Lower Lenvatinib Starting Doses Show No Clear Advantage in Advanced Endometrial Cancer

When endometrial cancer returns after initial treatment or spreads beyond the uterus, one of the most powerful drug combinations available to oncologists pairs lenvatinib, an oral multikinase inhibitor that starves tumors of their blood supply, with pembrolizumab, an immune checkpoint blocker that reactivates T cells against malignant cells. The regimen, approved after the pivotal KEYNOTE-775 trial demonstrated a survival advantage over conventional chemotherapy, has become a standard option for patients with previously treated advanced or recurrent disease. Yet the approved lenvatinib dose of 20 milligrams per day is notoriously difficult to tolerate, and many clinicians in everyday practice deliberately start patients on lower doses, betting that reduced exposure early on will prevent the hypertension, fatigue, diarrhea, and protein loss that frequently force dose cuts later. A new systematic review and meta-analysis published in Frontiers in Oncology has now tested that bet against the available comparative evidence, and the results suggest the strategy rests on surprisingly shaky ground.

The research team, registered prospectively in the PROSPERO database under identifier CRD420261458417, conducted a comprehensive search of PubMed/MEDLINE, the Web of Science Core Collection, and the Cochrane Central Register of Controlled Trials from database inception through 11 July 2026, with source-specific updates completed on 13 and 14 July 2026. Their question was deceptively simple: among adults treated with the lenvatinib-pembrolizumab combination for advanced or recurrent endometrial cancer, do patients who begin lenvatinib at doses below 20 milligrams fare differently from those who start at the full 20-milligram dose? Because no randomized trial has directly addressed this question, the investigators had to rely on comparative observational studies, a class of evidence that comes with well-known vulnerabilities to confounding and selection bias.

To be eligible, studies had to compare patients on different lenvatinib starting doses within the combination regimen. The researchers defined their primary estimand as the contrast between starting doses below 20 milligrams and the standard 20-milligram starting dose, and they deliberately kept a secondary 10-milligram threshold contrast and a separate measure of relative dose intensity during the first eight weeks of treatment apart from the main analysis, recognizing that pooling these distinct exposure definitions would blur clinically meaningful differences. Binary outcomes such as tumor response and adverse events were summarized as risk ratios, while time-to-event outcomes such as progression-free survival were expressed as hazard ratios. This methodological discipline matters, because dose intensity and starting dose are related but not interchangeable concepts, and conflating them is a common pitfall in dose-optimization research.

The search yielded ten studies representing ten independent cohorts and a total of 1,000 patients. Of these, five cohorts encompassing 703 patients provided data for the primary starting-dose comparison, with the specific denominators varying by outcome depending on how completely each study reported its results. The remaining cohorts informed the secondary contrasts or descriptive syntheses. The scale of the evidence base is modest by the standards of modern oncology, where practice-changing trials often enroll thousands of participants, and this limitation echoes throughout the findings.

The headline result concerns antitumor effectiveness. Pooling across the available cohorts, the risk ratio for objective response, meaning measurable tumor shrinkage, was 0.97 for patients starting below 20 milligrams compared with those starting at 20 milligrams, with a 95 percent confidence interval of 0.70 to 1.34 and minimal heterogeneity between studies, indicated by an I-squared value of 11.9 percent. For disease control, which combines complete responses, partial responses, and stable disease, the pooled risk ratio was 0.96 with a confidence interval of 0.83 to 1.11 and essentially no heterogeneity at all, with I-squared at 0 percent. In plain terms, patients who started at reduced doses appeared to derive neither more nor less tumor control than those who started at the full dose, at least as far as these observational data can tell.

The tolerability picture was equally inconclusive. The pooled risk ratio for subsequent lenvatinib dose reduction, an indirect marker of how often patients needed their doses lowered after starting treatment, was 0.66 with a confidence interval spanning 0.34 to 1.27 and modest heterogeneity of 27.4 percent. Although the point estimate hints that reduced-dose starters needed fewer later dose reductions, the confidence interval crosses 1.0 by a wide margin, meaning the data cannot rule out anything from a substantial benefit to a substantial harm. Estimates for treatment interruption, permanent discontinuation, and grade 3 or higher adverse events were similarly imprecise, and two outcomes of obvious practical importance, hospitalization and treatment persistence, could not be pooled at all because the underlying studies reported them in forms too heterogeneous to combine statistically.

Perhaps the most sobering aspect of the analysis is the quality assessment. The researchers evaluated risk of bias using ROBINS-I, a tool specifically designed for non-randomized studies of interventions, and found that every evaluated result carried a serious overall risk of bias. They used ROB-ME to assess bias arising from missing evidence and applied the GRADE framework to rate certainty, concluding that the certainty of evidence was very low for all pooled comparisons. In the GRADE hierarchy, very low certainty means the true effect could differ substantially from the estimate, and the findings should be regarded as tentative at best. This is not a criticism of the review itself but a reflection of the observational evidence it had to work with: clinicians choose starting doses based on patient characteristics such as body weight, comorbidities, and baseline blood pressure, and those same characteristics independently influence outcomes, creating confounding that statistical adjustment can only partially remove.

The authors’ conclusion is measured but pointed: comparative observational evidence does not establish that starting lenvatinib below 20 milligrams preserves antitumor effectiveness or improves tolerability relative to the standard 20-milligram starting dose. They call for prospective, confounder-aware comparative studies before a reduced starting-dose strategy can be recommended broadly. The distinction between

Subject of Research: Lenvatinib starting doses below 20 mg versus 20 mg with pembrolizumab in advanced or recurrent endometrial cancer: a systematic review and meta-analysis of comparative observational evidence

Article Title: Lenvatinib starting doses below 20 mg versus 20 mg with pembrolizumab in advanced or recurrent endometrial cancer: a systematic review and meta-analysis of comparative observational evidence

Article References: Wu, Y., Wei, J., Chang, C., Li, M., Xu, R., Dong, S., & Jiang, M. (2026). Lenvatinib starting doses below 20 mg versus 20 mg with pembrolizumab in advanced or recurrent endometrial cancer: a systematic review and meta-analysis of comparative observational evidence. Frontiers in Oncology, 16, Article 1965694. https://doi.org/10.3389/fonc.2026.1965694

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1965694

Keywords: endometrial cancer, lenvatinib, pembrolizumab, dose optimization, meta-analysis, systematic review, immunotherapy, observational evidence, progression-free survival, drug tolerability, oncology, GRADE certainty