Predictive Value of Corrected 18 F-FDG PET/CT Baseline Parameters for Primary DLBCL Prognosis: A Single-center Study

Objective  The purpose of this study was to evaluate the prognostic significance of corrected baseline metabolic parameters in fluorodeoxyglucose positron emission tomography imaging ( 18 F-FDG PET/CT) for 3-year progression-free survival (PFS) in patients with primary diffuse large B cell lymphoma (DLBCL). Patients and Methods  Retrospective clinical and pathological data were collected for 199 patients of DLBCL diagnosed between January 2018 and January 2021. All patients underwent 18 F-FDG PET/CT scans without any form of treatment. The corrected maximum standardized uptake value (corSUVmax), corrected mean standardized uptake value (corSUVmean), corrected whole-body tumor metabolic volume sum (corMTVsum), and corrected total lesion glycolysis of whole body (corTLGtotal) were corrected using the SUVmean in a 1-cm diameter mediastinal blood pool (MBP) from the descending thoracic aorta of patients. Kaplan–Meier survival curves and Cox regression were used to examine the predictive significance of corrected baseline metabolic parameters on 3-year PFS of patients. The incremental values of corrected baseline metabolic parameters were evaluated by using Harrell's C-indices, receiver operating characteristic, and Decision Curve Analysis. Results  The multivariate analysis revealed that only the National Comprehensive Cancer Network (NCCN)-International Prognostic Index (IPI) and corMTVsum had an effect on 3-year PFS of patients ( p  < 0.05, respectively). The Kaplan–Meier survival analysis demonstrated significant differences in PFS between the risk groups classified by corSUVsum, corMTVsum, and corTLGtotal (log-rank test, p  < 0.05). The predictive model composed of corMTVsum and corTLGtotal surpasses the predictive performance of the model incorporating MTVsum and TLGtotal. The optimal performance was observed when corMTVsum was combined with NCCN-IPI, resulting in a Harrell's C index of 0.785 and area under the curve values of 0.863, 0.891, and 0.947 for the 1-, 2-, and 3-year PFS rates, respectively. Conclusion  The corMTVsum offers significant prognostic value for patients with DLBCL. Furthermore, the combination of corMTVsum with the NCCN-IPI can provide an accurate prediction of the prognosis.


Introduction
Diffuse large B cell lymphoma (DLBCL) is the most prevalent form of non-Hodgkin lymphoma, constituting approximately 30 to 35% of cases. 1,2It is a complex and heterogeneous malignancy characterized by a diffuse pattern of growth and aggressive behavior.While standard treatment with rituximab, cyclosphosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) induces complete response in approximately 75 to 80% of patients diagnosed with DLBCL, a subset of approximately 10 to 15% of patients will present with primary refractory disease, and an additional 20 to 25% will experience relapse after initial response. 3he International Prognostic Index (IPI) 4 and its modifications, age-adjusted IPI (aa-IPI), 5 and the National Comprehensive Cancer Network-IPI (NCCN-IPI), 6 have served as established prognostic models for DLBCL, helping clinicians make informed decisions and guiding treatment strategies for patients.While other prognostic factors, activated B cell-like subtype historically carries a less favorable prognosis 7 and protein expression patterns of MYC, BCL2 and BCL6, and MYC rearrangements were predictive of outcome and provided prognostic information independent of the IPI for overall survival (OS) and event-free survival. 8However, most of the above indices were based on serum or fluorescence in situ hybridization, required a combination of several factors.
][11] Various parameters that were measured during this process, such as maximum standard uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), have demonstrated significant prognostic value in DLBCL.][14][15] Accurate prognostic markers are essential for prognostic stratification, guiding timely interventions with novel treatments to improve their prognosis.Standardizing segmentation thresholds and measurement methods may also have significant impacts on clinical decision-making and ultimately improve patient outcomes with DLBCL.The principal objectives of our study were to evaluate the prognostic performance of corrected baseline tumor metabolic parameters and to determine the additional value in DLBCL patients, compared with the uncorrected baseline tumor metabolic parameters and NCCN-IPI.

