Radiologists Sound the Alarm on the World Standard for Staging Childhood Liver Cancer

For more than three decades, nearly every child diagnosed with a malignant liver tumor anywhere in the world has been staged with the same system. Known as PRETEXT, short for PRETreatment EXTent of disease, it divides the liver into four sections and asks a deceptively simple question: how many of those sections are free of tumor? The answer, expressed as a group from I to IV, drives decisions about chemotherapy, surgery, and even liver transplantation. Yet a new position paper from the SIOPEL Radiology Group, the imaging arm of the International Childhood Liver Tumours Strategy Group in Europe, argues that this cornerstone of pediatric oncology is showing its age. Published in Pediatric Radiology, the paper catalogs a series of technical, definitional, and clinical challenges that the authors say can lead to inconsistent staging, misallocated treatment, and biased research data.

The history of PRETEXT explains both its power and its problems. Before 1990, there was no common language for staging malignant liver tumors in children; the extent of disease was judged from what surgeons found in the operating room or from the resected specimen afterward. A joint working group spanning the International Society of Pediatric Oncology, the German Pediatric Oncology Group, the Japanese Liver study, and Canadian representatives agreed in 1992 to adopt the imaging-based grouping system first used in the SIOPEL 1 trial as a global standard. Crucially, this shift to pretreatment imaging meant that children too sick for immediate surgery could finally be included in trials. The system was revised in 2005 and again in 2017, when a new set of definitions was crafted to serve the Paediatric Hepatic International Tumor Trial, the largest collaborative effort yet undertaken against these rare cancers.

The modern system has two components. The PRETEXT group, from I to IV, reflects how many contiguous liver sections are free of tumor, with the liver divided into left lateral, left medial, right anterior, and right posterior sections using hepatic and portal venous anatomy as landmarks. Involvement of the caudate lobe, segment I, is by convention classified as at least PRETEXT II. Alongside the group sit eight annotation factors, memorably summarized as VPEFRCNM: venous involvement, portal vein involvement, extrahepatic extension, multifocality, tumor rupture, caudate lobe involvement, lymph node involvement, and distant metastases. Each is recorded as simply present or absent. After neoadjuvant chemotherapy, the same framework is reapplied and called POSTTEXT. The whole edifice feeds directly into the CHIC risk-stratification scheme that determines whether a child receives standard, intermediate, or high-risk chemotherapy.

The first major weakness, the authors argue, lies upstream of interpretation entirely: image acquisition. Staging requires multiplanar, multiphase, high-quality MRI or CT, but imaging small children is notoriously difficult. Motion artifacts plague upper abdominal scans in young patients, small body volumes demand high spatial resolution, and sedation or general anesthesia adds risk and complexity. Consensus imaging guidelines from the American College of Radiology’s Pediatric LI-RADS Working Group have helped, but the SIOPEL group reports that implementation and adherence vary considerably across Europe. A recent study of 358 CT scans and 144 MRI scans from 105 institutions in 226 patients, part of the Children’s Oncology Group AHEP0731 study, documented considerable heterogeneity even within the United States. In Europe, marked variation persists, with significant ongoing use of CT despite guideline recommendations favoring MRI, reflecting differences in expertise, resources, scanner technology, and access to appropriate contrast agents. Notably, the European Society of Paediatric Radiology has yet to publish dedicated guidance on pediatric liver imaging.

Then comes the problem of human disagreement. The system’s binary nature, with sections and annotation factors marked only as involved or uninvolved, leaves little room for nuance when tumors push against boundaries or ambiguously abut blood vessels. Studies comparing PRETEXT assignments between local and central radiologists have found kappa values of 0.57 to 0.79, indicating moderate to substantial agreement, and there has been no apparent improvement over time despite better scanners and formal imaging recommendations. Agreement appears higher for POSTTEXT, with kappa values of 0.82 and 0.93 in two contemporary studies, but annotation factors remain the weak point. Hepatic venous involvement shows the lowest inter-observer agreement of all, and local radiologists have reported particularly high false-positive rates for vascular involvement, extrahepatic extension, rupture, and lymph node involvement. Rarity compounds the problem: hepatoblastoma strikes only about 1.5 children per million per year, so many hospitals encounter these tumors too infrequently to build deep expertise. Even within the SIOPEL Radiology Group itself, a 2024 survey found many members struggled with the system.

Some of the deepest difficulties are anatomical rather than technical. PRETEXT boundaries are theoretical planes that are hard to delineate in practice, especially when large tumors distort the liver. The boundary between the left medial and right anterior sections, Cantlie’s line, runs from the middle hepatic vein to the middle of the gallbladder fossa, but a bulky tumor can deform the plane while the landmarks stay put, making it genuinely hard to distinguish true invasion from mere mass effect. Small interpretive differences can flip a child from PRETEXT II to III and change management entirely. The consequences are measurable: one study comparing imaging-based PRETEXT with pathology found the stage correct in only 51 percent of cases, over-staged in 37 percent, and under-staged in 12 percent. The high over-staging rate partly reflects an official recommendation to assign the highest group when uncertain, a rule that trades surgical conservatism for statistical noise.

