6  Systematic review and meta-analysis of Real-World Evidence

Authors

Affiliations

Dimitris Mavridis

University of Ioannina

Thomas Debray

Smart Data Analysis and Statistics B.V.

6.1 Introduction

We first load the required packages

library(dplyr)
library(gemtc)
library(netmeta)

6.2 Pairwise meta-analysis of clinical trials

6.2.1 Toculizumab for coronavirus disease 2019

In this example, we consider the results from a systematic literature review of clinical trials investigating any pharmacological in hosptialized patients with coronavirus disease 2019 (Selvarajan et al. 2022). A total of 23 randomized controlled trials were included and studied seven different interventions: dexamethasone, remdesivir, tocilizumab, hydroxychloroquine, combination of lopinavir/ritonavir, favipiravir and interferon-β. We here focus on the synthesis of 7 trials that comparted toculizumab (TOCI) to standard care (STD) and collected mortality data.

studlab	treat1	treat2	event1	n1	event2	n2
Hermine et al	TOCI	STD	7	63	8	67
Rosas et al	TOCI	STD	58	294	28	144
Salama et al	TOCI	STD	26	249	11	128
Salvarini et al	TOCI	STD	2	60	1	66
Stone et al	TOCI	STD	9	161	3	82
Veiga et al	TOCI	STD	14	65	6	64

We now conduct a pairwise meta-analysis to assess the pooled effect of tocilizumab versus standard care. For each study, the log odds ratio and corresponding standard error is derived after which the corresponding estimates are pooled using the Mantel-Haenszel method.

results.TOCI <- metabin(event1,n1,event2,n2,studlab,data=tocilizumab,
                        sm = "OR", main = "tocilizumab vs standard care", 
                        prediction = TRUE)
forest(results.TOCI, leftcols = "studlab", rightcols = "effect.ci")

Altough a random effects meta-analysis was conducted, no heterogeneity was found (\(\tau\)=0, with a 95% confidence interval ranging from 0 to 0.85).

6.2.2 Remdesivir for coronavirus disease 2019

In aforementioned example, a total of 4 trials compared remdesivir to standard care:

6.3 Network meta-analysis of clinical trials

We here use the R packages netmeta for conducting a frequentist network meta-analysis. A detailed tutorial on the use of netmeta is available from the book Doing Meta-Analysis with R: A Hands-On Guide.

6.3.1 Interventions for coronavirus disease 2019

We here consider data from a study which aimed to assess the comparative effectiveness of remdesivir and tocilizumab for reducing mortality in hospitalised COVID-19 patients. 80 trials were identified from two published network meta-analyses (Selvarajan et al. 2022), (Siemieniuk et al. 2020), a living COVID-19 trial database (COVID-NMA Initiative) [Covid-NMA.com], and a clinical trial database [clinicaltrials.gov]. Trials were included in this study if the patient population included hospitalized COVID-19 patients, active treatment was remdesivir or tocilizumab, comparator treatment was placebo or standard care, short-term mortality data was available, and the trial was published. 21 trials were included. For included trials, a risk of bias score was extracted from the COVID-NMA Initiative.

studlab	treat1	treat2	event1	n1	event2	n2
Ader	REM	STD	34	414	37	418
Beigel (ACTT-1)	REM	STD	59	541	77	521
Broman	TOCI	STD	1	57	0	29
Criner	REM	STD	4	384	4	200
Declerq (COV-AID)	TOCI	STD	10	81	9	74
Gordon (REMAP-CAP)	TOCI	STD	83	353	116	358
Hermine (CORIMUNO)	TOCI	STD	7	63	8	67
Horby (RECOVERY)	TOCI	STD	621	2022	729	2094
Islam	REM	STD	0	30	0	30
Mahajan	REM	STD	5	34	3	36
Pan (WHO Solidarity)	REM	STD	602	4146	643	4129
Rosas (COVACTA)	TOCI	STD	58	294	28	144
Rutgers	TOCI	STD	21	174	34	180
Salama (EMPACTA)	TOCI	STD	26	249	11	128
Salvarani	TOCI	STD	2	60	1	63
Soin (COVINTOC)	TOCI	STD	11	92	15	88
Spinner	REM	STD	5	384	4	200
Stone (BACC-BAY)	TOCI	STD	9	161	4	82
Talaschian	TOCI	STD	5	17	4	19
Veiga (TOCIBRAS)	TOCI	STD	14	65	6	64
Wang	REM	STD	22	158	10	78

