Meta Analysis - Cognition, Amyloid, Tau

1 Libraries

library(meta)
library(metasens)
library(readxl)
library(metafor)
library(DescTools)
library(psych)
library(tidyverse)
library(metaviz)
library(table1)
library(stargazer)
remotes::install_github("MathiasHarrer/dmetar")

2 Data

#download data and subset it
data_full <- read_xlsx("/Users/carolinaferreiraatuesta/Documents/H-LAND/Meta Analysis/data_feb.xlsx", sheet= "representatives_global_memory")
data <- subset(data_full, table1 ==1 & exclude !=1) 

3 Tables 1

data_table1 <- subset(data_full, table1 == 1 & exclude != 1)
#labels
label(data_table1$memory_composite) <- "Memory"
label(data_table1$global_composite) <- 'Global'
label(data_table1$author) <- 'Author'
label(data_table1$year) <- 'Year '
label(data_table1$amyloid) <- 'Amyloid'
label(data_table1$tau) <- 'Tau'
label(data_table1$interaction) <- 'Interaction'
label(data_table1$addition) <- 'Addition'
label(data_table1$selected_tau_amyloid) <- 'Selected ligand'
label(data_table1$ligand) <- 'Ligand'
label(data_table1$ligand_other) <- 'Ligand other'
label(data_table1$csf) <- 'CSF'
label(data_table1$csf_other) <- 'CSF other'
label(data_table1$serum) <- 'Serum'
label(data_table1$type_study) <- 'Type study'
label(data_table1$neurodeg) <- 'Neurodegeneration'
label(data_table1$method) <- 'Method'
label(data_table1$controlled) <- 'Controlled'
label(data_table1$cov_time) <- 'Time'
label(data_table1$sex) <- 'Sex'
label(data_table1$education) <- 'Education'
label(data_table1$apoe) <- 'Apoe'
label(data_table1$covaratiate_others) <- 'Other cov'
label(data_table1$cohort_database) <- 'Cohort'
label(data_table1$age_type) <- 'Age type'
#label(data_table1$age) <- 'Age'
#label(data_table1$age_variance) <- 'Age variance'
#abel(data_table1$total_nc) <- 'Total N'
#label(data_table1$nc_pos) <- 'N positive'
#label(data_table1$nc_neg) <- 'N negative'
label(data_table1$roi_lateral_temporal) <- 'ROI temporal lateral'
label(data_table1$roi_entorhinal_hippocampus) <-
'ROI entorhinal or hipocampus'
label(data_table1$roi_lateral_parietal) <- 'ROI parietal lateral'
label(data_table1$roi_cingulate) <- 'ROI cingulate'
label(data_table1$roi_frontal) <- 'ROI frontal'
label(data_table1$roi_precuneus) <- 'ROI precuneus'
label(data_table1$roi_occipital) <- 'ROI occipital'
label(data_table1$suvr_dvr) <- 'SUVR/DVR'
#label(data_table1$type_cognitive) <- ''
label(data_table1$cross_long) <- 'CS or Longitudinal'
label(data_table1$all_r) <- 'R value'
label(data_table1$exclude) <- ''
label(data_table1$fisherz) <- 'Fisher z'
label(data_table1$weight) <- 'Weighted (n-3)'
label(data_table1$weight_se) <- 'Weigted SE'

#as factors

data_table1$memory_composite <-
as.factor(data_table1$memory_composite)
data_table1$global_composite <-
as.factor(data_table1$global_composite)
data_table1$amyloid <- as.factor(data_table1$amyloid)
data_table1$tau <- as.factor(data_table1$tau)
data_table1$interaction <- as.factor(data_table1$interaction)
data_table1$addition <- as.factor(data_table1$addition)
data_table1$selected_tau_amyloid <-
as.factor(data_table1$selected_tau_amyloid)
data_table1$type_study <- as.factor(data_table1$type_study)
data_table1$neurodeg <- as.factor(data_table1$neurodeg)
data_table1$sex <- as.factor(data_table1$sex)
data_table1$education <- as.factor(data_table1$education)
data_table1$apoe <- as.factor(data_table1$apoe)
data_table1$age_type <- as.factor(data_table1$age_type)
data_table1$roi_lateral_temporal <-
as.factor(data_table1$roi_lateral_temporal)
data_table1$roi_entorhinal_hippocampus <-
as.factor(data_table1$roi_entorhinal_hippocampus)
data_table1$roi_lateral_parietal <-
as.factor(data_table1$roi_lateral_parietal)
data_table1$roi_cingulate <- as.factor(data_table1$roi_cingulate)
data_table1$roi_frontal <- as.factor(data_table1$roi_frontal)
data_table1$roi_precuneus <- as.factor(data_table1$roi_precuneus)
data_table1$roi_occipital <- as.factor(data_table1$roi_occipital)
data_table1$suvr_dvr <- as.factor(data_table1$suvr_dvr)
data_table1$type_cognitive <- as.factor(data_table1$type_cognitive)
data_table1$cross_long <- as.factor(data_table1$cross_long)

3.1 Table descriptive amyloid

3.2 Table descriptive tau

3.3 Table summary included papers

newtab <-
  data.frame(
  data_table1$author,
  data_table1$year,
  data_table1$cohort_database,
  data_table1$method,
  data_table1$cross_long,
  data_table1$cov_time,
  data_table1$controlled,
  data_table1$total_nc,
  data_table1$Global_composite,
  data_table1$Memory_composite
  )
  
  names(newtab)[names(newtab) == "data_table1.author"] <- "Author"
  names(newtab)[names(newtab) == "data_table1.method"] <- "Method"
  names(newtab)[names(newtab) == "data_table1.year"] <- "Year"
  names(newtab)[names(newtab) == "data_table1.cohort_database"] <- "Cohort"
  names(newtab)[names(newtab) == "data_table1.cross_long"] <- "Type of study"
  names(newtab)[names(newtab) == "data_table1.cov_time"] <- "Controlled by time of follow-up"
  names(newtab)[names(newtab) == "data_table1.controlled"] <- "Controlled"
  names(newtab)[names(newtab) == "data_table1.total_nc"] <- "N"
  names(newtab)[names(newtab) == "data_table1.Global_composite"] <- "Global"
  names(newtab)[names(newtab) == "data_table1.Memory_composite"] <- "Memory"
  
 
  kableExtra::kbl(newtab) %>%
  kableExtra::kable_paper(bootstrap_options = "striped", full_width = F)

