Background:

Various studies show an inverse relation between Alzheimer disease and cancer, but findings are likely to be biased by surveillance and survival bias. Plasma amyloid-β (Aβ) is defined as a preclinical feature of Alzheimer disease, with lower levels of Aβ42 being associated with a higher risk of Alzheimer disease. To get more insight into the biological link between Alzheimer disease and cancer, we investigated plasma Aβ levels in relation to the risk of cancer.

Methods:

Between 2002 and 2005, we measured plasma Aβ40 and Aβ42 levels in 3,949 participants from the population-based Rotterdam Study. These participants were followed until the onset of cancer, all-cause dementia, death, loss to follow-up, or January 1, 2014, whichever came first. We used Cox proportional hazards models to investigate the association between plasma Aβ40 and Aβ42 levels, and the risk of cancer. Analyses were stratified by cancer site.

Results:

During a median (interquartile range) follow-up of 9.0 years (6.9–10.1), 560 participants were diagnosed with cancer. Higher levels of log2 plasma Aβ40 and Aβ42 were associated with a higher risk of cancer [hazard ratio per standard deviation increase for Aβ40 = 1.12 (95% confidence interval, CI = 1.02–1.23) and Aβ42 = 1.12 (95% CI = 1.03–1.23)]. These effect estimates were most pronounced for hematologic cancers, urinary tract cancers, and cancers of unknown primary origin.

Conclusions:

We found that higher levels of both plasma Aβ40 and Aβ42 were associated with a higher risk of cancer.

Impact:

Our study suggests a potential biological link between Alzheimer disease and cancer. The pathophysiologic role of Aβ in cancer and its causality warrant further investigation.

This article is featured in Highlights of This Issue, p. 1841

Alzheimer disease and cancer are common diseases in the elderly population that pose a high burden of morbidity and mortality on societies (1, 2). Various observational studies have suggested that patients with Alzheimer disease have a lower risk of non–central nervous system (CNS) cancer and vice versa (3–7), which was not driven by a specific cancer type. However, methodologic issues including surveillance and survival bias may drive the association toward an inverse direction (8). In fact, recent evidence points toward the possibility of a positive link (9, 10), which is supported by overlapping risk factors for Alzheimer disease and cancer, such as age and smoking, and by overlapping pathways, including genome instability and inflammation (11). Against this background, we recently showed that persons with mild cognitive impairment (MCI), a preclinical stage of Alzheimer disease, tended to have a higher risk of cancer [hazard ratio (HR) = 1.25, 95% confidence interval (CI) = 0.99–1.58]. This risk was statistically significantly higher than the decreased risk of cancer of patients with Alzheimer disease (HR = 0.52, 95%CI = 0.39–0.69; ref. 10).

Investigating the preclinical stage of Alzheimer disease and linking it to cancer could further unravel the association between these diseases. Accumulation of plaques containing amyloid-β (Aβ) in the brain is a defining feature of Alzheimer disease pathology (12). Aβ is currently the earliest detectable pathologic change in the preclinical stage of Alzheimer disease (13) and can be measured noninvasively in plasma. Although endothelial cells of blood vessels and platelets are the main source of circulating Aβ (14), Aβ is also produced by neuronal cells and is subsequently transported across the blood–brain barrier (15). During the earliest stages of Alzheimer disease, neuronal Aβ production is first increased, resulting in higher plasma Aβ levels. Plasma Aβ40 and Aβ42 levels decrease during the incipient clinical phase of the disease as a result of Aβ deposition in the brain (16, 17). A substantial amount of Aβ accumulates before the manifestation of clinical symptoms—thus also before MCI—but the exact onset of deposition is as yet unknown (13). In turn, lower plasma Aβ42 levels are associated with a higher risk of dementia (18–21), although not all persons with Aβ accumulation will eventually have clinically manifested dementia (22). Previous work on Aβ in the oncology field showed that patients with cancer, in particular, those with hepatic cancer, have higher plasma Aβ40 and Aβ42 levels compared with controls (23). Yet, to understand the role of Aβ in the context of the relation between Alzheimer disease and cancer, it is pivotal to study plasma Aβ before cancer diagnosis.

We hypothesized that if there is a biological link between Alzheimer disease and cancer, this should probably extend through all preclinical stages of Alzheimer disease. The life expectancy of persons in the preclinical stage of Alzheimer disease is longer than that of patients with clinically manifested Alzheimer disease, thereby limiting the effects of survival bias. We have previously shown in the prospective population-based Rotterdam Study that higher levels of plasma Aβ42 are associated with a lower risk of Alzheimer disease (21). Given that early-stage Alzheimer disease is characterized by higher plasma Aβ levels, whereas Aβ levels decrease during disease progression, any association between plasma Aβ and cancer may support a biological association between Alzheimer disease and cancer. A positive association might suggest that the very early stage of Alzheimer disease is related to cancer, whereas a negative association might reflect a link between a later preclinical stage of Alzheimer disease and cancer. We therefore determined the association between plasma Aβ40 and Aβ42 levels and the risk of cancer in the Rotterdam Study.

Study population

This study is embedded within the Rotterdam Study, a prospective population-based cohort designed to study the occurrence and determinants of diseases in the elderly population. The details of this cohort have been described in detail previously (24). Briefly, in 1990, all inhabitants ages 55 years and over from the Ommoord area, a suburb in Rotterdam, the Netherlands, were invited to participate. This initial cohort (RS-I) consisted of 7,983 participants and was subsequently extended with a second subcohort (RS-II) in 2000 with 3,011 participants who had reached the age of 55 years or moved into the study area. In 2006, the cohort was further expanded (RS-III) with 3,932 participants ages 45 years and over. In total, the Rotterdam Study comprises 14,926 participants (overall response rate 72%).

The Rotterdam Study complies with the Declaration of Helsinki and has been approved by the Medical Ethics Committee of Erasmus Medical Center and by the board of The Netherlands Ministry of Health, Welfare, and Sports. A written informed consent was obtained from all participants.

