Validation of prognostic scores in patients with <scp>HIV</scp> ‐related diffuse large B‐cell lymphoma: The value of peripheral blood parameters—A study from the Spanish Lymphoma Group <scp>GELTAMO</scp>

To the Editor, Although combined antiretroviral therapy (cART) has dramatically reduced the incidence of acquired immunodeficiency syndrome (AIDS)-defining malignancies, Human Immunodeficiency Virus (HIV)-related diffuse large B-cell lymphoma (DLBCL) remains a leading cause of cancer-related morbidity and mortality among people with HIV (PWH).1 Standard first-line therapy is the same as that of the general population, typically CHOP (cyclophosphamide, doxorubicin hydrochloride, oncovin, and prednisone) or dose-adjusted EPOCH (etoposide, prednisone, oncovin, cyclophosphamide, and doxorubicin hydrochloride) regimens combined with rituximab.1 The widespread use of chemoimmunotherapy combined with cART in HIV-related DLBCL has narrowed the outcome gap between PWH and HIV-negative patients, but survival still remains heterogeneous, underscoring the need for reliable prognostic tools for this population.2 In the pre-cART era, prognosis was determined largely by HIV-related parameters, such as CD4+ T-cell count, viral load and prior AIDS-defining illnesses.3 With current cART, these factors play a less important role, and outcomes are increasingly influenced by lymphoma-related variables.4 Consequently, it is essential to reassess whether the classic prognostic scores used in the general population—based on tumour and host clinical parameters—are equally valid in HIV-related DLBCL. The International Prognostic Index (IPI), introduced in 1993, remains the cornerstone for risk stratification in DLBCL, combining age, stage, lactate dehydrogenase (LDH), performance status and extranodal involvement.5 The revised IPI (R-IPI) adapted this model to the rituximab era, grouping patients into three risk categories while maintaining strong prognostic discrimination.6 The National Comprehensive Cancer Network (NCCN)-IPI refined these parameters by introducing LDH stratification and weighting specific extranodal sites, offering improved survival prediction.7 The Spanish Lymphoma Group (GELTAMO)-IPI added β2-microglobulin (β2M) as a marker of tumour burden,8 and the more recent GELTAMO-IPI-2 integrated peripheral blood parameters—lymphocyte-to-monocyte ratio (LMR) and red cell distribution width (RDW).9 Although these scores are widely validated in HIV-negative DLBCL patients, none had been tested in PWH. We conducted a multicentre retrospective study of 125 patients with HIV-related DLBCL from 17 Spanish centres (2002–2021). Eligibility criteria included histologically confirmed DLBCL, HIV infection and first-line treatment with rituximab-CHOP and cART. Radiotherapy was administered when indicated. Clinical and biological variables at diagnosis were collected, including age, sex, Ann Arbor stage, B symptoms, bulky disease, Eastern Cooperative Oncology Group (ECOG) performance status, LDH, β2M, RDW, absolute lymphocyte and monocyte counts (to calculate LMR), CD4+ T-cell count and HIV viral load. All participating centres are members of the Spanish GELTAMO network, which applies uniform diagnostic and therapeutic criteria for DLBCL. Laboratory parameters were obtained using validated local assays, and quantitative variables were categorized according to each centre's reference ranges to account for inter-laboratory variability. Therefore, no additional data harmonization was required. Each prognostic score—IPI,5 R-IPI,6 NCCN-IPI,7 GELTAMO-IPI8 and GELTAMO-IPI-29—was calculated as per original definitions. Complete remission (CR) was assessed after first-line therapy. Overall survival (OS) and progression-free survival (PFS) were estimated using Kaplan–Meier methods and compared by log-rank tests. Logistic regression identified predictors of CR, and Cox models were applied for OS and PFS. Model discrimination was measured by Harrell's C-index. Relationship among the five prognostic scoring systems was assessed using Spearman's rank correlation for ordinal comparisons and Cohen's Kappa coefficient for measuring agreement in dichotomized categories (low/low–intermediate versus high/intermediate–high). This study was approved by the Ethics Committee of Germans