Cerebrospinal Fluid Biomarkers of Alzheimer’s Disease in Chinese Patients

Abstract

In view of the paucity of data on cerebrospinal fluid (CSF) biomarkers in Chinese patients, we evaluated the validity of tau, phosphorylated tau 181 (p-tau-181), amyloid β 42 (Aβ42), and Aβ40 proteins in Chinese patients with Alzheimer’s disease (AD). We recruited 24 patients with AD, 12 nondemented controls, and 12 patients with non-AD dementia. We found the CSF levels of Aβ42, tau, p-tau, Aβ42–tau, and Aβ42–p-tau ratios, except the Aβ40 protein level, were significantly different among the 3 groups of patients. Patients with AD had higher levels of CSF tau and p-tau but lower levels of Aβ42 proteins, Aβ42–tau, and Aβ42–p-tau ratios than the nondemented controls. In the diagnosis of AD versus nondementia, the sensitivity and specificity of the ratios of Aβ42–tau and Aβ42–p-tau were 96% and 83% versus 92% and 83%, respectively. Only the Aβ42–p-tau ratio showed satisfactory sensitivity and specificity in the diagnosis of AD versus other dementia.

Introduction

With the aging of the world population, the number of dementia cases will increase, that is, from 35.56 million in 2010 to 65.69 million in 2030. Of them, 60% have Alzheimer’s disease (AD). ¹ The 2 neuropathological hallmarks of AD are the accumulation of extracellular amyloid β (Aβ) plaques and intracellular neurofibrillary tangles. The Aβ is derived from proteolytic cleavage of amyloid precursor protein. The Aβ42 and Aβ40 are common monomeric Aβ isoforms. ² Previous studies have shown that the levels of cerebrospinal fluid (CSF) total tau (tau), phosphorylated tau (p-tau), and Aβ42 proteins can improve the diagnostic sensitivity and specificity of AD in caucasian populations. Typically, patients with AD have low levels of CSF Aβ42 but high levels of tau and p-tau proteins. ^3–6

At present, there is a paucity of studies in Chinese populations regarding the levels of CSF protein in AD. In a previous Chinese study, patients with AD had higher levels of CSF tau but lower levels of CSF Aβ42 proteins than non-AD patients. ⁷ To the best of our knowledge, there is no previous Chinese study on the levels of CSF p-tau proteins in Chinese populations. In addition, recent studies reported different cutoff values of these CSF biomarkers. ⁸ This has led to the recent recommendation from the Alzheimer’s Association Quality Control Program Work Group, which recommended every clinical site to establish the validity of CSF biomarkers in each site.⁸ Therefore, the aim of the present study was to evaluate the diagnostic validity of tau, p-tau-181, and Aβ42 proteins in Chinese patients with AD and to determine the cutoff values of the valid CSF biomarkers for future clinical use in the diagnosis of AD in Chinese patients in our clinical site and laboratory.

Methods

Patients

This was a cross-sectional case–control study. This research protocol was approved by the institutional review board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (Hong Kong). We recruited all patients from the Geriatric Memory Clinic and Medical Geriatric Wards of the Queen Mary Hospital, The University of Hong Kong. We obtained written informed consents from all study patients and their family members. The inclusion criteria of the patients included Chinese ethnicity, age between 56 and 90 years, and availability of a family member. Patients with AD were diagnosed to have dementia by the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) (DSM-IV) criteria ⁹ and AD by the National Institute of Neurological Communicative Disorders and Stroke-Alzheimer Disease and Related Disorders Association (NINCDS-ADRDA) criteria. ¹⁰ Two comparison groups of patients were recruited. The first comparison group consisted of nondemented control patients who had no dementia by the DSM-IV criteria. ⁹ The second comparison group consisted of patients who had other types of dementia (non-AD dementias) or other neurological diseases with cognitive impairment. They met the DSM-IV criteria for dementia ⁹ but did not fulfill the NINCDS-ADRDA criteria for AD. ¹⁰ The exclusion criteria included age less than 56 years or greater than 90 years, active infection, end-stage illness (eg, renal failure and heart failure), cancer within 5 years, deafness, and other communication barriers. We also excluded patients who had mild cognitive impairment, ^11,12 because some of them might have underling preclinical AD. ^11–14 We performed detailed clinical history, neurological examination, computed tomography brain imaging, and cognitive assessment, including the Chinese version of the Mini-Mental State Examination (MMSE), the Chinese version of Alzheimer’s Diseases Assessment Scale Cognitive Subscale, and Delayed 10-Word Recall Test (words recalled after a delay of 30 minutes). ^15,16 Details of these assessments have been published previously. ^17–19 Two clinicians, who were blinded to the CSF biomarkers’ results, rated the clinical diagnoses of all patients in a consensus meeting. This was done prior to assay of these CSF proteins. Apolipoprotein E (ApoE) genotyping was assayed by standard methods as reported in our previous studies. ^17–19 (Refer to the supplementary materials for details of the assays.)

