The objective function value for the ILP problem can never

be as good as the optimal solution to its LP relaxation $($variables need not be integers$).$

be as poor as the optimal solution to its LP relaxation $($variables need not be integers$).$

be worse than the optimal solution to its LP relaxation $($variables need not be integers$).$

be better than the optimal solution to its LP relaxation $($variables need not be integers$).$