The corresponding network is displayed below:

Evidence network of the 21 coronavirus-19 trials

We use the following command to calculate the log odds ratios and corresponding standard errors for each study:

covid <- pairwise(treat = treat, 
                  event = event, 
                  n = n, 
                  studlab = studlab, 
                  sm = "OR")
head(covid)

TE	seTE	studlab	treat1	treat2	event1	n1	event2	n2	incr	allstudies
-0.0819293	0.2483849	Ader	REM	STD	34	414	37	418	0.0	FALSE
-0.3483875	0.1851030	Beigel (ACTT-1)	REM	STD	59	541	77	521	0.0	FALSE
0.4487619	1.6487159	Broman	TOCI	STD	1	57	0	29	0.5	FALSE
-0.6620566	0.7125543	Criner	REM	STD	4	384	4	200	0.0	FALSE
0.0170679	0.4904898	Declerq (COV-AID)	TOCI	STD	10	81	9	74	0.0	FALSE
-0.4442338	0.1688337	Gordon (REMAP-CAP)	TOCI	STD	83	353	116	358	0.0	FALSE

Below, we conduct a random effects network meta-analysis where we consider standard care (STD) as the control treatment. Note that we have one study where zero cell counts occur, this study will not contribute to the NMA as the log odds ratio and its standard error cannot be determined.

NMA.covid <- netmeta(TE = TE, seTE = seTE, treat1 = treat1, treat2 = treat2,
                     studlab = studlab, data = covid, sm = "OR", ref = "STD",
                     comb.random = TRUE, common = FALSE, warn = FALSE)
NMA.covid 

Number of studies: k = 20
Number of pairwise comparisons: m = 20
Number of treatments: n = 3
Number of designs: d = 2

Random effects model

Treatment estimate (sm = 'OR', comparison: other treatments vs 'STD'):
         OR           95%-CI     z p-value
REM  0.8999 [0.8067; 1.0039] -1.89  0.0588
STD       .                .     .       .
TOCI 0.8301 [0.7434; 0.9268] -3.31  0.0009

Quantifying heterogeneity / inconsistency:
tau^2 = 0; tau = 0; I^2 = 0% [0.0%; 48.9%]

Tests of heterogeneity (within designs) and inconsistency (between designs):
                    Q d.f. p-value
Total           16.38   18  0.5663
Within designs  16.38   18  0.5663
Between designs  0.00    0      --

A league table of the treatment effect estimates is given below:

netleague(NMA.covid)

League table (random effects model):
                                                                        
                     REM 0.8999 [0.8067; 1.0039]                       .
 0.8999 [0.8067; 1.0039]                     STD 1.2047 [1.0789; 1.3451]
 1.0842 [0.9282; 1.2663] 1.2047 [1.0789; 1.3451]                    TOCI

We can also present the results in a forest plot:

We now consider a Bayesian random effects network meta-analysis that analyzes the observed event counts using a binomial link function.

bdata <- data.frame(study = studlab,
                    treatment = treat,
                    responders = event,
                    sampleSize = n)

network <- mtc.network(data.ab  = bdata)

model <- mtc.model(network,
                   likelihood = "binom",
                   link = "log",
                   linearModel = "random",
                   n.chain = 3)

# Adaptation
mcmc1 <- mtc.run(model, n.adapt = 1000, n.iter = 1000, thin = 10)

Compiling model graph
   Resolving undeclared variables
   Allocating nodes
Graph information:
   Observed stochastic nodes: 42
   Unobserved stochastic nodes: 45
   Total graph size: 930

Initializing model

# Sampling
mcmc2 <- mtc.run(model, n.adapt = 10000, n.iter = 100000, thin = 10)

Compiling model graph
   Resolving undeclared variables
   Allocating nodes
Graph information:
   Observed stochastic nodes: 42
   Unobserved stochastic nodes: 45
   Total graph size: 930

Initializing model

We can extract the pooled treatment effect estimates from the posterior distribution. When using STD as control group, we have:

summary(relative.effect(mcmc2, t1 = "STD"))

