Complex Math Model Could Be Simple Way to Predict Bladder Cancer Recurrence 

Finding more accurate ways to predict the recurrence of cancer would be a boon to doctors and patients, and is something Devin Koestler, PhD, a member of the Cancer Biology Program at The University of Kansas Cancer Center, is hoping to make a reality.

Dr. Koestler, also on staff with the University of Kansas School of Medicine’s Biostatistics Department, is creating a statistical prediction model that could help doctors determine when noninvasive bladder cancer will return. The model is based on a series of clinical and pathological markers as well as biomarkers.

“I’m hoping it will be something that will let clinicians do their jobs more successfully,” explained Dr. Koestler. “There are a lot of different factors that go into predicting recurrence, and human beings might not be the best at making use of that information all at once, where a prediction model can do it easily.”

The model he’s creating, called a nomogram, is already in use to predict cancer recurrence; however, the current process uses easily accessible clinical data such as age, family history, tumor size and location, and cancer stage. Dr. Koestler is taking it up a notch by also using a biomarker called DNA methylation in the construction of his nomogram.

The nomogram combines complex statistics and math to come to a simple conclusion: Approximately when is this person’s bladder cancer likely to return?

Bladder cancer's tricky timeline

Dr. Koestler chose bladder cancer for his nomogram because it has such a high rate of recurrence, even if the cancer hasn’t spread outside the bladder and the tumor is removed surgically. Bladder cancer is the sixth most common cancer (an estimated 74,600 new cases in 2014 alone), yet the long-term survival rate is about 70 percent when the cancer is found before it spreads outside the bladder. Even so, about 80 percent of patients see a recurrence within three years of their initial diagnosis and treatment, according to the National Cancer Institute.

“These patients need to be coming into the doctor more frequently and over a long period of time,” said Dr. Koestler. “That is not only a physical and societal burden, but also a huge economic cost. Bladder cancer is No. 1 in terms of per-patient treatment cost, and it’s because people live a long time and need to be monitored the entire time.”

A potential smartphone app layout in which a bladder cancer nomogram could help doctors easily predict cancer recurrence.

If the nomogram could accurately predict recurrence, low-risk patients wouldn’t need to come into the doctor’s office as often, and high-risk patients could be identified early on for potential clinical trials and receive priority testing.

Dr. Koestler is hoping to add more clinical value to his nomogram by adding DNA methylation biomarkers into the equation. As of now, he doesn’t know of any bladder cancer prediction model using such information.

How gene changes predict recurrence

DNA methylation is a change in nucleotides of DNA that can affect the expression of a gene, should it occur in a key region.  If DNA methylation is occurring in a tumor suppressor gene, which prevents us from developing tumors, then the methylation could silence the gene and potentially set the stage for cancer development.

Though there are at least 27,000 possible locations for DNA methylation to occur, Dr. Koestler and his team narrowed it down to just 11. These were the places his team thought would be most predictive in bladder cancer recurrence.

To determine if adding in the DNA methylation biomarkers made the nomogram a better predictor than just the basic patient and tumor information, Dr. Koestler used what is called a concordance index, which measures how well the prediction model actually works. The index measurements are from a 0.5 to 1, with 0.5 meaning the prediction model is no better than a coin toss.

A previous nomogram for bladder cancer that used only clinical and pathological data – no biomarkers – scored a 0.67. Dr. Koestler also made a nomogram using his data without the biomarkers, and it scored a 0.65. This was encouraging, as both nomograms used the same type of predictors and achieved similar results, he said.

When Dr. Koestler added in the biomarkers, the concordance index jumped up to 0.75, which is considered a statistically significant jump.

“These biomarkers do seem to have value for predicting bladder cancer recurrence above and beyond the traditional clinical and pathological predictors,” said Dr. Koestler. “That was really exciting to us, especially because we only used a small number of the biomarkers available to us.”

So far the preliminary results have been promising, but there are still other biomarkers to be tested within the nomogram to potentially increase its accuracy.

In the future, Dr. Koestler is hoping this prediction tool could be something patients can easily access and use. He would like to develop a mobile app that would let nurses and doctors use the prediction tool right in the room with their patients. All the information about the patient, such as the location of the tumor, where it has spread, whether the patient has pain and other predictors, could be entered immediately. Biomarkers and other lab data could be entered later.

“My ultimate goal is to allow clinicians to make real-time predictions and be able to distill all this data into a simple tool,” said Dr. Koestler.

This fall, Dr. Koestler and his team will apply for another grant to extend their bladder cancer nomogram to include more participants. Testing the nomogram using other populations is vital to validating how well the prediction model works and proving it could be used in a broad clinical setting.

“There’s a perception that [biostatisticians] are all about math and writing code, so it feels good to have this model that could end up contributing to better clinical practice and improved decision making by doctors,” said Dr. Koestler. “It’s a tool that will support a doctor’s judgment or give an alternate point of view.”

Funding sources

  • K-INBRE Recruitment package: “Development of prediction models for bladder cancer recurrence using clinical, pathological, and molecular data”
  • Pilot Award, The University of Kansas Cancer Center: “Development of prediction models for bladder cancer recurrence using clinical, pathological, and molecular data”

Relevant publications

  • Marsit CJ, Koestler DC, Christensen BC, Karagas MR, Houseman EA, Kelsey KT. DNA methylation array analysis identifies profiles of blood-derived DNA methylation associated with bladder cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2011;29(9):1133-9.
  • Wilhelm-Benartzi CS, Koestler DC, Karagas MR, Flanagan JM, Christensen BC, Kelsey KT, et al. Review of processing and analysis methods for DNA methylation array data. British journal of cancer. 2013;109(6):1394-402.
  • Cicek MS, Koestler DC, Fridley BL, Kalli KR, Armasu SM, Larson MC, et al. Epigenome-wide ovarian cancer analysis identifies a methylation profile differentiating clear-cell histology with epigenetic silencing of the HERG K+ channel. Hum Mol Genet. 2013. Epub 2013/04/11.
  • Koestler DC, Christensen B, Karagas MR, Marsit CJ, Langevin SM, Kelsey KT, et al. Blood-based profiles of DNA methylation predict the underlying distribution of cell types: A validation analysis. Epigenetics : official journal of the DNA Methylation Society. 2013;8(8).

Types of Bladder Cancer

  • Transitional Cell Carcinoma
  • The cancer starts in the cells that line the bladder. This is the most common type of bladder cancer.
  • Squamous Cell Carcinoma
    Squamous cells typically form after a long-term infection, and can become cancerous over time. It is rare in the U.S. but more common in parts of the world with particular parasites that cause bladder infections.
  • Adenocarcinoma
  • The cancer cells begin in the glandular cells, which secrete mucus. This is also a rare type of bladder cancer.