Pre-diagnostic prescription patterns in pancreatic cancer: a retrospective cohort study.

Background Prescribing patterns in primary care could demonstrate early clinical features of cancer and windows of opportunity for timely investigation. Aim Analyse primary care prescription patterns prior to a pancreatic cancer diagnosis. Design and Setting Retrospective cohort study using linked primary care and cancer registry data from patients diagnosed with pancreatic cancer in England between 2011-2018. Method Prescription records registered in the Clinical Practice Research Datalink were analysed in the five years pre-diagnosis. Eight categories of prescriptions which may be used to treat clinical features of pancreatic cancer were included (anti-emetics, anti-reflux medications, insulin, other hypoglycaemic agents, opioids, non-opioid analgesics, neuropathic analgesics and non-steroidal anti-inflammatories). Poisson regression was used to estimate the inflection points for increased prescribing above baseline. Results Among 12,990 patients, 669,287 prescriptions were analysed. Insulin was the least common prescription (7% patients), anti-reflux the most common (53% patients). Insulin prescribing increased 19 months pre-diagnosis (95% confidence interval [CI]=14.2-23.8), rising earlier in females (25 months; 95% CI=17.4-32.5) than males (11 months; 95% CI=5.8-16.2). Prescriptions for other hypoglycaemic agents increased 13 months (95% CI=7.7-18.5), anti-reflux and opioid analgesic prescribing 7 months (95% CI=5.4-8.6 and 4.4-9.6, respectively) and anti-emetics and non-opioid analgesics 5 months (95% CI=2.9-7.1 and 3.2-6.8, respectively) prior to diagnosis. Conclusion The early increase in insulin prescribing suggests tumour-induced type 3c diabetes, highlighting an opportunity for earlier diagnosis in a small proportion of patients. Opportunities for earlier diagnosis through investigation and referral also exist in patients prescribed anti-emetic, anti-reflux, and analgesic medications in primary care.