Cerebrospinal Fluid Sampling and Enzyme-Linked Immunosorbent Assay of Biomarkers

Cerebrospinal fluid samples were obtained by lumbar punctures in the morning after an overnight fast. The collected CSF samples were then centrifuged for 10 minutes at 2500 rpm at 4°C. All CSF samples were then stored (in aliquots) at −80°C. The CSF tau, p-tau, and Aβ42 and Aβ40 protein levels were determined by commercial enzyme-linked immunosorbent assay (ELISA) kits (INNOTEST; Innogenetics NV, Belgium). For the CSF Aβ40 protein assay, we used another commercial ELISA kit (BioSource, Int, Camarillo, California). (Refer to the supplementary materials for details of the assays.)

Statistical Analyses

The characteristics of the patients were analyzed by descriptive statistics and compared among the 3 groups. The chi-square (χ ² ) statistics and 1-way analysis of variance (ANOVA) with subgroup comparisons (Bonferonni correction method) were used for categorical (eg, gender) and continuous variables (eg, age), respectively. In bivariate analyses, we used the Kruskal-Wallis test to compare the medians (with interquartile range as a measure of the distribution of the variable) of each CSF biomarker among the 3 groups. The Mann-Whitney U test was used for comparisons of each CSF biomarker between the 2 subgroups. For the adjustment of the potential confounding effect of age, the effect of AD versus nondemented controls was analyzed by the general linear model. We used the receiver–operator characteristic (ROC) curve analyses to determine the area under curve (AUC) and the sensitivity and specificity of each CSF biomarker for the diagnosis of AD versus no dementia and then for the diagnosis of AD versus non-AD dementia. All P values <.05 were taken as statistically significant. We used the SPSS software (version 17.0; SPSS Inc, Chicago, Illinois) for all data analyses, and the MedCal program (version 9.4.2; Schoonjans, Belgium) for the ROC curve analyses.

Results

Characteristic of the Patients

A total of 48 patients, all of Chinese ethnicity, were recruited from January 2012 to May 2013. Of the patients, 37.5% were females. The mean (standard deviation [SD]) age and education level were 75.8 (10.8) and 6.1 (5.4) years, respectively. In all, 24 of them had AD, 12 had no dementia (nondemented controls), and 12 had non-AD dementia or cognitive impairment. The latter group included 1 patient with vascular dementia, 2 patients with Parkinson’s disease dementia, 2 patients with dementia due to normal pressure hydrocephalus, 1 alcoholic dementia patient, 2 patients with cognitive impairment related to end-stage renal failure, 2 patients with frontotemporal lobe dementia, 1 patient with Creutzfeldt-Jakob disease (CJD), and 1 patient with cryptoccoal meningitis.

There were no significant differences in gender and education among the 3 groups (ie, patients with AD, nondemented controls, and non-AD dementia patients). Age and the MMSE score were significantly different among the 3 groups (P = .006 and <.001, respectively; 1-way ANOVA). Subgroup comparisons showed that nondemented patients were younger than patients of the other 2 groups. Also, patients with AD had the lowest MMSE score, while patients with non-AD dementia and nondemented controls had the intermediate and highest MMSE scores, respectively (Table 1). In the AD group, 33.33% (n = 8) were in severe stage with MMSE scores between 5 and 10, 45.83% (n = 11) were in moderate stage with MMSE scores between 11 and 20, and 20.83% (n = 5) were in mild stage with MMSE scores between 21 and 29.

