Frontal Assessment Battery distinguishes frontotemporal dementia from Alzheimer’s disease

A simple bedside test that has been used for decades to probe the frontal lobes may be less powerful at separating the two most confusing dementias from each other than clinicians have long hoped—at least until the disease has advanced. That is the central finding of a large new study from a consortium of Milan-based research institutions, published in the Journal of Neurology, which examined whether the Frontal Assessment Battery, a widely used fifteen-minute cognitive screen, can reliably distinguish behavioral variant frontotemporal dementia from Alzheimer’s disease. The answer, drawn from one of the largest cohorts ever assembled for this question, is nuanced: the test is superb at detecting that something is wrong, but only modest at telling clinicians which of the two diseases is responsible, and its discriminatory power shifts dramatically with the stage of illness.

The Frontal Assessment Battery, known universally in neurology clinics as the FAB, was introduced in 2000 by Bruno Dubois and colleagues as a rapid bedside instrument for evaluating executive function—the family of mental abilities managed largely by the prefrontal cortex, including planning, inhibition, mental flexibility and the retrieval of words. The battery consists of six brief subtests: similarities, in which patients must identify abstract concepts shared by pairs of words; lexical fluency, in which they generate as many words as possible beginning with a given letter; motor series, a Luria-style sequencing task requiring the reproduction of a fist-edge-palm pattern; conflicting instructions, which demands that patients respond opposite to an examiner’s command; go-no-go, which measures impulse control; and prehension behavior, which probes for primitive grasping reflexes. Each subtest is scored out of three, giving a maximum of eighteen points. Because it is quick, cheap and requires no equipment, the FAB has been translated into dozens of languages and deployed across an extraordinary range of neurological conditions, from Parkinson’s disease to amyotrophic lateral sclerosis to Huntington’s disease.

The clinical problem the Italian team set out to address is one of the most persistent in dementia medicine. Alzheimer’s disease, the most common dementia, is classically a disorder of the temporal and parietal lobes, presenting first with memory loss for recent events. Behavioral variant frontotemporal dementia, or bvFTD, arises from degeneration of the frontal and anterior temporal lobes and announces itself not with forgetfulness but with a slow erosion of personality: disinhibition, apathy, compulsive behaviors, loss of empathy, dietary changes and poor judgment. In textbooks the two conditions look opposites, and the FAB was proposed early on as a discriminator—bvFTD patients were expected to collapse on frontal tests while Alzheimer’s patients, at least early on, would pass them. Yet real patients rarely read textbooks. Behaviorally disturbed patients can turn out to have Alzheimer’s pathology, and memory complaints can accompany frontotemporal degeneration. Several earlier studies, with smaller samples, came to conflicting conclusions about whether the FAB could separate the two diseases at all.

To resolve the question with adequate statistical power, the researchers assembled a retrospective cohort of 595 participants: 153 patients with Alzheimer’s disease whose diagnosis was supported by biomarkers—the fluid or imaging measures of amyloid and tau pathology that are now considered the biological gold standard—96 patients with probable bvFTD diagnosed under international consensus criteria, and 346 healthy controls, all of whom scored within the normal range on the Montreal Cognitive Assessment. This biomarker anchoring of the Alzheimer’s group is a methodological strength, since many previous studies relied on clinical diagnosis alone, a practice that can misclassify patients whose pathology does not match their symptoms. Disease severity in the patient groups was staged using Clinical Dementia Rating levels derived retrospectively from Mini-Mental State Examination scores, a validated mapping that allowed the team to ask a question rarely posed at this scale: does the FAB’s diagnostic value depend on how far the disease has progressed?

