Identifying patients with suspected lung cancer in primary care: derivation and validation of an algorithm

Our plain-English summary of Identifying patients with suspected lung cancer in primary care: derivation and validation of an algorithm, by J. Hippisley-Cox, C. Coupland (The British Journal of General Practice, 2011).

The study

• Authors: J. Hippisley-Cox, C. Coupland
• Journal: The British Journal of General Practice
• Published: 2011;61(592):e715-23
• Cited by: 84 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Background: Lung cancer has one of the lowest survival outcomes of any cancer because more then two-thirds of patients are diagnosed when curative treatment is not possible. The challenge is to help earlier diagnosis of lung cancer and hence improve prognosis.

Aim: To derive and validate an algorithm incorporating information on symptoms, to estimate the absolute risk of having lung cancer

Design And Setting: Cohort study of 375 UK QResearch® general practices for development, and 189 for validation.

Method: Selected patients were aged 30-84 years and free of lung cancer at baseline and haemoptysis, loss of appetite, or weight loss in previous 12 months. Primary outcome was incident diagnosis of lung cancer recorded in the next 2 years. Risk factors examined were: haemoptysis, appetite loss, weight loss, cough, dyspnoea, tiredness, hoarseness, smoking, body mass index, deprivation score, family history of lung cancer, other cancers, asthma, chronic obstructive airways disease, pneumonia, asbestos exposure, and anaemia. Cox proportional hazards models with age as the underlying time variable were used to develop separate risk equations in males and females. Measures of calibration and discrimination assessed performance in the validation cohort.

Results: There were 3785 incident cases of lung cancer arising from 4 289 282 person-years in the derivation cohort. Independent predictors were haemoptysis, appetite loss, weight loss, cough, body mass index, deprivation score, smoking status, chronic obstructive airways disease, anaemia, and prior cancer (females only). On validation, the algorithms explained 72% of the variation. The receiver operating characteristic (ROC) statistics were 0.92 for both females and males. The D statistic was 3.25 for females and 3.29 for males. The 10% of patients with the highest predicted risks included 77% of all lung cancers diagnosed over the subsequent 2 years.

Conclusion: The algorithm has good discrimination and calibration and could potentially be used to identify those at highest risk of lung cancer, to facilitate early referral and investigation.

In plain English

The authors built and tested a tool estimating the chance of a lung cancer diagnosis within 2 years, using UK general practice records from 2000 to 2010. Practices in England and Wales were randomly split: patients aged 30 to 84 at 375 practices (2 406 127 patients) were used to build the tool, and 1 267 151 patients at 189 practices to test it. Candidate factors included symptoms, smoking, deprivation, other illnesses and asbestos exposure. Asbestos exposure was recorded for about 0.1% of patients and was not kept in the final models for women or men; the article gives no separate risk figure. Compared with its absence, current coughing up blood (haemoptysis) had the largest hazard ratio in the final models: 23.9 (95% CI 20.6 to 27.6) in women and 21.5 (19.3 to 23.9) in men. In the test group, the 10% with the highest predicted risk included 1697 of 2196 lung cancers, a sensitivity of 77.3% (95% CI 75.5 to 79.0). The authors concluded the tool had good discrimination and calibration, noting that cancer stage was not recorded.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Prospective cohort study using a UK primary care database (QResearch), with the risk tool developed in one group of practices and tested in a separate group
• Who was studied: Patients aged 30 to 84 registered with general practices in England and Wales between 1 January 2000 and 30 September 2010: 2 406 127 patients at 375 practices for development and 1 267 151 patients at 189 practices for validation
• Exposure: Candidate predictors of lung cancer: symptoms (haemoptysis, appetite loss, weight loss, cough, dyspnoea, tiredness, hoarseness), smoking, body mass index, deprivation, family history, other cancers, asthma, chronic obstructive airways disease, pneumonia, anaemia, and ever-recorded asbestos exposure (yes/no). Asbestos exposure was recorded for 1649 (0.1%) patients in the development cohort and 928 (0.1%) in the validation cohort. The fiber type was not specified
• Main results:
  • Development cohort: 3785 new cases of lung cancer over 4 289 282 person-years, a rate of 88.2 per 100 000 person-years
  • Asbestos exposure was not included in the final model for women or men. Variables that were not kept were described as not independent risk factors. No hazard ratio for asbestos was reported
  • Compared with people without it, current haemoptysis had an adjusted hazard ratio of 23.9 (95% CI 20.6 to 27.6) in women and 21.5 (95% CI 19.3 to 23.9) in men. These were the largest hazard ratios in the final models. Age was used as the time variable and has no hazard ratio
  • Compared with non-smokers, heavy smokers (20 or more cigarettes a day) had an adjusted hazard ratio of 10.6 (95% CI 8.49 to 13.2) in women and 6.35 (95% CI 5.43 to 7.43) in men
  • In the validation cohort, the ROC statistic was 0.92 (95% CI 0.91 to 0.93) for both women and men. The tool explained 71.7% of the variation in time to diagnosis in women and 72.1% in men
  • In the validation cohort, the top 10% of predicted risk (2-year risk above 0.37%) included 1697 of 2196 new lung cancers. This was a sensitivity of 77.3% (95% CI 75.5 to 79.0), with a positive predictive value of 1.34% (95% CI 1.28 to 1.40)
  • In the validation cohort, only 18.4% of lung cancers occurred in current or ex-smokers aged 40 and over who presented with a first onset of haemoptysis, which is the approach in NICE guidance. The positive predictive value in this group was 9.7%
• The authors' conclusion: The authors concluded that the algorithm has good discrimination and calibration and could potentially be used to identify those at highest risk of lung cancer, to help with early referral and investigation
• Limitations the authors state:
  • Outcomes were not formally adjudicated, and there was information bias and missing data
  • The stage of lung cancer at diagnosis was not recorded
  • The study gives no evidence on whether using the tool would lead to earlier identification of lung cancer at a stage when curative treatment is more likely
  • Symptoms may be recorded less completely or accurately than diagnoses
  • Family history of lung cancer may be under-recorded, and some patients may misreport their smoking

Topics

The National Library of Medicine indexes this study under subjects our site covers: Lung cancer.

Medical terms in this study, explained in our Lexicon: Lung Cancer, Prognosis, Cohort Study, Asbestos Exposure.

Related studies

• Guideline for referral of patients with suspected lung cancer by family physicians and other primary care providers — M.E. Del Giudice, S.M. Young, et al. (2014)
• Comparison of the ultra-low-dose Veo algorithm with the gold standard filtered back projection for detecting pulmonary asbestos-related conditions: a clinical observational study — M. Tekath, F. Dutheil, et al. (2014)
• Combining a gravitational search algorithm, particle swarm optimization, and fuzzy rules to improve the classification performance of a feed-forward neural network — M.L. Huang, Y.C. Chou (2019)
• Pleural fluid cytology samples in cases of suspected lung cancer: An experience from a tertiary care centre — G. Kaur, R. Nijhawan, et al. (2017)
• Prospective study of asbestos-related diseases incidence cases in primary health care in an area of Barcelona province — M. Rosell-Murphy, R. Abós-Herràndiz, et al. (2010)
• Prediction of Radiation Pneumonitis With Machine Learning in Stage III Lung Cancer: A Pilot Study — M. Yakar, D. Etiz, et al. (2021)

Browse return to medical research.