Long term clinical performance and analytical robustness of the Lumipulse plasma p‐tau217 assay in routine clinical practice

Abstract

INTRODUCTION

Plasma phosphorylated tau 217 (p‑tau217) immunoassays are increasingly being used to assess Alzheimer's disease (AD) pathology despite limited data affirming test reliability over time.

METHODS

Clinical performance of the Fujirebio Lumipulse G p‑tau217 assay was assessed in 744 Mayo Clinic patients presenting for evaluation of cognitive impairment between July 2024 and December 2025. Cerebrospinal fluid p‑tau181/amyloid beta 42 or amyloid positron emission tomography, within 1 year of p‐tau217 testing, were used as reference. Analytical robustness was evaluated by tracking result frequencies over time, conducting quality control checks, performing 10 lot‑to‑lot comparisons, and comparing two versions of reagents.

RESULTS

Diagnostic accuracy for detection of AD pathology was 93% (95% confidence interval 91–95) with 23% intermediate results. Reduced estimated glomerular filtration rate was associated with increased intermediate results among amyloid‐negative individuals. Result frequencies remained stable, average imprecision was 8%, and analytical bias between lots and version changes were ≤ 11%.

DISCUSSION

The findings support plasma p‐tau217 reliability in clinical practice for evaluation of cognitively impaired individuals.

Highlights

• Plasma phosphorylated tau (p‐tau)217 clinical performance was evaluated in routine clinical practice over 18 months.

• Plasma p‐tau217 demonstrated 93% accuracy in predicting amyloid pathology.

• Reduced kidney function increased intermediate results among amyloid‐negative patients.

• Analytical performance of the Lumipulse p‐tau217 assay remained stable over an 18‐month period.

1. BACKGROUND

Plasma phosphorylated tau at position 217 (p‐tau217) has emerged as a leading blood biomarker for detecting Alzheimer's disease (AD) pathology ¹ , ² and is increasingly applied in the diagnostic evaluation of individuals with cognitive impairment. Plasma p‐tau217 assays are rapidly being adopted in clinical laboratory workflows, using fully automated immunoassay analyzers that deliver affordable, precise, and fast results. Platforms such as Fujirebio Lumipulse (p‐tau217/amyloid beta [Aβ]42) and Roche Elecsys (p‐tau181) have received US Food and Drug Administration (FDA) clearance for p‐tau–based assays for the assessment of amyloid pathology. ³ , ⁴ In addition, most recently, several plasma p‐tau217 immunoassay methods have received regulatory approval in various countries, including assays from Fujirebio Lumipulse, Roche Elecsys, and Beckman Coulter. ⁵ , ⁶ , ⁷ , ⁸

Successful clinical use of these assays as routine clinical chemistry tests requires not only careful evaluation of initial test performance and selection of appropriate interpretive cut‐points, but also ongoing monitoring of assay performance over time in routine clinical settings. The importance of ongoing validation is emphasized by reports documenting a significant deviation from the manufacturer's clinical performance claims for the FDA‐cleared Fujirebio Lumipulse p‐tau217/Aβ42 ratio during the initial evaluation of this assay in a clinical cohort. These findings resulted in a temporary assay recall by the manufacturer with unclear consequences for patients and providers whose care and counselling may have been informed by inaccurate results. ⁹ , ¹⁰

Prior to regulatory approval of these assays, several clinical laboratories provided the Lumipulse G pTau217 assay as laboratory‐developed tests (LDTs) using laboratory‐established thresholds for interpretation. ¹¹ , ¹² , ¹³ The Lumipulse G pTau217 assay, when applied using a two cut‐point model to classify results as negative, intermediate, or positive, has demonstrated high diagnostic accuracy for predicting brain amyloid pathology. ² , ¹¹ , ¹⁴ Although prior published work is foundational for establishing expected assay performance, the transition to routine clinical practice introduces new variables, including patient populations with multiple medical comorbidities, ¹⁵ varying pre‐analytical handling, ¹⁶ and reagent lot variability, ¹² , ¹⁷ all of which may impact assay performance and the diagnostic accuracy of previously established clinical cut‐points.

