Comparative efficacy and safety of traditional Chinese medicine injections combined with chemotherapy in ovarian...

Traditional Chinese medicine injections (TCMIs) are widely used as adjunctive agents in ovarian cancer chemotherapy across East Asia, yet no head-to-head evidence exists comparing their relative efficacy and safety.

This study aimed to estimate the effects of ten traditional Chinese medicine injections combined with chemotherapy versus chemotherapy alone in ovarian cancer and, as a secondary objective, to summarize the uncertainty of indirect treatment rankings.

Eight databases were searched from inception through March 2026 for randomized controlled trials (RCTs) comparing a TCMI combined with platinum-based chemotherapy versus chemotherapy alone in ovarian cancer. The primary outcomes were objective response rate (ORR) and KPS-defined functional-status improvement (reported as quality-of-life improvement in the original trials); secondary outcomes included gastrointestinal adverse events, bone marrow suppression, hepatic and renal function impairment, and serum CA125. Random-effects frequentist network meta-analysis used risk ratios (RR), odds ratios (OR), or standardized mean differences (SMD) with 95% confidence intervals (CI). Relative effects were the primary inferential output; SUCRA, probability of being best, and mean rank were secondary descriptive summaries reported with trial counts, node sample sizes, and CINeMA certainty. All networks were connected through chemotherapy but were star-shaped and contained no closed loops, so incoherence could not be statistically evaluated. Sensitivity analyses excluded nodes supported by no more than two trials and compared random-effects with fixed-effect models. The study is registered with PROSPERO (CRD420261341791).

Fifty-nine RCTs encompassing 4483 patients and ten TCMIs were included. For ORR, XPI + Chemo (RR 1.46, 95% CI 1.17-1.82), CKI + Chemo (1.35, 1.22-1.49), and SFI + Chemo (1.34, 1.17-1.54) showed favorable estimates versus chemotherapy alone, but all comparisons between TCMIs were indirect. In the analysis restricted to nodes with at least three trials, ORR estimates remained favorable for CKI, SFI, ADI, KAI, and SMI versus chemotherapy, while the probability of being best was distributed across CKI and SFI. For KPS improvement, the one-trial ZTOI and CI nodes had the largest full-network point estimates; after sparse nodes were removed, only ADI, KAI, and SFI remained and best-rank probabilities were dispersed. The corrected CA125 analysis included six RCTs (498 patients); SFI + Chemo was associated with a lower CA125 level than chemotherapy alone (SMD -4.34, -7.51 to -1.17), with extreme heterogeneity (τ² = 5.087; I² = 97.4%). Several safety estimates were below the null, but precision and certainty varied and no head-to-head safety hierarchy was supported.

A connected star-shaped network can estimate indirect relative effects through the common chemotherapy comparator, but it cannot empirically test incoherence and does not provide head-to-head evidence. Adjunctive TCMIs showed several signals versus chemotherapy alone; however, ORR is a short-term surrogate and no progression-free or overall survival data were available. Sparse nodes, unmeasured effect modifiers, absence of trials at overall low risk of bias, and predominantly low-to-moderate certainty preclude comparative treatment recommendations. Rankings are secondary, exploratory summaries and require confirmation in adequately powered head-to-head randomized trials.