Diagnostic performance of salivary markers of Alzheimer's disease: A systematic review

Abstract

High heterogeneity of diagnostic accuracy have been reported for salivary markers of Alzheimer's disease (AD), but the reasons remain unclear. This systematic review aims to evaluate the potential sources of heterogeneity in the diagnostic performance of salivary biomarkers for the identification of AD. We systematically reviewed four databases for studies from inception to 2025. We evaluated biomarker sensitivity, specificity, and area under the curve (AUC). This study was conducted according to the Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) statement. Of 3118 studies, 18 met the inclusion criteria. AUC data were available only for amyloid beta (Aβ)42 and lactoferrin. Pre‐analytical properties were a major source of heterogeneity, comprehending prior orientation, time of collection, recipient material, and centrifugation methods. The main source of variability likely stems from substantial differences in pre‐analytical procedures across studies. Further studies on salivary biomarkers in AD implementing standardized protocols are warranted.

Highlights

• Previous meta‐analysis showed that Lactoferrin may differentiate individuals with cognitive impairment (CU) from those with Alzheimer's disease (AD)

• Pre‐analytical methods were highly variable among studies

• Salivary markers of AD showed substantial heterogeneity in distinguishing CU from AD

• No study matched guidelines for saliva sampling, processing, and analysis

• Further studies should harmonize procedures according to standardized protocols

1. INTRODUCTION

Alzheimer's disease (AD) is the most common neurodegenerative disease affecting the elderly population. ¹ Early diagnosis of AD may enable timely interventions and management, which can potentially lead to reduced treatment costs and help slow or modify the disease progression. AD core biomarkers have been integrated into clinical practice owing to their high sensitivity and specificity in detecting AD core pathology, such as amyloid beta (Aβ) deposition and tau neurofibrillary tangles, even in the preclinical stages of the disease. The ability to identify these AD‐related pathologies at an early phase is pivotal for guiding the development of disease‐modifying therapies aimed at altering the trajectory of AD.

In clinical settings, positron emission tomography (PET) and cerebrospinal fluid (CSF) analysis are gold standard methods for assessing AD pathology in patients. ² However, a biomarker‐assisted diagnosis in AD based on PET and CSF results faces several barriers to clinical implementation. The challenges are related to the high costs associated with these exams, which are often unaffordable for public health systems. Furthermore, not all patients are eligible for lumbar puncture or imaging procedures, underscoring the importance of more accessible approaches such as blood‐based biomarkers. In fact, very recently, a blood test has been approved by the U.S. Food and Drug Administration (FDA) for AD diagnosis, given its demonstrated high accuracy in identifying amyloid pathology in both research and clinical settings. ³ , ⁴ Resource‐limited countries, especially those in the Global South, frequently lag in the advances of AD diagnosis, highlighting the urgent need for new and more accessible diagnostic methods. In an effort to increase accessibility, blood‐based biomarkers have been largely explored in recent years and have demonstrated high accuracy in identifying amyloid beta (Aβ) positivity in research ⁵ and clinical settings. ⁶ , ⁷ , ⁸

Recent evidence also suggests that saliva may serve as an alternative biological matrix for measuring AD‐related biomarkers. Saliva collection offers several advantages over blood, including its noninvasive nature, stress‐free procedures, and repeatability. ⁹ Saliva is an easily accessible bodily fluid, and its composition may change under various pathophysiological conditions. ¹⁰ Its clinical utility is already established for diagnosing oral cancer ¹¹ and severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) infection and for drug testing. ¹² The potential to detect and quantify biomarkers in saliva samples is highly attractive for research and clinical applications and for noninvasive biomarkers aimed at early diagnosis and continuous monitoring of AD. However, saliva may be subject to substantial variations, such as comorbidities and hyposalivation in the elderly, different from the more stable biological matrices such as blood or CSF. Salivary composition involves a two‐step filtered production: from the brain to the blood and from the blood to the saliva. Indeed, a recent study evaluated salivary, plasma, and CSF levels of core AD biomarkers, showing that salivary markers did not correlate with other biofluids. ¹³ In the context of AD, recent studies using saliva as a biological matrix for biomarkers have yielded conflicting results, raising questions about their clinical utility. Although previous meta‐analyses have also evaluated diagnostic performance, a critical gap remains: their underlying sources of heterogeneity were not fully addressed. Herein this study builds upon prior work by evaluating the potential pre‐analytical and analytical sources of heterogeneity in diagnostic performance of salivary biomarkers in identifying AD. This systematic evaluation serves as a practical roadmap for the standardization of future salivary biomarker protocols.