Author	Year	Cohort	Method	Type of study	Controlled by time of follow-up	Controlled	N	Global	Memory
Gu	2012	WHICAP	Ab42	C	No	Yes	813	NA	0.0670742
Horn MM	2018	Dallas	PiB	C	No	Yes	195	NA	0.2500000
Hoscheidt SM	2016	IMPACT	ab42/ab40/ptau/ttau	C	No	Yes	70	NA	0.0964365
Adamczuk K	2014	Leuven	Flutemetamol	C	No	Yes	56	0.2271847	NA
Adamczuk K	2014	Leuven	Flutemetamol	C	No	Yes	56	NA	0.2150714
Aizenstein HJ	2008	PITT	PiB	C	No	No	38	-0.1215034	NA
Aizenstein HJ	2008	PITT	PiB	C	No	No	34	NA	-0.3791941
Pomara N	2005	Nathan Kline	Ab42	L	No	No	34	0.4100000	NA
Radanovic M	2019	Brazil	ab42/ptau181	C	No	Yes	54	0.2070000	NA
Radanovic M	2019	Brazil	Ab42	C	No	Yes	54	0.2110000	NA
Radanovic M	2019	Brazil	ptau181	C	No	Yes	54	0.2840000	NA
Resnick SM	2010	BLSA	PiB	L	Yes	Yes	51	NA	0.0163589
Resnick SM	2010	BLSA	PiB	L	Yes	Yes	51	0.0135846	NA
Roberts RO	2018	MCSA	PiB	L	No	Yes	1492	-0.0367629	NA
Roberts RO	2018	MCSA	PiB	L	No	Yes	1492	NA	-0.0385758
Roberts RO	2018	MCSA	PiB	C	No	No	1492	0.2087571	NA
Roberts RO	2018	MCSA	PiB	C	No	No	1492	NA	0.1569886
Roe CM	2013	Knight	ptau181	L	No	Yes	430	0.0548775	NA
Roe CM	2013	Knight	ptau181/ab42	L	No	Yes	430	0.0926467	NA
Roe CM	2013	Knight	ab42	L	No	Yes	430	0.0678207	NA
Rosenberg PB	2013	John Hopkns	Florbetapir	C	No	No	15	NA	0.1500000
Rosenberg PB	2013	John Hopkns	Florbetapir	C	No	No	15	0.1650000	NA
Schindler SE	2017	Knight	Ab42	C	Yes	Yes	233	NA	0.0695833
Schindler SE	2017	Knight	Ab42	C	Yes	Yes	233	0.0755000	NA
Schindler SE	2017	Knight	ptau	C	Yes	Yes	233	NA	0.0282500
Schindler SE	2017	Knight	ptau	C	Yes	Yes	233	0.0025000	NA
Schindler SE	2017	Knight	ptau/ab42	C	Yes	Yes	233	NA	0.0950833
Schindler SE	2017	Knight	ptau/ab42	C	Yes	Yes	233	0.0695000	NA
Schindler SE	2017	Knight	Ab42	L	Yes	Yes	233	0.0405000	NA
Schindler SE	2017	Knight	Ab42	L	Yes	Yes	233	NA	0.0110000
Schindler SE	2017	Knight	ptau	L	Yes	Yes	233	0.0135000	NA
Schindler SE	2017	Knight	ptau	L	Yes	Yes	233	NA	0.0175000
Schindler SE	2017	Knight	ptau/ab42	L	Yes	Yes	233	0.0385000	NA
Schindler SE	2017	Knight	ptau/ab42	L	Yes	Yes	233	NA	0.0210000
Sala-Llonch R	2017	Norway	Ab42	C	No	No	89	NA	-0.0524737
Sala-Llonch R	2017	Norway	Ab42	C	No	No	89	-0.0467196	NA
Sala-Llonch R	2017	Norway	Ab42	L	No	No	89	0.1102283	NA
Sierra-Rio A	2015	Barcelona	ab42/ptau	C	No	No	55	0.0475204	NA
Song Z	2016	DLBS	Florbetapir	C	No	No	82	NA	0.1287696
Song Z	2016	DLBS	Florbetapir	C	No	No	82	0.1225236	NA
Song Z	2015	UCSF	Florbetapir	C	No	Yes	50	NA	0.0040000
Song Z	2015	UCSF	Florbetapir	C	No	Yes	50	0.0200000	NA
Stomrud E	2010	CMRU	Ab42	C	No	Yes	37	-0.0005000	NA
Stomrud E	2010	CMRU	ptau 181	C	No	Yes	37	-0.0320000	NA
Stomrud E	2010	CMRU	ptau 181	L	No	Yes	37	0.0600000	NA
Stomrud E	2010	CMRU	Ab42	L	No	Yes	37	0.2300000	NA
Stomrud E	2010	CMRU	ptau 181	C	No	Yes	37	NA	-0.1280000
Stomrud E	2010	CMRU	Ab42	C	No	Yes	37	NA	0.1340000
Stomrud E	2010	CMRU	ptau 181	L	No	Yes	37	NA	-0.0940000
Stomrud E	2010	CMRU	Ab42	L	No	Yes	37	NA	0.4370000
Timmers T	2019	SCIENCe	Florbetapir	C	No	Yes	107	0.7456244	NA
Timmers T	2019	SCIENCe	Florbetapir	C	No	Yes	107	NA	0.6711077
Timmers T	2019	SCIENCe	Florbetapir	L	No	Yes	107	0.4860464	NA
Timmers T	2019	SCIENCe	Florbetapir	L	No	Yes	107	NA	0.3085927
Storandt M	2009	Knight	PiB	C	No	No	135	0.1209122	NA
Storandt M	2009	Knight	PiB	L	Yes	Yes	135	0.0041069	NA
Storandt M	2009	Knight	PiB	L	Yes	Yes	135	NA	0.0041069
Tardif CL	2017	INTREPAD	Ab42	C	No	No	46	-0.1107440	NA
Tardif CL	2017	INTREPAD	ttau	C	No	No	46	0.3570145	NA
Teipel SJ	2017	INSIGHT	Florbetapir	C	No	Yes	318	0.1100000	NA
Teipel SJ	2017	INSIGHT	Florbetapir	C	No	Yes	318	NA	0.1220000
van Bergen JMG	2018	zurich	Flutemetamol	C	No	Yes	116	-0.0447848	NA
van Harten AC	2013	Amsterdam	Ab42	C	No	No	132	0.1643990	NA
van Harten AC	2013	Amsterdam	ptau	C	No	No	132	NA	0.6000000
van Harten AC	2013	Amsterdam	ptau	C	No	No	132	0.4472136	NA
