A randomized trial of a digitally delivered, home-based neuromodulation and mindfulness intervention for pain management in older adults with knee osteoarthritis

Abstract

Knee osteoarthritis (OA) is a major cause of chronic pain and disability in older adults, yet scalable home-based interventions remain limited, partly due to the lack of clinically effective digital solutions. This study is the first fully powered randomized, double-blind, sham-controlled trial to test a digitally delivered, home-based protocol combining transcranial direct current stimulation (tDCS) and mindfulness-based meditation (MBM) for knee OA pain. A total of 208 participants were randomized to active tDCS + MBM, active tDCS + sham MBM, sham tDCS + active MBM, or double sham; they completed ten 20-min sessions over 2 weeks. Knee pain intensity was measured at baseline, post treatment, and monthly for 3 months. Although both active tDCS groups improved, the difference between the combined tDCS + MBM intervention and tDCS alone was not statistically significant. Benefits were not sustained at 3 months. These findings offer the first evidence that a remotely supervised, digitally delivered tDCS + MBM intervention can rapidly reduce knee OA pain, supporting future personalized and extended treatment studies. Clinical trial registration: ClinicalTrials.gov NCT04375072 (registered May 7, 2020).

Introduction

Knee osteoarthritis (OA), characterized by joint degeneration, is one of the most common causes of chronic pain and disability in older adults. It significantly impairs mobility, reduces daily functioning, and negatively affects quality of life^1,2. Conventional treatments, including pharmacological and joint-targeted procedures and systemic analgesics^3,4, often provide only modest pain relief and carry risks of adverse effects, such as gastrointestinal issues and cardiovascular complications⁵. These limitations underscore the need for effective, scalable, and safe nonpharmacological intervention for knee OA pain management in aging populations⁶.

Transcranial direct current stimulation (tDCS) and mindfulness-based meditation (MBM) have each demonstrated clinically meaningful benefits in addressing both physiological and psychological dimensions of knee OA pain^7,8. tDCS, a noninvasive, safe, and affordable digital stimulation technique⁹, applies low-intensity electrical currents to cortical regions involved in pain regulation, including the motor, somatosensory, and prefrontal cortices^10–12. This stimulation modulates cortical excitability and enhances descending pain inhibition pathways^13–17. MBM complements this neural approach by engaging attentional and emotional regulation pathways, which can reduce pain perception and improve coping strategies^18–20. Meta-analyses have reported significant pain reductions with tDCS in persons with knee OA²¹ and mindfulness-based interventions have shown similar benefits in other chronic pain conditions²².

Combining interventions with distinct neural and behavioral mechanisms may yield additive or synergistic effects that exceed either modality alone^23–25. Delivered together, these modalities may enhance knee pain relief by reinforcing neuroplastic changes and behavioral adaptations²⁶. Early studies suggest feasibility²³, but rigorous, fully powered evidence, particularly in home-based, digitally delivered formats, is lacking^24,25. Given the growing emphasis on patient-centered, accessible care for older adults with mobility limitations, a digitally delivered, home-based protocol integrating tDCS and MBM may offer a scalable solution²³. Most prior studies have been conducted in clinic settings, which can be burdensome for this population^27,28. In contrast, remotely supervised home-based interventions can maintain treatment fidelity while enhancing access and adherence²³.

Prior clinic-based studies have demonstrated that combining tDCS with mindfulness-based approaches is safe and effective for pain management. For example, McCallion et al.²⁹ and Pimenta et al.³⁰ reported improvements in pain and attention outcomes when mindfulness practices were paired with anodal tDCS in clinical settings. Building on this work, our group^23,31,32 has demonstrated that home-based, remotely supervised tDCS, administered with or without concurrent MBM, is feasible, safe, and yields clinically meaningful reductions in knee OA pain comparable to in-clinic protocols, though in smaller samples. More specifically, these remotely delivered protocols achieved high session completion rates, no serious adverse events, and clinically meaningful reductions in pain intensity and osteoarthritis-related symptoms, with effect sizes comparable to or exceeding those reported in traditional in-clinic tDCS trials^33,34.

