Archived tumor specimens, particularly those collected by large cooperative groups and trials, provide a wealth of material for post hoc clinical investigation. As these tissues are rigorously collected and preserved for many decades, subsequent use of the specimens to answer clinical questions must rely on the assumption that expression and detection of target biomarkers are not degraded with time. To test this assumption, we measured the expression of estrogen receptor (ER), human epidermal growth receptor 2 (HER2), and Ki67 in human breast carcinoma using quantitative immunofluorescence (QIF) in a series of formalin-fixed paraffin-embedded (FFPE) tissues from 1295 individual patients preserved for 7 to 53 years in four cohorts on tissue microarrays. Protein expression was measured using the automated quantitative analysis method for QIF. Change in quantitative protein expression over time was estimated in positive cases using both Pearson's correlation and a polynomial regression analysis with a random effects model. The average signal decreased with preservation time for all biomarkers measured. For ER and HER2, there was an average of 10% signal loss after 9.9 years and 8.5 years, respectively, compared with the most recent tissue. Detection of Ki67 expression was lost more rapidly, with 10% signal loss in just 4.5 years. Overall, these results demonstrate the need for adjustment of tissue age when studying FFPE biospecimens. The rate of antigenicity loss is biomarker specific and should be considered as an important variable for studies using archived tissues.