van Harten AC	2013	Amsterdam	ab42/ttau	C	No	No	132	NA	0.5547002
van Harten AC	2013	Amsterdam	ab42/ttau	C	No	No	132	0.0000000	NA
van Harten AC	2013	Amsterdam	Ab42	C	No	No	132	NA	0.8516583
van Harten AC	2013	Amsterdam	Ab42	L	Yes	No	132	0.8574929	NA
van Harten AC	2013	Amsterdam	ptau	L	Yes	No	132	NA	0.4472136
van Harten AC	2013	Amsterdam	ptau	L	Yes	No	132	0.1643990	NA
van Harten AC	2013	Amsterdam	ab42/ttau	L	Yes	No	132	NA	0.8682431
van Harten AC	2013	Amsterdam	ab42/ttau	L	Yes	No	132	0.8574929	NA
van Harten AC	2013	Amsterdam	Ab42	L	Yes	No	132	NA	0.5547002
Visser PJ	2009	DESCRIPA	ab42/ttau	L	No	Yes	58	0.0142401	NA
Visser PJ	2009	DESCRIPA	ab42/ttau	C	No	Yes	60	-0.0576322	NA
Visser PJ	2009	DESCRIPA	ab42	L	No	Yes	58	NA	0.0449278
Xiong C	2016	ACS	ptau	C	No	Yes	209	0.0600000	NA
Xiong C	2016	ACS	PiB	C	No	Yes	209	0.1500000	NA
Xiong C	2016	ACS	ptau	L	No	Yes	209	0.3300000	NA
Xiong C	2016	ACS	PiB	L	No	Yes	209	0.2700000	NA
Xiong C	2016	ACS	Ab42	C	No	Yes	209	0.0200000	NA
Xiong C	2016	ACS	Ab42	L	No	Yes	209	0.2200000	NA
Huang KL	2018	Taiwan	Florbetapir	C	No	Yes	11	NA	-0.5600000
Insel PS	2019	BioFINDER	ab42/ab40	C	No	Yes	329	0.0458154	NA
Janelidze S	2018	Mälmo Diet Cancer Study +BioFINDER	ab42/ab40	C	No	Yes	508	0.1161625	NA
Jansen WJ	2017	ABS	PiB	C	No	Yes	2908	0.0709225	NA
Jansen WJ	2017	ABS	PiB	C	No	Yes	2908	NA	0.0884409
Kang JM	2017	Gachon, Korea	THK5351	C	No	Yes	43	NA	0.0677500
Kawas CH	2012	90 + study	Florbetapir	C	No	No	13	0.3097950	NA
Kawas CH	2012	90 + study	Florbetapir	C	No	No	13	NA	NA
Kawas CH	2012	90 + study	Florbetapir	L	Yes	No	7	0.9540240	NA
Kawas CH	2012	90 + study	Florbetapir	L	Yes	No	7	NA	NA
Kemppainen N	2017	FINGER	PiB	C	No	Yes	48	0.1034905	NA
Kemppainen N	2017	FINGER	PiB	C	No	Yes	48	NA	0.0313408
Konijnenberg E	2019	European Information Framework for AD-PreclinAD	Flutemetamol	C	No	Yes	196	NA	0.0775520
Konijnenberg E	2019	European Information Framework for AD-PreclinAD	ab42/ab40	C	No	Yes	126	NA	0.0345742
Kristofikova Z	2014	Czech Republic.	Ab42	C	No	No	15	0.1720000	NA
Kristofikova Z	2014	Czech Republic.	ptau	C	No	No	15	0.4450000	NA
Lafirdeen ASM	2019	Multicenter France	ptau/ttau	C	No	Yes	3562	0.1733867	NA
Lafirdeen ASM	2019	Multicenter France	ab42	C	No	Yes	3562	0.2219107	NA
Leahey TM	2008	Kent USA	Ab40	C	No	Yes	35	0.3400000	NA
Liguori C	2017	Rome	Ab42	C	No	No	50	NA	0.7250000
Lilamand M	2019	MAPT	Florbetapir	C	No	No	269	0.1056808	NA
Lilamand M	2019	MAPT	Florbetapir	L	Yes	Yes	269	0.4713930	NA
Lilamand M	2019	MAPT	Florbetapir	L	Yes	Yes	269	NA	0.1433694
Lim YY	2015	Rhode Island	Florbetapir	L	No	No	63	0.0606041	NA
Lim YY	2015	Rhode Island	Florbetapir	C	No	No	63	-0.0452078	NA
Lim YY	2015	Rhode Island	Florbetapir	C	No	No	63	NA	0.1836993
Llado-Saz S	2015	Seville	Florbetapir	C	No	Yes	120	NA	0.0285706
Lu K	2019	DRC-UK	Florbetapir	C	No	Yes	502	0.2050000	NA
Lu K	2019	DRC-UK	Florbetapir	C	Yes	Yes	502	NA	0.1200000
Martikainen IK	2018	Finnish Geriatric Intervention Study	PiB	C	No	Yes	40	0.0304178	NA
Martikainen IK	2018	Finnish Geriatric Intervention Study	PiB	C	No	Yes	40	NA	-0.1323262
McMillan CT	2016	PPMI	ab42/ptau/ttau	C	No	Yes	174	0.3727037	NA
McMillan CT	2016	PPMI	Ab42	C	No	Yes	174	NA	-0.3465076
Mecca AP	2017	Yale	PiB	C	No	Yes	45	NA	0.1270000
Aschenbrenner AJ	2014	ADRC	ptau	C	No	Yes	113	NA	0.0250000
Aschenbrenner AJ	2014	ADRC	Ab42	C	No	Yes	113	NA	-0.0650000
Aschenbrenner AJ	2014	ADRC	PiB	C	No	Yes	113	NA	0.0550000
Berenguer RG	2014	Alicante	Ab42	C	No	No	39	NA	0.3770000
Berenguer RG	2014	Alicante	ab42/ptau	C	No	No	39	NA	0.3440000
Besson FL	2015	IMAP	0	C	No	No	54	-0.0504891	NA
Besson FL	2015	IMAP	0	C	No	No	54	NA	-0.1110823
Bilgel M	2018	BLSA	PiB	L	Yes	Yes	127	NA	0.0574776
Bilgel M	2018	BLSA	PiB	L	Yes	No	127	0.8434402	NA
Casaletto KB	2017	Wisconsin ADRC	ptau	C	No	No	132	NA	-0.1400000
Casaletto KB	2017	Wisconsin ADRC	Ab42	C	No	No	132	NA	0.0200000
