Introduction: The objective of this study was to investigate the evidence for positive predictive value (PPV) of alarm symptoms and combinations of symptoms for colorectal cancer, breast cancer, prostate cancer and lung cancer in general practice.
Methods: This study is based on a literature search performed in PubMed, Embase, the Cochrane database and at ClinicalTrials.gov in accordance with the PRISMA guidelines. The main outcome measure used was PPV.
Results: A total of 16 eligible studies were identified. The intervals in the brackets refer to the variation of the results in the studies. Colorectal cancer: The PPV of “rectal bleeding” was high for patients > 60 years (6.6-21.2%), but much lower in younger age groups. For “change in bowel habits” and “significant general symptoms”, the PPV was 3.5-8.5%. Breast cancer: “Palpable suspected tumour” was well supported (8.1-24%). No studies on the predictive value of “pitting of the skin”, “papil-areola eczema/ulceration” and “suspect axillary lymph nodes” were found. Prostate cancer: One study showed a high PPV for positive rectal examination (12%). The value for “lower urinary tract symptoms” was low (1.0-3.0%). PPV for “perianal pain” and “haemospermia” were not found. Lung cancer: For “haemoptysis” the PPV increased from 8.4 in patients aged 55 years to 20.4 at the age of > 85 years. PPV for “cough”, “pain in the thorax”, “dyspnoea” and “general symptoms” were low (0.4-1.1%). Using a new algorithm that estimates the PPV of combinations of symptoms and risk factors, a higher PPV may be achieved.
Conclusion: A few of the alarm symptoms show a high PPV, whereas the PPV for some symptoms currently remains unknown. To improve the GPs’ diagnostic judgment, a new algorithm for calculating the PPV for combinations of symptoms and risk factors seems promising.
Waiting a long time for a diagnosis is associated with a poorer prognosis for cancer patients [1, 2]. In Denmark 2008 saw the introduction of accelerated, well-planned, fast-track diagnostic and treatment pathways, coined cancer packages, for patients suspected of having cancer. For each cancer type, the Danish Health and Medicines Authority developed a manual for general practitioners (GPs) describing the symptoms that should engender a reasonable suspicion of malignancy (alarm symptoms) and trigger the initiation of the diagnostic pathway. Naturally, the relevance of the manual depends on the ability of the alarm symptoms to predict cancer in a general practice population. Low predictive values may imply a waste of resources and unnecessary anxiety for patients. The aim of this review was to examine the evidence for PPV of the alarm symptoms described by the Danish Health and Medicines Authority in relation to the four most commonly occurring cancers in Denmark: colorectal cancer, lung cancer, prostate cancer and breast cancer. Furthermore, tools for calculation of PPV using combinations of symptoms and risk factors were examined.
The MeSH-terms shown in Figure 1 were combined. The literature search was performed in PubMed, Embase, the Cochrane database and at ClinicalTrials.gov in accordance with the PRISMA guidelines . The last search was performed on 9 April 2014. The reference lists of the papers we identified were examined for other relevant articles. Each of the identified papers was read independently by the authors and assessed for eligibility. Data from the included studies were extracted and transferred into data sheets (Table 1).
Original studies in English or Danish using an unselected population from general practice with a newly recognised alarm symptom were included. The studies had to have positive predictive values (PPV) or likelihood-ratios (LR) for any of the alarm symptoms included in the guidance from the Danish Health and Medicines Authority. The PPV is expressed as the percentage of patients with a given symptom who actually have cancer.
The PPV depends on the prevalence of the disease, which should be kept in mind when interpreting the results. The LR of a symptom is the probability of finding the symptom in patients with cancer divided by the probability of the same finding in patients without cancer. Thus, an LR > 1 indicates an association between the symptom and cancer.
If a paper described an alarm symptom in terms of more specific symptoms, e.g. fatigue, weight loss and loss of appetite instead of using the term “significant general symptoms”, the results for these categories were included.
Studies originating from secondary health-care services or screening programmes were excluded.
