The datasets

All the datasets used in the book are available from this page.

Please remember that these are chosen for learning purposes: ease of understanding and a clear statistical pattern. They generally do not represent up-to-date evidence on the topics.

ACE inihibitors versus angiotensin receptor blockers

This is drawn from the 2014 Cochrane review by Li, Heran & Wright (doi:
10.1002/14651858.CD009096.pub2), analysis 1.4. The studies are quite compatible with common effect meta-analysis, or at least a very small heterogeneity variance.

As some studies also have small numbers of adverse events, it is also a example of of fitting a binomial “exact” likelihood in an arm-based model.

The numerators are in variables ARB_AE and ACEI_AE. The denominators are in ARB_n and ACEI_n. The study “Sozen2009” did not report adverse events and should be set aside. The Cochrane analysis showed subgroups (pp39-40) but we ignore those.

A deeper reading of the papers would identify differences in the definitions of adverse events and how they were recorded. Notably, ONTARGET has fewer AEs than deaths, in contrast to other studies. This, as they say, is left as an exercise to the reader.

Study,ARB_AE,ARB_d,ARB_n,ACEI_AE,ACEI_d,ACEI_n,ramipril_telmisartan_subgroup,
Fogari2011,3,,132,7,,130,1,
ONTARGET2008,465,570,4711,535,575,4687,1,
Bremner1997,37,2,334,30,2,167,0,
DETAIL2004,16,6,100,24,6,102,0,
Fogari2008,1,,122,5,,124,0,
Fogari2012,3,,102,7,,103,0,
Lacourciere2000,2,0,52,1,0,51,0,
deSpoelstra2006,3,,24,1,,22,0,
Sozen2009,,0,22,,0,22,0,

CSV file download

For contrast-based meta-analyses, we convert these observed counts into a log odds ratio, and its standard error, for each study. Here is some simple (non-Tidy) R code for it:

ava <- read.csv("ARBvACEI.csv")

ava$ACEI_odds <- ava$ACEI_AE / 
                (ava$ACEI_n - ava$ACEI_AE)
ava$ARB_odds <- ava$ARB_AE / (ava$ARB_n - ava$ARB_AE)

ava$logor <- log(ava$ACEI_odds / ava$ARB_odds)
ava$se <- sqrt((1 / ava$ACEI_AE) + 
               (1 / (ava$ACEI_n - ava$ACEI_AE)) +
               (1 / ava$ARB_AE) + 
               (1 / (ava$ARB_n - ava$ARB_AE)))

# drop one study that did not report AEs
ava <- ava[!is.na(ava$logor),]

Here is the equivalent in Stata, using the -meta esize- command to obtain the effect estimates in _meta_es and the standard errors in _meta_se:

import delimited "ARBvACEI.csv", clear

gen acei_no_ae = acei_n - acei_ae
gen arb_no_ae = arb_n - arb_ae

meta esize acei_ae acei_no_ae ///
           arb_ae arb_no_ae, ///
           studylabel(study) common(ivariance)

rename _meta_es logor
gen se2 = _meta_se^2

Green tea for weight loss

This comes from the 2012 Cochrane review by Jurgens et al. (doi:
10.1002/14651858.CD008650.pub2), analysis 1.4. This is an all-purpose example for meta-analysis with mean differences (the outcome is weight change in units of BMI: kilograms per square meter). It has the benefit of being easy to understand the subject matter, so we use it as a thought exercise for readers to consider the individual study characteristics.

Each arm (*_tea and *_control) has a number of particpants (n_*), an observed mean (mean_*), and an observed standard deviation (sd_*). The variables named cochrane_* can be ignored unless you want to compare the Bayesian and frequentist processes. We consider the “japan” variable for sub-groups in Chapter 4, and the variables “catechin” and “baseline_bmi” in the section on meta-regression.