2. METHODS

2.1. Search strategy and selection criteria

The study protocol was registered in the Open Science Framework (registration number: fve4u). ¹⁴ This systematic review was conducted based on the records published from inception to January 2025, according to the Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) statement guidelines. We searched Embase, PubMed, Scopus, and Web of Science. The search queries included Boolean operators and the following key terms: Alzheimer's disease, saliva*, and similar terms (Table S1).

Records were screened by the title, abstract, and full text by two independent investigators (P.V.B. and L.M.), and by a third reviewer when discrepancies occurred (W.V.B.). Studies included in this review matched all the eligibility criteria (Table S2). Studies selected for review included original, full‐text articles published in any language, investigating biomarkers for AD in saliva. In addition, a manual search was conducted for gray literature.

Articles were excluded if: (1) they did not describe a specific group of cognitively unimpaired (CU) individuals, or mild cognitive impairment (MCI) due to AD, or dementia due to AD according to international guidelines ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ ; (2) cohorts with fewer than 10 individuals; (3) reported concentration of the salivary biomarkers data in a format other than mean (standard deviation [SD]) or median (interquartile range); (4) had used nonquantitative methods to assess biomarker results; (5) had cohorts of individuals younger than 18 years of age; and (6) had studies without appropriately referenced methods. 

FIGURE 1.

(A) Preferred Reporting Items for Systematic reviews and Meta‐Analyses (PRISMA) flowchart of studies included and (B) distribution of studies retrieved worldwide. AD, Alzheimer's disease.

2.2. Data analysis and quality assessment

By screening the titles and the abstracts based on the eligibility criteria listed above, studies were selected for further data extraction by two authors (P.V.B. and W.V.B.) according to a data extraction sheet, including biomarker identification, age, and sex for CU and AD individuals, inclusion and exclusion criteria, diagnostic criteria, analytical method, diagnostic performance, descriptive area under the curve (AUC), and sensitivity and specificity of the biomarkers evaluated in the study. For studies with incomplete or missing data, corresponding authors were contacted via email (two attempts per study) to request additional information. Data were taken from cross‐sectional and cohort studies and baseline measurements in longitudinal studies with clinical follow‐up. All data retrieved were checked for accuracy by two researchers (P.V.B. and W.V.B.). The methodological quality of the included studies was assessed using the Newcastle–Ottawa Scale (NOS; Table S3), with scores ranging from 0 to 9. ¹⁹

2.3. Evaluation of diagnostic performance of salivary biomarkers

We carefully evaluated pre‐analytical methods to identify orientations of pre‐collection of saliva, the time of saliva collection, the method of collection, the recipient material, centrifugation procedures, and storage. Pre‐analytical methods were evaluated qualitatively to identify convergence or divergence of methods used in included studies. Analytical methods were also evaluated according to the platform reported in each study.

3. RESULTS

We identified 3118 articles from the initial database search, and 2505 remained after duplicate removal (Figure 1). Eighteen studies matched the eligibility criteria for this systematic review. The included studies were published between 2010 and 2025. Studies collected data from 12 different countries, and most studies were from the European continent (11/18, 61.1%). The included studies comprised 726 CU individuals, 361 MCI, and 668 AD dementia. Studies that included CU individuals had sample sizes ranging from 10 to 160, MCI ranged from 15 to 68, whereas the sample size for AD individuals ranged from 11 to 80. Table 1 summarizes the data collected from the studies included in this systematic review. Overall, most studies had an unclear risk of bias/medium quality, with NOS scores ranging from 5 to 6 (Table S3).

TABLE 1.

