Federico, Massimo

Abstract Background: Although progression-free survival (PFS) in early-stage (ES) classic Hodgkin Lymphoma (cHL) is high, there remains a subset of patients (pts) at higher risk of worse outcomes, particularly in those with unfavorable risk disease. Leveraging pt-level data from within the HoLISTIC (Hodgkin Lymphoma International STudy for Individual Care) Consortium, the E-HIPI was developed in clinical trial pts and validated in real-world registry pts to predict 2-year (y) PFS (Rodday. NEJM Evidence 2025). Continuous, objective, and readily measurable pre-treatment variables were considered, with sex, maximum tumor diameter, albumin, and hemoglobin included as significant predictors. Using the E-HIPI, interactive calculators for pt risk, comparison, and stratification were created (https://rtools.mayo.edu/holistic_ehipi/). Although the E-HIPI was developed and validated in both unfavorable and favorable ES cHL pts, the clinical predictors are primarily markers of unfavorable disease. In the development and initial validation cohorts, 2y PFS was 93.7% and 90.3%, respectively, the C-statistic was 0.63 in both cohorts, calibration was strong with slight underprediction of events in the validation cohort, and the E-HIPI outperformed the historic binary classification of patients as favorable or unfavorable using EORTC criteria. However, since pts in the original E-HIPI development and validation were primarily treated with ABVD chemotherapy protocols, additional validation is needed, particularly among pts with unfavorable ES cHL who were treated initially with more intensive chemotherapeutic platforms (e.g., BEACOPP-based). Thus, we conducted an external validation of the E-HIPI in pts treated on recent unfavorable ES cHL GHSG trials. Methods: The validation cohort included 2367 adults (ages 18 to 60y) with newly-diagnosed unfavorable cHL from randomized phase 3 trials conducted by the GHSG (HD14 von Tresckow. JCO 2012; and HD17 Borchmann. Lancet 2021). We calculated the predicted probability of 2y PFS events based on the E-HIPI model equation (scale 1-100%). Model performance was assessed using discrimination and calibration. Discrimination, which is the model's ability to separate high and low risk pts, was assessed using Harrell's C-statistic. Calibration, which is the model's ability to accurately predict absolute risk, was assessed by comparing observed and predicted probabilities of 2y PFS events within quintiles of predicted probabilities and estimating calibration-in-the-large (predicted minus observed event rate) and the calibration slope. Results: For pts treated on unfavorable GHSG trials, mean age was 33y, 95% had stage II disease, mean baseline E-HIPI score (i.e., predicted 2y PFS event rate) was 6.0% (SD=2.3%), and the observed 2y PFS rate was 96.2% (95%CI: 95.4-97.0%). The C-statistic was 0.65 (95% CI: 0.59-0.70), calibration-in-the-large was 0.02, and the calibration slope was 1.08 (95% CI: 0.52, 1.64). Within the lowest risk E-HIPI quintile, the observed 2y PFS event rate was 1.6%, compared with 6.1% in the highest risk quintile. Conclusions: In external validation of the recently published E-HIPI in pts with cHL treated on unfavorable GHSG trials using more intensive BEACOPP-based protocols, discrimination was similar to the original development and validation cohorts. Collectively, the E-HIPI is a treatment-agnostic prediction tool across varied treatment regimens for unfavorable ES cHL pts. Better 2y PFS in the GHSG trials than in the E-HIPI development and initial validation cohorts may have contributed to mild misclassification, likely reflecting differences in treatment regimens. This may be resolved with recalibration or incorporating granular treatment variables into future models. Additional research is needed to understand E-HIPI performance in pts with favorable disease who have even lower event rates as well as to more fully predict individualized pt outcomes across varied chemotherapy platforms and radiation received for all ES cHL pts.