Study,n_tea,mean_tea,sd_tea,n_control,mean_control,sd_control,cochrane_weight,cochrane_mean_diff,cochrane_lci,cochrane_uci,japan,catechin,baseline_bmi
Kozuma 2005,107,-1,0.6,119,0.3,0.4,9.51%,-1.3,-1.43,-1.17,1,540,27
Takase 2008,44,-1.2,0.5,45,0,0.3,9.4%,-1.2,-1.37,-1.03,1,540,28
Nagao 2007,123,-0.6,0.6,117,0,0.6,9.46%,-0.6,-0.75,-0.45,1,583,27
Takeshita 2008,40,-0.5,0.4,41,-0.1,0.5,9.31%,-0.4,-0.6,-0.2,1,548,28
Kajimoto 2005,129,-0.2,0.5,66,0.2,0.8,9.25%,-0.4,-0.61,-0.19,1,545,26
Suzuki 2009,18,-0.2,0.5,20,0,0.6,8.57%,-0.2,-0.55,0.15,1,392,26
Kataoka 2004,71,-0.4,0.8,71,-0.3,0.8,9.03%,-0.1,-0.36,0.16,1,560,25
Takashima 2004,10,-0.5,0.6,9,-0.5,0.6,7.38%,0,-0.54,0.54,1,570,24
Auvichayapat 2008,30,-3,1.7,30,-1.9,1.8,5.26%,-1.07,-1.96,-0.18,0,141,28
Hill 2007,19,0,0.3,19,0.2,0.4,9.2%,-0.17,-0.39,0.05,0,300,31
Hsu 2008,41,-0.1,2.8,37,0.0,0.8,5.21%,-0.05,-0.94,0.84,0,614,31
Diepvens 2005,23,-1.5,0.7,23,-1.5,0.6,8.41%,0,-0.38,0.38,0,1207,28

Contrasts are the mean difference and its standard error, and they are not hard to calculate. Here’s R (non-Tidy) code:

library(cmdstanr)

greentea <- read.csv("cochrane_green_tea_weight_loss.csv")
greentea$md <- greentea$mean_tea-greentea$mean_control
greentea$pooled_var <- (((greentea$n_tea-1)*
                         (greentea$sd_tea^2))+
                        ((greentea$n_control-1)*
                        (greentea$sd_control^2)))
                        /
                  (greentea$n_tea+greentea$n_control-2)
greentea$var_md <- (greentea$n_tea+greentea$n_control)*
                    greentea$pooled_var 
                    /
                   (greentea$n_tea*greentea$n_control)
greentea$se_md <- sqrt(greentea$var_md)

…and here’s equivalent Stata code:

import delimited "cochrane_green_tea_weight_loss.csv", varnames(1) delimiter(",") clear
label define japanlab 0 "Outside Japan" 1 "Japan"
label values japan japanlab

// calculate the mean difference
meta esize n_tea mean_tea sd_tea ///
           n_control mean_control sd_control, ///
           esize(mdiff) random(dlaird) ///
           studylabel(study) ///
           eslabel("BMI change (kg/m²)")
// the superscript 2 is Unicode, which is available in Stata from version 14 up

gen md = _meta_es 
gen se2 = _meta_se^2
gen id = _n

CSV file download and another version with stricter quotations (which JASP needs).

Thrombolytics and angioplasty network MA

This is a widely-used example used in teaching network meta-analysis, rather like the irises or Titanic datasets in elementary statistics. In our NMA chapter, we closely follow the key concepts of the book Network Meta-Analysis for Decision Making by Sofia Dias and colleagues, where they use it with BUGS code.

The dataset was first compiled by Caldwell, Ades and Higgins (2005) (doi: 10.1136/bmj.331.7521.897), comparing 6 thrombolytic drugs alongside PTCA (angioplasty).

Each study has a studyid integer, and within each studym each arm has an arm integer. narms shows how many arms are in each study (and can be used for looping within studies). treatment contains an integer id for the treatment used in each arm (study 1, for example, has 3 arms using treatments 1, 3, and 4). baseline is the treatment integer which is used as comparator in each study.