The population for this study was defined by availability of −80°C stored plasma samples obtained from participants during the fourth visit of RS-I between January 2002 and July 2004, and the second visit of RS-II between July 2002 and December 2005. From this selection of 5,094 participants with plasma samples available, we excluded participants who did not provide informed consent to access medical records and hospital discharge letters during follow-up (n = 6), those with a history of cancer at blood sample draw (n = 408), participants with a history of all-cause dementia (n = 25) or insufficient data to determine their cognitive status (n = 1), and participants with missing (n = 471) or invalid test results for plasma Aβ40 and Aβ42 levels (n = 234). Missing and invalid test results were missing at random. As a result, the final sample included 3,949 participants for analyses.

Assessment of plasma Aβ40 and Aβ42

Blood samples of participants were collected during the research center visit. Blood was sampled in tubes containing ethylenediaminetetraacetic acid and was centrifuged, of which subsequently plasma was aliquoted and frozen at −80°C according to standard procedures. Measurements of plasma Aβ levels were performed at Quanterix (Lexington) on a Simoa HD-1 analyzer platform using the Simoa Human Neurology 3-Plex A assay (25). Measurements were done in two separate batches and samples were tested in duplicate. Two quality control samples were run on each plate for each analyte. When duplicate measurements were missing (e.g., blood sample was not good) or inconsistent, or if the concentration coefficient of variation exceeded 20%, participant's samples were not valid and were therefore not included.

Assessment of cancer

Registration of prevalent and incident cancer diagnoses was based on medical records of general practitioners (including hospital discharge letters) and through linkage with the Netherlands Cancer Registry, Dutch Hospital Data, and histology and cytopathology registries in the region (PALGA). Each diagnosis of cancer was coded independently by two physicians and classified according to the International Classification of Diseases, 10th revision. In case of discrepancy between the two physicians, consensus was sought through consultation with a physician specialized in internal medicine. Cancer diagnosis was defined as any primary malignant tumor that was confirmed by pathology, excluding nonmelanoma skin cancer. Date of diagnosis was based on date of biopsy for solid tumors, laboratory assessment for hematologic tumors, or—if information on these dates was unavailable—date of hospital admission or hospital discharge letter. Follow-up of cancer registration was completed up to January 1, 2014. Only non-CNS cancers were included in the primary analysis.

Other assessments

Participants provided information on educational level, smoking habits, and alcohol use during the home interview. Educational level was classified into primary, lower (lower or intermediate general education or lower vocational education), intermediate (intermediate vocational education or higher general education), or higher (higher vocational education or university). Smoking habits were classified into never, current, or former smoker. Alcohol use was categorized as any use or no use.

Height and weight were measured at the research center from which the body mass index (kg/m2) was computed. Hypertension was defined as a systolic blood pressure ≥140 mm Hg, a diastolic blood pressure ≥90 mm Hg, or use of antihypertensive medication (26). Hypercholesterolemia was defined as a serum total cholesterol >6.5 mmol/L or use of lipid-lowering medication. Diabetes mellitus was defined as a fasting serum glucose level ≥7.1 mmol/L, a random serum glucose level ≥11.1 mmol/L, or use of antidiabetic medication (27). Symptoms of depression were assessed using the Center for Epidemiologic Studies Depression scale (CES-D) and were converted into a sum score. Apolipoprotein E (APOE) genotype was determined using polymerase chain reaction on coded DNA samples in the subcohort RS-I and with a biallelic TaqMan assay in the subcohort RS-II (28, 29). APOE ϵ4 carrier status was defined as carrier of at least one APOE ϵ4 allele. Granulocyte, lymphocyte, and platelet counts were quantified using the COULTER Ac·T diff2 Hematology Analyzer (Beckman Coulter). Granulocyte-to-lymphocyte ratio (GLR) was calculated as the ratio of granulocyte to lymphocyte count and platelet-to-lymphocyte ratio (PLR) as the ratio of platelet to lymphocyte count. Systemic immune-inflammation index (SII) was calculated as platelet count times GLR. Serum creatinine was measured with an enzymatic assay method and was subsequently standardized to isotope–dilution mass spectrometry–traceable measurements (30).

Incidence of all-cause dementia was evaluated by screening of participants at the research center visit using the Mini-Mental State Examination and the Geriatric Mental Schedule organic level. Participants with a Mini-Mental State Examination score below 26 or Geriatric Mental Schedule score of at least one underwent further investigation, including the Cambridge Examination for Mental Disorders of the Elderly. In addition, the cohort was electronically linked with medical records from general practitioners and the regional institute for outpatient mental health care to ensure continuous surveillance for all-cause dementia (31).

Statistical analysis

We used Cox proportional hazards models to obtain HRs and 95% CIs to investigate the association between plasma Aβ40 and Aβ42 levels with risk of non-CNS cancer. Aβ levels were log2 transformed to reach a normal distribution and were subsequently standardized. The standard deviation (SD) increase in Aβ level on the log2 scale can then be multiplied with 2 to obtain the corresponding SD increase in Aβ level on the original scale. As yet, neither reference values for plasma Aβ nor thresholds for preclinical Alzheimer disease are available. We therefore investigated the association with continuous Aβ and Aβ levels divided into quartiles, using two nested models: model I was adjusted for age at blood sample draw, sex, and assay batch number; model II was model I plus additional adjustment for covariates related to both Alzheimer disease and cancer, including education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking status, alcohol use, depression (as measured with CES-D sum score), and inflammation (using the SII). Because of collinearity, we only adjusted for SII and not for GLR and PLR. Follow-up time was used as timescale. Follow-up time started at date of blood sample draw (i.e., at the fourth visit of participants of RS-I and at the second visit of participants of RS-II) until date of incident cancer, all-cause dementia, death, loss to follow-up, or January 1, 2014, whichever came first. Participants were censored at date of CNS cancer diagnosis because we hypothesized that the potential mechanisms underlying any association between Aβ and cancer would differ between non-CNS and CNS cancers, given that CNS cancer can cause direct damage to the brain (32). We repeated analyses using age as timescale instead of follow-up to verify that the choice of the timescale did not affect the results. We checked the proportional hazards assumption by visual inspection of the Schoenfeld residuals.

To further evaluate the robustness of our findings, we performed two sensitivity analyses: (i) adjusting for creatinine to minimize the effect of impaired kidney function because plasma Aβ is partly cleared by the kidney and (ii) excluding the first 2 and 5 years of follow-up time to examine reverse causation (i.e., the possible effect of subclinical cancer on plasma Aβ). In this latter analysis, only those participants with a follow-up time of more than, respectively, 2 and 5 years were included.

