This was a three-arm, parallel, and multicenter RCT conducted at four teaching hospitals in China between October 2017 and November 2020. Eligible patients were recruited from the outpatient departments of Acupuncture and Moxibustion, Integrated Chinese-Western Medicine, and Rehabilitation Medicine. This trial was approved by the ethics review board of the Bioethics Subcommittee of West China Hospital, Sichuan University (Approval No. 228), registered at ClinicalTrials.gov (No. NCT03299439), and overseen by an independent data monitoring committee (DMC). We conducted a pilot study to inform the design and feasibility of the current trial, suggesting the need to include a waiting-list control group. We thus used a three-arm design to help evaluate assay sensitivity of this trial. The trial protocol was previously published. We followed the CONSORT and STRICTA guidelines to report this study.
Patients were included if they were 40 years or older and diagnosed with mild or moderate KOA (Grade 0-III, according to Kellgren-Lawrence criteria). The diagnostic criteria were followed the Chinese Guideline for the Medical Management of KOA. All patients signed the informed consent form to participate in the trial.
Patients with any of the following conditions were excluded: diagnosed with tuberculosis, tumors, rheumatism of the knee joint, and rheumatoid arthritis; sprain or trauma in the lower limb; present with mental disorders; present with comorbidities that severe cardiovascular disease, liver or kidney impairment, immunodeficiency, diabetes mellitus, blood disorder or skin disease; pregnancy or lactation; use of physiotherapy for osteoarthritis knee pain in the immeidiate past month; use of intra-articular injection of glucocorticoid or viscosupplementation in the past 6 months; received knee-replacement surgery; and positive floating patella test.
Randomization and Blinding
Patients were randomly allocated, at a ratio of 1:1:1, into a lower PT group (LPT group, that is, acupuncture at acupoints with lower PT), a higher PT group (HPT group, that is, acupuncture acupoints with higher PT) or a waiting-list group (no acupuncture). The randomization was onducted via a central randomization system, and the randomization sequence was generated in a block size of 3 or 6 and stratified by participating sites. Assignment of patients was performed by an independent coordinator. Patients in LPT and HPT groups were blinded to acupuncture treatment (patients in LPT group and HPT group were informed receving the same acupuncture treatment), and were required not to release their treatment information to outcome assessors during the study. Outcome assessors and data analysts were blinded to treatment allocation.
Identification, measurement and selection of acupoints
Identification of acupoints
We identified 13 acupoints around knee joint for treating KOA according to literature and expert consensus. The acupoints included Heding (EX-LE2), Neixiyan (EX-LE4), Dubi (ST35), Xuehai (SP10), Liangqiu (ST34), Yinlingquan (SP9), Yanglingquan (GB34), Zusanli (ST36), Weizhong (BL40), Yingu (KI10), Xiguan (LR7), Ququan (LR8) and Weiyang (BL39). In addition, we also identified ashi point from 12 testing areas around knee based on anatomical structure and expert consensus (eFigure 1 in Supplement 1).
Measurement of PT of acupoints
Trained acupuncturists measured PT with the electronic von Frey detector (2390 series, IITC Life Science). Each point was tested twice at an interval of 2 minutes. If the difference between the two values was greater than 15g / N, a third measure was made at this point. The average of two values with the smallest difference was recorded as the final PT of the tested acupoint.
Selection of acupoints for interventions
The acupoints were ranked by the PT value. The five points with lowest PT were identified as lower PT acupoints (correspoinding to LPT group), and the five with the highest PT as higher PT acupoints (corresponding to HPT group).
Sterile, single-use needles (Hwato Needles, Sino-foreign Joint Venture Suzhou Hwato Medical Instruments, China) with a length of 40 mm and a diameter of 0.30 mm were used. Acupuncture was performed by an acupuncturist who did not participate in identification and measurement of acupoints. The acupuncturists were specialists in Traditional Chinese Medicine at the hospitals and received specialized acupuncture training.
In the LPT group, patients received acupuncture treatment at five lower PT acupoints. After skin disinfection, the needles were inserted vertically into the acupoints with a depth of 15-30 mm. The stimulation was performed with lifting and thrusting combined with rotating to induce the sensation known as de qi (sensation of soreness, numbness, distention, or radiating). Patients received 12 sessions of acupuncture (three sessions per week or every other day) for four consecutive weeks. Each session lasted 30 minutes and acupuncture needle manipulation was performed every 15 min.
