Translate this page into:
AIIMS Mantle Cell Lymphoma Overall Survival (AMOS) & AIIMS Mantle Cell Lymphoma Event-Free Survival (AMES) scores: Development of novel indigenous survival prediction models & risk stratification for mantle cell lymphoma in the Indian population
For correspondence: Dr Ajay Gogia, Department of Medical Oncology, All India Institute of Medical Sciences, New Delhi 110029, India e-mail: ajaygogia@gmail.com
-
Received: ,
Accepted: ,
Abstract
Background & objectives
Mantle cell lymphoma (MCL) is an aggressive subtype of non-Hodgkin lymphoma (NHL) with considerable heterogeneity in clinical outcomes. Existing prognostic models, such as the MCL International Prognostic Index (MIPI), demonstrate limited applicability in the Indian population due to differences in demographic and clinical profiles. This study was designed to develop indigenous prognostic tools to predict the survival of MCL patients from India.
Methods
We conducted a retrospective analysis of 92 patients with MCL diagnosed and treated over the past decade. Clinical, laboratory, and treatment parameters were reviewed to identify predictors of survival outcomes. Two prognostic models, the AIIMS MCL Overall Survival (AMOS) and Event-Free Survival (AMES) scores, were developed using multivariate logistic regression. Model performance was evaluated using receiver operator characteristic (ROC) curve analysis, with stratification into risk categories based on empiric score cut-offs.
Results
AMOS and AMES models demonstrated robust predictive performance with area under ROC curve values of 0.75 and 0.82, respectively. The AMOS score incorporated Eastern Co-operative Oncology Group performance status (ECOG PS) and splenomegaly to predict two yr overall survival, while the AMES score included ECOG PS, B-symptoms, advanced stage, and rituximab maintenance (RM) to predict two yr event-free survival. Risk stratification revealed significant survival differences, with a median overall survival of nine months in the high-risk AMOS group compared to 75.5 months in the low-risk group. Similar trends were observed for event-free survival.
Interpretation & conclusions
AMOS and AMES scores are novel indigenous MCL survival prediction models. They are effective, pragmatic tools for risk stratification of MCL patients.
Keywords
AMOS
AMES
NHL
mantle cell lymphoma
rituximab maintenance
Mantle cell lymphoma (MCL) is an aggressive subtype of non-Hodgkin lymphoma (NHL). It accounts for 3-10 per cent of all NHL cases worldwide1,2. Indian retrospective series have reported a frequency of 7 per cent of NHL3. Overexpression of cyclin D1 due to translocation t(11;14)(q13;q32) is central in pathogenesis4. Despite recent advances in therapeutic strategies, including the use of rituximab and targeted agents, the prognosis of MCL remains heterogeneous. Overall survival varies widely depending on disease presentation and treatment response5. Globally, prognostic models such as MCL International Prognostic Index (MIPI) and its variants have been pivotal in stratifying patients and guiding treatment decisions6. MIPI incorporates clinical parameters such as age, Eastern Co-operative Oncology Group performance status (ECOG PS), lactate dehydrogenase (LDH) levels, and white blood cell count to predict survival outcomes7. However, these models, primarily developed in Western populations, exhibit limitations when applied to diverse ethnic and socio-economic groups. Unique demographic and clinical characteristics with limited access to rituximab maintenance (RM) therapy due to financial constraints, reduce their applicability in the Indian context8-11. While some studies have attempted to modify existing prognostic indices to better suit local populations, such efforts are unseen from this part of the world. Consequently, there is a pressing need for indigenous prognostic tools that account for the unique clinical and socio-economic realities of the Indian healthcare system.
This study primarily aimed to develop novel prognostic models: AIIMS MCL Overall Survival (AMOS) and Event-Free Survival (AMES) scores, tailored specifically for the Indian MCL population. We also intended to develop risk categories to stratify prognosis.