Results on the Log Risk Ratio scale

Iterations = 10010:110000
Thinning interval = 10 
Number of chains = 3 
Sample size per chain = 10000 

1. Empirical mean and standard deviation for each variable,
   plus standard error of the mean:

              Mean      SD  Naive SE Time-series SE
d.STD.REM  -0.1069 0.09691 0.0005595      0.0008094
d.STD.TOCI -0.1120 0.08259 0.0004768      0.0008746
sd.d        0.1122 0.08815 0.0005089      0.0017902

2. Quantiles for each variable:

                2.5%      25%      50%      75%   97.5%
d.STD.REM  -0.314930 -0.16017 -0.10346 -0.04992 0.08408
d.STD.TOCI -0.259309 -0.16297 -0.11920 -0.06839 0.07738
sd.d        0.003287  0.04449  0.09368  0.15868 0.32866

The corresponding odds ratios are as follows:

Comparison	95% CrI
REM vs. STD	0.9 (0.73; 1.09)
TOCI vs. STD	0.89 (0.77; 1.08)
REM vs. TOCI	1.02 (0.75; 1.27)

Finally, we expand the COVID-19 network with trials investigating the effectiveness of hydroxychloroquine (HCQ), lopinavir/ritonavir (LOPI), dexamethasone (DEXA) or interferon-\(\beta\) (INTB) (Selvarajan et al. 2022). The corresponding network is displayed below:

Evidence network of the 33 coronavirus-19 trials

We conducted a random effects network meta-analysis, results are depicted below:

Number of studies: k = 33
Number of pairwise comparisons: m = 33
Number of treatments: n = 7
Number of designs: d = 6

Random effects model

Treatment estimate (sm = 'OR', comparison: other treatments vs 'STD'):
         OR           95%-CI     z p-value            95%-PI
DEXA 0.8557 [0.7558; 0.9688] -2.46  0.0139  [0.7463; 0.9812]
HCQ  1.1809 [0.8934; 1.5610]  1.17  0.2428  [0.8786; 1.5872]
INTB 1.1606 [0.9732; 1.3841]  1.66  0.0973  [0.9604; 1.4026]
LOPI 1.0072 [0.8906; 1.1392]  0.11  0.9085  [0.8794; 1.1537]
REM  0.8983 [0.8014; 1.0070] -1.84  0.0658  [0.7913; 1.0199]
STD       .                .     .       .                 .
TOCI 0.8304 [0.7410; 0.9306] -3.20  0.0014  [0.7316; 0.9426]

Quantifying heterogeneity / inconsistency:
tau^2 = 0.0004; tau = 0.0205; I^2 = 0.6% [0.0%; 42.3%]

Tests of heterogeneity (within designs) and inconsistency (between designs):
                    Q d.f. p-value
Total           27.18   27  0.4543
Within designs  27.18   27  0.4543
Between designs  0.00    0      --

We can calculate the P score for each treatment as follows:

netrank(NMA.covidf)

     P-score
TOCI  0.9070
DEXA  0.8357
REM   0.7143
STD   0.4027
LOPI  0.3899
HCQ   0.1336
INTB  0.1166

6.3.2 Pharmacologic treatments for chronic obstructive pulmonary disease

In this example, we consider the resuls from a systematic review of randomized controlled trials on pharmacologic treatments for chronic obstructive pulmonary disease (Baker, Baker, and Coleman 2009). The primary outcome, occurrence of one or more episodes of COPD exacerbation, is binary (yes / no). For this outcome, five drug treatments (fluticasone, budesonide, salmeterol, formoterol, tiotropium) and two combinations (fluticasone + salmeterol, budesonide + formoterol) were compared to placebo. The authors considered the two combinations as separate treatments instead of evaluating the individual components.

data(Baker2009)

study	year	id	treatment	exac	total
Llewellyn-Jones 1996	1996	1	Fluticasone	0	8
Llewellyn-Jones 1996	1996	1	Placebo	3	8
Boyd 1997	1997	2	Salmeterol	47	229
Boyd 1997	1997	2	Placebo	59	227
Paggiaro 1998	1998	3	Fluticasone	45	142
Paggiaro 1998	1998	3	Placebo	51	139

Baker <- pairwise(treat = treatment,
                  event = exac,
                  n = total,
                  studlab = id,
                  sm = "OR",
                  data = Baker2009)

NMA.COPD <- netmeta(TE = TE, seTE = seTE, treat1 = treat1, treat2 = treat2,
                    studlab = studlab, data = Baker, sm = "OR", ref = "Placebo",
                    comb.random = TRUE)

Warning: Comparisons with missing TE / seTE or zero seTE not considered in
network meta-analysis.