Studies included in this systematic review.

Author	Year	Country	Cases (n, F/M, mean age)	CU (n, F/M, mean age)	Main findings regarding salivary markers
Ashton et al.	2018	Sweden, UK	AD: 53, 30/23, 81.4 ± 6.6 MCI: 68, 35/33, 79.8 ± 7.4	160, 94/66, 78.0 ± 6.7	Levels of t‐tau were similar between AD, MCI, and CU
Bermejo‐Pareja et al.	2010	Spain	AD: 70, 49/21, 77.2	56, 39/17, 74.3	Levels of Aβ42 in AD patients were similar to those in PD and CU
Boschi et al.	2022	Italy	AD: 18, 10/8, 72.13 ± 5.5	18, 11/7, 65.7 ± 12	Levels of Aβ42 in AD patients were higher than in CU. There was a negative correlation between CSF and salivary Aβ42
Carro et al.	2017	Spain	AD: 80, 49/31, 76.2 ± 5.3 aMCI: 44, 25/19, 75.2 ± 5.1	91, 59/32, 73.7 ± 6.9	Lactoferrin showed exceptional diagnostic performance in distinguishing AD from CU
Cui et al.	2022	China	AD: 30, 15/15, NA	15, 7/8, 67.5 ± 3.2	Aβ42/Aβ40 and p‐tau/t‐tau combined were accurate in distinguishing AD from CU. All four markers combined showed higher accuracy
François et al.	2024	Australia	AD: 20, 12/8, 78 ± 6.5	40, 21/19, 74.9 ± 7.2	Stratifin was decreased in AD compared to CU. Metabolites distinguished AD from CU
Gleerup et al. (a)	2021	Denmark	AD: 71, 41/30, 72.1 ± 7.3 MCI: 56, 27/29, 70.4±8.2	20, 8/12, 65.7 ± 10.1	Lactoferrin did not differentiate AD from CU
Gleerup et al. (b)	2021	Denmark	AD: 49, 29/20, 72.7 ± 7.5 MCI: 47, 22/25, 71.1 ± 8.2	17, 6/11, 68.4 ± 8.3	No statistically significant differences were found in salivary NfL concentration across the diagnostic groups
González‐Sánchez et al.	2020	Spain	AD: 25, 12/13, 67.2 ± 9.2 MCI: 21, 8/13, 68.8 ± 7.5	48, 33/15, 66.9 ± 5.9	Lactoferrin levels were lower in MCI and AD groups compared to CU
Katsipis et al.	2021	Greece	AD: 20, 9/11, 75 ± 5.5 MCI: 20, 12/8, 75 ± 7.2	20, 11/9, 79 ± 4.7	GFAP levels were lower in AD or MCI than in CU
McNicholas et al.	2022	Australia	AD: 16, 6/10, 79 ± 6 MCI: 15, 8/7, 76 ± 6	29, 14/15, 74 ± 8	Panel of salivary proteins (CST‐C, IL‐1RN, SFN, MMP‐9, and HP) was higher in AD than in CU
Pekeles et al.	2019	Canada	AD: 46, 22/24, Median: 80 (IQR: 9) aMCI: 55, 32/23, Median: 78 (IQR: 14)	47, 32/15, Median: 73 (IQR: 6).	Levels of p‐tau/t‐tau for the S396 phosphorylation site were higher in AD than in CU
Peña‑Bautista et al.	2020	Spain	AD: 31, 18/13, Median: 69 (67, 74)	12, 4/8, Median: 69 (60, 70)	Amino acids and derivatives of neurotransmission impairment were higher in AD than in CU
Ralbovsky et al.	2019	USA	AD: 11, NA, NA MCI: 18, NA, NA	10, NA, NA	Raman hyperspectroscopy showed 99% accuracy for differentiating AD from CU
Ryu et al.	2023	South Korea	AD: 27, 12/15, 72.59 ± 6.90	13, 11/2, 75.5 ± 6.6	The concentration of miRNA‐485‐3p was higher in AD than in CU
Sabaei et al.	2023	Iran	AD: 24, 10/14, 73.5 ± 9.8	22, 13/9, 64.1 ± 9.2	Aβ42 and p‐tau levels were higher in AD than in CU. Salivary α‐syn was lower in AD than in CU
Tvarijonaviciute et al.	2020	Spain	AD: 69, NA, NA	83, NA, NA	Aβ42 and C4 levels were similar between AD and CU
Zalewska et al.	2021	Poland	AD: 25, 15/10, 81.2 ± 6.8	25, 15/10, 82.1 ± 6.7	Salivary oxidative stress index was higher in AD compared to CU. Lactoferrin and Aβ42 were similar between AD and CU