Quality assessment of the studies
The studies identified were assessed using the Newcastle-Ottawa Quality Assessment Scales (NOQAS)  for cohort and case-control studies. An article was rated according to three elements: the selection of study groups, the comparability of these groups, and the evaluation of exposure and outcome, graded from 1 to 9 stars.
The search yielded 46 studies, and 16 studies met the inclusion criteria and were included in the final analysis (Figure 2). The results from studies of individual symptoms will be presented first, followed by combinations of symptoms.
In the guidance from the Danish Health and Medicines Authority aiming to assist access to the pathways for fast track cancer diagnosis, the following alarm symptoms are described: rectal bleeding, changes in bowel habits and significant general symptoms in patients more than 40 years old [5, 6].
Several studies showed a significant association between rectal bleeding and colorectal cancer [7-19] (Table 1). A total of seven prospective cohort studies, four retrospective cohort studies and one case-control study were identified. The range in PPV was considerable; from 0.22 to 21.1. The highest PPV was demonstrated in a Belgian retrospective study of people aged 70-79 years . The same study found that younger people aged < 50 and 50-59 years only showed a PPV of 0.7 and 1.7,
respectively. Wauters and Jones [8, 9] described a decreasing PPV at the age of > 75 and 80 years. Conversely, Lawrenson  found an increasing PPV in the group aged > 80. Jones describes a difference in gender with men having a higher PPV, particularly in the age group 75-84 years. In the study, only small differences in PPV were found in relation to colorectal cancer diagnosed
after six months and diagnosed after three years from the first presentation of rectal bleeding with the GP. One study detected an elevated PPV on the first occasion of bleeding compared to later events (14.8 compared to 4.4). Two studies on rectal bleeding and development of colorectal cancer achieved the highest rating of 9, and seven studies were rated from 3 to 7 (Table 1).
Changes in bowel habits
Two studies, a case control study and a retrospective cohort study, presented PPV for later diagnosis of colorectal cancer [7, 11, 18]. The studies showed a PPV from 3.5 to 8.5 in the age groups 60-69, 70-79 and 80-89 years. Both studies showed a trend towards a higher PPV in the 70-79 years age range than in the other age groups. Six studies described a change in bowel habits with an LR ranging from 1 to 2.9 [13, 15-17]. Hamilton  found low PPVs for constipation and diarrhoea, amounting to respectively 0.42 and 0.94, without taking age into account. Most studies were of good quality (Table 1).
Significant general symptoms
Four studies described LR for abdominal pain [13, 15, 17]. A variation from 0.7 to 2.2 was found. Fitjen et al  detected decreased appetite and nausea with an LR of 0.7 and 0.4, respectively. Nørrelund & Nørrelund  found discomfort with a PPV of 1.3 and 0.9. Weight loss was described in nine studies [7, 8, 13, 15-19] and LR varied from 1.3 to 5.1. The studies were of variable quality (Table 1).
In relation to breast cancer, the following alarm symptoms were described: Suspected tumour at palpation, new papil retraction, new indentation of the skin, eczema or ulceration of the papil/areola and clinically suspect axillary lymph nodes [20, 21].
Suspected tumour at palpation
Three retrospective cohort studies described PPV of suspected tumour at palpation [22-24]. The variation of PPV was substantial, from 8.1- 24. The quality of the studies was quite good (Table 1).
No studies were found. A single study of moderate quality quantified LR for nipple complaint  and found an LR of 3.1.
The PPV of newly onset papil retraction, indentation of the skin, eczema or ulceration of the papil/areola and clinically suspect axillary lymph nodes was not examined in the literature; therefore no conclusion could be drawn.
The following alarm symptoms were described: symptoms from the lower urinary tract, positive rectal exploration, perianal pain and haemospermia [25, 26]
Symptoms from the lower urinary tract
Hamilton et al  described PPV for more symptoms originating from the lower urinary tract. The symptom hesitancy has a PPV of 3.0, nocturia 2.2 and haematuria 1.0. LR of urinary retention reached 9. The quality of the studies was very good.
