Study	Reason for exclusion
Altomare 2016	MCI diagnosis at baseline was not made with any of our accepted definitions by protocol for MCI participants. Dr Altomare kindly responded to some questions regarding the method of his study (mail received 16/06/2017).
Apostolova 2016	Not having data for constructing a 2 x 2 table. The study was focused on the development of neuropsychiatric symptoms and not on Alzheimer's disease or dementia progression.
Brendel 2014	Not having data for constructing a 2 x 2 table. The study was focused on longitudinal quantitative analyses of 18F‐florbetapir PET and their association with progression of dementia.
Brendel 2015	Not having data for constructing a 2 x 2 table. The study was focused on testing the effects of different reference regions and atrophy‐based partial volume effects on the discriminatory power and longitudinal performance of amyloid PET.
Cheewakriengkrai 2014	Not having data for constructing a 2 x 2 table. The study was focused on the relationship between regional distributions of brain fibrillar amyloid deposition, neurodegenerative biomarkers in CSF (CSF Aβ 1‐42, t‐tau, p‐tau) and cognitive function (ADAS‐cog) at 24 months follow‐up.
Chen 2015a	Not having data for constructing a 2 x 2 table. The study compared the power of template‐based cerebellar, pontine, and cerebral white matter reference regions to track 24‐month florbetapir standardized uptake value (SUV) ratio (SUVR) changes; and to relate those changes to 24‐month clinical declines
Chen 2015b	Not having data for constructing a 2 x 2 table. The study was focused in the diagnostic potential of FDG PET, florbetapir, PiB and CSF biomarkers in monitoring the progression from mild cognitive impairment (MCI) to Alzheimer’s disease (ADD) and cognitively normal (NC) to MCI in a longitudinal study
Chincarini 2015	Not having data for constructing a 2 x 2 table. The study was focused on examining different approaches to amyloid‐PET quantification and a longitudinal analyses of Aβ deposition.
Chincarini 2016	The study focused on the evaluation of brain amyloidosis (ELBA) with a new method on imaging of the 18F‐florbetapir PET scan. We did not include this study because we preferred to include the Schreiber study for the following reasons: There was a high risk of duplication of participants with the Schreiber study, due to both studies using the same ADNI database. The Schreiber study had more participants: 401 MCI participants compared to 62 in the Chincarini study. The reason why there were no participants with MCI at baseline who maintained their condition at the follow‐up was not clear.
Durkanova 2015	Not having data for constructing a 2 x 2 table. The study was focused in evaluate five different test strategies for integrating use of florbetapir and FDG PET information to predict rates of cognitive and functional decline over 2 years.
Fan 2015	Not having data for constructing a 2 x 2 table. The study was focused on investigating whether different translocator protein genotypes influenced cognitive function, amyloid load, and disease progression over time.
Greenia 2014	Not having data for constructing a 2 x 2 table. The study was focused on evaluating the 18F‐florbetapir PET and the relationship with cognitive decline in the oldest‐old.
Hochstetler 2014	Not having data for constructing a 2 x 2 table. The study was focused on trying to define trajectories of cognitive and functional decline, and characteristics associated with distinct trajectories, using Growth Mixture Modeling.
Joshi 2014	Not having data for constructing a 2 x 2 table. The study was focused on the estimation of longitudinal change in Aβ burden over 2 years.
Klein 2015	Not having data for constructing a 2 x 2 table. The study was focused on the evaluation of native space compared to SPM template methods and a variety of possible SUVR reference regions with highest longitudinal change in the SUVR at 24 months.
Landau 2014	Not having data for constructing a 2 x 2 table. The study was focused on the 18F‐florbetapir PET longitudinal evaluation in cognitively normal, MCI, and ADD participants, examining characteristics of normal individuals with subthreshold florbetapir retention and the influence of reference region selection on estimated trajectories across the entire range of amyloid measurements.
