Can androgen deprivation with leuprolide be predicted from histology alone? If not, why not?

Androgen deprivation therapy with analogues of luteinizing hormone-releasing hormone produces distinctive histological changes in neoplastic and nonneoplastic prostate tissue. Others have described these features in cases in which treatment status was known. To our knowledge the specificity and sensitivity of luteinizing hormone-releasing hormone effects based only on histology and possible reasons for interobserver variation have not been addressed previously.

Slides from 97 prostatectomies performed in a 3-year period were reviewed by 2 observers blinded to knowledge of previous hormonal treatment. The observers evaluated each case, recording the presence or absence of 14 features associated with androgen deprivation therapy, and then made an overall assessment regarding treatment status.

Of the 97 patients 31 had received androgen deprivation therapy with the luteinizing hormone-releasing hormone agonist leuprolide. A luteinizing hormone-releasing hormone effect was identified by the 2 observers in 26 and 28, respectively, of the 31 cases (83.9 versus 90.3% sensitivity and 92.4 versus 80.3% specificity). Including focal changes as consistent with androgen deprivation therapy increased sensitivity but decreased specificity. Of the 14 features 12 had a significant association for predicting treatment status.

Interobserver variations in interpretation occurred although both examiners were experienced in the evaluation of luteinizing hormone-releasing hormone effects and they used exactly the same criteria. These variations were apparently due to differences in the value (mental weight) given by each observer to the features assessed in each case. Predicting treatment status was optimized by noting a luteinizing hormone-releasing hormone effect only when changes were diffuse, improving specificity significantly with only a modest decrease in sensitivity.