Materials & Methods
This retrospective study was undertaken by the departments of Medicine and Medical Oncology, All India Institute of Medical Sciences (AIIMS), New Delhi, India after obtaining ethical clearance from the Institute Ethics Committee. All the participants were informed about the plan of the study, and necessary consent was taken to utilise their clinico-pathological and laboratory data for analyses.
Study setting and participants
Data on MCL patients diagnosed and treated at AIIMS-Institute Rotary Cancer Hospital (IRCH) from January 2013 to December 2022 (10 yr) were retrieved from digital records. Ninety-two patients, for whom a histological diagnosis was made and who completed scheduled treatment according to the institutional protocol, were included in the study. All patients were followed up till June 2024. Patients with incomplete case records and those who did not complete treatment and were lost to follow up were excluded.
Patient evaluation
Demographics, clinical history including B-symptoms, physical examination, assessment of international prognostic index (IPI) and MCL IPI (MIPI) scores, and ECOG PS were reviewed at baseline. Clinical staging was performed using the conventional Ann Arbor criteria. Laboratory parameters included complete blood count, LDH, and serum albumin at presentation. Number and sites of involved nodal and extra-nodal (EN) regions were analysed by computed tomography of chest, abdomen, pelvis, and/or whole-body positron emission tomography scan. Patients were assessed to have received either rituximab (R)-based or non-R-based induction chemotherapy. Use of RM was noted. Overall survival (OS) was defined as treatment initiation to death from any cause or last follow up. Event-free survival (EFS) was defined as the time from treatment initiation to relapse or death from any cause.
Statistical analysis and development of survival prediction models
Data were analysed by STATA version 14.0 software (StataCorp, College Station, TX). Multivariate logistic regression analysis of clinical, laboratory, and treatment variables, death (OS), and death or relapse (EFS) at two yr, was conducted using the stepwise inclusion method. The threshold value for addition to the models was P=0.05. The prognostic significance of individual predictors was expressed as adjusted odds ratio (aOR) with 95% confidence interval (CI). All regression coefficients were divided by the lowest coefficient and multiplied by 10 to obtain a score for each predictor (to get the lowest score as 10). Using derived individual scores for predictors, total survival prediction scores (AMOS and AMES) were obtained for each subject. Receiver operator characteristic (ROC) curve analysis was used to evaluate the discriminatory power of models and establish optimal cut-offs. Performance was reported using the area under the ROC curve (AUROC), sensitivity, specificity, predictive values, and likelihood ratios. Patients were divided into risk categories based on empiric score cut-offs. Kaplan-Meier analysis estimated survival. Difference in survival across risk categories was assessed by the log-rank test. P value <0.05 defined statistical significance.
Results
Baseline patient characteristics
Median age of presentation was 60 yr (interquartile range; IQR: 52.5–66 yr, range: 36-83 yr), three times more common in males. Generalised lymphadenopathy, B-symptoms, bulky disease (lymph node size ≥7.5 cm), and splenomegaly were seen in 80 (87%), 48 (52%), 18 (20%), and 47 (51%) patients, respectively. EN involvement was observed in 71 (77%) patients; more than one site of involvement was seen in 38 (41%) patients. Bone marrow was the most common EN site involved, followed by the colon. The majority of patients, 85 (92%), were in the advanced stages (III/IV) of the disease at presentation. Thirty-four (37%) had ECOG PS ≥2. Fifteen (16%), 50 (54%), and 41 (45%) patients presented with high-risk IPI, MIPI, and revised-MIPI (R-MIPI) scores, respectively. The median R-MIPI score was 5.9 (IQR: 5.4 – 6.5). Of 92 patients, 25 (27%) had anaemia (haemoglobin <10 g/dL), 31 (34%) had thrombocytopenia (platelet count <1,50,000/cmm), and 31 (34%) had leucocytosis (total leucocyte count >10,000/cmm). High LDH (≥ULN; upper limit of normal) and hypoalbuminemia (serum albumin <3.5 g/dL) were seen in 23 (25%) and 29 (32%) cases, respectively. R-based induction chemotherapy was used in 80 (87%) patients, while only 30 (33%) received RM due to financial constraints.