Patients
Between January 2018 and January 2021, we retrospectively examined 199 patients diagnosed with DLBCL through lymph node or target lesion biopsy with postoperative pathological confirmation.All patients had complete clinical and follow-up data with a first-line R-CHOP chemotherapy regimen administered after the definitive diagnosis.Exclusion criteria included (1) a previous association with other malignancies, (2) progression or transformation from inert lymphoma, (3) prior malignancy-related treatment (including surgery, radiotherapy, and targeted drug therapy), ( 4) DLBCL of the central nervous system, (5) and comorbidities of acute and chronic active infectious diseases.
Follow-up information was obtained based on routine follow-up clinical examination, CT, and/or FDG-PET/CT scans.Progression-free survival (PFS) was defined as the duration from the initial diagnosis of the disease to the time of disease progression, recurrence, or death.

Clinical/Histological Parameters
In this retrospective study, we evaluated the gender, age, erythrocyte sedimentation rate (ESR), hemoglobin level, number of extranodal invasion sites, lactate dehydrogenase (LDH), and ferritin levels of 199 patients diagnosed with DLBCL.Disease stages were classified according to the Ann Arbor staging criteria as I to IV, while patients were classified as having either A or B symptoms based on the presence or absence of unexplained fever of 38°C or higher, nocturnal sweating, and weight loss of 10% or more within 6 months.Physical status was evaluated using the Zubrod-Eastern Clinical Oncology Group (ECOG)-WHO (ZPS, 5-point scale) criterion, with scores ranging from 0 to 5. The IPI, aa-IPI, and the NCCN-IPI were calculated based on different criteria.

Positron Emission Tomography Imaging Image Acquisition
Prior to undergoing an Ingenuity TF 128 PET/CT examination, patients were required to fast for 6 to 8 hours and maintain fasting blood glucose levels below 11.1 mmol/L.The imaging agent 18 F-FDG was injected via a superficial vein on the back of the hand at a dose of 3.7 MBq/kg, following which patients were instructed to rest calmly in the PET/CT lounge for approximately 50 to 60 minutes before the whole-body examination.Patients were advised to drink 500 mL of water and urinate prior to the exam.A whole-body PET/CT scan was conducted in the supine position, scanning from the cranial vault to the middle femur.This was followed by a 15-to 20-index of 0.785 and area under the curve values of 0.863, 0.891, and 0.947 for the 1-, 2-, and 3-year PFS rates, respectively.Conclusion The corMTVsum offers significant prognostic value for patients with DLBCL.Furthermore, the combination of corMTVsum with the NCCN-IPI can provide an accurate prediction of the prognosis.
minutes PET scan while still in the same posture, followed by a CT scan with 128 rows at 120 kV, 100 to 150 mA, 0.75-mm pitch, and 3.75-mm layer thickness.PET data were corrected for attenuation and iteratively reconstructed using the selfcontained software once the scan had been completed.Finally, all PET and CT images were transferred to a MEDEX computer workstation (Phillips Medical Systems), which constructs fused PET/CT images for further analysis.

Positron Emission Tomography Imaging Image Analysis
To ensure accuracy, two diagnostic PET/CT physicians with more than 10 years of experience at the associate senior level or above collaborated to measure the SUVmean of systemic lesions as well as the SUVmean in the MBP of the descending thoracic aorta, which is 1 cm in diameter on PET/CT images.In the event of a disagreement, a consensus was achieved through consultation.The SUVmean of the lesion was corrected by the PET/CT diagnostician using the SUVmean of the patients' MBP of the descending thoracic aorta.The corSUVmean was calculated using the following formula: corSUVmean ¼ patient whole-body lesion SUVmean Â average MBP SUVmean/patients MBP SUVmean.The ratio of corSUVmean to SUVmean was used to determine corSUVmax.The region of interest (ROI) was determined using the threshold of 41% of corSUVmax proposed by the European Society of Nuclear Medicine.Corrected whole-body tumor metabolic volume sum (corMTVsum) was measured, and corrected total lesion glycolysis of whole body (corTLGtotal) ¼ Σ (corSUVmean Â corMTVsum).Lastly, based on the median, baseline metabolic parameters were separated into low-and high-risk groups.

Statistical Analysis
This study conducted a statistical evaluation using R (version.4.2.1).Continuous variables were presented as mean AE standard deviation or median with interquartile range (IQR), and categorical variables were reported as frequency and proportions.The R packages "survival" were used for Kaplan-Meier survival analyses and "rms" for Cox proportional hazards regression.Univariate regression analysis was conducted to obtain variables significantly related to PFS.In the uncorrected or corrected baseline parameters, significant variables identified through univariate regression analysis were used to construct the uncorrected and corrected model.All significant variables through univariate regression analysis were integrated into a multivariate regression analysis, then selected the significant variables to construct the combined models.Harrell's C-indices, ROC curve and Decision Curve Analysis (DCA) were used to compare the performance of the prediction.All tests were two-sided, with a significance level set at p-value less than 0.05.

