Esophageal squamous cell carcinoma is one of the most common and lethal malignancies worldwide, and it is particularly prevalent in parts of East Asia. Treatment decisions increasingly hinge on PD-L1, a protein that tumor cells use to dampen the immune system’s attack. When PD-L1 is abundant on the tumor surface, immune checkpoint inhibitors, drugs that block the PD-1/PD-L1 interaction, tend to work well. The standard way to measure PD-L1 is immunohistochemistry, a laboratory technique in which a biopsy sample is stained with antibodies and scored using a combined positive score, or CPS, that reflects how many tumor and immune cells display the protein. Patients whose tumors score at or above a defined threshold are typically considered candidates for immunotherapy.

The problem is that this assessment requires tissue, and obtaining it is not always straightforward. Esophageal tumors sit deep in the chest, biopsies sample only a fragment of a heterogeneous tumor, and PD-L1 expression can vary across space and change over time. A noninvasive surrogate that could be read off a scan performed before treatment would be enormously valuable, allowing clinicians to stratify patients earlier and perhaps to repeat the assessment as the disease evolves. That is the premise behind the new study, led by Jiazhong Ren and Zheng Fu of the Department of Medical Imaging and Yongbin Cui and Yong Yin of the Department of Radiation Oncology Physics and Technology at Shandong Cancer Hospital and Institute, affiliated with Shandong First Medical University and the Shandong Academy of Medical Sciences.

The physics behind the approach is elegant. 18F-FDG is a radioactive analog of glucose that is taken up by metabolically active cells, and tumors, with their voracious appetite for energy, typically accumulate far more of the tracer than surrounding healthy tissue. The PET scanner converts this accumulation into standardized uptake values, or SUVs, which quantify how much tracer concentrates in a given region. The researchers measured a family of these parameters: the maximum standardized uptake value (SUVmax), which captures the single hottest pixel in the tumor; the mean (SUVmean) and peak (SUVpeak) values, which describe average and locally averaged intensity; the standard deviation (SUVsd), which reflects how heterogeneous the uptake is; and two volumetric measures, the metabolic tumor volume (MTV), the physical volume of metabolically active tumor, and total lesion glycolysis (TLG), which combines volume with intensity.

The team enrolled 89 patients with esophageal squamous cell carcinoma in this retrospective study. All patients underwent 18F-FDG PET/CT and had PD-L1 expression evaluated by immunohistochemistry on primary tumor tissue before receiving any anti-tumor treatment. The investigators also recorded a comprehensive set of clinicopathological characteristics, including age, gender, smoking status, drinking status, tumor location, vertical tumor length, TNM stage, and pathological grade. For the analysis, patients were divided into two groups: those with PD-L1-negative tumors, defined as a combined positive score below 10, and those with PD-L1-positive tumors, defined as a CPS of 10 or higher, a cutoff commonly used in clinical decision-making.

The results showed a clear statistical pattern. Four of the six metabolic parameters, SUVmax, SUVmean, SUVpeak, and SUVsd, were significantly higher in patients whose tumors expressed PD-L1 than in those whose tumors did not, with all comparisons reaching statistical significance at P values below 0.05. In other words, tumors that glowed more brightly and more unevenly on the PET scan were more likely to be PD-L1 positive. The volumetric measures, MTV and TLG, did not show the same association, suggesting that it is the intensity and heterogeneity of glucose metabolism, rather than the sheer bulk of metabolically active tissue, that tracks with this immune marker.

To quantify how well these imaging features could actually predict PD-L1 status, the researchers turned to receiver operating characteristic curve analysis, a standard statistical technique that plots true positive rates against false positive rates across all possible cutoff values. The area under the curve, or AUC, summarizes discriminative ability on a scale from 0.5, equivalent to a coin flip, to 1.0, perfect prediction. SUVmax achieved an AUC of 0.636 with a P value of 0.029, and SUVmean achieved an AUC of 0.628 with a P value of 0.044, both statistically significant but modest in magnitude. The optimal cutoff values were 19.74 for SUVmax and 10.01 for SUVmean, thresholds that could be applied to future patients, though with considerable uncertainty.

The team then used univariate logistic regression to examine each parameter’s independent association with PD-L1 positivity. SUVmax was the strongest predictor with a P value of 0.006, followed by SUVmean at 0.011, SUVpeak at 0.022, and SUVsd at 0.024. However, a technical complication emerged: these four SUV-derived parameters are all computed from the same underlying uptake distribution and are therefore severely multicollinear, meaning they carry heavily overlapping information and cannot be meaningfully combined in a single statistical model. Because of this severe multicollinearity, the researchers selected only SUVmax, the strongest individual predictor, for the final single-predictor logistic model, which retained its significance at P = 0.006.

The authors are explicit about what these numbers do and do not mean. An AUC of 0.636 represents discriminative performance that is only modestly better than chance, far below what would be needed for a reliable diagnostic test. Consequently, the study concludes that SUVmax should serve only as an auxiliary imaging reference rather than a dependable noninvasive substitute for pathological PD-L1 detection in patients with esophageal squamous cell carcinoma. This honest framing matters, because the temptation to overread an imaging-biomarker association is real, and premature clinical adoption of an imprecise surrogate could misdirect treatment decisions with serious consequences.

Nevertheless, the research adds to a growing body of work exploring how imaging phenotypes reflect molecular tumor characteristics, a field sometimes described as radiomics or imaging genomics. The biological plausibility is compelling: PD-L1 expression is often driven by inflammatory and oncogenic signaling pathways that also reprogram tumor metabolism, so a link between glucose avidity and immune checkpoint ligand expression is not surprising. The study was supported by the National Natural Science Foundation of China under grants 12275162 and 12575365, was approved by the Ethics Committee of the Cancer Hospital Affiliated to Shandong First Medical University in accordance with the Declaration of Helsinki, and was published open access on 29 September 2026. Future studies with larger cohorts, prospective designs, and possibly machine learning approaches that combine multiple imaging features may refine the predictive power enough to bring scan-based biomarker assessment closer to the clinic.

Subject of Research: The association between 18F-FDG PET/CT metabolic imaging parameters and PD-L1 expression in esophageal squamous cell carcinoma

Article Title: The predictive value of 18F-FDG PET/CT metabolic parameters for PD-L1 expression in esophageal squamous cell carcinoma

Article References: Ren, J., Cui, Y., Fu, Z., & Yin, Y. (2026). The predictive value of 18F-FDG PET/CT metabolic parameters for PD-L1 expression in esophageal squamous cell carcinoma. BMC Cancer. https://doi.org/10.1186/s12885-026-17026-7

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17026-7

Keywords: 18F-FDG PET/CT, PD-L1, esophageal squamous cell carcinoma, SUVmax, metabolic parameters, immunotherapy biomarker, immunohistochemistry, combined positive score, cancer imaging, radiomics, tumor metabolism, BMC Cancer