Hamilton et al  have described PPVs for this symptom in patients over 40 years of age. Hamilton found a PPV of 12 for a positive exploration and 2.8 for a negative one in a high-quality study (Table 1).
Perianal pain and haemospermia
No studies were identified and PPV is unknown.
The following alarm symptoms were described: cough, haemoptysis (> 50 years), thoracal pain (> 50 years), unexplained dyspnoea and general symptoms > 40 years [28, 29] (Table 1).
Only one study was found in the literature search. The PPV of cough for lung cancer was 0.4 in this study of good quality .
In a study by Jones et al  the PPV of haemoptysis was very low for younger people (< 45 years: males 0.21 and females 0.36), increasing with age to high levels (males > 85 years: 20.4, females 75-84 years: 10.5, > 85 years: 2.6) Hence, a significant gender difference was shown. This may be explained by higher prevalence of lung cancer in men than in women. In the study, a higher PPV was also described for development of lung cancer after 3 years than after 6 months from the first presentation of haemoptysis to the GP. In another study, Hamilton et al found a PPV of 2.4 for development of lung cancer .
Only one study was found. A low PPV of 0.82 for thoracal pain was demonstrated .
Unexplained dyspnoea and general symptoms
Low PPV was found in a single high-quality study by Hamilton . For dyspnoea, the value was 0.66, weight loss 1.1 and reduced appetite 0.87 [18, 30].
No studies were identified therefore no conclusion can be drawn.
Combination of more symptoms
Three systematic reviews concerning PPV for a combination of symptoms relating to colorectal cancer have recently been published. Rectal bleeding, weight loss, abdominal pain, change in bowel habits and perianal symptoms were combined. These studies showed a lack of evidence from the primary care sector, and unconvincing diagnostic values were presented [31-33].
One recent very comprehensive review concerning symptoms, signs and diagnostic tests for colorectal cancer, urological cancer, lung cancer, oesophageal cancer, breast cancer and gynaecological cancer was found. In this review, it was not possible to make any significant conclusion for patients presenting with combinations of symptoms. This was due to the small size of the study populations and the unpersuasive quality of the studies included in this comprehensive review . Hamilton et al conducted studies that are examples of this. In these studies, he described PPV for combinations of two symptoms, but did not calculate confidence intervals due to a small sample size .
Hippisley-Cox et al have recently developed a new algorithm to estimate the PPV for cancer of combinations of symptoms. It is based on data capture in 452 general practices in England and Wales including 3.8 million patients [35, 36]. The estimation incorporates both symptoms (e.g. lump in the breast, haemoptysis and weight loss) and risk factors (e.g. age, gender, smoking, family history and social determinants) in a multivariate analysis. The algorithm has been used to estimate the absolute risk of having a lung cancer and colorectal cancer. Meanwhile, an extensive presentation of calculations of PPV has not been published, but some examples have [37, 38]. For haemoptysis as the only symptom without any other symptoms and risk factors included in the calculation, the PPV is 6%. A 78-year old female ex-smoker with haemoptysis, anaemia, cough and chronic obstructive pulmonary disease (COLD) has an estimated risk of 37% of having lung cancer. If the patient also has loss of appetite and weight loss, the estimated risk increases to 76% . For a patient with rectal bleeding alone, the PPV for colorectal cancer is 3%. A 60-year-old male with rectal bleeding, recent change in bowel
habits, anaemia, loss of appetite and a positive family history of gastrointestinal cancer has a 49% risk of colorectal cancer . At a website connected to the project , a so-called risk calculator is publicly available for calculation of PPVs given different symptoms and risk factors, e.g. for a 69-year-old female with a breast lump and nipple discharge, who is a light drinker and an ex-smoker, the cancer risk is 74% of which 73% is due to breast cancer. If the patient only has a lump, the PPV for breast cancer is 21%.
