Supplement

Palmqvist et al., 2018

Supplement Accurate risk estimation of β-amyloid positivity to identify prodromal Alzheimer’s disease: crossvalidation study of practical algorithms METHODS OF THE BIOFINDER STUDY Inclusion criteria were that patients (1) were referred to any of the three participating memory clinics because of cognitive complaints, (2) did not fulfil the criteria for dementia[1], (3) had a Mini-Mental State Examination (MMSE) score of 24 to 30 points[2], (4) were aged 60 to 80 years, and (5) were fluent in Swedish. The exclusion criteria were (1) cognitive impairment that without doubt could be explained by another condition (other than prodromal dementia), (2) severe somatic disease, and (3) refusing lumbar puncture or neuropsychological investigation. Only cross-sectional data were used to establish and test the Aβ accuracy of the models. The study was approved by the ethical review board in Lund, Sweden, and all participants gave their written informed consent.

METHODS OF THE ADNI STUDY STUDY DESIGN The ADNI was launched in 2003 by the National Institute on Aging (NI A), the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the Food and Drug Administration (FDA), private pharmaceutical companies and non-profit organizations, as a $60 million, 5-year public-private partnership. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early Alzheimer’s disease (AD). Determination of sensitive and specific markers of very early AD progression is intended to aid researchers and clinicians to develop new treatments and monitor their effectiveness, as well as lessen the time and cost of clinical trials. The Principal Investigator of this initiative is Michael W. Weiner, MD, VA Medical Center and University of California – San Francisco. ADNI is the result of efforts of many co-investigators from a broad range of academic institutions and private corporations, and subjects have been recruited from over 50 sites across the U.S. and Canada. The initial goal of ADNI was to recruit 800 subjects but ADNI has been followed by ADNI-GO and ADNI-2. To date these three protocols have recruited over 1500 adults, ages 55 to 90, to 1


participate in the research, consisting of cognitively normal older individuals, people with early or late MCI, and people with early AD. The follow up duration of each group is specified in the protocols for ADNI-1, ADNI-2 and ADNI-GO. Subjects originally recruited for ADNI-1 and ADNI-GO had the option to be followed in ADNI-2. For up-to-date information, see www.adni-info.org. PARTICIPANTS According to our aim, we selected only non-demented subjects with cognitive symptoms. This included participants with early (47%) and late (38%) MCI from the MCI cohort and participants from the healthy control cohort who had significant memory concerns (SMC; 15%). Inclusion/exclusion criteria are described in detail at www.adni-info.org. Briefly, all subjects in the present study were between the ages of 55 and 91 years, had completed at least 6 years of education, were fluent in Spanish or English, and were free of any significant neurologic disease other than AD. Subjects with SMC had Mini Mental State Examination score (MMSE) ≥24[2] and Clinical Dementia Rating (CDR) score 0[3], but expressed concerns of memory impairment. Subjects classified as MCI had MMSE ≥24, objective memory loss as shown on scores on delayed recall on the Wechsler Memory Scale Logical Memory II, CDR 0.5, preserved activities of daily living, and absence of dementia. Early and late MCI was differentiated based on the score of Wechsler Memory Scale Logical Memory II (cutoffs ranging from 2 to 8 depending on education level). ADNI DATA Data was downloaded on Oct 28 2016 from the ADNI database http://ida.loni.ucla.edu. Plasma NfL and tau were not significant in the BioFINDER models. NfL and tau data from ADNI were thus not included since this would have reduced the ADNI population further due to missing data.

RESULTS CHARACTERISTICS OF THE TRAINING COHORT (BIOFINDER) The characteristics of BioFINDER are shown in Table 1. Around 50% of the patients were Aβ+. Compared to the Aβ- group, the Aβ+ group were older, had more errors on the 10-word list delayed recall test, scored poorer on the MMSE (including the orientation and memory subscore), had higher prevalence of the APOE ε4 allele and lower prevalence of the ε2 and ε3 alleles. In addition, the Aβ+ group had a lower plasma Aβ42/40 ratio and increased levels of plasma NfL. CHARACTERISTICS OF THE VALIDATION COHORT (ADNI)

2


The most prominent difference compared to BioFINDER was years of education (on average 4.5 more years in ADNI). In the total ADNI population, MMSE and age was slightly higher. Significantly lower plasma Aβ42/40 ratio was seen in ADNI, probably reflecting differences in assays rather than actual differences in plasma concentrations (Table 1).