Periodic assessment of analytical robustness is critical to maintaining high‐quality clinical laboratory testing over time. This is especially important following modifications or other changes to reagent lots, which may affect assay performance. ¹⁸ In this study, we aimed to evaluate the clinical performance and analytical robustness of the Fujirebio Lumipulse G p‐tau217 immunoassay across time in patients assessed at the Mayo Clinic. Specifically, we sought to demonstrate the consistency of clinical results using our previously established cut‐points ¹¹ and to assess assay stability, including the effects of assay version changes and lot‐to‐lot variability over an 18‐month period.

2. METHODS

2.1. Participants

This retrospective study included 744 unique patients assessed at the Mayo Clinic in Rochester (Minnesota), Jacksonville (Florida), or Scottsdale (Arizona) between July 2024 and December 2025 for symptoms of cognitive impairment. The sample set included two cohorts. Cohort 1 included 275 patients who underwent amyloid positron emission tomography (PET) and/or cerebrospinal fluid (CSF) p‐tau181/Aβ42 as part of their clinical evaluation and were asked to provide a blood sample for plasma p‐tau217 testing under research protocols. Cohort 2 included 469 patients who underwent plasma p‐tau217 testing as part of their clinical evaluation, with test results extracted from the electronic medical records. Inclusion criteria for both cohorts required confirmatory amyloid status assessment using either CSF p‐tau181/Aβ42 and/or amyloid PET within 1 year of blood collection. Patients’ demographics and clinical characteristics were obtained from electronic medical records. Further details on sample selection and exclusions are described in Figure S1 in supporting information. The study was approved by the Mayo Clinic Institutional Review Board (IRB protocol numbers 23‐011176, 23‐007112, and 26‐000654). All study procedures were conducted in accordance with institutional requirements and applicable ethical standards.

2.2. Samples

K₂ ethylenediaminetetraacetic acid (EDTA) plasma samples were processed and stored according to our internal standard operating procedures. Briefly, blood specimens were centrifuged, and plasma separated within 2 hours of collection. Samples from Cohort 1 were stored frozen at −80°C until testing occurred within 2 weeks of sample collection. For Cohort 2, samples were immediately transferred to the laboratory with plasma measures obtained within 24 hours of collection. Prior to testing, all samples were mixed and centrifuged for 5 minutes at 4000 × g, consistent with validated standard operating procedures for the clinical LDT.

2.3. Plasma p‐tau217 measurements

Plasma p‐tau217 concentrations were measured using the Fujirebio Lumipulse G p‐tau217 immunoassay on the Lumipulse G1200 automated immunoassay analyzer (Fujirebio, Inc.). This test has been analytically and clinically validated as a LDT per College of American Pathologists (CAP)/Clinical Laboratory Improvement Amendments (CLIA) requirements. Clinical performance of the test has been previously described. ¹¹ During the study period, two versions of the assay were used: version 1 (v1; catalog: 81472; 6 lots) from May 2024 until August 2025, and version 2 (v2; catalog: 81550 [IVD] and 81472 [RUO]; 4 lots) from August 2025 through December 2025. Testing was performed in a CLIA‐certified, CAP‐accredited clinical laboratory at Mayo Clinic, Rochester, Minnesota, USA. Results are expressed in pg/mL. Before introducing the assay in clinical practice, cut‐points were selected to achieve ≥ 90% sensitivity and ≥ 95% specificity ¹¹ and were subsequently validated in an independent cohort. ¹⁴ Results are classified as negative, intermediate, or positive using previously established clinical assay cut‐points: negative ≤ 0.185 pg/mL, intermediate 0.186 to 0.324 pg/mL, positive ≥ 0.325 pg/mL. ¹¹

RESEARCH IN CONTEXT

Systematic review: The authors reviewed the literature to identify studies evaluating the clinical and analytical performance of phosphorylated tau (p‐tau)217 in routine clinical practice. Data describing clinical performance and analytical robustness of the Fujirebio Lumipulse plasma p‐tau217 assay in routine clinical practice remain limited.

Interpretation: In 744 patients evaluated for cognitive impairment with amyloid status established by cerebrospinal fluid and/or amyloid positron emission tomography, plasma p‐tau217 showed high diagnostic accuracy for amyloid pathology and supported clinically actionable results. Reduced kidney function was associated with more intermediate results among amyloid‐negative individuals, indicating renal function is an important consideration when interpreting results. Quality control, lot‐to‐lot, and reagent version assessments supported stable analytical performance over 18 months of clinical implementation.

