Model to Determine Risk of Pancreatic Cancer in Patients With New-Onset Diabetes
- Study type
- Observational study
- Study name
- ENDPAC
- Design
- Retrospective cohort study to develop and validate a risk prediction model (exploratory cohort and independent population-based validation cohort)
- Sample size
- 1561 patients with new-onset diabetes in 4 independent, non-overlapping cohorts; independent validation cohort: 1096 patients
- Source
- Gastroenterology · 2018
In plain words
Pancreatic cancer is usually found late, and it is sometimes accompanied by the sudden appearance of diabetes. Researchers built a model called ENDPAC that uses the age at which diabetes started, the change in weight and the change in blood sugar to estimate the risk of pancreatic cancer in people with newly diagnosed diabetes. This model is not yet a definite diagnostic tool; it only shows who needs closer follow-up. A high score does not mean you have cancer, and recent steroid use, for example, can also raise the score.
Key results
Treatment arms and regimens
No therapeutic intervention (model development study). 64 patients with pancreatic cancer and 192 patients with type 2 diabetes; the three factors most associated with pancreatic cancer (change in weight, change in blood glucose and age at onset of diabetes) were selected to build the ENDPAC model.
No therapeutic intervention. The locked model and its cut-off were validated in an independent population-based cohort of 1096 patients with diabetes.
Source of the regimen details: Abstract of the paper
Doses are given for the information of specialists, as in this study’s protocol, and are not a prescription. Every drug must be prescribed by the treating physician for the individual patient.
Study population
Patients with new-onset diabetes (based on glycaemic status) in the Rochester Epidemiology Project from 1 January 2000 to 31 December 2015. Data were collected at the time of diabetes diagnosis and 12 months before. The exploratory cohort included 64 patients with pancreatic cancer and 192 patients with type 2 diabetes. The validation cohort included 1096 patients with diabetes, 9 of whom developed pancreatic cancer within 3 years of meeting the criteria for new-onset diabetes.
Endpoints
Aim: to develop and validate a model to determine the risk of pancreatic cancer in patients with new-onset diabetes. Measures included the area under the receiver operating characteristic curve (AUC), sensitivity and specificity at the score cut-off, prevalence of pancreatic cancer in patients with a high score and the ability to identify patients more than 6 months before the diagnosis of pancreatic cancer.
Safety and side effects
This study is a risk prediction model, and no safety data or adverse effects of a therapeutic intervention are reported in the abstract. In patients without pancreatic cancer, a high score (false positive) was often related to factors such as recent steroid use or another malignancy.
Clinical significance and place in guidelines
This model may help identify patients with new-onset diabetes who are at higher risk of pancreatic cancer and may aid its early detection. According to the abstract, an independent prospective study is needed for further validation; routine use in clinical practice therefore cannot yet be considered established. Its status in clinical guidelines should be checked separately.
Limitations
The study was retrospective and, as the authors conclude, independent prospective validation is needed. The number of patients with pancreatic cancer in the validation cohort was small (9 patients). A high score in some patients without pancreatic cancer (false positives) was related to factors such as recent steroid use or another malignancy. In the abstract, the figure "9 patients (82%)" in the description of the validation cohort of 1096 patients appears inconsistent and should be checked against the full text of the paper.
This summary was prepared by the editorial team from the published paper; the original paper is the final reference. Read the original abstract on PubMed.