Chatterjee P	2019	KARVIAH	Florbetaben	C	No	No	100	-0.1245904	NA
Cosentino SA	2010	WHICAP	Ab42	C	No	Yes	481	0.0200000	NA
Cosentino SA	2010	WHICAP	Ab43	L	No	Yes	481	0.0100000	NA
Cosentino SA	2010	WHICAP	Ab42	C	Yes	Yes	481	NA	0.0233333
Cosentino SA	2010	WHICAP	Ab42	L	Yes	Yes	481	NA	0.0200000
Donohue MC	2017	ADNI	Florbetapir, PiB	C	No	No	445	0.0398969	NA
Donohue MC	2017	ADNI	Florbetapir, PiB	C	No	No	445	NA	0.0764715
Donohue MC	2017	ADNI	Florbetapir, PiB	L	Yes	Yes	445	0.0100000	NA
Donohue MC	2017	ADNI	Florbetapir, PiB	L	Yes	Yes	445	NA	0.2400000
Doraiswamy PM	2014	cohort	Florbetapir	C	No	No	67	0.0693774	NA
Doraiswamy PM	2014	cohort	Florbetapir	C	No	No	67	NA	0.2334509
Doraiswamy PM	2014	cohort	Florbetapir	L	Yes	Yes	67	0.1237229	NA
Dubois B	2018	INSIGHT	Florbetapir	C	No	Yes	318	0.1213773	NA
Dubois B	2018	INSIGHT	Florbetapir	C	No	Yes	318	NA	0.0090362
Ecay-Torres M	2018	GAP	Ab42	C	No	Yes	238	0.0010000	NA
Farrell ME	2017	DLBS	Florbetapir	L	Yes	No	123	NA	0.8046629
Farrell ME	2017	DLBS	Florbetapir	L	Yes	No	123	0.7180338	NA
Franzmeier N	2018	DELCODE	Florbetapir	C	No	No	49	0.2613050	NA
Franzmeier N	2018	DELCODE	Florbetapir	C	No	Yes	49	NA	0.1203751
Gangishetti U	2018	Emory, Penn, Washu	Florbetapir	C	No	No	44	0.1008711	NA
Haapalinna F	2018	Finland	Ab42	C	No	No	57	NA	0.2860000
Hamelin L	2018	IMABio3	PiB	L	No	No	17	0.1672821	NA
Hamelin L	2018	IMABio3	PiB	C	No	No	17	0.1396219	NA
Hamelin L	2018	IMABio3	PiB	L	No	No	17	NA	0.2762891
Zhao Y	2017	GEM	PiB	C	No	No	175	NA	0.0819069
Yaffe K	2011	Healthy ABC	Ab42	C	No	No	658	-0.0323689	NA
Yaffe K	2011	Healthy ABC	Ab42	L	No	No	658	-0.1112537	NA
Yaffe K	2011	Healthy ABC	ab42/ab40	C	No	No	659	0.0198521	NA
Yaffe K	2011	Healthy ABC	ab42/ab40	L	No	No	659	-0.0800631	NA
Meng Y	2015	Peking	APL1b28	C	No	No	35	0.0250556	NA
Mielke MM	2017	MCSA	PiB	L	Yes	Yes	115	NA	0.0058281
Mielke MM	2017	MCSA	PiB	L	Yes	Yes	115	0.0116555	NA
Mielke MM	2012	MCSA	PiB	C	No	Yes	483	-0.0270166	NA
Mielke MM	2012	MCSA	PiB	C	No	Yes	483	NA	-0.0220700
Mok VC	2016	Hong Kong	PiB	C	No	Yes	75	0.1400470	NA
Mok VC	2016	Hong Kong	PiB	L	No	Yes	53	0.0726713	NA
Mok VC	2016	Hong Kong	PiB	L	No	Yes	53	0.1850980	NA
Mok VC	2016	Hong Kong	PiB	L	No	Yes	53	0.3220456	NA
Molinuevo JL	2014	Barcelona	ab42/ptau/ttau	C	No	Yes	38	0.1724884	NA
Molinuevo JL	2014	Barcelona	ab42/ptau/ttau	C	No	Yes	38	NA	0.0722052
Moon YS	2011	Korean	ab42	C	No	Yes	123	0.2500000	NA
Mueller SG	2018	UCSF	Florbetapir	C	No	No	51	0.1524001	NA
Mueller SG	2018	UCSF	Florbetapir	C	No	No	51	NA	0.0639220
Nakamura A	2018	MULNIAD	PiB	C	No	Yes	38	-0.0189523	NA
Nakamura A	2018	MULNIAD	PiB	C	No	Yes	38	NA	-0.0141493
Nebes RD	2013	PITT	PiB	C	No	Yes	71	NA	0.1295316
Ossenkoppele R	2013	BACS	PiB	C	No	Yes	81	0.0572249	NA
Ossenkoppele R	2013	BACS	PiB	C	No	Yes	81	NA	0.0833677
Ossenkoppele R	2013	BACS	PiB	C	No	Yes	81	-0.2292357	NA
Palmqvist S	2014	BioFINDER	Flutemetamol	C	No	Yes	118	0.3200000	NA
Palmqvist S	2014	BioFINDER	Flutemetamol	C	No	Yes	118	NA	0.2800000
Hammers DB	2017	Utah	Flutemetamol	C	No	Yes	27	NA	0.3002590
Hammers DB	2017	Utah	Flutemetamol	C	No	Yes	27	0.1912685	NA
Hanseeuw BJ	2017	HABS	PiB	L	No	No	277	0.1777283	NA
Hanseeuw BJ	2017	HABS	PiB	L	No	No	277	NA	-0.0411176
Hanseeuw BJ	2017	HABS	PiB	C	No	No	277	0.1317225	NA
Hanseeuw BJ	2017	HABS	PiB	C	No	No	277	NA	0.0337453
Harrington KD	2018	AIBL	Florbetapir, PiB, FDDNP	C	No	No	494	0.0829341	NA
Harrington KD	2018	AIBL	Florbetapir, PiB, FDDNP	C	No	Yes	494	NA	0.1105152
Harrington MG	2013	Pasadena	ab42/ttau	C	No	No	46	NA	-0.0663486
Jacobs HIL	2018	HABS	PiB	C	No	Yes	256	0.0759581	NA
Jacobs HIL	2018	HABS	PiB	C	No	Yes	256	NA	-0.0011623
Tolboom	2009	Amsterdam	PiB	C	No	Yes	13	0.7500000	NA
Tolboom	2009	Amsterdam	FDDNP	C	No	Yes	13	0.3900000	NA
Schott JM	2010	ADNI	Ab42	C	No	Yes	105	0.1331823	NA
Schott JM	2010	ADNI	Ab42	C	No	Yes	105	NA	-0.0303373