Background: Although advanced stage cHL typically has excellent disease outcomes, 20-30% of patients (pts) experience treatment failure. In addition, decision-making is challenging given varying treatment options and incomplete prognostic data across the disease course. Pre-treatment factors, such as those incorporated into the Advanced Stage Hodgkin Lymphoma International Prognostication Index (A-HIPI), predict 5 year (y) overall survival (OS) and progression-free survival (PFS) (Rodday JCO 2023). Other than interim PET (iPET) results, it is unknown what factors influence cHL disease states mid-therapy. Using multistate modeling (MSM) and individual pt data from the HoLISTIC Consortium, we refined prognostication across the cHL disease course, specifically assessing the relationships between A-HIPI, iPET and end of treatment (EOT) response, and whether the A-HIPI and iPET provide independent prognostic information. Methods: We analyzed 1240 pts aged 18-65y with newly diagnosed Stage IIB, III or IV cHL treated on 2 advanced stage PET-adapted trials (RATHL, SWOG0816). All pts first received 2 cycles of ABVD; those with a negative iPET (Deauville score (DS) ≤3) received ABVD or AVD, and those with a positive iPET (DS >3) received BEACOPP. In contrast to other methods, MSMs incorporate multiple disease states into one model, account for censoring and competing risks, and estimate transitions between disease states using different covariates. Our MSM had 6 states: diagnosis, negative iPET, positive iPET, EOT non-progression (e.g., DS ≤ 3), treatment failure, and death. Treatment failure included progression or relapse. All pts start at the diagnosis state and can transition to other states without return to prior states. Transition states with <5 events were excluded. The MSM was estimated using a multivariable Cox model with censoring at 5y. Covariates were the 5y PFS A-HIPI (comprised of stage and continuous age, lymphocyte count and albumin), which was included for all transitions, and iPET, which was included for transitions from EOT non-progression and from treatment failure. Higher scores on the 5y PFS A-HIPI indicate lower risk of progression or death; the A-HIPI was modeled for a 1 standard deviation (SD) increase. We calculated probabilities of 5y treatment failure for 4 sample pts based on low (70) and high (85) A-HIPI scores, positive and negative iPET, and EOT non-progression. Results: Median age was 33y (Q1=25, Q3=45), 56% were male, 23% were stage IIB, 40% III and 37% IV, and mean A-HIPI was 77 (SD=7). Median follow-up was 75 months (Q1=53, Q3=92). At iPET, 80% (95% CI 78%, 83%) were negative and 16% (95% CI 14%, 18%) were positive. At EOT, 90% (95% CI 89%, 92%) were in the non-progression state. At 5y, 75% (95% CI 73%, 78%) were in the non-progression state, 16% (95% CI 14%, 19%) had experienced treatment failure, and 9% (7%, 11%) had died. Better A-HIPI scores (per SD) were associated with lower rates of transitioning from diagnosis to positive iPET (HR=0.80, p<0.01) and from positive iPET to treatment failure (HR=0.58, p=0.04). Better A-HIPI scores (per SD) were also associated with lower rates of transitioning from EOT non-progression to treatment failure (HR=0.78, p<0.01) and from treatment failure to death (HR=0.64, p<0.01), adjusting for iPET. Positive iPET was associated with an increased rate of transitioning from EOT non-progression to treatment failure (HR=1.90, p<0.01), adjusting for A-HIPI. The probability of 5y treatment failure for 4 sample pts was: 25% for low A-HIPI and positive iPET; 20% for high A-HIPI and positive iPET; 13% for low A-HIPI and negative iPET; and 12% for high A-HIPI and negative iPET. Conclusions: We created a novel MSM that refines prognostication across the cHL disease course by incorporating pre- and post-treatment factors (e.g., iPET) to estimate transitions to future disease states. Although the A-HIPI was developed using pre-treatment factors to predict 5y OS and PFS, we found that it was also associated with transitions to interim disease states, including iPET and EOT response. The A-HIPI and iPET each provided independent prognostic information, supporting the use of both in estimating cHL outcomes. Future analyses will develop clinical prediction models using MSMs that also incorporate varying frontline and salvage treatments and treatment-specific late effects, with the goal of providing individualized information across the disease course.