We studied effect modification for median age, sex, education, smoking status, APOE ϵ4 carrier status, and inflammatory ratios by stratification and by adding multiplicative interaction terms to the model. Finally, we stratified analyses by primary cancer site.

We used multiple imputation for missing covariates (maximum of 3.2%), with five imputed datasets based on the covariates and outcome. We used Rubin method for pooled HRs and 95% CIs (33). Statistical analyses were performed using the R package “survival” in RStudio Version 3.3.2 (34).

Characteristics of the study population are presented in Table 1. The median [interquartile range (IQR)] age at blood sample draw was 70.4 years (65.8–76.4) and 57.7% were women. During a median (IQR) follow-up of 9.0 years (6.9–10.1), 560 of 3,949 participants were diagnosed with cancer. In the same follow-up period, 303 participants developed all-cause dementia (of whom 247 with Alzheimer disease) and 712 participants died. Most frequently diagnosed cancer sites were breast (30.4% among women), male genital organs (29.1% among men), colorectal (16.3%), and lung (14.5%).

Table 1.

Characteristics of the study population.

Included participantsExcluded participants
Characteristic(N = 3,949)(N = 705)a
Ageb 70.4 years (65.8–76.4) 75.6 years (69.7–80.9) 
Sex 
 Women 2,277 (57.7) 442 (62.7) 
 Men 1,672 (42.3) 263 (37.3) 
Educational level 
 Primary 421 (10.7) 118 (16.7) 
 Lower 1,711 (43.3) 318 (45.1) 
 Intermediate 1,200 (30.4) 194 (27.5) 
 Higher 557 (14.1) 69 (9.8) 
Body mass indexc 27.6 kg/m2 (4.1) 27.6 kg/m2 (4.4) 
Hypertension 
 No 855 (21.7) 109 (15.5) 
 Yes 3,091 (78.3) 595 (84.4) 
Hypercholesterolemia 
 No 2,346 (59.4) 434 (61.6) 
 Yes 1,580 (40.0) 263 (37.3) 
Diabetes mellitus 
 No 3,400 (86.1) 562 (79.7) 
 Yes 421 (10.7) 100 (14.2) 
Smoking 
 Never 1,159 (29.3) 221 (31.3) 
 Former 2,222 (56.3) 394 (55.9) 
 Current 491 (12.4) 70 (9.9) 
Alcohol use 
 No 553 (14.0) 120 (17.0) 
 Yes 3,319 (84.0) 565 (80.1) 
CES-D sum scoreb 3 (1–8) 5 (1–10) 
APOE ϵ4 carrier status 
 Noncarrier 2,787 (70.6) 494 (70.1) 
 Carrier 1,044 (26.4) 186 (26.4) 
Inflammatory ratiosb 
 Granulocyte-to-lymphocyte ratio 1.8 (1.4–2.4) 2.0 (1.5–2.5) 
 Platelet-to-lymphocyte ratio 120 (96–153) 128 (100–161) 
 Systemic immune-inflammation index 455 (338–621) 497 (355–682) 
Plasma Aβ40b 258 pg/mL (229–291)  
Plasma Aβ42b 10.3 pg/mL (8.9–11.9)  
Included participantsExcluded participants
Characteristic(N = 3,949)(N = 705)a
Ageb 70.4 years (65.8–76.4) 75.6 years (69.7–80.9) 
Sex 
 Women 2,277 (57.7) 442 (62.7) 
 Men 1,672 (42.3) 263 (37.3) 
Educational level 
 Primary 421 (10.7) 118 (16.7) 
 Lower 1,711 (43.3) 318 (45.1) 
 Intermediate 1,200 (30.4) 194 (27.5) 
 Higher 557 (14.1) 69 (9.8) 
Body mass indexc 27.6 kg/m2 (4.1) 27.6 kg/m2 (4.4) 
Hypertension 
 No 855 (21.7) 109 (15.5) 
 Yes 3,091 (78.3) 595 (84.4) 
Hypercholesterolemia 
 No 2,346 (59.4) 434 (61.6) 
 Yes 1,580 (40.0) 263 (37.3) 
Diabetes mellitus 
 No 3,400 (86.1) 562 (79.7) 
 Yes 421 (10.7) 100 (14.2) 
Smoking 
 Never 1,159 (29.3) 221 (31.3) 
 Former 2,222 (56.3) 394 (55.9) 
 Current 491 (12.4) 70 (9.9) 
Alcohol use 
 No 553 (14.0) 120 (17.0) 
 Yes 3,319 (84.0) 565 (80.1) 
CES-D sum scoreb 3 (1–8) 5 (1–10) 
APOE ϵ4 carrier status 
 Noncarrier 2,787 (70.6) 494 (70.1) 
 Carrier 1,044 (26.4) 186 (26.4) 
Inflammatory ratiosb 
 Granulocyte-to-lymphocyte ratio 1.8 (1.4–2.4) 2.0 (1.5–2.5) 
 Platelet-to-lymphocyte ratio 120 (96–153) 128 (100–161) 
 Systemic immune-inflammation index 455 (338–621) 497 (355–682) 
Plasma Aβ40b 258 pg/mL (229–291)  
Plasma Aβ42b 10.3 pg/mL (8.9–11.9)  

Note: Characteristics were measured during the visit of the blood sample draw (i.e., at the fourth visit of participants of RS-I and at the second visit of participants of RS-II). Data are presented as frequency (percentage) unless otherwise indicated. Missing values are not imputed, and, therefore, numbers do not always sum up to 100%.

aExcluded participants in this table includes only those participants who were excluded because of missing test results (n = 471) or invalid test results (n = 234).

bPresented as median (interquartile range).

cPresented as mean (SD).

Higher levels of log2 plasma Aβ40 and Aβ42 were associated with a higher risk of cancer [HR per SD increase in Aβ40 = 1.12 (95% CI = 1.02;1.23) and Aβ42 = 1.12 (95% CI = 1.03–1.23); Table 2]. Given that the unadjusted HR for the risk of cancer per 1 year increase in age is 1.01 (95% CI = 1.00–1.02), the HR per SD increase in Aβ corresponds with a risk increase in cancer of at least 10.2 years of age (the unrounded HR for age is 1.0117 and for Aβ40 is 1.1193, thus 0.1193/0.0117 = 10.2). Participants with plasma Aβ40 and Aβ42 levels in the highest quartile had a higher risk of cancer than those with levels in the lowest quartile [HR for Aβ40 = 1.43 (95% CI = 1.11–1.83) and Aβ42 = 1.38 (95% CI = 1.09–1.76)]. Additional adjustment for creatinine did not meaningfully change the estimated HRs (Table 3). Also, exclusion of the first 2 and 5 years of follow-up time did not affect the risk estimates (Table 3).

