The 2025 Alzheimer’s Association guideline addresses blood-based biomarkers for detecting Alzheimer’s disease pathology in patients with objective cognitive impairment, including mild cognitive impairment or dementia, undergoing evaluation in specialized memory-care settings. Based on a systematic review of 49 observational studies evaluating 31 biomarker-assay combinations, the panel issued two conditional recommendations, both supported by low-certainty evidence: blood tests with ≥90% sensitivity and ≥75% specificity may be used for triage, with positive results requiring confirmation by cerebrospinal fluid biomarkers or amyloid positron emission tomography; tests with ≥90% sensitivity and ≥90% specificity may be used as confirmatory tests, substituting for these established methods. The recommendations are based on test performance rather than specific brands. [1]
Diagnostic accuracy varied substantially, and many commercially available tests did not meet these thresholds, particularly when using a single cutoff. Evidence limitations included nonprespecified cutoffs, unclear blinding of test interpretation, and variable consistency and precision across assays. The guideline’s good practice statement specifies that testing should follow a comprehensive clinical evaluation, with results interpreted in the context of the patient’s presentation and pretest probability of Alzheimer’s pathology. The recommendations do not address screening cognitively unimpaired individuals, patients with only subjective memory complaints, or routine testing in nonspecialty settings. [1]
Blood testing may offer a less invasive, more accessible, and less costly diagnostic option, although cost-effectiveness remains uncertain and implementation depends on reimbursement, laboratory quality control, and appropriate follow-up. Extra caution is advised when other conditions or medications may affect biomarker concentrations. Biomarker combinations and two-cutoff strategies, which classify results as negative, positive, or indeterminate, were not formally evaluated in this initial guideline and were identified as priorities for future updates. [1]
A 2026 Cochrane systematic review evaluated amyloid-beta-targeting monoclonal antibodies in patients with mild cognitive impairment or mild dementia due to Alzheimer’s disease. It included 17 placebo-controlled randomized trials involving 20,342 participants and seven agents: aducanumab, bapineuzumab, crenezumab, donanemab, gantenerumab, lecanemab, and solanezumab. All trials were industry-funded, and most lasted 18 months. At 18 months, treatment produced little to no clinically meaningful difference in cognitive function (moderate-certainty evidence) or dementia severity (low-certainty evidence). Pooled differences favored treatment by 0.85 points on the Alzheimer’s Disease Assessment Scale-Cognitive and 0.29 points on the Clinical Dementia Rating-Sum of Boxes, both below the clinically important differences cited by the authors. Effects on functional ability were trivial to small, supported by moderate- or low-certainty evidence, depending on the assessment scale. [2]
Treatment increased amyloid-related imaging abnormalities involving edema by approximately 107 additional cases per 1,000 patients at 18 months (95% confidence interval, 77 to 148 additional cases; moderate-certainty evidence). Symptomatic edema occurred in approximately 29 additional patients per 1,000, although the authors classified this absolute effect as little to no difference. Results for imaging abnormalities involving hemorrhage were too heterogeneous to pool. There was no meaningful increase in overall serious adverse events or mortality at 18 months, supported by high-certainty evidence. Important limitations included inconsistent reporting of symptomatic imaging abnormalities, limited longer-term evidence, restrictive trial populations, and possible functional unblinding from recognizable adverse effects. [2]
The authors concluded that clinical benefits were trivial for cognition and dementia severity and small at best for functional ability, despite successful amyloid removal. Considering these effects, imaging-abnormality risks, and the resources required for administration and monitoring, they judged the benefit-risk balance unfavorable. [2]
Two 2025 meta-analyses evaluated blood-based biomarkers for detecting Alzheimer’s disease pathology against established biological reference standards. One analysis included 113 studies comprising 29,625 unique individuals and identified plasma phosphorylated tau (p-tau)217 as the highest-performing p-tau biomarker, with pooled sensitivity of 88.1%, specificity of 88.7%, and area under the receiver operating characteristic curve of 91.1%, each supported by moderate-certainty evidence; p-tau sensitivity was higher in cognitively impaired than in unimpaired individuals, whereas specificity did not differ. Approximately 90% of included studies were rated at high risk of bias because they lacked predefined or externally derived thresholds, and the authors called for prospective studies evaluating effects on diagnosis and clinical management. The other paper evaluated 49 observational studies examining 31 blood-based tests, including p-tau217, percentage p-tau217, p-tau181, p-tau231, and amyloid-β42/40, in individuals with mild cognitive impairment or dementia within specialized care settings. Using a single cutoff per test, pooled sensitivity ranged from 49.3% to 91.4%, and specificity from 61.5% to 96.7%; evidence certainty ranged from moderate to very low, and most studies were judged at high risk of bias. This review informed an accompanying Alzheimer’s Association clinical practice guideline, and the authors advised interpreting results according to the specific assay and integrating them with a comprehensive clinical assessment. [3], [4]