This dataset reports arm-based data with counts of a binary event. The examples we give in the book use an exact binomial likelihood, but in due course we will add a contrast-based version with log odds ratios.

studyid,arm,narms,treatment,baseline,nevents,npatients,studyname,year
1,1,3,1,1,1472,20251,GUSTO-1,1993
1,2,3,3,1,652,10396,GUSTO-1,1993
1,3,3,4,1,723,10374,GUSTO-1,1993
2,1,2,1,1,3,65,ECSG,1985
2,2,2,2,1,3,64,ECSG,1985
3,1,2,1,1,12,159,TIMI-1,1987
3,2,2,2,1,7,157,TIMI-1,1987
4,1,2,1,1,7,85,PAIMS,1989
4,2,2,2,1,4,86,PAIMS,1989
5,1,2,1,1,10,135,White,1989
5,2,2,2,1,5,135,White,1989
6,1,2,1,1,887,10396,GISSI-2,1990
6,2,2,2,1,929,10372,GISSI-2,1990
7,1,2,1,1,5,63,Cherng,1992
7,2,2,2,1,2,59,Cherng,1992
8,1,2,1,1,1455,13780,ISIS-3,1992
8,2,2,2,1,1418,13746,ISIS-3,1992
9,1,2,1,1,9,130,CI,1993
9,2,2,2,1,6,123,CI,1993
10,1,2,1,1,4,107,KAMIT,1991
10,2,2,4,1,6,109,KAMIT,1991
11,1,2,1,1,285,3004,INJECT,1995
11,2,2,5,1,270,3006,INJECT,1995
12,1,2,1,1,11,149,Zijlstra,1993
12,2,2,7,1,2,152,Zijlstra,1993
13,1,2,1,1,1,50,Riberio,1993
13,2,2,7,1,3,50,Riberio,1993
14,1,2,1,1,8,58,Grinfeld,1996
14,2,2,7,1,5,54,Grinfeld,1996
15,1,2,1,1,1,53,Zijlstra,1997
15,2,2,7,1,1,47,Zijlstra,1997
16,1,2,1,1,4,45,Akhras,1997
16,2,2,7,1,0,42,Akhras,1997
17,1,2,1,1,14,99,Widimsky,2000
17,2,2,7,1,7,101,Widimsky,2000
18,1,2,1,1,9,41,DeBoer,2002
18,2,2,7,1,3,46,DeBoer,2002
19,1,2,1,1,42,421,Widimsky,2002
19,2,2,7,1,29,429,Widimsky,2002
20,1,2,2,2,2,44,DeWood,1990
20,2,2,7,2,3,46,DeWood,1990
21,1,2,2,2,13,200,Grines,1993
21,2,2,7,2,5,195,Grines,1993
22,1,2,2,2,2,56,Gibbons,1993
22,2,2,7,2,2,47,Gibbons,1993
23,1,2,3,3,13,155,RAPID-2,1996
23,2,2,5,3,7,169,RAPID-2,1996
24,1,2,3,3,356,4921,GUSTO-3,1997
24,2,2,5,3,757,10138,GUSTO-3,1997
25,1,2,3,3,522,8488,ASSENT-2,1999
25,2,2,6,3,523,8461,ASSENT-2,1999
26,1,2,3,3,3,55,Ribichini,1996
26,2,2,7,3,1,55,Ribichini,1996
27,1,2,3,3,10,94,Garcia,1997
27,2,2,7,3,3,95,Garcia,1997
28,1,2,3,3,40,573,GUSTO-2,1997
28,2,2,7,3,32,565,GUSTO-2,1997
29,1,2,3,3,5,75,Vermeer,1999
29,2,2,7,3,5,75,Vermeer,1999
30,1,2,3,3,5,69,Schomig,2000
30,2,2,7,3,3,71,Schomig,2000
31,1,2,3,3,2,61,LeMay,2001
31,2,2,7,3,3,62,LeMay,2001
32,1,2,3,3,19,419,Bonnefoy,2002
32,2,2,7,3,20,421,Bonnefoy,2002
33,1,2,3,3,59,782,Andersen,2002
33,2,2,7,3,52,790,Andersen,2002
34,1,2,3,3,5,81,Kastrati,2002
34,2,2,7,3,2,81,Kastrati,2002
35,1,2,3,3,16,226,Aversano,2002
35,2,2,7,3,12,225,Aversano,2002
36,1,2,3,3,8,66,Grines,2002
36,2,2,7,3,6,71,Grines,2002

CSV file download

We will add some more example datasets for practising NMA here soon.

