What Is PPV in Mammography? PPV1, PPV2, and PPV3 Explained

Positive predictive value (PPV) in mammography is the proportion of positive examinations that yield a tissue diagnosis of cancer within a defined follow-up window, typically one year. Under the Mammography Quality Standards Act (MQSA), the U.S. Food and Drug Administration (FDA) requires every facility to use a defined imaging assessment lexicon for the overall assessment of findings. The required categories include Negative, Benign, Probably Benign, Suspicious, Highly Suggestive of Malignancy, and Incomplete (Need Additional Imaging Evaluation). Positive examinations for audit purposes are those assessed as Suspicious or Highly Suggestive of Malignancy. The American College of Radiology (ACR) Breast Imaging Reporting and Data System (BI-RADS) Atlas supplies the widely adopted optional numbered lexicon (categories 0 through 6) that maps directly onto the FDA-MQSA assessment requirements and enables standardized calculation of three sequential PPVs. These distinctions are not academic; they expose where the pathway succeeds or fails and where data gaps distort facility performance metrics required under MQSA.

Introduction

Familiar assumption holds that "PPV" is a single number reflecting how often a positive mammogram means cancer. The unexpected distinction is that the FDA-MQSA assessment framework, operationalized through ACR BI-RADS categories, deliberately creates three sequential stopping points: PPV1 at the abnormal screening interpretation, PPV2 at the recommendation for tissue diagnosis, and PPV3 at the biopsy actually performed. The deeper mechanism is that each metric isolates a different segment of the pathway so that facilities can locate the precise operational failure: excess recalls, incomplete biopsy conversion, or lost-to-follow-up. Supporting evidence from the FDA MQSA regulations, the ACR BI-RADS Atlas (5th edition), and Breast Cancer Surveillance Consortium (BCSC) data confirms the definitions and expected ranges. Patient consequence is delayed or missed cancer diagnosis when any segment leaks. Operational implication is that incomplete ascertainment or open-loop tracking renders the numbers incomplete and non-actionable. The leadership question is whether your facility's audit system can calculate each PPV with closed-loop cancer status for every positive examination under the FDA-MQSA lexicon.

Core Definitions Anchored in FDA-MQSA Assessment Categories

MQSA requires facilities to assign an overall assessment of findings using the FDA-defined lexicon and to track outcomes for all positive mammograms (Suspicious or Highly Suggestive of Malignancy). The ACR BI-RADS Atlas maps these requirements into numbered categories that most U.S. facilities use for both reporting and audit: category 0 (Incomplete), category 3 (Probably Benign), category 4 (Suspicious), and category 5 (Highly Suggestive of Malignancy). The three PPVs are calculated from these mapped categories with explicit one-year cancer ascertainment rules.

True positive (TP) is defined as a positive examination followed by tissue diagnosis of invasive cancer or ductal carcinoma in situ within one year. The one-year window is non-negotiable; metrics cannot be finalized until the ascertainment period closes.

PPV1 (Abnormal Screening Interpretation)

Percentage of all positive screening examinations (BI-RADS categories 0, 3, 4, or 5) that result in a tissue diagnosis of cancer within one year. This metric reflects the yield of the entire screening recall process under the FDA-MQSA incomplete and positive assessment framework.

PPV1 = TP / (TP + FP1)

where FP1 = positive screening examinations with no known tissue diagnosis of cancer within one year.

PPV2 (Biopsy Recommended)

Percentage of examinations recommended for tissue diagnosis or surgical consultation (primarily BI-RADS 4 and 5, mapping to FDA Suspicious or Highly Suggestive of Malignancy) that result in a tissue diagnosis of cancer within one year. This metric isolates the yield after diagnostic work-up has refined the recommendation.

PPV2 = TP / (TP + FP2)

where FP2 = examinations recommended for tissue diagnosis with no known tissue diagnosis of cancer within one year.

PPV3 (Biopsy Performed / Positive Biopsy Rate)

Percentage of known biopsies performed as a result of positive diagnostic examinations (BI-RADS 4 or 5) that result in cancer. Also called the positive biopsy rate. This is the metric most closely watched in facility audits because it reflects actual tissue sampling outcomes.

PPV3 = TP / Number of biopsies = TP / (TP + FP3)

where FP3 = biopsies performed that yielded a concordant benign result (or discordant benign result with no subsequent cancer diagnosis) within one year.

Population mix (screening versus diagnostic, breast density distribution, underlying risk) shifts expected values, so benchmarks must be applied with local context.

Benchmark Ranges and Supporting Data

ACR BI-RADS and BCSC data supply the reference ranges most commonly used in audits that fulfill FDA-MQSA requirements. For screening mammography, acceptable PPV1 typically falls in the 3-8% range; PPV2 in the 20-40% range. BCSC data from large multi-site cohorts show median PPV1 near 4.4%, median PPV2 near 25-26%, and median PPV3 near 28-31% for digital screening. Diagnostic mammography yields higher absolute cancer rates and correspondingly higher PPV2 and PPV3 values, often in the mid-to-high 20s to low 30s percent range, again with wide facility-to-facility variation.