A 2026 meta-analysis examined how patient characteristics influence the efficacy and safety of lecanemab and donanemab (anti-amyloid monoclonal antibodies for early Alzheimer’s disease). The analysis included 16 publications representing six randomized clinical trials with 5,633 participants, with literature included through July 30, 2025. Participants had mild cognitive impairment or mild dementia due to Alzheimer’s disease and were generally followed for 18 months or 76 weeks. Both treatments significantly slowed cognitive decline compared with placebo, although the authors noted that the treatment differences did not reach the cited 1-point threshold for clinically important change on the Clinical Dementia Rating-Sum of Boxes (CDR-SB). The comparisons between lecanemab and donanemab were drawn from separate trials rather than a direct head-to-head cognitive efficacy trial. [5]
However, treatment effects varied across subgroups. On the CDR-SB, where negative treatment-placebo differences indicate less worsening, lecanemab produced differences of -0.35 points in mild cognitive impairment and -0.62 points in mild dementia, compared with -0.29 and -0.68 points, respectively, for donanemab. Lecanemab’s treatment difference was larger in men than women (-0.73 versus -0.20 points), with no statistically significant benefit identified in women; donanemab showed significant benefit in both sexes (-0.51 points in men and -0.79 points in women). Lecanemab showed significant benefit in participants aged 65 years or older, whereas donanemab showed benefit across the reported age groups. By apolipoprotein E4 (ApoE4) genotype, treatment differences for lecanemab were -0.75 points in noncarriers, -0.50 in heterozygotes, and +0.28 in homozygotes, with no demonstrated benefit in homozygotes. Corresponding differences for donanemab were -0.76, -0.73, and -0.41 points, with benefit reported across genotype groups. [5]
Safety analysis found an increased risk of amyloid-related imaging abnormalities (ARIA), especially among ApoE4 carriers. Compared with placebo, pooled relative risks for ARIA with edema or effusion (ARIA-E) were 7.96 for lecanemab (95% confidence interval [CI] 4.84-13.11) and 13.08 for donanemab (95% CI 8.16-20.96). For ARIA with microhemorrhages (ARIA-H), relative risks were 1.84 (95% CI, 1.45-2.34) and 2.65 (95% CI, 2.04-3.44), respectively. Across treatments, ApoE4 carriers had 2.19 times the ARIA-E risk and 3.45 times the ARIA-H risk of noncarriers. Pooled ARIA-E/ARIA-H prevalence was 10%/9% in noncarriers, 17%/22% in heterozygotes, and 40%/46% in homozygotes. Overall adverse-event and serious-adverse-event risks were comparable to placebo in the pooled analyses; infusion reactions were particularly common with lecanemab, whereas ARIA was generally more prevalent with donanemab. [5]
Quality-of-life evidence was limited to lecanemab and showed statistically significant improvements in selected outcomes among participants aged 65-74 years and ApoE4 heterozygotes, alongside smaller increases in caregiver burden across several subgroups. The paper’s main contribution to recent treatment developments is its emphasis on variation in benefit and risk by patient characteristics, particularly ApoE4 genotype. However, cognitive and quality-of-life outcomes could not be pooled because subgroup data were insufficient, racial and ethnic minority populations were underrepresented, and long-term evidence remained limited. [5]
Pertaining to disease pathology, a 2026 meta-analysis including 42 independent cohorts evaluating soluble triggering receptor expressed on myeloid cells 2 (sTREM2) and 44 evaluating chitinase-3-like protein 1 (YKL-40), found that these neuroinflammatory biomarkers were more strongly associated with tau pathology and neurodegeneration than with amyloid accumulation. Pooled correlations for sTREM2 versus YKL-40 were 0.41 versus 0.50 for phosphorylated tau, 0.41 versus 0.53 for total tau, and 0.31 versus 0.55 for neurofilament light chain. Associations with amyloid were weaker and generally attenuated after adjustment. Relationships with tau weakened from cognitively unimpaired individuals to mild cognitive impairment and Alzheimer’s disease, particularly for sTREM2, suggesting that protective glial responses may become less effective as disease progresses. Limited longitudinal evidence linked higher sTREM2 to increasing soluble tau but slower accumulation of amyloid and tau aggregates. For treatment development, the authors proposed investigating stage-specific therapies that restore protective glial function, with potential future use of these biomarkers for patient selection and treatment monitoring. However, the review did not evaluate treatment efficacy, and its observational findings do not establish causality. Substantial heterogeneity, limited longitudinal evidence, and variable assay methods require further validation before clinical application. [6]
A 2025 Food and Drug Administration (FDA) announcement reported the clearance of the Lumipulse G pTau217/β-Amyloid 1-42 Plasma Ratio, the first blood-based in vitro diagnostic device cleared to aid in diagnosing Alzheimer’s disease in adults aged 55 years and older exhibiting signs and symptoms of the disease. The test measures plasma pTau217 and β-amyloid 1-42 and calculates their ratio to assess the presence or absence of amyloid plaques. The FDA evaluated a multicenter clinical study involving 499 plasma samples from cognitively impaired adults, comparing blood test results with amyloid positron emission tomography (PET) or cerebrospinal fluid (CSF) test results. Among individuals with positive blood test results, 91.7% had amyloid plaques indicated by PET or CSF testing; among those with negative results, 97.3% had negative PET or CSF results. Fewer than 20% received indeterminate results. The test is intended for patients presenting to specialized care settings with signs and symptoms of cognitive decline, and results must be interpreted alongside other clinical information. It is not intended for screening or as a stand-alone diagnostic test; false-positive and false-negative results may lead to inappropriate diagnosis or treatment or delay effective treatment. [7]