Landau 2016	This study was focused on comparing participants with amyloid beta negative MCI and participants with ADD enrolled in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) with their Aβ amyloid positive counterparts on a number of clinical, neuropsychological, and biomarker characteristics with an average available follow‐up time for longitudinal cognitive measurements of 1.4 + 0.8 years. The conversion rate in those MCI participants with PET negative was 11% and the conversion in those with PET positive was 45%. We did not include this study, and we preferred Schreiber 2015 to be included for the following reasons: There was a high risk of duplication of participants with the Schreiber study, due to the use of the same ADNI database and Landau was the second author of the Schreiber study. The Schreiber study had more participants: 401 MCI participants compared to 217 in the Landau study. The follow‐up was longer in the Schreiber study: 1.6 + 0.7 years and in the Landau study it was 1.4 + 0.8 years.
Lee 2015	Not having data for constructing a 2 x 2 table. This study was focused in the correlation between florbetapir and FDG PET and cognition measured by MMSE at follow‐up.
Lim 2014	Not having data for constructing a 2 x 2 table. This study was focused on evaluating the florbetapir status at baseline and different cognitive composite measures at 36 months.
Manitsirikul 2015	Not having data for constructing a 2 x 2 table. The study was focused on the relationship between regional distributions of brain fibrillar amyloid deposition, neurodegenerative biomarkers in brain (FDG) and CSF (tau), brain structural change, and cognitive function at 24‐month follow‐up.
Margolin 2013	Not having data for constructing a 2 x 2 table. The study was focused on evaluating the 18F‐florbetapir PET and the relationship with cognitive decline at follow‐up.
Mathotaarachchi 2015	Not having data for constructing a 2 x 2 table. The study was focused on the regional effects of amyloid retention measured by the 18F‐florbetapir PET scan on the rate of hypometabolism measured by FDG PET scan over the follow‐up.
Mattsson 2014a	This study was focused on comparing the diagnostic test accuracy with CSF Aβ42 and the 18F‐florbetapir PET scan in three different groups, healthy controls, Alzheimer's disease dementia, and MCI (progressive vs stable MCI) participants. We did not include this study, as we preferred Schreiber 2015 to be included for the following reasons: There was a high risk of duplication of participants with the Schreiber study, due to both studies using the same ADNI database and two authors from the Schreiber study (Landau and Jagust) also worked in the Mattsson study. The Schreiber study had more participants, 401 MCI participants compared to 224 in the Mattson study. The follow‐up was similar: Schreiber study: 1.6 + 0.7 years; Mattsson study: in those with stable MCI, the follow‐up was 2.2 + 0.3 years and in those with progressive MCI, the follow‐up was 1.7 + 0.6 years.
Mattsson 2014b	Not having data for constructing a 2 x 2 table. This study was focused in determine the extent to which CSF and 18F‐florbetapir PET contribute independent diagnostic information in AD studies, and to determine the nature and degree of pathology in discordantly classified individuals in healthy controls, ADD patients, and MCI participants.
Mattsson 2015a	Not having data for constructing a 2 x 2 table. The study was focused on testing if CSF and amyloid beta PET scan biomarkers were independently related to other Alzheimer's disease markers, and to examine individuals who were discordantly classified by these two biomarker modalities with a follow‐up for up to three years.
Mattsson 2015b	Not having data for constructing a 2 x 2 table. The study was focused on relationships in a large number of brain regions in MCI participants with cognitive evaluations for up to three years with Logical Memory delayed recall and Rey Auditory Verbal Learning Test delayed recall.
Ming 2015	Not having data for constructing a 2 x 2 table. The study was focused on MCI participants and 18F‐florbetapir at baseline and follow‐up for up to three years with cognitive evaluations with MMSE, ADAS11 and CDR sum of boxes.
Mohades 2014	Not having data for constructing a 2 x 2 table. The study was focused on comparing neurodegeneration in 18F‐florbetapir accumulators and nonaccumulators based on a 24‐month assessment.