Table 2.

The association between standardized log2-transformed plasma Aβ40 and Aβ42 levels with risk of cancer.

Cancer
Model IModel II
Plasma assessment (pg/mL)an/NHR (95% CI)HR (95% CI)
Aβ40 
Per SD increase 560/3,949 1.13 (1.03–1.24) 1.12 (1.02–1.23) 
Quartiles (range)    
 1st quartile (−8.05 to −0.61) 126/988 1.00 1.00 
 2nd quartile (−0.61 to −0.01) 138/987 1.10 (0.86–1.40) 1.11 (0.87–1.42) 
 3rd quartile (−0.01 to 0.59) 136/987 1.12 (0.87–1.43) 1.13 (0.88–1.44) 
 4th quartile (0.59–4.24) 160/987 1.47 (1.14–1.88) 1.43 (1.11–1.83) 
Ptrend  .004 .005 
Aβ42 
Per SD increase 560/3,949 1.12 (1.03–1.23) 1.12 (1.03–1.23) 
Quartiles (range)    
 1st quartile (−11.4 to −0.53) 124/988 1.00 1.00 
 2nd quartile (−0.53 to 0.03) 136/987 1.11 (0.87–1.42) 1.13 (0.88–1.44) 
 3rd quartile (0.03–0.59) 140/987 1.14 (0.89–1.45) 1.16 (0.91–1.47) 
 4th quartile (0.59–8.46) 160/987 1.38 (1.09–1.75) 1.38 (1.09–1.76) 
Ptrend  .009 .009 
Cancer
Model IModel II
Plasma assessment (pg/mL)an/NHR (95% CI)HR (95% CI)
Aβ40 
Per SD increase 560/3,949 1.13 (1.03–1.24) 1.12 (1.02–1.23) 
Quartiles (range)    
 1st quartile (−8.05 to −0.61) 126/988 1.00 1.00 
 2nd quartile (−0.61 to −0.01) 138/987 1.10 (0.86–1.40) 1.11 (0.87–1.42) 
 3rd quartile (−0.01 to 0.59) 136/987 1.12 (0.87–1.43) 1.13 (0.88–1.44) 
 4th quartile (0.59–4.24) 160/987 1.47 (1.14–1.88) 1.43 (1.11–1.83) 
Ptrend  .004 .005 
Aβ42 
Per SD increase 560/3,949 1.12 (1.03–1.23) 1.12 (1.03–1.23) 
Quartiles (range)    
 1st quartile (−11.4 to −0.53) 124/988 1.00 1.00 
 2nd quartile (−0.53 to 0.03) 136/987 1.11 (0.87–1.42) 1.13 (0.88–1.44) 
 3rd quartile (0.03–0.59) 140/987 1.14 (0.89–1.45) 1.16 (0.91–1.47) 
 4th quartile (0.59–8.46) 160/987 1.38 (1.09–1.75) 1.38 (1.09–1.76) 
Ptrend  .009 .009 

Note: HRs in model I are adjusted for age at blood draw, sex, and assay batch number. HRs in model II are adjusted for covariates in model I plus adjustment for education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index.

Abbreviations: n, number of participants with incident cancer; N, total number for participants.

aPlasma assessments are log2 transformed and standardized.

Table 3.

Sensitivity analyses for the association between plasma Aβ40 and Aβ42 levels with risk of cancer.

Cancer
Plasma assessment (pg/mL)an/NHR (95% CI)
Additional adjustment for creatinine level (μmol/L)b 
Aβ40 229/1,514 1.13 (0.96–1.33) 
Aβ42 229/1,514 1.10 (0.94–1.28) 
Excluding first 2 years of follow-up 
Aβ40 438/3,672 1.08 (0.97–1.20) 
Aβ42 438/3,672 1.07 (0.97–1.18) 
Excluding first 5 years of follow-up 
Aβ40 242/3,271 1.12 (0.96–1.29) 
Aβ42 242/3,271 1.15 (1.01–1.32) 
Cancer
Plasma assessment (pg/mL)an/NHR (95% CI)
Additional adjustment for creatinine level (μmol/L)b 
Aβ40 229/1,514 1.13 (0.96–1.33) 
Aβ42 229/1,514 1.10 (0.94–1.28) 
Excluding first 2 years of follow-up 
Aβ40 438/3,672 1.08 (0.97–1.20) 
Aβ42 438/3,672 1.07 (0.97–1.18) 
Excluding first 5 years of follow-up 
Aβ40 242/3,271 1.12 (0.96–1.29) 
Aβ42 242/3,271 1.15 (1.01–1.32) 

Note: HRs are adjusted for age at blood draw, sex, assay batch number, education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index.

Abbreviations: n, number of participants with incident cancer; N, total number for participants.

aPlasma assessments are log2 transformed and standardized.

bCreatinine levels were measured in a random sample of 1,514 of 3,949 participants.

Stratified analyses showed that the association between plasma Aβ and cancer was more profound in older participants, men, former smokers, participants with an intermediate education level, and APOE ϵ4 carriers (Fig. 1). Regarding inflammatory status, participants with higher GLR had a higher risk of cancer compared with those with a lower GLR, which was not observed for PLR and SII (Supplementary Fig. S1). All interactions were tested on the multiplicative scale and did not reach statistical significance.

Figure 1.

Forest plot of association between log2-transformed and standardized plasma Aβ40 and Aβ42 levels with risk of cancer, stratified by median age at blood draw, sex, education, smoking status, and APOE ϵ4 carrier status. HRs are adjusted for age at blood draw, sex, assay batch number, education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index. APOE ϵ4 carrier status was missing for 118 participants. Missing values of education and smoking were imputed, and, therefore, the total number of participants per smoking category is higher than the number of participants presented in Table 1. n, number of participants with incident cancer; N, total number for participants.

Figure 1.