Gangliosides after stroke

This is another antiquated dataset with an unusually clear statistical pattern, from a Cochrane review by Candelise and Ciccone (1997). It was then used as an example by Duval in Chapter 8 of the book Publication Bias in Meta-analysis, edited by Rothstein, Sutton, and Borenstein.

The dataset for our purposes (publication bias visualisation and selection model) is extremely simple, having only two variables: the log odds ratio in logor, and its standard error in se. There are eleven studies.

logor,se
-0.2,0.41
-0.07,0.18
0.04,0.3
0.16,0.53
0.21,0.51
0.27,0.33
0.53,0.74
0.56,1.08
0.8,0.62
1.08,0.66
2.11,1.55

CSV file download

Exercise for osteoarthritis

This was not used as an example with code in the book, but was discussed in terms of data extraction and standardisation. We intend to expand this topic as the website develops.

It builds on the dataset used in the Cochrane review by Hurley et al (2018), analysis 1.1. That used final arm statistics in each study. Comparing it with the Cochrane forest plot, you will find that some studies (Focht, Hurley, French, and Keefe) had more than one active arm, and in keeping with Cochrane rules, we split the control group. In a modelling framework, we don’t have to do that.

The dataset here arose from complete re-extraction from source papers by Robert. Further, some later papers from an updated search (by Rachel Hallett) are included. This is an extended data extraction from all outcomes at all time points. This reflects the extra data extraction necessary to capitalise on all the information given in published studies. There is the potential for time-response curves to be added to the meta-analytic model.

