Testing the role of P-glycoprotein expression in clinical trials: applying pharmacological principles and best methods for detection together with good clinical trials methodology.

P-gp (Pgp) is a cell surface ATPase which confers resistance to many of the most active chemotherapy drugs, including taxol, doxorubicin, and vinca alkaloids. Pgp can be detected in human cancers by immunohistochemistry, RNA probes, or by functional assays utilizing transported fluorescent dyes such as rhodamine. The expression of Pgp in untreated human cancers is highly variable, being almost universal in colon, hepatocellular carcinoma, and renal cell cancers, less common in breast, ovarian, and lymphoid malignancies. At least part of the heterogeneity is attributable to different definitions of positivity even with a given method of detection. In chemotherapy naive cancers, resistant cells may not occur very frequently. Whilst the Goldie-Coldman hypothesis predicts treatment failure if 1 cell in 10(6) expresses a resistance mechanism, no method of detection yet described can reliably achieve this. The field has reached a stage in which it may be possible to detect Pgp accurately in advanced cancers which have failed chemotherapy allowing phase II clinical trials to be performed in Pgp-positive tumors. In terms of which Pgp inhibitors are selected for clinical study it is likely that selection of Pgp inhibitors with nM potency to bind to Pgp will be important. Such drugs should undergo extensive phase I trial evaluation to assess pharmacokinetic interactions with a range of cytotoxic drugs before entering randomized trials. In randomized clinical trials Pgp detection may be less important, as disease-free survival and overall survival would be the key end-points, but the Pgp positivity of relapsed disease would indicate if treatment with inhibitors of Pgp-eliminated Pgp-expressing clones. The accurate detection of Pgp in human cancers is being refined and will be an essential component of future Pgp inhibitor clinical trials. Finally, these trails must be of sufficient size (> 500 patients per arm) to reliably detect clinically meaningful differences.