combining results obtained from 3 different methods

I have received data tables where 3 methods have been used on 54 different variables.
Method 1 (M1): multivariate model of logistic regression, to see if the 54 variables influence the outcome (only 24 do)
Method 2 (M2): a sampling approach
Method 3 (M3): a circular permutation approach

I have p-values from method 1, and odds ratios (OR) from methods 2 and 3 for each variable.
I'd like help in identifying the most important variables in each method and integrating that result to the most important variables across all three methods.
So far, I have thought of ranking them by respective ORs and P-values. For M1 I plan to rank by p-value (lowest pval gets lowest rank). For M2 and M3, the highest OR will get the lowest rank and so on. Then I plan on taking the rank sum or average rank sum and then choosing the variables within the top 50th percentile as the key variables represented by all three methods.
Does this make sense or can any one suggest a better way?