  # stargazer(
  #   newtab,
  #   summary = FALSE,
  #   single.row = TRUE,
  #   no.space = TRUE,
  #   digits = 2,
  #   column.sep.width = "1pt",
  #   font.size = "small", 
  #   type = 'text',
  #   out= "summary.text"
  #   )

4 Meta analysis

4.1 Grouping by cognitive domain, tracer, time of analysis

#random-effects-model for r values
#data_global <- subset(data, global_composite ==1)
data$memory_composite <- factor(data$memory_composite,
                                levels = c(0, 1),
                                labels = c("", "Memory"))
data$global_composite <- factor(data$global_composite,
                                levels = c(0, 1),
                                labels = c("", "Global"))
data$cross_long <- factor(data$cross_long,
                          levels = c("C", "L"),
                          labels = c("Baseline", "Followup"))
data$amyloid <- factor(data$amyloid,
                       levels = c(0, 1),
                       labels = c("", "Amyloid "))
data$tau <- factor(data$tau,
                   levels = c(0, 1),
                   labels = c("", "Tau"))
data$interaction <- factor(data$interaction,
                           levels = c(0, 1),
                           labels = c("", "Interaction"))
first <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data,
  n = data$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR",
 
  #The summary measure we want to calculate- fisher z
  byvar = paste(
    memory_composite,
    global_composite,
    cross_long,
    amyloid,
    tau
    
  ),
  #grouping variable
  print.byvar = gs("print.byvar")
) #print grouping variable

summary(first, digits = 2, pval=TRUE)

## Number of studies combined: k = 194
## 
##                         COR            95%-CI    t  p-value
## Random effects model 0.1679 [ 0.1252; 0.2099] 7.67 < 0.0001
## Prediction interval         [-0.3909; 0.6362]              
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0867 [0.0626; 0.1038]; tau = 0.2944 [0.2502; 0.3221];
##  I^2 = 90.4% [89.3%; 91.3%]; H = 3.22 [3.06; 3.40]
## 
## Quantifying residual heterogeneity:
##  I^2 = 90.7% [89.7%; 91.7%]; H = 3.28 [3.11; 3.47]
## 
## Test of heterogeneity:
##        Q d.f.  p-value
##  1996.26  192 < 0.0001
## 
## Results for subgroups (random effects model):
##                                         k    COR            95%-CI  tau^2
## byvar = Memory  Baseline Amyloid       45 0.1285 [ 0.0449; 0.2103] 0.0683
## byvar = Memory  Baseline Amyloid  Tau   9 0.1332 [-0.0867; 0.3407] 0.0667
## byvar =  Global Baseline Amyloid       56 0.1068 [ 0.0590; 0.1541] 0.0257
## byvar =  Global Followup Amyloid       27 0.2795 [ 0.1186; 0.4261] 0.1721
## byvar =  Global Baseline Amyloid  Tau   9 0.1357 [ 0.0197; 0.2481] 0.0124
## byvar =  Global Baseline  Tau           8 0.1994 [ 0.0383; 0.3504] 0.0281
## byvar = Memory  Followup Amyloid       12 0.2210 [ 0.0064; 0.4162] 0.1077
## byvar =  Global Followup  Tau           5 0.1308 [-0.0408; 0.2949] 0.0128
## byvar =  Global Followup Amyloid  Tau   5 0.2812 [-0.3828; 0.7536] 0.3010
## byvar = Memory  Baseline  Tau           6 0.0982 [-0.2262; 0.4030] 0.0844
## byvar = Memory  Followup  Tau           3 0.2205 [-0.3714; 0.6850] 0.0472
## byvar = Memory  Followup Amyloid  Tau   4 0.3371 [-0.5961; 0.8829] 0.4132
## byvar =   Followup Amyloid              3 0.7027 [-0.9718; 0.9991] 1.3238
## byvar =   Baseline Amyloid              2 0.0792 [-0.1372; 0.2884] 0.0003
##                                          tau      Q   I^2
## byvar = Memory  Baseline Amyloid      0.2614 323.14 86.4%
## byvar = Memory  Baseline Amyloid  Tau 0.2583  77.68 89.7%
## byvar =  Global Baseline Amyloid      0.1604 233.05 76.8%
## byvar =  Global Followup Amyloid      0.4149 601.12 95.7%
## byvar =  Global Baseline Amyloid  Tau 0.1115  18.26 56.2%
## byvar =  Global Baseline  Tau         0.1677  26.58 73.7%
## byvar = Memory  Followup Amyloid      0.3281 211.30 94.8%
## byvar =  Global Followup  Tau         0.1133  15.03 73.4%
## byvar =  Global Followup Amyloid  Tau 0.5486 167.60 97.6%
## byvar = Memory  Baseline  Tau         0.2905  56.29 91.1%
## byvar = Memory  Followup  Tau         0.2172  17.77 88.7%
## byvar = Memory  Followup Amyloid  Tau 0.6428 166.09 98.2%
## byvar =   Followup Amyloid            1.1506  27.47 92.7%
## byvar =   Baseline Amyloid            0.0181   0.12  0.0%
## 
## Test for subgroup differences (random effects model):
##                      Q d.f. p-value
## Between groups   14.58   13  0.3341
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

