Dear editor
Pan et al present a timely systematic review of acupuncture for chemotherapy-induced peripheral neuropathy (CIPN).1 The use of trial sequential analysis (TSA) is a useful addition because it addresses whether the accumulated sample size is adequate. I was particularly interested in the reported boundary crossing for FACT-Ntx, EORTC QLQ-CIPN20, and clinical effectiveness rate. A few reporting details would help readers assess how stable these findings are.
The main issue concerns the assumptions used to calculate the required information size. The article reports a two-sided type I error of 5%, type II error of 10%, a DerSimonian-Laird random-effects model, diversity adjustment, and a continuity correction. However, the numerical MCID and variance used for FACT-Ntx and EORTC QLQ-CIPN20 are not provided; they are described only as being derived from the current meta-analysis.1 The PROSPERO record confirms that TSA was planned for these outcomes, but it does not give the MCID, anticipated effect, variance, control event rate, or error assumptions.2 Similarly, the article calls a 25% relative increase in clinical effectiveness rate prespecified, yet that value is not shown in the public record. This does not establish that the choices were made after the analysis, but it prevents independent verification of when and how they were selected.
I would therefore ask the authors to report the exact input values for each TSA and the source of each value. Sensitivity analyses using externally derived effect thresholds would also be informative. This is relevant for FACT/GOG-Ntx because published MCID estimates differ by method: distribution-based values range from 1.38 to 3.68 points, while anchor-based estimates of about 3.54 points for improvement and 4.84 points for worsening have been reported.3–5 Showing whether the monitoring boundary remains crossed across several plausible MCIDs would make the TSA conclusion easier to interpret clinically.
The handling of influential studies also needs clarification. One study was excluded from the FACT-Ntx analysis and one from the EORTC QLQ-CIPN20 analysis after leave-one-out sensitivity analysis, with heterogeneity subsequently reported as I2 = 0%.1 Leave-one-out analysis can identify influential studies, but influence alone is not usually a reason to remove an eligible trial from the primary meta-analysis. The excluded studies and the reasons for exclusion should be stated explicitly. Unless there was an eligibility problem or data error, the analysis containing all eligible trials would be the more transparent primary result, with the exclusions presented as sensitivity analyses. The corresponding TSA results for both datasets would be helpful.
The pooled clinical effectiveness rate deserves similar caution. In the review, this outcome was defined according to each original trial.1 A response event may therefore represent different symptom thresholds, scales, or assessment times across studies. TSA can address random error in the cumulative evidence, but it cannot make non-equivalent outcome definitions interchangeable. A table listing each trial’s response definition and assessment time, followed by stratified analyses where definitions are compatible, would make the pooled risk ratio more clinically meaningful. This is especially important given the reported publication-bias signal and the very low certainty ratings.
Future reviews in this area would benefit from registering the numerical TSA assumptions, rather than only the intention to perform TSA. Future trials should also use consistent patient-reported outcomes and anchor-based thresholds that reflect how patients feel and function. These steps would make it easier to separate statistical boundary crossing from a clinically important treatment effect.
The review addresses an important supportive-care problem and brings together a substantial body of randomized evidence. Providing the missing TSA inputs, retaining all eligible studies in the primary analyses, and clarifying the response definitions would make its conclusions more reproducible and more useful to clinicians.
Funding Statement
The author received no financial support for the research, authorship, or publication of this letter.
Author Contributions
Wenhui Shi was responsible for conceptualization, investigation, and preparation of the original draft, and reviewed and approved the final version of the manuscript.
Disclosure
The author reports no conflicts of interest in this communication.
References
- 1.Pan Z, Wu J, Xie Y, et al. Acupuncture for chemotherapy-induced peripheral neuropathy: a systematic review and meta-analysis with trial sequential analysis. J Pain Res. 2026;19:631984. doi: 10.2147/JPR.S631984 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pan Z, Guo Z. Efficacy of acupuncture in treating chemotherapy-induced peripheral neuropathy: a systematic review and meta-analysis. Prospero. 2025:CRD420251272891. https://www.crd.york.ac.uk/PROSPERO/view/CRD420251272891 [Google Scholar]
- 3.Wetterslev J, Jakobsen JC, Gluud C. Trial Sequential Analysis in systematic reviews with meta-analysis. BMC Med Res Methodol. 2017;17(1):39. doi: 10.1186/s12874-017-0315-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Cheng HL, Lopez V, Lam SC, et al. Psychometric testing of the Functional Assessment of Cancer Therapy/Gynecologic Oncology Group—Neurotoxicity (FACT/GOG-Ntx) subscale in a longitudinal study of cancer patients treated with chemotherapy. Health Qual Life Outcomes. 2020;18(1):246. doi: 10.1186/s12955-020-01493-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Li T, Park SB, Chung NC, et al. Assessing clinically meaningful changes in patient-reported outcomes for chemotherapy-induced peripheral neuropathy. Support Care Cancer. 2023;31(4):236. doi: 10.1007/s00520-023-07671-2 [DOI] [Google Scholar]