Patient Characteristics
The main characteristics of the patients are summarized in ►Table 1.This study included a total of 199 patients with  corMTVsum (HR 7.099, 95% CI 2.642-19.078)were identified as the significant factors that influenced the 3-year PFS of patients, as shown in ►Table 2.

Survival Analysis
Kaplan-Meier estimates showed that patients could be stratified into low-or high-risk according to median of corrected baseline tumor parameters.the highest AUC (AUC 0.863, 0.891, and 0.947 for 1-, 2-, and 3-year PFS).Notably, all these models above superior performance compared with NCCN-IPI (AUC 0.768, 0.818, and 0.898 for 1-, 2-, and 3-year PFS).Detailed results are provided in ►Fig. 3and ►Table 3. DCA demonstrated a notably higher net benefit for both the combined and corrected models compared with the uncorrected model and NCCN-IPI across diverse threshold probabilities in predicting PFS (►Fig. 4).

Discussion
The MTV and TLG volumetric parameters have been recognized as reliable indicators for survival prediction in DLBCL patients.However, the lack of a standardized methodology for calculating these values hinders their broader clinical application.To improve accuracy and provide more meaningful results in practice, it is crucial to develop a normalized methodology for calculating these parameters.In this study, we have corrected baseline metabolic parameters to take into account the radiochemical purity and dose of each patient.Our study showed that baseline corSUVsum, corMTVsum, and corTLGtotal were significantly correlated with DLBCL.Improved prediction accuracy (Harrell's C rose from 0.731 to 0.769) for PFS was achieved by incorporating corMTVsum and corTLGtotal over MTVsum and TLGtotal.It also revealed that the combined model incorporating corMTVsum and NCCN-IPI provided superior predictive performance MTV has been regarded as a potential biomarker for the development of initial PET-adapted approaches in follicular lymphoma. 16However, debates have arisen regarding the predictive value of baseline metabolic parameters in patients   This study utilizes a hybrid measurement method that combines the fixed threshold of corSUVmax method with the threshold of 41% corSUVmax method to calculate baseline parameters.The determination of the fixed threshold for corSUVmax is based on the average SUVmean in the MBP form the Deauville criteria.It is a 5-point scale model, implemented to standardize the visual assessment of treatment response in DLBCL through interim and post-treatment FDG-PET/CT. 20In this scale model, the FDG uptake in the MBP and liver is referenced. 21Barrington et al 22 delineated the methodology and location for measuring SUV in reference regions, but there remains a lack of uniformity in the measurement details.Variations include the choice between SUVmean and SUVmax, as well as the use of ROI versus volume of interest.To enhance reproducibility, we adopted a 1-cm diameter ROI within the MBP from the descending thoracic aorta to correct the SUVmean.In contrast, the SUVmean of the liver was not used to normalize the SUVmean of the whole-body lesions since many patients had liver involvement, and the SUVmean of the affected liver was substantially higher than the normal liver.Additional correction methods, such as partial volume correction (PVC), have demonstrated enhanced accuracy in quantifying PET images in cancer studies. 23,24A previous investigation, the application of PVC improved the accuracy of TLG for predicting survival in patients with relapsing/refractory lymphoma before and after radioimmunotherapy. 25Our study utilized the SUVmean of the MBP to eliminate the heterogeneity caused due to unequal radioactivity during the FDG injection or varying doses of patients owing to their body weight or other factors.After correction, the corSUVmax emerged as a predictive variable for PFS in univariate regression analysis than SUVmax without correction.With a 100% follow-up rate, cor SUVmax, corMTVsum, and corTLGtotal were all significantly correlated with PFS in DLBCL before R-CHOP treatment.
To evaluate 3-year PFS in patients with primary DLBCL, we combined all available clinical variables that could impact prognosis with PET/CT baseline metabolic parameters.B symptoms, 26 anemia, 27 elevated levels of ESR 28 and ferritin 29 have been proved to be poor prognostic factors in DLBCL patients.However, conflicting results have been reported regarding their impact on prognosis.Our study confirms that anemia and elevated levels of ferritin are associated with a poor prognosis, but we did not observe B symptoms and ESR levels as significant predictors of outcomes.
Our study has several limitations that should be noted.This study was a single-center retrospective study, a small sample size and heterogeneous population may have introduced potential bias and confounding variables that could have influenced the identified clinical prognostic factors.Additionally, we did not perform molecular pathology analysis, including assessments of MYC, BCL-2, and BCL-6 protein expression and gene rearrangements, known as prognostic factors.This limitation arose from the unavailability of data for all patients.To further explore the potential of combining baseline tumor metabolic characteristics and radiomics for personalized prognosis in DLBCL patients and to assess whether it could eventually replace the NCCN-IPI, future prospective research is warranted.