Forest plot of association between log2-transformed and standardized plasma Aβ40 and Aβ42 levels with risk of cancer, stratified by median age at blood draw, sex, education, smoking status, and APOE ϵ4 carrier status. HRs are adjusted for age at blood draw, sex, assay batch number, education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index. APOE ϵ4 carrier status was missing for 118 participants. Missing values of education and smoking were imputed, and, therefore, the total number of participants per smoking category is higher than the number of participants presented in Table 1. n, number of participants with incident cancer; N, total number for participants.

Close modal

Analyses per cancer site showed that the association was most pronounced between plasma Aβ40 and Aβ42 and hematologic cancer [HR per SD increase in log2 Aβ40 = 1.56 (95% CI = 1.12–2.17) and Aβ42 = 1.30 (95% CI = 0.94–1.79); Fig. 2]. The association was also stronger for cancer of unknown primary origin and cancer in the urinary tract, esophagus and stomach (only for Aβ40), and head and neck (only for Aβ42), albeit not statistically significantly. In a post hoc analysis, we found a strong association with pancreatic cancer, although the power was limited by the number of participants with pancreatic cancer [n = 13, HR per SD increase in log2 Aβ40 = 1.52 (95% CI = 0.83–2.79) and Aβ42 = 1.51 (95% CI = 0.98;2.31)].

Figure 2.

Association between log2-transformed and standardized plasma Aβ40 and Aβ42 levels with different cancer sites. HRs are adjusted for age at blood draw, sex, assay batch number, education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index. *Thirteen of 22 participants were diagnosed with pancreatic cancer. Higher levels of Aβ40 and Aβ42 were associated with a higher risk of pancreatic cancer [HR for Aβ40 = 1.52 (95% CI = 0.83–2.79) and Aβ42 = 1.51 (95% CI = 0.98–2.31)]. n, number of participants with incident cancer; N, total number for participants.

Figure 2.

Association between log2-transformed and standardized plasma Aβ40 and Aβ42 levels with different cancer sites. HRs are adjusted for age at blood draw, sex, assay batch number, education, body mass index, hypertension, hypercholesterolemia, diabetes mellitus, smoking, alcohol use, CES-D sum score, and systemic immune-inflammation index. *Thirteen of 22 participants were diagnosed with pancreatic cancer. Higher levels of Aβ40 and Aβ42 were associated with a higher risk of pancreatic cancer [HR for Aβ40 = 1.52 (95% CI = 0.83–2.79) and Aβ42 = 1.51 (95% CI = 0.98–2.31)]. n, number of participants with incident cancer; N, total number for participants.

Close modal

In this population-based cohort study, we found that higher plasma levels of both Aβ40 and Aβ42 were associated with a higher risk of cancer. This indicates that Aβ could be involved in the pathophysiology of cancer and may further support a potential biological link between Alzheimer disease and cancer. The direction of this link, that is, inverse or positive, and the causal effect of Aβ on cancer warrant further investigation.

Our finding is in line with a previously conducted cross-sectional study showing that individuals who have been diagnosed with cancer have higher plasma Aβ levels (23). To further understand this association and its interpretation regarding the link between Alzheimer disease and cancer, it is necessary to first elaborate on the Aβ pathway. Aβ is the product of cleavage of amyloid precursor protein (APP), which is expressed in neuronal and nonneuronal tissues including the kidney, lung, and pancreas. Aβ is formed as product of APP cleavage by α-secretase, followed by β-secretase. APP cleavage can result in different isoforms of Aβ, depending on the number of amino acids (i.e., 38, 40, and 42), with longer isoforms being more prone to aggregation (35). Although many tissues contain APP, Aβ can only be produced in cells that also express β-secretase, for example, neuronal cells, muscle cells, platelets, and vascular wall endothelial cells. Plasma Aβ can therefore have different sources.

Against this background, there are different explanations for the association between Aβ and the risk of cancer. First, the plasma Aβ can be of neuronal origin. Aβ might be reflective of an underlying early-stage Alzheimer disease process and our findings might be pointing toward a shared causal predisposition between cancer and Alzheimer disease. Although effect estimates hardly changed when excluding the first 2 and 5 years of follow-up, we cannot fully exclude reversed causation. An alternative explanation therefore is that the blood–brain barrier permeability can increase due to subclinical cancer (36). Different animal and in vitro studies have shown that systemic inflammation can disrupt the integrity of the blood–brain barrier (37, 38). Because higher inflammatory markers are associated with a higher risk of cancer (39), the integrity of the blood–brain barrier might be altered by subclinical cancer due to systemic inflammation. This could result in leakage of neuronal Aβ to the peripheral circulation. Such reverse causality is compatible with higher Aβ levels reflective of a higher risk of getting diagnosed with clinical cancer. Second, the plasma Aβ can be of nonneuronal origin. It is conceivable that platelets and vascular wall endothelial cells—the main sources of circulating Aβ (14)—are activated as response to subclinical cancer (40), resulting in more Aβ production before the cancer is diagnosed. For instance, platelets were indeed more activated in patients with multiple myeloma than in healthy controls (41). In addition, platelets promoted proliferation of multiple myeloma cells and that of acute myeloid leukemia blasts in vitro (42). This may partly explain the strong association with hematologic cancers, but it should be noted that the group of hematologic cancers is composed of different types of leukemia and lymphoma. In the same context of an extraneuronal origin of Aβ, Aβ might be produced by organs that express APP, but not β-secretase, for instance, pancreas, kidney, and lung (43). Cancer cells in these organs might be mutated such that β-secretase expression gets enhanced, resulting in Aβ production. We indeed found that the relation between plasma Aβ and cancer risk was stronger for those cancer sites with cells that express APP. Interestingly, APP is upregulated in pancreatic cancer cells (44). In light of this, we examined the association between Aβ and pancreatic cancer in a post hoc analysis. Although limited by power, we found that higher plasma Aβ levels were associated with a higher risk of pancreatic cancer. Third, shared mechanisms such as inflammation can cause both higher levels of plasma Aβ and cancer (23). Higher inflammatory ratios are associated with a higher risk of cancer, indicating a proinflammatory state before cancer diagnosis, and interestingly have also been linked with Alzheimer disease (40, 45). This might also explain why the association between plasma Aβ and cancer was more pronounced—albeit not statistically significantly—in APOE ϵ4 carriers than in noncarriers. Human cell studies and animal studies have shown that APOE4 may predispose cells to inflammation and may promote a greater inflammatory response following immune activation than other APOE isoforms, resulting in the secretion of inflammatory factors (46). Figure 3 summarizes the potential biological mechanisms underlying the association between plasma Aβ and cancer. Finally, we cannot completely rule out that the association is partly driven by methodologic bias. We have previously shown that lower plasma Aβ are associated with a higher risk of Alzheimer disease (21). Given this association, we might expect that—if cancer and Alzheimer disease are positively associated—lower plasma Aβ levels would be associated with a higher risk of cancer. The finding that higher plasma Aβ levels are related to a higher risk of cancer might be explained because those persons with low plasma Aβ levels are more likely to develop dementia before they might have been diagnosed with cancer. Consequently, higher plasma Aβ levels are associated with a higher risk of cancer.

