When my 61-year-old mother went to her GP with a walnut-sized lump on her neck and unaccountable tiredness, he reassured her that it was linked to arthritis and sent her home. Unconvinced, she contacted her rheumatologist, who immediately referred her to a cancer consultant. The neck lump was a secondary skin tumour of a cancer whose origin was never discovered and which galloped unstoppably through her body.

Many of the symptoms associated with cancer - lumps, aches and pains, tiredness, weight loss - usually reflect something less serious. Patients often seek medical advice for these, yet most GPs see only eight or nine cases of cancer a year. How can we expect them to decide who needs further investigation while also correctly diagnosing our diverse repertoire of physical and mental malaises?

Dr Jon Emery, a GP and Cancer Research UK clinician scientist at Cambridge University, is working with the National Institute for Clinical Excellence to set guidelines for what should alert GPs to possible cancer. "Diagnosing cancer presents particular problems," he says. "It's hard to be prescriptive about it, to build an evidence base about the key signs. Lots of common symptoms alert you to the possibility, but more often than not the patient doesn't have cancer."

Emery explains that two-week referral clinics, established in 1999 for suspected breast cancer but now expanding, allow GPs to fast-track patients with specific sets of symptoms. But do they work? He says: "Officially, yes. But the problem is lack of clarity about what the signs actually are. For example, 90% of people fast-tracked to colon clinics don't have cancer."

While Emery says delays in diagnosis sometimes result from the patient's denial that they may have a serious problem, the common theme in countless anecdotes of cancers that got away is the message that doctors should have listened more assiduously to their patients.

John Harrison was 46 when he developed back pain while teaching in Italy. At first he attributed it to long hours standing in class, but then he noticed other changes. A keen runner, Harrison was surprised to find himself getting breathless: "I thought maybe age was catching up. I also had an acute problem with trapped wind - unsurprising in view of what transpired." Back in England, he consulted a GP who suggested an indigestion remedy. "I felt it was something more serious but that I was somehow expected to prove it."

Harrison decided to rest more, but got no better and returned to see a more senior partner in the same practice. "He was immediately concerned," he recalls. "When my blood test showed signs of major inflammation, he decided I should have an endoscopy [internal gut examination]. I said I sensed constriction in my chest, but the GP started with the stomach. It was only when I began coughing up blood weeks later that I was sent for a chest x-ray."

Harrison had a lymphoma [a cancer that forms in the network of tubes and organs of the body's lymphatic system] the size of a grapefruit in his chest. Looking back now, after intensive chemo- and radiotherapy have given him a clean bill of health, Harrison says: "The first doctor could have seen from my notes that I was normally healthy and rarely saw a doctor. When I pointed out a series of veins that had appeared on my chest, she showed little interest. I felt I was being brushed off. The second GP has been excellent, but lost time with the endoscopy when I felt it was a chest problem. I was heartened by his honesty when he admitted I could have been diagnosed two months earlier if he'd done a chest x-ray." Reflecting on why the process went so wrong, Harrison adds: "The GP said mine was only the second lymphoma he had dealt with."

There are a number of initiatives designed to improve GPs' knowledge in specialist areas, among them a collaboration between the Royal College of General Practitioners and the Department of Health. This has committed to training up to 1,000 "GPs with special interests" by July 2004, who will take referrals from fellow GPs across many areas.

Dr James Mackay, consultant clinical geneticist in oncology at North East Thames Regional Genetics Unit, is helping GPs understand important clues from family history. "About 30% of GPs have a reasonable grasp of the importance of inherited factors, but others ignore vital clues," he says. "For example, if a woman in her 40s has a sister, mother and maternal aunt who developed breast cancer in their 30s, it's a sure sign she is at increased risk. If she develops a lump, she should see a surgeon urgently. Conversely, a woman with a lump whose granny had one in her 80s is highly unlikely to have familial breast cancer.

"I've seen many women anxious about increased risk of breast cancer because their GP couldn't reassure them that their family history wasn't relevant, while getting it right can be a big help towards early recognition, referral and timely treatment." Other approaches to help GPs decide what to do include that being pioneered within the London centre of Cancer Research UK. The NHS is testing their web-based ERA, or Early Referrals Application. Professor John Fox, head of the research team, said: "There is now so much knowledge about diseases, their management and diagnosis that it is humanly impossible for doctors to keep it all in mind and act in their patients' best interests. If targets such as two-week referrals for cancer are to be reliable, doctors need what's called clinical decision support. ERA and other systems we are developing provide just this."

But is the real problem, limited resources with GPs under pressure not to over-refer? "The solution is not to investigate everything more rigorously," says Emery. "Some countries now offer whole-body scans but this exposes people to dangerous radiation and the risk of getting false positives. Similarly, if you did a colonoscopy for everyone with rectal bleeding, you'd put a vast number of people who don't have a serious problem through the horrible preparation for the procedure."

Harrison fought his way to diagnosis and successful treatment. My mother left it too late, perhaps through fear, and I doubt that even a better GP could have saved her. We all make mistakes, even doctors, and programs like ERA will only ever be as good as the information elicited by a sensitive doctor from a patient not too scared to provide it.