You should not blindly use principal components (PCs) without justification. For example, what justification do you have for using 20 PCs? - most of those PCs may be meaningless...
You should explore whether or not any PCs actually stratify your cohort in a way that could confound your results, and then include the PCs that are likely to confound. This may likely be just PC1, PC2, and PC3. You can explore this via bi-plots, like here: Produce PCA bi-plot for 1000 Genomes Phase III - Version 2
PCs are used as covariates to adjust for population stratification. You may not even have to use them in your cohort.