4.2 Forest plot

  forest(
  first,
  print.I2.ci = TRUE,
  digits.sd = 2,
  layout = "subgroup",
  #layout = "subgroup",
  squaresize = 0.8,
  plotwidth = unit(8, "cm"),
  fontsize = 5,
  addspace = FALSE,
  xlab = "Fisher z",
  calcwidth.hetstat = TRUE
)

#option 1 -all studies
# fun <- trimfill(first)
# funnel(
#   fun,
#   pch = ifelse(fun$trimfill, 1, 16),
#   level = 0.9,
#   comb.random = FALSE
# )
# 
# funnel(first, 
#        main = "Standard Error")
# funnel(first, 
#        yaxis = "invvar")
# funnel(first,
#        yaxis = "size",
#        main = "Sampling size",
#        title = "hsjs")

5 Memory

5.1 Amyloid

5.1.1 Memory amyloid baseline

data <- subset(data_full, table1 == 1 & exclude != 1)
data_memory_amyloid <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$amyloid == "1" & data$tau != "1" & data$cross_long == "C" )
MA <- metacor(
  as.numeric(all_r),
  #column with r values
  data = data_memory_amyloid,
  n = data_memory_amyloid$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
)

summary(MA, digits = 2, pval=TRUE)

## Number of studies combined: k = 45
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1285 [ 0.0449; 0.2103] 3.09  0.0035
## Prediction interval         [-0.3840; 0.5804]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0683 [0.0400; 0.1143]; tau = 0.2614 [0.1999; 0.3381];
##  I^2 = 86.4% [82.6%; 89.3%]; H = 2.71 [2.40; 3.06]
## 
## Test of heterogeneity:
##       Q d.f.  p-value
##  323.14   44 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.1.1.1 Funnel

funnel(
 MA,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MA)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      0.428 -0.92 - 1.77 0.624 0.5358138
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

Timmers T doi: 10.1016/j.pdpdt.2019.04.022 , Amsterdam, 107 NC , Florbetepan - checked Merrill DA doi: 10.3233/JAD-121903, 75 NC, FRS, FDDNP - checked Liguori doi: 10.1093/sleep/zsx011, 50 NC with, 35 with obstructuve sleep apnea, CSF ab42 - checked van Harten

5.1.1.2 Memory amyloid baseline - no outliers

data <- subset(data_full, table1 == 1 &  exclude != 1)
data_memory_amyloid <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$amyloid == "1" & data$tau != "1" & data$cross_long == "C" & data$outliers == "0")
MA_out <- metacor(
  as.numeric(all_r),
  #column with r values
  data = data_memory_amyloid,
  n = data_memory_amyloid$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
)

summary(MA_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 42
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.0707 [ 0.0335; 0.1076] 3.84  0.0004
## Prediction interval         [-0.1560; 0.2902]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0124 [0.0001; 0.0137]; tau = 0.1113 [0.0119; 0.1169];
##  I^2 = 34.7% [4.6%; 55.3%]; H = 1.24 [1.02; 1.50]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  62.78   41  0.0159
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.1.1.3 Funnel

funnel(
 MA_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MA_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI      t         p
##     -0.406 -1.03 - 0.22 -1.278 0.2085363
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

5.1.2 Memory amyloid longitudinal

data <- subset(data_full, table1 == 1 & exclude != 1)
data_memory_amyloidL <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$amyloid == "1" & data$tau != "1" & data$cross_long == "L")
MAL <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_amyloidL,
  n = data_memory_amyloidL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(MAL, digits = 2, pval=TRUE)

## Number of studies combined: k = 12
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.2210 [ 0.0064; 0.4162] 2.27  0.0447
## Prediction interval         [-0.4923; 0.7568]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.1077 [0.0503; 0.3254]; tau = 0.3281 [0.2243; 0.5705];
##  I^2 = 94.8% [92.5%; 96.4%]; H = 4.38 [3.66; 5.25]
## 
## Test of heterogeneity:
##       Q d.f.  p-value
##  211.30   11 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.1.2.1 Funnel

funnel(
 MAL,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MAL)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI    t         p
##      4.028 -0.46 - 8.52 1.76 0.1089903
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

Farrell, doi: 10.1001/jamaneurol.2017.0892, 123 NC, florbetapir) van Harten

5.1.2.2 Memory amyloid longitudinal - no outliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_memory_amyloidL <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$amyloid == "1" & data$tau != "1" & data$cross_long == "L" & data$outliers =="0")
MAL_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_amyloidL,
  n = data_memory_amyloidL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(MAL_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 9
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.0579 [-0.0275; 0.1425] 1.56  0.1564
## Prediction interval         [-0.1692; 0.2793]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0080 [0.0017; 0.0369]; tau = 0.0894 [0.0408; 0.1921];
##  I^2 = 76.1% [54.1%; 87.5%]; H = 2.04 [1.48; 2.83]
## 
## Test of heterogeneity:
##      Q d.f.  p-value
##  33.43    8 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.1.2.3 Funnel

funnel(
 MAL_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MAL_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##       1.25 -1.34 - 3.84 0.945 0.3759873
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

5.2 Amyloid and tau

5.2.1 Memory amyloid and tau baseline

data <- subset(data_full, table1 == 1 & exclude != 1)
data_memory_tau <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$tau == "1" & data$amyloid == "1" & data$cross_long == "C")

MT <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_tau ,
  n = data_memory_tau $total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(MT, digits = 2, pval=TRUE)

## Number of studies combined: k = 9
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1332 [-0.0867; 0.3407] 1.40  0.1994
## Prediction interval         [-0.4757; 0.6558]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0667 [0.0256; 0.2697]; tau = 0.2583 [0.1600; 0.5193];
##  I^2 = 89.7% [82.7%; 93.9%]; H = 3.12 [2.40; 4.04]
## 
## Test of heterogeneity:
##      Q d.f.  p-value
##  77.68    8 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.2.1.1 Funnel

funnel(
 MT,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MT)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      1.629 -4.69 - 7.94 0.506 0.6285625
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

