TALENTOP
In selected patients with locally advanced HCC and macrovascular invasion who achieve partial response or stable disease after induction atezolizumab plus bevacizumab, does liver resection followed by systemic therapy improve outcomes compared with continued maintenance atezolizumab plus bevacizumab?
Absolute Effect
Absolute event difference: treatment failure occurred in 37% vs 59%, an absolute reduction of 22 percentage points.
NNT: not quantifiable. The reported 37% versus 59% treatment-failure proportions are cumulative event rates at the interim analysis rather than a prespecified fixed-time landmark, so a formal time-horizon NNT would be potentially misleading. The absolute 22-point event difference is reported directly.
Absolute increase: 18 percentage points.
NNH: 6, based on grade 3–4 treatment-related adverse events in the reported safety populations. This captures treatment burden but does not encompass the full risk of major surgery.
The big takeaway
Of 489 patients entering induction treatment, 201 were randomized after achieving partial response or stable disease and being considered suitable for resection. In the intention-to-treat population, median time to treatment failure was 20.4 months with surgery versus 11.8 months with maintenance therapy (HR 0.60, 95% CI 0.39–0.91; P=.015). Treatment failure occurred in 37% versus 59%.
Among the 101 patients assigned to surgery, 83 underwent resection; 77 achieved R0 resection (93%), and 23 had a pathological complete response (28%). These results demonstrate feasibility in a highly selected population rather than making all patients with macrovascular invasion surgical candidates.
Overall survival
OS was immature at the reported analysis. There were 15 deaths in the surgery group and 23 in the maintenance group; the hazard ratio for death was 0.67 (95% CI 0.35–1.29). The confidence interval crosses 1, so an OS benefit has not been demonstrated.
Estimated OS at 12, 18 and 24 months was 86%, 86% and 83% with surgery versus 91%, 78% and 67% with maintenance therapy. Longer follow-up is required before concluding that the improvement in treatment failure translates into a survival benefit.
Safety and treatment burden
Grade 3–4 treatment-related adverse events occurred in 32/83 (39%) of patients who underwent resection versus 21/100 (21%) receiving maintenance therapy. The most common were increased alanine aminotransferase (8% vs 1%), reduced platelet count (8% vs 3%), and proteinuria (4% vs 7%).
Two treatment-related deaths occurred in the surgery group: one from abnormal liver function considered related to atezolizumab and one from liver failure considered related to atezolizumab, bevacizumab, or surgery. Grade ≥3 surgical complications occurred in 21.7% in the reported surgical analysis.
NNH: 6, based on grade 3–4 treatment-related adverse events. This should not be interpreted as the complete harm profile of surgery; it is a deliberately defined trial-level toxicity NNH.
Clinical interpretation
TALENTOP is best understood as a conversion-surgery trial, not a trial establishing resection for unselected advanced HCC. Only 201 of 489 induction-treated patients were randomized. The evidence therefore applies to patients with macrovascular invasion, no extrahepatic disease, preserved liver function, disease control after atezolizumab plus bevacizumab, and anatomy judged suitable for resection by a multidisciplinary team.
The 22-point absolute reduction in treatment failure is clinically substantial, but it should not be converted into an NNT without a defensible fixed time horizon. The benefit is accompanied by an 18-point increase in grade 3–4 treatment-related adverse events and operative risk, while OS remains immature. The appropriate clinical conclusion is therefore that post-systemic-therapy reassessment for resectability should be considered in carefully selected patients—not that all such patients should undergo surgery.
This is an important Absolute Effect lesson: a large improvement in a disease-control endpoint can coexist with substantial treatment burden and uncertain survival benefit.
Limitations
- Primary endpoint was time to treatment failure; OS was immature.
- Only 201 of 489 induction-treated patients were randomized, creating substantial selection before the comparative phase.
- The trial was conducted at 24 hospitals in China; most patients had hepatitis B-associated disease, limiting generalizability to other HCC populations.
- Open-label treatment and the surgical selection process may influence treatment decisions and applicability.
- A formal NNT is not reported because the treatment-failure event rates were not presented as a fixed-time landmark suitable for a defensible NNT.