These ranges are not rigid pass/fail thresholds. High recall rates dilute PPV1 even when cancer detection remains adequate. Incomplete biopsy follow-up or patients who decline recommended tissue sampling distort PPV3 downward. Facilities that track only completed biopsies while ignoring recommended-but-not-performed cases produce an artificially elevated PPV3 that hides navigation failures.

Why the Three Stopping Points Matter Operationally

Each PPV functions as a diagnostic probe into a different segment of the pathway defined by the FDA-MQSA assessment lexicon. PPV1 tests the efficiency of the screening interpretation and recall decision (Incomplete or positive assessments). An elevated recall rate with low PPV1 signals over-calling of Probably Benign or Incomplete findings. PPV2 tests the quality of diagnostic work-up: does additional imaging and clinical correlation appropriately triage which cases proceed to tissue diagnosis under Suspicious or Highly Suggestive of Malignancy assessments? PPV3 tests the final conversion of recommendation into tissue and the accuracy of the recommendation itself.

Lost-to-follow-up cases create the largest data gap. When cancer status remains unknown for a positive examination, that case cannot contribute cleanly to either the numerator or the denominator. MQSA requires facilities to establish a system to collect and review outcome data for all positive mammograms and to correlate pathology results with the interpreting physician's findings. The 2023 amendments to the MQSA regulations, effective September 2024, explicitly require calculation of positive predictive value, cancer detection rate, and recall rate as part of the medical outcomes audit. Incomplete ascertainment therefore violates both the letter of the regulation and the analytic integrity of the metrics.

Population differences further modulate expected values. Screening populations with higher average density or higher underlying risk produce lower PPV1 at any given recall rate. Diagnostic populations that include symptomatic patients or known high-risk individuals produce higher PPV2 and PPV3. Facilities that pool screening and diagnostic data without stratification obscure these differences and lose the ability to act on the correct segment of the pathway.

Audit Maturity and Practical Calculation Requirements

A mature audit program calculates all three PPVs with closed-loop cancer status while remaining fully compliant with the FDA-MQSA assessment lexicon. This requires:

• Systematic capture of every Incomplete (category 0), Probably Benign (category 3), Suspicious (category 4), and Highly Suggestive of Malignancy (category 5) assessment.
• Documented recommendation for tissue diagnosis or surgical consultation.
• Confirmation that biopsy was performed (or explicit documentation of patient refusal or transfer of care).
• Linkage to pathology results or cancer registry data within the one-year window.
• Separate calculation for screening and diagnostic examinations when volumes permit.

Incomplete biopsy tracking is the most common operational failure that inflates PPV3 while hiding real patient risk. Facilities that rely solely on pathology reports received in-house miss the cases that left the system. Cross-checking against state cancer registries or structured navigation logs closes that loop.

Counterarguments exist. Some facilities argue that PPV3 alone is sufficient because it reflects actual tissue outcomes and is easiest to calculate from local pathology data. That position is incomplete. PPV3 cannot detect excess recalls that never reach biopsy recommendation, nor can it detect recommended biopsies that never occur. A facility with low PPV1 and high PPV3 may simply be recalling too many women and then correctly biopsying the subset that remains; the excess anxiety and cost remain invisible if only PPV3 is monitored. Conversely, a facility with acceptable PPV1 and PPV2 but low PPV3 may have excellent interpretation yet poor navigation to tissue diagnosis. Only the three-metric set, grounded in the FDA-MQSA lexicon and operationalized through ACR BI-RADS, reveals the true location of the failure.

Operational Implication and Leadership Question

The three PPVs are sequential filters on the same patient pathway defined by the FDA-MQSA imaging assessment lexicon. High recall lowers PPV1; incomplete conversion of recommendation to biopsy lowers the effective yield that PPV3 can capture; lost-to-follow-up makes every calculation provisional until the one-year window closes and status is known. MQSA now requires the core metrics, but the regulation does not automatically produce closed-loop data. That remains an operational design problem.

Does your facility's audit system calculate PPV1, PPV2, and PPV3 with documented cancer status for every positive examination under the FDA-MQSA assessment categories, or does it still rely on incomplete pathology returns that leave the most clinically important cases invisible?

References

1. US Food and Drug Administration. Mammography Quality Standards Act; quality standards. 21 CFR 900.12(c) and (f). Final rule amendments effective September 10, 2024.
2. Sickles EA, D'Orsi CJ. ACR BI-RADS follow-up and outcome monitoring. In: ACR BI-RADS Atlas, 5th edition. Reston, VA: American College of Radiology; 2013.
3. Lehman CD, Arao RF, Sprague BL, et al. National performance benchmarks for modern screening digital mammography: update from the Breast Cancer Surveillance Consortium. Radiology. 2017;283(1):49-58. doi:10.1148/radiol.2016161174
4. Sprague BL, et al. National performance benchmarks for modern diagnostic digital mammography: update from the Breast Cancer Surveillance Consortium. Radiology. 2017;283(1):59-69.
5. Breast Cancer Surveillance Consortium. Performance measures and data definitions. Available at: https://www.bcsc-research.org. Updated periodically.
6. American College of Radiology. BI-RADS Atlas, 5th edition, Follow-up and Outcome Monitoring section. Reston, VA: ACR; 2013.