New examination of prostate cancer data identifies ‘detection bias’ in prostate studies
Population studies that identify the causes of disease—known as epidemiological studies—are crucial for pinpointing cancer risks and advancing public health policy to combat them. But it’s also crucial that researchers understand that who does—and does not—get screened for cancer may color our perception of risk factors.
For instance, if overweight men are screened less or get fewer biopsies, they may seem to have less prostate cancer. But does that truly give us a clear picture of an overweight man’s risk for the disease?
A new study led by public health researchers at Fred Hutchinson Cancer Research Center suggests variation in biopsy patterns among screened populations may have influenced how we measure prostate cancer risks.
Fred Hutch biostatistician Catherine Tangen, lead author of a study published last fall in the Journal of Clinical Oncology, said she and colleagues looked at two large prevention trials and found discrepancies between known risk factors—which have not changed—and who actually got the disease.
“We assumed that prostate cancers are diagnosed uniformly, but that’s not true. We found a lot of variation,” she said. “Risk and reality often didn’t line up. As a public health researcher, this concerns me. Bias can warp our understanding of the prostate cancer disease process—and misdirect our efforts to test new risk factors and prevention strategies in future research.”
It’s all about individual versus population-level risk, she said. It’s key for a man and his doctor to look at individual risk factors—age, weight, race, general health, family history of the disease—and make a personalized decision about screening and/or biopsy. A clear understanding of risk factors can help men make more informed decisions.
“But at a population level, bias can lead us to screen and biopsy men who are at low risk and lead us to fail to screen and biopsy men who are at higher risk,” she said. “In research, bias can cause us to waste precious time and money investigating the wrong risk factors.”
Funded by the National Cancer Institute and conducted by the cancer clinical trials network SWOG, the study drilled down into data from two of the largest prostate cancer prevention trials ever conducted in the U.S., PCPT and SELECT. Tangen partnered with fellow Fred Hutch researchers Phyllis Goodman, Cathee Till and epidemiologist Jeannette Schenk as well as researchers from the University of Texas Health Science Center at San Antonio and the University of Colorado Denver School of Medicine for the SWOG study. SWOG’s Statistical Center is housed at Fred Hutch.
The Prostate Cancer Prevention Trial, or PCPT, was conducted between 1994 and 1997 and offered researchers an unusual edge because most participants in the trial received end-of-study biopsies. The uniform biopsy provided unbiased data on actual cancer risk and was considered a “gold standard.”
“We had this great opportunity with the PCPT study where we actually biopsied lots of men regardless of what their PSA looked like, men from all walks of life,” Tangen said. “We have an estimate of what the true association was with different risk factors for prostate cancer.”
The Selenium and Vitamin E Chemoprevention Trial, conducted between 2001 and 2008, operated a little differently. Men in this trial received PSA testing and digital rectal exams based on local medical standards. Biopsies were done at the discretion of the doctor and based on the men’s own preferences. These men did not receive automatic end-of-study biopsies as with the other cohort.
“With SELECT, we followed the men and some had PSAs and some didn’t. Some got biopsies and some didn’t,” Tangen said. “It was more like what happens in the general community.”
Using data from the prevention trials’ placebo arms (men who received no preventive intervention), researchers created two cohorts totaling 17,000-plus men, including more than 2,200 who were actually diagnosed with prostate cancer during the course of the trials. And they found that among men with the same PSA values, certain men underwent biopsies more than others. Younger, healthier, better educated men; married men; and those with a family history of prostate cancer were more likely to get biopsies, at least in the U.S. Overweight men, men who smoked and men who were diabetic were less likely to get a biopsy.
The investigators then evaluated the associations of risk factors with prostate cancer in SELECT (reflecting treatment of the general population) and then in PCPT (the gold standard) and checked to see if the results were consistent.
They found many discrepancies. Men who took statin drugs in the SELECT trial, for example, had a significantly reduced risk of prostate cancer. But when they looked at statin use in the PCPT trial, there was no association. They concluded that individual doctor and patient biopsy decisions probably accounted for these very different results. On the flip side, researchers confirmed that men with a family history of prostate cancer and men with elevated PSA levels— both proven risk factors—did indeed develop cancer at higher rates than their peers. However, that association wasn’t as strong in PCPT when the “biopsy bias” was reduced.
“We don’t know what forces are at work behind the scenes,” said Tangen. “Do people who get statins have a better relationship with their doctor so they’re getting screened more for prostate cancer? Or do people on statins tend to be sicker and have higher BMI and their doctor is less likely to recommend a biopsy? We scratched the surface and showed that there are a lot of factors that impact whether someone gets a prostate biopsy. But how all of those factors work together for a given study, we don’t know.”
Tangen said these discrepancies are important because they influence screening recommendations for the general population.
“The research community sometimes makes faulty conclusions about prostate cancer risk—assumptions that may lead to flawed screening decisions and incorrect choices about research directions,” she said. “Detection bias is introduced when screening information is not acted upon uniformly.”
Fred Hutch biostatistician Ruth Etzioni, who was not involved with the report, said the findings are probably more important for scientists than for the general public.
“The results have implications for researchers who do these kinds of studies, but the implications for the public are not as clear,” she said. “Still, the public can learn from this why different studies get different results. It can be confusing sometimes because one study will say one thing and another study will say another thing. That’s because they’re done in different populations and the screening and biopsy patterns can really vary.”
Family history, for instance, is a big risk factor for prostate cancer, and most men with a family history of the disease know this.
“So people with a family history are more likely to comply,” she said. “If they get a positive PSA test, they’re going to get a biopsy. They’ll think, ‘My brother died of this. I need to go.'”
The problem is, a cohort full of these types of men will skew the outcome.
“It will make the risk factor of family history look more important,” she said. “That’s the bias. If the risk factors are high, they’ll look even higher with this cohort because they’re more likely to have a biopsy.”
Both Tangen and co-author Ian Thompson of the University of Texas Health Science Center said it’s all about getting to the truth.
“In medicine, we don’t want to do the wrong thing and in research, we don’t want to look in the wrong places,” said Thompson. “Our work shows we may be doing both in prostate cancer. So the message is this: Risk factors for cancer are very complex. Before we all leap to conclusions, we need to collect very detailed research data on who is screened and why and be very rigorous in our analysis of that data and what it might mean for clinical care.”
Tangen agreed, adding this was particularly important since screening also picks up indolent, asymptomatic cancers which may not require treatment.
“In cancer research, we all want the truth,” she said. “But we’ve found that we may just have to work harder to find it. There is a lot of noise out there.”
