Network meta-analysis: trials with two groups with same vaccine

[Joaquin Baruch]

Hello all,

Situation: I'm running network meta-analysis and I have trials that have two arms that use the same vaccine. An example would be a trial with a control group, a "vaccine A" group, another "vaccine A" group, and a "vaccine B" group.

Problem: when I run the network setup command, I get an error message that indicates that my study variable and my treatment variable don't uniquely identify observations.

Question: Should I add the two groups of "vaccine A" together? that sounds like losing power, but do I have another option?

Code: network setup number_sick number_animals, studyvar (Study) trtvar (vaccine_) or

Error: variables Study vaccine_ do not uniquely identify the observations

Thank you,
Joaquin

Data (just as an example):

Study	number_animals	number_sick	vaccine_
1	48	20	Control
1	48	16	A
1	48	16	A
1	126	1	B
2	107	1	Control
2	107	5	B
2	107	1	C
2	107	4	D

[William Lisowski]

How do the two groups of Study 1 who both receive vaccine A differ?

If combining the two groups would "lose power" then it seems to me you could divide them into 32 groups of 1 apiece and "gain power".

[Joaquin Baruch]

Hi William,

The two groups of vaccine A within study 1 differ on things that the researcher was interested to evaluate but we chose to ignore (i.e., vaccinated on day 1 or vaccinated on day 3).

Should I add them together then?

Thank you for your response,
Joaquin

[William Lisowski]

Yes, you should combine groups within a study that are - for the purposes of your analysis - treated identically. The example code below shows how to have Stata do this for you using the collapse command.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte study int number_animals byte number_sick str7 vaccine_
1  48 20 "Control"
1  48 16 "A"      
1  48 16 "A"      
1 126  1 "B"      
2 107  1 "Control"
2 107  5 "B"      
2 107  1 "C"      
2 107  4 "D"      
end

collapse (sum) number_animals number_sick, by(study vaccine_)
list, abbreviate(20) sepby(study)

. list, abbreviate(20) sepby(study)

     +-------------------------------------------------+
     | study   vaccine_   number_animals   number_sick |
     |-------------------------------------------------|
  1. |     1          A               96            32 |
  2. |     1          B              126             1 |
  3. |     1    Control               48            20 |
     |-------------------------------------------------|
  4. |     2          B              107             5 |
  5. |     2          C              107             1 |
  6. |     2    Control              107             1 |
  7. |     2          D              107             4 |
     +-------------------------------------------------+

[Joaquin Baruch]

Thank you for the information and the code provided, I really appreciate it! I will make sure to use dataextnext time.

Thank you,
Joaquin