Comparisons not considered in network meta-analysis:
 studlab                 treat1     treat2 TE seTE
      39 Fluticasone+Salmeterol    Placebo NA   NA
      39 Fluticasone+Salmeterol Salmeterol NA   NA
      39             Salmeterol    Placebo NA   NA

netgraph(NMA.COPD)

We can test for local network inconsistency as follows:

netsplit(NMA.COPD)

By default, indirect estimates are derived using the back-calculation method. A comparison of direct and indirect treatment effect estimates (depicted as odds ratio) from the random effects network meta-analysis model is given in the table below:

comparison	k	nma	direct	indirect	RoR	p.RoR
Budesonide:Budesonide+Formoterol	2	1.20	1.20	745.35	0.00	0.28
Budesonide:Fluticasone	0	0.93	.	0.93	.	.
Budesonide:Fluticasone+Salmeterol	0	1.07	.	1.07	.	.
Budesonide:Formoterol	2	0.80	0.74	1.39	0.53	0.18
Budesonide:Placebo	2	0.80	0.85	0.52	1.63	0.33
Budesonide:Salmeterol	0	1.00	.	1.00	.	.
Budesonide:Tiotropium	0	1.17	.	1.17	.	.
Budesonide+Formoterol:Fluticasone	0	0.77	.	0.77	.	.
Budesonide+Formoterol:Fluticasone+Salmeterol	0	0.89	.	0.89	.	.
Budesonide+Formoterol:Formoterol	2	0.67	0.63	0.94	0.67	0.37
Budesonide+Formoterol:Placebo	2	0.67	0.72	0.39	1.83	0.21
Budesonide+Formoterol:Salmeterol	0	0.83	.	0.83	.	.
Budesonide+Formoterol:Tiotropium	0	0.97	.	0.97	.	.
Fluticasone:Fluticasone+Salmeterol	4	1.15	1.09	1.43	0.77	0.20
Fluticasone:Formoterol	0	0.86	.	0.86	.	.
Fluticasone:Placebo	8	0.86	0.90	0.67	1.34	0.15
Fluticasone:Salmeterol	4	1.08	1.07	1.09	0.98	0.90
Fluticasone:Tiotropium	0	1.26	.	1.26	.	.
Fluticasone+Salmeterol:Formoterol	0	0.75	.	0.75	.	.
Fluticasone+Salmeterol:Placebo	6	0.75	0.80	0.67	1.19	0.23
Fluticasone+Salmeterol:Salmeterol	6	0.94	0.91	1.01	0.90	0.53
Fluticasone+Salmeterol:Tiotropium	1	1.09	1.14	1.08	1.05	0.77
Formoterol:Placebo	4	1.00	0.96	2.79	0.34	0.09
Formoterol:Salmeterol	0	1.25	.	1.25	.	.
Formoterol:Tiotropium	1	1.46	4.20	1.39	3.03	0.07
Salmeterol:Placebo	15	0.80	0.79	0.85	0.93	0.65
Tiotropium:Placebo	12	0.69	0.70	0.63	1.11	0.50
Salmeterol:Tiotropium	3	1.17	1.14	1.18	0.96	0.80

6.3.3 Advanced Therapies for Ulcerative Colitis

In this example, we consider a systematic literature review of Phase 3 randomized controlled trials investigating the following advanced therapies: infliximab, adalimumab, vedolizumab, golimumab, tofacitinib, ustekinumab, filgotinib, ozanimod, and upadacitinib (Panaccione et al. 2023). This review included 48 RCTs, from which 23 were found eligible for inclusion in a network meta-analysis. The included RCT populations were largely comparable in their baseline characteristics, though some heterogeneity was noted in weight, disease duration, extent of disease, and concomitant medications. A risk of bias assessment showed a low risk of bias for all included RCTs, which were all industry sponsored.

