Accounting for only 3% of all new cancer cases in the UK, pancreatic cancer is the fifth leading cause of cancer death.1 Diagnosis usually occurs at advanced stages, with 61% of patients diagnosed at stage IV in England in 2020.1,2 Outcomes for pancreatic cancer in the UK remain poor, with only a quarter of patients surviving 1 year.3 Almost half of diagnoses in England are made via attendances to emergency departments, which is a diagnostic route associated with worse survival.3 Only 15% of diagnoses are diagnosed via the urgent suspected cancer (USC, previously known as ‘2-week wait’) pathway, which allows GPs to refer patients for urgent specialist assessment.3 This indicates that identifying symptomatic patients with possible pancreatic cancer is particularly challenging in primary care.
CHALLENGES IN DIAGNOSIS
Early diagnosis is challenging since pancreatic cancer typically presents with non- specific symptoms that have low positive predictive values (PPVs) for cancer, such as abdominal pain, back pain, and weight loss.4 Consequently, time to referral can be long, as indicated by the relatively high proportion of patients with pancreatic cancer having multiple GP consultations before diagnosis.5 Jaundice is the sole alarm symptom,4 though its reported frequencies range from 12% to 43%, and tends to be a late sign of disease.6 Given the non-specific nature of many presentations, multidisciplinary centres (sometimes known as ‘vague’ or ‘non- specific’ symptom pathways) may improve timely diagnosis of pancreatic cancer. Indeed, an analysis from five such centres showed that, of 241 cancers diagnosed, 25 (10%) were pancreatic cancers.7
In individual patients, the low likelihood that common symptoms such as weight loss are due to pancreatic cancer mean that CT, the definitive modality to identify the disease, is not usually warranted.4 Current National Institute for Health and Care Excellence (NICE) guidelines advise GPs to ‘consider’, rather than ‘offer’, urgent CT to patients aged 60 years and over with unexplained symptoms including weight loss and any of diarrhoea, back pain, abdominal pain, nausea, vomiting, constipation, or new- onset diabetes.8 England’s NHS has recently announced that GPs will have direct access to imaging, including urgent CT, but challenges around implementation are anticipated in relation to radiology capacity, which is restricted by workforce and equipment constraints.9 Consequently, it is not yet clear whether timely direct GP access to CT imaging will become widely available in the near future.9 Recognising the barriers that GPs face in obtaining direct access to CT imaging, the guidance permits ultrasound if CT is not available, despite the relatively poor accuracy of that modality, with a sensitivity and specificity of 75%–89% and 90%–99%, respectively.10 We do not know what proportion of patients who have symptoms for which NICE recommends consideration of investigation actually receive these tests.
Worsening control of blood sugar in those who have known diabetes or new-onset diabetes is an important presentation of pancreatic cancer.11 Given the ubiquity of type 2 diabetes, the possibility of pancreatic cancer can be easily overlooked. A model has been developed to estimate pancreatic cancer risk in patients in the US with new- onset diabetes, while a case-control study using English primary care records has utilised HbA1c alongside other blood test results, body mass index, comorbidities, and medications.12,13 In the UK, clinical decision support tools such as QCancer are also available in general practice.14 However, despite retrospective evidence supporting the performance of these models in risk prediction for pancreatic cancer, prospective evaluation regarding their utility and uptake in routine care remain limited.
NOVEL DIAGNOSTIC AND MULTI-CANCER TESTS
Historically, tests and tools have been developed to detect a single cancer type, but new multi-cancer early diagnostic tools (MCEDs) such as Galleri, a blood test that is currently being evaluated in the NHS, are instead intended to identify a range of cancers. This approach has promise for cancers, such as pancreatic, with non-specific symptoms that overlap with those of other cancers. For example, while abdominal bloating has a relatively low PPV for pancreatic cancer (0.05%), the PPV for any cancer is higher at 1.7%.15 However, important questions remain regarding MCEDs, including their performance for specific intended purposes during the diagnostic process (for example, for risk stratification for further testing or for cancer diagnosis), and their utility in identifying early-stage disease amenable to treatment.16 Only a small number of pancreatic cancer diagnoses have so far been documented in studies on Galleri, and the prospects of the test for expediting diagnosis of the disease needs to be further evaluated.17
SHOULD THE ROLE OF CANCER ANTIGEN 19-9 IN PRIMARY CARE BE RE-EVALUATED?