5.2.1.2 Memory amyloid and tau baseline - no outliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_memory_tau <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$tau == "1" & data$amyloid == "1" & data$cross_long == "C" & data$outliers =="0")

MT_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_tau ,
  n = data_memory_tau $total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
)

summary(MT_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 7
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1302 [ 0.0123; 0.2445] 2.70  0.0356
## Prediction interval         [-0.1396; 0.3820]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0088 [0.0000; 0.0757]; tau = 0.0938 [0.0000; 0.2752];
##  I^2 = 0.0% [0.0%; 67.8%]; H = 1.00 [1.00; 1.76]
## 
## Test of heterogeneity:
##     Q d.f. p-value
##  5.44    6  0.4892
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.2.1.3 Funnel

funnel(
 MT_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MT_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      0.602 -1.52 - 2.73 0.556 0.6020448
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

McMillan - Parkisnons

5.2.2 Memory amyloid tau longitudinal

data <- subset(data_full, table1 == 1 &  exclude != 1) 
data_memory_tauL <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$tau == "1" & data$amyloid == "1" & data$cross_long == "L")

MTL <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_tauL ,
  n = data_memory_tauL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(MTL, digits = 2, pval=TRUE)

## Number of studies combined: k = 4
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.3371 [-0.5961; 0.8829] 1.08  0.3609
## Prediction interval         [-0.9919; 0.9980]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.4132 [0.1278; 5.8867]; tau = 0.6428 [0.3575; 2.4263];
##  I^2 = 98.2% [97.0%; 98.9%]; H = 7.44 [5.82; 9.52]
## 
## Test of heterogeneity:
##       Q d.f.  p-value
##  166.09    3 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.2.2.1 Funnel

funnel(
 MTL,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MTL)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept         95% CI     t         p
##      2.687 -34.25 - 39.63 0.143 0.8996864
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

5.2.2.2 Memory amyloid tau longitudinal - no outliers

data <- subset(data_full, table1 == 1 &  exclude != 1) 
data_memory_tauL <-
  subset(data, table1== "1" & data$memory_composite == "1" & data$tau == "1" & data$amyloid == "1" & data$cross_long == "L"& data$outliers =="0")

MTL_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_memory_tauL ,
  n = data_memory_tauL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(MTL_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 3
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.0200 [-0.0164; 0.0564] 2.36  0.1417
## Prediction interval         [-0.0927; 0.1322]             
## 
## Quantifying heterogeneity:
##  tau^2 < 0.0001 [0.0000; 0.0016]; tau = 0.0027 [0.0000; 0.0402];
##  I^2 = 0.0% [0.0%; 0.0%]; H = 1.00 [1.00; 1.00]
## 
## Test of heterogeneity:
##     Q d.f. p-value
##  0.06    2  0.9719
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

5.2.2.3 Funnel

funnel(
 MTL_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = MTL_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept      95% CI     t         p
##      0.213 -0.67 - 1.1 0.474 0.7182071
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

6 Global

6.1 Amyloid

6.1.1 Global amyloid baseline

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_amyloid <-
  subset(data, table1== "1" & data$global_composite == "1" & data$amyloid == "1" & data$tau != "0" & data$cross_long == "C")

GA <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_amyloid,
  n = data_global_amyloid$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GA, digits = 2, pval=TRUE)

## Number of studies combined: k = 9
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1357 [ 0.0197; 0.2481] 2.70  0.0273
## Prediction interval         [-0.1519; 0.4020]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0124 [0.0003; 0.0615]; tau = 0.1115 [0.0173; 0.2480];
##  I^2 = 56.2% [7.7%; 79.2%]; H = 1.51 [1.04; 2.19]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  18.26    8  0.0194
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.1.1.1 Funnel

funnel(
 GA,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GA)      

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI      t         p
##     -0.381 -3.54 - 2.77 -0.237 0.8196795
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

Timmers T doi: 10.1016/j.pdpdt.2019.04.022 , Amsterdam, 107 NC , Florbetepan - checked Merrill DA doi: 10.3233/JAD-121903, 75 NC, FRS, FDDNP - checked Roberts, Tolboom, 9 NC individuals

6.1.1.2 Global amyloid baseline - no outliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_amyloid <-
  subset(data, table1== "1" & data$global_composite == "1" & data$amyloid == "1" & data$tau != "0" & data$cross_long == "C" & data$outliers =="0")

GA_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_amyloid,
  n = data_global_amyloid$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GA_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 7
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1007 [ 0.0068; 0.1928] 2.62  0.0394
## Prediction interval         [-0.0932; 0.2872]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0042 [0.0000; 0.0393]; tau = 0.0651 [0.0000; 0.1982];
##  I^2 = 0.0% [0.0%; 54.6%]; H = 1.00 [1.00; 1.48]
## 
## Test of heterogeneity:
##     Q d.f. p-value
##  3.86    6  0.6957
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.1.1.3 Funnel

funnel(
 GA_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GA_out)      

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      0.764 -0.99 - 2.52 0.851 0.4334595
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

6.1.2 Global amyloid longitudinal

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_amyloidL <-
  subset(data, table1== "1" & data$global_composite == "1" & data$amyloid == "1" & data$tau != "0" & data$cross_long == "L")

GAL <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_amyloidL,
  n = data_global_amyloidL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
  #The summary 
) #print grouping variable

summary(GAL, digits = 2, pval=TRUE)

## Number of studies combined: k = 5
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.2812 [-0.3828; 0.7536] 1.16  0.3110
## Prediction interval         [-0.9259; 0.9761]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.3010 [0.1035; 2.5583]; tau = 0.5486 [0.3217; 1.5995];
##  I^2 = 97.6% [96.2%; 98.5%]; H = 6.47 [5.12; 8.18]
## 
## Test of heterogeneity:
##       Q d.f.  p-value
##  167.60    4 < 0.0001
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.1.2.1 Funnel

funnel(
 GAL,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GAL)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept         95% CI     t         p
##      4.664 -14.54 - 23.87 0.476 0.6665892
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

Kawas, 13 NC Farrell, doi: 10.1001/jamaneurol.2017.0892, 123 NC, florbetapir) Timmers T doi: 10.1016/j.pdpdt.2019.04.022 , Amsterdam, 107 NC , Florbetepan - checked Lilamnad van Harten Bilgel, doi: 10.1093/brain/awy150, 171 NC BLSA, CVR thresholdwas 1.062 which is the lowest for DVR taht we have.

