Standard Pathology Tests Outperform Molecular Subtyping in Bladder Cancer Tumor Classification

The deadliness of muscle invasive bladder cancer has made the development of new diagnostic and treatment methods imperative. One strategy involves the use of molecular genetic testing to subtype tumors for aggressiveness. But now a new study published in the Journal of Urology, has found that molecular subtyping of bladder tumors was consistently outperformed by standard tests that pathologists have long used to characterize cancer as low- or high-grade and determine the extent of its invasion into the bladder wall and beyond.

Molecular Subtyping

Molecular subtyping involves the use of RNA sequencing (RNA-Seq) to compile databanks on gene expression and mutations present in a cancer type to find patterns of gene expression. The data are then used to subtype tumors that “pathologically look similar” but are molecularly different. The theory is that molecular subtypes are better equipped to indicate which cancer is more or less aggressive and to help steer treatment options, e.g., whether to use chemotherapy before surgery to remove a cancerous bladder. Originally, there were two subtypes for muscle invasive bladder cancer (there are now six):

• Luminal, which betokens better survival; and
• Basal, which predicts poor prognosis.

An Accidental Discovery

Information obtained from molecular profiling is no doubt helpful in guiding treatment. But the process is also complex and expensive. So, researchers from the Medical College of Georgia at Augusta University set out to find a simpler, cheaper and widely available test that could provide similar insight. They reviewed several sets of data on cancer specimens from patients with muscle invasive bladder cancer, including the one used to determine emerging molecular subtypes. They also had information about the patients’ outcomes.

The researchers planned to base their new subtyping panel test, which they dubbed MCG-1, on the 11 genes they identified as being common in all subtype classification methods. But instead of doing RNA-Seq, which costs several thousand dollars, they used the readily available reverse transcription quantitative PCR method which also looks at gene expression and is actually used to verify RNA-Seq data, but costs less than $10.

First, they looked at their own cohort of 52 bladder cancer patients, of whom 39 had muscle invasive disease. They found that MCG-1 was only 31% to 36% accurate in predicting key diagnostic and prognosis indicators, including:

• Likelihood of metastasis;
• Disease specific survival, i.e., surviving bladder cancer; and
• Overall survival, i.e., survival from all causes of death from the time of cancer diagnosis or start of treatment until the end of the study.

To make up for the relative smallness of their dataset and lack of RNA-Seq use for analysis, the researchers then used three patient datasets from the cancer database ONCOMINE which had more patients, including 151 with muscle invasive bladder cancer, and used RNA-Seq to assess gene expression.

They found that their own MCG-1 test was incapable of predicting disease-specific mortality. On some patients in this dataset, information on response to chemotherapy was available but subtypes could not predict chemotherapy response either, the researchers found.

So, the researchers looked at the dataset that has been used to identify the subtypes, The Cancer Genome Atlas (TCGA), a National Cancer Institute and National Human Genome Research Institute project that includes genetic material for 33 different cancers and routine pathology information on 402 specimens from patients with muscle invasive bladder cancer, including the patients’ overall survival and recurrence-free survival.

“Up until this point, we had been looking at patients that other groups had not looked at,” explains study author Vinata B. Lokeshwar, chair of the College’s Department of Biochemistry and Molecular Biology. In this dataset, MCG-1 predicted overall survival similar to findings reported from subtypes in several high profile publications. “We were intrigued why MCG-1 could not predict anything in our cohort or ONCOMINE dataset but predicted overall survival in the TCGA dataset,” Lokeshwar notes.

So, once more they looked at the 402 patients whose specimens were in the dataset and found that 21 of the patients’ tumors were actually low-grade. When they removed the low-grade cases from the TCGA dataset, MCG-1 accurately predicted essentially nothing, not even overall survival. But after including some patients with low-grade tumors into their own dataset, they found that MCG-1 was now able to predict metastasis and disease specific survival.

All the existing subtypes are categorized as bad or better based on the cancer prognosis, the researchers say. The presence of the low-grade tumors in the classification of subtypes skewed the data to make it look like subtypes were predicting overall survival when really it was the grade of the cancer itself that was predictive.

More Evidence of Problems with Molecular Subtyping

Of course, one study is one study. But the Medical College of Georgia at Augusta University study is consistent with another recent study by investigators at Sweden’s Lund University published in the journal Urologic Oncology finding that subtypes were not associated with cancer-specific survival based on study of 519 patients who had their bladders removed because of bladder cancer.

Takeaway

Use of molecular subtyping to classify tumors is tricky business. One potential problem with subtyping, according to Daley S. Morera, co-author of the Georgia study, is that the inherent heterogeneity in the gene expression of tumors, within both the same tumor type, e.g., bladder cancer, and different parts of the same tumor, makes it difficult to categorize any tumor into a single subtype. Adding to the problem is that the pattern of heterogeneity can change during both tumor growth and treatment. “Just because it’s bladder cancer does not mean it’s the same in all patients,” adds Lokeshwar.