Table 2 Systematic review of 23 studies deriving tau PET cut-points
From: What’s the cut-point?: a systematic investigation of tau PET thresholding methods
Study | Sample | Reference region | PVC | ROI(s) | Analytic approach | Cut-point |
|---|---|---|---|---|---|---|
[19] Botha et al. (2018), Brain | 24 CI OAs from MCSA and ADRC | Cerebellar crus gray | Not reported | Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) | ROC maximizing sensitivity/specificity between YAs and A+ CI OAs | 1.33 |
[20] Cho et al. (2018), Neurobiol Aging | 220 OAs (all dx) from memory clinic | Cerebellar cortex | Yes using region-based voxelwise method (and no) | 25 individual ROIs | Mean + 2.5 SDs above CU with minimal cortical Abeta (<1.4) and EC tau (<1.2) | Not reported |
[21] Dodich et al. (2020), Eur J Nucl Med | 81 CI OAs from memory clinic | Cerebellar crus | Not reported | Meta-ROI (medial temporal lobe, lateral occipital, ITG) | ROC between A−/N− vs other | 1.24 |
[22] Guo et al. (2020), Alz Res Ther | 341 OAs (all dx) from ADNI | Inferior cerebellar cortex | Yes using GTM | (1) Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) (2) EC ROI | (A) ROC with Youden index between A− CU and A+ C (B) Mean +2 SDs above A− CU | (1A) 1.25 (2A) 1.21 (1B) 1.34 (2B) 1.31 |
[23] Jack et al. (2017), Alz Dem | 49 YAs and 153 OAs (all dx) from MCSA and ADRC | Cerebellar crus gray median | “Most likely CSF” voxels were removed | Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) | (1) Maximize specificity (95%) based on YAs (2) Maximize sensitivity (10%) based on A+ CI OAs (3) ROC maximizing accuracy between YAs and A+ CI OAs (4) ROC maximizing accuracy between A− CU OAs and A+ CI OAs | (1) 1.19 (2) 1.21 (3) 1.21 (4) 1.32 |
[24] Jack et al. (2017), Lancet Neuro | 435 OAs from MCSA (YAs not reported) | Cerebellar crus gray median | No | Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) | ROC maximizing accuracy between YAs and A+ CI OAs | 1.23 |
[25] Jack et al. (2019), JAMA | 480 OAs without dementia from MCSA (YAs not reported) | Cerebellar crus gray | No (and yes using two compartment method) | Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) | (1) ROC maximizing accuracy between YAs and A+ CI OAs (2) ROC maximizing accuracy between A− CU and A+ CI OAs | (1) 1.25 (2) 1.33 |
[26] Jang et al. (2019), Eur J Nucl Med imaging | 60 OAs with SVCI from medical center and 49 OAs (all dx) from hospital | Cerebellar gray | Yes | (1) Braak V/VI (2) Braak III/IV (3) Braak I/II | CIDT (criterion variable unspecified); T+ considered III/IV or above | (1) 1.58 (2) 1.33 (3) 1.29 |
[27] Lowe et al. (2018), Alz Dem: DADM | 112 YAs from MCSA and 576 OAs from MCSA and ADRC | Cerebellar Crus | No | Temporal meta-ROI (EC, parahippocampal, hippocampus); 26 individual ROIs also assessed | 95% above A− YAs | Not reported |
[28] Lowe et al. (2018), Brain | 98 YAs, 601 A− CU OAs, 86 A+ CI OAs from MCSA and ADRC | Cerebellar crus gray | No (and yes using two compartment method) | 47 individual ROIs | 95% above A− YAs for each ROI | Not reported |
[29] Lowe et al. (2019), Neurology | 117 YAs; 579 CU OAs from MCSA | Cerebellar crus | No (and yes using GTM) | 43 individual ROIs | 95% percent above A− YAs | 1.17 in EC |
[30] Lowe et al. (2020), Alz Dem | 26 OAs from MCSA and ADRC | Cerebellar Crus | No (and yes using GTM) | (1) Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) (2) EC ROI | ROC with Youden index between AD and non-AD spectrum pathology | * (1) 1.29 (2) 1.27 |
