MALT lymphoma has long been considered one of the more indolent members of the non-Hodgkin lymphoma family. Arising from mucosa-associated lymphoid tissue in the stomach, lungs, salivary glands, and other barrier organs, it often grows slowly and responds well to treatment. Yet clinicians have repeatedly observed that a subset of patients experiences early progression or relapse despite the disease’s reputation for indolence. The challenge has always been knowing, at the moment of diagnosis, which patients belong to that unlucky subset. A new study from Union Hospital, Tongji Medical College of Huazhong University of Science and Technology in Wuhan, published in BMC Medical Imaging, offers a fresh answer by fusing metabolic imaging data with standard clinical and pathological variables into a single predictive tool.
The research team, led by Yanmei Han and corresponding authors Zairong Gao and Xiao Zhang, set out to address a well-recognized gap in lymphoma prognostication. Existing risk stratification systems for MALT lymphoma, including the MALT lymphoma International Prognostic Index, rely on clinical parameters such as age, disease stage, and lactate dehydrogenase levels, but they do not incorporate information from metabolic imaging. That omission matters, because fluorodeoxyglucose positron emission tomography combined with computed tomography captures something that blood tests and pathology slides cannot: the glucose appetite of the tumor in living tissue. The maximum standardized uptake value, or SUVmax, quantifies how avidly the most active region of a lesion consumes the radiolabeled glucose analog, providing a whole-body metabolic fingerprint of the disease at baseline.
To build their model, the investigators retrospectively analyzed 98 patients with histologically confirmed MALT lymphoma who underwent baseline FDG PET/CT scanning between January 2020 and December 2024. From the pool of available clinical, laboratory, pathological, and imaging measurements, they assembled 17 candidate predictors. Rather than forcing all of them into a regression equation, the team applied least absolute shrinkage and selection operator regression, a statistical technique known as LASSO that penalizes model complexity and drives the coefficients of uninformative variables to zero. Ten-fold cross-validation was used to tune the procedure, meaning the data were repeatedly split so that the variable selection would not simply memorize the quirks of one particular sample.
Out of the 17 candidates, five variables survived the LASSO filter: patient age, the primary site of disease, SUVmax, the Ki-67 proliferation index, and Eastern Cooperative Oncology Group performance status. Each of these carries a plausible biological story. Age and performance status reflect the resilience of the patient and the aggressiveness of the clinical course. Ki-67 measures the fraction of tumor cells actively dividing, a direct readout of proliferative drive. The primary disease site has long been associated with differing behaviors across MALT lymphoma locations, with gastric and non-gastric presentations following distinct trajectories. SUVmax, the sole imaging variable to make the cut, adds the metabolic dimension that conventional indices have ignored.
These five predictors were then combined in a multivariable Cox regression, the standard framework for modeling time-to-event outcomes such as progression-free survival. The result was distilled into a nomogram, a graphical scoring instrument that allows a clinician to assign points to each patient characteristic and read off an estimated probability of remaining progression-free. On the data used to build the model, the five-variable score achieved a concordance index of 0.814, with a 95 percent confidence interval of 0.738 to 0.891. The concordance index, or C-index, expresses the probability that a patient who progresses earlier is correctly assigned a higher risk score than a patient who progresses later, with 0.5 representing random guessing and 1.0 representing perfect ranking.
The authors were careful not to stop at that flattering apparent figure. Apparent performance is notoriously optimistic, because the same data that build a model inevitably flatter it. To correct for this, the team ran a full-pipeline bootstrap procedure with 500 resamples, in which the entire modeling process, from variable selection onward, was repeated on each resampled dataset. This approach estimates how much the model’s measured accuracy is inflated by chance fitting, and the optimism-corrected C-index settled at 0.740, with a 95 percent uncertainty interval of 0.662 to 0.757. In practical terms, the model retains moderate discriminative ability after accounting for statistical overfitting, a level of honesty that many published prognostic models never report.
Time-dependent receiver operating characteristic analysis added a temporal dimension to the evaluation. The area under the time-ROC curve was 0.766 at one year, 0.872 at two years, and 0.883 at three years, indicating that the model’s ability to separate patients who progress from those who remain in remission actually strengthens over the first few years of follow-up. That pattern is clinically meaningful for an indolent lymphoma, where the most consequential question is often not whether relapse will ever occur but whether it will arrive early enough to justify intensified surveillance or alternative first-line strategies.
Perhaps the most intriguing result emerged from an exploratory re-stratification exercise. The team focused on patients classified as intermediate risk by the MALT-IPI, the current standard prognostic index, and asked whether the new integrated model could split this apparently homogeneous group into meaningful subgroups. It could. Within the intermediate-risk category, the nomogram separated patients into lower-risk and higher-risk groups that showed clearly different observed progression-free survival. If confirmed in larger and independent cohorts, this finding suggests that metabolic imaging can refine the coarse resolution of existing clinical indices, potentially sparing lower-risk intermediate patients unnecessary treatment intensity while flagging higher-risk ones for closer monitoring.
The study’s limitations are those inherent to its design. It was retrospective, single-center, and modest in size, with 28 progression events among 98 patients, and the validation was internal rather than external. The authors themselves describe the risk separation within the intermediate-risk subgroup as exploratory. External validation on independent cohorts from different institutions and populations will be essential before the nomogram can influence routine clinical decisions. Nevertheless, the methodological rigor, including the full-pipeline bootstrap and the transparent reporting of optimism-corrected performance, sets a standard that prognostic modeling studies in oncology imaging do not always meet.
Beyond its immediate findings, the work signals a broader shift in how nuclear medicine and hematology are converging. FDG PET/CT has already transformed staging and response assessment in aggressive lymphomas, and quantitative metabolic parameters are increasingly being treated as biomarkers in their own right rather than as pictures for the radiologist’s eye alone. By demonstrating that a single imaging metric can stand alongside age, proliferation index, and performance status in a validated survival model, the Wuhan team provides a template for integrating molecular imaging into prognostic algorithms across indolent cancers. For patients with MALT lymphoma, the hope is that a routine baseline scan, already performed in many centers, will one day do double duty: mapping the disease and forecasting its course, so that the small minority destined for early progression can be identified while their options are still widest.
Subject of Research: An integrated FDG PET/CT and clinicopathological nomogram for predicting progression-free survival in MALT lymphoma
Article Title: Development and internal evaluation of an integrated [18F]FDG PET/CT-clinicopathological nomogram for predicting progression-free survival in MALT lymphoma
Article References: Han, Y., Zhang, Y., Wu, R., Hu, F., Gao, Z., & Zhang, X. (2026). Development and internal evaluation of an integrated [18F]FDG PET/CT-clinicopathological nomogram for predicting progression-free survival in MALT lymphoma. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-02847-6
Image Credits: AI Generated
DOI: 10.1186/s12880-026-02847-6
Keywords: MALT lymphoma, FDG PET/CT, nomogram, SUVmax, progression-free survival, prognosis, Ki-67 index, LASSO regression, Cox regression, bootstrap validation, MALT-IPI, nuclear medicine
News Source: Nathaniel Bowman. (October 4, 2026). PET Scan Data Joins Clinical Factors to Predict Relapse in MALT Lymphoma. Scienmag.