Trias i Pujol Hospital, Badalona, Spain (PI-20-330). The median age was 47 years (range 19–76); 85% were younger than 60; and 15% were ≥60 years. Most were male (83%) and presented with advanced disease (stage III/IV: 80%), B symptoms (58%) and ECOG ≥2 (37%). About half had detectable HIV viraemia and CD4+ T-cell counts <0.2 × 109/L at lymphoma diagnosis. Baseline characteristics are shown in Table S1. All patients received a median of six R-CHOP cycles, and 13 received additional radiotherapy. The overall CR rate was 72%. After a median follow-up of 7.3 years, the 5-year OS (95% confidence interval [CI]) and PFS (95% CI) were 62% (52%–70%) and 56% (47%–65%) respectively. In univariable analyses, B symptoms absence, ECOG <2, absence of bulky mass, normal LDH, normal lymphocyte count, LMR ≥3 and normal RDW were associated with higher CR probability (Table S2). In multivariable analysis, ECOG <2 (p = 0.009), absence of bulky disease (p = 0.011) and LMR ≥3 (p = 0.004) remained independently significant. The impact of LMR <3 on OS and PFS is shown in Figure 1. For survival outcomes, ECOG ≥2, advanced stage, lymphopenia and LMR <3 predicted worse OS and PFS (Table S3). Bulky disease, >1 extranodal site and high RDW predicted inferior PFS (Table S3). Outcomes did not differ between patients diagnosed in the early (2002–2010) and late (2011–2021) cART eras. Multivariable models confirmed ECOG ≥2 and LMR <3 as independent predictors of shorter OS and ECOG ≥2, bulky disease and LMR <3 as independent predictors of shorter PFS (Table S4). All five prognostic indices stratified patients into groups with distinct OS and PFS, though categories were merged into low/low–intermediate versus high/intermediate–high due to the limited sample sizes. Figure 2 and Figure S1 show the risk groups for OS and PFS, respectively, according to the five clinical score systems. Among them, GELTAMO-IPI-2 showed the best discriminative performance for both OS and PFS, with C-index 0.654. According to this model, patients in the high/intermediate–high-risk group had a 4.7-fold higher risk of death (95% CI, 2.36–9.29) and a 5.8-fold higher risk of progression (95% CI, 3.04–11.14) compared with those in the low/low–intermediate group (Table S5). Sensitivity analyses restricted to complete cases and using multiple imputation yielded consistent results, confirming the robustness of the main findings. Association analyses demonstrated statistically significant correlation and concordance among all five prognostic indices. The strongest correlation was observed between IPI and R-IPI (ρ = 0.985), followed by IPI and NCCN-IPI (ρ = 0.834). Similar results were obtained by the κ index. As expected, GELTAMO-IPI and GELTAMO-IPI-2 showed slightly lower but still significant correlations with the classical indices, reflecting their distinct variable composition. This study validates the prognostic utility of five established indices—IPI, R-IPI, NCCN-IPI, GELTAMO-IPI and GELTAMO-IPI-2—in patients with HIV-related DLBCL treated with rituximab-containing immunochemotherapy and cART. Despite the challenges of immunosuppression, our data demonstrate that the same clinical variables driving prognosis in the general population retain significance in PWH. Moreover, models incorporating host immune and inflammatory markers provided superior risk discrimination, probably emphasizing the dual contribution of lymphoma biology and immune competence to outcomes. ECOG performance status ≥2 and LMR <3 emerged as the most robust independent predictors of poor OS and PFS, consistent with findings in HIV-negative cohorts.10-12 ECOG remains a universal marker of frailty and treatment tolerance, while the prognostic impact of LMR likely reflects the balance between lymphocyte-mediated antitumour immunity and monocyte-driven inflammation and immunosuppression. In the context of HIV-related immune dysregulation, LMR may be particularly relevant, as impaired lymphocyte function further compromises immune surveillance and favours a more permissive tumour microenvironment.13 The integration of LMR into GELTAMO-IPI-2 likely explains its better predictive capacity compared with traditional indices. Bulky disease also retained independent prognostic significance for PFS, supporting its inclusion in modern risk