Synthesis: PD-L1 testing and assay heterogeneity in TNBC

Evidence grades (GRADE-adapted): A high — multiple well-conducted RCTs or systematic reviews converge. B moderate — single pivotal RCT or consistent observational evidence. C limited — single observational study, mechanistic, or expert consensus. D preclinical / hypothesis-generating.

Why PD-L1 testing matters in metastatic TNBC

Pembrolizumab plus chemotherapy is the standard first-line systemic therapy for metastatic TNBC with PD-L1 expression at or above a defined threshold; chemotherapy alone is the default for PD-L1-negative metastatic disease. The decision rests on a single immunohistochemistry result interpreted with one of two scoring systems on one of two FDA-cleared assays, and the alignment between the assays is imperfect. The same physical tumor can be eligible for first-line pembrolizumab under one assay-scoring combination and ineligible under another. This is the operationally hardest companion-diagnostic situation in current breast oncology, and the source of substantial cross-trial and cross-institution variability in care.

Two antibodies, two scoring systems

Four PD-L1 IHC assays have been used in pivotal TNBC trials. They differ on two orthogonal dimensions: the antibody clone (and matched detection chemistry), and the scoring system applied to the resulting stain pattern.

Antibody clones

• SP142 (Ventana) — the assay used in IMpassion130 (atezolizumab + nab-paclitaxel). Companion diagnostic to atezolizumab as originally approved. Known to produce sparser, more focal staining than other clones at the same biological signal level.
• 22C3 (Dako) — the assay used in KEYNOTE-355 and KEYNOTE-522 (pembrolizumab combinations). Companion diagnostic to pembrolizumab. Produces somewhat stronger and broader staining than SP142 in the same tissue.
• SP263 (Ventana) — an alternative PD-L1 clone with reasonable concordance to 22C3 across multiple tumor types but with no formal companion-diagnostic role in TNBC.
• 28-8 (Dako) — primarily relevant outside breast cancer; included in some cross-assay comparisons.

Scoring systems

• IC (immune cell) score — the percent of the tumor area occupied by PD-L1-stained immune cells. Used with SP142 in IMpassion130. Threshold for atezolizumab eligibility: IC ≥ 1%.
• CPS (combined positive score) — the count of PD-L1-stained cells (tumor cells, lymphocytes, macrophages) divided by the total number of viable tumor cells, multiplied by 100. Used with 22C3 in KEYNOTE-355. Threshold for pembrolizumab eligibility in metastatic disease: CPS ≥ 10.

These are not interconvertible by formula. Patients whose tumor scores IC ≥ 1% on SP142 do not necessarily score CPS ≥ 10 on 22C3, and vice versa. The two assays measure overlapping but non-identical biology, and the two scoring systems weight it differently.

IMpassion130 — the atezolizumab evidence base

Schmid and colleagues randomized 902 patients with previously untreated metastatic TNBC to atezolizumab + nab-paclitaxel vs placebo + nab-paclitaxel. The primary endpoints were investigator-assessed PFS in the ITT population and in the PD-L1-positive (SP142 IC ≥ 1%) subgroup^[1]A:

• ITT PFS: 7.2 vs 5.5 months (HR 0.80).
• PD-L1+ subgroup PFS: 7.5 vs 5.0 months (HR 0.62).
• OS update (Schmid 2020 Lancet Oncol): ITT OS HR 0.87 (not significant); PD-L1+ subgroup median OS 25.4 vs 17.9 months (numerical benefit, statistical hierarchy precluded formal testing)^[2]A.

The FDA granted accelerated approval to atezolizumab + nab-paclitaxel for PD-L1–positive (IC ≥ 1% by SP142) metastatic TNBC in March 2019. The companion diagnostic was the SP142 assay specifically; the approval did not authorize use of 22C3 or any other assay as the eligibility test.

IMpassion131 — the confirmatory failure

IMpassion131 was the confirmatory phase III trial required by the accelerated approval, designed to test atezolizumab + paclitaxel (not nab-paclitaxel) vs placebo + paclitaxel in the same patient population. The trial did not meet its primary endpoint: PFS in the PD-L1–positive (SP142 IC ≥ 1%) subgroup was 6.0 vs 5.7 months (HR 0.82, p=0.20)^[3]A. OS in the same subgroup was numerically worse with atezolizumab (22.1 vs 28.3 months).