Morbelli 2015	Not having data for constructing a 2 x 2 table. The study was focused on MCI participants that had longitudinal evaluation with the 18F‐florbetapir PET scan over two years and different methods to establish the PET positivity.
Pascoal 2016	Not having data for constructing a 2 x 2 table. The study was focused on neuropsychological and clinical decline in participants with MCI and if they were associated with brain amyloid‐beta deposition and tau hyperphosphorylation.
Pascoal 2017	The study was focused on amnestic MCI individuals and whether the synergism between Aβ aggregation and tau hyperphosphorylation could determine the progression from amnestic MCI to ADD dementia. We did not include this study because we preferred the Schreiber study to be included for the following reasons: They used the same ADNI database and 279 of 314 MCI participants in Pascoal 2017 were also included in Schreiber 2015. The Schreiber study had more participants: 401 MCI participants compared to 314 in the Pascoal study. Dr Pascoal kindly responded to some questions regarding the method of his study and provided the ADNI identification code of the participants (mail received 16/06/2017).
Pontecorvo 2011	Not having data for constructing a 2 x 2 table. The study was focused on the evaluation of the correlation of florbetapir SUVR with cognitive change from baseline to month 24 in MCI and cognitively normal participants, PET PiB, and CSF amyloid and tau levels.
Risacher 2014	Not having data for constructing a 2 x 2 table. The study was focused on the comparative assessment of two‐year change in amyloid deposition, glucose metabolism, and hippocampal atrophy in healthy controls, MCI and ADD participants.
Shokouhi 2016	Not having data for constructing a 2 x 2 table. The study was focused on evaluating the effect of reference tissue normalization in a test–retest 18F‐florbetapir SUVR study using different reference regions and evaluating the correlation between 18F‐florbetapir PET and concurrent CSF Aβ1–42 levels in a MCI cohort over the course of 2 years.
Siderowf 2013	Not having data for constructing a 2 x 2 table. The study was focused on evaluating cognitive decline measured by ADAS‐cog in participants with negative and positive 18F‐florbetapir PET scan imaging with a clinical follow‐up of 18 months.
Teipel 2015	Not having data for constructing a 2 x 2 table. The study was focused on comparing penalized regression analysis, with more classical unregularised regression models in respect to predicting conversion from MCI to ADD in 127 MCI subjects who had a clinical follow‐up between 6 and 31 months.
Toledo 2015	Not having data for constructing a 2 x 2 table. The study was focused on determining the association between CSF and PET amyloid biomarkers (cross‐sectional and longitudinal measures) and comparing the cut‐offs for these measures.
Wisse 2015	Not having data for constructing a 2 x 2 table. The study was focused on characterising MCI participants separated into four groups according to their abnormal amyloid‐beta 42 levels and abnormal hippocampal volume or hypometabolism using fluorodeoxyglucose PET and the conversion rate at 24 months.
Xu 2016	The study was focused on exploring the contribution of different neuroimaging modalities in their predictive power and characterised the sensitive biomarkers from each modality. We did not include this study, as we preferred the Schreiber study to be included for the following reasons: They used the same ADNI database and 70 of 110 MCI participants in Xu 2016 were also included in Schreiber 2015. Schreiber had more participants: 401 MCI participants compared to 110 in the Xu study.

^{Aβ: Amyloid Beta
 ADAS11: Alzheimer's disease assessment scale‐11
 ADAScog: Alzheimer's Disease Assessment Scale‐Cognitive subscale
 ADD: Alzheimer's disease dementia
 ADNI: Alzheimer's Disease Neuroimaging Initiative
 CDR: Clinical dementia rating
 CSF: Cerebrospinal fluid
 ELBA: Evaluation of brain amyloidosis
 FDG: Fluorodeoxyglucose
 MCI: Mild cognitive impairment
 MMSE: Mini‐mental state examination
 PET: Positron emission tomography
 PiB: Pittsburgh compound B
 SPM: statistical parametric mapping
 SUV: Standardised uptake value
 SUVR: Standardised uptake value ratio}