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; aMCI, amnestic mild cognitive impairment; CSF, cerebrospinal fluid; CU, cognitively unimpaired; CST‐C, cystatin‐C; C4, complement C4; GC‐MS, gas chromatography mass spectrometry; GFAP, glial fibrillary acidic protein; Hp, haptoglobin; IL‐1RN, interleukin‐1 receptor antagonist protein; IQR, interquartile range; MCI, mild cognitive impairment; MMP‐9, matrix metalloproteinase 9; miRNA, micro Ribonucleic Acid; n, number of individuals; NA, Not Available; NfL, neurofilament light chain; p‐tau, hyperphosphorylated tau; PD, Parkinson's disease; PET, positron emission tomography; Simoa, single molecule array; SFN, stratifin; t‐tau, total tau; UK, United Kingdom; UPLC‐MS, ultra‐performance liquid chromatography‐mass spectrometry; MS, mass spectrometry. Age is shown with mean and Standard Deviations (SD).

Eighteen studies evaluated the diagnostic performance of salivary markers, comprising 47 different salivary biomarkers (Table 2). Proteomics and metabolomics analyses were conducted in five studies. ²⁰ , ²¹ , ²² , ²³ , ²⁴ Six studies (33.3%) evaluated salivary Aβ42 levels, four studies (22.2%) assessed lactoferrin and phosphorylated tau (p‐tau), three studies (16.6%) evaluated total tau (t‐tau), and two studies (11.1%) examined the p‐tau/t‐tau ratio. In addition, six studies investigated other markers such as adenosine deaminase (ADA), cholinesterase (ChE), ²⁵ Raman spectra, ²⁶ or microRNAs (miRNAs). ²²

TABLE 2.

Diagnostic performance of salivary biomarkers evaluated.

				Diagnostic performance of the most accurate marker
Author	Year	Salivary markers analyzed	Most discriminant markers	AUC	Sensitivity	Specificity
Ashton et al.	2018	t‐tau	None	NS	NS	NS
Bermejo‐Pareja et al.	2010	Aβ42	None	NS	NS	NS
Boschi et al.	2022	Aβ42	Aβ42	0.81	0.84	0.68
Carro et al.	2017	Lactoferrin	Lactoferrin	1	1	1
Cui et al.	2022	Aβ40, Aβ42, t‐tau, p‐tau, p‐tau/t‐tau ratio	Panel: Aβ42, p‐tau, t‐tau, Aβ42/Aβ40 ratio, p‐tau/t‐tau ratio	0.92	NA	NA
François et al.	2024	Untarget proteomics and metabolomics	Stratifin, Mucin‐5B, and beta‐hexosaminidase	0.98	0.95	0.95
Gleerup, et al. (a)	2021	Lactoferrin and total protein	None	NS	NS	NS
Gleerup, et al. (b)	2021	Total protein and NfL	None	NS	NS	NS
González‐Sánchez et al.	2020	Lactoferrin	Lactoferrin	0.93	0.87	0.92
Katsipis et al.	2021	Aβ42, GFAP, p‐tau, IL‐1β, IL‐6, TNF‐α, caspase‐8, and COX‐2	GFAP (Dot Blot)	1	0.85	0.75
McNicholas et al.	2022	CST‐C, IL‐1RN, SFN, MMP‐9, and HP	Panel: CST‐C, IL‐1RN, MMP‐9, total protein	0.97	NA	NA
Pekeles et al.	2019	p‐tau/t‐tau in different phosphorylation sites (T181, S396, S404, S400, T403, T404)	p‐tau/t‐tau (S396 site)	NA	0.73	0.5
Peña‑Bautista et al.	2020	Aspartic acid, glutamic acid, glutamine, GABA, creatine, taurine, N‐acetyl aspartate, myo‐inositol, and acetylcholine	Panel: Myo‐inositol, glutamine, creatine, and acetylcholine	0.81	0.61	0.92
Ralbovsky et al.	2019	Spectra	Raman Spectra	0.99	0.99	0.99
Ryu et al.	2023	miRNA‐485‐3p	miRNA‐485‐3p	0.89	0.74	0.92
Sabaei et al.	2023	Aβ42 and p‐tau	Aβ42	0.81	0.62	0.91
Tvarijonaviciute et al.	2020	Aβ42, Aβ40, t‐tau, p‐tau, FRAP, ADA, ChE, Hp, CRP, PEDF, SAP, MIP‐4, CC4, and α1‐antitrypsin	None	NS	NS	NS
Zalewska et al.	2021	Lactoferrin, oxidative stress index, advanced glycation end products, Aβ cross‐structure, and IL‐1β	Oxidative stress index	0.94	0.90	0.92