Table 1.

Characteristics of Patients.

	AD Group (n = 24)	Nondemented Control Group (n = 12)	Non-AD Dementia Group (n = 12)	P a	P b
					a. AD vs No Dementia
					b. AD vs Non-AD Dementia
					c. No Dementia vs  Non-AD Dementia
Gender: female, %	41.7	25	41.7	.57	–
Age, years (mean ± SD)	78.3 ± 11.0	60.5 ± 9.6	79.0 ± 7.1	.006	a. .01
					b. 1.0
					c. .019
Education, years (mean ± SD)	6.0 ± 5.7	6.5 ± 5.2	5.9 ± 5.1	.96	a. 1.0
					b. 1.0
					c. 1.0
ApoE∊4+ve (versus ε4-ve)	45.5% (n = 22)	9.1% (n = 11)	0% (n = 11)	.002
MMSE score (mean ± SD)	13.7 ± 6.0	24.2 ± 4.4	16.1 ± 9.8	<.001	a. <.001
					b. .99
					c. .017
ADAS-cog score (mean ± SD)	29.27 ± 10.57	14.96 ± 6.05	25.63 ± 15.62	.016	a. .014
					b. 1.00
					c. .16
Delayed 10-Word Recall Test (DWRT; no. of words recalled; mean ± SD)	0.25 ± 0.58	2.78 ± 3.15	0.56 ± 1.33	.006	a. .006
					b. 1.00
					c. .035

Abbreviations: AD, Alzheimer’s disease; ANOVA, analysis of variance; ApoE, apolipoprotein E; ADAS-cog, Alzheimer’s Diseases Assessment Scale Cognitive Subscale; MMSE, Mini-Mental State Examination (Cantonese Chinese version); NS, nonsignificant; SD, standard deviation.

^a Chi-square statistics for gender and ApoE genotype, and 1-way ANOVA for other variables.

^b Subgroups’ comparison (Bonferroni correction method); P > .05 regarded as NS.

Cerebrospinal Fluid Biomarkers Results

We found significant differences in the CSF Aβ42, tau, and p-tau levels (P = .035, <.001, and <.001, respectively; Kruskal-Wallis test) and the Aβ42–tau and Aβ42–p-tau ratios (P < .001, for both, Kruskal-Wallis test) among the 3 groups. However, there was no significant difference in the CSF Aβ40 level among the 3 groups (Table 2).

Table 2.

Comparison of CSF Biomarkers Among AD, No Dementia, and Non-AD Dementia Groups.

	AD Group (n = 24)	Nondemented Control Group (n = 12)	Non-AD Dementia Group (n = 12)	P a	P b
					a. AD vs No Dementia
					b. AD vs Non-AD Dementia
					c. No Dementia vs Non-AD  Dementia
CSF tau, median (interquartile range), pg/mL	660.22 (394.65)	224.61 (132.66)	357.44 (609.22)	<.001	a. <.001
					b. .112
					c. .078
CSF p-tau, median (interquartile range), pg/mL	78.13 (44.35)	35.53 (20.53)	45.54 (44.51)	<.001	a. <.001
					b. .022
					c. .128
CSF Aβ42, median (interquartile range), pg/mL	278.11 (181.64)	458.90 (417.55)	426.93 (307.59)	.035	a. .022
					b. .067
					c. .478
CSF Aβ40, median (interquartile range), pg/mL	3033.0 (636.8)	2925.1 (562.1)	2611.2 (1233.3)	.31	a. .379
					b. .146
					c. .755
CSF Aβ42–t-tau ratio, median (interquartile range)	0.442 (0.650)	3.12 (1.96)	1.30 (1.48)	<.001	a. <.001
					b. .018
					c. .010
CSF Aβ42–p-tau ratio, median (interquartile range)	3.69 (3.82)	19.54 (10.71)	11.42 (7.50)	<.001	a. <.001
					b. .003
					c. .045

Abbreviations: Aβ, amyloid β; AD, Alzheimer’s disease; CSF, cerebrospinal fluid; p-tau, phosphorylated tau; t-tau, total tau.

