Help with test for tendency of 2 categorical data taking into account the time

[delao victor]

I have a question about how to carry out the following analysis: I need to evaluate if there are differences between 2 drugs (A and B) over time collected through 3 categorical variables: RECIST1 (improvement/no improvement); RECIST2 (improvement/no improvement); RECIST3 (improves/does not improve).

I have believed in doing a prtest, but it only shows me the differences of 1 proportion compared to another. However, my intention is to assess whether over the 3 visits these proportions in RECIST1, RECIST2 and RECIST3 in drug A and B differ (significantly or not) over time.

Here is some information in case it helps:
trat (0/1): 55/53
RECIST1, RECIST2, RECIST3 (0/1)

bysort trat: tab1 RECIST*,m 
                
--------------------------------------------------------------------------------------------------------------------------------------------------------------
-> trat = A

-> tabulation of RECIST1

                  RECIST1 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         43       78.18       78.18
               progresion |          4        7.27       85.45
                        . |          8       14.55      100.00
--------------------------+-----------------------------------
                    Total |         55      100.00

-> tabulation of RECIST2

                  RECIST2 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         39       70.91       70.91
               progresion |          4        7.27       78.18
                        . |         12       21.82      100.00
--------------------------+-----------------------------------
                    Total |         55      100.00

-> tabulation of RECIST3

                  RECIST3 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         28       50.91       50.91
               progresion |          4        7.27       58.18
                        . |         23       41.82      100.00
--------------------------+-----------------------------------
                    Total |         55      100.00

--------------------------------------------------------------------------------------------------------------------------------------------------------------
-> trat = B

-> tabulation of RECIST1

                  RECIST1 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         40       75.47       75.47
               progresion |          6       11.32       86.79
                        . |          7       13.21      100.00
--------------------------+-----------------------------------
                    Total |         53      100.00

-> tabulation of RECIST2

                  RECIST2 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         34       64.15       64.15
               progresion |          9       16.98       81.13
                        . |         10       18.87      100.00
--------------------------+-----------------------------------
                    Total |         53      100.00

-> tabulation of RECIST3

                  RECIST3 |      Freq.     Percent        Cum.
--------------------------+-----------------------------------
respuesta + estabilizcion |         24       45.28       45.28
               progresion |          7       13.21       58.49
                        . |         22       41.51      100.00
--------------------------+-----------------------------------
                    Total |         53      100.00

Thanks!

[Bruce Weaver]

Hello delao victor. If I follow, your data file looks like this:

    trat   RECIST1   RECIST2   RECIST3  
       A         0         0         1  
       A         1         1         0  
       A         0         0         0  
       A         0         0         0  
       A         0         0         1  
       A         0         0         0  
       B         0         0         0  
       B         0         0         0  
       B         0         0         0  
       B         0         0         0  
       B         0         0         1  
etc.

And the 1, 2, and 3 on the RECIST variables refer to 3 time points. If I have understood correctly, I think you need to do the following:

1. Use -reshape- to restructure the file from WIDE to LONG;
2. Use -melogit- or -xtlogit- to estimate a logit model that takes into account the correlated nature of the repeated measures.

You should find helpful examples in the help for those two commands.

[delao victor]

Thank you very much Bruce!
I have converted my database to long format and I have performed a multilevel mixed-effects logistic regression model fitting vis and treatment as interaction (code attached). Then I have performanced a contrasts and linear hypothesis tests after estimation to obtain de pvalue.

melogit recist b1.vis##i.trat || code:
margins , over(trat vis)
contrast trat#vis
contrast trat@vis
contrast r.trat#r.vis