Figure 3.

Overview of potential mechanisms underlying the association between plasma Aβ and the risk of cancer. There are different sources of plasma Aβ: neuronal cells (A) and nonneuronal cells (B and C). Neuronal Aβ production might reflect an underlying Alzheimer disease process (A; this image was modified from Servier Medical Art, licensed under a Creative Common Attribution 3.0 Generic License. http://smart.servier.com/). Extraneuronal Aβ production can be caused by activation of platelets and vascular wall cells (C). This activation might be due to subclinical cancer. In addition, subclinical cancer cells might produce Aβ if they also express β-secretase due to mutations. Inflammation can stimulate the pathogenesis of both Alzheimer disease and cancer (B), and could therefore indirectly lead to higher plasma Aβ levels (D). In turn, higher plasma Aβ levels are associated with a higher risk of cancer (E). However, the causality of this association warrants further investigation.

Figure 3.

Overview of potential mechanisms underlying the association between plasma Aβ and the risk of cancer. There are different sources of plasma Aβ: neuronal cells (A) and nonneuronal cells (B and C). Neuronal Aβ production might reflect an underlying Alzheimer disease process (A; this image was modified from Servier Medical Art, licensed under a Creative Common Attribution 3.0 Generic License. http://smart.servier.com/). Extraneuronal Aβ production can be caused by activation of platelets and vascular wall cells (C). This activation might be due to subclinical cancer. In addition, subclinical cancer cells might produce Aβ if they also express β-secretase due to mutations. Inflammation can stimulate the pathogenesis of both Alzheimer disease and cancer (B), and could therefore indirectly lead to higher plasma Aβ levels (D). In turn, higher plasma Aβ levels are associated with a higher risk of cancer (E). However, the causality of this association warrants further investigation.

Close modal

Some limitations of this study need to be addressed. First, we had no measurements of Aβ within the brain, for instance, measured in cerebrospinal fluid (CSF) or by amyloid PET neuroimaging. However, previous studies showed correlation between plasma Aβ and CSF or amyloid PET (16, 47). Second, we cannot determine the origin of the measured plasma Aβ levels. Third, although we tried to take the effect of reverse causation into account by excluding the first 2 and 5 years of follow-up, we cannot determine the causal effect of Aβ on cancer. The length of the latency period between cancer initiation and manifestation differs per cancer site and can range from 5 to 40 years for solid tumors (48). Fourth, it should be noted that although the missing and invalid plasma Aβ levels were missing at random, the characteristics of participants with known plasma Aβ levels differed from those of participants with unknown or invalid plasma Aβ levels. This, as well as that most of the included participants were middle class persons of European descent (98.5%), could possibly limit the generalizability of our findings to other populations. Strengths include using Aβ as proxy for preclinical Alzheimer disease to circumvent surveillance and survival bias, the large sample size, and the inclusion of participants with different cancer sites. Although this enabled us to explore the association between plasma Aβ and different cancer sites, the groups of different cancer sites were heterogeneous and analyses were limited by the low number of cases per cancer site.

In conclusion, we found that higher plasma Aβ40 and Aβ42 levels are associated with a higher risk of cancer. This finding may support a potential biological link between Alzheimer disease and cancer. Also, this association may indicate a potential pathophysiologic role of Aβ in cancer, outside the context of Alzheimer disease. The causality of this association warrant further investigation, for instance, by investigating the trajectory of plasma Aβ levels before cancer diagnosis.

No potential conflicts of interest were disclosed.

The funding source had no role in study design, collection, analysis, interpretation of data, writing of the report, or decision to submit the article for publication.

K.D. van der Willik: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft. M. Ghanbari: Data curation, supervision, investigation, methodology, writing–review and editing. L. Fani: Investigation, writing–review and editing. A. Compter: Supervision, investigation, writing–review and editing. R. Ruiter: Data curation, supervision, writing–review and editing. B.H.Ch. Stricker: Data curation, supervision, writing–review and editing. S.B. Schagen: Conceptualization, supervision, funding acquisition, investigation, methodology, writing–review and editing. M.A. Ikram: Conceptualization, supervision, investigation, methodology, writing–review and editing.

The authors gratefully thank all Rotterdam Study participants and staff for their time and commitment to the study. The authors acknowledge Frank van Rooij as data manager, and Jolande Verkroost-van Heemst for her contribution to the collection of the data.

