Abstract
Objective: To investigate if patterns of CSF biomarkers (T-tau, P-tau, and Aβ42) can predict cognitive progression, outcome of cholinesterase inhibitor (ChEI) treatment, and mortality in Alzheimer disease (AD).
Methods: We included outpatients with AD (n = 151) from a prospective treatment study with ChEI. At baseline, patients underwent cognitive assessments and lumbar puncture. The patients were assessed longitudinally. The 5-year survival rate was evaluated. CSF-Aβ42, T-tau, and P-tau were analyzed at baseline. K-means cluster analysis including the 3 CSF biomarkers was carried out.
Results: Cluster 1 contained 87 patients with low levels of Aβ42 and relatively low levels of T-tau and P-tau. Cluster 2 contained 52 patients with low levels of Aβ42 and intermediate levels of T-tau and P-tau. Cluster 3 contained 12 patients with low levels of Aβ42 and very high levels of CSF T-tau and P-tau. There were no differences between the clusters regarding age, gender, years of education, baseline instrumental activities of daily living, or APOE genotype. Even though there was no difference between cluster 3 and the other clusters in disease duration or global rating, the patients in cluster 3 performed worse on cognitive tests already at baseline. Patients in cluster 3 exhibited a very poor outcome of ChEI treatment. Finally, cognition deteriorated faster over time and the mortality rate was substantially increased in cluster 3.
Conclusion: A subgroup of patients with Alzheimer disease with extreme levels of CSF biomarkers exhibits worse clinical outcomes over time, including faster progression of cognitive deficits, no response to ChEI treatment, and a higher mortality.
Glossary
- Aβ42=
- β-amyloid 1-42;
- AD=
- Alzheimer disease;
- ADAS-cog=
- Alzheimer's Disease Assessment Scale–Cognitive Subscale;
- ChEI=
- cholinesterase inhibitor;
- CI=
- confidence interval;
- DSM-IV=
- Diagnostic and Statistical Manual of Mental Disorders, 4th edition;
- IADL=
- Instrumental Activities Of Daily Living scale;
- MMSE=
- Mini-Mental State Examination;
- P-tau=
- phosphorylated tau;
- SATS=
- Swedish Alzheimer Treatment Study;
- T-tau=
- total tau.
The diagnosis of Alzheimer disease (AD) is currently based on the identification of dementia1 and specific clinical symptoms suggesting AD.2 To enhance the diagnostic accuracy of AD, and to enable an earlier diagnosis, measurements of CSF biomarkers have been suggested to complement the diagnostic criteria.3 In AD, the level of CSF β-amyloid 1-42 (Aβ42) is lower, while the levels of total tau (T-tau) and phosphorylated tau (P-tau) are higher than in healthy controls.4 However, whether various CSF biomarker patterns in patients with AD are associated with different clinical presentations is unclear. Recently, a cross-sectional study showed that a subgroup of patients with extreme levels of CSF T-tau and P-tau exhibited more severely impaired memory, mental speed, and executive functions, which could not be explained by disease severity.5 However, it would be interesting to study whether CSF biomarkers can identify subgroups of patients with AD with a more malignant disease course over time.
We investigated 151 patients with AD who participated in a prospective and longitudinal 3-year treatment study with cholinesterase inhibitors. Patients underwent lumbar puncture at baseline. The baseline CSF levels of Aβ42, T-tau, and P-tau were analyzed and K-means cluster analysis, including the 3 CSF biomarkers, was carried out and 3 clusters were acquired. The patients were followed longitudinally with clinical assessments and 5-year survival. We found that a certain pattern of the CSF biomarkers is associated with poor treatment outcome, rapid cognitive deterioration over time, and high mortality rate.
METHODS
Participants and clinical assessments.