The analytical approach relied on receiver operating characteristic analysis, a statistical framework that evaluates a test by plotting its sensitivity against its specificity across all possible cutoff scores. The area under the resulting curve, or AUC, quantifies discriminatory accuracy on a scale from 0.5, equivalent to a coin flip, to 1.0, equivalent to perfect classification. Values between 0.7 and 0.8 are conventionally deemed adequate, 0.8 to 0.9 good, and above 0.9 excellent. The team calculated demographically adjusted FAB scores for every participant, then ran pairwise comparisons: each patient group against the healthy controls, and—critically—the two patient groups against each other, both as a whole and within severity strata. Complementary binary logistic regression models, adjusted for age and sex, were used to determine whether any individual subtest carried independent diagnostic information.

The first result was emphatic. In separating patients from healthy controls, the FAB performed excellently, confirming its value as a sensitive screen for the executive dysfunction that accompanies both Alzheimer’s disease and bvFTD. But the second result was sobering. When the comparison was restricted to patients only—Alzheimer’s versus bvFTD—the overall discriminatory accuracy was modest, falling well short of the thresholds clinicians would want before making a diagnosis on the basis of the test. The crucial twist was that this accuracy was not constant. Stratified by severity, the FAB distinguished the two diseases adequately only at the moderate-to-severe stages of dementia. In the earliest phases—patients with mild cognitive impairment or questionable dementia—the total score offered essentially no help in telling the two conditions apart.

The subtest analyses sharpened the picture further. Of the six components of the battery, only lexical fluency showed a statistically significant difference between the patient groups, with performance adjusted for demographic variables, and even that signal emerged specifically in patients at the earliest stage and in those at the moderate-to-severe stage. In the FAB’s lexical fluency task, patients must produce words beginning with a designated letter in sixty seconds, a task that loads heavily on left frontal-lobe systems governing strategic word retrieval and cognitive control over language output. Verbal fluency deficits have long been associated with frontotemporal degeneration, and prior FDG-PET imaging work has linked FAB performance in Alzheimer’s disease to frontal hypometabolism, but the new findings suggest that within this brief bedside battery, it is the fluency subtest alone—not the inhibition tasks, not the motor sequencing—that carries any differentiating weight between these two diagnoses. No significant differences emerged for any of the remaining five subtests.

The authors’ conclusion, stated with clinical candor, is that the FAB is an excellent screening test for executive dysfunction in both conditions but that its contribution to differential diagnosis emerges only at moderate-to-severe stages. The finding carries practical weight for clinicians and families alike. In real-world memory clinics, the moment of maximum diagnostic confusion is early in the illness, precisely when the FAB is least able to arbitrate. A first-degree relative watching a parent become uncharacteristically rude or passive wants an answer quickly; the new data indicate that a low FAB score confirms frontal impairment but cannot substitute for biomarker workup, structural imaging or multidisciplinary assessment when the question is which pathology is driving the change. Conversely, once dementia has progressed to a moderate or severe stage, the pattern of frontal test performance begins to align with the underlying disease in a way that can usefully support the clinical impression.

The study also adds to a growing body of evidence that the neat dichotomy between a “memory dementia” and a “behavior dementia” softens as both diseases advance. Neuroimaging studies have documented overlapping patterns of atrophy and white-matter deterioration in advanced Alzheimer’s and bvFTD, and recent biomarker-positive comparative studies have found the two conditions more alike in cognition and cortical thinning than their textbook portraits suggest. The prefrontal cortex, it appears, is not spared indefinitely in Alzheimer’s disease; as pathology spreads, executive dysfunction becomes the shared currency of both disorders, narrowing the behavioral and cognitive gap that a frontal bedside test is asked to detect.

For the field of dementia diagnostics, the results land amid an era of rapid change, with blood-based biomarkers and disease-modifying therapies reshaping early diagnosis. Yet brief cognitive instruments remain the first gate through which nearly every patient passes, and studies like this one—large, biomarker-anchored and severity-stratified—serve as essential calibration for what those instruments can and cannot promise. The FAB will continue to earn its place on the clinic cart as a sensitive detector of frontal-lobe dysfunction. What it will not do, the Milan consortium shows, is settle the Alzheimer’s-versus-bvFTD question early in the disease course, when that answer matters most.