Abbreviations: Aβ: amyloid beta; AD: Alzheimer's disease; ADA: adenosine deaminase; Che: cholinesterase; COX‐2: cyclooxygenase 2; CRP: plasma C‐reactive protein; CST‐C: cystatin‐C; C4: complement C4; FRAP: ferric reducing ability of plasma; GFAP: glial fibrillary acidic protein; Hp: haptoglobin; IL‐1RN: interleukin‐1 receptor antagonist protein; IL‐6: interleukin 6; MMP‐9: matrix metalloproteinase 9; MIP‐4: macrophage inflammatory protein‐4; NA: not available; NfL: neurofilament light chain; NS: not significant; PEDF: pigment epithelium derived protein; SAP: salivary amyloid A; SFN: Stratifin; TNF‐α: Tumor necrosis factor‐alpha.

Regarding the diagnostic performance of different salivary biomarkers in studies retrieved, the results varied substantially. The AUC at or above 0.99 was reported in three salivary markers (glial fibrillary acidic protein [GFAP], Raman spectra, and Lactoferrin) in distinguishing CU from AD (Table 2). ²¹ , ²⁶ , ²⁷ The other five studies did not demonstrate a statistically significant difference in differentiated AD by the salivary markers (Table 2). ²⁰ , ²⁵ , ²⁸ , ²⁹ , ³⁰ Divergent findings were reported for Aβ42, Lactoferrin, and p‐tau (Table 2). ³¹ , ³² Meta‐analysis was not possible due to incomplete diagnostic performance data (i.e., lack of AUC, sensitivity, or specificity) or an insufficient number of studies for other markers (Sup. Table 4).

We also evaluated the diagnostic accuracy of salivary markers in distinguishing MCI from other groups (CU and AD), which yielded inconsistent findings. One study reported that salivary lactoferrin levels were higher in CU individuals than in those with MCI, whereas levels were similar between MCI‐PET+ and AD patients. ⁴⁰ Other markers were found to be altered in individual studies—specifically salivary GFAP, ²² stratifin [36], and Raman spectra. ²⁷ Conversely, three studies reported no significant differences between the CU and MCI groups. ⁵ , ²¹ , ³¹

