Why PD-L1 immunohistochemistry left an opening for antibody PET
PD-L1 immunohistochemistry became the practical companion diagnostic for checkpoint blockade, but its limitations were recognised early: single-site biopsies miss intratumoural and interlesional heterogeneity, staining thresholds and antibody conditions vary, and PD-L1 expression is dynamic over a patient's course [2][5]. In gastroesophageal cancer, 18F-BMS-986229 PET detected tracer accumulation in some lesions while 71% of patients with any accumulation also had lesions without uptake, directly visualising intrapatient heterogeneity that a single biopsy cannot capture [5]. The same review notes that PD-L1 combined positive score evolves during disease and that repeated biopsies are invasive and not always feasible [5]. These constraints motivated whole-body, non-invasive PD-L1 imaging rather than replacing IHC outright.
The first-in-human step was 89Zr-atezolizumab PET, which used a tracer-only antibody dose of about 1 mg followed by scans at 1 hour and days 2, 4 and 7, establishing that the labelled antibody could be given safely and imaged over time [3]. That precursor defined the prior frontier: it showed where the tracer distributed before therapy, but it could not say whether therapeutic atezolizumab dosing occupies the same binding sites. The anchor study was designed to close exactly that gap by imaging during treatment [1].
On-treatment uptake is lower than pretreatment, but not abolished
In the anchor study, patients received 10 mg 89Zr-atezolizumab after the initial two cycles of 1200 mg atezolizumab, with PET/CT on days 4 and 7 and tumour biopsies around day 7 of both cycles [1]. Tumour SUVmax on day 7 was essentially unchanged between cycle 1 (geometric mean 5.5, 95% CI 4.4–6.9) and cycle 2 (5.4, 95% CI 4.3–6.8; N = 13; P = 0.88), and tumour-to-background ratio likewise did not change significantly (1.67 vs 1.53; N = 11; P = 0.32) [1]. Compared with previously reported pretreatment 89Zr-atezolizumab PET/CT data, on-treatment uptake was lower in lesions and spleen, with a higher blood pool SUVmean [1]. Autoradiography signal was still present in tumour biopsies, and most lesions remained conspicuous against background [1].
The authors read this as partial PD-L1 tumour saturation: a single 1200 mg dose appears sufficient to reach a plateau, but not to block tracer binding completely [1]. They argue against a static saturable PD-L1 pool and propose a dynamic pool in which newly surfaced PD-L1 outpaces antibody binding, partly limited by slow extravasation, so increasing antibody dose yields diminishing returns [1]. This interpretation is a hypothesis about mechanism, not a direct measurement of unoccupied binding sites, and the study did not demonstrate full saturation at any dose level [1].
The partial-saturation signal recurs across tracers and antibodies
The strongest external comparison comes from 89Zr-durvalumab PET in non-small cell lung cancer, where 50 tracer-positive lesions were identified before treatment and only 17 during treatment in 11 patients, with two new lesions appearing; the authors of the anchor paper cite this as also suggesting partial saturation during the first durvalumab cycle [1]. Limited 89Zr-pembrolizumab data in NSCLC point in the same direction, with reduced but still present uptake during therapy, although the anchor authors note the very small sample size and lack of quantitative comparison [1]. A peptide-based tracer, 68Ga-HF12, showed the same directional pattern in CHO-hPD-L1 xenografts: tumour uptake fell from 5.33 ± 0.25 %ID/g before atezolizumab to 3.2 ± 0.17 %ID/g after two treatments and stabilised at 3.16 ± 0.15 %ID/g after five, while PBS controls showed no change [4]. That convergence across a large antibody tracer, a smaller antibody tracer, and a peptide tracer strengthens the interpretation that therapeutic PD-L1 blockade leaves measurable available target [1][4].
The comparison also exposes a conflict with preclinical expectations. Mouse PET experiments with radiolabelled checkpoint inhibitors had suggested complete tumour saturation, but the anchor authors explicitly note the limitations of translating those models to the clinic [1]. A separate murine study using 68Ga-HF12 likewise reported reduced tracer uptake after atezolizumab, consistent with decreased available PD-L1 binding sites rather than complete occupancy [4]. The disagreement is therefore not about whether blockade occurs, but about whether the clinical 1200 mg regimen achieves the near-complete saturation predicted by animal models [1][4].