GPs are familiar with using first-line tests (‘triage tests’), such as faecal immunochemical test for colorectal cancer, chest X-ray for lung cancer, and cancer antigen 125 (CA 125) for ovarian cancer, prior to further referral or more definitive investigations.18 By contrast, the long- established blood-based biomarker for pancreatic cancer, cancer antigen 19-9 (CA 19-9), remains little used for this purpose, with CA 19-9 currently not part of NICE’s recommendations for suspected pancreatic cancer.8,19
Evidence on the performance of CA 19-9 in symptomatic general practice populations is limited, but a systematic review primarily drawn from hospital patients has reported sensitivity and specificity of 79% (95% confidence interval [CI] = 70 to 90) and 82% (95% CI = 68 to 91).19 If evaluated in primary care populations, with lower prevalence of disease, test performance would be expected to diminish, but might still be broadly comparable to existing triage tests such as chest X-ray for lung cancer, which has a sensitivity and specificity of 75% (95% CI = 68 to 83) and 90% (95% CI = 90 to 91).20 Although the diagnostic accuracy of these tests may be suboptimal in primary care, CA 19-9, like chest X-ray, might still have utility to inform consideration of further CT imaging for some patients who are judged to be at low risk, particularly given that such patients may not otherwise have testing for pancreatic cancer at all.
An important concern arising from such a strategy would be the substantial proportion of false positive results from CA 19-9 if applied to a low prevalence population, for which there are currently limited data available. Assuming a prevalence of 0.1% in those tested with CA 19-9 and a specificity of 90%, 10% of results could be false positives leading to a substantial number of additional CT investigations. Specificity could well be found to be substantially lower; consequently, it may be worth investigating combining CA 19-9 with other biomarkers, such as CA 125 and carcinoembryonic antigen (CEA), or within panels of novel biomarkers.21,22 As well as research on performance and cost-effectiveness of these biomarkers on symptomatic populations in primary care, health economic analysis could inform whether the cost and harms could be offset by earlier detection and define the threshold of benefits and/or the specific populations for which such a strategy might be beneficial.
RETHINKING IMAGING FOR PANCREATIC CANCER
Although CT with contrast is the recommended investigation for suspected pancreatic cancer, the limited availability of this test, particularly in the NHS, means that GPs have to steward this radiological resource carefully. Consequently, direct-access investigation with CT where available, or via urgent suspected cancer referrals, are likely to be reserved predominantly for patients deemed to be at substantial risk manifested by more advanced symptomatology. Rather than imaging the whole abdomen, the possibility of instituting a bespoke protocol whereby only the pancreas is imaged could be investigated. This may reduce the resources and time needed to perform imaging and reporting, potentially making it more accessible. Thus patients with relatively common, low-risk symptoms could have a rapid and accessible imaging test, akin to the current use of chest X-ray for prolonged cough. Such a departure from conventional diagnostic practices would require comprehensive evaluation of diagnostic performance, feasibility, acceptability, and cost-effectiveness.
CONCLUSION
Despite some reductions in late-stage diagnosis rates, outcomes for pancreatic cancer remain poor.1,2 If we are to achieve earlier diagnosis and improve outcomes, innovative approaches are needed. This could include the use of CA 19-9 in new ways or in biomarker panels, devising imaging approaches that can be delivered with adequate volume for patients with common symptoms, and evaluating novel technologies such as MCEDs. Early pancreatic diagnosis remains a significant challenge, but a combined research effort exploring both how we can use existing tests to best effect and examining the added value of novel diagnostics within the pathway is likely to give us the best chance of improving outcomes for patients in the near future.
Notes
Provenance
Freely submitted; externally peer reviewed.
Competing interests
Stephen H Bradley is clinical lead for cancer for the Leeds office of the West Yorkshire integrated care board. The opinions expressed in this editorial are the authors’ alone and do not necessarily reflect those of employers. Garth Funston is a member of the editorial board of the BJGP. The other authors have declared no competing interests.
- © British Journal of General Practice 2023