Survival estimates
Median duration of follow up was 63 (IQR: 25.5 – 85.5) months. Half of the patients were surviving beyond 43 months. Median EFS was 22 (IQR: 15 – 30) months. Estimated 1 yr, 2 yr, 3 yr overall survival was 81 per cent, 59 per cent, and 49 per cent, respectively. Event-free survival at corresponding time intervals was 70 per cent, 71 per cent, and 33 per cent, respectively (Figs. 1A-B).
- Kaplan-Meier (A) overall survival estimate, and (B) event-free survival estimate.
Derivation of AMOS and AMES scores
In multivariate logistic regression analysis, ECOG PS ≥2 (a) and splenomegaly (b) were associated with death at two yr, while ECOG PS ≥2 (a), B-symptoms (c), advanced stage (d), non-use of RM (e) were associated death or relapse at two yr. Regression analysis yielded specific score against each predictor from their aOR and regression coefficient (Table I). The AMOS score was derived as 12.3 (a) + 10.0 (b). The AMES score was derived as 21.3 (a) + 12.3 (c) + 14.3 (d) + 10.0 (e). Value for presence or absence of a predictor (a) through (e) was considered in the formula as 1 or 0, respectively.
| Predictors | Death at 2 yr | |||
|---|---|---|---|---|
| uOR* | aOR* | RC* | CS** | |
| ECOG PS ≥2 (a) | 5.3 (2.1 – 13.4) | 4.8 (1.8 – 12.8) | 1.6 (0.6 – 2.6) | 12.3 |
| Splenomegaly (b) | 4 (1.7 – 9.5) | 3.6 (1.4 – 9.2) | 1.3 (0.3 – 2.2) | 10 |
| Predictors | Death or relapse at 2 yr | |||
| uOR* | aOR* | RC* | CS** | |
| ECOG PS ≥2 (a) | 9.9 (3.1 – 31.8) | 4.9 (1.4 – 17.6) | 1.6 (0.3 – 2.9) | 21.3 |
| B-symptoms (c) | 4.2 (1.7 – 10.2) | 3.9 (1.4 – 11.1) | 1.4 (0.3 – 2.4) | 12.3 |
| Advanced stage (d) | 10.5 (1.2 – 90.9) | 10.6 (1 – 109.8) | 2.4 (0 – 4.7) | 14.3 |
| Non-use of RM (e) | 3.4 (1.4 – 8.4) | 3.0 (1 – 9.2) | 1.1 (0 – 2.2) | 10 |
Performance of AMOS and AMES models
The mean AMOS score was 9.7 ± 8.4. The optimal cut-off was 11.2, with an AUROC of 0.75 (Fig. 2A). AMOS score >11.2 predicted death by two yr with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 57.1 per cent, 80 per cent, 70.6 per cent, and 69 per cent, respectively. The mean AMES score was 38±14. The optimal cut-off was 34.6, with an AUROC of 0.82 (Fig. 2B). AMES score >34.6 predicted death or relapse by two years with sensitivity, specificity, PPV, and NPV of 70.9 per cent, 81.1 per cent, 84.8 per cent, and 65.2 per cent, respectively. Detailed performance metrics have been shown in table II.
- Receiver operator characteristic (ROC) curve for (A) AMOS model, and (B) AMES model.