Future directions: Multi‐center studies are needed to confirm generalizability across settings with different amyloid prevalence, pre‐analytical workflows, and reader approaches for amyloid pathology assessment.

2.4. Reference methods for amyloid status

Amyloid status was determined by CSF p‐tau181/Aβ42 (Roche Diagnostics; ratio > 0.028 considered positive) or 18F‐florbetapir amyloid PET imaging (positivity assessment by visual read). In participants who completed testing with both modalities, a positive outcome from either method was regarded as amyloid positive (A+). ¹⁹ , ²⁰

2.5. Assessment of kidney function

Serum creatinine and estimated glomerular filtration rate (eGFR) measurements performed within 6 months of plasma p‐tau217 collection were extracted from the medical records. eGFR calculation was based on the 2021 Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) creatinine equation. ²¹ The association between eGFR and plasma p‐tau217 concentrations were assessed using the Spearman correlation coefficient. An eGFR of 45 and 60 mL/min/1.73 m² was used to assess the effect of kidney function on plasma p‐tau217 result classifications. ²² A Fisher exact test was used to determine statistical differences between plasma p‐tau217 intermediate and positive results in amyloid‐negative (A−) individuals with eGFR < 45 mL/min/1.73 m², those with eGFR 45 to 59, and those with eGFR 60+ mL/min/1.73 m². P values were adjusted using a Bonferroni correction. ²³

2.6. Quality control measurements

Two levels of manufacturer‐provided quality control (QC) materials (4 lots) and one level of in‐house plasma pool control (2 lots), run twice daily, were analyzed from May 2024 through December 2025. Levey–Jennings charts were generated, and monthly means, standard deviations (SDs), and coefficients of variation (%CV) were calculated for each quality control level. Additionally, %CV was calculated per reagent lot before and after the removal of > 3 SD results for each QC lot.

2.7. Frequency monitoring

To evaluate the assay performance over time, the mean frequency of negative, intermediate, and positive p‐tau217 results was plotted from May 2024 to December 2025, spanning 10 different reagent lots and two different assay versions. Expected frequencies were set to 33%, 20%, and 47% for negative, intermediate, and positive results, respectively, based on the initial clinical validation study. ¹¹

2.8. Reagent lot and version difference evaluation

To evaluate lot‐to‐lot consistency, data from the 10 different reagent lots used during the study period were compared. New lots were evaluated using clinical residual plasma EDTA samples run by both reagent lots (n > 10). Passing–Bablok regression analysis was performed, and the regression equation and Pearson r were evaluated to assess for significant bias or lack of correlation between p‐tau217 lots.

To evaluate differences between assay versions, 187 plasma samples with concentrations spanning the analytical measuring range were tested on both v1 and v2 of the assay. Passing–Bablok regression analysis was performed, and individual differences were plotted to assess agreement between versions.

2.9. Statistical analysis

The diagnostic performance of plasma p‐tau217 was assessed, referencing amyloid status (positive/negative) established by amyloid PET (visual read) or CSF p‐tau181/Aβ42 (> 0.028 positive). Clinical sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and prevalence were determined. Intermediate biomarker concentrations were excluded from these analyses, implying intermediate results require further evaluation. Additionally, overall, positive, and negative percent agreement (OPA, PPA, and NPA) between amyloid PET and CSF p‐tau181/Aβ42 were determined. Statistical analyses were performed using Analyse‐it version 6.15 for Microsoft Excel, or the R language and environment for statistical computing version 4.5.2.

3. RESULTS

3.1. Participant characteristics

Patient demographics are shown in Table 1. Of the 744 individuals presenting with symptoms of cognitive decline, 500 (67%) were A+ and 244 (33%) were A−. The median age was 73 years (range 36–90), and 68% were male. Amyloid pathology was determined by amyloid PET in 486 (65%) and CSF p‐tau181/Aβ42 in 334 (45%), with 76 participants having both modalities performed. Global Clinical Dementia Rating (CDR) was available in a subset of patients (n = 325; 44%), among whom 94% (305) had a CDR Global between 0.5 and 1. eGFR data were available from 634 participants; 96% had eGFR ≥ 45 mL/min/1.73 m²; 4% had eGFR < 45. Nearly all clinical orders for plasma p‐tau217 testing originated from neurology departments (722/744, 97%).

TABLE 1.

Patients demographic and clinical characteristics.

