Inside the Real-World Evidence Picture for Hematologic Cancers: TriNetX at ISPOR 2026

May 21, 2026 | Blog

Hematologic cancers continue to demand a clearer view of how patients actually fare once they leave the controlled environment of a clinical trial. Treatment landscapes for diffuse large B-cell lymphoma (DLBCL) and chronic lymphocytic leukemia (CLL) are evolving quickly, yet the data underpinning health technology assessment (HTA) and regulatory decisions has not always kept pace with that complexity. Routine practice, eligibility realities, and population estimates all shape whether a new therapy reaches the patients it was designed to help. 

At ISPOR 2026 held May 17–20, I had the pleasure of presenting two posters from our TriNetX team that addressed exactly these challenges: one looking at outcomes in relapsed/refractory DLBCL, the other introducing a reproducible method for generating population estimates robust enough to stand up in HTA and regulatory submissions. Both drew on real-world data (RWD) from Germany. 

Here is a closer look at what each contribution explored, and what the findings suggested for the wider Health Economics and Outcomes Research (HEOR) community. 

Poster #CO180: Real-World Outcomes in Relapsed/Refractory DLBCL in Germany

DLBCL is curable with front-line therapy for many patients, but those who relapse or become refractory (r/r) move into a markedly more difficult second-line landscape, one that may involve chemotherapy, stem cell transplantation (SCT), and CAR-T therapy. Understanding how these patients fare in routine practice, and which factors predict worse outcomes, is essential both for day-to-day clinical decision-making and for positioning the novel therapies coming through the pipeline. 

A closer look at second-line practice

Together with our co-authors, we analyzed data from 289 patients who received ≥2L+ therapies in 2022, drawn from non-Hodgkin lymphoma treatment centers across Germany. We collected retrospective data via certified electronic case report forms extending back to initial diagnosis, which allowed us to examine the full arc from front-line therapy through subsequent treatment lines. 

The central question we set out to answer was whether the front-line treatment-free period (TFP1) — the time between front-line therapy and relapse — meaningfully shapes what happens next. 

Key real-world insights

  • An older, sicker population than trial data alone might suggest. The 2L+ population was predominantly male (65%), with a median age of 72 years. 27% had an ECOG score ≥2, and comorbidities were common: hypertension in 44%, diabetes in 20%, and coronary disease in 18%. 
  • Early relapse was strikingly common. 39.4% of patients experienced an early relapse (TFP1 <12 months), and within that group, 20% had TFP1 <2 months — consistent with primary refractory disease. 
  • Early-relapse patients looked clinically distinct. They were significantly younger, more symptomatic, and more often diagnosed with Ann Arbor Stage IV disease (p<0.01), with a trend toward more BCL6 translocations (p<0.1). 
  • Among SCT recipients, late relapse meant longer time to next treatment. Patients with late relapse had a considerably longer 2L time to next treatment (29.6 months) than early-relapse patients (17.0 months). 
  • Overall survival was significantly shorter in the early-relapse group (log-rank p<0.0001). 

Implications for clinical development

Taken together, I think these findings underline a persistent unmet need among r/r DLBCL patients who relapse early after front-line therapy. They are sicker at baseline, they reach subsequent treatments faster, and they die sooner. That profile reinforces the case for prioritizing their inclusion in clinical trials of novel agents and for evaluating new therapies against the realities of this population, not only the cleaner cohorts typically captured in pivotal studies. 

Poster #RWD55: A Novel Method for Robust Population Estimates in HTA and Regulatory Submissions

Population estimates (prevalence, incidence, eligible-population figures) sit at the heart of nearly every HTA and regulatory dossier. Yet producing estimates robust enough to withstand scrutiny remains a recurring challenge. In my experience, small inconsistencies between data sources, shifting case definitions, and evolving treatment practice all introduce uncertainty that is difficult to quantify and even harder to defend in a submission. 

A two-step methodology

In this poster, we introduced a structured approach that combines public epidemiological data with targeted epidemiologic surveys conducted at a representative sample of care-relevant centers. The method rests on two key components: 

  • An ABC analysis that ranks treating centers by patient volume and identifies the subset accounting for roughly 80% of care. 
  • Stratification by institution type (university hospitals (UH), non-university hospitals (NUH), and office-based practices (OBP)) combined with proportional, weighted sampling to extrapolate to national estimates. 

We chose chronic lymphocytic leukemia (CLL) in Germany as our worked example, a particularly useful test case given its heterogeneous treatment landscape and active drug development pipeline. 

Key real-world insights

  • A concentrated treatment landscape. We identified 321 CLL-treating centers (6% UH, 13% NUH, 81% OBP), with 51% of hospitals treating 80% of hospitalized patients. 
  • A solid evidence base from a representative sample. Our surveys drew on 61 centers and 1,631 patients, weighted to reflect the broader care landscape. 
  • National estimates with clearly defined uncertainty bounds. The approach yielded a national treated prevalence of 11,382–12,189 patients (13.66–14.69 per 100,000) and an incidence of 3,388–3,657 patients (4.06–4.39 per 100,000). 
  • External validation reinforced confidence. A chart review (n=682) supported the findings, with hospital distributions aligning with German Federal Joint Committee data and incidence estimates aligning with local projections. 

Why “treated” matters

One point I want to dwell on: local projections typically reflect diagnosed cases, whereas our approach estimates treated prevalence and incidence. The distinction is more than semantic. Treated populations are what actually drives resource allocation, cost-effectiveness modeling, and the interpretation of real-world datasets, and they are the ones HTA bodies and payers are ultimately weighing. We designed the method to be reproducible, transparent, and scalable across oncology indications and geographies, so the same framework can be applied wherever those questions matter most. 

Looking Ahead

Both posters spoke to a wider theme that ran through ISPOR 2026: that as treatment landscapes grow more complex, real-world evidence (RWE) has to do more than describe what happens. It has to support clinical development, ground HTA submissions in defensible population estimates, and connect routine practice to the decisions being made about access. 

Our DLBCL findings sharpened the case for designing trials around the patients with the greatest unmet need. Our CLL methodology offered a transferable framework for producing the kind of population estimates dossiers depend on.

I was glad to share this work with the ISPOR community in Philadelphia, and I look forward to continuing the conversations it sparked in DLBCL, in CLL, and across the wider set of HTA and regulatory questions that RWE is uniquely positioned to answer. 

More from TriNetX at ISPOR 2026

If you’d like to see more of what our team brought to ISPOR 2026, five additional posters are worth a look:

Explore how TriNetX high-quality real-world evidence can strengthen your research. 

About Zuzana Dostalova 

Zuzana helps transform real-world health data into meaningful insights that support better research and patient care. She guides scientific strategy, strengthens data quality standards, and collaborates with global partners to promote responsible, transparent, and impactful use of real-world evidence.