The negative result was interpreted variably: chemotherapy backbone difference (paclitaxel vs nab-paclitaxel) was the most-cited hypothesis, with the suggestion that the corticosteroid premedication required for solvent-based paclitaxel might blunt the IO response. Whether the failure reflects a true biological difference between the chemotherapy backbones or chance variation in a moderately-powered trial remains debated.

Regulatory consequences were sharp. Genentech voluntarily withdrew the atezolizumab indication for TNBC in the US in August 2021, and the EMA followed shortly thereafter. Atezolizumab is no longer available for first-line metastatic TNBC outside of a clinical trial. Pembrolizumab is now the only PD-1/PD-L1 agent approved for this setting.

KEYNOTE-355 — the pembrolizumab evidence base

KEYNOTE-355 ran in parallel with IMpassion130 and was the trial that established pembrolizumab in this setting. Cortes and colleagues randomized 847 patients with previously untreated metastatic TNBC to pembrolizumab + chemotherapy (investigator's choice of nab-paclitaxel, paclitaxel, or gemcitabine + carboplatin) vs placebo + chemotherapy. PD-L1 was measured with the 22C3 assay; pre-specified analyses tested benefit at CPS ≥ 1, CPS ≥ 10, and ITT^[4]A:

• CPS ≥ 10 subgroup PFS: 9.7 vs 5.6 months (HR 0.65).
• CPS ≥ 1 subgroup PFS: 7.6 vs 5.6 months (HR 0.74).
• ITT PFS: 7.5 vs 5.6 months (numerical benefit; statistical hierarchy precluded formal testing).
• OS update (Cortes 2022 NEJM): CPS ≥ 10 subgroup median OS 23.0 vs 16.1 months (HR 0.73; statistically significant by pre-specified hierarchy)^[5]A.

Both PFS and OS benefits concentrated in the CPS ≥ 10 subgroup. The FDA approval (November 2020) specified CPS ≥ 10 by 22C3 as the eligibility threshold, and that remains the standard companion-diagnostic configuration today. KEYNOTE-119 in second-line monotherapy was a separate trial that explored CPS thresholds of 1, 10, and 20 in pembrolizumab single-agent therapy; benefit increased monotonically with the CPS cutoff, suggesting that the dose-response relationship between PD-L1 expression and IO benefit is real, not artifact^[6]A.

The cross-assay concordance problem

The clinical-utility question that has occupied the field since 2019: given a tumor scored on SP142 (IC), what is the probability it would score CPS ≥ 10 on 22C3 if re-tested?

Rugo and colleagues conducted a post hoc analysis of IMpassion130 in which 614 archival tumor samples were re-tested with 22C3 and SP263 alongside the original SP142^[7]A. Headline findings:

• SP142 IC ≥ 1% positivity: 46% of the cohort.
• 22C3 CPS ≥ 1% positivity: 81% — SP142 is much sparser at the lowest threshold.
• 22C3 CPS ≥ 10 positivity: 39% — closer to SP142 IC ≥ 1% by total positivity, but the patients identified are not the same set.
• Concordance (positive vs negative) between SP142 IC ≥ 1% and 22C3 CPS ≥ 10: ~64%. Approximately 36% of cases were discordant.

A separate prospective multi-institutional study (Reisenbichler 2020) reported substantially lower inter-pathologist agreement on SP142 IC scoring (kappa ~0.27) than on 22C3 CPS scoring (kappa ~0.39 — still poor, but better)^[8]B. The SP142 IC scoring is particularly hard for pathologists to do reproducibly because immune cells are mobile, scattered, and present at variable density across tumor regions.

The combined picture: the two main assays disagree on roughly a third of cases at their respective trial-defined thresholds, and the SP142 assay is the harder of the two to score reproducibly. This drove the field toward 22C3 / CPS as the de facto standard once atezolizumab was withdrawn.

Preanalytical and analytical sources of variability

Beyond the assay-clone differences, the standard preanalytical variability story applies to PD-L1 testing as it does to ER/PR/HER2 (see IHC for ER/PR/HER2 and the TNBC definition for the foundational analysis):

• Tissue age. PD-L1 antigenicity degrades over months to years in FFPE blocks. Archival tissue more than 3 years old shows systematically lower positivity rates — relevant when re-testing for first-line metastatic disease using primary-tumor blocks.
• Fixation and processing. The standard 6–72 hour neutral buffered formalin fixation window applies; outside it, results are unreliable.
• Intratumoral heterogeneity. PD-L1 expression varies across tumor regions and between primary tumors and metastases. Discordance between matched primary and metastatic samples is documented at 15–25%; testing on the most recent biopsy (typically the metastatic site at progression) is preferable when available.
• Prior therapy effects. Cytotoxic chemotherapy can transiently induce PD-L1 expression in tumor cells; recent chemotherapy exposure may inflate apparent PD-L1 positivity. The KEYNOTE-355 protocol required testing on archival pre-treatment tissue to avoid this confound.