Background: Prognostic modeling allows personalized risk prediction for individual patients (pt). The A-HIPI model in advanced stage classical Hodgkin lymphoma (AS-HL), developed and validated by the HoLISTIC Consortium (www.hodgkinconsortium.org), generates the individualized probability of a progression-free survival (PFS) event or death (OS) within the first 5 years (y) from diagnosis in pts based on continuous variables (www.qxmd.com/calculate/calculator_869/a-hipi). Clinical prognostic tools in lymphoma (eg, IPS, IPI, FLIPI, etc) typically use groupings of categorical values to define risk. Grouping a continuous value often results in loss of information, and most tools are not predictive for individual pts. However, discrete groupings have clinical utility & practicality for a) defining pt populations for clinical trials & real world studies, b) stratification within clinical trials, and c) crafting treatment guidelines. We studied potential approaches for utilizing the A-HIPI model to generate risk groups with input on strengths & limitations from the HoLISTIC modeling team & clinical experts. Methods: The A-HIPI model was developed via TRIPOD guidelines on 4,022 pts treated on 8 international clinical trials for AS-HL (Rodday. JCO 2023). External validation was performed on a dataset of 1,431 pts from 4 prospective registries. The 5y PFS (PFS5) in the A-HIPI development dataset was 77% (95% CI: 76-78); the 5y OS (OS5) was 92% (95% CI: 91-93). This represented the average outcome for a pt with AS-HL pt naïve to other clinical data at presentation. The distribution of PFS5 & OS5 predictions were heavily skewed (ie, asymmetric distribution) in both the A-HIPI discovery and validation dataset. While not unexpected due to the excellent PFS & OS in this disease setting, this presents challenges in the delineation of risk groups as depicted below . Three approaches were examined for the generation of A-HIPI risk groups. Proposed cutoffs were defined using the distribution of A-HIPI risk scores and data from the model-building cohort (ie, clinical trials). Validation was done using the A-HIPI validation cohort (ie, HL registries). Results: Approach 1: Risk groups based on clinical thresholds. Clinicians were queried what estimates of PFS5 would constitute high vs low risk. The positive, right-skewed distribution of A-HIPI risk scores limited this approach ( Figure), as cutoffs of PFS5 <70 and PFS5> 90 would only identify 15% and <1% of the population, respectively. Approach 2: Risk groups based on deviation from “average” pt. The 5-y PFS was 77% (95% CI: 76-78). We explored defining “standard risk” based on this confidence interval as well as clinical boundaries, with pts above or below this classified as decreased or increased risk, respectively. Results of a +/-5% clinical boundary are presented in the Table. Approach 3: Risk groups based on “ranking” of pts. Here we ranked the A-HIPI risk scores of the 4022 AS-HL pts in the model building cohort and used the distribution of the risk scores as a benchmark. The risk profile for a future pt is then compared to this distribution (eg, how do you rank compared to your peers). Continuous results are often presented this way (eg, tertiles or quartiles). It also allows flexibility for the user to define the size of the risk groups and/or clinical threshold of interest. Application of this approach showed good alignment between the predicted model percentiles and the observed distribution of scores in the validation cohort ( Table). This is also reflected in calibration curves presented in the primary manuscript. An online application (eg, R-Shiny) will be provided at the meeting to allow users to define their own cutoffs to aid in pt prognostication as well as identify populations for clinical trial development. Conclusions: There are challenges with defining risk groups from individual risk prediction modeling in AS-HL. Different applications and purposes, the skewed distribution of events/risk estimates, as well as varying clinical definitions of high risk, make it challenging to define consensus expert-based risk groupings for AS-HL. A flexible “rank-based” approach appeared to provide the most clinical utility & data granularity, which may be leveraged for clinical trial design and pt stratification. Further analysis and discussion of how to optimally define high-risk and low-risk populations in AS-HL will be needed as new therapeutic options emerge.