The PPV of alarm symptoms formulated by the Danish Health and Medicines Authority shows large variation among the published studies. Often the results have no segmentation for gender or age. Some alarm symptoms have no published evidence to support them as no
studies were found and no conclusions could be drawn about the value of these symptoms in daily clinical practice. However, significant and good quality evidence was found describing rectal bleeding and colorectal, suspect tumour at palpation and cancer mamma, positive findings from rectal exploration and cancer of the prostate as well as haemoptysis and lung cancer. A few studies have interesting results concerning age and gender. Rectal bleeding, changing of bowel habits and haemoptysis are more predictive for cancer in men than in women. Rectal bleeding is more predictive for colorectal cancer in age groups over 70 years, and haemoptysis is more predictive for lung cancer in age groups over 55 years. However, many of these methodological problems have been addressed in the method developed by Julia Hippisley-Cox et al allowing inclusion of many variables in the estimation of PPV of symptoms [35, 36]. As mentioned above, the results are based on data from a very large sample of patients and GPs which contributes to a high statistical power in the calculations.
Most of the studies identified were of good quality, but some weaknesses were apparent. The pattern for seeking a GP’s help may vary between countries. A low iatropic threshold of patients could imply a lower PPV. The missing information about gender and age could lead to results that are difficult to use in practice. For elderly people with the largest risk of cancer, the estimated PPV will be too low if younger people are included in the studies without controlling for it. In this context, it must be remembered that the PPV is dependent on the prevalence of the disease. The follow-up period was often missing in the studies. This could lead to underestimation of the PPV. Publication bias is probably a smaller problem in this research subject. However, selective publishing of results with high PPVs will lead to an overestimation of the real PPV. Some studies may have been overlooked, as only papers in Danish and English were included.
The quality of patients’ records in general practice
In England, uncertainty and lack of agreement when using data collected in general practice has been reported . The quality of the data has been questioned. GPs may report patients with more serious symptoms, or the clinical examination may include other symptoms. This selection bias implies an overestimation of PPV for a single symptom. Some researchers have used prospective computerised coding of symptoms. Others have made a retrospective reading of the patient record looking for alarm symptoms. The completeness of the latter
method is probably lower than that of the former. It may imply overestimation of PPV if the GP does not register the symptoms in the patient record at the time of the clinical examination.
Alarm symptoms in general practice
Knowing alarm symptoms is necessary for the GP in daily clinical decision-making. However, it is very important to know the PPV in order to ensure that these clinical decisions are evidence-based. Selecting a patient for diagnostic fast track has a great impact on the resources used in health care. Admitting patients to fast track services with a low PPV for their symptoms could lead to unnecessary anxiety and medicalisation. Many diagnostic methods have side effects and a mortality rate, e.g. colonoscopy . In the clinical situation, the physician typically does not make a judgement based on one symptom, but on a combination of factors, e.g. all symptoms together, the way the symptom is presented, how long it has been present, etc. Until recently, it has been disappointing that the published literature makes it difficult to determine the importance of a combination of symptoms. The methods developed by Julia Hippisley-Cox et al [35, 36] for risk estimation seem very promising. The public risk calculator established in relation to the method may be useful in daily practice.
It is clear that research estimating PPV or LR is needed for symptoms traditionally interpreted as serious as well as for other symptoms. The diagnosis of advanced cancer is relatively easy; the challenge is to detect cancers early. The generalist knowledge of GPs, combined with the high-level communication skills needed to get the necessary information from patients, is vital.
In this review, one large review conducted by Shapley  has investigated the prognostic factors of symptoms. However, he only included symptoms with a PPV > 5.
Our aim was to investigate the evidence supporting the symptoms described in the manual developed by The Danish Health and Medicines Authority for the four largest cancers. This gave us 17 different symptoms that needed to be investigated. The selection of these symptoms as entry points to the packages has not been addressed. Other symptoms could be relevant. Use of methods for combinations of symptoms and risk factors would be relevant.
Based on evidence, we were looking for the level of PPV that should trigger a diagnostic fast track. This is also studied by Shapley, who argues that a PPV of 5 should be considered as highly predictive . But 19 out of 20 patients admitted to a cancer package following examination by their GP would not have a cancer in this instance. Meanwhile, compared to the values found in screening programmes, a PPV of 5 is much higher . In this context, a fact-based ethical discussion is needed concerning the number of patients that should be referred to diagnostic fast track to find an early cancer.
