Network Import - Network Meta Analysis Utility Command for Stata14

[elisabeth louise]

I am trying to use 'network import' command to conduct a multiple treatment network meta-analysis. Some of the studies have multiple doses/treatment arms. I am setting up my data as follows:

study	treatment1	treatment2	effect_size	stderr
1	placebo	druga	.5	.2
2	placebo	drugb_1	.3	.1
2	placebo	drugb_2	.9	.05
3	placebo	drugc_1	.4	.2
3	placebo	drugc_2	.5	.3

Where study 1 is comparing drug a to place. study 2 is a multi-arm trial comparing drug b, dose 1 to placebo and drug b, dose 2 to placebo. study 3 is a multiarm trial comparing drug c, dose 1 to placebo and drug c, dose 2 to placebo. effect size and stderr reflect the estimates for mean effect and standard error for each comparison. I am setting up the data this way so that I do not have to estimate the covariance structure for the multi-arm trials, since the stata help file indicates that the network import command will do this in the background.

However, I am running into an error where stata says:

. network import, study(study) treat(trt1 trt2) effect(fev_effect) stde(fev_se)
Importing from pairs format
Reference treatment: drugc_2
All treatments: drugc_2 drugc_1 drugb_2 drugb_1 druga
1 contradiction in 5 observations
assertion is false

So, it is selecting drugc_2 instead of placebo as my main reference comparator, which doesn't allow for comparisons across other studies.

Does anyone know how to circumvent this issue? I am wanting to have placebo as the reference treatment for all studies.

Thanks in advance!

[Federico Tedeschi]

My understadning of the "network import" command in the "pairs" format is that you need all possible pairwise contrasts. Thus, you'd need "drugb1 vs drugb2" in study 2, and "drugc1 vs drugc2" in study 3. The covariance structure may only be calculated if all the three standard errors are available.
Alternatively, you could use the "augmented" approach, by showing contrasts, but it seems to me that, in that case, you should have one raw per study. More importantly, you should add the covariance between the first and the second contrast.
The point is that, to calculate the variance of the "drug1 vs drug2" comparisons, the information on the variance of the "drug1 vs placebo" and "drug2 vs placebo" is not enough: you also need their covariances.

[MYVIZHI KANNAN]

same problem but different. I recently practice a NMA specifically survival data what i have is ln(HR) and SE(lnHR). based on "network import" command i set a network data format

 
 network import, tr(t1 t2) eff(lnhr) study(study_name) stderr(selnhr)

network map i get it using this code

 
 network map

The issue i have is in that one study was multi-arm trail(3 arm) and study name is Llovet et al. 2002 due to this i cant able to run further codes like

 
 network meta consistency network meta consistency, fixed network forest, msize (*0.15) diamond eform xlabel (0.1 1 10 100) colors (black blue red) list intervalplot, eform null (1) labels (Network Meta Analysis Covid) textsize (2) xlabel (0.01 0.1 1 10) network rank min, line cumulative xlabel (1/4) seed (10000) reps (10000) meanrank network rank min, seed(48106) all line cumul reps(10000) meanrank network rank max, all zero gen(prob) sucra prob*, rankogr

this is reference articel where i took the data https://doi.org/10.1371/journal.pone.0184597 And below table is data
and i also attached both excel file and error picture

study_name t1 t2 n1 n2 logHR SE Pelletier et al. 1990 1 9 21 21 0.2971 0.3988 Madden et al. 1993 1 9 25 25 0.0276 0.3487 Groupe d'Etude et al. 1995 1 9 50 46 -0.3407 0.225 Pelletier et al. 1998 1 9 37 36 -0.0834 0.2694 Llovet et al. 2002 1 9 40 35 -0.755 0.3371 Lo et al. 2002 1 9 40 39 -0.6932 0.2461 FFCD 9402 et al. 2208 1 9 62 61 -0.0773 0.1873 Mabed et al. 2009 1 9 50 50 -0.4185 0.239 Lin et al. 1988 4 9 42 21 -0.9239 0.3291 Bruix et al. 1998 4 9 40 40 0.0926 0.2582 Llovet et al. 2002 4 9 37 35 -0.5621 0.3068 Raoul et al. 1994 3 9 14 13 -1.1394 0.3217 Kawai et al. 1991 4 1 137 145 -0.2114 0.1683 Chang et al. 1994 4 1 24 22 -0.3147 0.2803 Meyer et al. 2013 4 1 42 44 -0.0943 0.2943 Yu et al. 2014 4 1 45 45 -0.1898 0.2901 Raoul et al. 1997 3 1 73 69 -0.2157 0.2396 Kolligs et al. 2015 3 1 13 15 0.758 0.6565 Salem et al. 2016 3 1 24 21 -0.0058 0.4595 Sacco et al. 2011 2 1 33 34 0.0268 0.5346 Golfieri et al. 2014 2 1 89 88 -0.0151 0.2343 Malagari et al. 2009 2 4 41 43 0.0334 0.6022 Brown et al. 2015 2 4 50 51 -0.1044 0.228 Pitton et al. 2015 2 3 12 12 0.0489 0.5345 Li et al. 2009 8 1 108 108 -0.1092 0.0368 Kudo et al. 2011 8 1 229 227 0.0583 0.2209 Britten et al. 2011 8 1 15 15 0.734 0.5855 Inaba et al. 2013 8 1 50 51 0.0583 0.3092 Kudo et al. 2014 8 1 249 253 -0.1054 0.1588 Printer et al. 2015 8 1 16 16 0.5306 0.3837 Wang et al. 2015 8 1 61 64 -0.9943 0.3159 Lencioni et al. 2016 7 2 154 153 -0.1076 0.2005 Bartolozzi et al. 1995 6 1 26 27 -0.734 0.4059 Yamamoto et al. 1997 6 1 50 50 -0.6162 0.2069 Wu et al. 1998 6 1 50 52 -0.9416 0.2728 Xu et al. 2002 6 1 23 22 -0.4943 0.1798 Becker et al. 2005 6 1 27 25 -0.3567 0.2657 Yang et al. 2008 6 1 24 11 -0.5798 0.4793 Liu et al. 2009 6 1 39 39 -0.6539 0.378 Zhao et al. 2011 6 1 23 24 -0.6349 0.2551 Huang et al. 2017 6 1 60 60 -0.6349 0.1972 Xue et al. 1995 5 1 21 20 -0.6733 0.3062 Leng et al. 2000 5 1 36 39 -0.478 0.2351 Peng et al. 2000 5 1 43 48 -0.5108 0.1588 Wang et al. 2000 5 1 20 20 -0.2357 0.2042 Li et al. 2003 5 1 41 41 -0.6539 0.2151 Zhao et al. 2006 5 1 49 47 -0.7133 0.206 Wang et al. 2006 5 1 54 54 -0.5621 0.1927 Shang et al. 2007 5 1 40 36 -0.478 0.226 Xiao et al. 2008 5 1 30 30 -0.4155 0.2604 Liao et al. 2010 5 1 24 24 -0.3285 0.2247 Zhang et al. 2012 5 1 135 124 -0.5276 0.1525

[MYVIZHI KANNAN]

I tried changing the data format also but it was not working

 
 network convert augmented, large(1000) ref(1) network meta consistency

And i also attached the error what i get