We here focus on the synthesis of 18 trials that contributed efficacy data for induction in bio-naive populations. The following FDA- and/or EMA-approved biologic or SMD doses were investigated:

• Adalimumab subcutaneous 160 mg at week 0, 80 mg at week 2, and 40 mg at week 4 (ADA160/80)
• Infliximab intravenous 5 mg/kg (INF5) at weeks 0, 2, and 6 then every 8 weeks
• Infliximab intravenous 10 mg/kg (INF10) at weeks 0, 2, and 6 then every 8 weeks
• Filgotinib oral 100 mg once daily (FIL100)
• Filgotinib oral 200 mg once daily (FIL200)
• Golimumab subcutaneous 200 mg at week 0 and 100 mg at week 2 (GOL200/100)
• Ozanimod oral 0.23 mg once daily for 4 days, 0.46 mg once daily for 3 days, then 0.92 mg once daily (OZA0.92)
• Tofacitinib oral 10 mg twice daily for 8 weeks (TOF10)
• Upadacitinib oral 45 mg once daily for 8 weeks (UPA45)
• Ustekinumab intravenous 6 mg/kg at week 0 (UST6)
• Vedolizumab intravenous 300 mg at weeks 0, 2, and 6 (VED300)

The reference treatment is placebo (PBO).

Efficacy outcomes (i.e., clinical remission) data of induction bio-naïve populations

studlab	treat1	treat2	event1	n1	event2	n2
ACT-1	INF10	INF5	39	122	47	121
ACT-1	INF10	PBO	39	122	18	121
ACT-1	INF5	PBO	47	121	18	121
ACT-2	INF10	INF5	33	120	41	121
ACT-2	INF10	PBO	33	120	7	123
ACT-2	INF5	PBO	41	121	7	123
GEMINI 1	VED300	PBO	30	130	5	76
Japic CTI-060298	INF5	PBO	21	104	11	104
Jiang 2015	INF5	PBO	22	41	9	41
M10-447	ADA160/80	PBO	9	90	11	96
NCT01551290	INF5	PBO	11	50	5	49
NCT02039505	VED300	PBO	22	79	6	41
OCTAVE 1	TOF10	PBO	56	222	9	57
OCTAVE 2	TOF10	PBO	43	195	4	47
PURSUIT-SC	GOL200/100	PBO	45	253	16	251
SELECTION	FIL100	FIL200	47	277	60	245
SELECTION	FIL100	PBO	47	277	17	137
SELECTION	FIL200	PBO	60	245	17	137
TRUE NORTH	OZA0.92	PBO	66	299	10	151
U-ACCOMPLISH	UPA45	PBO	54	166	3	81
U-ACHIEVE Study 2	UPA45	PBO	41	145	4	72
ULTRA-1	ADA160/80	PBO	24	130	12	130
ULTRA-2	ADA160/80	PBO	32	150	16	145
UNIFI	UST6	PBO	27	147	15	151

The corresponding network is displayed below:

Evidence network of 18 trials that contributed efficacy data for induction in bio-naive populations

Below, we conduct a random effects network meta-analysis of the reported study effects (expressed as odds ratio) and consider placebo (treat = "PBO") as the control treatment.

NMA.uc <- netmeta(TE = TE, seTE = seTE, treat1 = treat1, treat2 = treat2,
                  studlab = studlab, data = UlcerativeColitis, sm = "OR", 
                  ref = "PBO", common = FALSE, comb.random = TRUE)
NMA.uc

All treatments except FIL100 and UST6 are significantly more efficacious than PBO at inducing clinical remission. We can now estimate the probabilities of each treatment being at each possible rank and the SUCRAs (Surface Under the Cumulative RAnking curve):

sucra.uc <- rankogram(NMA.uc, nsim = 100, random = TRUE, common = FALSE, 
                      small.values = "undesirable")

# Exctract the SUCRA values
sucra.uc$ranking.random

 ADA160/80     FIL100     FIL200 GOL200/100      INF10       INF5    OZA0.92 
0.28090909 0.14363636 0.43272727 0.65454545 0.61727273 0.73727273 0.78909091 
       PBO      TOF10      UPA45       UST6     VED300 
0.01727273 0.37363636 0.98454545 0.37363636 0.59545455 

These results indicate that 98.5% of the evaluated treatments are worse than UPA45.