^a Kruskal Wallis test.

^b Mann-Whitney U test.

Subgroup comparisons showed that patients with AD had significantly higher levels of CSF tau and p-tau than nondemented control patients (P < .001 for both). Also, patients with AD had significantly lower CSF Aβ42 levels than nondemented controls (P = .022). The Aβ42–tau and Aβ42–p-tau ratios were both significantly lower in the AD than in nondemented control groups (P < .001, for both, Mann-Whitney U test). Also, all subgroup comparisons showed no significant differences in the CSF Aβ40 level (Table 2). For the adjustment of the potential confounding effect of age, the effect of AD versus nondemented controls was analyzed by the general linear model. Separate analyses were done for each of the 5 CSF biomarkers (ie, tau, p-tau, Aβ42, Aβ42–tau, and Aβ42–p-tau ratios). The dependent variable was each of the 5 CSF biomarkers, and the independent variable (fixed factor) was the diagnosis (ie, AD vs nondemented control). Age was included as a covariate in the model. Age was statistically nonsignificant (P > .05) in all the general linear model analyses on tau (then individually for p-tau, Aβ42, Aβ42–tau, and Aβ42–p-tau ratios); only the diagnosis (AD versus no dementia) was significant in these adjusted models.

Subgroup comparisons between the patients with AD and patients with non-AD dementia demonstrated higher p-tau levels and lower Aβ42–tau and Aβ42–p-tau ratios in patients with AD than patients with non-AD dementia. However, there were no significant differences in the levels of CSF tau and Aβ42 between these 2 groups of patients (Table 2).

The ApoE genotype results were available in 44 (91.7%) patients. The prevalence of ApoE∊4 genotype was 45.5% in AD, 9.1% in nondemented controls, and 0% in other dementias. In the subgroup analyses in patients with AD (and then in nondemented controls), the presence of ApoE∊4 genotype has no significant effect (E4+ve vs E4-ve; all P values >.05, Mann-Whitney U test) on each of the CSF biomarkers (tau, p-tau, Aβ42, Aβ42 Aβ40, Aβ42–tau, and Aβ42–p-tau ratios; Table 1).

Results of the Receiver–Operator Characteristic Curve Analyses

For the diagnosis of AD versus nondemented controls, the AUC in ROC curve analyses for the CSF levels of Aβ42, tau, p-tau, Aβ42–tau, and Aβ42–p-tau ratios ranged from 0.74 to 0.93. Among these CSF biomarkers, the ratios of Aβ42–tau and Aβ42–p-tau had the best combinations of sensitivity and specificity, which were 96% and 83% for the Aβ42–tau ratio and 92% and 83% for the Aβ42–tau ratio, respectively (Table 3).

Table 3.

The ROC Curve Analysis Results of CSF Biomarkers in the Diagnosis of Patients With AD Versus Nondemented Controls.

CSF Biomarkers	Cutoff	AUC (95% Confidence Interval)	Sensitivity, %	Specificity, %
tau	>325.7 pg/mL	0.93 (0.80-0.99)	83	91
p-tau	>44.25 pg/mL	0.89 (0.74-0.97)	79	92
Aβ42	≤357.1 pg/mL	0.74 (0.56-0.87)	75	83
Aβ40	>331.2 pg/mL	0.59 (0.42-0.75)	46	83
Aβ42–t-tau ratio	≤1.54	0.92 (0.78-0.98)	96	83
Aβ42–p-tau ratio	≤9.84	0.89 (0.74-0.97)	92	83

Abbreviations: Aβ, amyloid β; AD, Alzheimer’s disease; ROC, receiver–operator characteristic; AUC, area under curve; CSF, cerebrospinal fluid; p-tau, phosphorylated tau; t-tau, total tau.