Characteristic	Total cohort (n = 744)	Amyloid negative (n = 244)	Amyloid positive (n = 500)
Median age, years (range)	73 (36–90)	70 (36–88)	73 (42–90)
Male, n (%)	507 (68%)	175 (72%)	332 (66%)
APOE ɛ4 allele, n (%)	393 (53%)	47 (19%)	346 (69%)
ε4 null	133 (18%)	29 (12%)	104 (21%)
ε4 heterozygous	209 (28%)	16 (7%)	193 (39%)
ε4 homozygous	51 (7%)	2 (1%)	49 (10%)
p‐tau217, pg/mL, median (IQR)	0.368 (0.202–0.681)	0.154 (0.104–0.241)	0.548 [(0.340–0.859)
Amyloid PET (visual read) – N/A, n (%)	258 (35%)	162 (66%)	96 (19%)
Amyloid PET (visual read) – positive, n (%)	400 (54%)	0 (0%)	400 (80%)
Amyloid PET (visual read) – negative, n (%)	86 (12%)	82 (34%)	4 (1%)
CSF p‐tau181/Aβ42 ratio – N/A, n (%)	410 (55%)	73 (30%)	337 (67%)
CSF p‐tau181/Aβ42 ratio – positive, n (%)	157 (21%)	0 (0%)	157 (31%)
CSF p‐tau181/Aβ42 ratio – negative, n (%)	177 (24%)	171 (70%)	6 (1%)
CDR Global, n (%)	325 (44%)	53 (22%)	272 (54%)
0, n (%)	7 (2%)	2 (4%)	5 (2%)
0.5 or 1, n (%)	305 (94%)	49 (92%)	256 (94%)
> 1, n (%)	13 (4%)	2 (4%)	11 (4%)
MMSE test, n (%)	486 (65%)	139 (57%)	347 (69%)
MMSE score, median (IQR)	26 (23‐27) 23 , 24 , 25 , 26 , 27	27 (24‐28)	25 (23‐27) 23 , 24 , 25 , 26 , 27
eGFR ≥45, n (%)	611(96%)	197 (95%)	414 (97%)
eGFR < 45, n (%)	23 (4%)	10 (5%)	13 (3%)

Abbreviations: APOE, apolipoprotein E; Aβ, amyloid beta; CDR, Clinical Dementia Rating; CSF, cerebrospinal fluid; eGFR, estimated glomerular filtration rate; IQR, interquartile range; MMSE, Mini‐Mental State Examination; PET, positron emission tomography; p‐tau, phosphorylated tau.

3.2. Diagnostic performance

Plasma p‐tau217 concentrations were positive (≥ 0.325 pg/mL) in 421/744 (57%) patients, negative (≤ 0.185 pg/mL) in 154/744 (21%) patients, and intermediate (0.186–0.324 pg/mL) in 169/744 (23%) patients (Table 2). Limiting analyses to patients with definitive results, the combined dataset showed a sensitivity of 98% (95% confidence interval [CI]: 95–99) and specificity of 84% (95% CI: 77–89). At a 67% amyloid prevalence, the PPV was 92% (95% CI: 90–95), NPV was 94% (95% CI: 90–97), with an accuracy of 93% (95% CI: 91–95). Considering amyloid PET only as the reference method, sensitivity was 99% (95% CI: 97–100) and specificity 69% (95% CI: 55–81). Considering CSF p‐tau181/Aβ42 as the reference, sensitivity was 95% (95% CI: 90–98) and specificity 85% (95% CI: 78–91). For the subset of samples with amyloid PET and CSF p‐tau181/Aβ42 available (n = 76), agreement between these two modalities was 90% (57/63) PPA and 69% (9/13) NPA, with an OPA of 87% (66/76; Table S1 in supporting information).

TABLE 2.

Clinical performance of the plasma p‐tau217 assay.

	Amyloid status			Amyloid PET			CSF p‐tau181/Aβ42
	n = 744			n = 486			n = 334
Plasma p‐tau217	A−	A+	%	A−	A+	%	A−	A+	%
Negative	144	10	21%	36	4	8%	115	6	36%
Intermediate	72	97	23%	34	72	22%	42	35	23%
Positive	28	393	57%	16	324	70%	20	116	41%
Sensitivity (95% CI)	98% (95–99)			99% (97–100)			95% (90–98)
Specificity (95% CI)	84% (77–89)			69% (55–81)			85% (78–91)
PPV (95% CI)	92% (90–95)			94% (91–96)			85% (79–90)
NPV (95% CI)	94% (90–97)			92% (82–97)			95% (90–98)
Accuracy (95% CI)	93% (91–95)			94% (91–96)			90% (85–93)
Prevalence	67%			82%			47%

Note: There were 76 patients with both amyloid PET and CSF p‐tau181/Aβ42 available. These are included in their respective categories.