[14] Maass et al. (2017), Neuroimage | Sample 1: 12 YAs, 74 CU OAs from BACS, and 61 CI OAs from ADRC Sample 2: 42 CU and 28 CI OAs from ADNI | Inferior cerebellar gay | (1) Yes using GTM (2) No | (A) Temporal meta-ROI (B) ITG (Also examined whole brain, AD-vulnerable mask, and Braak stage composite regions) | ROC with Youden index between A− CU OAs and A+ CI OAs) (Also used CIDT with age/diagnostic group as input variable for Braak composite staging) | (1A) 1.47 (BACS/UCSF) and 1.34 (ADNI) (1B) 1.27 (BACS/UCSF) and 1.20 (ADNI) (2A) 1.46 (BACS/UCSF) and 1.40 (ADNI) (2B) 1.30 (BACS/UCSF) and 1.23 (ADNI) (other cut-points omitted from table but included in results section) |
[31] Mattsson-Carlgren et al. (2020), Sci Adv | 131 OAs (all dx) from BioFINDER | Inferior cerebellar gray | No (and yes using GTM in EC) | (1) EC (2) ITG (3) Braak V/VI | Mean +2 SDs above independent A− CU group | (1) 1.39 (2) 1.31 (3) 1.28 |
[32] Meyer et al. (2020), JAMA Neuro | 322 OAs (all dx) from ADNI | Inferior cerebellar gray | No (and yes using GTM) | Temporal meta-ROI (amygdala, EC, fusiform, parahippocampal, ITG, MTG) | ROC with maximum percentage correct classification between A− CU OAs and A+ CI OAs | * 1.37 |
[33] Mishra et al. (2017), Neuroimage | 97 OAs (all dx) from ADRC | (1) Whole cerebellum (2) Cerebellar cortex | Yes using RSF | Meta-ROI (EC, lateral occipital, ITG, amygdala) | SKM method to cluster into high and low tau groups; midpoint taken | (1) 1.25 (2) 1.22 |
[34] Ossenkoppele et al. (2018), JAMA | 719 OAs (all dx) from ADRC, BioFINDER, and memory clinic | Inferior cerebellar gray | No (and yes using GTM) | 5 ROIs examined: EC, ITG, temporal meta-ROI, temporoparietal cortex, Braak V/VI | (1) Mean +2 SDs above CU (2) ROC Youden index between controls and AD | * For temporal meta-ROI: (1) 1.34 (2) 1.27 (other cut-points omitted from table but included in results section) |
[35] Rafii et al. (2017), JAD | 9 adults with Down syndrome from DSBI | (1) Cerebellar cortex (bottom slice removed and edges eroded) (2) Subcortical WM | Yes using Muller-Gartner method | Braak I-VI average | Quantitative discrimination of A− vs A+ | (1) 1.2 (2) 1.05 |
[36] Schöll et al. (2016), Neuron | 5 YAs and 33 CU OAs from BACS; 15 CI OAs from ADRC | Cerebellar gray | Yes using GTM | (1) Braak V/VI (2) Braak III/IV (3) Braak I/II | CIDT using age/diagnostic group as input variable | (1) 2.79 (2) 1.73 (3) 1.40 |
[37] Schwarz et al. (2016), Brain | 14 YAs and 173 OAs (all dx) from a clinical study | Cerebellar crus | Not reported | 7 ROIs corresponding to Braak histological stages | Mean +2.5 SDs above YAs | Ranged from 1.22 (STG) to 1.36 (fusiform) |
[38] Schwarz et al. (2018), Alz Dem | 14 YA, 21 OAs (all dx) for test-retest, and 98 OAs (all dx) from ADNI | Cerebellar gray | Not reported | (1) Meta-ROI (hippocampus, TEC, fusiform, MTG, STG, extrastriate, striate) (2) Simplified meta-ROI (medial temporal, lateral temporal, STG, primary visual cortex) (3) Lobar (average temporal, parietal, and frontal lobes) (ROIs gray matter masked) | (1) Mean +2.5 SDs above T− YAs for each ROI within meta-ROI; T+ includes those above threshold in hippocampus, TEC, fusiform, MTG, and extrastriate (2) Mean +3 SDs above T− YAs to obtain same average as (1); T+ includes those above threshold in MTL and lateral temporal (3) Mean +3 SDs above T− YAs to obtain same average as (1); T+ includes those above threshold in temporal lobe | * 1.28 |
[39] Wang et al. (2016), JAMA Neuro | 59 OAs (all dx) from ADRC | Cerebellar cortex | No (and yes using linear regression) | (1) Hippocampus (2) Meta-ROI (medial temporal, ITG, lateral temporal, inferior parietal, PCC, precuneus, SPL) | (A) ROC with Youden index between A− CU and A+ AD (B) ROC with Youden index between A+ CU and A+ AD | * (1A) 1.36 (2A) 1.19 (1B) 1.36 (2B) 1.33 |
[40] Weigand et al. (2020), Brain Comms | 523 OAs (all dx) from ADNI | Inferior cerebellar gray | Yes using GTM | (1) Braak V/VI (2) Braak III/IV (3) Braak I/II | CIDT with MMSE as input variable | (1) 1.96 (2) 1.51 (3) 1.18 |