- Design
- Retrospective cohort, external validation in a national primary care network
- Population
- 70,050 adults aged 50 or over with new-onset diabetes in England
- Primary outcome
- Discrimination and accuracy of ENDPAC for pancreatic cancer
- Effect
- AUC 0.734; at ≥3: sensitivity 62.6%, specificity 72.5%, PPV 0.6%
ENDPAC combines weight change, change in glycaemic control and age at diabetes onset to flag new-onset diabetes that may be caused by pancreatic cancer. This BJGP study validated it in an English primary care sentinel network, in 70,050 adults aged 50 or over with new-onset diabetes.
185 (0.26%) went on to develop pancreatic cancer; they were older, had higher HbA1c and more weight loss at diagnosis. The score's area under the curve was 0.73. A cut-off of 3 or more labelled 27.8% of patients high risk, catching 62.6% of cancers, with a specificity of 72.5%, positive predictive value of 0.6% and negative predictive value of 99.9%.
The positive predictive value is too low to send everyone flagged for imaging. Its value is as a first filter in a stepped pathway, and as a prompt: diabetes that arrives after 50 with weight loss rather than weight gain is not the usual type 2 story.
- Ask about weight loss in anyone over 50 diagnosed with new diabetes.
- Be suspicious when new diabetes comes with weight loss rather than gain.
- Look for abdominal or back pain, jaundice or steatorrhoea alongside new-onset diabetes.
- Do not image every flagged patient; the positive predictive value is under 1%.
Why it matters
New diabetes is common and pancreatic cancer is rare, so primary care needs a way to tell the unusual case apart.
Don't overread it
A validation of a risk score, not evidence that using it saves lives.
The statistics, in plain English
An area under the curve of 0.73 means the score ranks a future cancer above a non-cancer patient about three times in four. A positive predictive value of 0.6% means about 1 in 170 flagged patients will have cancer. The 99.9% negative predictive value mostly reflects how rare the cancer is.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for family medicine, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free