Abbreviations: Aβ, amyloid beta; CI, confidence interval; CSF, cerebrospinal fluid; NPV, negative predictive value; PET, positron emission tomography; PPV, positive predictive value; p‐tau, phosphorylated tau.

The effect of kidney function on p‐tau217 concentrations was evaluated (Figure 1 and Figure S2 in supporting information). A modest correlation was identified between p‐tau217 and eGFR across all individuals (n = 634) (Spearman ρ = −0.188, P < 0.001). Reduced eGFR resulted in an increased intermediate and positive plasma p‑tau217 result classifications among A− individuals (Figure 1). The proportion of intermediate results in A− individuals with eGFR < 45 versus those with eGFR 60+ showed a statistically significant difference (Bonferroni‐adjusted P = 0.008), whereas the proportion of intermediate or positive results in A– individuals were not significant after Bonferroni adjustment when comparing other groups: intermediate eGFR < 45 versus 45 to 59, P = 0.0569; intermediate eGFR 45 to 59 versus 60+, P = 1.00; positive eGFR < 45 versus 60+, P = 1.00; positive eGFR < 45 versus 45 to 59, P = 1.00; positive eGFR 45 to 59 versus 60+, P = 0.131.

FIGURE 1.

Effect of kidney function on plasma p‐tau217 result classification. Correlation between eGFR (mL/min/1.73m²) and plasma p‐tau217 concentrations (pg/mL) are shown. One result with a very high p‐tau217 of 5.294 pg/mL was omitted due to truncation of the y axis; for full figure see Figure S2 in supporting information. Red dotted lines denote p‐tau217 cut‐points (≤ 0.185 and ≥ 0.325 pg/mL) and eGFRs of 45 and 60 mL/min/1.73m². The Spearman correlation coefficient (ρ) was −0.188 (95% CI: −0.265 to −0.110, P ≤ 0.0001). Result classification is shown for all individuals with an eGFR result within 6 months of p‐tau217 collection. Result classification differences are shown between individuals with an eGFR < 45, 45 to 59, and 60+. Intermediate results in A− eGFR < 45 versus eGFR 60+ showed a significant difference, P = 0.008. CI, confidence interval; eGFR, estimated glomerular filtration rate; IQR, interquartile range; MMSE, Mini‐Mental State Examination; PET, positron emission tomography; p‐tau, phosphorylated tau

3.3. Analytical robustness

QC performance across four manufacturer quality control lots and two in‐house QC lots exhibited %CVs ranging from 5% to 8% for quality control level 1 with mean concentrations of 0.452 to 0.472 pg/mL and 6% to 8% for quality control level 2 with mean concentrations of 3.6 to 3.9 pg/mL (Figure 2). An in‐house QC pool targeting a concentration close to the negative cut‐point of 0.186 pg/mL (mean concentrations 0.148–0.194 pg/mL) exhibited %CV of 12% to 14% over the 18‐month period. Lot‐specific QC analysis suggests that certain reagent lots contribute disproportionately to the observed imprecision, particularly for the in‐house QC (Table S2 in supporting information). Consistency across reagent lot changes (n = 10) was evaluated using clinical residual samples run by both lots. Passing–Bablok regression analysis showed excellent correlation between reagent lots (R ² > 0.990) and biases ranging from ± 2% to 11% as assessed by slopes ranging from 0.98 to 1.11 (Table S3 in supporting information). In addition, the effect of an assay version change introduced by the manufacturer in Q3 2025 was evaluated prior to clinical implementation. Method comparison (n = 187) showed a Passing–Bablok regression equation of (p‐tau217 v2) = 0.979 × (p‐tau217 v1) − 0.00132, R ² = 0.99, and a mean −4.35% bias between versions (Figure 3), indicating that a clinically meaningful bias was not introduced with the reagent version change.

FIGURE 2.