MODELS DEVELOPED SEPERATELY IN SCD AND MCI POPULATIONS In situations where the specific diagnostic entity is known, i.e. SCD or MCI, it can be favourable to use models specifically developed for that population, in contrast to the previous models that were developed in a population consisting of non-demented patients with cognitive symptoms (SCD and MCI). When running the LASSO analysis in SCD participants in BioFINDER, gender and education were also selected in addition to the previously selected variables (MMSE/delayed recall, APOE and age). The AUC from the logistic regression analysis was 0.82 (95% CI 0.75-0.88) for the delayed recall model and 0.81 (95% CI 0.74-0.87) for the MMSE model (Supplementary Table 3). In MCI participants, only 10-word delayed recall and APOE were chosen in the delayed recall model (AUC 0.83, 95% CI 0.78-0.88). In the MMSE model, no difference in selected variables was seen compared to the previous analysis (i.e. MMSE orientation and memory, APOE, and age; AUC 0.82, 95% CI 0.76-0.87) (Supplementary Table 3). The specific estimates of the models are shown in Supplementary Fig. 3. The BioFINDER SCD models were not tested in ADNI, since ADNI lacks a clearly defined SCD population. The MCI models, on the other hand, were replicated with similar accuracies. In MCI participants in the plasma dataset (n=170), the delayed recall model had an AUC of 0.82 (95% CI 0.75-0.88) and the MMSE model an AUC of 0.81 (95% CI 0.74-0.87). The corresponding AUCs in the total MCI population in ADNI (n=560) were 0.79 (95% CI 0.75.0.82) and 0.80 (95% CI 0.76-0.83), respectively.

3


SUPPLEMENTARY TABLES Supplementary Table 1. Prediction Of Amyloid Positivity In BioFINDER Cumulative AUC (95% CI) (AIC)

Univariate AUC (95% CI) Delayed recall model

Predictor variable

Age

0.61 (0.55-0.66)

MMSE orientation and memory

0.63 (0.58–0.68)

MMSE model

Delayed recall MMSE model model + plasma + plasma Aβ42/40 Aβ42/40

0.61 (0.55-0.66) 0.61 (0.55-0.66) 0.61 (0.55-0.66) 0.61 (0.55-0.66) (530) (530) (530) (530) 0.67 (0.62– 0.73) (509

0.67 (0.62–0.73) (509)

10-word list 0.71 (0.65-0.76) delayed recall (0-10 0.71 (0.66-0.76) (487) errors)

0.71 (0.65-0.76) (487)

APOE ε2ε2/ε2ε3

0.56 (0.53–0.58)

0.74 (0.69-0.79) 0.71 (0.66–0.76) 0.74 (0.69-0.79) (472) (493) (472)

0.71 (0.66– 0.76) (493)

APOE ε2ε4/ε3ε4

0.64 (0.59-0.69)

0.78 (0.73-0.82) 0.74 (0.69–0.79) 0.78 (0.73-0.82) (452) (472) (452))

0.74 (0.69– 0.79) (472)

APOE ε4ε4

0.58 (0.55-0.61)

0.83 (0.79-0.87) 0.81 (0.77-0.85) 0.83 (0.79-0.87) 0.81 (0.77-0.85) (409) (426) (409) (426)

Plasma Aβ42/40

0.74 (0.69-0.79)

0.85 (0.81-0.89) 0.83 (0.79-0.87) (392) (410)

The table corresponds to the AUCs shown in Fig. 1. Logistic regression analysis was performed using the variables selected from the LASSO regression (variables with estimates not shrunken to zero). All analyses are on the same 391 patients. Data are shown as AUC (95% CI of AUC, where >0.5 in the lower bound indicates that the model significantly predicts Ab positivity). Below (in parenthesis) is the AIC. Lower AIC equals better model fit in relation to the complexity (number of variables) where a decreased AIC of ≥2 indicates a significantly better model. The univariate AUC shows the accuracy using single variable prediction of Ab positivity. The cumulative AUC shows the increase in AUC when adding the present variables to those above. The AUC of the final models are thus shown at the bottom (in bold). List of predictor variables in the LASSO analysis: 10-word list delayed recall (from ADAS-cog), MMSE total score, MMSE orientation and memory, AQT (including ratios with the other cognitive measures), white-matter lesions (ARWMC scale), and all APOE groups. In the reduced set of variables (for the MMSE model), white-matter lesions, delayed recall, and AQT were excluded. In the secondary analyses, plasma NfL, plasma Aβ42/40 ratio, and plasma tau were added to the two sets of variables.

4


Supplementary Table 2. Replication Of The Amyloid Risk Models Populations

Delayed recall model

MMSE model

Delayed recall model + plasma Aβ42/40

MMSE model + plasma Aβ42/40

BioFINDER (n=391)

0.83 (0.79-0.87)

0.81 (0.77-0.85)

0.85 (0.81-0.89)

0.83 (0.79-0.87)

SCD (n=178)

0.79 (0.72-0.86)

0.78 (0.71-0.85)

0.82 (0.75-0.88)

0.81 (0.75-0.88)

MCI (n=213)

0.83 (0.77-0.88)

0.81 (0.75-0.87)

0.85 (0.79-90)

0.83 (0.77-0.88)

0.85 (0.80-91)

0.84 (0.79-0.90)

0.87 (0.82-0.90)

0.86 (0.81-0.91)

0.81 (0.74-0.0.87)

0.76 (0.70-0.83)

0.81 (0.75-0.87)

0.80 (0.74-0.86)

0.82 (0.75-0.89)

0.81 (0.75-0.88)

0.83 (0.77-0.89)

0.83 (0.76-0.89)

0.83 (0.75-92)

0.82 (0.73-91)

0.84 (0.76-92)

0.83 (0.75-92)

0.0.78 (60.8-0.80.8)

0.82 (0.72-93)

0.82 (0.72-92)

0.83 (0.73-93)

Age