As this paper shows, there is a lack of knowledge about the significance of symptoms in relation to cancer. In particular, more research is needed in the general practitioner’s judgment of the entire clinical picture. Also, the level at which PPVs determine further diagnostic investigation calls for research into the costs and ethics of fast tracks. It may be worthwhile to determine if some clinicians are particularly good at diagnosing cancer and, if so, what information they use. Electronic data capture in general practice with classification of the patient’s reason for the contact and the diagnosis managed in the domains of the GP, with transmission to central databases offers good opportunities for research into the significance of symptoms  as shown in English large scale studies [35, 36].
Correspondence: John Sahl Andersen, Section of General Practice, Research Unit for General Practice and the University Clinic, University of Copenhagen, 1014 Copenhagen K, Denmark. E-mail: email@example.com
Accepted: 16 January 2015
Conflicts of interest:Disclosure forms provided by the authors are available with the full text of this article at www.danmedj.dk
Richards MA. The size of the prize for earlier diagnosis of cancer in England. Br J Cancer 2009;101(suppl 2):S125-S129.
Olesen F, Hansen RP, Vedsted P. Delay in diagnosis: the experience in Denmark. Br J Cancer 2009;101(suppl 2):S5-S8.
Moher D, Liberati A, Tetzlaff J et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. BMJ 2009;339:b2535.
Wells GA SB, O’Connell D, Peterson J et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality if nonrandomized studies in meta-analyses. www.ohri.ca/programs/clinical_epidemiology/oxford.htm (6 Mar 2015).
Sundhedsstyrelsen. Pakkeforløb for kræft i tyk- & endetarm. Sundhedsstyrelsen, 2009;1,1:39.
Praksiskonsulentordningen. PKO-NYT Tyk - og endetarmskræft. Hillerød: Praksiskonsulentordningen, Region Hovedstad, 2009.
Hamilton W, Lancashire R, Sharp D et al. The risk of colorectal cancer with symptoms at different ages and between the sexes: a case-control study. BMC Med 2009;7:17.
Wauters H, Van Casteren V, Buntinx F. Rectal bleeding and colorectal cancer in general practice: diagnostic study. BMJ 2000;321:998-9.
Jones R, Latinovic R, Charlton J et al. Alarm symptoms in early diagnosis of cancer in primary care: cohort study using General Practice Research Database. BMJ 2007;334:1040.
Parker C, Hippisley-Cox J, Coupland C et al. Rectal and postmenopausal bleeding: consultation and referral of patients with and without severe mental health problems. Br J Gen Pract 2007;57:371-6.
Lawrenson R, Logie J, Marks C. Risk of colorectal cancer in general practice patients presenting with rectal bleeding, change in bowel habit or anaemia. Eur J Cancer Care (Engl) 2006;15:267-71.
du Toit J, Hamilton W, Barraclough K. Risk in primary care of colorectal cancer from new onset rectal bleeding: 10 year prospective study. BMJ 2006;333:69-70.
Heintze C, Matysiak-Klose D, Krohn T et al. Diagnostic work-up of rectal bleeding in general practice. Br J Gen Pract 2005;55:14-9, discussion 8.
Ellis BG, Thompson MR. Factors identifying higher risk rectal bleeding in general practice. Br J Gen Pract 2005;55:949-55.
Metcalf JV, Smith J, Jones R et al. Incidence and causes of rectal bleeding in general practice as detected by colonoscopy. Br J Gen Pract 1996;46:
Fijten GH, Starmans R, Muris JW et al. Predictive value of signs and symptoms for colorectal cancer in patients with rectal bleeding in general practice. Fam Pract 1995;12:279-86.
Norrelund N, Norrelund H. Colorectal cancer and polyps in patients aged 40 years and over who consult a GP with rectal bleeding. Fam Pract 1996;
Hamilton W. The CAPER studies: five case-control studies aimed at identifying and quantifying the risk of cancer in symptomatic primary care patients. Br J Cancer 2009;101(suppl 2):S80-S86.