Quality control performance over 18 months. Two levels and four different lots of manufacturer QC material as well as two lots of one in‐house patient pooled QC level were analyzed twice per day (start and end of testing) during clinical testing days from May 2024 through December 2025. Each level is plotted over time with the mean, 1 SD, 2 SD, and 3 SD shown for each QC lot. Vertical black dashed lines denote when new lots of p‐tau217 reagent went into use. Table shows the QC mean, SD, %CV, and number of measurements for each QC lot used. CV, coefficient of variation; p‐tau, phosphorylated tau; QC, quality control; SD, standard deviation

FIGURE 3.

Comparison between version 1 and 2 of the plasma p‐tau217 reagent. Passing–Bablok regression and difference plot analysis was performed to evaluate the differences between reagent lot versions. The solid red line denotes the Passing–Bablok fit, n = 187, y = 0.979x−0.00132, R ² = 0.99, while dashed red lines denote the 95% confidence interval. The blue solid line in the difference plot denotes the mean % difference of −4.35% (bias). p‐tau, phosphorylated tau

The stable performance demonstrated by these quality measures translated to stable result frequencies over the 18‐month period (Figure 4). Mean negative, intermediate, and positive frequencies were 36%, 18%, and 46%, respectively, with an average difference of 3%, 2%, and 1% from the target value month over month, respectively.

FIGURE 4.

Result classification frequencies over the 18‐month period. The frequency (percent) of patients with negative, intermediate, and positive results is plotted by month over the study period from May 2024 through December 2025. The dashed red line indicates the expected % of results in each category based on Figdore et al. ¹¹

4. DISCUSSION

This study reports on the reliability, clinical performance, and analytical robustness of the Fujirebio Lumipulse G p‐tau217 immunoassay in a routine clinical practice setting at the Mayo Clinic over an 18‐month period. We demonstrate that the assay delivered consistent results in patients with cognitive impairment, despite changes in reagent lots and assay versions. ,  ,  ,  

The overall diagnostic performance (98% sensitivity and 84% specificity) observed in this routine clinical practice cohort was consistent with performance reported earlier in an outpatient memory clinic setting (95% sensitivity and 82% specificity). ¹⁴ Considering the subgroup of individuals with amyloid PET as the comparator, the observed clinical sensitivity and specificity were lower than reported by Figdore et al. ¹¹ A potential reason for these discrepancies is the use of a different amyloid PET tracer (18F‐florbetapir) in this study versus ¹¹ C‐Pittsburgh compound B, and differences in interpretation (visual read in routine clinical practice versus Centiloid quantification in a more controlled research setting). Discordance between Centiloid and visual‐read interpretation of amyloid PET has been reported to range from 7% to 14% in recent studies. ²⁴ , ²⁵ Furthermore, the agreement between CSF and visual‐read amyloid PET in this study showed a relatively low NPA of 69% and an OPA of 87%. Our findings are in line with large cohort studies that have shown 10% to 20% discordance between amyloid PET and CSF biomarkers, particularly in early or heterogeneous disease states. ²⁶ , ²⁷ Discrepancies between CSF and amyloid PET may reflect biological differences in amyloid pathology. Whereas CSF biomarker changes are often detected earlier than amyloid PET changes (reflecting soluble amyloid metabolism and clearance), amyloid PET reflects fibrillar plaque deposition. ²⁸ These differences may inflate the apparent false‐positive or false‐negative rates of blood biomarkers and confound reported diagnostic performance. Furthermore, when blood biomarker performance is benchmarked against one reference method but not the other, the apparent sensitivity, specificity, and “false‐positive” rates can change substantially—not because the blood test is inaccurate, but because the reference methods measure partially different processes of amyloid pathology.

The diagnostic accuracy of plasma p‐tau217 measures was 93% in our cohort with amyloid pathology prevalence of 67%, with a PPV of 92% and NPV of 94%, based on pre‐established plasma p‐tau217 cut‐points and meeting the minimum performance requirements for blood‐based biomarkers. ²⁰ , ²⁹ , ³⁰ These results are consistent with the performance reported in two additional cohorts with similar amyloid prevalence. ¹¹ , ¹⁴ The high PPV and NPV observed in this study provide clinicians with confidence when applying p‐tau217 to rule in or out underlying amyloid pathology in a clinical population, supporting the clinical utility of plasma p‐tau217 as an effective front‐line biomarker to guide further diagnostic evaluation and patient management. Intermediate results were more frequent in the Mayo Clinic cohort than among all samples tested in the laboratory during the study period (23% vs. 18%, respectively). Although the reason for this difference was not formally investigated, it highlights the importance of understanding factors that may shift p‐tau217 concentrations into the intermediate range. Reduced kidney function is one such factor and may have contributed to the increase in intermediate results.

