Hamilton et al, 200513–15 | Hippisley-Cox and Coupland, 201116 | Hippisley-Cox and Coupland, 201317,18Tool for women/men | Iyen-Omofoman et al, 201312 | Jones et al, 200719 | Jordan et al, 201320 | |
---|---|---|---|---|---|---|
Outcome | Lung cancer risk within 2 years | Lung cancer risk within 2 years | Lung cancer risk within 2 years | Lung cancer risk within 4–12 months | Lung cancer risk within 3 years | Lung cancer risk within 0–1, 2–5, and 6–10 years |
Cases | 247 | 3785 | 2043a/3351a | 12 074 | 301 | 239 |
Controls | 1235 | 2 402 342 | 1 240 864b/1 263 071b | 120 731 | 4511 | 17 451 |
Predictor variables examined | All symptoms, physical signs, and investigation results from the primary care records recorded in the 2 years prior to the cases’ lung cancer diagnosis. Entered into the analyses if they occurred in ≥2.5% of cases or controls | Haemoptysis, appetite loss, weight loss, cough, dyspnoea, tiredness, hoarseness, BMI, smoking status, chronic obstructive airways disease, Townsend deprivation score, family history of lung cancer, previous diagnosis of cancer apart from lung cancer, asthma, pneumonia, asbestos exposure, and anaemia | A large number of symptoms of any cancer and risk factors, including smoking, alcohol intake, age, BMI, haemoptysis, appetite loss, weight loss, cough, dyspnoea, tiredness, anaemia, abdominal pain, dysphagia, indigestion, neck lump, night sweats, venous thromboembolism, COPDc | Age, sex, SES, smoking status, cough, haemoptysis, dyspnoea, chest/shoulder pain, weight loss, hoarseness, URTI, LRTI, non-specific chest infections, constipation, depressive disorders, COPD, outcome of blood tests, and number of GP consultations. All recorded 4–12 and 13–24 months before diagnosis | Age, sex, and haemoptysis | Age, sex, BMI, smoking status, drinking status, deprivation, and comorbidity, and musculoskeletal pain in the back, neck, shoulder, and hip |
Predictor variables in final tool | Haemoptysis, cough, fatigue, dyspnoea, chest pain, weight loss and appetite loss, thrombocytosis, and abnormal spirometry | Haemoptysis, appetite loss, weight loss, cough, anaemia, BMI, smoking status, chronic, obstructive airways disease, Townsend deprivation score, and previous diagnosis of cancer apart from lung cancer (the latter was only included in the tool for females) | In both tools: haemoptysis, appetite loss, weight loss, cough, anaemia, dysphagia, indigestion, neck lump, venous thromboembolism, COPD, smoking status, Townsend deprivation score. In tool for males only: abdominal pain, night sweats | Age, sex, SES, smoking status, cough, weight loss, haemoptysis, chest/shoulder pain, dyspnoea, hoarseness, URTI, LRTI, non-specific chest infections, COPD, and number of GP consultations. Recorded 4–12 months before lung cancer diagnosis | Age, sex, and haemoptysis | Age, sex, BMI, smoking status, drinking status, deprivation, and comorbidity, and musculoskeletal pain in the back, neck, shoulder, and hip |
Missing data handling | Unclear, but no imputation appears to have been performed | Multiple imputation to replace missing values for smoking status and BMI | Multiple imputation to replace missing values for smoking status, alcohol status, and BMI | No imputation has been performed. Low levels of missing data | No imputation appears to have been performed. All patients appear to be accounted for | Unclear, but no imputation appear to have been performed |
Tool development | Multivariate analysis with univariate pruning. Used PPVs as the risk measure | Cox regression analysis with univariate pruning, age used as the underlying time variable | Multinomial logistic regression. Used RRs as the risk measure | Multivariate logistic regression with univariate pruning; used ORs as the risk measure | Calculation of PPVs for haemoptysis split by age group and sex | Cox proportional hazards regression analysis |
Tool presentation | Tabular presentation of two tools (all patients, smokers) of the risks associated with single symptoms, repeat presentation of single symptoms, and symptom pairs. The PPVs ranged from 0.4% (cough, fatigue, both in all patients) to 17% (repeat presentation of haemoptysis in all patients) | Tabular presentation of two tools (males, females) with adjusted HRs for each predictor variable. The tools are also available on a website as a risk calculator | Tabular presentation of two tools (males, females) with adjusted RRs for each predictor variable. The tools are also available on a website as a risk calculator | As an equation with all the necessary β-coefficients for patients aged ≥40 years | Tabular format with the PPVs for haemoptysis split by age group and sex. These risk measures varied from 0.21% (in males aged <45 years) to 20.43% (in males aged ≥85 years) | Tabular format of adjusted HRs of musculoskeletal pain at the four locations. HRs were adjusted for age, sex, BMI, smoking status drinking status, deprivation, and comorbidity. Only back pain within the first year of follow up was associated with an increased risk of lung cancer (HR 1.67). |
Tool availability | Fully available | Underlying computer code, or all the numbers underlying the tool do not appear to be readily available | Underlying computer code, or all the numbers underlying the tool do not appear to be readily available | Fully available, apart from the intercept | Fully available | HRs adjusted for other variables associated with risk of lung cancer, but their individual effects not reported |
Validation | None | Internal | Internal | Internal | None | None |
Impact study | Before-and-after study | No | No | No | No | No |
BMI = body mass index. COPD = chronic obstructive pulmonary disease. HR = hazard ratio. LRTI = lower respiratory tract infection. OR = odds ratio. PPV = positive predictive value. RR = risk ratio. SES = socio-economic status. URTI = upper respiratory tract infection.
↵a Not reported separately for development and validation cohorts, so this is the total for both cohorts.
↵b Includes the number of cases (see a).
↵c Supplementary data available from the authors on request.