This work was supported by the Dutch Cancer Society (to S.B. Schagen, grant number NKI-20157737). The Rotterdam Study is funded by Erasmus Medical Center and Erasmus University Rotterdam; the Netherlands Organization for the Health Research and Development (ZonMw); the Research Institute for Diseases in the Elderly (RIDE); the Ministry of Education, Culture and Science; the Ministry for Health, Welfare and Sports; the European Commission (DG XII); and the Municipality of Rotterdam.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Heron
M
. 
Deaths: leading causes for 2016
.
Natl Vital Stat Rep
2018
;
67
:
1
77
.
2.
Siegel
RL
,
Miller
KD
,
Jemal
A
. 
Cancer statistics, 2019
.
CA Cancer J Clin
2019
;
69
:
7
34
.
3.
Shi
HB
,
Tang
B
,
Liu
YW
,
Wang
XF
,
Chen
GJ
. 
Alzheimer disease and cancer risk: a meta-analysis
.
J Cancer Res Clin Oncol
2015
;
141
:
485
94
.
4.
Freedman
DM
,
Wu
J
,
Chen
H
,
Kuncl
RW
,
Enewold
LR
,
Engels
EA
, et al
Associations between cancer and Alzheimer's disease in a U.S. Medicare population
.
Cancer Med
2016
;
5
:
2965
76
.
5.
Bowles
EJA
,
Walker
RL
,
Anderson
ML
,
Dublin
S
,
Crane
PK
,
Larson
EB
. 
Risk of Alzheimer's disease or dementia following a cancer diagnosis
.
PLoS One
2017
;
12
:
e0179857
.
6.
Frain
L
,
Swanson
D
,
Cho
K
,
Gagnon
D
,
Lu
KP
,
Betensky
RA
, et al
Association of cancer and Alzheimer's disease risk in a national cohort of veterans
.
Alzheimers Dement
2017
;
13
:
1364
70
.
7.
Schmidt
SA
,
Ording
AG
,
Horvath-Puho
E
,
Sorensen
HT
,
Henderson
VW
. 
Non-melanoma skin cancer and risk of Alzheimer's disease and all-cause dementia
.
PLoS One
2017
;
12
:
e0171527
.
8.
Hanson
HA
,
Horn
KP
,
Rasmussen
KM
,
Hoffman
JM
,
Smith
KR
. 
Is cancer protective for subsequent Alzheimer's disease risk? Evidence from the Utah Population Database
.
J Gerontol B Psychol Sci Soc Sci
2017
;
72
:
1032
43
.
9.
van der Willik
KD
,
Schagen
SB
,
Ikram
MA
. 
Cancer and dementia: two sides of the same coin?
Eur J Clin Invest
2018
;
48
:
e13019
.
10.
van der Willik
KD
,
Ruiter
R
,
Wolters
FJ
,
Ikram
MK
,
Stricker
BH
,
Hauptmann
M
, et al
Mild cognitive impairment and dementia show contrasting associations with risk of cancer
.
Neuroepidemiology
2018
;
50
:
207
15
.
11.
Nudelman
KNH
,
McDonald
BC
,
Lahiri
DK
,
Saykin
AJ
. 
Biological Hallmarks of cancer in Alzheimer's disease
.
Mol Neurobiol
2019
;
56
:
7173
87
.
12.
Serrano-Pozo
A
,
Frosch
MP
,
Masliah
E
,
Hyman
BT
. 
Neuropathological alterations in Alzheimer disease
.
Cold Spring Harb Perspect Med
2011
;
1
:
a006189
.
13.
Sperling
RA
,
Aisen
PS
,
Beckett
LA
,
Bennett
DA
,
Craft
S
,
Fagan
AM
, et al
Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease
.
Alzheimers Dement
2011
;
7
:
280
92
.
14.
Chen
M
,
Inestrosa
NC
,
Ross
GS
,
Fernandez
HL
. 
Platelets are the primary source of amyloid β-peptide in human blood
.
Biochem Biophys Res Commun
1995
;
213
:
96
103
.
15.
Shibata
M
,
Yamada
S
,
Kumar
SR
,
Calero
M
,
Bading
J
,
Frangione
B
, et al
Clearance of Alzheimer's amyloid-ss(1–40) peptide from brain by LDL receptor-related protein-1 at the blood-brain barrier
.
J Clin Invest
2000
;
106
:
1489
99
.
16.
Janelidze
S
,
Stomrud
E
,
Palmqvist
S
,
Zetterberg
H
,
van Westen
D
,
Jeromin
A
, et al
Plasma β-amyloid in Alzheimer's disease and vascular disease
.
Sci Rep
2016
;
6
:
26801
.
17.
Kawarabayashi
T
,
Younkin
LH
,
Saido
TC
,
Shoji
M
,
Ashe
KH
,
Younkin
SG
. 
Age-dependent changes in brain, CSF, and plasma amyloid (beta) protein in the Tg2576 transgenic mouse model of Alzheimer's disease
.
J Neurosci
2001
;
21
:
372
81
.
18.
van Oijen
M
,
Hofman
A
,
Soares
HD
,
Koudstaal
PJ
,
Breteler
MM
. 
Plasma Abeta(1-40) and Abeta(1-42) and the risk of dementia: a prospective case-cohort study
.
Lancet Neurol
2006
;
5
:
655
60
.
19.
Koyama
A
,
Okereke
OI
,
Yang
T
,
Blacker
D
,
Selkoe
DJ
,
Grodstein
F
. 
Plasma amyloid-β as a predictor of dementia and cognitive decline: a systematic review and meta-analysis
.
Arch Neurol
2012
;
69
:
824
31
.
20.
Hilal
S
,
Wolters
FJ
,
Verbeek
MM
,
Vanderstichele
H
,
Ikram
MK
,
Stoops
E
, et al
Plasma amyloid-beta levels, cerebral atrophy and risk of dementia: a population-based study
.
Alzheimers Res Ther
2018
;
10
:
63
.
21.
de Wolf
F
,
Ghanbari
M
,
Licher
S
,
McRae-McKee
K
,
Gras
L
,
Weverling
GJ
, et al
Plasma tau, neurofilament light chain and amyloid-β levels and risk of dementia; a population-based cohort study
.
Brain
2020
;
143
:
1220
32
.
22.
Rodrigue
KM
,
Kennedy
KM
,
Devous
MD
 Sr
,
Rieck
JR
,
Hebrank
AC
,
Diaz-Arrastia
R
, et al
β-Amyloid burden in healthy aging: regional distribution and cognitive consequences
.
Neurology
2012
;
78
:
387
95
.
23.
Jin
WS
,
Bu
XL
,
Liu
YH
,
Shen
LL
,
Zhuang
ZQ
,
Jiao
SS
, et al
Plasma amyloid-beta levels in patients with different types of cancer
.
Neurotox Res