To evaluate the potential sources of variability, we assessed pre‐analytical procedures parameters in each study (Figure 2). Most studies (15, 83.3%) reported providing prior orientation to sample collection, such as refraining from smoking, cleaning teeth, eating or drinking, following a specific diet before sample collection, rinsing mouth with water before collection, or fasting ⁵ , ²⁰ , ²¹ , ²⁵ , ²⁶ , ²⁸ , ³⁰ , ³¹ , ³² , ³³ , ³⁴ (Table 3). Regarding the specification of tubes, five studies (27.7%) used polypropylene tubes without any solution, four (22.2%) used polypropylene with sodium azide solution, one used an RNA Pro‐SAL kit, one study used an absorbent pad, one used a phosphate‐buffered saline (PBS) solution, one used a preweighed sterile plastic container, and five did not specify the recipient material. Regarding the time of collection, 12 studies (66.6%) reported having conducted the collection at a predefined time, comprising 6 studies collecting saliva in the morning, 3 in the afternoon, and three studies collecting saliva in the morning/noon. There were 14 different centrifugation methods reported (Table 3), with only three using the same parameters (600 g, 10 min, at 4°C), and one did not centrifugate. ³⁵ The most frequent method of collection was unstimulated saliva (12, 66.6%), whereas 3 (16.7%) studies used a stimulated method, 1 (5.6%) used both stimulated and unstimulated methods, and 2 studies did not report the method of saliva collection. ²⁶ , ²⁸ Only nine studies (52.9%) reported that samples were placed on ice immediately after sample collection. Sixteen studies (88.9%) reported storing samples at –80°C (16 studies), one study (5.6%) used a storage temperature of –20°C, and one (5.6%) did not report the storage method. Only one study specified the salivary glands collected. ³⁴

FIGURE 2.

(A) Pre‐analytical procedures evaluated in the included studies, including prior orientation, recipient material, time of collection, centrifugation protocols, method of collection, and storage. (B) Analytical methods used to evaluate salivary biomarkers in studies included in this review. Percentages are shown regarding the total number of studies included (n = 18).

TABLE 3.

Preanalytical and analytical procedures of salivary biomarkers included in this review.

Author	Year	Prior orientation	Time of collection	Method of collection	Recipient materials	Centrifugation	Method of analysis	Best AUC reported
Ashton et al.	2018	Overnight fasting, mouth rinse	NA	Unstimulated	Preweighed sterile plastic 30 mL containers	10 min (500 g at 4°C)	Simoa	NS
Bermejo‐Pareja et al.	2010	4 h fasting	Afternoon	NA	2% sodium azide solution in polypropylene	5 min (1500 rpm)	ELISA	NS
Boschi et al.	2022	8 h fasting, mouth rinse	NA	Unstimulated	10 µL of thioflavin S and 2 µL of sodium azide in polypropylene	10 min (600 g at 4°C)	ELISA	0.81
Carro et al.	2017	4 h fasting	Afternoon	Unstimulated	2% sodium azide solution in polypropylene	10 min (600 g at 4°C)	ELISA	1
Cui et al.	2022	30 min fasting (smoking included), mouth rinse	Morning	Unstimulated and stimulated (Citric acid)	Polypropylene without solution	15 min (1000 g at 4°C)	ELISA	0.92
François et al.	2024	None (2 h fasting reported, not oriented)	Noon and Morning	Unstimulated	RNA Pro‐SAL kit	15 min (14,000 g)	GC‐MS and MS/MS	0.98
Gleerup et al. (a)	2021	30 min fasting, mouth rinse	Noon and Morning	Unstimulated	15 mL polypropylene without solution	10 min (2000 rpm at 4°C)	ELISA	NS
Gleerup et al. (b)	2021	Mouth rinse	Noon and Morning	Unstimulated	15 mL polypropylene without solution	10 min (2000 g at 4°C)	Simoa	NS
González‐Sánchez et al.	2020	4 h fasting	Afternoon	Unstimulated	2% sodium azide solution pre‐coated sterile plastic tube in polypropylene	10 min (600 g at 4°C)	ELISA	0.93
Katsipis et al.	2021	Mouth rinse	Morning	Unstimulated	Unspecified	15 min (13,500 rpm)	ELISA, Dot Blot, electrophoresis, and western blot	1
McNicholas et al.	2022	Avoid eating and drinking before saliva (unspecified time)	NA	Unstimulated	Absorbent pad	Not centrifuged	Bradford, BSA, ELISA, and LC‐MS using Orbitrap Fusion MS	0.97
Pekeles et al.	2019	NA	Morning	Unstimulated	50 mL polypropylene without solution	10 min (5000 rpm or 10,000 rpm at 4°C)	ELISA	NA
Peña‑Bautista et al.	2020	Mouth rinse, 30‐min after breakfast	Morning	Unstimulated	Unspecified	5 min (1200 g at 4°C)	UPLC‐MS	0.81
Ralbovsky et al.	2019	1 h fasting (12h for alcohol)	NA	NA	Unspecified	10 min (10,000 rpm)	Raman hyperspectroscopy	0.99
Ryu et al.	2023	NA	NA	Stimulated (Oral swabs)	1 mL of phosphate‐buffered saline in polypropylene	10 min (12,000 g at 4°C)	Real‐time PCR analysis of miRNA‐485‐3p	0.89
Sabaei et al.	2023	5‐day low‐protein diet, 4 h fasting, oral hygiene	NA	Stimulated (Moist rolls)	Unspecified	5 min (1500 rpm)	ELISA	0.81
Tvarijonaviciute et al.	2020	2 h fasting	Morning	Unstimulated	Polypropylene without solution	10 min (3000 g at 4°C)	MILLIPLEX MAP Human Amyloid Beta and Tau Magnetic Bead Panel	NS
Zalewska et al.	2021	Unspecified fasting and oral hygiene	Morning	Stimulated (Citric acid)	Unspecified	20 min (5000 g at 4°C)	ELISA, Colorimetric, spectrofluorimetric, spectrophotometric methods, thioflavin T fluorescence	0.94