study_name,active,point_name,outcome,scale_lo,scale_hi,duration,follow_up,n_bayes,mean_bayes,sd_bayes
Aglamis,0,control baseline,WOMAC pain 0-20,0,20,0,0,14,6.1,1.8
Aglamis,1,exercise baseline,WOMAC pain 0-20,0,20,0,0,17,4.2,2.5
Aglamis,0,control interim,WOMAC pain 0-20,0,20,6,0,9,5.3,2.5
Aglamis,1,exercise interim,WOMAC pain 0-20,0,20,6,0,16,2.6,1.7
Aglamis,0,control endpoint,WOMAC pain 0-20,0,20,12,0,9,6.7,1.5
Aglamis,1,exercise endpoint,WOMAC pain 0-20,0,20,12,0,16,5.7,1.9
Baker,0,control baseline ITT,WOMAC pain 0-500,0,500,0,0,22,209,98
Baker,1,exercise baseline ITT,WOMAC pain 0-500,0,500,0,0,22,207,95
Baker,0,control endpoint ITT,WOMAC pain 0-500,0,500,16,0,22,189,116
Baker,1,exercise endpoint ITT,WOMAC pain 0-500,0,500,16,0,22,128,99
Fernandes,1,exercise baseline,WOMAC pain 0-100,0,100,0,0,54,26,16.1
Fernandes,0,control baseline,WOMAC pain 0-100,0,100,0,0,54,27.3,17.9
Fernandes,1,exercise interim 4mo,WOMAC pain 0-100,0,100,12,4,55,20.6,17.2
Fernandes,0,control interim 4mo,WOMAC pain 0-100,0,100,12,4,54,25.3,18.5
Fernandes,1,exercise interim 10mo,WOMAC pain 0-100,0,100,12,31,42,16.8,17.7
Fernandes,0,control interim 10mo,WOMAC pain 0-100,0,100,12,31,42,23.4,19.6
Fernandes,1,exercise 16mo endpoint,WOMAC pain 0-100,0,100,12,57,42,17.3,14.5
Fernandes,0,control 16mo endpoint,WOMAC pain 0-100,0,100,12,57,36,22.3,18.4
Focht,0,Healthy lifestyle baseline,WOMAC pain 0-20,0,20,0,0,78,7.25,3.44
Focht,0,Healthy lifestyle 6mo,WOMAC pain 0-20,0,20,26,0,70,6.19,3.85
Focht,0,Healthy lifestyle 18mo,WOMAC pain 0-20,0,20,78,0,67,6.02,3.68
Focht,1,Exercise only baseline,WOMAC pain 0-20,0,20,0,0,80,6.64,3.49
Focht,1,Exercise only 6mo,WOMAC pain 0-20,0,20,26,0,70,6.22,3.76
Focht,1,Exercise only 18mo,WOMAC pain 0-20,0,20,78,0,64,6.24,3.76
Fransen,1,tai chi baseline,WOMAC pain 0-100,0,100,0,0,56,40.3,19
Fransen,1,tai chi 12wk,WOMAC pain 0-100,0,100,12,0,52,30.7,18.9
Fransen,0,waiting list baseline,WOMAC pain 0-100,0,100,0,0,41,44.4,17
Fransen,0,waiting list 12 wks,WOMAC pain 0-100,0,100,12,0,41,40,16.2
Hurley,0,usual care baseline,WOMAC pain 0-20,0,20,0,0,140,7.7,4
Hurley,0,usual care 6wks,WOMAC pain 0-20,0,20,6,0,128,7.1,4.1
Hurley,0,usual care 6mo,WOMAC pain 0-20,0,20,6,20,113,6.5,4.47
Hurley,1,indiv-rehab baseline,WOMAC pain 0-20,0,20,0,0,146,7.4,4
Hurley,1,indiv-rehab 6wks,WOMAC pain 0-20,0,20,6,0,128,5,3.4
Hurley,1,indiv-rehab 6mo,WOMAC pain 0-20,0,20,6,20,121,5.5,3.83
Hurley,1,group-rehab baseline,WOMAC pain 0-20,0,20,0,0,132,7.69,4.1
Hurley,1,group-rehab 6wks,WOMAC pain 0-20,0,20,6,0,111,5.5,3.9
Hurley,1,group-rehab 6mo,WOMAC pain 0-20,0,20,6,20,108,5.91,4.45
French,1,exercise baseline,NRS pain on activity,0,10,0,0,45,6,4
French,1,exercise 9 wks,NRS pain on activity,0,10,8,1,45,3,4
French,0,control baseline,NRS pain on activity,0,10,0,0,43,5,4
French,0,control 9 wks,NRS pain on activity,0,10,8,1,43,6,5
Hopman-Rock,0,control baseline,VAS pain intolerance,0,100,0,0,44,29.4,20
Hopman-Rock,0,control posttest,VAS pain intolerance,0,100,6,0,44,25.2,23.5
