📄 medicine

Clinical Phenotypes and Prognosis in Patients with Venous Thromboembolism and Cancer: A Cluster Analysis

This retrospective cohort study of 654 Colombian patients with venous thromboembolism and cancer demonstrates that clustering based on non-oncologic clinical characteristics identifies distinct prognostic subgroups with varying survival outcomes, independent of traditional tumor-related thrombotic risk factors.

Original authors: Arturo D. Mora, Juan Pablo Gómez-Arellano, David Fernando Narvaez-De Los Rios, Manuela Estrada-Schweineberg, Andrés Domínguez-Durán, Valeria Azcarate-Rodríguez, Juan Andres Muñoz, Diana C. Carrillo-Gó

Published 2026-07-30

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Arturo D. Mora, Juan Pablo Gómez-Arellano, David Fernando Narvaez-De Los Rios, Manuela Estrada-Schweineberg, Andrés Domínguez-Durán, Valeria Azcarate-Rodríguez, Juan Andres Muñoz, Diana C. Carrillo-Gómez, Julián David Yara, Hoover León-Giraldo, Frank Douglas Cañón-Estrada, Mario Miguel Barbosa-Rengifo, Jessica Largo-Ocampo, Juan Esteban Gómez-Mesa

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the human body as a bustling, high-tech city. In this city, blood vessels are the highways, and blood cells are the delivery trucks carrying oxygen and nutrients to every neighborhood. Usually, traffic flows smoothly. But sometimes, a delivery truck gets stuck, forming a traffic jam that blocks the road. In medical terms, this is a blood clot. When these clots travel to the lungs or get stuck in deep veins, it's called Venous Thromboembolism (VTE). It's a serious event that can stop the city's power grid.

Now, imagine that this city is also under construction because of a massive, chaotic project called "Cancer." Cancer is like a rogue construction crew that not only messes up the city's layout but also makes the roads sticky and prone to jams. Doctors have long known that people with cancer are much more likely to get these blood clots. For a long time, the way doctors tried to guess how sick a patient would get was by looking only at the construction crew: What kind of cancer is it? How big is the site? How aggressive is the crew? They used a "map" based entirely on the tumor to predict the future. But just like two cities with the same construction project can have very different outcomes depending on the weather, the age of the buildings, or the health of the local population, two cancer patients might have very different journeys even if they have the same type of tumor. This study asks a simple but revolutionary question: What if we stopped looking only at the construction crew and started looking at the city itself?

The Study: Mapping the City, Not Just the Construction

A team of researchers from Colombia decided to take a fresh look at 654 patients who had both cancer and a blood clot. Instead of just asking, "What kind of cancer do you have?" they asked, "Who are you as a person?" They gathered a huge list of details about the patients' lives: their age, whether they had other health issues like high blood pressure or diabetes, if they had recently had surgery, and what their overall health looked like. They deliberately ignored the specific type of cancer and the cancer treatments during the first part of their analysis. They wanted to see if the "person" behind the disease had a story to tell that the "disease" alone couldn't.

Using a fancy computer method called "cluster analysis" (think of it as a super-smart sorting machine that groups things that look alike), they sorted these 654 people into three distinct groups, or "phenotypes." It's like sorting a mixed bag of marbles not by color, but by how heavy they are and what they're made of.

The Three Groups Discovered

The "Young and Fighting" Group: This group had 367 people. They were generally younger (average age around 59) and were actively fighting cancer. They didn't have many other health problems, but they were in the thick of the battle.
The "Older, Recently Operated" Group: This group had 211 people. They were older (average age around 76) and had recently had surgery. They had some other health issues, but not a ton.
The "Older with Many Health Issues" Group: This was the smallest group with 76 people. They were the oldest (average age around 78) and were carrying a heavy load of other health problems, like heart disease, diabetes, and high blood pressure.

The Big Surprise: Who Survived Longer?

Here is where the story gets twisty. You might guess that the "Young and Fighting" group would live the longest because they are younger and healthier. But the data told a different story.

The researchers tracked how long each group lived after their blood clot was found.

The Young and Fighting group had a median survival of 2.4 years.
The Older, Recently Operated group had a median survival of 6.2 years.
The Older with Many Health Issues group had a median survival of 3.8 years.

Wait, the older group that had just had surgery lived the longest? Yes! The study suggests that being young and actively fighting cancer might actually be a sign of a more aggressive disease or a more complex situation, leading to a shorter survival time in this specific context. Meanwhile, the older group that had surgery seemed to have a more stable path, surviving much longer.

Why This Matters

The researchers ran the numbers again to make sure this wasn't just a fluke caused by the type of cancer. They adjusted for the "Khorana score," which is the standard way doctors currently guess risk based on the tumor. Even after accounting for how dangerous the cancer was, the "person" still mattered. The three groups had different survival rates regardless of their tumor type.

This suggests that the old way of looking only at the cancer isn't the whole picture. The patient's own body, their age, and their other health issues act like a second map that changes the destination. The study doesn't say we should stop treating the cancer, but it does suggest that doctors need to look at the whole patient, not just the tumor, to understand who is at the highest risk.