This study contributes to the digital transformation of chronic knee OA pain care by evaluating a highly-deployable and home-based neuromodulation intervention that is remotely supervised and digitally supported. It represents the first fully powered, randomized, double-blind trial to assess a digitally delivered protocol integrating tDCS and MBM for chronic knee OA pain. This scalable, technology-enabled approach offers a promising alternative for individuals with mobility limitations. We conducted a randomized, double-blind, sham-controlled factorial trial comparing four groups: active tDCS + active MBM, active tDCS + sham MBM, sham tDCS + active MBM, and sham tDCS + sham MBM. The primary objective was to examine the combined analgesic effects (short-term and sustained) of tDCS and MBM on knee pain intensity. The secondary objective was to investigate whether the treatment effects were associated with covariates such as BMI, age, or sex.

Results

Of the 221 individuals screened, 13 declined participation and were excluded prior to randomization, resulting in a final sample of 208 participants analyzed under the ITT principle (Fig. 1). Participants were recruited and completed the 3-month follow-up data collection from August 2020 to January 2025. Of the 208 randomized participants, data from 8 were missing, and multiple outliers in the NRS outcome were retained in analyses to ensure conclusion robustness. Based on Shapiro-Wilk test and skewness calculation (e.g., for the NRS difference between Day 10 and baseline in the active tDCS + active MBM group, test statistic: 0.95; p value: 0.04; skewness: –0.68), the knee pain intensity outcome followed a skewed normal distribution rather than a symmetric one. No missing data were reported among the remaining 200 participants during the study follow-up. All participants completed every intervention session, demonstrating 100% adherence to the tDCS protocol. No adverse events were observed, and all participants tolerated the interventions well. No device- or study-related adverse events occurred.

Fig. 1. Participant flow through screening, randomization, and analysis.

Flow diagram illustrating participant progression through the study. A total of 221 individuals were assessed for eligibility, of whom 208 were randomized. Participants were allocated to four groups: active transcranial direct current stimulation combined with active mindfulness-based meditation (tDCS+MBM; n = 53), active tDCS with sham MBM (n = 50), active MBM with sham tDCS (n = 53), and sham tDCS with sham MBM (n = 52). Withdrawals occurred in each group as indicated (n = 3, 0, 3, and 2, respectively), and 50 participants per group were included in the final analysis. MBM mindfulness-based meditation, tDCS transcranial direct current stimulation.

Baseline demographic and clinical characteristics of the 208 randomized participants are summarized in Table 1. The mean age was 67.64 years; 68.1% were female and 82.6% were Caucasian. Groups were comparable in age, sex, race, marital status, and education (all p > 0.05).

Table 1.

Demographic and baseline clinical characteristics of the participants^a (N = 208)