In the HPT group, patients received acupuncture at five higher PT acupoints, and all the other treatment settings were the same with those in the LPT group.
In the LPT and HPT groups, patients with unilateral KOA were treated with acupuncture on the affected side. Patients with bilateral KOA were treated and assessed on their most painful side, and the non-trial affected low limbs were provided with acupuncture treatment on ST35, EX-LE4, GB34, ST36 and SP10. Patients in the waiting-list group did not receive any acupuncture from the beginning of the trial but were informed that they would be offered with 12 sessions of acupuncture treatment for free after the study.
All the patients were advised not to use any other treatments for KOA. However, if the patient had intolerable pain and the outcome assessment was not scheduled within the next 48 hours, NSAIDs were allowed. For patients in the waiting-list group, non-acupuncture treatments, such as application of medicinal liquor on the knee, massage and moxibustion, were allowed if they requested treatment. All the above treatments were recorded, including the name, dosage, and duration of treatment.
The primary outcome was the change of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) total score (a composite total score of pain, stiffness, and physical function) from baseline to 16 weeks. The WOMAC, a disease-specific scale with high reliability and validity, has been translated into different languages and used widely in clinical trials for KOA. The Chinese version of WOMAC contains 24 items that measure pain (5 items, scored 0-50), stiffness (2 items, scored 0-20) and physical function (17 items, scored 0-170), with a total score ranging from 0 to 240. Each WOMAC item is rated on a Visual Analogue Scale (VAS) of 0 to 10, with a high score indicating a worse symptom.
The secondary outcome included the change of WOMAC subscale score (pain, stiffness and physical function), 12-item Short Form Health Survey (SF-12), active knee ranges of motion (ROM) and adverse events. The validated Chinese version of SF-12 consists of eight domains and may generate two separate summary scores including physical component score (PCS) and mental component score (MCS) , this self-report questionnaire measures quality of life. The active ROM were assessed by using a standard goniometer, including flexion, extension, internal rotation and external rotation. Any adverse events, such as bleeding, fainting, bruising and soreness at needle sites were documented and followed up during the study.
All outcome measures were performed at baseline, 4, 8, 12 and 16 weeks. The outcome assessments, following a standard protocol, were performed in a separate room at the outpatient department of the research site by trained nurses, who were blinded to treatment allocation.
Sample Size Estimation
The sample size calculation was based on the mean difference of WOMAC total score changes from baseline to 16 weeks according to the pilot trial. A sample size 189 patients per group was required to achieve 90% power to detect a mean difference of WOMAC total score of 12 between LPT group and HPT group when the standard deviation (SD) was 33, and the significance level was 0.025 (adjusted for multiple testing) for a two-sided test. This sample size was sufficient to detect the difference between LPT group and waiting-list group, given that the treatment effect between LPT group and HPT group would be smaller than that between LPT group and waiting-list group. The sample size of 222 individuals per group (total of 666) was allowed considering 15% loss to follow up.
An independent data monitoring committee (DMC), consisting of five members with epidemiologist, biostatistics, acupuncturist and specialist physicians, was developed to monitor the quality and regulatory compliance of the trial, and ensure the safety of participating patients. We developed a procedural document for the DMC meeting, and strictly followed the document. Two DMC regular meetings were held during this study.
Baseline characteristics and clinical outcomes were summarized using means and SDs for continuous variables, and counts and percentages for categorical variables. The primary outcome was analyzed according to the intent-to-treat principle (i.e., full analysis set), which included all patients randomized. Missing data on primary outcome were multiply imputed using chained equations with predictive mean matching under missing at random assumption, and the estimates from 200 imputed datasets were combined by rubin rules. The change from baseline in WOMAC total score over time was analyzed by fitting a linear mixed-effect model that included baseline value as a covariate; modeled treatment, visit and treatment × visit interaction as fixed effects; and treated sites and individuals as random effects. The same approach was used for secondary continuous outcomes.
Four prespecified subgroup analyses of Kellgren-Lawrence criteria (Grade 0, Grade I, Grade II and Grade III), BMI (<18.5, 18.5-24 and >24 kg/m2), duration of disease (<5, 5-10 and >10 years), and unilateral/bilateral KOA were conducted for the primary outcome that comparing LPT group versus HPT group by adding interaction term (ie, BMI × group) into the linear mixed-effect model.
A sensitivity analysis was performed for the primary outcome basing on the per-protocol set, which included patients who adhere to the treatment sessions and follow ups. All analyses were implemented using R software (version 3.6.3).