Study design
This prospective, multi-institutional study was conducted at 11 facilities (Supplementary Table 1, Additional file 1). The study (Study on Diagnosis of Alzheimer’s disease with FDG-PET: SDAF-PET) was first reviewed and approved by the Institutional Review Board of the National Center for Geriatrics and Gerontology, and registered with UMIN (UMIN 000016427) on February 3, 2015. Subsequently, with the enactment of the Clinical Research Act in Japan, it was again reviewed and approved by the Nagoya University Clinical Research Review Board and registered with JRCT (jRCTs041180098) on March 14, 2019. The study protocol was prepared in agreement with the PMDA in Japan and approved by Japan's Ministry of Health, Labour and Welfare under the advanced healthcare services system.
Participants whose written consent was obtained and who were clinically diagnosed with AD or FTLD by clinical, neuropsychological, and MRI examinations were enrolled in the study. FDG-PET and CSF examinations were performed within 4 weeks of enrollment. The target number of patients at enrollment was 190 (AD group, 150; FTLD group, 40).
In addition, 12 months after enrollment, neuropsychological and MRI examinations were performed to clinically diagnose AD and FTLD; this diagnosis was used as the gold standard (final clinical diagnosis).
The primary endpoint of this study was the difference in sensitivity between FDG-PET and p-tau181 in CSF for differential diagnosis of AD and FTLD. The sensitivity was calculated by comparing the results of the image interpretation and the results of p-tau181 in CSF with the gold standard.
Secondary endpoints were as follows: 1) difference in accuracy of FDG-PET and p-tau181 in CSF for the differential diagnosis between AD and FTLD; 2) diagnostic performance of FDG-PET based on clinical diagnosis reflecting CSF Aß42; 3) region of interest (ROI)-based automated analysis for 18F-FDG PET; and 4) group analysis of CSF biomarkers and MRI between AD and FTLD.
The integrity of the data and compliance with the procedures were verified by third-party monitoring at each facility. Subsequently, an overall confirmation was made by a third-party audit.
Participating subjects
The inclusion criteria for the study were as follows: 1) patients with AD or FTLD whose native language is Japanese; 2) participants must sign a written consent form. If the subject is deemed incapable of consenting, a substitute must give his/her consent and sign on his/her behalf; 3) participants must be accompanied by a research assistant who is able to assess and evaluate the subject's situation; 4) they must be between 55 and 84 years of age (at the time of obtaining consent); and 5) they must be able to undergo a PET scan.
The criteria for AD were as follows: 1) Mini Mental State Examination (MMSE) score between 20 and 26; 2) objective evidence of memory impairment, taking into account age and education; 3) Wechsler Memory Scale-Revised (WMS-R) Logical Memory II subscale (Delayed Replay Task of Logical Memory: maximum score 25) must be below the cutoff value by years of education; 4) Clinical Dementia Rating (CDR) of 0.5 or 1; and 5) the patient must meet the NINCDS-ADRDA Criteria for Probable AD (diseases and causes of dementia other than AD must be excluded) [8].
The criteria for FTLD were as follows: 1) MMSE score between 20 and 26 points; 2) CDR of 0.5 or 1; and 3) FTLD clinical diagnostic criteria must be met (diseases and causes of dementia other than FTLD must be excluded). However, FDG-PET is excluded from the clinical diagnostic criteria for FTLD. Details of the FTLD clinical diagnostic criteria were based on the diagnostic criteria for each type [9].
The exclusion criteria for this study were as follows: 1) patients with a history of or undergoing treatment for alcoholism; 2) patients with a history of or undergoing treatment for epilepsy; 3) patients with less than 6 years of education; 4) patients with diabetes and on insulin therapy; 5) patients receiving antidepressants, antipsychotics, or long-term hypnotic sedatives (including anxiolytics); 6) patients who have been diagnosed with major depressive or bipolar disorder within the past year, patients with a history of schizophrenia, or patients with severe psychiatric symptoms such as anxiety or agitation within the last 3 months who are unlikely to be able to complete the protocol; 7) patients with serious complications (malignancy, heart failure, hepatic disorder, renal disorder, endocrine disorder, and so on); and 8) patients whose MRI reveals cerebral infarction or other focal lesions that may affect cognitive function.
Determining the number of cases
With reference to previous studies [10,11], assuming "sensitivity of FDG-PET: 90%, specificity: 70%; sensitivity of p-tau: 75%, specificity: 70%" and α (type 1 error) and β (type 2 error) as 0.05 and 0.2, respectively, the number of AD cases needed to assess sensitivity, the primary endpoint, is 97 cases. Assuming a dropout rate of approximately 20%, the number of cases required would be approximately 125.
Based on the sensitivity and specificity of FDG-PET and p-tau, and the accuracy calculated by assuming an 80%:20% prevalence ratio of AD and FTLD, the number of cases required to assess the accuracy, a secondary endpoint, is 170 cases. The breakdown of the required number of cases is 136 AD cases and 34 FTLD cases.
In addition to the primary endpoint of sensitivity, we set the number of AD and FTLD cases to 150 and 40, respectively, to allow for a dropout rate of approximately 10% for the secondary endpoint of accuracy.
Neuropsychological test battery
Patients were, initially and 12 months after enrollment, assessed using a neuropsychological test battery, including the MMSE, Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS), CDR, Frontal Assessment Battery, and WMS-R LM.
