Äijälä 2026 — Distance-Based Evaluation of Tumor Budding in Colorectal Cancer

Evaluation of TB-distance (mean distance from tumor bulk to the three farthest buds, cutoff ≥123 µm) across 1,876 CRC patients in two independent cohorts — correlates with aggressive features and poor survival, but adds no independent prognostic value beyond conventional ITBCC bud counting.
Author

Äijälä VK, Sirniö P, Karjalainen H, Kastinen M, Tapiainen VV, Elomaa H, Ahtiainen M, Pohjanen VM, Mattila TT, Lindgren O, Helminen O, Wirta EV, Rintala J, Meriläinen S, Saarnio J, Rautio T, Seppälä TT, Böhm J, Mecklin JP, Tuomisto A, Mäkinen MJ, Väyrynen JP

Doi

10.1007/s00428-026-04471-9

Measuring how far tumor buds extend from the tumor bulk (TB-distance, cutoff \(\ge 123\,\mu\text{m}\)) correlates with aggressive clinicopathologic features and worse survival across 1,876 colorectal cancers, but does not provide independent prognostic value beyond standard ITBCC bud counts — an informative negative result indicating spatial dispersion is collinear with the dissociation phenotype already captured by conventional grading.

Read status: full text (Open Access). Captures the full two-cohort study design, digital measurement definitions, ROC cutoff derivation, multivariable Cox models, and subgroup stratification.

The clinical and biological question

Conventional [[lugli-2017-itbcc-tumor-budding|ITBCC assessment]] focuses exclusively on bud density (counting single cells or clusters of \(\le 4\) cells within a normalized 0.785 mm² hotspot). However, epithelial–mesenchymal transition (EMT) and active stromal invasion involve both detachment and migration.

This study asks: 1. Does the distance that tumor buds travel into peritumoral stroma away from the tumor bulk (TB-distance) reflect a more aggressive invasive phenotype and predict cancer-specific survival? 2. Can TB-distance salvage prognostic discrimination in tumors with low (Bd1, 0–4 buds) or intermediate (Bd2, 5–9 buds) budding that might behave more aggressively than expected?

The four fields

Dataset 1,876 stage I–IV CRC patients across two independent Finnish hospital cohorts, excluding neoadjuvant therapy and perioperative deaths (\(\le 30\) days):
• Cohort 1 (Oulu University Hospital, 2006–2020): \(N = 776\) evaluated (\(N = 771\) in survival models; median follow-up 7.0 years).
• Cohort 2 (Central Hospital of Central Finland, 2000–2015): \(N = 1,100\) evaluated (\(N = 1,063\) in survival models; median follow-up 10.0 years).
Stain / scanner / magnification Routine H&E whole-slide digital images (WSI). Invasive front reviewed systematically; digital caliper distance measured from the nearest tumor bulk edge to the 3 farthest tumor buds and averaged (TB-distance). Large tumor islets, glands, cribriform structures, and mucin pools counted as bulk; poorly differentiated clusters (\(\ge 5\) cells) and intravascular cells excluded.
Validation Cutoff derived in Cohort 1 via ROC analysis and Youden’s index for 10-year cancer-specific survival (\(123\,\mu\text{m}\); AUC 0.63, 95% CI 0.58–0.68). Continuous reproducibility between 2 observers across 30 cases: Spearman’s \(\rho = 0.58\); binary agreement: Cohen’s \(\kappa = 0.43\).
External validation Yes — the exact \(123\,\mu\text{m}\) cutoff was applied unchanged to independent Cohort 2 (\(N = 1,100\)).

Findings

1. Clinicopathologic associations

  • High TB-distance (\(\ge 123\,\mu\text{m}\)) occurred in 27% of Cohort 1 (\(209/776\)) and 18% of Cohort 2 (\(199/1,100\)).
  • In both cohorts, high TB-distance was strongly associated with aggressive features:
    • High WHO grade (\(p < 0.001\))
    • Advanced UICC stage (\(p < 0.001\))
    • Lymphovascular invasion (\(p < 0.001\))
    • High ITBCC tumor budding grade (\(p < 0.001\))
    • MMR-proficient status (\(p = 0.002\) in Cohort 1, \(p = 0.003\) in Cohort 2).

2. Standalone survival prognosticator

In multivariable Cox regression adjusted for age, sex, operation year, tumor location, stage, WHO grade, LVI, MMR status, and BRAF mutation: - Cohort 1: Multivariable HR 1.47 (95% CI 1.04–2.09, \(p = 0.030\)) for cancer-specific survival (CSS). - Cohort 2: Multivariable HR 1.34 (95% CI 1.04–1.74, \(p = 0.026\)) for CSS.

3. The decisive negative finding: zero incremental value over ITBCC

Despite standalone prognostic significance: - Within ITBCC grade strata: When evaluating TB-distance strictly within Bd1, Bd2, or Bd3 subgroups, it failed to provide statistically significant prognostic stratification in multivariable analyses across both cohorts. - Direct joint model: When conventional ITBCC budding grade and TB-distance were entered into the same multivariable Cox model, conventional budding remained robustly significant, while TB-distance lost all independent prognostic value. - Conclusion: TB-distance is collinear with conventional bud density. It reflects the degree of tumor dissociation already captured by counting buds in the ITBCC 0.785 mm² hotspot, rather than an autonomous invasive axis.

Transferable takeaways for pathology and computational models

  1. No routine clinical measuring: There is no clinical or evidentiary basis to burden pathologists with measuring bud distances in micrometers during routine synoptic reporting. Standard ITBCC count/grade remains the evidence-based benchmark.
  2. Visual prompt for hotspot scanning: While not a standalone reportable metric, noting tumor buds dispersed deep into peritumoral fat or stroma serves as a useful practical cue for the pathologist to re-scan the invasive front for a higher-density budding hotspot.
  3. Spatial feature for computational pathology: Unlike manual workflows where digital caliper measurements are time-consuming, automated WSI pipelines (cell detection, segmentation, and graph neural networks) can extract spatial dispersion metrics effortlessly. TB-distance provides an interpretable spatial metric for automated invasive-front phenotyping.

How it connects