Version info

This chapter was rendered using the following version of R and its packages:

R version 4.3.0 (2023-04-21)
Platform: aarch64-apple-darwin20 (64-bit)
Running under: macOS 14.2.1

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRblas.0.dylib 
LAPACK: /Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

time zone: Europe/Amsterdam
tzcode source: internal

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] netmeta_2.8-2    meta_6.5-0       gemtc_1.0-2      coda_0.19-4     
[5] dplyr_1.1.4      kableExtra_1.3.4

loaded via a namespace (and not attached):
 [1] magic_1.6-1         utf8_1.2.4          generics_0.1.3     
 [4] xml2_1.3.6          stringi_1.8.2       lattice_0.22-5     
 [7] lme4_1.1-35.1       digest_0.6.33       magrittr_2.0.3     
[10] evaluate_0.23       grid_4.3.0          mvtnorm_1.2-4      
[13] CompQuadForm_1.4.3  fastmap_1.1.1       plyr_1.8.9         
[16] jsonlite_1.8.8      Matrix_1.6-4        httr_1.4.7         
[19] rvest_1.0.3         fansi_1.0.6         viridisLite_0.4.2  
[22] scales_1.3.0        codetools_0.2-19    numDeriv_2016.8-1.1
[25] abind_1.4-5         cli_3.6.1           rlang_1.1.2        
[28] munsell_0.5.0       splines_4.3.0       withr_2.5.2        
[31] yaml_2.3.7          tools_4.3.0         nloptr_2.0.3       
[34] metafor_4.4-0       minqa_1.2.6         ggplot2_3.4.4      
[37] colorspace_2.1-0    mathjaxr_1.6-0      webshot_0.5.5      
[40] boot_1.3-28.1       vctrs_0.6.5         R6_2.5.1           
[43] lifecycle_1.0.4     stringr_1.5.1       htmlwidgets_1.6.4  
[46] MASS_7.3-60         pkgconfig_2.0.3     gtable_0.3.4       
[49] pillar_1.9.0        glue_1.6.2          Rcpp_1.0.11        
[52] systemfonts_1.0.5   highr_0.10          xfun_0.41          
[55] tibble_3.2.1        tidyselect_1.2.0    rstudioapi_0.15.0  
[58] knitr_1.45          htmltools_0.5.7     nlme_3.1-164       
[61] igraph_1.5.1        rmarkdown_2.25      svglite_2.1.3      
[64] compiler_4.3.0      metadat_1.2-0      

References

Baker, William L, Erica L Baker, and Craig I Coleman. 2009. “Pharmacologic Treatments for Chronic Obstructive Pulmonary Disease: A Mixed-Treatment Comparison Meta-Analysis.” Pharmacotherapy 29 (8): 891–905. https://doi.org/10.1592/phco.29.8.891.

Panaccione, Remo, Eric B Collins, Gil Y Melmed, Severine Vermeire, Silvio Danese, Peter D R Higgins, Christina S Kwon, et al. 2023. “Efficacy and Safety of Advanced Therapies for Moderately to Severely Active Ulcerative Colitis at Induction and Maintenance: An Indirect Treatment Comparison Using Bayesian Network Meta-Analysis.” Crohn’s & Colitis 360 5 (2). https://doi.org/10.1093/crocol/otad009.

Selvarajan, Sandhiya, Annuja Anandaradje, Santhosh Shivabasappa, Deepthy Melepurakkal Sadanandan, N. Sreekumaran Nair, and Melvin George. 2022. “Efficacy of Pharmacological Interventions in COVID-19: A Network Meta-Analysis.” British Journal of Clinical Pharmacology 88 (9): 4080–91. https://doi.org/10.1111/bcp.15338.

Siemieniuk, Reed AC, Jessica J Bartoszko, Dena Zeraatkar, Elena Kum, Anila Qasim, Juan Pablo Dı́az Martinez, Ariel Izcovich, et al. 2020. “Drug Treatments for Covid-19: Living Systematic Review and Network Meta-Analysis.” BMJ, July, m2980. https://doi.org/10.1136/bmj.m2980.