Ideally, you would run a standard curve on the same ELISA plate using known concentrations of your protein of interest. Then you would read off the optical density (OD) readings of your unknown samples using your standard curve to obtain your protein concentration. Failing that, you could also normalise your treated samples to an untreated control, and express your OD values as a ratio/pecentage of the untreated control. Ultimately, you would want to convert the OD readings obtained from your plate reader to a protein concentration using a standard curve, or normalising to an untreated control would allow you to see increases/decreases in protein level relative to your control.
It depends on your goal. If it is a quantitative ELISA you need a standard curve to calculate the concentration of your samples. If not you can establish a cut-off value to clasify the samples as positive or negatives. Alternatively, you could compare the relative absorbance values.