Vascular assessment exposes perhaps the sharpest mismatch between the staging system and modern surgery. Tumor proximity to the hepatic veins, portal vein, and inferior vena cava determines whether a tumor can be resected or whether transplantation is needed, yet V and P status alone do not capture resectability. A PRETEXT I or II tumor that abuts the portal vein bifurcation by less than 180 degrees is technically P-negative but may still require an extreme resection or a transplant. Conversely, tumor thrombus extending into the contralateral portal vein or the draining veins of unaffected sections renders the whole liver effectively involved, leading surgeons to treat such tumors as functionally PRETEXT IV even when the group assignment says otherwise. The system also ignores details that matter enormously in the operating room, such as thrombus extension into the superior mesenteric vein, which can necessitate combined liver and small bowel transplantation, or tumor emboli that travel into the pulmonary arteries, which the authors suggest reporting separately as embolic M-positive disease.

The annotation factors each carry their own traps. Extrahepatic spread occurs in fewer than 5 percent of children at diagnosis but carries a poor prognosis, and distinguishing contiguous invasion from simple proximity or reactive inflammation remains a diagnostic challenge; targeted ultrasound can help by showing whether tumor and adjacent organs move together or freely. Rupture is defined narrowly as free intraperitoneal fluid with evidence of tumor-associated hemorrhage, excluding contained subcapsular ruptures that still demand altered biopsy and surgical planning. Multifocality, present in up to 20 percent of hepatoblastoma cases, can emerge or vanish between PRETEXT and POSTTEXT as chemotherapy shrinks tumors and separates confluent foci, yet sites of disease that become radiologically occult may still harbor viable microscopic tumor and remain critical for curative surgery. Lymph node assessment relies on size and morphology criteria that are imperfect in both directions, and lung metastasis assessment is hampered by the fact that up to two-thirds of resected suspected pulmonary metastases in these children prove benign.

POSTTEXT, the post-treatment counterpart, inherits every one of these problems and adds its own. Imaging cannot reliably indicate tumor viability, so apparent abnormalities may not correspond to residual disease, while occult tumor can persist where images look clean. Current risk stratification and treatment allocation still rest on the initial PRETEXT, and the role of POSTTEXT in guiding decisions or predicting outcomes has not been established. This creates a troubling possibility the authors highlight: children whose tumors show marked radiological downstaging may still undergo resections planned according to the original tumor distribution, exposing them to more extensive surgery, and more morbidity, than strictly necessary to control their disease.

The paper closes with a program of remedies. The SIOPEL group proposes standardized imaging protocols, structured reporting templates that explicitly document uncertainty and surgery-specific details such as future liver remnant volume, routine multidisciplinary tumor board discussion, double and central review for complex cases, and a secure cloud-based platform for consensus review and training. They call for validation studies quantifying inter-observer agreement under standardized protocols, refinement of ambiguous definitions such as vascular encasement thresholds, and eventual inclusion of radiology-specific research questions in trial protocols. Emerging tools, from improved MRI sequences that enable free-breathing imaging to AI-based segmentation and deep learning methods, may help, though the authors caution that robust pediatric evidence remains limited. No formal revision of PRETEXT is currently planned, and any such effort would require global consensus on its primary aims. Until then, the message to the worldwide pediatric oncology community is clear: the system that has unified childhood liver cancer staging since 1990 still works, but only if everyone using it understands exactly where it bends.

Subject of Research: Limitations of the PRETEXT imaging staging system for pediatric malignant liver tumors

Article Title: Challenges with the PRETEXT staging system: a SIOPEL Radiology Group position paper

Article References: Lohne, S. M. H., Hebelka, H., Prince, J. F., Bohte, A. E., Littooij, A. S., de Lange, C., Deganello, A., Costa Dias, S., Chambers, G., Toso, S., Habre, C., Pace, E., Franchi-Abella, S., Woodley, H., & McGuirk, S. P. (2026). Challenges with the PRETEXT staging system: a SIOPEL Radiology Group position paper. Pediatric Radiology. https://doi.org/10.1007/s00247-026-06786-w

Image Credits: AI Generated

DOI: 10.1007/s00247-026-06786-w

Keywords: PRETEXT, hepatoblastoma, pediatric liver tumors, radiology, cancer staging, SIOPEL, MRI, POSTTEXT, risk stratification, liver transplantation, inter-observer variability, structured reporting