Hopman-Rock,0,control followup,VAS pain intolerance,0,100,6,20,44,37.9,20.3
Hopman-Rock,1,active baseline,VAS pain intolerance,0,100,0,0,55,33,22.4
Hopman-Rock,1,active posttest,VAS pain intolerance,0,100,6,0,55,27.2,21.4
Hopman-Rock,1,active followup,VAS pain intolerance,0,100,6,20,55,34.7,20.3
Kao,0,routine care baseline,Pain index,0,100,0,0,91,63.8,17.9
Kao,0,routine care 4 wks,Pain index,0,100,4,0,91,64.4,19.8
Kao,0,routine care 8 wks,Pain index,0,100,4,4,91,60.5,18.5
Kao,1,exercise baseline,Pain index,0,100,0,0,114,70.5,18.7
Kao,1,exercise 4 wks,Pain index,0,100,4,0,114,70,16.7
Kao,1,exercise 8 wks,Pain index,0,100,4,4,114,70,16.7
KeefeSolo,0,control baseline,AIMS pain,0,10,0,0,18,3.91,1.73
KeefeSolo,0,control endpoint,AIMS pain,0,10,12,0,18,4.03,2.08
KeefeSpouse,0,spouse baseline,AIMS pain,0,10,0,0,18,5.44,1.88
KeefeSpouse,0,spouse endpoint,AIMS pain,0,10,12,0,18,4,1.56
KeefeSolo,1,exercise baseline,AIMS pain,0,10,0,0,16,3.91,1.64
KeefeSolo,1,exercise endpoint,AIMS pain,0,10,12,0,16,3.19,1.85
KeefeSpouse,1,exercise+spouse baseline,AIMS pain,0,10,0,0,20,5.2,1.2
KeefeSpouse,1,exercise+spouse endpoint,AIMS pain,0,10,12,0,20,4.26,1.45
Pelonquin,0,education 3mo,AIMS2 pain,0,10,12,0,65,3.94,2.22
Pelonquin,1,exercise 3mo,AIMS2 pain,0,10,12,0,59,3.09,1.54
Sullivan,0,control baseline,AIMS pain,0,10,0,0,23,5.5,2.39
Sullivan,0,control 8wks,AIMS pain,0,10,8,0,23,4.93,2.08
Sullivan,0,control 1yr,AIMS pain,0,10,8,44,23,5.5,2.07
Sullivan,1,exercise+education baseline,AIMS pain,0,10,0,0,29,4.86,2.06
Sullivan,1,exercise+education 8wks,AIMS pain,0,10,8,0,29,3.76,2.07
Sullivan,1,exercise+education 1yr,AIMS pain,0,10,8,44,29,4.59,2.4
Yip,0,control baseline,Pain VAS,0,100,0,0,94,44.26,24.42
Yip,0,control 7 wk,Pain VAS,0,100,6,1,70,44.41,23.23
Yip,0,control 22 wk,Pain VAS,0,100,6,16,53,42.5,23.67
Yip,1,exercise baseline,Pain VAS,0,100,0,0,88,50.45,20.81
Yip,1,exercise 7 wk,Pain VAS,0,100,6,1,79,37.33,21.06
Yip,1,exercise 22 wk,Pain VAS,0,100,6,16,67,38.58,22.01
McCarthy,1,home,WOMAC pain 0-20,0,20,0,0,71,9.99,3.71
McCarthy,1,home,WOMAC pain 0-20,0,20,8,0,71,9.04,3.84
McCarthy,1,home,WOMAC pain 0-20,0,20,8,18,71,9.13,3.99
McCarthy,1,home,WOMAC pain 0-20,0,20,8,44,71,9.38,3.53
McCarthy,1,class,WOMAC pain 0-20,0,20,0,0,80,9.63,3.69
McCarthy,1,class,WOMAC pain 0-20,0,20,8,0,80,7.5,3.95
McCarthy,1,class,WOMAC pain 0-20,0,20,8,18,80,8.04,3.6
McCarthy,1,class,WOMAC pain 0-20,0,20,8,44,80,8.05,3.81
Hughes,1,fit_strong,WOMAC pain,0,20,0,0,80,5.9,3.9
Hughes,1,fit_strong,WOMAC pain,0,20,8,0,68,4.9,3.4
Hughes,1,fit_strong,WOMAC pain,0,20,8,18,60,5.1,3.7
Hughes,0,waiting,WOMAC pain,0,20,0,0,70,6.5,3.9
Hughes,0,waiting,WOMAC pain,0,20,8,0,43,6.2,4.3
Hughes,0,waiting,WOMAC pain,0,20,8,18,36,6.7,3.9
Foley,1,gym,WOMAC pain 0-20,0,20,0,0,26,8,4
Foley,1,gym,WOMAC pain 0-20,0,20,6,0,26,8,5
Foley,0,control,WOMAC pain 0-20,0,20,0,0,32,10,4
Foley,0,control,WOMAC pain 0-20,0,20,6,0,32,10,4
Hay,1,physio_advice baseline,WOMAC pain 0-20,0,20,0,0,109,9.1,4.1
Hay,1,physio_advice 3mo,WOMAC pain 0-20,0,20,13,0,93,7.36,4.3