A total of 151 patients from the Clinical Memory Research Unit in Malmo, Sweden, were included in this prospective and longitudinal study (Swedish Alzheimer Treatment Study [SATS]). Investigation at baseline included medical history taking, physical and neurologic examination, cognitive testing, laboratory tests, and CT or MRI of the brain. Patients fulfilled the clinical criteria of dementia as defined by the DSM-IV1 and probable or possible AD according to the criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association.2 Patients were investigated with clinical and cognitive assessments as well as CSF taps before cholinesterase inhibitor treatment (donepezil, rivastigmine, or galantamine) was initiated. The treatment and follow-up protocol of the SATS study have previously been described in detail.6 Briefly, SATS is an open, prospective, longitudinal study assessing patients from a routine clinical setting for a total of 3 years. The inclusion criteria are patients with AD older than 40 years, living at home at the time of diagnosis, having a caregiver, assessable with Mini-Mental State Examination (MMSE) at baseline, and giving their informed consent to participate. Cognitive tests used were MMSE (0–30 points, lower score indicates greater impairment)7 and Alzheimer's Disease Assessment Scale–Cognitive Subscale (ADAS-cog) (0–70, higher score indicates greater impairment).8 Functional status was measured by the Instrumental Activities of Daily living (IADL) scale,9 ranging from 8 to 31 points with the higher value indicating more impaired function. The assessment of disease level at baseline was made with a 7-point scale that varies from 1 (normal) to 7 (very severe).10,11 Information on the patients' MMSE score before treatment (>2 months before) was obtained and expressed as a linear coefficient (pretreatment progression rate in MMSE/time).
Patients were followed longitudinally with MMSE 2 months after treatment start and every 6 months from baseline for a total of 3 years. Dropout was monitored during the study. At the 1-year follow-up, 83% remained for cognitive assessment; at 2 years, 62%; and after 3 years, the completion rate was 42%. The most common cause of dropout was addition of memantine treatment or change of ChEI inhibitor. The individuals' annual cognitive deterioration rates were calculated using the last MMSE value observed before dropout divided by the time from baseline to the time when this observation was carried out (median follow-up time of cognitive testing was 2.5 years). All patients, except 3 who were lost to follow-up, were followed for a minimum of 5 years to monitor the exact time of the event of death. In the 5-year time span, 45 patients out of the 151 died (30%).
Standard protocol approvals, registrations, and patient consents.
The study was approved by the Ethics Committee of the University of Lund. All patients and their closest relative/caregiver gave written informed consent to participate in the study, which was carried out in accordance with the Helsinki Declaration.
CSF analysis and APOE genotype.
CSF taps were obtained from the patients and taken at the lumbar L3/L4 or L4/L5 interspace. The CSF samples were frozen at −80°C and stored in polypropylene tubes. Total-tau (T-tau), tau phosphorylated at threonine 181 (P-tau181), and Aβ42 levels were determined by the xMAP Luminex technology using the INNO-BIA AlzBio3 kit (Innogenetics, Ghent, Belgium), as previously described in detail.12
APOE (gene map locus 19q13.2) genotyping was performed by minisequencing as described previously in detail.13
Statistics.
SPSS program version 17.0 was used. The level of significance was defined as p < 0.05 if not otherwise specified. To avoid the possibility of skewed distributions, nonparametric methods were used. The Spearman correlation coefficient was used to investigate possible bivariate linear associations. Kruskal-Wallis analysis of variance was performed for analyzing the difference among 3 or more independent groups, followed by Mann-Whitney U test with Bonferroni correction when significant. χ2 Test was used for categorical variables. Pairwise comparisons of 2 related samples were made by the Wilcoxon matched-pairs signed ranks test.
A hierarchical cluster analysis (Ward method) was applied to assess the different solutions based on minimum variance within clusters and relatively equal cluster sizes. The sequence of mergers in the dendrogram suggested a 3-cluster solution. A K-mean cluster analysis defining 3 clusters based on the CSF biomarkers Aβ42, T-tau, and P-tau was performed. The K-mean cluster analysis forms groups to obtain maximal differences in cluster variables between clusters and minimal differences within groups and is preferred when non-normally distributed data are analyzed. The baseline status of the 3 clusters was investigated as well as potential differences between the clusters.
