Intended for healthcare professionals


Alarm symptoms in primary care

BMJ 2007; 334 doi: (Published 17 May 2007) Cite this as: BMJ 2007;334:1013
  1. Robin Fox, general practitioner,
  2. John Fletcher, general practitioner, clinical epidemiologist, BMJ
  1. Health Centre, Bicester, Oxon OX26 6AT
  1. Robin.Fox{at}

    These greatly increase the risk of cancer, but the diagnosis is still rare

    A new diagnosis of cancer is rare in primary care, and the role of general practitioners (GPs) in diagnosing cancer can be challenging. In the United Kingdom, a GP with a list size of 1500 will on average see only 1.39 cases of lung cancer, 0.23 oesophageal cancers, 0.99 colorectal cancers, and 0.45 renal or bladder cancers each year.1 We seek to diagnose the few patients with cancer out of the many who are concerned about it or who have non-specific problems. In this week's BMJ, a cohort study by Jones and colleagues assesses the association between alarm symptoms and a subsequent diagnosis of cancer in just under 800 325 patients in primary care.2

    Diagnostic errors are one of the leading causes of medicolegal claims against GPs,3 and they can damage the faith of our patients. However, we have a role as gatekeepers of health resources and more recently the added responsibility of managing a budget. Over-referral to secondary care can unnecessarily raise patients' anxiety while awaiting investigation and waste precious resources.

    What are we to make of alarm symptoms? Are certain symptoms or signs so suggestive of cancer that no further consideration is needed apart from how to write the urgent referral letter? On the face of it, the high positive likelihood ratios for cancer reported by Jones and colleagues,2 which range from 75 for rectal bleeding to around 300 for dysphagia, might suggest this. Put simply, the presence of dysphagia makes it 300 times more likely that a patient has cancer. But strangely, even this is not enough for the GP to refer because most patients with such symptoms will not have cancer. The positive predictive value (PPV) of dysphagia for cancer is only 2% in women and 5% in men; that is, more than 95% will not have cancer.

    Further difficulties arise when we analyse what doctors mean when they code symptoms. Current National Institute for Health and Clinical Excellence (NICE) guidelines4 define dysphagia as interference with the swallowing mechanism that occurs within five seconds of having started swallowing. It advises urgent referral of dyspeptic patients with dysphagia who have “suspected cancer.” But dysphagia has been reported as a symptom in 37% of patients with erosive oesophagitis, and this resolves in most (83%) patients after treatment with a proton pump inhibitor.5

    The recent Montreal definition and classification of gastro-oesophageal reflux disease highlights this problem.6 It defines “troublesome dysphagia” as dysphagia that causes patients to alter their eating patterns or have symptoms of solid food getting impacted. Dysphagia is troublesome only in a minority of patients with gastro-oesophageal reflux disease. The Montreal classification suggests that troublesome and worsening dysphagia, especially for solids, is an alarm symptom and should be investigated. Jones and colleagues found that the PPV of dysphagia for cancer was only 0.16-0.21% if patients were less than 45 years old. GPs have to decide whether to treat young patients at lower risk who have non-troublesome dysphagia initially with a one month trial of proton pump inhibitors or automatically to refer them all.

    While Jones and colleagues found the PPV of haematuria was high for urological cancer (5.5%. for men, 2.5% for women), age and sex have a strong effect—the PPV is only 0.22% for women under 45 years. If a 40 year old woman presents with a first episode of cystitis-like symptoms and haematuria, a urinary tract infection may be the most likely diagnosis, but this should be confirmed by a midstream urine specimen. In a 70 year old man, similar symptoms should be viewed with high suspicion as the PPV for urological cancer is 11.2% in such patients,2 and this is not altered by the presence or absence of dysuria.7 8 This supports the NICE guidelines, which suggest urgent referral of adults with painless macroscopic haematuria.4 Patients with symptoms suggestive of a urinary infection and macroscopic haematuria should be referred urgently if infection is not confirmed by investigation. Patients aged 40 years or more who present with recurrent or persistent urinary tract infection associated with haematuria should also be referred urgently, as urological cancer can present in this way.4

    NICE guidelines suggest haemoptysis should be investigated by chest radiography.4 If the results are negative, those aged 40 or more should be referred urgently if haemoptysis persists. Secondary care studies suggest 6-21% may have lung cancer when investigated further, and these cancers may be smaller and more curable than those detected on radiography.9 This is supported by the findings of Jones and colleagues, where the PPV was 4.1-20.4% in patients over 55 but only 0.21-0.36% in those under 45. The PPV for the younger patients in particular may be an overestimate because this is a General Practice Research Database study, which is dependent on GPs correctly coding haemoptysis. GPs may be more likely to do this if they plan to make a referral than if a small amount of blood is mixed with sputum in a young patient with a presumed chest infection.

    Rectal bleeding is the most common alarm symptom in primary care identified by Jones and colleagues. It has the lowest PPV overall—only around 2%—which highlights the difficulties GPs face when presented with this symptom. Many patients with rectal bleeding fear they have bowel cancer but do not quite fit the criteria for urgent referral. Hopefully, the national bowel screening programme will improve things for the future.

    The take home message is that alarm symptoms need to be considered seriously. The 2005 NICE guidelines on referral for suspected cancer4 provide a valuable and pragmatic tool that can help GPs make realistic referral decisions. These guidelines are now supported by evidence from primary care.2 However, the action that a GP takes will depend on their intimate knowledge of the patient and his or her wishes.


    • Competing interests: None declared.

    • Provenance and peer review: Commissioned; not externally peer reviewed.