Recent hyperglycaemia helps detect pancreatic cancer early

Pancreatic cancer may be identified early if assessment will be steered towards the point of recent hyperglycaemia based on elevated HbA1c, suggests a retrospective cohort study, noting that such early detection may then improve survival in cancer patients.

“New-onset diabetes (NOD) has been suggested as an early indicator of pancreatic cancer,” the researchers said. “However, the definition of NOD by the American Diabetes Association requires two simultaneous or consecutive elevated glycaemic measures.”

On this note, the research team applied a machine-learning approach using electronic health records to predict the risk of pancreatic cancer in patients with recent-onset hyperglycaemia. They identified a total of 109,266 health plan enrolees (mean age 63.6 years) who had an elevated HbA1c (6.5 percent) tested in January 2010 to September 2018 with recent-onset hyperglycaemia.

The researchers then extracted 102 potential predictors. They also used generated 10 imputation datasets to handle the missing data. Risk models were developed and validated using the random survival forests approach. Finally, performance was assessed via c-index, calibration plot, sensitivity, specificity, and positive predictive value.

The incidence rate at 3 years was 1.4 (95 percent confidence interval, 1.3‒1.6) per 1,000 person-years of follow-up. Out of the 50 training samples, the three models which included age, weight change in 1 year, HbA1c, and one of the three variables (HbA1c change in 1 year, HbA1c in the prior 6 months, or HbA1c in the prior 18 months) appeared most often.

The c-indexes ranged from 0.81 to 0.82. In patients who had the top 20 percent of the predicted risks, sensitivity was 56 percent to 60 percent, specificity 80 percent, and positive predictive value 2.5 percent to 2.6 percent.