6.1.2.2 Global amyloid longitudinal - no outliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_amyloidL <-
  subset(data, table1== "1" & data$global_composite == "1" & data$amyloid == "1" & data$tau != "0" & data$cross_long == "L" & data$outliers ==0)

GAL_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_amyloidL,
  n = data_global_amyloidL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) #print grouping variable

summary(GAL_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 4
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.0602 [-0.0022; 0.1222] 3.07  0.0545
## Prediction interval         [-0.0493; 0.1683]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0003 [0.0000; 0.0144]; tau = 0.0162 [0.0000; 0.1201];
##  I^2 = 0.0% [0.0%; 51.3%]; H = 1.00 [1.00; 1.43]
## 
## Test of heterogeneity:
##     Q d.f. p-value
##  0.94    3  0.8148
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.1.2.3 Funnel

funnel(
 GAL_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GAL_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept        95% CI      t         p
##     -1.187 -2.21 - -0.16 -2.276 0.1505914
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

6.2 Amyloid and tau

6.2.1 Global amyloid tau baseline

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_tau <-
  subset(data, table1== "1" & data$global_composite == "1" & data$tau == "1" & data$amyloid != "1" & data$cross_long == "C")


GT <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_tau,
  n = data_global_tau$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GT, digits = 2, pval=TRUE)

## Number of studies combined: k = 8
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1994 [ 0.0383; 0.3504] 2.92  0.0224
## Prediction interval         [-0.2373; 0.5691]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0281 [0.0051; 0.1535]; tau = 0.1677 [0.0714; 0.3918];
##  I^2 = 73.7% [46.5%; 87.0%]; H = 1.95 [1.37; 2.78]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  26.58    7  0.0004
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.2.1.1 Funnel

funnel(
 GT,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GT)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t        p
##      0.352 -1.59 - 2.29 0.355 0.734564
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

van harten

6.2.1.2 Global amyloid tau baseline - no outliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_tau <-
  subset(data, table1== "1" & data$global_composite == "1" & data$tau == "1" & data$amyloid != "1" & data$cross_long == "C" & data$outliers =="0")


GT_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_tau,
  n = data_global_tau$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GT_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 6
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1448 [-0.0519; 0.3307] 1.90  0.1166
## Prediction interval         [-0.3274; 0.5591]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0247 [0.0000; 0.2363]; tau = 0.1571 [0.0000; 0.4861];
##  I^2 = 48.9% [0.0%; 79.7%]; H = 1.40 [1.00; 2.22]
## 
## Test of heterogeneity:
##     Q d.f. p-value
##  9.78    5  0.0817
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.2.1.3 Funnel

funnel(
 GT_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GT_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept      95% CI     t         p
##      2.044 0.11 - 3.98 2.074 0.1067998
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

6.2.2 Global amyloid tau longitudinal

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_tauL <-
  subset(data, table1== "1" & data$global_composite == "1" & data$tau == "1" & data$amyloid != "1" & data$cross_long == "L")


GTL <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_tauL,
  n = data_global_tauL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GTL, digits = 2, pval=TRUE)

## Number of studies combined: k = 5
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1308 [-0.0408; 0.2949] 2.12  0.1015
## Prediction interval         [-0.2725; 0.4950]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0128 [0.0015; 0.1398]; tau = 0.1133 [0.0392; 0.3739];
##  I^2 = 73.4% [33.6%; 89.3%]; H = 1.94 [1.23; 3.06]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  15.03    4  0.0046
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.2.2.1 Funnel

funnel(
 GTL,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GTL)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      0.996 -4.67 - 6.67 0.344 0.7532244
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.

Xiong

6.2.2.2 Global amyloid tau longitudinal - no ouliers

data <- subset(data_full, table1 == 1 & exclude != 1)
data_global_tauL <-
  subset(data, table1== "1" & data$global_composite == "1" & data$tau == "1" & data$amyloid != "1" & data$cross_long == "L" & data$outliers=="0")


GTL_out <- metacor(
  as.numeric(all_r),
  #column with r values
  #pooled_se, #this tells R to use the seTE column to retrieve the standard error for each study
  data = data_global_tauL,
  n = data_global_tauL$total_nc,
  #sample sizes
  studlab = paste(author),
  # labels for each study
  comb.fixed = FALSE,
  #Whether to use a fixed-effects model
  comb.random = TRUE,
  #Whether to use a random-effects model
  method.tau = "SJ",
  #Which estimator to use for the between-study variance
  hakn = TRUE,
  #Which estimator to use for the between-study variance
  prediction = TRUE,
  #Whether to print a prediction interval for the effect of future studies based on present evidence
  sm = "ZCOR"
) 

summary(GTL_out, digits = 2, pval=TRUE)

## Number of studies combined: k = 4
## 
##                         COR            95%-CI    t p-value
## Random effects model 0.1227 [-0.1270; 0.3578] 1.56  0.2160
## Prediction interval         [-0.4936; 0.6569]             
## 
## Quantifying heterogeneity:
##  tau^2 = 0.0176 [0.0025; 0.3130]; tau = 0.1327 [0.0504; 0.5595];
##  I^2 = 79.6% [45.7%; 92.3%]; H = 2.21 [1.36; 3.61]
## 
## Test of heterogeneity:
##      Q d.f. p-value
##  14.69    3  0.0021
## 
## Details on meta-analytical method:
## - Inverse variance method
## - Sidik-Jonkman estimator for tau^2
## - Q-profile method for confidence interval of tau^2 and tau
## - Hartung-Knapp adjustment for random effects model
## - Fisher's z transformation of correlations

6.2.2.3 Funnel

funnel(
 GTL_out,
  yaxis= "se",
 legend = TRUE,
 main = "Standard Error",
 studlab = TRUE,
 xlim = c(-1,1),
 level=0.95, contour=c(0.9, 0.95, 0.99))$col.contour

## [1] "#808080" "#B3B3B3" "#E6E6E6"

dmetar::eggers.test(x = GTL_out)

## Eggers' test of the intercept 
## ============================= 
## 
##  intercept       95% CI     t         p
##      0.825 -6.58 - 8.23 0.218 0.8474085
## 
## Eggers' test does not indicate the presence of funnel plot asymmetry.