What clinicians can actually do today

Pragmatic guidance for the clinician facing a patient with metastatic TNBC and a PD-L1 result on the "wrong" assay:

• If your lab reports CPS by 22C3 and the threshold is CPS ≥ 10: standard. Use the result directly for pembrolizumab eligibility.
• If your lab reports IC by SP142: the result was informative for atezolizumab eligibility, which no longer matters in the US. For pembrolizumab eligibility, request 22C3 CPS testing on the same tissue block. If the lab cannot do 22C3, consider repeat biopsy if clinically feasible.
• If your patient is CPS < 10 (PD-L1-negative for pembrolizumab): chemotherapy alone (or BRCA-mutated → PARP inhibitor) is the standard. Consider clinical trial enrollment for the substantial subset of PD-L1-negative patients in whom novel-IO-combination strategies are being tested.
• If your patient is CPS 5–10 (just below threshold): not eligible for first-line pembrolizumab per the FDA approval; some clinicians and patients consider it borderline given the imperfect assay reproducibility, but off-label use is not supported by trial evidence and not generally reimbursed.

The cleaner long-term solution — assay harmonization, or a digital-pathology AI scoring tool that abstracts away the assay specifics — remains an active investigational goal but is not yet clinical-grade.

Alternative biomarkers under investigation

The unreliability of PD-L1 IHC has motivated several alternative-biomarker programs:

• Tumor-infiltrating lymphocytes (TILs) per the International TILs Working Group standard are the most-mature alternative biomarker. TILs are prognostic in early-stage TNBC^[9]A; their predictive value for IO benefit specifically is supported by retrospective analyses of KEYNOTE-522 and IMpassion130 but has not been prospectively validated as a treatment-selection tool^[10]B.
• Tumor mutational burden (TMB). Pembrolizumab carries a tissue-agnostic indication for TMB ≥ 10 mutations/Mb (KEYNOTE-158); approximately 5–10% of TNBC tumors meet this threshold. The indication is meaningful for the TMB-high subset but does not generalize.
• MSI / mismatch-repair deficiency. Very rare in breast cancer (< 2% of TNBC) but carries a tissue-agnostic pembrolizumab indication when present.
• Gene-expression-based immune signatures (T-cell-inflamed signature, immune checkpoint signatures). Largely retrospective; not yet used clinically.
• ctDNA-based tumor immunogenicity markers. Investigational; potential advantage of decoupling from tissue-based assay reproducibility issues.

Evidence table — PD-L1 assays in pivotal TNBC trials

Note: KEYNOTE-522 did not use PD-L1 for enrollment selection; the early-stage benefit appears uniform across PD-L1 strata. PD-L1 is therefore a metastatic-setting decision tool only.

Open questions and active investigation

• Why did IMpassion131 fail when IMpassion130 succeeded? The chemotherapy-backbone hypothesis (steroid premedication for solvent-based paclitaxel) is plausible but unproven; alternative hypotheses include cohort drift, chance variation in moderately-powered trials, and unspecified differences in supportive care. A direct comparison of atezolizumab + nab-paclitaxel vs atezolizumab + paclitaxel would require a head-to-head trial that is unlikely to be funded now that atezolizumab is withdrawn from this indication.
• Will assay harmonization succeed? The Blueprint PD-L1 Assay Comparison Project has produced multiple cross-tumor-type assay-concordance datasets; a TNBC-specific harmonization effort comparable to the lung-cancer PD-L1 Blueprint would substantially clarify clinical practice. As of 2025 this has not been formalized.
• Digital-pathology AI assistance. Multiple groups have reported CNN-based PD-L1 scoring with reproducibility better than expert pathologists; none is FDA-cleared as a standalone scoring device. Pathologist-assist deployment is plausible within 2–3 years.
• Beyond-PD-L1 IO combinations. Several agent classes are in development for the PD-L1-negative metastatic TNBC population: TIGIT inhibitors, LAG-3 inhibitors, novel chemo + IO combinations, ADC + IO combinations. None has yet reached pivotal-trial readout in TNBC specifically.
• Liquid-biopsy PD-L1. ctDNA-based PD-L1 measurement is technically feasible but lags tissue IHC in clinical maturity. If validated, it could bypass the entire assay-reproducibility problem.
• Integrating PD-L1 with TILs and TMB. A composite immune-context biomarker is a long-discussed but unimplemented direction; the CONTACT or BLUEPRINT-style multi-marker scoring frameworks remain academic exercises rather than clinical tools.