Consistent with prior studies, reduced kidney function resulted in increased plasma p‐tau217 concentrations, including among individuals classified as amyloid negative by methods that are not expected to be affected by kidney impairment, such as amyloid PET and CSF biomarkers. ¹⁴ , ²² , ³¹ The use of the two cut‐point model resulted in most A− individuals with an eGFR < 45 mL/min/1.73 m² being classified as intermediate rather than positive. This demonstrates an advantage to the use of a two cut‐point model over a single cut‐point in routine clinical practice, as it may mitigate false‐positive interpretations related to renal dysfunction by triggering confirmatory testing.

Stable analytical test performance and reagent stability are as critical as the clinical performance of the assay. Despite using two versions of Fujirebio Lumipulse research use only reagents, the analytical data presented here (lot‐to‐lot performance, method comparison, and result frequencies) suggest that the manufacturing process of the research use only reagents has been stable over time and has not impacted test performance. In addition, our QC evaluation showed good analytical precision over the course of clinical testing, even when using an in‐house plasma QC pool that is more reflective of the patient sample matrix compared to manufacturer‐provided quality control material.

Our findings expand on the recent 6‐month real‐world evaluation by Arslan et al. of the Lumipulse G plasma p‐tau217 assay in a specialized clinical neurochemistry laboratory. ¹² In 121 individuals with paired CSF and plasma results, the authors evaluated the agreement between plasma p‐tau217 and CSF Aβ42/40 and reported sensitivity and specificity of 85% and 88%, respectively, after exclusion of indeterminate results. In comparison, our analysis based on CSF p‐tau181/Aβ42 as a reference method showed 95% sensitivity and 85% specificity. Differences in the CSF reference methods, cohort differences, and the use of locally established cut‐points may contribute to the observed differences in sensitivity, but nevertheless both studies support robust real‐world clinical performance of Lumipulse G plasma p‐tau217 in specialized laboratory settings. Stable analytical performance during routine use was observed in both laboratories with coefficients of variation < 8% for the manufacturer‐provided QC material and no clinically meaningful reagent lot‐to‐lot variability. By evaluating a larger clinically characterized cohort (n = 744) over 18 months, across 10 reagent lots and two assay versions, our study provides complementary evidence supporting the robustness of the assay in routine clinical practice.

Our study was conducted in a tertiary specialty center, with most testing requested by highly specialized behavioral neurologists. Undoubtedly, the testing environment influences pretest probability of amyloid positivity and thereby the interpretation of blood biomarker testing results and generalization of findings. ³² Our findings may not be fully generalizable to patient populations seen in less specialized neurology practices, including populations with lower prevalence of amyloid positivity. As PPV and NPV are strongly influenced by disease prevalence, the reported diagnostic performance may not directly translate to primary care settings or more heterogeneous patient populations with lower pre‐test probability. Another reality is that amyloid PET visual‐read interpretations were performed at three different Mayo Clinic campuses (Minnesota, Florida, Arizona). The multisite design introduces variability based on differences in local expertise, although it better mirrors application and interpretation of test findings in practice. Additionally, amyloid status was determined using two different reference standards, which measure overlapping but distinct aspects of amyloid pathology and therefore introduce unavoidable heterogeneity into performance estimates, and discordant PET and CSF cases were classified as amyloid positive without further adjudication, which may overestimate false‐positive rates. Last, although plasma remains the most commonly used specimen type, serum p‐tau217 has been shown to correlate strongly with plasma p‐tau217 and to follow comparable patterns across the AD continuum. ³³ Because absolute concentrations differ by specimen type, appropriately validated matrix‐specific cutoffs are required, and serum and plasma results should not be used interchangeably.

