Correlation of tumor-profiling results with drug selection in ovarian cancer patients

 
Citation:

J Clin Oncol 30, 2012 (suppl; abstr e15555)
Author(s):
Laura K. Shawver, Deborah A. Zajchowski; Clearity Foundation, San Diego, CA

Background:

With the aim of informing therapy-selection decisions for ovarian cancer patients, we have profiled tumors from 190 ovarian cancer patients since 2008 for expression of 25 biomarkers associated with drug response. For 48 patients with mature follow-up data, we asked how frequently the drugs designated as likely to be of clinical benefit based on the tumor profile were administered.

Methods:

The tumor profiles and therapeutic choices of 48 patients that received a median of 2 (range 1-9) prior therapies for recurrent disease were extracted from our privacy-protected database. Protein levels of drug sensitivity (ER/PR, hormonal agents; TOPO1, topotecan/irinotecan (T-I); TOP2A, liposomal doxorubicin/etoposide (PLD-E); SPARC, nab-paclitaxel, N-P) and resistance (TS, pemetrexed/capecitabine (PC); RRM1, gemcitabine (G)) markers in each tumor were compared to expression in the ovarian cancer population to assign drugs likely to be associated with clinical benefit (values = 25th percentile: sensitivity; = 25th percentile: resistance).

Results:

A median of two drugs (range 1-5) were prioritized for each patient based on marker expression. 38 patients received one of the 6 drug classes evaluated by the profile. Of those 38, 22 (58%) received a therapy that correlated with the profile and 16 (42%) received drugs that were not prioritized by the profile. G and PLD were most frequently administered both before (68%) and after (48%) profiling but were supported by the tumor profile in only 40% (4/10: G) and 30% (3/10: PLD-E) of the cases. These frequencies are similar to the patient fraction (26-34%) in which those drugs were each prioritized by profiling, suggesting that the profile is not influencing therapy choice. In contrast, 11 of the 14 patients who received N-P, T-I, or hormonal agents had tumor profiles that prioritized those agents, a frequency of profile match to the treatment received that is higher than expected based on the patient fraction assigned to each therapy.

Conclusions:

These results require confirmation in a much larger study, but suggest that tumor profiling may guide clinical decision-making for some therapies, although frequently used drugs are often given regardless of the tumor profiling evidence.