models.9 Conversely, age >60 years did not influence survival in our cohort, likely due to the predominance of younger patients. Despite its established role in DLBCL, β2M showed only a trend towards shorter PFS, possibly confounded by HIV-related inflammation.8 In our cohort of PWH, RDW was associated with adverse lymphoma outcomes, consistent with its previously described role as a marker of systemic inflammation in HIV-negative patients with lymphoma.14 Among the five tested indices, GELTAMO-IPI-2 showed the most consistent trend towards better discrimination for both OS and PFS. Its incorporation of β2M, LMR and RDW may reflect the influence of tumour burden and host immune-inflammatory status, factors of particular relevance in PWH. Classical indices (IPI, R-IPI, NCCN-IPI) remained robust tools for general risk stratification, though they might be less responsive to immunologic variations. Our results align with prior work suggesting that immune recovery under cART reduces the prognostic weight of HIV-specific parameters while reinforcing the role of lymphoma-intrinsic and host clinical variables.2, 4 The demonstration that, in HIV-related DLBCL, both classic and new indices perform comparably to the general population supports the application of therapeutic standards to this population as well as their inclusion in clinical trials. The superior performance of GELTAMO-IPI-2 highlights the need for future models that explicitly integrate immune and inflammatory dynamics. This study has several limitations. Its retrospective nature and the lack of centralized pathology review may have introduced heterogeneity in diagnostic and laboratory assessments. Missing data for some variables, particularly β2M, LMR and RDW, limited the calculation of all prognostic scores in every patient, potentially affecting comparative analyses. Moreover, the study spanned nearly two decades (2002–2021), during which cART regimens and HIV management evolved substantially, possibly influencing immune recovery and survival.15 Nevertheless, the homogeneity of lymphoma treatment mitigates some of these limitations and supports the internal consistency of the results. In conclusion, the present study confirms that classical prognostic models (IPI, R-IPI, NCCN-IPI, GELTAMO-IPI and GELTAMO-IPI-2) are applicable to HIV-related DLBCL in the cART era, with GELTAMO-IPI-2 providing the best survival discrimination. ECOG performance status and LMR remain independent prognostic factors, underscoring the importance of host immunity in disease control. These data reinforce that PWH treated with cART with DLBCL should be risk-stratified and treated according to the same principles as HIV-negative patients and highlight GELTAMO-IPI-2 as a practical and biologically relevant model for this setting. MH, MM, JMR, and JTN designed the study and wrote the paper; MH, AR-D, MB-O, RC, AS, FC-V, MA, BF-L, AJ-U, AG, MI, SH, PA, DC, CM, AM, SG-V, JMR and JTN looked after patients and contributed with data collection and report; MM performed the statistical analysis. All authors reviewed the study data, read and approved the manuscript. The authors would like to thank the members of the GELTAMO group for their collaboration and continued support. We are also grateful to the patients and their families. This work was supported by Josep Carreras Leukaemia Foundation, Spain. The data that support the findings of this study are available from the corresponding author upon reasonable request. Figure S1. Progression-free survival for risk groups (risk groups are unified in two risk categories) defined by five clinical scoring systems. (A) International Prognostic Index (IPI). (B) Revised International Prognostic Index (R-IPI). (C) NCCN-IPI. (D) GELTAMO-IPI. (E) GELTAMO-IPI-2. Table S1. Baseline characteristics of the patients in the series and risk stratification according to the different scoring systems. Table S2. Univariable analysis for complete remission. Table S3. Univariable analysis for overall survival and progression-free survival. Table S4. Multivariable models for overall survival (OS) and progression-free survival (PFS). Table S5. Models for overall survival (OS) and progression-free survival (PFS) by clinical scoring systems. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.