Background: Earlier detection of pancreatic ductal adenocarcinoma (PDAC) with clinically meaningful lead time (>4 months) can improve survival. Retrospective studies show increased 3-year risk of PDAC in new-onset diabetes (NOD), with highest risk being in the first year; these findings need validation in prospective studies. Clinical diagnosis of NOD is delayed >1 year in 30% of those with glycemic evidence of NOD (G-NOD). Real time G-NOD identification will require an automated approach using passive surveillance of glycemic and other electronic health record (EHR) parameters.
METHODS: G-NOD was defined as elevation in glycated hemoglobin >6.5% or blood glucose values meeting American Diabetes Association criteria for diabetes with confirmed lack of such elevations or treatment for diabetes in the previous 18 months. For real time identification of G-NOD in subjects >50 years of age we developed an algorithm using EHR parameters. We validated it at 23 sites across the United States. Subjects identified at the four largest participating sites between 09/2018 to 05/2022 form the Recent Elevation in Glucose and Recent Diabetes (REGARD) Cohort which is being followed prospectively for 3 years for development of PDAC. In this interim analysis, we report the standardized incidence ratio (SIR) (ratio of observed to expected PDAC [1 in 2500 person years (PY) of follow-up]). We also report the 2-year incidence of PDAC, the proportion with clinical diagnosis of PDAC >4-month after G-NOD date, and the effect on PDAC risk estimation of a 6-month delay in NOD diagnosis.
RESULTS: Of 18,944 G-NOD subjects in the REGARD Cohort (51.0% male, mean age 63.1±8.4 years), 82 (0.43%) have been diagnosed with PDAC (59.8% male, mean age, 69.8±8.3 years) during a median follow-up of 2.3 years. The SIR for PDAC in Non-Hispanic Whites (n= 6,567) was 8.2 (95% CI: 5.8 - 10.7, 2-year incidence 0.7%). SIR in the remaining 12,377 subjects was 3.1 (95% CI: 2.2-4.4, 2-year incidence 0.3%). The risk in Hispanics and African Americans was significantly lower than in non-Hispanic Whites (p=0.0025 and 0.0170, respectively) (Table). 65% of PDAC were diagnosed >4 months after G-NOD date. A 6-month delay in NOD diagnosis would lead to underestimation of PDAC risk from 4.9-fold to 3.2-fold (difference = 1.8-fold (95% CI: 0.96-2.54).
SUMMARY AND CONCLUSIONS: An algorithm for passive surveillance of electronic health records identifies G-NOD in real time. G-NOD is associated with increased risk of PDAC with highest risk in non-Hispanic Whites (~8-fold higher risk, 2-year PDAC incidence 0.7%). Clinically meaningful lead time from G-NOD date to PDAC diagnosis occurs in 65% of subjects. A delay in NOD diagnosis would lead to significatly diminished ability to identify PDAC cases. Electronic identification and prompt clinical work up of G-NOD could lead to earlier detection of PDAC.

Table:<br /> Incidence of PDAC in REGARD Cohort by Time (months) from G-NOD date in White non-Hispanics (n=6567), Hispanics (n=6015), African Americans (n=2200) and All REGARD Subjects (n=18944)

Table:
Incidence of PDAC in REGARD Cohort by Time (months) from G-NOD date in White non-Hispanics (n=6567), Hispanics (n=6015), African Americans (n=2200) and All REGARD Subjects (n=18944)

Ying-Qi Zhao

Appearances