This study confirms the consistent clinical performance and analytical robustness over time of the Fujirebio Lumipulse G p‐tau217 immunoassay in routine clinical practice. The stability of results, even with an assay version change and across multiple reagent lots, supports the continued use of established cut‐points for amyloid pathology assessment and reinforces the continued need for close QC monitoring and robust validation after assay changes. These findings provide valuable evidence for the reliability and utility of plasma p‐tau217 in a routine clinical setting as a critical biomarker in the diagnosis and management of patients with symptomatic AD.

CONFLICT OF INTEREST STATEMENT

D.J. Figdore has received speaker honoraria from Beckman Coulter. J.A. Bornhorst has received honorarium from Roche Diagnostics and travel funds from Siemens Healthineers. P.M. Vanderboom has no disclosures. His research is supported by NIH (R01AI196126‐1), DOD (HT9425‐25‐1‐0155), and the Minnesota Office of Higher Education (238183). G.S. Day reports no competing interests directly relevant to this work. His research is supported by NIH (R01AG089380, U01AG057195, U01NS120901, U19AG032438, P30AG062677). He serves as a Topic Editor (Dementia) for DynaMed (EBSCO). He is a co‐Project PI for a clinical trial in anti‐NMDAR encephalitis, which receives support from NIH/NINDS (U01NS120901) and Amgen Pharmaceuticals. G.S. Day's institution has received in‐kind contributions for radiotracer precursors for tau PET neuroimaging in studies of memory and aging (via Avid Radiopharmaceuticals, a wholly owned subsidiary of Eli Lilly). He owns stock in ANI Pharmaceuticals. C. Lachner reports no disclosures. His research is supported by NIH (UH3AG083186, P30AG062677). N.R. Graff‐Radford reports no conflicts of interest. His research is supported by NIH. He has participated in multicenter therapy studies sponsored by Biogen, Eisai, and Lilly. C.R. Jack Jr. receives no personal compensation from any commercial entity; he receives research support from NIH and the Alexander Family Alzheimer's Disease Research Professorship of the Mayo Clinic. D. Knopman serves on a Data Safety Monitoring Board for the Dominantly Inherited Alzheimer Network Treatment Unit study sponsored by Washington University St. Louis, the SMART‐HS clinical trial (Univ of Kentucky), the CRANE study (Univ of Michigan), and for Roche TRONTIER phase 3 studies. He receives personal compensation for all. He has served as a consultant for Cognito Therapeutics, AriBio, and Alzeca Biosciences but receives no personal compensation. He receives funding from the NIH. J. Graff‐Radford serves as associate editor of JAMA neurology. He is consultant to open evidence, and have received travel support from the Alzheimer's Association, honoraria from American Academy of Neurology, and IMPACT‐AD. R.C. Petersen serves as a consultant for Roche, Genentech, Eli Lilly, Eisai, Novartis, and Novo Nordisk; receives royalties from Oxford University Press and UpToDate; and receives NIH funding. V. Ramanan has received research funding from the NIH, the Kogod Center for Aging, and the Mangurian Foundation for Lewy Body disease research; has provided CME and other educational content for Medscape, Expert Perspectives in Alzheimer's Disease, Clinical Care Options, PeerView Institute (CME activity supported by an educational grant from Eli Lilly), and the Association of Diagnostic and Laboratory Medicine (webinar supported by an educational grant from Roche); has received speaker and conference session honoraria from the American Academy of Neurology Institute; has served on and chaired a Data Safety Monitoring Board for a clinical trial supported by the Weston Family Foundation; is PI for a clinical trial supported by the Alzheimer's Association; is site Co‐PI for the Alzheimer's Clinical Trials Consortium; and is a site clinician for clinical trials supported by Eisai, Cognition Therapeutics, the Alzheimer's Treatment and Research Institute at USC, and Transposon Therapeutics, Inc. A. Algeciras‐Schimnich has participated in advisory boards for Roche Diagnostics, Fujirebio Diagnostics, Siemens Healthineers, and Beckman Coulter; and has received speaker honoraria from Roche Diagnostics, Beckman Coulter, and Eli Lilly. All other authors report no conflicts of interest. Author disclosures are available in the Supporting Information.

FUNDING INFORMATION

The authors have nothing to report.

CONSENT STATEMENT

This study was approved by the Mayo Clinic Institutional Review Board. All individuals had provided prior research authorization for use of their clinical and laboratory data for research purposes. The study was conducted in accordance with the Declaration of Helsinki and applicable regulatory requirements.

Supporting information

Supporting Information

Supporting Information