	mean (SD) or %
	Active tDCS + Active MBM (n = 53)	Active tDCS + Sham MBM (n = 50)	Sham tDCS + Active MBM (n = 53)	Sham tDCS + Sham MBM (n = 52)	Total (N = 208)
Age, years	66.49 (7.74)	68.06 (7.42)	67.19 (7.89)	68.88 (6.92)	67.64 (7.51)
Gender
female	34 (64.2%)	34 (68.0%)	40 (75.5%)	34 (65.4%)	142 (68.1%)
male	19 (35.8%)	16 (32.0%)	13 (24.5%)	18 (34.6%)	66 (31.7%)
Race	Op
African American	7 (13.2%)	3 (6.0%)	4 (5.7%)	6 (11.5%)	20 (9.6%)
Asian	0 (0.0%)	0 (0.0%)	5 (9.4%)	1 (1.9%)	6 (2.9%)
Hispanic or Latino	2 (3.8%)	4 (8.0%)	3 (5.7%)	2 (3.8%)	11 (5.3%)
White	44 (83.0%)	43 (86.0%)	41 (77.4%)	43 (82.7%)	171 (82.6%)
BMI, kg/m2	31.83 (8.72)	29.03 (6.95)	31.40 (8.41)	31.16 (7.69)	30.85 (8.00)
Marital Status
divorced	8 (15.1%)	9 (18.0%)	11 (20.8%)	12 (23.1%)	40 (19.2%)
living with partner	0 (0.0%)	2 (4.0%)	1 (1.9%)	1 (1.9%)	4 (1.9%)
married	35 (66.0%)	35 (70.0%)	28 (52.8%)	30 (57.7%)	128 (61.5%)
never married	6 (11.3%)	0 (0.0%)	7 (13.2%)	5 (9.6%)	18 (8.7%)
widowed	4 (7.5%)	4 (8.0%)	6 (11.3%)	4 (7.7%)	18 (8.7%)
Education
< high school	2 (3.8%)	0 (0.0%)	0 (0.0%)	0 (0.0%)	2 (1.0%)
high school	9 (17.0%)	10 (20.0%)	10 (18.9%)	12 (23.1%)	41 (19.7%)
2-year college	10 (18.9%)	6 (12.0%)	11 (20.8%)	10 (19.2%)	37 (17.8%)
4-year college	21 (39.6%)	16 (32.0%)	17 (32.1%)	12 (23.1%)	66 (31.7%)
master’s degree	10 (18.9%)	11 (22.0%)	9 (17.0%)	10 (19.2%)	40 (19.2%)
doctoral degree	1 (1.9%)	7 (14.0%)	6 (11.3%)	8 (15.4%)	22 (10.6%)
Average duration of osteoarthritis (Month)	62.49 (59.01)	34.86 (39.97)	57.10 (69.63)	50.92 (58.58)	51.56 (58.44)
Index knee
left	22 (41.5%)	26(52.0%)	28 (53.8%)	24 (46.2%)	100 (48.1%)
right	31 (58.5%)	24 (48.0%)	24 (46.2%)	28 (53.8%)	108 (51.9%)
Kellgren-Lawrence score (index knee)	2.91 (1.34)	2.86 (1.22)	2.96 (1.44)	2.80 (1.50)	2.88 (1.37)
NRS	42.40 (25.79)	44.94 (25.33)	41.67 (22.16)	42.13 (24.17)	42.80 (24.26)

MBM mindfulness-based meditation, NRS numeric rating scale, SD standard deviation, tDCS Transcranial Direct Current Stimulation.

^aKruskal-Wallis test was used for continuous variables and Chi-square test was used for discrete variables.

The mean BMI was 30.85 kg/m², consistent with Obesity Class I³⁵ and reflecting known associations between obesity and knee OA^36,37. The active tDCS + sham MBM group reported a shorter knee OA duration (34.86 ± 39.97 months) than other groups ( > 50 months), which indicated the necessity to include this variable in evaluation of intervention effects. Left- and right-knee OA were evenly distributed (48.1% vs. 51.9%). Baseline Kellgren-Lawrence scores (2.80–2.96) and NRS pain scores (mean = 42.80 ± 24.26) did not differ significantly across groups.

Figure 2 presents NRS scores for all four groups across six time points. Baseline NRS distributions did not differ significantly across groups (Kruskal-Wallis p = 0.886; Table 1, Fig. 2). By Day 5, mean NRS scores decreased in all groups: active tDCS + active MBM ( − 30.5%), active tDCS + sham MBM ( − 29.8%), sham tDCS + active MBM ( − 11.9%), and sham MBM + sham tDCS ( − 8.4%). Reductions continued by Day 10: active tDCS + active MBM ( − 19.6 points; 43.5%), active tDCS + sham MBM ( − 16.0 points; 37.3%), sham tDCS + active MBM ( − 6.3 points; 15.2%), and sham tDCS + sham MBM ( − 5.8 points; 13%).

Fig. 2. Pain intensity trajectories across intervention groups over time.

Line plot showing Numeric Rating Scale (NRS) pain scores for all four groups across six time points: baseline, Day 5, Day 10, and Months 1–3 post-intervention. Each line represents the mean NRS score for a treatment group (sham, transcranial direct current stimulation [tDCS], meditation, and tDCS combined with meditation), with different colors corresponding to each group. Error bars represent standard errors of the mean.