For the diagnosis of patients with AD versus patients with other dementia, the AUC in ROC curve analyses for the CSF levels of Aβ42, tau, p-tau, Aβ42–tau, and Aβ42–p-tau ratios ranged from 0.67 to 0.80. Only the ratio of Aβ42–p-tau showed a satisfactory combination of sensitivity and specificity of 79% and 75%, respectively (Table 4).

Table 4.

The ROC Curve Analysis Results of CSF Biomarkers in the Diagnosis of AD Versus Non-AD Dementia.

CSF Biomarkers	Cutoff	AUC (95% Confidence Interval)	Sensitivity, %	Specificity, %
tau	>370.2 pg/mL	0.67 (0.49-0.81)	83	58
p-tau	>31.51 pg/mL	0.74 (0.56-0.87)	100	42
Aβ42	≤301.6 pg/mL	0.69 (0.51-0.83)	63	83
Aβ40	>235.7 pg/mL	0.65 (0.48-0.80)	92	50
Aβ42–t-tau ratio	≤1.54	0.74 (0.57-0.87)	96	50
Aβ42–p-tau ratio	≤6.87	0.80 (0.64-0.92)	79	75

Abbreviations: Aβ, amyloid β; AD, Alzheimer’s disease; AUC, area under curve; CSF, cerebrospinal fluid; p-tau, phosphorylated tau; ROC, receiver–operator characteristic; t-tau, total tau.

Discussion

To the best of our knowledge, our present study was the first Chinese study in patients with AD, which investigated the validity of the level of CSF p-tau, Aβ40, and the ratios of Aβ42–tau and Aβ42–p-tau in the diagnosis of AD. Previously, only the CSF Aβ42 and tau protein levels were reported in a Chinese study in China. ⁷ In the present study, we found Chinese patients with AD had higher levels of CSF tau and p-tau but lower levels of CSF Aβ42 than nondemented persons. The CSF tau and p-tau levels in patients with AD were approximately 3 and 2 times of those found in nondemented controls, while the CSF Aβ42 level in patients with AD was approximately 60% of that detected in nondemented controls. Encouragingly, these findings were in accord with previous major studies in caucasian populations. ^3–6 Regarding the CSF Aβ40 level, this was not a discriminative CSF biomarker for AD, as there were no significant differences among the 3 groups and all subgroup comparisons.

We found high sensitivity and specificity of the Aβ42–tau and Aβ42–p-tau ratios for the diagnosis of AD versus nondemented controls in Chinese, when compared to those of the individual biomarker (ie, Aβ42, tau, and p-tau). The sensitivity and specificity of the Aβ42–tau ratio were 96% and 83%, respectively, while the corresponding values of Aβ42–p-tau ratio were 92% and 83%, respectively. Our findings were very comparable to those reported in previous studies, indicating no major ethnic difference in the diagnostic utility of these biomarkers ^{3–6,20–21} (Table 5). For the first time, our present study provided strong evidences to support the use of a combination of these 3 CSF biomarkers as well as their ratios for the diagnosis of AD in Chinese populations.

Table 5.

Comparison of Cutoff Values of CSF Biomarkers in the Present Study With Previous Studies.^a

Caucasian Study	CSF Biomarker	Cutoff	Sensitivity, %	Specificity, %
Shojib et al 20	Tau	>375 pg/mL	59.1	80.4
Lewczukb et al 21	p-tau	>47.9 pg/mL	77.4	73.3
Lewczukb et al 21	Aβ42	≤197.7 pg/mL	75.5	60
Kapakic et al 6	Aβ42–tau	≤2	90	83
de Souzad et al 5	Aβ42–p-tau	≤4.739	91.7	92.6
Shaw 4 (xMAP Luminex platform)e	tau, pg/mL	>93	70	92
	p-tau, pg/mL	>23	68	73
	Aβ42, pg/mL	<192	96	77
	Aβ42–tau ratiof	2.56	86	85
	Aβ42–p-tau ratiof	10.0	91	71
Present studye	tau	>325.7 pg/mL	83	91
	p-tau	>44.25 pg/mL	79	92
	Aβ42	≤357.1 pg/mL	75	83
	Aβ42–tau ratio	≤1.54	96	83
	Aβ42–p-tau ratio	≤9.84	92	83

Abbreviations: Aβ, amyloid β; AD, Alzheimer’s disease; CSF, cerebrospinal fluid; ELISA, enzyme-linked immunosorbent assay; p-tau, phosphorylated tau.