The cognitive outcome at 2 months and the annual change in MMSE were analyzed with general linear models between the CSF clusters, adjusting for the baseline MMSE level, age, gender, and presence of APOE ε4 allele.
Kaplan-Meier survival analysis was used to examine the distribution of time from start of treatment (baseline) until a defined endpoint (death) in different groups. Log rank and Breslow tests were performed to analyze the equality of the survival distributions for the different groups (clusters). Multivariate Cox proportional hazards regression models were used to estimate the effects of the 3 different clusters on the relative hazard of death within 5 years. The analyses were carried out with adjustment for potential confounding of the baseline demographic variables, i.e., age, gender, baseline MMSE, and APOE ε4 carrier status.
RESULTS
All patients.
The 151 patients with AD were characterized at baseline. Age was 75.8 ± 6.3 years (mean ± SD) and 104 patients were female (69%). Median level of education was 9.2 ± 3.2 years (range 7–16). The duration of disease was 2.8 ± 2.0 years (mean ± SD). At baseline, the mean score of MMSE was 21.7 ± 4.5 (range 6–28) and of ADAS-cog 20.6 ± 9.9 points (range 1–70) (mean ± SD). The mean baseline score of IADL was 16.0 ± 5.8 and the global level of severity was 3.7 ± 0.8 (mean ± SD). The levels of the CSF biomarkers were Aβ42 398 ± 97 (mean ± SD, ng/L), T-tau 620 ± 349 (mean ± SD, ng/L), and P-tau 77 ± 33 (mean ± SD, ng/L). The longitudinal cognitive outcome in MMSE (expressed as a mean annual change) was −2.0 ± 4.1 points/year (mean ± SD).
Spearman correlation analyses were performed on data obtained at baseline. CSF T-tau exhibited a slight correlation with baseline ADAS-cog (r = 0.287, p < 0.0001) and global severity (r = 0.222, p < 0.01) and a correlation with baseline MMSE (r = −0.306, p < 0.0001), indicating that higher CSF T-tau levels are associated with more pronounced cognitive deficits at baseline. Higher CSF P-tau levels were associated with worse performance in baseline MMSE (r = −0.213, p < 0.01). CSF Aβ42 was not associated with any of the tests.
Cluster analysis.
Three clusters were identified using K-means cluster analysis. Clinical characteristics of CSF clusters are displayed in detail in the table. The 3 clusters' distribution of CSF Aβ42, T-tau, and P-tau are displayed in figure 1. Cluster 1 consisted of 87 patients with low levels of Aβ42 (402 ± 109, mean ± SD, ng/L) and relatively low levels of T-tau (397 ± 113, mean ± SD, ng/L) and P-tau (60 ± 19, mean ± SD, ng/L). Cluster 2 consisted of 52 patients with intermediate level CSF biomarkers. Cluster 3 consisted of patients with low levels of Aβ42 (362 ± 66, mean ± SD, ng/L) and very high levels of CSF T-tau (1,501 ± 292, mean ± SD, ng/L) and P-tau (139 ± 39, mean ± SD, ng/L) (table, figure 1). The 3 clusters differed in CSF T-tau (p < 0.0001) and CSF P-tau (p < 0.0001), but not in CSF Aβ42 (p = 0.43).
Table Clinical characteristics of CSF clusters (mean ± SD or %)
Figure 1 Distribution of CSF total tau (T-tau), phosphorylated tau (P-tau), and β-amyloid 1-42 (Aβ42) levels in ng/L in the 3 clusters
The 3 clusters' distribution of CSF T-tau (y-axis), P-tau (x-axis), and Aβ42 (z-axis) in ng/L are displayed. Cluster 1 contains 87 patients with low levels of Aβ42 and relatively low levels of T-tau and P-tau. Cluster 2 contains 52 patients with low levels of Aβ42 and intermediate levels of T-tau and P-tau. Cluster 3 contains 12 patients with low levels of Aβ42 and very high levels of CSF T-tau and P-tau.