Conclusion
Corrected PET-CT baseline metabolic parameters can serve as critical prognostic factors for primary DLBCL patients.A combined model incorporating corMTVsum with the NCCN-IPI can provide additional prognostic value to PFS in primary DLBCL patients.Ultimately, this approach may lead to the development of more precise and individualized treatment plans, thereby enhancing the accuracy of prognostication and improving patient outcomes in primary DLBCL patients.

Conflict of Interest
None declared.
with DLBCL and the existing studies have presented conflicting results.Shagera et al17 discovered that prechemotherapy MTV was an independent prognostic factor in addition to NCCN-IPI.The combination of baseline TMTV and NCCN-IPI may have the potential to improve the prognosis of DLBCL patients with high-risk NCCN-IPI group.On the other hand, Sasanelli et al18 found that MTV was the only independent factor in the prognosis of DLBCL patients, while TLG was not significant.Zhou et al15 reported that baseline TLG was an independent predictor of survival in DLBCL patients before R-CHOP treatment.Nevertheless, diverging from these results, Gallicchio et al19 revealed that only SUVmax was a predictor of PFS in DLBCL patients, and lower SUVmax values were associated with worse outcomes.To ensure the reliability and comparability of PET imaging parameters across centers, we propose a collaborative effort similar to the Image Biomarker Standardization Initiative.

Fig. 4
Fig. 4 Decision curve analysis for the NCCN-IPI, uncorrected model, corrected model, and combined prediction model in the progressionfree survival.Net benefit was calculated as follows: Net benefit ¼ true positive rate À (false positive rate Â weighting factor).Weighting factor ¼ threshold probability/1 À threshold probability.NCCN-IPI, National Comprehensive Cancer Network-International Prognostic Index.

Table 1
Baseline demographic and clinical characteristics of patients duration of 19.5 (IQR 13.0-27.0)months.A total of 25 patients died, with 3-year PFS and OS rates of 45.7 and 12.6%, respectively.Extranodal involvement was observed in 147 cases, with the gastrointestinal region being the most commonly affected (52 cases), followed by the bone (34 cases) the spleen (22 cases).A patient with multiple extranodal involvement is detailed in ►Fig. 1.
World Journal of Nuclear Medicine Vol.23 No. 1/2024 © 2024.The Author(s).Predict Primary DLBCL Prognosis Li and Liu et al. 35 confirmed diagnosis of DLBCL, with a mean age of 62.91 AE 11.73 years and a range of 15 to 87 years.Complete follow-up information was obtained for all patients in the trial as of January 2021.Of the all patients, 91 patients experienced progression, relapse, or death, with a median time of 6.0 (IQR 4.0-9.0)months.In total of 108 patients did not experience such an event by the follow-up time, which had a median According to univariate regression analysis, sex, age, ESR, B symptoms, number of extralymph node invasions, SUVmax and SUVmean and corSUVmean were not the factors influencing patients 3-year PFS (p > 0.05).Conversely, lactate dehydrogenase, hemoglobin, ferritin, Eastern Clinical Oncology Group Performance Status, Ann Arbor staging, IPI, aa-IPI, NCCN-IPI, MTVsum, and TLGtotal, corSUVmax, corMTVsum, and corTLGtotal were identified as significant factors that impacted PFS (p < 0.05).Following the univariate regression analysis, only significant variables were subjected to further multivariate regression analysis.Based on the results, it was noted that NCCN-IPI (hazard ratio [HR] 3.077, 95% confidence interval [CI] 1.147-8.250)and

Table 3
Receiver operating characteristic analysis curve for prediction of 1-, 2-, 3-year progression-free survival in prediction models Abbreviations: AUC, area under the curve; CI, confidence interval; corMTVsum, corrected whole-body tumor metabolic volume sum; corTLGtotal, corrected total lesion glycolysis of total body; NCCN-IPI, National Comprehensive Cancer Network-International Prognostic Index; PFS, progression-free survival.