Acupuncture Trials for Diabetes and Obesity Fail Rigor Tests, Landmark Audit Finds

A sweeping new audit of acupuncture research for diabetic obesity has delivered an uncomfortable verdict: the evidence base supporting one of the most widely used complementary therapies for two of the world’s fastest-growing metabolic conditions is riddled with methodological weaknesses and incomplete reporting. The systematic review, published in BMC Complementary Medicine and Therapies by a team at the First Hospital of Hunan University of Chinese Medicine, applied nine separate quality-assessment instruments to 109 studies and found that neither individual clinical trials nor the meta-analyses that pool them meet the standards that modern evidence-based medicine demands.

The scale of the analysis is itself notable. The researchers searched seven major databases from their inception through 24 July 2025, capturing 81 randomized controlled trials, 25 non-randomized studies of interventions, and 3 meta-analyses. For the randomized trials, they used the STRICTA 2010 checklist, the CONSORT 2025 statement, and the TREND statement to judge how completely the studies were reported, while the Cochrane Risk of Bias tool version 2.0, the MINORS criteria, and the ROBINS-I V2 2025 tool assessed the underlying methodological soundness. The three meta-analyses were judged against AMSTAR 2, PRISMA 2020, and the GRADE system for certainty of evidence.

The central finding is stark: overall, both the reporting and the methodological quality of studies in this field require substantial improvement. Of the six assessment tools applied to clinical trials, only the STRICTA 2010 checklist, which is specific to acupuncture interventions, and the MINORS criteria achieved relatively high compliance. That pattern is telling, because STRICTA focuses on how the acupuncture itself is described, such as needling depth, stimulation, and practitioner details, rather than on the trial design features that determine whether results can be trusted.

The weakest areas identified by the audit are precisely the ones that matter most for causal inference. Compliance related to open science practices, randomization procedures, and blinding implementation scored lowest across the clinical trials. Randomization and blinding are the twin pillars of the controlled trial: randomization ensures that treatment and comparison groups are comparable at baseline, while blinding prevents expectations of patients and assessors from contaminating outcome measurements. In acupuncture research, blinding is notoriously difficult because patients can usually tell whether they have been needled, but sham-needle designs and blinded outcome assessors exist precisely to mitigate this problem, and the audit suggests they are not being deployed or documented adequately.

Open science practices fared worst of all. This category encompasses trial registration before enrollment, publication of protocols, and sharing of data and analysis code. When trials are not prospectively registered, researchers can selectively report the outcomes that flatter the intervention, a practice known as outcome reporting bias. When protocols are not published, there is no way to verify that the analysis performed matches the analysis planned. The low compliance in this domain means that for most of the 106 clinical trials examined, readers and future meta-analysts cannot rule out that negative findings were quietly discarded.

The meta-analyses, which are supposed to distill the trial literature into reliable summary estimates, showed a different but equally concerning profile. They scored relatively well only on reporting quality as measured by the PRISMA 2020 guidelines, meaning they describe what they did in a transparent manner. But the audit found severe omissions in three technical areas: publication bias reporting, heterogeneity exploration, and sensitivity analyses. Publication bias assessment, typically through funnel plots and statistical tests for asymmetry, reveals whether small positive studies have been preferentially published while larger or null studies went unreported. Heterogeneity exploration examines whether the pooled studies differ in ways that make a single summary estimate misleading. Sensitivity analyses test whether the conclusions survive when individual studies or analytic choices are varied. Skipping all three leaves the headline conclusions of a meta-analysis resting on unexamined assumptions.

The clinical stakes are considerable. Diabetes and obesity frequently co-occur, a combination often described as diabetic obesity, and the International Diabetes Federation and World Obesity Federation both project continued growth in the burden of these conditions. Acupuncture has demonstrated distinct clinical efficacy in the intervention of diabetic obesity, according to the review’s authors, which is exactly why the quality of the underlying evidence matters. A therapy that appears promising in poorly designed trials may be overestimated, underestimated, or entirely artifactual once bias is accounted for. Clinicians considering acupuncture as an adjunct to lifestyle intervention and pharmacotherapy currently have no way to weigh the evidence with confidence.

The study’s methodological approach deserves attention in its own right. Applying nine instruments simultaneously is unusual and reflects a growing recognition that no single checklist captures the full picture. CONSORT and PRISMA measure whether authors reported what they did; RoB 2.0 and ROBINS-I measure whether what they did was trustworthy; AMSTAR 2 evaluates the rigor of the synthesis itself; and GRADE translates all of it into a judgment about how certain the resulting evidence is. By running all of these lenses over the same corpus, the researchers could triangulate where the evidence pipeline breaks down, from trial design through publication to synthesis, and their multidimensional analyses aimed to characterize the overall status of methodological, reporting, and evidence quality in the field.

The authors’ prescription is straightforward: future studies should strictly adhere to internationally recognized reporting guidelines and carry out multi-stage self-assessment together with iterative protocol refinement. In practice, that means registering trials before the first participant is enrolled, publishing the protocol, predefining outcomes, documenting the randomization mechanism, implementing sham controls and assessor blinding wherever feasible, and then running the CONSORT and STRICTA checklists against the manuscript before submission. For systematic reviewers, it means prospectively registering review protocols, formally assessing publication bias, quantifying and exploring heterogeneity, and stress-testing conclusions with sensitivity analyses before declaring that acupuncture works or does not.

For the broader field of complementary medicine research, the audit functions as both a warning and a roadmap. It demonstrates that a large and active literature can accumulate impressive volume while remaining fragile at its foundations, and it shows that the tools to diagnose that fragility already exist and are freely available. Whether the acupuncture-for-diabetic-obesity literature responds by upgrading its methods will determine if the next decade of meta-analyses can finally offer clinicians and patients a trustworthy answer about where, whether, and how needling belongs in the management of two of the defining metabolic diseases of our time.

Subject of Research: Methodological and reporting quality of acupuncture trials and meta-analyses for diabetic obesity

Article Title: Methodological and reporting quality of acupuncture for diabetes and obesity in clinical trials and meta-analyses: a quality assessment using nine tools

Article References: Fang, Y., Shan, S., Zhang, X., Xiang, K., Wang, K., Fan, Y., Chen, Y., Li, Y., Chen, Y., & Ke, C. (2026). Methodological and reporting quality of acupuncture for diabetes and obesity in clinical trials and meta-analyses: a quality assessment using nine tools. BMC Complementary Medicine and Therapies. https://doi.org/10.1186/s12906-026-05629-3

Image Credits: AI Generated

DOI: 10.1186/s12906-026-05629-3

Keywords: acupuncture, diabetic obesity, clinical trials, meta-analysis, reporting quality, methodological quality, risk of bias, CONSORT, STRICTA, PRISMA, AMSTAR 2, GRADE