The 3 clusters did not differ in gender, APOE genotype, education, age at baseline, age at debut, disease duration at baseline, or IADL level at baseline (table). The 3 clusters differed in MMSE levels at baseline (p < 0.05), ADAS-cog at baseline (p < 0.05), and global level at baseline (p < 0.05) (table). Patients in the third cluster, with extreme levels of CSF T-tau and P-tau, performed worse on cognitive tests at baseline, but were not worse in the global testing compared to the other clusters (table).
Treatment outcome.
The outcome of treatment with choline esterase inhibitors was compared among the 3 clusters. The change observed in MMSE scores after 2 months of treatment differed between clusters (Kruskal-Wallis, p < 0.05). Patients in cluster 3 did not respond as well to treatment as cluster 1 (p < 0.01) and cluster 2 (p < 0.05) (Mann-Whitney U test) (figure 2).
Figure 2 Mini-Mental State Examination (MMSE) change after treatment in the 3 clusters
The figure depicts the outcome of cholinesterase inhibitor (ChEI) treatment and compares the 2-month outcome in the 3 different clusters. Both the clusters' change in MMSE before treatment (pretreatment progression rate) and MMSE change after treatment are displayed. Patients with Alzheimer disease in cluster 3 performed worse at 2 months than in cluster 1 (p < 0.01) and cluster 2 (p < 0.05) (Mann-Whitney).
Using Wilcoxon signed rank test to investigate the difference between the individuals' total baseline and 2-month MMSE scores, clusters 1 and 2 performed better on MMSE after 2 months of treatment compared to baseline (p < 0.05). In contrast, cluster 3 had a tendency to perform less well after 2 months of treatment compared to baseline. In a general linear model, the MMSE change after 2 months of treatment differed between clusters even after adjustment for age, gender, presence of APOE ε4 allele, and baseline MMSE level (p < 0.0001). After Bonferroni correction of this model, the treatment outcome of cluster 3 differed compared to clusters 1 and 2 (p < 0.001).
Long-term cognitive deterioration.
The mean annual change in MMSE after baseline was worse in cluster 3 patients (p < 0.05), who exhibited a deterioration rate of 4.9 points per year compared to less than 2 points per year in the other 2 clusters (table). Since cluster 3 had a lower MMSE at baseline, this could be a confounding factor. However, there was no correlation between annual change in MMSE and baseline MMSE score. In a general linear model, the progression rate differed between clusters even after adjustment for age, gender, presence of APOE ε4 allele, and baseline MMSE level (p < 0.05). After Bonferroni correction of this model, the cognitive deterioration in cluster 3 differed compared to cluster 1 (p < 0.05) but not to cluster 2 (p = 0.087).
Five-year survival.
The mortality after 5 years was different in the 3 clusters (Pearson χ2 test, p < 0.01). In cluster 3, 67% died within the 5-year span, compared to 30% in cluster 1 and 21% in cluster 2 (table).
Figure 3 depicts the 5-year survival rates of all 3 clusters. In a Kaplan-Meier survival analysis with the time limit of 5 years, the 3 clusters differed in survival (log-rank test, p < 0.01, Breslow test, p < 0.01) (figure 3). Multivariate Cox proportional hazards models revealed that patients in cluster 3 had an increased risk of death within 5 years compared to patients in cluster 1 (hazard ratio 3.4, 95% confidence interval [CI] 1.4–7.6) and cluster 2 (hazard ratio 5.3, 95% CI 2.0–13.9), even after adjustment for age and gender. Similar and significant results were obtained when correcting the Cox analyzes simultaneously for age, gender, baseline MMSE score, and APOE genotype (data not shown).