2017
;
31
:
283
8
.
24.
Ikram
MA
,
Brusselle
G
,
Ghanbari
M
,
Goedegebure
A
,
Ikram
MK
,
Kavousi
M
, et al
Objectives, design and main findings until 2020 from the Rotterdam Study
.
Eur J Epidemiol
2020
;
35
:
483
517
.
25.
Rissin
DM
,
Fournier
DR
,
Piech
T
,
Kan
CW
,
Campbell
TG
,
Song
L
, et al
Simultaneous detection of single molecules and singulated ensembles of molecules enables immunoassays with broad dynamic range
.
Anal Chem
2011
;
83
:
2279
85
.
26.
Engberink
MF
,
Bakker
SJ
,
Brink
EJ
,
van Baak
MA
,
van Rooij
FJ
,
Hofman
A
, et al
Dietary acid load and risk of hypertension: the Rotterdam Study
.
Am J Clin Nutr
2012
;
95
:
1438
44
.
27.
Diabetes mellitus. Report of a WHO study group
.
World Health Organ Tech Rep Ser
1985
;
727
:
1
113
.
28.
Wingbermuhle
R
,
Wen
KX
,
Wolters
FJ
,
Ikram
MA
,
Bos
D
. 
Smoking, APOE genotype, and cognitive decline: the Rotterdam Study
.
J Alzheimers Dis
2017
;
57
:
1191
5
.
29.
Slooter
AJ
,
Cruts
M
,
Kalmijn
S
,
Hofman
A
,
Breteler
MM
,
Van Broeckhoven
C
, et al
Risk estimates of dementia by apolipoprotein E genotypes from a population-based incidence study: the Rotterdam Study
.
Arch Neurol
1998
;
55
:
964
8
.
30.
Sedaghat
S
,
de Vries
PS
,
Boender
J
,
Sonneveld
MA
,
Hoorn
EJ
,
Hofman
A
, et al
von Willebrand factor, ADAMTS13 activity, and decline in kidney function: a population-based cohort study
.
Am J Kidney Dis
2016
;
68
:
726
32
.
31.
de Bruijn
RF
,
Bos
MJ
,
Portegies
ML
,
Hofman
A
,
Franco
OH
,
Koudstaal
PJ
, et al
The potential for prevention of dementia across two decades: the prospective, population-based Rotterdam Study
.
BMC Med
2015
;
13
:
132
.
32.
Coomans
MB
,
van der Linden
SD
,
Gehring
K
,
Taphoorn
MJB
. 
Treatment of cognitive deficits in brain tumour patients: current status and future directions
.
Curr Opin Oncol
2019
;
31
:
540
7
.
33.
Rubin
DB
.
Multiple imputation for nonresponse in surveys
.
Hoboken (NJ)
:
John Wiley & Sons, Inc.
; 
1987
.
34.
Therneau
TM
.
A package for survival analysis in R
.
New York
:
Springer
; 
2005
.
35.
Bitan
G
,
Kirkitadze
MD
,
Lomakin
A
,
Vollers
SS
,
Benedek
GB
,
Teplow
DB
. 
Amyloid beta -protein (Abeta) assembly: Abeta 40 and Abeta 42 oligomerize through distinct pathways
.
Proc Natl Acad Sci U S A
2003
;
100
:
330
5
.
36.
Olson
B
,
Marks
DL
. 
Pretreatment cancer-related cognitive impairment-mechanisms and outlook
.
Cancers
2019
;
11
:
687
.
37.
Banks
WA
,
Gray
AM
,
Erickson
MA
,
Salameh
TS
,
Damodarasamy
M
,
Sheibani
N
, et al
Lipopolysaccharide-induced blood-brain barrier disruption: roles of cyclooxygenase, oxidative stress, neuroinflammation, and elements of the neurovascular unit
.
J Neuroinflammation
2015
;
12
:
223
.
38.
de Vries
HE
,
Blom-Roosemalen
MC
,
van Oosten
M
,
de Boer
AG
,
van Berkel
TJ
,
Breimer
DD
, et al
The influence of cytokines on the integrity of the blood-brain barrier in vitro
.
J Neuroimmunol
1996
;
64
:
37
43
.
39.
Fest
J
,
Ruiter
R
,
Mulder
M
,
Groot Koerkamp
B
,
Ikram
MA
,
Stricker
BH
, et al
The systemic immune-inflammation index is associated with an increased risk of incident cancer-a population-based cohort study
.
Int J Cancer
2020
;
146
:
692
8
.
40.
Buergy
D
,
Wenz
F
,
Groden
C
,
Brockmann
MA
. 
Tumor-platelet interaction in solid tumors
.
Int J Cancer
2012
;
130
:
2747
60
.
41.
Takagi
S
,
Tsukamoto
S
,
Park
J
,
Johnson
KE
,
Kawano
Y
,
Moschetta
M
, et al
Platelets enhance multiple myeloma progression via IL-1β upregulation
.
Clin Cancer Res
2018
;
24
:
2430
9
.
42.
Yan
M
,
Jurasz
P
. 
The role of platelets in the tumor microenvironment: from solid tumors to leukemia
.
Biochim Biophys Acta
2016
;
1863
:
392
400
.
43.
Arai
H
,
Lee
VM
,
Messinger
ML
,
Greenberg
BD
,
Lowery
DE
,
Trojanowski
JQ
. 
Expression patterns of beta-amyloid precursor protein (beta-APP) in neural and nonneural human tissues from Alzheimer's disease and control subjects
.
Ann Neurol
1991
;
30
:
686
93
.
44.
Hansel
DE
,
Rahman
A
,
Wehner
S
,
Herzog
V
,
Yeo
CJ
,
Maitra
A
. 
Increased expression and processing of the Alzheimer amyloid precursor protein in pancreatic cancer may influence cellular proliferation
.
Cancer Res
2003
;
63
:
7032
7
.
45.
van der Willik
KD
,
Fani
L
,
Rizopoulos
D
,
Licher
S
,
Fest
J
,
Schagen
SB
, et al
Balance between innate versus adaptive immune system and the risk of dementia: a population-based cohort study
.
J Neuroinflammation
2019
;
16
:
68
.
46.
Dorey
E
,
Chang
N
,
Liu
QY
,
Yang
Z
,
Zhang
W
. 
Apolipoprotein E, amyloid-beta, and neuroinflammation in Alzheimer's disease
.
Neurosci Bull
2014
;
30
:
317
30
.
47.
Blennow
K
. 
A review of fluid biomarkers for Alzheimer's disease: moving from CSF to blood
.
Neurol Ther
2017
;
6
:
15
24
.
48.
Nadler
D
,
Zurbenko
I
. 
Developing a Weibull model extension to estimate cancer latency
.
ISRN Epidemiology
2012
;
2013
:
750857
.