and often undervalued — it redirects the assessment towards the other causes, several of which are treatable.
The exclusion list, and why it is not a formality
Every new cognitive assessment includes a defined set of blood tests and a medicines review. They are cheap, they are quick, and they find the answers that matter most, which is why they come first rather than last.
| Cause | Test | If it is found |
|---|---|---|
| Vitamin B12 and folate deficiency | Serum B12, folate, and where the picture is unclear methylmalonic acid | Replacement. The cognitive change does not always fully reverse, which is the argument for testing early |
| Thyroid disease | Thyroid function testing | Treatment of the thyroid disorder, and reassessment afterwards |
| Metabolic and organ failure | Electrolytes, calcium, glucose, kidney and liver function, full blood count | Correction of the abnormality, which often produces a surprising improvement in attention |
| Infection and inflammation | Directed by the history and examination; serology where indicated | Treatment of the underlying infection. This is a small group and it is not one to miss |
| Medicine burden | A review that calculates the anticholinergic and sedative load rather than listing drugs | Reduction and substitution, with the cognitive effect reviewed weeks later |
| Sleep-disordered breathing | Sleep history, and a sleep study where snoring or witnessed apnoea is present | Treatment of the apnoea, which improves attention, mood and vascular risk together |
| Depression | A structured mood assessment, which is part of the assessment rather than an extra | Treatment, and a repeat of the cognitive testing afterwards before any label is settled |
| Normal pressure hydrocephalus | Imaging, plus cerebrospinal fluid assessment | Referral for the treatment decision. Of this list, the one most likely to improve substantially |
| Structural lesions | Structural MRI, and CT where MRI is not possible | Treatment of what is found; a subdural collection after a forgotten fall is the recurring example |
| Autoimmune encephalitis | Antibody testing and cerebrospinal fluid analysis, where the tempo or the phenotype warrants it | Immunotherapy rather than symptomatic treatment. Uncommon, treatable, and the reason tempo matters |
Where the picture is incomplete, the honest position is an active differential with a date to review it rather than a premature label. That position has a real advantage: it keeps the exclusion work running, and the exclusion work is what finds the treatable conditions. What to send ahead of an opinion is set out on the practical questions page.