NEW YORK – Pancreatic cancer risk prediction models based on clinical factors may be refined and improved by bringing in other types of data, including germline genetic risk variant and blood-based biomarker data, according to an international team led by investigators at the Harvard T.H. Chan School and Public Health and the Dana-Farber Cancer Institute.
"Like most cancers, pancreatic cancer is multifactorial," co-senior author Peter Kraft, an epidemiology researcher at Harvard, said in a statement. "The more we are able to combine information from multiple domains, the better we will become at identifying those who could benefit from screening."
Using a nested case-control approach, the researchers brought together genetic, clinical, and other data for 500 individuals with primary pancreatic adenocarcinoma and 1,091 matched control individuals, enrolled through four prospective studies, using these data to develop several pancreatic cancer risk models based on clinical data alone, genetic and clinical factors, or genetic, clinical, and biomarker data.
When the team applied those models to almost 1,000 more cases or controls followed for up to 10 years, it found that the integrated model provided the most accurate pancreatic risk prediction, surpassing risk estimates done using clinical data without genetic or blood biomarker insights. The results were reported in the journal Cancer Epidemiology, Biomarkers, and Prevention on Wednesday.
"The final integrated model has improved risk discrimination over those that include clinical factors alone and successfully identify a small segment of the general population at elevated risk of pancreatic cancer," the authors wrote. "Further refinement and validation in independent samples will be necessary to make these models clinically actionable and impact survival of patient with pancreatic cancer."
The vast majority of individuals with pancreatic cancer are not diagnosed until they have reached an advanced form of the disease, the team noted, which is thought to contribute to the high mortality of the disease and poor outcomes that are often seen in pancreatic cancer patients.
In an effort to come up with more accurate risk models to help find at-risk individuals at earlier stages of the disease, the investigators came up with models based on clinical, genetic, and other potential risk factors, comparing them with more conventional models based exclusively on clinical factors such as body mass index or past diabetes diagnoses.
"These factors have been investigated individually," Kraft explained, "and in this study, we wanted to examine the combined effect of clinical factors, common genetic predisposition variants, and circulating biomarkers."
Along with clinical data, medical comorbidities, and lifestyle factors such as smoking or alcohol use, their "absolute risk models" included blood plasma biomarkers related to insulin resistance and other processes implicated in pancreatic cancer, the researchers noted, as well as susceptibility loci identified through past genome-wide association studies.
"[B]ecause all of our subjects were enrolled in prospective cohorts, all risk factor data and circulating markers were measured before the cases' diagnosis of pancreatic cancer," the authors wrote, noting that their study design "faithfully recapitulates the situation faced by primary care physicians, where decisions related to disease screening are made in the prediagnostic setting using data collected in the several years prior to cancer diagnosis."
The team found that the individuals who scored in the top percentile on models that combined genetic risk factors, circulating biomarkers, clinical data had a 4 percent risk of developing pancreatic cancer, on average, before reaching 80 years old.
The absolute risk model also provided a look at the individuals whose pancreatic cancer risk in the coming decade was jacked up at least three-fold compared to the average population risk, the researchers noted, including 3.7 percent of the men and 2.6 percent of women assessed.
"The final integrated model identified a subset of approximately 2 percent of individuals who have three-fold higher risk than the average in the general US population," the authors suggested, adding that "individuals with the top 1 percent of pancreatic risk as determined by the final integrated model carried a 4 percent lifetime risk of pancreatic cancer and a 2 percent 10-year risk at age 70 years."
In contrast, models based on clinical data or clinical and genetic data without blood biomarkers clues flagged fewer at-risk men or women, the team found, though it noted that the current models require additional "refinement and validation in independent samples" before they can be considered clinically actionable.