| Parameters* | AMOS score | AMES score |
|---|---|---|
| Sn (%) | 57.1 (41-72.3) | 70.9 (57.1-82.4) |
| Sp (%) | 80.0 (66.3-90) | 81.1 (64.8-92) |
| PPV (%) | 70.6 (52.5-84.9) | 84.8 (71.1-93.7) |
| NPV (%) | 69.0 (55.5-80.5) | 65.2 (49.8-78.6) |
| PLR | 2.86 (1.55-5.27) | 3.75 (1.88-7.46) |
| NLR | 0.54 (0.37-0.78) | 0.36 (0.23-0.56) |
| AUROC | 0.75 (0.65-0.84) | 0.82 (0.73-0.91) |
Development of risk categories for AMOS and AMES models
Maximum, minimum AMOS and AMES scores were 0, 22.3, and 0, 57.9, respectively. Patients were divided into low (LR), intermediate (IR), and high-risk (HR) for overall survival based on AMOS score: 0 – 10, >10 – 12.3, and >12.3–22.3, respectively. Similar risk categories were developed for event-free survival based on AMES score: 0–14.3, >14.3–36.6, >36.6–57.9, respectively. Median OS was significantly poor in HR [9 (IQR: 4–23) months] compared to IR [26 (IQR: 14.5 – not reached; NR) months] and LR [75.5 (IQR: 43 – NR) months] AMOS groups, log-rank P<0.0001 (Fig. 3A). Similarly, median EFS was poor in HR [12 (IQR: 7–22) months] compared to IR [24 (IQR: 13.5–33.5) months] and LR [49 (IQR: 25 – NR) months] AMES groups, log-rank P<0.0001 (Fig. 3B).
- Stratification of (A) overall survival based on AMOS risk categories, and (B) event-free survival based on AMES risk categories. LR, low-risk; IR, intermediate-risk; HR, high-risk.
Discussion
MCL is a rare, distinct subtype of NHL, characterised by aggressive behaviour and heterogeneity in clinical outcomes9,11,12. Prognostic scoring systems are critical for stratifying patients, predicting outcomes, and tailoring therapeutic approaches. In this study, we developed two novel prognostic scores, AMOS and AMES, specifically designed for the Indian population. These models addressed the unmet need for indigenous tools calibrated to this demographic. They incorporated clinical and treatment variables reflective of the unique presentation and therapeutic landscape in India.
AMOS and AMES scores demonstrated robust prognostic accuracy in our study cohort. The AMOS model incorporated ECOG PS and splenomegaly; it effectively predicted two yr overall survival with a PPV of 71 per cent. AMES model, which included ECOG PS, B-symptoms, advanced stage, and non-use of RM, demonstrated even stronger predictive ability for two yr event-free survival with PPV of 85 per cent. Patients scoring above the AMOS and AMES cut-offs were three times more likely to experience mortality and four times more likely to experience mortality or relapse, respectively, within two yr. These findings underscore the utility of both models in real-world settings.
Stratification of patients across AMOS and AMES risk categories demonstrated a significant difference in survival even beyond two yr. High-risk AMOS had a survival probability of 15 per cent at 3 years, compared to 33 per cent and 78 per cent in intermediate and low-risk patients. Three-year probability of death or relapse was 69 per cent, 32 per cent, and 12 per cent for low, intermediate, and high-risk AMES categories, respectively. This reflected the aggressiveness of the lymphoma; median OS was nine months in the high-risk AMOS group, and median EFS was 12 months in the high-risk AMES group.
MIPI remains the benchmark for MCL worldwide1. While effective, MIPI has shown variable applicability across diverse populations13. In Indian settings, its discriminatory power is often limited due to differences in baseline clinical characteristics and treatment accessibility. Patients from this part of the world present, a decade earlier, at an advanced stage with significant extra-nodal involvement8,10. In addition, financial constraints limiting the use of rituximab maintenance (RM), as observed in our cohort, are inadequately accounted for by MIPI.
While IPI and National Comprehensive Cancer Network (NCCN)-IPI are valuable in aggressive lymphomas, their utility in MCL-specific prognosis remains suboptimal6,7. In contrast, AMOS and AMES relied on readily available clinical and treatment variables at presentation, making them pragmatic tools for the Indian healthcare context. Revised MIPI (MIPI-c) incorporates Ki-67 proliferation index; though it has higher sensitivity (75% vs. 57%) in predicting OS compared to AMOS score, its specificity is lower (54% vs. 80%)4. AMES gave special weightage to RM. This increased robustness of the model, as use of RM conferred improved EFS across MCL cohorts5,14.
