Positive Predicted Value
What is Positive Predictive Value?
The Breast Cancer Surveillance Consortium has three separate definitions for Positive Predictive Value (PPV) (n.d.).:
- PPV1: "PPV1 is the proportion of exams with a positive initial assessment that had a cancer diagnosis during follow-up. PPV1 should be computed using the initial assessment."
- PPV2: "PPV2 is the proportion of exams with a positive final result that had a cancer diagnosis during follow-up."
- PPV3: "PPV3 is the proportion of exams with positive final result and biopsy performed within 1 year of the exam that had a cancer diagnosis during follow-up."
What is the difference between PPV1, PPV2, and PPV3?
PPV1 is used for patients who had abnormal findings during a screening mammogram. PPV2 is used when a patient is recommended for a biopsy procedure. PPV3 is used when a biopsy is performed, and the result is indicative of cancer. Both PPV2 and PPV3 are used to assess diagnostic mammograms.
What are the desirable performance benchmarks for Positive Predictive Values 1, 2, and 3?
The desirable performance benchmarks provided by the BCSC are as follows:
- PPV1- Screening: 4.4%; Diagnostic: N/A
- PPV2- Screening: 25.6%; Diagnostic: 27.5%
- PPV3- Screening: 28.6%; Diagnostic: 30.4%
How do you calculate Positive Predictive Values 1, 2, and 3?
The formulas the BCSC provides for calculating PPV1, 2, and 3 are as follows:
- PPV1: TP / (TP + FP1)
- PPV2: TP / (TP + FP2)
- PPV3: TP / (TP + FP3)
Variables as defined by the BCSC:
- TP: True Positives
- FP1: False-Positives 1, "no known tissue diagnosis within 1 year after positive screening exam (initial assessment of 0, 3, 4, 5)."
- FP2: False-Positives 2, "no known tissue diagnosis within 1 year after positive final result."
- FP3: False-Positives 3, "concordant benign tissue diagnosis (or discordant benign tissue diagnosis and no known tissue diagnosis of cancer) within 1 year after positive final result."
Why is calculating Positive Predictive Values important for mammography auditing?
PPV is important in "finding those cancers with an acceptable range of recommendations for additional imaging and biopsies," which is a primary goal of the mammography audit (Funaro et al., 2021). Additionally, PPV is useful for mammographic recommendations for both individual radiologists and facilities alike. In the same journal article, Funaro et al. give the following example: a high PPV2 rate indicates "a high number of true positives in those studies recommended for biopsy, which may be a sign of poor performance because only lesions with a high likelihood of being cancer are recommended for biopsy, so smaller and more subtle malignancies may be getting missed" (2021).
For screening mammograms, PPV1 is useful in determining their effectiveness, including cancer detection and recall rates. For a more general view of practitioner and facility performance, PPV3 is helpful, as it reveals if the recommended clinical care was provided to the patient.
Sources:
Breast Cancer Surveillance Consortium (BCSC). (n.d.). BCSC Standard Definitions, Version 3.1. BCSC. https://www.bcsc-research.org/application/files/1516/9505/9808/BCSC_Data_Definitions_v3.1__2023.09.15.pdf
Funaro, K., Ataya, D., & Niell, B. (2021). Understanding the mammography audit. Radiologic Clinics of North America, 59(1), 41–55. https://doi.org/10.1016/j.rcl.2020.09.009