Hamilton W, Round A, Sharp D et al. Clinical features of colorectal cancer before diagnosis: a population-based case-control study. Br J Cancer 2005;93:399-405.
Sundhedsstyrelsen. Pakkeforløb for brystkræft. Sundhedsstyrelsen, 2009;1,1:35.
Praksiskonsulentordningen. PKO-NYT Brystkræft. 2011.
Eberl MM, Phillips RL, Jr., Lamberts H, et al. Characterizing breast symptoms in family practice. Ann Fam Med 2008;6:528-33.
Barton MB, Elmore JG, Fletcher SW. Breast symptoms among women enrolled in a health maintenance organization: frequency, evaluation, and outcome. Ann Intern Med 1999;130:651-7.
Bywaters JL. The incidence and management of female breast disease in a general practice. J R Coll Gen Pract 1977;27:353-7.
Praksiskonsulentordningen. PKO-NYT Kræft i prostata, penis og testikler. 2009.
Sundhedsstyrelsen. Pakkeforløb for kræft i prostata, penis og testikler. Sundhedsstyrelsen, 2009;1,1:71.
Hamilton W, Sharp DJ, Peters TJ et al. Clinical features of prostate cancer before diagnosis: a population-based, case-control study. Br J Gen Pract 2006;56:756-62.
Sundhedsstyrelsen. Pakkeforløb for lungekræft. Sundhedsstyrelsen, 2009;1,1:35.
Praksiskonsulentordningen. PKO-NYT Lungekræft. 2010.
Hamilton W, Peters TJ, Round A et al. What are the clinical features of lung cancer before the diagnosis is made? A population based case-control study. Thorax 2005;60:1059-65.
Ford AC, Veldhuyzen van Zanten SJ, Rodgers CC et al. Diagnostic utility of alarm features for colorectal cancer: systematic review and meta-analysis. Gut 2008;57:1545-53.
Olde Bekkink M, McCowan C, Falk GA et al. Diagnostic accuracy systematic review of rectal bleeding in combination with other symptoms, signs and tests in relation to colorectal cancer. Br J Cancer. 2010;102:48-58.
Jellema P, van der Windt DA, Bruinvels DJ et al. Value of symptoms and additional diagnostic tests for colorectal cancer in primary care: systematic review and meta-analysis. BMJ 2010;340:c1269.
Shapley M, Mansell G, Jordan JL et al. Positive predictive values of ≥5% in primary care for cancer: systematic review. Br J Gen Pract 2010;60:366-77.
Hippisley-Cox J, Coupland C. Symptoms and risk factors to identify women with suspected cancer in primary care: derivation and validation of an algorithm. Br J Gen Pract. 2013;63:11-21.
Hippisley-Cox J, Coupland C. Symptoms and risk factors to identify men with suspected cancer in primary care: derivation and validation of an algorithm. Br J Gen Pract 2013;63:1-10.
Hippisley-Cox J, Coupland C Identifying patients with suspected lung cancer in primary care: derivation and validation of an algorithm Br J Gen Pract 2011;61:653-4
Hippisley-Cox J, Coupland C Identifying patients with suspected colorectal cancer in primary care: derivation and validation of an algorithm. Br J Gen Pract 2012;62:29-37
www.qcancer.org (6 Mar 2015).
Toward Consensus for best practice: Use of patient records from general practice for research. London: Welcome Trust, 2009.
Luning TH, Keemers-Gels ME, Barendregt WB et al. Colonoscopic perforations: a review of 30,366 patients. Surg Endosc 2007;21:994-7.
Morikawa T, Kato J, Yamaji Y et al. A comparison of the immunochemical fecal occult blood test and total colonoscopy in the asymptomatic population. Gastroenterology 2005;129:422-8.
www.dak-e.dk/flx/english/dak_e_it/sentinel_data_capture/ (6 Mar 2015).