BACKGROUND: Pancreatic cancer is the tenth most common cancer in the UK; however, outcomes are poor, in part due to late diagnosis. We aimed to identify symptoms and other clinical and sociodemographic factors associated with pancreatic cancer diagnosis and diagnostic intervals. METHODS: We did this prospective cohort study at seven hospitals in two regions in England. We recruited participants aged 40 years or older who were referred for suspicion of pancreatic cancer. Data were collected by use of a patient questionnaire and primary care and hospital records. Descriptive and regression analyses were done to examine associations between symptoms and patient factors with the total diagnostic interval (time from onset of the first symptom to the date of diagnosis), comprising patient interval (time from first symptom to first presentation) and health system interval (time from first presentation to diagnosis). FINDINGS: We recruited 391 participants between Jan 1, 2011, and Dec 31, 2014 (24% response rate). 119 (30%) participants were diagnosed with pancreatic cancer (41 [34%] had metastatic disease), 47 (12%) with other cancers, and 225 (58%) with no cancer. 212 (54%) patients had multiple first symptoms whereas 161 (41%) patients had a solitary first symptom. In this referred population, no initial symptoms were reported more frequently by patients with cancer than by those with no cancer. Several subsequent symptoms predicted pancreatic cancer: jaundice (51 [49%] of 105 patients with pancreatic cancer vs 25 [12%] of 211 patients with no cancer; p<0·0001), fatigue (48/95 [51%] vs 40/155 [26%]; p=0·0001), change in bowel habit (36/87 [41%] vs 28/175 [16%]; p<0·0001), weight loss (55/100 [55%] vs 41/184 [22%]; p<0·0001), and decreased appetite (41/86 [48%] vs 41/156 [26%]; p=0·0011). There was no difference in any interval between patients with pancreatic cancer and those with no cancer (total diagnostic interval: median 117 days [IQR 57-234] vs 131 days [IQR 66-284]; p=0·32; patient interval 18 days [0-37] vs 15 days [1-62]; p=0·22; health system interval 76 days [28-161] vs 79 days [30-156]; p=0·68). Total diagnostic intervals were shorter when jaundice (hazard ratio [HR] 1·38, 95% CI 1·07-1·78; p=0·013) and decreased appetite (1·42, 1·11-1·82; p=0·0058) were reported as symptoms, and longer in patients presenting with indigestion (0·71, 0·56-0·89; p=0·0033), back pain (0·77, 0·59-0·99; p=0·040), diabetes (0·71, 0·52-0·97; p=0·029), and self-reported anxiety or depression, or both (0·67, 0·49-0·91; p=0·011). Health system intervals were likewise longer with indigestion (0·74, 0·58-0·95; p=0·0018), back pain (0·76, 0·58-0·99; p=0·044), diabetes (0·63, 0·45-0·89; p=0·0082), and self-reported anxiety or depression, or both (0·63, 0·46-0·88; p=0·0064), but were shorter with male sex (1·41, 1·1-1·81; p=0·0072) and decreased appetite (1·56, 1·19-2·06; p=0·0015). Weight loss was associated with longer patient intervals (HR 0·69, 95% CI 0·54-0·89; p=0·0047). INTERPRETATION: Although we identified no initial symptoms that differentiated people diagnosed with pancreatic cancer from those without pancreatic cancer, key additional symptoms might signal the disease. Health-care professionals should be vigilant to the possibility of pancreatic cancer in patients with evolving gastrointestinal and systemic symptoms, particularly in those with diabetes or mental health comorbidities. FUNDING: National Institute for Health Research and Pancreatic Cancer Action.