For the upstream IHC framework that defines TNBC itself, see IHC for ER/PR/HER2 and the TNBC definition. For the KEYNOTE-522 early-stage trial that does not use PD-L1 for selection (a useful contrast), see KEYNOTE-522 and the neoadjuvant chemo+IO standard. For the patient-layer companion describing treatment selection in plain language, see Treatment options.

References

Each citation links to the original publication via DOI. The same records are searchable in the evidence library by title or DOI.

1. Schmid P, Adams S, Rugo HS, et al. Atezolizumab and Nab-Paclitaxel in Advanced Triple-Negative Breast Cancer (IMpassion130). N Engl J Med. 2018;379(22):2108–2121. doi:10.1056/NEJMoa1809615. ↩
2. Schmid P, Rugo HS, Adams S, et al. Atezolizumab plus nab-paclitaxel as first-line treatment for unresectable, locally advanced or metastatic triple-negative breast cancer (IMpassion130): updated efficacy results from a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol. 2020;21(1):44–59. doi:10.1016/S1470-2045(19)30689-8. ↩
3. Miles D, Gligorov J, André F, et al. Primary results from IMpassion131, a double-blind, placebo-controlled, randomised phase III trial of first-line paclitaxel with or without atezolizumab for unresectable locally advanced/metastatic triple-negative breast cancer. Ann Oncol. 2021;32(8):994–1004. doi:10.1016/j.annonc.2021.05.801. ↩
4. Cortes J, Cescon DW, Rugo HS, et al. Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial. Lancet. 2020;396(10265):1817–1828. doi:10.1016/S0140-6736(20)32531-9. ↩
5. Cortes J, Rugo HS, Cescon DW, et al. Pembrolizumab plus Chemotherapy in Advanced Triple-Negative Breast Cancer. N Engl J Med. 2022;387(3):217–226. doi:10.1056/NEJMoa2202809. ↩
6. Winer EP, Lipatov O, Im SA, et al. Pembrolizumab versus investigator-choice chemotherapy for metastatic triple-negative breast cancer (KEYNOTE-119): a randomised, open-label, phase 3 trial. Lancet Oncol. 2021;22(4):499–511. doi:10.1016/S1470-2045(20)30754-3. ↩
7. Rugo HS, Loi S, Adams S, et al. PD-L1 Immunohistochemistry Assay Comparison in Atezolizumab Plus Nab-Paclitaxel–Treated Advanced Triple-Negative Breast Cancer. J Natl Cancer Inst. 2021;113(12):1733–1743. doi:10.1093/jnci/djab108. ↩
8. Reisenbichler ES, Han G, Bellizzi A, et al. Prospective multi-institutional evaluation of pathologist assessment of PD-L1 assays for patient selection in triple negative breast cancer. Mod Pathol. 2020;33(9):1746–1752. doi:10.1038/s41379-020-0540-1. ↩
9. Loi S, Drubay D, Adams S, et al. Tumor-Infiltrating Lymphocytes and Prognosis: A Pooled Individual Patient Analysis of Early-Stage Triple-Negative Breast Cancers. J Clin Oncol. 2019;37(7):559–569. doi:10.1200/JCO.18.01010. ↩
10. Salgado R, Denkert C, Demaria S, et al. The evaluation of tumor-infiltrating lymphocytes (TILs) in breast cancer: recommendations by an International TILs Working Group 2014. Ann Oncol. 2015;26(2):259–271. doi:10.1093/annonc/mdu450. ↩

Last reviewed: 2026-05-31. Researcher-layer synthesis page. Evidence grades follow the GRADE-adapted rubric defined at the top of this page. Citations are anchored to the full bibliographic entries above; click the ↩ arrow next to any reference to return to its first citation in the prose.