This JAMA Guide to Statistics and Methods describes how a decision curve analysis can be used to evaluate the benefits of a diagnostic test, such as 3 prostate biopsy strategies.
Decision curve analysis (DCA) is a method for evaluating the benefits of a diagnostic test across a range of patient preferences for accepting risk of undertreatment and overtreatment to facilitate decisions about test selection and use.1 For example, Siddiqui and colleagues2 used DCA to evaluate 3 prostate biopsy strategies: targeted magnetic resonance/ultrasound fusion biopsy, standard extended-sextant biopsy, or a combination, for establishing the diagnosis of intermediate- to high-risk prostate cancer. Their goal was to identify the best biopsy strategy to ensure prostatectomy is offered to patients with intermediate- and high-risk tumors and avoided for patients with low-risk tumors.
When patients have signs or symptoms suggestive of but not diagnostic of a disease, they and their physician must decide whether to (1) treat empirically, (2) not treat, or (3) perform further diagnostic testing before deciding between options 1 and 2. The decision to treat depends on how confident the clinician is that the disease is present, the effectiveness and complications of treatment if the disease is present, and the patient's willingness to accept the risks and burden of a treatment that might not be necessary. A diagnostic test may provide additional information on whether the disease is present.3 Decision curve analysis is a method to assess the value of information provided by a diagnostic test by considering the likely range of a patient's risk and benefit preferences, without the need for actually measuring these preferences for a particular patient.1
A key concept in DCA is that of a probability threshold, namely, a level of diagnostic certainty above which the patient would choose to be treated. The probability threshold used in DCA captures the relative value the patient places on receiving treatment for the disease, if present, to the value of avoiding treatment if the disease is not present. If the treatment has high efficacy and minimal cost, inconvenience, and adverse effects (eg, oral antibiotics for community-acquired pneumonia), then the probability threshold will be low; conversely, if the treatment is minimally effective or associated with substantial morbidity (eg, radiation for a malignant brain tumor), then the probability threshold will be high.
The net benefit, or “benefit score,” is determined by calculating the difference between the expected benefit and the expected harm associated with each proposed testing and treatment strategy. The expected benefit is represented by the number of patients who have the disease and who will receive treatment (true positives) using the proposed strategy.
The expected harm is represented by number of patients without the disease who would be treated in error (false positives) multiplied ...