FDG-PET
Prior to baseline FDG-PET scanning, all participants fasted for at least 4 hrs. Intravenous administration of 18F-FDG (185 ± 37 MBq) was followed by a resting period of 30 min in a dimly lit, quiet room. A 3D dynamic scan with six 300-second frames was performed 30 minutes after 18F-FDG injection. Attenuation was corrected using a CT scan for PET/CT. Supplementary Table 2 (Additional file 2) lists the PET/CT devices and reconstruction conditions.
The FDG-PET images were processed using the 3-dimensional stereotactic surface projection (3D-SSP) technique to generate z-score maps, using iSSP software version 3.5 (Nihon Medi-Physics Co. Ltd., Tokyo, Japan). The normal database used for generating the z-score maps was constructed based on 50 normal subjects (20 males and 30 females, average age = 67.8 y) from five participating institutions. The healthy subjects in the normal database had no memory complaints and no history of neurological or psychiatric disorders. The results of their neurological and brain imaging examinations (MR or CT) were normal, and their cognitive function was judged to be normal by experienced neurologists (MMSE score, 26–30).
PET image interpretation
Three nuclear medicine experts, blinded to the clinical information, independently evaluated the reconstructed PET images with reference to the 3D-SSP z-score map (tomogram + 3D-SSP) to classify the images into different dementia patterns of AD and FTLD according to the imaging diagnostic criteria of a previous study [10].
This interpretation method attributes frontal and anterior temporal hypometabolism to FTLD and parietal, posterior cingulate, and lateral temporal hypometabolism to AD. It guides the interpreter to imagine a fulcrum in the middle of the brain. If the visible "weight" of the hypometabolism is predominantly anterior, then the scan is classified as showing FTLD, whereas if it is predominantly posterior, then the scan is classified as showing AD. Examples of FDG PET scan patterns are shown in Figure 1. When the evaluation of the three raters was not in complete agreement, the evaluation agreed upon by the two raters was considered the final evaluation, and in cases where the evaluation of all three raters differed, the final evaluation was "undecided.
CSF examination
CSF was collected once within 4 weeks of case enrollment during fasting. Total tau, p-tau181 (ELISA assay by Innogenetics), and Aβ42 (ELISA assay by WAKO) were measured centrally at Hirosaki University. The reference values for positive and negative tau, p-tau181, and Aß42 in CSF were 297, 61, and 687 pg/mL, respectively, based on the reference values of the central cerebrospinal fluid assay laboratory.
ZSAM
We investigated the diagnostic performance of FDG-PET using ZSAM [12, 13] in addition to visual reading.
ZSAM is an ROI-based automated analysis method for FDG PET and includes the following procedures:
1. The FDG-PET images of normal individuals were analyzed using 3D-SSP to generate Z-score images. Summed positive Z-scores (Z-sum) were obtained from template ROIs (parietal lobe, posterior cingulate gyrus, precuneus, medial surface of the occipital lobe, and lateral surface of the occipital lobe), which are regions of characteristic glucose metabolism reduction in the Z-score images of AD and dementia with Lewy bodies; DLB. This is because each ROI includes negative and positive Z-scores, and reduced glucose metabolism appears as a positive Z-score.
2. The Z-sum of some normal individuals was averaged for each ROI, and the normal threshold of the Z-sum with the mean value and standard deviation was calculated.
3. The Z-sum of patients processed by the same procedure as #1 is compared with the normal threshold calculated above in #2.
4. Glucose metabolism is considered to be reduced if the Z-sum of the patient exceeds the normal threshold.
In this study, when the Z value exceeded the threshold in any one of the four ROIs (bilateral parietal lobe, and bilateral posterior cingulate gyrus/precuneus), the patient was diagnosed with AD.
MRI
All participants were scanned using a 3T or 1.5T MRI system. A T1-weighted fast-field echo sequence was used. Supplementary Table 3 (Additional file 3) lists the MRI devices and imaging parameters. T1-weighted 3-dimensional sagittal sections of the brains were acquired and analyzed on a PC using a voxel-based specific regional analysis system for Alzheimer’s disease (VSRAD® advance, Eisai Co., Ltd., Tokyo, Japan), which was developed based on the voxel-based morphometry method and is now freely available [14–18]. First, the equalization of voxel sizes and linear and nonlinear transformations were performed. Next, images of the gray matter, white matter, and cerebrospinal fluid were separated, and the gray matter images were standardized and smoothed onto templates using the DARTEL software (Wellcome Department of Imaging Neuroscience, London, UK). Using the z-score method, a comparative statistical analysis of the voxels was performed using a healthy control database. The database for the healthy controls contained data from 40 men and 40 women, each aged 54–86 (mean 70.2 ± 7.3) years. In this study, the average positive z-score in the target volume of interest for the medial temporal structures, including the entorhinal cortex, head-to-tail of the hippocampus, and amygdala, was used for further analyses.
Statistical analysis
The sensitivities and accuracies of FDG-PET and p-tau181 in CSF were compared using a sign test. CSF biomarkers, MRIs, and neuropsychological tests were compared between the AD and FTLD groups. The mean value was used as the evaluation index for the items on the measurement scale, and a two-sample t-test was used to compare the groups. As a rule, the significance level of the test was 5% two-sided when using the normal, t, hypergeometric, and binomial distributions, and 5% one-sided when using the χ2 distribution and F distribution. In principle, the confidence coefficient for the estimation was two-sided at 95%.
Statistical analysis software, SPSS version 22 or higher and R version 3.4.4 or higher, were used for analysis.