The research team set out to answer a deceptively simple question: can the immune cells already present inside a bladder tumor, before any BCG is given, reveal how a patient will fare after therapy? Current clinical practice relies heavily on clinicopathological features such as tumor grade, stage, size and the number of tumors to estimate recurrence risk, most famously through risk tables developed by the European Organization for Research and Treatment of Cancer (EORTC) and the Club Urologico Español de Tratamiento Oncológico (CUETO). These tools are useful, but they capture nothing about the biology of the immune response that BCG is supposed to provoke. The Chinese team suspected that the missing ingredient lay in the tumor immune microenvironment, the dense ecosystem of immune cells, signaling molecules and structural cells that surrounds and infiltrates every malignancy.

To test that hypothesis, the investigators assembled a retrospective cohort of 80 patients with non-muscle invasive bladder cancer who had been treated with BCG instillation. From each patient, tumor tissue collected before the start of therapy was available, allowing the researchers to examine the pre-treatment immune context rather than the post-treatment state, which is far more informative for prediction. Using immunohistochemical staining, a laboratory technique that uses antibodies to mark specific cell types on tissue sections so they can be counted under a microscope, the team quantified the densities of key immune cell populations across different tumor compartments, distinguishing between cells located within the tumor epithelium and those in the surrounding stroma, the supportive connective tissue that frames the malignancy.

The analytical pipeline combined classical survival statistics with modern variable-selection methods. The researchers first applied univariate Cox regression, a statistical technique that estimates the association between each candidate marker and the time to recurrence, to identify which immune measurements carried prognostic weight. They then used Lasso analysis, a regression method that shrinks the coefficients of weak predictors toward zero and thereby filters out noise in datasets where the number of variables can easily overwhelm the number of patients. This two-step approach converged on two independent prognostic markers: the total density of CD4-positive T cells across the tumor, and the density of CD68-positive macrophages specifically within the tumor stroma, abbreviated CD68s.

Both markers make biological sense in the context of BCG therapy. CD4-positive T cells are helper lymphocytes that orchestrate adaptive immune responses, coordinating the attack on infected or malignant cells and sustaining immunological memory. BCG instillation is thought to work partly by recruiting and activating these cells within the bladder wall, so a tumor already rich in CD4-positive cells may be primed to respond. CD68-positive macrophages, meanwhile, are phagocytic innate immune cells whose behavior in tumors is famously double-edged: some subsets support anti-tumor inflammation while others suppress immunity and promote tissue remodeling. Their density in the stromal compartment, the corridor through which immune cells migrate into the tumor, appears to encode information about how the tumor microenvironment will handle the inflammatory storm that BCG triggers.

Rather than relying on either marker alone, the team built a combined risk score, a weighted mathematical function of CD4 and CD68s values, and found that this composite measure demonstrated better predictive accuracy and discriminatory power than either marker in isolation. Discrimination, in the language of survival analysis, refers to a model’s ability to correctly rank patients by their risk, separating those who will recur from those who will remain disease-free. The improvement from combining two immune markers of different lineages, one adaptive and one innate, suggests that recurrence after BCG is not governed by a single cell type but by the interplay between the two arms of the immune system within the tumor microenvironment.

The next step was to translate the risk score into something a clinician could actually use at the bedside. The researchers constructed a nomogram, a graphical calculating tool that assigns points to each predictor and sums them into a probability estimate, integrating the immune-based risk score with standard clinicopathological features. When benchmarked against the established EORTC, CUETO and European Association of Urology (EAU) risk stratification models, the new nomogram showed slightly improved predictive performance and, importantly, greater net clinical benefit. Net benefit is a decision-analytic measure that weighs the consequences of true and false predictions, asking not merely whether the model is statistically accurate but whether acting on its predictions would help more patients than it would harm.

The clinical implications of such a tool are substantial. Patients identified before therapy as likely to be BCG-unresponsive could be steered toward alternative strategies earlier, avoiding months of ineffective instillations during which their disease might progress. Conversely, patients predicted to respond well could be spared more aggressive interventions, such as early radical cystectomy, the surgical removal of the bladder, which is sometimes recommended for high-risk disease but carries significant morbidity and a lasting impact on quality of life. Because the required measurements come from immunohistochemical staining of routinely collected biopsy or resection specimens, the approach could in principle be implemented in pathology laboratories without exotic equipment, though the authors’ cohort of 80 patients is modest and the findings will need validation in larger, independent and ideally prospective cohorts before entering clinical guidelines.

The study also contributes to a broader shift in oncology, in which the tumor immune microenvironment has moved from a subject of basic research to a source of actionable biomarkers. Similar immune-contexture scoring systems have transformed prognostication in colorectal cancer, melanoma and other malignancies, and immune checkpoint inhibitors have made the composition of immune infiltrates a central consideration in treatment selection across many tumor types. Applying this lens to BCG therapy is particularly apt, because BCG is, in essence, one of the oldest forms of immunotherapy, and its variable success has long hinted at the importance of individual immune differences. By showing that two simple, pre-treatment immune measurements can sharpen recurrence prediction beyond what tumor stage and grade alone can offer, the work points toward a more personalized approach to a disease that affects millions worldwide.

For now, the message from Guangzhou is one of cautious optimism. The model built by Lu, Zhong and their colleagues does not replace the established clinical risk tools, but it refines them, adding a biological dimension that those tools have always lacked. If larger studies confirm that CD4 and stromal CD68 densities reliably forecast BCG response, urologists may soon order a panel of immune stains alongside the routine pathology workup, and the decision of whether to proceed with BCG, escalate surveillance or consider more radical options could be made with far greater confidence. In a field where treatment failure is discovered only in retrospect, a test that sees it coming could change the trajectory of care for a substantial share of bladder cancer patients.

Subject of Research: An immune microenvironment-based predictive model for recurrence after BCG therapy in non-muscle invasive bladder cancer

Article Title: Development of an immune microenvironment-based model to predict recurrence after BCG therapy in non-muscle invasive bladder cancer

Article References: Lu, J.-L., Zhong, L., Wen, Y.-L., Cai, T.-N., Qin, Z.-K., Tan, L., & Ye, Y.-L. (2026). Development of an immune microenvironment-based model to predict recurrence after BCG therapy in non-muscle invasive bladder cancer. BMC Cancer. https://doi.org/10.1186/s12885-026-17083-y

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17083-y

Keywords: bladder cancer, BCG therapy, tumor immune microenvironment, CD4 T cells, CD68 macrophages, recurrence prediction, nomogram, immunohistochemistry, biomarkers, EORTC, prognostic model, urothelial carcinoma