Genetic Profiling Disappoints for CUP Survival

One-year survival did not improve with treatment based on comprehensive gene-expression profiling in metastatic cancers of unknown primary site (CUP), Japanese researchers found.

Compared with empirical carboplatin/paclitaxel combination therapy, which is the usual treatment for CUP, regimens based on genetic site prediction led to a 1-year survival rate of just 44.0% versus 54.9% for empiric treatment in the randomized phase II trial.

Predicting the tumor’s original site via molecular profiling, however, may be useful at the prognostic level, stated Kazuhiko Nakagawa, MD, PhD, of Kindai University in Osaka, and colleagues. “The observation that patients predicted to have more responsive tumor types had a better OS [overall survival] and PFS [progression-free survival] than those with less-responsive tumor types suggests that prediction of original tumor site has prognostic value.”

In the study, published online in the Journal of Clinical Oncology, a post-hoc analysis showed that the overall median OS and PFS rates were 9.81 and 5.8 months for the site-specific group versus 12.5 and 4.8 months for the empirical group (P=0.896 and 0.550, respectively). But the median OS of 16.7 versus 10.6 months (P=0.116) and median PFS of 5.5 versus 3.9 months (P=0.018) were better for tumor types predicted by profiling to be more rather than less responsive to treatment, the researchers reported.

Nakagawa and co-authors noted that tumors considered to be relatively sensitive to treatment include colorectal, breast, ovarian, kidney, prostate, bladder, germ cell, and non-small-cell lung, as well as lymphoma. There was some imbalance in the number of responsive-type cancers between the two trial arms, with, for example, 17 lymphomas and five ovarian cancers in the site-specific group versus nine and one, respectively, in the empiric group.

Although the survival advantage observed for more responsive tumors did not translate to an OS advantage with site-specific therapy, the results paralleled those of a 2013 study by John Hainsworth and colleagues, in which patients with CUP predicted to have more responsive tumor types had longer survival compared with patients with less responsive tumor types, leading those researchers to recommend molecular tumor profiling as part of standard patient evaluation. Since tumor type affects treatment outcomes, Hainsworth and co-authors also recommended that predicted tumor type be considered as a stratification factor in trials.

Study Details

For the new study, Nakagawa and colleagues randomized 130 demographically similar patients at 14 Japanese centers from October 2008 to February 2015. The median ages in the site-specific and empirical groups were 67 and 65, and 58.5% in both arms were male.

About 40% in either group had metastases only in the lymph nodes. All had tissue sent for genome-wide molecular analysis for tumor type but not for targetable genetic mutations. A total of 50 patients were assigned to site-specific therapy and 51 to empirical therapy, with 15 and 14 patients, respectively, ultimately not receiving site-specific or empirical therapy. Data were collected to May 2017.

An algorithm based on microarray analysis successfully predicted a tissue of origin for all patients, with 16 different sites predicted. Identified cancers included pancreatic, gastric, lymphoma, urothelium, cervix, ovary, biliary tract, head and neck, endometrium, breast, liver, non-small-cell lung, kidney, prostate, and thyroid. The most common cancers were pancreatic (n=27, 20.8%), gastric (n=27, 20.8%), and lymphoma (n=26, 20%), followed by urothelium (n=8,6.2%), cervix (n=7, 5.4%), and ovary (n=6, 4/6%). The least common type was thyroid cancer (n=1, 0.8%).

The authors noted that even with the growing number of effective targeted molecular therapies, the gap in treatment advances for CUP versus cancers with a known primary site is widening. “Improvement in the prognosis of several types of cancer have the potential to also contribute to better outcomes of site-prediction therapy for patients with CUP, but additional studies of site-specific therapy that are based on profiling of gene expression or genetic alterations with a larger number of patients are needed,” the researchers wrote.

Asked for his perspective on the study, Charles, L. Shapiro MD of the Tisch Cancer Institute at the Icahn School of Medicine at Mount Sinai, in New York City, who was not involved in the research, said the results show that chemotherapies for solid tumors “are marginal at best with abysmal outcomes, but that paclitaxel-carboplatinum is a pretty good cocktail and shows activity for a variety of solid tumors.”

Noting that CUP is present in very few cancer diagnoses (about 2%), Shapiro said the study results suggest that “since genomic profiling is of uncertain predictive value, although it has some prognostic value, you could go straight to an empirical cocktail without profiling and not be disadvantaged.” He also noted that the study’s microarray analysis was designed only to identify a large range of tumor types, but not specific genetic abnormalities that might be targeted with precision therapy.

Nakagawa and co-authors noted several study limitations, including the relatively small size and the loss of 29 patients after randomization and treatment initiation. With only 101 of 130 patients evaluated for efficacy, the study fell short of the 114 targeted for statistical analysis. Another limitation was the imbalance in responsive versus non-responsive tumor types between the two arms.

last updated 01.24.2019