Abbreviations: ELISA, enzyme‐linked immunosorbent assay; NA, not available; NS, not significant.

Among analytical procedures, enzyme‐linked immunosorbent assay (ELISA) prevailed in 11 studies (64.7%), with 8 using only ELISA and 3 using ELISA associated with other methods, such as blotting ²¹ and proteomic platforms. ³⁵ , ³⁶ Two studies (11.7%) used single molecule array (Simoa), three (16.6%) used proteomic platforms with mass spectrometry associated with other methods, and two studies (11.7%) used other methods, including Raman spectrometry ²⁶ and real‐time polymerase chain reaction (PCR). ²²

4. DISCUSSION

This systematic review evaluated the diagnostic performance of salivary biomarkers of AD and investigated potential sources of heterogeneity in pre‐analytical and analytical properties. Salivary Lactoferrin and Aβ42 reportedly distinguished CU from AD groups. Previous meta‐analyses analyzing that salivary Lactoferrin and Aβ42 reported sufficient diagnostic performance, though not evaluating pre‐analytical properties. ³⁷ , ³⁸

Pre‐analytical properties varied widely across studies. More specifically, we identified a considerable variability in sample collection and processing: (1) fasting before the saliva sampling seems to increase diagnostic accuracy ²³ , ²⁴ , ²⁶ , ²⁷ , ³⁵ ; (2) unstimulated saliva was also associated with high diagnostic properties ²¹ , ²⁴ , ²⁷ ; and (3) storage at −80°C was reported by most of the studies presenting with high diagnostic performance. ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ , ³⁵ , ³⁶ , ³⁹ By contrast, the time of collection, recipient material, and centrifugation procedures seem to not be associated with the diagnostic performance. It is important to mention that low‐binding tubes were not reported or not used by the studies included in this review, although they reportedly impact amyloid levels. ⁴⁰ The five studies that failed to distinguish AD from CU presented pre‐analytical discrepancies: more specifically, one study with Lactoferrin (stimulated collection), one study with Aβ42 (non‐fasting), one study with t‐tau (500 g at 4°C centrifugation), and one study with neurofilament light (NfL; non‐fasting). ²⁰ , ²⁵ , ²⁸ , ²⁹ , ³⁰ Overall, our findings suggest that the diagnostic performance of salivary biomarkers is highly influenced by pre‐analytical procedures, descriptively prior orientation, time of collection, recipient material, and centrifugation methods. Indeed, variability in pre‐analytical procedures has been cited consistently as a major source of bias in AD biomarker research. ¹³ , ⁴¹ , ⁴² Especially analytes in extremely low concentrations are vulnerable to pre‐analytical variations, as is the case for some salivary biomarkers, such as GFAP and p‐tau species. ⁴² Without rigorous standardization, variability in results will likely continue to hinder research progress, limiting the development of reliable salivary biomarkers for AD and other neurological conditions.