Hay,1,physio_advice 6mo,WOMAC pain 0-20,0,20,26,0,91,7.51,4.8
Hay,1,physio_advice 12mo,WOMAC pain 0-20,0,20,52,0,90,7.41,4.4
Hay,0,control baseline,WOMAC pain 0-20,0,20,0,0,108,9.2,3.3
Hay,0,control 3mo,WOMAC pain 0-20,0,20,13,0,89,8.99,3.7
Hay,0,control 6mo,WOMAC pain 0-20,0,20,26,0,93,8.36,3.9
Hay,0,control 12mo,WOMAC pain 0-20,0,20,52,0,87,8.49,4.5
Cheung14,0,waiting list baseline,WOMAC pain 0-20,0,20,0,0,18,7.7,4.2
Cheung14,0,waiting list 8 wks,WOMAC pain 0-20,0,20,8,0,18,8.3,2.84
Cheung14,1,yoga baseline,WOMAC pain 0-20,0,20,0,0,18,9.3,4
Cheung14,1,yoga 8wks,WOMAC pain 0-20,0,20,8,0,18,7.4,2.84
Alkatan,1,cycling baseline,WOMAC pain 0-20,0,20,0,0,24,7.8,4.4
Alkatan,1,swimming baseline,WOMAC pain 0-20,0,20,0,0,24,6.9,3.4
Alkatan,1,cycling 12w,WOMAC pain 0-20,0,20,12,0,24,4.5,2.4
Alkatan,1,swimming 12w,WOMAC pain 0-20,0,20,12,0,24,4.2,2.4
Bieler,1,ST baseline,WOMAC pain 0-20,0,20,0,0,50,5.23,2.76
Bieler,1,ST 2mo,WOMAC pain 0-20,0,20,2,0,50,4.7,3.01
Bieler,1,ST 4mo,WOMAC pain 0-20,0,20,4,0,50,4.26,3.23
Bieler,1,ST 12mo,WOMAC pain 0-20,0,20,4,8,50,4.94,3.88
Bieler,1,NW baseline,WOMAC pain 0-20,0,20,0,0,50,5.77,3.59
Bieler,1,NW 2mo,WOMAC pain 0-20,0,20,2,0,50,5.06,4.44
Bieler,1,NW 4mo,WOMAC pain 0-20,0,20,4,0,50,4.17,4.53
Bieler,1,NW 12mo,WOMAC pain 0-20,0,20,4,8,50,4.37,5.68
Bieler,1,HE baseline,WOMAC pain 0-20,0,20,0,0,52,5.53,2.9
Bieler,1,HE 2mo,WOMAC pain 0-20,0,20,2,0,52,5.55,3.97
Bieler,1,HE 4mo,WOMAC pain 0-20,0,20,4,0,52,5.61,4.58
Bieler,1,HE 12mo,WOMAC pain 0-20,0,20,4,8,52,4.59,5.7
Cheung17,1,HY baseline,WOMAC pain 0-20,0,20,0,0,32,7.9,2.8
Cheung17,1,HY 8wks,WOMAC pain 0-20,0,20,8,0,32,5.1,2.7
Cheung17,1,HY 4wks,WOMAC pain 0-20,0,20,4,0,32,5.5,2.9
Cheung17,1,AS baseline,WOMAC pain 0-20,0,20,0,0,28,7.7,4.4
Cheung17,1,AS 8wks,WOMAC pain 0-20,0,20,8,0,28,6.5,2.5
Cheung17,1,AS 4wks,WOMAC pain 0-20,0,20,4,0,28,7.3,4
Cheung17,0,control baseline,WOMAC pain 0-20,0,20,0,0,23,6.3,3.1
Cheung17,0,control 8wks,WOMAC pain 0-20,0,20,8,0,23,6.5,2.7
Cheung17,0,control 4wks,WOMAC pain 0-20,0,20,4,0,23,6,3.4
Kuntz,1,yoga baseline,KOOS pain reversed,0,100,0,0,10,51.2,12.4
Kuntz,1,yoga 12wks,KOOS pain reversed,0,100,12,0,10,29.7,12.8
Kuntz,1,trad baseline,KOOS pain reversed,0,100,0,0,10,42.7,14.3
Kuntz,1,trad 12wks,KOOS pain reversed,0,100,12,0,10,34.4,13.7
Kuntz,0,none baseline,KOOS pain reversed,0,100,0,0,10,48,17.9
Kuntz,0,none 12wks,KOOS pain reversed,0,100,12,0,10,50.1,24.7
Marconcin,0,education baseline,KOOS pain reversed,0,100,0,0,32,38.6,20.4
Marconcin,0,education 12wks,KOOS pain reversed,0,100,12,0,32,32.6,18.2
Marconcin,1,SME baseline,KOOS pain reversed,0,100,0,0,35,47.9,18.9
Marconcin,1,SME 12wks,KOOS pain reversed,0,100,12,0,35,31.8,17.4
Park,0,AC baseline English,PROMIS pain,8,40,0,0,28,19.8,9.3
Park,0,AC 4wks English,PROMIS pain,8,40,4,0,28,18.5,8.6
Park,0,AC 8wks English,PROMIS pain,8,40,8,0,28,18.6,9.2
Park,0,AC 8wks + 1mo English,PROMIS pain,8,40,8,4,28,20.3,9
Park,0,AC 8wks + 3mo English,PROMIS pain,8,40,8,13,28,18.8,10.6
