To investigate the diagnostic power of computed tomography for discriminating malignant from nonmalignant causes to pleural effusions in consecutive patients with no malignant cells found at a cytological examination and a chest x-ray not suspicious of malignancy. The investigators hypothesised that the combination improves the chance of detecting the cause of the pleural effusion.

Objectives

1. Is CT-scanning better than chest x-ray plus a thoracocentesis for discriminating malignant from benign pleural effusions?
2. What is the added value of a CT-scanning to chest x-ray plus thoracocentesis for discriminating malignant from benign pleural effusions?

The above measured as, true positive, true negative, false positive, false negative, sensitivity, specificity, likelihood ratio+, likelihood ratio -, positive predictive value, negative predictive value and diagnostic accuracy.

Study design A retrospective diagnostic study including consecutive patients with a unilateral pleural effusion. Thoracocentesis, chest x-ray and CT findings are categorised as either normal (i.e. not suggestive of any aetiology of the unilateral pleural effusion), suggestive of other lung pathology or suggestive of malignancy (i.e. representing a possible aetiology of the unilateral pleural effusion). The final diagnosis is extracted from the patients' electronic medical records. When no diagnosis is found, two investigators agree on a consensus diagnosis based on all investigation results. If no reasonable diagnosis can be established based on the findings, the patient case is categorised as having no final diagnosis.

The investigators follow the STARD guideline for reporting diagnostic accuracy studies.

Patients The investigators will include consecutive patients presenting with a unilateral pleural effusion between 01 January 2013 and 31 December 2016. The patients will be identified searching the hospitals administrative patient system for the procedure code thoracocentesis (KTGA30).

Departments of Respiratory Medicine at Zealand University Hospital, Roskilde and Naestved Hospital, Region Zealand, Denmark, participate in this study. Both hospitals are a large tertiary hospital with specialised functions. Patients older than 16 years are included, irrespective of smoking history and comorbidities, if both thoracocentesis, chest x-ray and a CT-scanning are performed. Exclusion criteria are previously diagnosed lung cancer, thoracic malignancy or incomplete data.

Data Relevant data on demographics, descriptive data, symptoms, medical history and diagnostic workup will be extracted from electronic medical records, and registered in the individual patient's case report file (CRF). The CRF is stored using Excel on an encrypted USB-key, the USB-key will be stored behind double lock (room plus cupboard). Background-variables will be utilised to describe the study population and for explorative research. Only study coordinators will have access to data. Computers, as well as a back-up, will be stored and kept in a locked cupboard in a locked room. Only study coordinators possess relevant keys.

Statistics Data will be presented as frequencies and/or mean ± standard deviation (SD). Test characteristics will be compared using McNemar's test with a Bonferroni correction. An alpha level of 0.05 is considered significant. Test characteristics for thoracocentesis, chest x-ray and CT-scanning and the combination of the three will be calculated. Test characteristics are true positive, true negative, false positive, false negative, sensitivity, specificity, likelihood ratio+, likelihood ratio-, positive predictive value, negative predictive value and diagnostic accuracy.

The combined sensitivity of two parallel tests will be calculated using the formula:

The sensitivity of test A + sensitivity of test B - (sensitivity of test A x sensitivity of test B)

The combined specificity for two parallel tests will be calculated using the formula:

The specificity of test A x specificity of test B

The combined sensitivity of three parallel tests will be calculated using the formula:

The sensitivity of test A + sensitivity of test B + sensitivity of test C - (sensitivity of test A x sensitivity of test B x sensitivity of test C)

The combined specificity for three parallel tests will be calculated using the formula:

The specificity of test A x specificity of test B x sensitivity of test C For data analysis, we use STATA (StataCorp LLC, Version 15.0, College Station, Texas, USA).

Ethics The study is retrospective and observational and thus will not affect the included patients. Before any study-related activity, the protocol and study must be approved by the Danish Data Protection Agency and the local Committee on Biomedical Research Ethics. All results will be stored and analysed by computer, and the investigators secure the patients' anonymity according to the national laws. The data are anonymized and stored after completion of the study and data containing personal identification numbers will be kept behind double-lock at the department.

Inclusion Criteria:

• Thoracocentesis
• Chest X-ray and
• CT-scanning are performed

Exclusion Criteria:

• Lung cancer
• Other thoracic malignancy

All patients with a unilateral pleural effusion of unknown origin