The Bottom Line

The study found that among people with cancer and blood clots, there are three distinct "personality types" of patients, and these types predict how long they might live better than just looking at the cancer alone. The "Young and Fighting" group faced a tougher road, while the "Older, Recently Operated" group had a longer journey ahead. The researchers are careful to say this is a suggestion based on their data, not a final rule, but it opens a new door. It tells us that to truly understand a patient's future, we need to listen to the story of their whole body, not just the story of their tumor.

Technical Summary: Clinical Phenotypes and Prognosis in Patients with Venous Thromboembolism and Cancer

Problem Statement
Current risk stratification for patients with venous thromboembolism (VTE) and cancer is predominantly tumor-centered, relying on models such as the Khorana score which focus on tumor type and stage. However, clinical outcomes in this population are heterogeneous and influenced by patient-level characteristics independent of tumor biology, including comorbidity burden, recent clinical exposures, and overall clinical status. Existing models may fail to capture this non-oncologic heterogeneity, leading to variability in prognostic assessment that is not explained by tumor-related thrombotic risk alone.

Methodology
The authors conducted a retrospective observational cohort study at a tertiary referral center in Colombia, analyzing 654 patients with objectively confirmed VTE and current or prior malignancy (excluding non-melanoma skin cancers) between January 2011 and January 2025.

Data Collection: Clinical and demographic data were extracted from electronic medical records, including oncologic characteristics, non-oncologic comorbidities, VTE presentation, and anticoagulation management.
Phenotype Identification: To identify clinical phenotypes based on non-oncologic characteristics, the authors employed an unsupervised machine learning approach. Sixteen variables reflecting patient clinical characteristics, comorbidities, and recent clinical exposures were selected. Crucially, tumor site and oncologic treatments were intentionally excluded from the clustering variables to prevent the groups from being driven primarily by tumor type.
- Technique: Multiple Correspondence Analysis (MCA) was performed, followed by hierarchical agglomerative clustering using Ward's minimum-variance method.
- Cluster Selection: While a two-cluster solution yielded the highest silhouette width, a three-cluster solution was selected based on greater clinical resolution and interpretability.
Statistical Analysis: Overall survival (OS) was estimated using Kaplan–Meier analysis. Multivariable Cox proportional hazards models were used to evaluate the association between the identified clinical phenotypes and mortality, adjusting for tumor-related thrombotic risk categories (standard, high, and very high risk per Khorana criteria).

Key Contributions
The study introduces a data-driven framework for characterizing patients with cancer-associated VTE that moves beyond tumor-centric models. By isolating non-oncologic variables, the authors demonstrate that patient clinical profiles constitute distinct prognostic subgroups. The study provides a granular view of how comorbidity burden, age, and recent clinical exposures (such as surgery) interact to influence survival outcomes independent of the underlying malignancy's thrombotic risk profile.

Results
The analysis identified three distinct clinical phenotypes:

Cluster 1 (Younger patients with active cancer): $n=367$ . Characterized by a median age of 59.3 years, a high prevalence of active cancer (81.7%), and a low comorbidity burden.
Cluster 2 (Older patients with recent surgical exposure): $n=211$ . Characterized by a median age of 75.9 years, recent surgical exposure, and an intermediate comorbidity burden.
Cluster 3 (Older patients with multimorbidity): $n=76$ . Characterized by a median age of 78.5 years, marked cardiovascular multimorbidity, and the highest HAS-BLED risk profile.

Survival Outcomes:

The median overall survival (OS) for the entire cohort was 3.7 years (95% CI 2.8–5.4).
Survival differed significantly across phenotypes (log-rank $p=0.044$ $p = 0.044$ ):
- Cluster 1: Median OS of 2.4 years.
- Cluster 2: Median OS of 6.2 years.
- Cluster 3: Median OS of 3.8 years.
Multivariable Analysis: After adjusting for tumor-related thrombotic risk, clinical phenotype remained independently associated with mortality. Compared to Cluster 2 (the reference), Cluster 1 was associated with a higher risk of mortality (HR 1.30; 95% CI 1.02–1.65; $p=0.032$ ). Cluster 3 showed no statistically significant difference in mortality risk relative to Cluster 2 (HR 1.07; 95% CI 0.74–1.54; $p=0.731$ ).
Secondary outcomes (major bleeding and recurrent VTE) occurred in 3.8% of patients each, with no significant differences across clusters, though bleeding rates were numerically higher in Cluster 3.

Significance and Claims
The authors claim that clinical phenotypes defined by non-oncologic characteristics identify prognostic groups beyond tumor-related thrombotic risk. The findings suggest that patient-level factors contribute significantly to outcome heterogeneity in this population. Specifically, the study highlights that younger patients with active cancer but low comorbidity (Cluster 1) may face worse survival outcomes compared to older patients with recent surgical exposure (Cluster 2), challenging the assumption that chronological age or active cancer status alone dictates prognosis.

The paper posits that integrating clinical phenotyping with tumor-based risk stratification offers a more nuanced understanding of prognosis. It suggests that recognizing these phenotypic profiles could support more individualized evaluation of anticoagulation management and follow-up intensity, particularly for patients with complex comorbidities or high bleeding risks. The authors emphasize that these findings are derived from real-world data and support the relevance of the patient's clinical profile in outcome variability, though they note that prospective validation is required to confirm reproducibility and determine if phenotype-informed strategies can refine risk assessment.

More like this