This PRISMA-NMA-compliant review compared progression-free survival (PFS) and overall survival (OS) across second-line and later HER2-targeted regimens after trastuzumab-based therapy. The authors used Bayesian network meta-analysis to summarize PFS and OS as hazard ratios (HRs). Within the connected overall networks, T-DXd ranked favorably for survival compared with several alternatives.
Direct results reported in the source show that relative to T-DXd, T-DM1 had less favorable outcomes: PFS HR 3.30 (95% credible interval [CrI] 2.06–5.28) and OS HR 1.52 (95% CrI 1.19–2.01). The manuscript cautions that T-DXd was not represented in the second-line-only OS network nor in Asia-dominant PFS/OS networks; therefore, the survival ranking for T-DXd primarily reflects evidence from trials that included mixed-line populations and were not Asia-dominant.
The authors emphasize careful interpretation of indirect survival comparisons because network connectivity and trial populations varied across analyses.
Confirmed objective response rate (ORR) was summarized as odds ratios (ORs) in the Bayesian network meta-analysis. T-DXd produced higher confirmed ORR than several comparators: compared with Lap-Cap (lapitinib-capecitabine) OR 7.43 (95% CrI 3.80–13.16), compared with PTN-Cap OR 3.35 (95% CrI 1.24–8.19), and compared with T-DM1 OR 6.31 (95% CrI 3.98–10.32).
Duration of response (DoR) was analyzed exploratorily using a fixed-effect Bayesian normal model after prespecified conversion of arm-level summary statistics. In the EMILIA/NALA network, DoR showed no clear between-treatment differences. Time to progression (TTP) could not be quantitatively synthesized due to unavailable or clinically non-equivalent summary measures.
Safety outcomes were summarized as odds ratios where feasible. Estimates for any-grade adverse events (AEs) did not indicate a safety advantage for T-DXd. Serious adverse event (SAE) comparisons were inconclusive within the sparse networks reported: the OR for T-DM1 versus Lap-Cap was 0.90 (95% CrI 0.67–1.21), and for T-DXd versus T-DM1 the OR was 1.30 (95% CrI 0.87–1.93).
A clinically important safety signal was a higher risk of ILD/pneumonitis with T-DXd compared with T-DM1: OR 5.55 (95% CrI 2.79–12.16). Differences in left ventricular dysfunction (LVD) between treatments were reported as unclear in the available networks.
The authors note that sparse event counts and use of fixed-effect models in limited safety networks reduce certainty and require cautious interpretation.
The review was prospectively registered with PROSPERO (CRD420251244047) and followed PRISMA-NMA guidance. The authors searched PubMed, Embase, Web of Science, and CENTRAL through July 24, 2026 for eligible randomized controlled trials. Phase II or III randomized trials conducted after trastuzumab-based therapy were eligible for inclusion.
A Bayesian network meta-analysis framework was used. PFS and OS were summarized as hazard ratios. Confirmed ORR and safety outcomes were summarized as odds ratios. Random-effects models were employed where heterogeneity could be estimated. For sparse safety networks (serious adverse events and ILD/pneumonitis) the authors used fixed-effect Bayesian models. DoR was examined exploratorily using a fixed-effect Bayesian normal model after converting arm-level summary statistics; TTP was not synthesized quantitatively when required summary measures were unavailable or clinically non-equivalent.
Sixteen reports representing seven independent randomized trial families were included, representing 3,339 unique randomized participants. The networks combined trials enrolling patients in second-line and later settings after prior trastuzumab-based therapy. The authors highlight that some networks were Asia-dominant or second-line-only and that T-DXd was not present in the second-line-only OS network nor in Asia-dominant survival networks; this affected the interpretability and applicability of some comparisons.
Within the connected overall networks, T-DXd demonstrated a favorable survival profile and the highest confirmed ORR, but these survival rankings depended on evidence from mixed-line, non–Asia-dominant trials. The increased risk of ILD/pneumonitis with T-DXd versus T-DM1 is clinically important. Any-grade AE and SAE comparisons were inconclusive in sparse safety networks. DoR did not show clear differences where analyzed, and TTP could not be pooled quantitatively.
The authors advise cautious interpretation of indirect survival comparisons and of safety results derived from sparse networks. Important limitations noted include variable network connectivity, the absence of T-DXd in certain focused networks (second-line-only, Asia-dominant), and limited safety event data that required fixed-effect modelling in some analyses.
No competing interests were declared by the authors. The analysis and reported effect estimates are as presented in the source article.