Organoid-guided adjuvant therapy for liver cancer enters phase II

What the work claims

This is a clinical-trial protocol, not a result. Its claim is prospective: that patient-derived tumor organoid drug sensitivity testing can guide adjuvant therapy after curative resection of hepatocellular carcinoma (HCC) well enough to improve recurrence-free survival.¹ The trial is registered as NCT07363356, is listed as not yet recruiting, and is planned to start on 2026-01-30 with an estimated completion date of 2028-11-30.

The design is a single-center, prospective, non-randomized, open-label phase II trial at the Third Affiliated Hospital of Naval Medical University. It plans to enroll 56 adults aged 18 to 70 with resectable primary HCC and high recurrence risk, defined by features such as tumor diameter greater than 5 cm, multiple tumors, microvascular invasion or poor differentiation. Resected tumor tissue will be used to establish patient-derived organoids. Successful organoids will be tested against a predefined panel: Apatinib, Atezolizumab plus Bevacizumab, Donafenib plus Toripalimab, Sintilimab, and FOLFOX. The most effective option by IC50 and area-under-the-curve values will be recommended for adjuvant therapy. Patients whose organoids fail to establish or for whom valid results are not available within one month will receive standard Lenvatinib as a control group.

How it works

The underlying model is that ex vivo organoid response predicts in vivo clinical response. HCC is a heterogeneous disease driven by viral hepatitis, metabolic dysfunction and diverse molecular alterations, and no single adjuvant regimen fits all patients. The approved and investigational drugs in the panel target different nodes: Apatinib is a vascular endothelial growth factor receptor 2 tyrosine kinase inhibitor; Atezolizumab plus Bevacizumab is an anti-PD-L1 and anti-VEGF combination; Donafenib plus Toripalimab combines a multikinase inhibitor with an anti-PD-1 antibody; Sintilimab is an anti-PD-1 antibody; and FOLFOX is a cytotoxic chemotherapy backbone. By measuring drug sensitivity directly in tumor organoids, the trial aims to bypass the need for a priori molecular biomarkers and assign each patient to the regimen most likely to prevent recurrence.

The primary endpoint is recurrence-free survival, defined as time from treatment initiation to tumor recurrence. Secondary endpoints are overall survival and safety measured by adverse events graded with NCI CTCAE version 5.0. Follow-up imaging is scheduled every 42 plus or minus 3 days initially, then every 3 months for the first year, and every 6 months thereafter. The sample size calculation assumes median recurrence-free survival will improve from 11.0 months in the empirical Lenvatinib group to 33.0 months in the organoid-guided group, with alpha 0.05, 80% power, 1:1 allocation and a 10% dropout rate.

Where a skeptic should push

The most obvious weakness is the non-randomized, open-label, single-center design. Without randomization, differences in recurrence-free survival could reflect patient selection, tumor biology or treating-physician behavior rather than the organoid test itself. The control group is not a parallel randomized arm; it is composed of patients for whom organoid culture fails or results are delayed. That group is likely to differ systematically from the organoid-guided group in tumor viability, sample quality and possibly disease biology, all of which influence prognosis.

The assumed effect size is also very large. Improving median recurrence-free survival from 11 to 33 months is a tripling, which is optimistic for any adjuvant strategy in HCC and is not supported by prior randomized data cited in the registry. A smaller true effect would make the 56-patient sample underpowered. In addition, the trial tests a fixed panel of drugs rather than a true all-comers strategy; if the best regimen for a particular patient lies outside the panel, the organoid guidance is constrained.

There are operational risks. Organoid establishment from resected HCC is not universally successful, and the one-month deadline for valid results is tight for adjuvant decision-making. Patients randomized in the empirical sense to Lenvatinib because of culture failure may have worse outcomes for reasons unrelated to the test. Finally, because the trial is listed as not yet recruiting, no patient has been treated yet; the protocol may change, and the actual accrual and feasibility are unknown.

What organoid-guided adjuvant therapy means for liver models

For organoid models of human organs and the drug-discovery work built on them, the non-obvious implication is that the field is moving from retrospective concordance studies to prospective clinical utility trials. Most organoid-oncology literature compares ex vivo drug response with patient response after the fact, which is valuable for validation but does not prove that acting on the organoid result changes outcomes. NCT07363356 is designed to answer the utility question directly: does treating patients according to organoid sensitivity improve recurrence-free survival compared with empirical choice? If positive, it would strengthen the case for organoid assays as clinically actionable decision-support tools, not just research models.

The opportunity is a blueprint for tumor-type-specific adjuvant trials. HCC is a sensible test case because recurrence is common, the adjuvant landscape is crowded with approved and investigational options, and there is no dominant molecular biomarker for regimen selection. The predefined panel mixes targeted therapy, immunotherapy and chemotherapy, which matches real clinical decision-making better than single-agent organoid studies. A successful trial could encourage similar designs in colorectal, pancreatic and gastric cancers, where adjuvant therapy is also empirically chosen and organoid culture is increasingly feasible.

The threat is that a positive result could be over-interpreted as proof that organoids are generically superior to molecular profiling, when this trial would actually test one assay in one center with one panel. A negative or inconclusive result, meanwhile, could be unfairly generalized to discredit organoid-based drug sensitivity testing as a whole. The single-center, non-randomized design makes either outcome hard to generalize. There is also a real risk that the trial succeeds on its primary endpoint for the wrong reason: if patients with successful organoid culture are simply healthier or have better-quality tumors, the apparent benefit may be selection bias rather than test-guided therapy.

For drug developers, the trial raises a subtler point about indication selection. The panel includes combinations that are approved or widely used in advanced HCC but not necessarily established in the adjuvant setting. If organoid guidance favors one combination over another, the trial could generate real-world evidence that accelerates or complicates adjuvant development programs. Conversely, if the organoid results cluster around drugs already known to be active, the test may add little incremental value beyond standard molecular stratification. The trial is therefore as much a test of the assay's clinical added value as it is of the drugs themselves.

The bottom line

Established with reasonable confidence: a 56-patient single-center phase II trial is planned to test whether patient-derived HCC organoid drug sensitivity testing can guide adjuvant therapy after curative resection better than empirical Lenvatinib, using recurrence-free survival as the primary endpoint. The trial design, patient population, drug panel and statistical assumptions are clearly specified in the registry. Not established: whether the trial will recruit on schedule, whether organoid culture will succeed reliably within one month, or whether organoid-guided therapy improves outcomes. The design is non-randomized and open-label, so even a positive result will require cautious interpretation and ideally confirmation in a randomized trial. The trial would be undercut if organoid establishment fails in a large fraction of patients, if the assumed recurrence-free survival benefit is not observed, or if the organoid-guided arm benefits primarily because of patient selection rather than the test itself.