Despite the promising findings, our study had several limitations. Being conducted at a single tertiary care center, the generalisability of results to other healthcare settings may be limited. Though our MCL cohort was one of the largest in India, the small sample size limited the performance of internal validation. Treatment heterogeneity was intrinsic as patients were treated with varying rituximab-based induction chemotherapy regimens. This reflected real-world diversity at the expense of potential confounders. Thirteen percent did not receive rituximab-based induction; however, this did not affect two yr overall and event-free survival after adjusting for other variables.
Previous attempts at prognostic modelling in Indian MCL patients have been sparse. To our knowledge, AMOS and AMES, respectively, are the first indigenous overall and event-free survival prediction models with good discriminatory power for Indian patients with MCL. Stratification based on model risk categories demonstrated a significant survival difference. The inclusion of clinical and treatment parameters makes them easy to use in routine outpatient and lymphoma clinics for patient prognostication. However, further validation of these models in larger multicentric prospective cohorts is needed to concretise the robustness. Adaptation of these models in other low and middle-income countries shall enhance their global relevance further.
Acknowledgment
Authors acknowledge Medical Records section of department of Medical Oncology, All India Institute of Medical Sciences, New Delhi for retrieving patient files and electronic data.
Declaration of previous presentation of the work
The study findings were presented as a poster at the Society of Hematologic Oncology (SOHO) 2024 Annual Meeting at Houston, Texas, USA held between September 4-7, 2024.
Financial support & sponsorship
None.
Conflicts of Interest
None.
Use of Artificial Intelligence (AI)-Assisted Technology for manuscript preparation
The authors confirm that there was no use of AI-assisted technology for assisting in the writing of the manuscript and no images were manipulated using AI.
References
- A new prognostic index (MIPI) for patients with advanced-stage mantle cell lymphoma. Blood. 2008;111:558-65.
- [Google Scholar]
- Mantle-cell lymphoma: a population-based clinical study. J Clin Oncol. 1996;14:1269-74.
- [Google Scholar]
- Detection of translocation t(11;14)(q13;q32) in mantle cell lymphoma by fluorescence in situ hybridization. Am J Pathol. 1999;154:1449-52.
- [Google Scholar]
- Prognostic value of Ki-67 Index, cytology, and growth pattern in mantle-cell lymphoma: Results from randomized trials of the European mantle cell lymphoma network. J Clin Oncol. 2016;34:1386-94.
- [Google Scholar]
- Treatment outcomes and roles of transplantation and maintenance rituximab in patients with previously untreated mantle cell lymphoma: results from large real-world cohorts. J Clin Oncol. 2023;41:541-54.
- [Google Scholar]
- A predictive model for aggressive non-Hodgkin’s lymphoma. N Engl J Med. 1993;329:987-94.
- [Google Scholar]
- International prognostic indices in diffuse large B-cell lymphoma: a comparison of IPI, r-IPI, and NCCN-IPI. Blood. 2020;135:2041-8.
- [Google Scholar]
- Mantle cell lymphoma: a North Indian tertiary care centre experience. Asian Pac J Cancer Prev. 2016;17:4583-6.
- [Google Scholar]
- Lymphoma subtypes in India: a tertiary care center review. Clin Exp Med. 2021;21:315-21.
- [Google Scholar]
- Clinicopathological and immunohistochemical profile of mantle cell lymphoma: an institutional experience. Cureus. 2021;13:e16534.
- [Google Scholar]
- Profile of non-Hodgkin lymphoma: an Indian perspective. South Asian J Cancer. 2018;7:162.
- [Google Scholar]
- SEER Cancer Statistics Review, 1975-2013. Bethesda, MD: National Cancer Institute; 2015.
- Prognostic model for mantle cell lymphoma in the rituximab era: a nationwide study in Japan. Br J Haematol. 2015;170:657-68.
- [Google Scholar]