^a ELISA in all studies except Shaw et al’s study, which used xMAP Luminex platform (Luminex Corp, Austin, TX).

^b AD versus non-AD dementia.

^c AD versus vascular dementia.

^d AD versus frontotemporal dementia.

^e AD versus nondemented controls.

^f Calculated from the reported data on tau/Aβ42 and p-tau/Aβ42.

For the differential diagnosis of AD versus other causes of dementia or cognitive impairment, the CSF p-tau level and Aβ42–tau ratio were also highly sensitive biomarkers, having sensitivity of 100% and 96%, respectively. In terms of specificity, the CSF Aβ42 level and Aβ42–p-tau ratio had good specificity of 83% and 75%, respectively. In a study of using CSF biomarkers in the differential diagnosis of AD versus non-AD dementias, Lewczuk et al found that the sensitivity and specificity of CSF p-tau and Aβ42 were 77% and 73% and 76% and 60%, respectively ²¹ (Table 5). Hence, our present findings were comparable. In addition, we clearly showed that using a single CSF biomarker was inferior to a combination of the 3 biomarkers, and the 2 derived Aβ42–tau and Aβ42–p-tau ratios in differentiating AD from other causes of dementia and cognitive impairment. We would therefore recommend using all 3 CSF biomarkers as well as their ratios to help the clinician to achieve an accurate diagnosis of AD and differential diagnosis of AD from other causes of dementias or cognitive impairment.

Despite comparable values in the sensitivity and specificity of these biomarkers, the cutoff values of the CSF biomarkers were quite different for the diagnosis of AD among different studies ^3–6,20,21 (Table 5). Therefore, we could not simply adopt the cutoff values of these biomarkers reported in previous overseas studies for clinical or research purposes. We must perform local validation studies in our own setting. The differences in the cutoff values of CSF biomarkers therefore strengthen the argument for global standardization studies on CSF biomarkers to establish the validity of international reference cutoff values of CSF biomarkers in the future. ²² Before valid global cutoffs of the latter are available, locally validated results would be very useful for clinicians to improve the accuracy of diagnosis of AD. ²³ In the revised diagnostic criteria of AD, the use of biomarkers, including these CSF biomarkers, are recommended for the diagnosis of AD, particularly in research studies. ¹⁴

We acknowledge several limitations of the present study. First, the sample size of our study was small. There was much difficulty with getting consent for CSF study, because Chinese patients did not like the invasive lumbar puncture procedure. However, our results were clearly significant and very consistent with previous similar studies. Nonetheless, we believe cautions are needed in interpretations of the present study. Further confirmation by a large sample study is needed. Second, all patients were recruited from a single center. It was possible that our patients might not be representative of all Chinese patients with AD. As our present study was the first study that investigated all the 3 CSF biomarkers, we could only compare our 2 findings on CSF tau and Aβ42 levels with a previous Chinese study. ⁷ We found our results on CSF tau and Aβ42 levels were similar to those reported in the latter Chinese study, supporting the argument that our patients of this study was quite similar to other Chinese patients with AD. Hence, our present findings would be applicable to other Chinese patients with AD. Third, we did not have neuropathologically confirmation of the AD clinical diagnoses in our patients.

In conclusion, Chinese patients with AD have low CSF Aβ42 levels and elevated CSF tau and p-tau levels. A combination of these 3 CSF biomarkers is useful in the diagnosis and differential diagnosis of AD in Chinese population. Overall, the CSF Aβ42–tau and Aβ42–p-tau ratios are sensitive and specific biomarkers for AD. They are recommended as diagnostic biomarkers for AD in our Chinese population. We would recommend a large prospective study on these biomarkers among Chinese persons with mild cognitive impairment. Future global standardization studies, involving multiethnic populations, are needed for the establishment of internationally valid cutoff values of CSF biomarkers for AD.