This study was prospectively registered in the prospero systematic review database, number crd42023484814.

As of November 30, 2023, we systematically searched the following databases: PubMed, Embase, and Cochrane Library. The search strategy included keywords and MeSH terms related to paclitaxel, nab-paclitaxel, etc. Finally, check to ensure no additional studies are missed. The detailed search terms and search strategy are presented in the attachment.

We conducted our research in accordance with the PRISMA guidelines. To guarantee a comprehensive search, we methodically scoured multiple databases for relevant studies, including PubMed, Embase, and Cochrane Library. Our search spanned up to November 30, 2023, ensuring the inclusion of the most up - to - date information. To enhance the search process, we combined MeSH terms or emtree terms with free - text terms. This allowed us to capture a wide array of articles that met our research criteria. The detailed search strategy, including the specific terms utilized, is presented in the attachment. Below are the relevant search terms for participants and interventions:The relevant search terms for the participants and intervention factors are as follows:

The inclusion criteria are as follows:(1) Prospective phase II and III clinical trials conducted in cancer patients; (2) To calculate OR and RR, the included studies must compare nab-paclitaxel with conventional paclitaxel, or compare nab-paclitaxel with the same chemotherapy agent with the same anti-paclitaxel and the same chemotherapy agent, with the same frequency of medication; (3) The search is limited to articles and abstracts published in English; (4) Available events or incidence of muscle pain and joint pain; (5) If multiple publications of the same trial are retrieved, only the latest publication is included. This study does not require approval from the ethics committee because meta-analysis, as a secondary statistical study, has no direct relationship with the subjects.

Exclusion criteria are as follows:(1) Review articles, conference papers, meta-analyses, case reports, animal experiments, and studies that do not meet the relevant literature;(2) Literature without main outcome indicators.

Two researchers (HBL and LW) independently read the relevant literature and extract data. If there is any disagreement, consult a third investigator (ZYW). After discovering missing data in the literature, try to contact the original author for supplementation. In the process of literature screening, the first step is to screen the title and abstract, and exclude obviously irrelevant literature; the second step is to screen the full text to determine whether it can be included in this study. Each eligible study includes the following information: the first author, publication year, treatment stage, trial design, type of cancer, median age, medication plan, medication cycle, and main outcome indicators (arthralgias, myalgias).

Meta-analysis was performed using RevMan 5.3 software. When i²>50%, the random effects model is used for heterogeneity test, and the odds ratio (OR) and 95% confidence interval are used for binary variables.

Two studies (Brufsky 2021, Zhang 2021) contained three arms (PTX, nab-PTX, and combination). Following Cochrane guidance, we selected the two relevant single-agent arms (nab-PTX and PTX) for direct comparison, following Cochrane guidance for multi-arm trials. This approach preserves the original patient counts and events without mathematical combination of different treatment arms.

We primarily reported odds ratios (ORs) in our main analyses. To ensure clinical interpretability and address potential overestimation of effect sizes when event rates are high, we supplementarily calculated risk ratios (RRs), which showed directionally consistent results (Supplementary Figures S1–S3). For clinical interpretability, we supplementarily calculated risk ratios (RRs) and present them in the supplementary figures. Both measures showed directionally consistent results, confirming the robustness of our findings.

We preliminarily screened out 1887 related literatures, with 104 in PubMed, 191 in Embase, and 1592 in Cochrane. Using EndNote X9 software to delete 272 duplicate publications, 1615 remaining; after the first screening through the title and abstract, 1561 publications were excluded, and 80 publications were left; according to the inclusion and exclusion criteria, after reading the full text, a total of 9 publications were included in this study. Notably, studies such as Mobus 2021 and Gradishar 2009 were excluded as they compared nab-paclitaxel to docetaxel rather than to paclitaxel (PTX). The screening process is shown in Figure 1.

A total of nine randomized controlled trials (RCTs) were included, involving 3,699 patients with solid tumors. The nab-PTX group included 1,817 patients, while the PTX group included 1,825 patients. The studies covered various solid tumors, including breast cancer, gastric cancer, urothelial cancer, and non-small cell lung cancer. Specifically, six studies focused on breast cancer, with one study each on gastric cancer, urothelial cancer, and non-small cell lung cancer. Five studies had an open-label design, and all studies used NCI-CTCAE v3.0–v4.0 for pain grading assessment. All studies were published with complete data. Detailed study characteristics are shown in Table 1. Figure 2 shows the risk of bias graph. The results of the risk of bias assessment are summarized in Supplementary Figure S4.

Nine studies were included, involving 3,699 patients (nab-PTX group: 1,817; PTX group: 1,825). Results showed no significant difference in arthralgia incidence between nab-PTX and PTX (p = 0.29, OR = 1.07, 95% CI 0.91–1.25, I²= 0%). The forest plot is shown in Figure 3. The funnel plot (Supplementary Figure S5) showed symmetry, suggesting a low risk of publication bias.

In subgroup analysis by dosing frequency, no significant differences in arthralgia incidence were found between the q3w and q4w schedules (test for subgroup differences: p = 0.34). The point estimates varied: the q3w schedule suggested a non-significant decreased risk with nab-PTX (OR = 0.87), while the q4w schedule showed a non-significant increased risk (OR = 1.10). The 95% CIs for both subgroups crossed 1, and heterogeneity was very low (I² = 0%), indicating that the risk of arthralgia was not significantly modulated by dosing frequency.The forest plot is shown in Supplementary Figure S6. The funnel plot (Supplementary Figure S7) showed symmetry, suggesting a low risk of publication bias.

We have now performed a meta-analysis of the arthralgia subgroup using RRs. The results (RR = 1.06, 95% CI: 0.91–1.24) are consistent with the OR analysis and confirm no significant difference in arthralgia risk. This new forest plot has been added as Supplementary Figure S1.

Seven studies were included, involving 2,895 patients (nab-PTX group: 1,442; PTX group: 1,453). Results showed a significantly higher incidence of myalgia in the nab-PTX group compared to PTX (p = 0.007, OR = 1.25, 95% CI 1.06–1.48, I²= 0%). The forest plot is shown in Figure 4, and The corresponding funnel plot showed no obvious asymmetry, suggesting a low risk of publication bias (Supplementary Figure S8).

In subgroup analysis by dosing frequency, the q4w (days 1, 8, 15) subgroup showed a significantly higher incidence of myalgia in the nab-PTX group compared to PTX (p = 0.005, OR = 1.26, 95% CI 1.06–1.49, I²= 0%). No significant difference was found in the q3w subgroup (p > 0.05, I²= 0%). The interaction test (p = 0.010) indicates a significant interaction between dosing frequency and myalgia risk.The forest plot is shown in Supplementary Figure S9. The funnel plot (Supplementary Figure S10) showed symmetry, suggesting a low risk of publication bias.

Treatment regimen subgroup: The risk of myalgia was slightly higher in combination chemotherapy regimens compared to monotherapy, but the test for subgroup differences indicated that this difference was not statistically significant (p = 0.44).