Figure 3 Five-year survival in the 3 clusters
In a Kaplan-Meier survival analysis, the 3 clusters differed in survival (log-rank test, p < 0.01, Breslow test, p < 0.01). Multivariate Cox regression analysis showed that cluster 3 had an increased risk of death compared to cluster 1 (hazard ratio 3.4, 95% confidence interval 1.4–7.6) and cluster 2 (hazard ratio 5.3, 95% confidence interval 2.0–13.9), even after adjustment for age and gender.
DISCUSSION
In the present longitudinal study, we identified a subgroup of patients with AD characterized by very high levels of CSF T-tau and P-tau and low levels of Aβ42. This subgroup of patients exhibited a more malignant disease course, including faster progression of cognitive deficits, worse outcome of ChEI treatment, and a higher mortality rate.
An advantage of the present study is that the data were obtained from a long-term clinical follow-up study and not only from observations at baseline. Moreover, it could be of importance that the follow- up data were collected prospectively and the evaluation was structured and standardized. The sample size was quite large and the investigations at baseline were thorough. A limitation is that no postmortem data were available to verify the clinical diagnoses. However, long-term clinical follow-up of patients with dementia disorders is known to increase the accuracy of the clinical diagnosis.
Cluster analysis of CSF biomarkers has previously been used to identify subgroups of patients with AD with different clinical characteristics.14 Recently, a cross-sectional study showed that clusters of CSF biomarker levels were related to distinct cognitive profiles in patients with AD.5 The authors of that article described a cluster of patients with AD with very high CSF T-tau and P-tau values that constituted about 10% of the total cohort, and it is interesting to note that the CSF cluster analysis of the present study produced very similar subgroups. In their cross-sectional study they showed that the subgroup with high tau levels exhibited a distinct cognitive profile with a more severe impairment of memory, mental speed, and executive function, which could not be explained by disease severity. These findings were supported by the baseline data of the present study. Patients in cluster 3 (with the highest levels of T-tau and P-tau) had at baseline a similar disease duration as well as a similar disease level, as measured with a global scale and IADL, but performed worse on the cognitive tests. In addition, several studies, including the present one, have reported that CSF tau levels correlate to some degree with worse cognitive performance at baseline.15–17 Together, these results indicate that patients with very high levels of CSF tau have more pronounced deficits of important cognitive domains already during the early phases of the disease.
Symptomatic treatments of AD, such as ChEI, result in an early response to treatment which can be measured with common cognitive tests. A peak in cognitive performance is usually observed 8 to 12 weeks after initiation of treatment, which is not seen in placebo-treated patients with AD.18–20 In the present study, we found that clusters 1 and 2 improved significantly in cognition after 8 weeks of treatment, which was not the case in cluster 3 (figure 2), indicating that patients with extreme levels of CSF biomarkers do not benefit from ChEI treatment. The present treatment study is long-term (more than 6 months) and thus no placebo arm was allowed due to ethical reasons, which could be a limitation when interpreting treatment outcome data. However, most previous AD treatment trials have shown that placebo effects after 8 to 12 weeks are absent when evaluated with MMSE18–20 or absent or minor when evaluated with ADAS-cog.21–23 Therefore, it is unlikely that the clear and positive cognitive effects seen in clusters 1 and 2 after 8 weeks of treatment (figure 2) are due to placebo effects that for some reason are absent in cluster 3. The clinical observation that subgroups of patients with AD do not appear to benefit from ChEI treatment is something many doctors have experienced. However, the fact that certain CSF biomarker patterns in AD might predict outcome of ChEI treatment has to our knowledge not been described before.
We found that patients in cluster 3 also exhibited significantly faster cognitive deterioration over time. This is in accordance with a longitudinal and retrospective study that included 49 patients with mild AD. That study found that patients with higher levels of CSF T-tau and lower levels of CSF Aβ42 at baseline deteriorate faster in cognition over time.24 Similarly, it has recently been shown that patients with AD with CSF T-tau levels above 800 ng/L at baseline exhibit a more rapid progress in MMSE score reduction.25 Interestingly, neuropathologic studies suggest that the abundance of tau-containing neurofibrillary tangles correlates with the duration and severity of dementia26,27 as well as cognitive symptoms.28 Moreover, widening ventricles over time in patients with AD are associated with higher levels of CSF T-tau and P-tau,29 supporting the idea that the intensity of the disease is associated with tau pathology.