Park,1,yoga baseline English,PROMIS pain,8,40,0,0,32,20.7,6.9
Park,1,yoga 4wks English,PROMIS pain,8,40,4,0,32,20,8.5
Park,1,yoga 8wks English,PROMIS pain,8,40,8,0,32,17,7.7
Park,1,yoga 8wks + 1mo English,PROMIS pain,8,40,8,4,32,18.9,7.3
Park,1,yoga 8wks+3mo English,PROMIS pain,8,40,8,13,32,16.9,7.3
Park,0,AC baseline Spanish,PROMIS pain,8,40,0,0,20,18.1,7.2
Park,0,AC 4wks Spanish,PROMIS pain,8,40,4,0,20,16.3,7
Park,0,AC 8wks Spanish,PROMIS pain,8,40,8,0,20,18.3,7.9
Park,0,AC 8wks + 1mo Spanish,PROMIS pain,8,40,8,4,20,18.6,8
Park,0,AC 8wks + 3mo Spanish,PROMIS pain,8,40,8,13,20,16.6,7.4
Park,1,yoga baseline Spanish,PROMIS pain,8,40,0,0,20,16,7.1
Park,1,yoga 4wks Spanish,PROMIS pain,8,40,4,0,20,15.4,7.6
Park,1,yoga 8wks Spanish,PROMIS pain,8,40,8,0,20,12.2,6.4
Park,1,yoga 8wks + 1mo Spanish,PROMIS pain,8,40,8,4,20,11.5,5.8
Park,1,yoga 8wks+3mo Spanish,PROMIS pain,8,40,8,13,20,13,7.1
Pazit,0,control baseline,WOMAC pain 0-500,0,500,0,0,9,153.7,91.1
Pazit,0,control 8wks,WOMAC pain 0-500,0,500,8,0,9,249.7,309.3
Pazit,0,HSRT baseline,WOMAC pain 0-500,0,500,0,0,9,181.8,111.4
Pazit,0,HSRT 8wks,WOMAC pain 0-500,0,500,8,0,9,117,132.6
Takacs,1,exercise baseline,NRS pain on activity,0,10,0,0,20,5,1.8
Takacs,1,exercise 10wks,NRS pain on activity,0,10,10,0,17,2.8,1.7
Takacs,1,control baseline,NRS pain on activity,0,10,0,0,20,4.8,2.2
Takacs,1,control 10wks,NRS pain on activity,0,10,10,0,19,4.6,2.3
Bennell14,1,exercise baseline,VAS pain,0,100,0,0,49,58.8,13.3
Bennell14,0,control baseline,VAS pain,0,100,0,0,53,58,11.6
Bennell14,1,exercise 13wks,VAS pain,0,100,13,0,46,40.1,24.6
Bennell14,0,control 13wks,VAS pain,0,100,13,0,50,35.2,21.4
Bennell14,1,exercise 36wks,VAS pain,0,100,13,23,39,43.7,24.8
Bennell14,0,control 36wks,VAS pain,0,100,13,23,44,39.4,25

The “active” variable indicates whether the arm is active exercise intervention or not, although, on deeper reading, the distinction is not that clear-cut. This again forms a recommended exercise for readers, although the papers are not necessarily open access. “point_name” is string naming the arm and timepoint, for example “control 12 wks”.

These study statistics all report pain, but using various scales. The most common are variations on the WOMAC scales, which all capture the same constructs in the same phrasing (although some used translated versions from the original English), even if the minimum-maximum values differ. The KOOS scale uses high values for low pain, so for consistency with others, it is reversed here. Standardising these scales, and choosing how to report the summary from the meta-analysis, are important challenges here. “scale_lo” and “scale_hi” report the minimum and maximum values.

“duration” is how many weeks of exercise intervention took place before the recorded statistic, and “follow_up” is how many weeks post-intervention. “n_bayes”, “mean_bayes”, and “sd_bayes” are the number of participants in the arm, the mean and the standard deviation, respectively.

CSV file download