Where the saturation evidence stops: response trends and dosing
The anchor study reported that higher tumour-to-background ratios showed a positive trend with best overall response, with a p for trend of 0.069, which does not meet conventional statistical significance [1]. The cohort was small: 24 patients were enrolled, 23 had evaluable PET/CT, and the paired cycle 1 versus cycle 2 tumour-to-background analysis rested on 11 patients with 27 paired lesions [1]. Median follow-up was 7.1 months, and the population was heterogeneous, spanning triple-negative breast cancer, MSI-high colon carcinoma, cholangiocarcinoma, cervical cancer, sarcoma and other tumour types [1]. The response-prediction signal is therefore a hypothesis-generating trend, not a validated biomarker [1].
The dosing implication is similarly bounded. The study shows that a single 1200 mg dose reaches a plateau of partial saturation and that a second cycle does not increase tumour saturation, but it did not test alternative doses or schedules, and the partial-saturation interpretation is an imaging observation rather than a pharmacokinetic or receptor-occupancy measurement [1]. The authors state that increasing antibody dose has diminishing returns approaching but never reaching full saturation, which is a mechanistic argument against simple dose escalation, not evidence that a different dose would improve outcomes [1]. For trialists, the practical message is that on-treatment PD-L1 PET can quantify available target and heterogeneity, but it cannot yet be used to adjust atezolizumab dosing or to replace biopsy-based PD-L1 assessment [1][5].
What changes for oncology imaging researchers and trialists
The anchor paper moves PD-L1 PET from a pretreatment biodistribution exercise to a serial, on-treatment pharmacodynamic readout [1][3]. It provides a concrete quantitative benchmark: day 7 tumour SUVmax around 5.4–5.5 and tumour-to-background ratios around 1.5–1.7 during therapy, with no cycle-to-cycle change [1]. It also shows that on-treatment imaging is feasible alongside therapeutic dosing, with no tracer-related adverse events and only one atezolizumab infusion-related reaction [1]. These are the kinds of operational details that enable larger imaging trials.
The boundary is that the study cannot distinguish partial saturation caused by limited extravasation from partial saturation caused by rapid PD-L1 turnover, and it cannot exclude that the lower on-treatment uptake partly reflects altered antibody pharmacokinetics or blood-pool effects [1]. The higher blood pool SUVmean during treatment and the approximately 10% decrease in liver uptake in cycle 2 are consistent with a changed biodistribution, which complicates direct comparison with pretreatment scans [1]. Until a larger study links on-treatment tracer availability to clinical outcomes with adequate statistical power, the saturation finding should be treated as a mechanistic observation with a plausible but unproven link to response [1].
About These Sources
This research page is built on 5 peer-reviewed studies — published from 2017 to 2026, 3 from 2024 or later — selected as the most relevant from 13 studies that passed quality screening, drawn from 51 papers retrieved from a database of over 500 million.
Sources used in this answer
Programmed death ligand 1 (PD-L1) PET/CT imaging to evaluate tumour saturation during atezolizumab treatment
Serial 89Zr-atezolizumab PET/CT during atezolizumab treatment showed day 7 tumour SUVmax of 5.5 in cycle 1 and 5.4 in cycle 2 (N = 13, P = 0.88), with lower on-treatment lesion and spleen uptake and higher blood pool than pretreatment imaging, suggesting partial PD-L1 tumour saturation [1].
Dynamic metrics-based biomarkers to predict responders to anti-PD-1 immunotherapy
This foundational paper argues that static PD-L1 immunohistochemistry is unreliable because PD-L1 expression and antitumour immune reactivity are dynamic, and proposes dynamic IFN-γ secretion metrics from peripheral lymphocytes as an alternative predictive strategy in a mouse melanoma model [2].
Abstract CT017: First-in-human PET imaging with the PD-L1 antibody 89Zr-atezolizumab
This precursor first-in-human study administered 89Zr-atezolizumab at about 1 mg and performed up to four PET scans at 1 hour and days 2, 4 and 7, establishing the feasibility of PD-L1 antibody PET before treatment [3].
68Ga-Labeled Peptide for Noninvasive Quantifying Tumor Exposure of PD-L1 Therapeutics.
This validation study developed the peptide tracer 68Ga-HF12 and showed that tumour uptake fell from 5.33 ± 0.25 %ID/g before atezolizumab to 3.2 ± 0.17 %ID/g after two treatments and stabilised at 3.16 ± 0.15 %ID/g after five, consistent with reduced available PD-L1 binding sites [5].
18F-BMS-986229 PET to Assess Programmed-Death Ligand 1 Status in Gastroesophageal Cancer.
This limitation study of 18F-BMS-986229 PET in 10 gastroesophageal cancer patients found concordance with pathologic PD-L1 assessment in 88% of biopsied lesions but also showed that 71% of patients with tracer accumulation had lesions without uptake, highlighting intrapatient heterogeneity [6].