AD is associated with increased risk of death compared to age-matched individuals without dementia.30 Factors outlined as increasing the risk of death in AD include high age and male gender.31,32 In the present study, patients with the extreme values of CSF T-tau and P-tau levels in cluster 3 showed higher mortality rates than the other 2 clusters even after correcting for age, gender, baseline MMSE, and APOE ε4 carrier status. Very little has previously been reported concerning AD mortality and CSF biomarkers. In one study, low prostaglandin E2 levels in CSF increased the risk of death but low levels were also associated with more advanced AD which in itself is a risk factor for death.16,33 In accordance with the present study, a pilot study from our group including 21 patients with AD has previously shown that high levels of CSF T-tau is associated with increased mortality.16
It is likely that AD is a heterogeneous condition. The findings of the present study may have significant clinical implications, as it is likely that the efficacy of different therapies may vary between subgroups of patients with AD, as suggested by the present study. Treatments for AD will probably have to be individualized in the future and results from the present study and others5,14 suggest that analysis of CSF biomarkers might be a meaningful way to stratify patients with AD for different types of therapies.
AUTHOR CONTRIBUTIONS
Statistical analysis was conducted by Dr. Åsa K. Wallin and Dr. Oskar Hansson.
ACKNOWLEDGMENT
The authors thank the patients and relatives for their cooperation in this study; the nursing staff Cecilia Dahl, Annacarin Björkman, Marie Person, Tarja Tikkanen, Eva Falk-Langebro, and Yvonne Sjöberg at the Memory Clinic in Malmö involved in the care of the patients in this study; and Roberta Boson for linguistic support and Carina Wattmo for statistical advice, both at the Memory Clinic in Malmö.
DISCLOSURE
Dr. Wallin has received funding for travel from Janssen-Cilag. Dr. Blennow has served on scientific advisory boards for Adlyfe Inc., Bayer Schering Pharma, Bristol-Myers Squibb, and Merz Pharmaceuticals GmbH; received a speaker honorarium from Pfizer Inc.; serves as a consultant for Wyeth, AstraZeneca, Bristol-Myers Squibb, and Eli Lilly and Company; and has received research support from Pfizer Inc., Innogenetics, the Swedish Research Council, Västra Götalandsregionen, Sweden, the Swedish Brain Power Project, the Swedish Council for Working Life and Social Research, the Swedish Alzheimer Foundation, Stiftelsen för Gamla Tjänarinnor, and the King Gustaf V and Queen Victoria Foundation. Dr. Zetterberg has served on a scientific advisory board for GlaxoSmithKline; serves as an Associate Editor for the Journal of Alzheimer's Disease; and receives research support from the Swedish Research Council, the Alzheimer's Association, and the Royal Swedish Academy of Sciences. Dr. Londos has received speaker honoraria from Lundbeck Inc. and Novartis and serves as a member of the Swedish Parkinson Academy. Dr. Minthon and Dr. Hansson report no disclosures.
Footnotes
-
Study funding: Supported by the Swedish Research Council (14002 and 2006-6227), the Alzheimer's Association (NIRG-08-90356), the Royal Swedish Academy of Sciences, Stiftelsen för Gamla Tjänarinnor, the Swedish Foundations of the National Board of Health and Welfare, the Swedish Alzheimer Foundation, the Swedish Society of Medicine, the Swedish Brain Power, the Trolle-Wachtmeister Foundation, and the regional agreement on medical training and clinical research (ALF) between Skåne County Council and Lund University.
Disclosure: Author disclosures are provided at the end of the article.
Received November 9, 2009. Accepted in final form February 11, 2010.
REFERENCES
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.