Regarding analytical methods conducted to identify the salivary levels of biomarkers of AD, a wide range of techniques were employed. More specifically, good accuracy was reported in different analytical methods such as enzyme‐linked immunosorbent assay (ELISA), ²¹ , ²⁷ , ³⁴ , ³⁶ , ³⁹ mass spectrometry, ²³ , ²⁴ and hyperspectroscopy. ²⁶ Of interest, studies using an ultrasensitive platform to identify salivary levels of t‐tau, ²⁹ p‐tau, and NfL ²⁰ did not demonstrate statistically significant results in distinguishing CU from AD groups. It is important to note that salivary markers had inconsistent diagnostic properties in distinguishing CU from MCI. One can suggest that t‐tau is not a good marker and that results with p‐tau or NfL have been influenced by lack of test validation in saliva. Another potential explanation is the reduced test accuracy in methods previously described. ⁴³ , ⁴⁴ Because we are dealing with low‐level analytes, future studies should focus on the diagnostic properties of salivary biomarkers with ultrasensitive technologies. Confounding factors should also be acknowledged when interpreting salivary biomarkers, such as alcohol use and the presence of Down syndrome in salivary levels of Lactoferrin. ⁴⁵ , ⁴⁶

Finally, a guideline for standardization of pre‐analytical variables for salivary biomarkers of AD was published by a Salivary Biomarkers for Dementia Research Working Group and is an important first step toward pre‐analytical and analytical harmonization. ⁴⁷ The implementation of the protocol of salivary collection for evaluating biomarkers of AD is essential to advance the investigation of its diagnostic performance but difficult to optimize without a high‐performing biomarker to use as a reference. For example, self‐guided sampling introduces uncontrolled pre‐analytical variability, and differences between stimulated and unstimulated saliva further complicate standardization. Optimal collection of saliva was proposed for sampling, storage, and analysis. It is intriguingly that none of the included studies would fit the full proposed protocol, which recommends (1) 2500 x g at 4°C for 15 min centrifugation; (2) fasting and mouth rinse prior to collection orientation; and (3) to not use cotton material for saliva sampling. These findings highlight that variability in diagnostic performance may possibly be due to a number of different pre‐analytical procedures reported.

Despite our efforts to review and integrate the data from various studies, significant limitations were identified. The methodological heterogeneity among studies challenged the study, and we may have selected only studies that reported positive results with AUC values. Although diagnostic criteria were an inclusion criterion, the diagnosis of AD has changed substantially over time, and diagnostic heterogeneity is also a possibility. Biomarkers of AD tested were used largely for blood or CSF, but validation for saliva was not reported. Sample size should also be considered when evaluating salivary biomarkers, although we tried to mitigate this issue by including only studies with ten or more individuals per group. Furthermore, the underrepresentation of low‐ and middle‐income countries in salivary biomarker research limits the generalization of these findings, particularly with respect to their global applicability.

In conclusion, pre‐analytical and analytical heterogeneity of studies on AD salivary biomarkers presents a major challenge to drawing definitive conclusions about the clinical utility of salivary markers. The variability in sample collection and processing methods, and the diverse analytical techniques employed, entangles biomarker comparisons and underscores the need for implementation of standardized operating procedures. Future research should focus on developing and adopting these standardized methods to provide consistent and robust evidence for the potential of salivary biomarkers in AD diagnosis.

CONFLICT OF INTEREST STATEMENT

W.V.B. has served on the scientific advisory board of Masima, and he is also a co‐founder and minority shareholder. E.R.Z. has served on the scientific advisory boards of Nintx, Novo Nordisk, and Masima. He is also a co‐founder and a minority shareholder at Masima. P.V.B., L.M., G.C.‐C., P.C.L.F., A.M.D.L.L., B.B., T.A.P., A.B., and N.J.A. do not have anything to disclose. Any author disclosures are available in the supporting information.

Supporting information

Supporting Information

Supporting Information