My data are log2 fold changes (FCs) between control and treatment usually coming from triple replicated affymetrix microarrays. I also have p-values for those data determined by t-test. As I currently have only those data available and not the replicates themselves, I wonder if it is sensible to use the FCs as input for the limma function lmFit?

The results I get seem reasonable but I have found no way to include the p-value information which makes me uneasy about the reliability. Is this possible in some way or is my only solution to obtain the raw replicate data?

Thank you for your insights.

Your best option would be to obtain the raw data (i.e., the raw intensities from each of the replicated array) and use limma on that. The log-fold change between control and treatment for each gene/probe does not provide any information about the variability of expression across replicates. In fact, it's not obvious to me how you would do the limma analysis at all, if you were to give it only a single log-fold change for each gene.