NEW YORK – Predicting the tumor origin site and detecting gene alterations for cancers of unknown primary (CUP) using next-generation sequencing may help to guide targeted treatment and improve patient outcomes, according to the results of a Phase II clinical trial in Japan.
The non-randomized trial, conducted across 19 sites in Japan, evaluated 97 patients with previously untreated metastatic carcinoma for whom the primary tumor site remained unknown following a standard diagnostic approach. The investigators performed RNA- and targeted DNA-based sequencing on the patients' tumor samples, then used an algorithm to predict the primary tumor sites based on the gene expression and gene alteration data. The results of the study were published in JAMA Oncology on Thursday.
Fifteen different primary tumor types were predicted across the 97 patients, the most common of which were lung, liver, kidney, and colorectal cancer. The most frequent gene alterations detected were in TP53, KRAS, and CDKN2A. Patients then received treatment based on these predictions and their gene alterations, including molecularly targeted therapy. The study's primary aim was to evaluate one-year survival probability. As secondary endpoints, the investigators looked into patients' median overall survival and progression-free survival, as well as their objective response rates and treatment safety, among other factors.
The one-year survival probability of 53 percent met the primary endpoint, according to the team, led by Hidetoshi Hayashi of Kindai University. The median overall survival was 13.7 months, median progression-free survival was 5.2 months, and the objective response rate was 39 percent.
The researchers found that among 68 patients predicted to have more responsive tumor types — lymphoma, non-small cell lung cancer, or colorectal, breast, ovarian, kidney, prostate, or bladder cancer — the median overall survival was 15.7 months, and progression-free survival was 5.5 months. Among 29 patients with cancers deemed less responsive, however, including biliary tract, pancreatic, gastroesophageal, liver, cervical, endometrial, and head and neck cancers, the median overall survival was 11 months and the median progression-free survival was 2.8 months.
The treatments that patients received varied based on their predicted tumor types and detected gene alterations, though the researchers pointed out that the patients' physicians were ultimately the ones to decide upon the treatment. This may have resulted in some variability that could have affected study results, they wrote. For example, though several patients found to have EGFR mutations were treated with EGFR-inhibitors such as afatinib (Boehringer Ingelheim's Gilotrif), some patients with specific EGFR mutation types deemed uncommon or non-activating were treated with chemotherapy rather than targeted treatment. This potential variability, the authors wrote, should be considered a limitation of the study, as should the fact that only approved and reimbursed targeted drugs were allowed to be administered. As a result, no one on the trial received immune checkpoint inhibitors, which were not approved for the predicted tumor types at the time. The non-randomized design of the study may also be a limitation, they wrote.
Nonetheless, the study's findings — specifically that NGS-based RNA and DNA profiling may allow for tumor type prediction and selection of effective targeted treatments for patients with CUP — were "suggestive of the clinical use of this strategy," the authors noted.
Interestingly, in a randomized study conducted in patients with CUP prior to this trial, the use of microarray-based gene expression analysis to determine site-specific therapy was not found to improve outcomes versus empirical chemotherapy. This, the researchers theorized, may have been due to the "insufficient predictive ability of microarray-based expression profiling." The fact that 20 percent of patients in the previous study were predicted to have lymphoma — a cancer type that is not common in studies of CUP, and was only predicted for a single patient in the current study — supported this hypothesis and bolstered the case for NGS versus microarray-based expression profiling.
Going forward, the researchers concluded, randomized studies will be needed to compare the efficacy of the NGS-based approach to empirical chemotherapy.