Compared to Day 10, NRS changes at Months 1–3 were sham tDCS + sham MBM (0.1%, 0.1%, –7.3%), active tDCS + active MBM (33.0%, 38.8%, 36.9%), active tDCS + sham MBM (11.9%, 19.5%, 3.4%), and sham tDCS + active MBM (–0.2%, –3.5%, –2.9%). Although the active tDCS + active MBM group and the active tDCS + sham MBM group had similar NRS means at Months 1, 2, and 3 post intervention, the percentage changes differed due to variations in pain levels at the end of treatment (Day 10).

Supplementary Figure 1 displays the trajectories of mean NRS scores over six time points for all four groups in this study. Knee pain in the active tDCS + active MBM group decreased most by Day 10, followed by a return toward baseline. In contrast, active tDCS + sham MBM showed moderate initial reduction by Day 10, gradual increase through Month 1 and Month 2, then the lowest score by Month 3. Sham tDCS + Active MBM exhibited a modest and stable decrease; the sham group showed minimal change.

For the primary analysis, an RLMM was used to compare changes in NRS scores from baseline to Day 10 across the four groups, adjusting for demographic and baseline clinical characteristics (Table 2). The goodness-of-fit metrics of RLMM used in the primary analysis were obtained, including a marginal R2 of 0.21, a conditional R2 of 0.75, an RMSE of 10.43, and a Sigma value of 13.07, all of which were numerically superior to those obtained using a regular LMM (i.e., a marginal R2 of 0.19, a conditional R2 of 0.73, an RMSE of 10.45, and a Sigma value of 13.75).

Table 2.

RLMM analysis for comparing NRS measures between baseline and day 10, with demographic and baseline characteristics included^a

	Time effect	p value	Group effect (between-group difference)	p value	Time × group interaction effect	p value
	Estimate [95% CI]		Estimate [95% CI]		Estimate [95% CI]
Time: Baseline vs. Day 10	–6.68 [–12.16, –1.20]	0.018*
Study group
tDCS + MBM (vs. Sham)			–2.02 [–12.28, 8.24]	0.700	–12.05 [-19.88, –4.21]	0.003*
tDCS (vs. Sham)			4.01 [–5.82, 13.84]	0.425	–10.09 [-17.72, –2.46]	0.010*
MBM (vs. Sham)			–2.06 [–11.99, 7.88]	0.685	–0.33 [–8.07, 7.42]	0.935

tDCS + MBM refers to active tDCS combined with active MBM; tDCS refers to active tDCS with sham MBM; MBM refers to sham tDCS with active MBM; and sham refers to sham tDCS with sham MBM.

RLMM Robust Linear Mixed Model, MBM Mindfulness-Based Meditation, NRS Numeric Rating Scale, tDCS Transcranial Direct Current Stimulation.

*p <0.025 (Bonferroni-corrected significance level).

^aThe significance level after Bonferroni correction is α = 0.025.

Average duration of OA was not significantly associated with NRS outcomes (p = 0.973), although baseline duration differed across groups (Supplementary Table 1). Among all demographic and clinical variables examined, only BMI was significantly and positively associated with NRS measures (estimate: 0.69, 95% CI [0.26–1.12]; p = 0.002), indicating that higher BMI was linked to increased knee pain scores (Supplementary Table 1).

The main effect of group was not significant (p > 0.4), indicating no overall difference in average pain across groups. The main effect of time was significant (estimate -6.68, 95% CI [ − 12.16, −1.20]; p = 0.018), reflecting overall significant NRS reduction from baseline to Day 10 across all groups. By Day 10, the active tDCS + active MBM group exhibited a significant reduction in knee pain (estimate = –12.05, 95% CI [ − 19.88, −4.21]; p = 0.003; Cohen’s d ≈ 0.82), as did the active tDCS + sham MBM group (estimate = –10.09, 95% CI [ − 17.72, −2.46]; p = 0.010; Cohen’s d ≈ 0.67). Both groups demonstrated significant short-term reductions in pain relative to sham, whereas the sham tDCS + active MBM group showed a modest and nonsignificant change (estimate = –0.33, 95% CI [ − 8.07, 7.42]; p = 0.935; Table 2).

Between-group comparisons indicated no statistically significant difference between active tDCS alone and combined active tDCS + MBM for short-term pain reduction. The sham tDCS + sham MBM group showed a mean pain change of –5.8 points ( ~ 13%) by Day 10, likely reflecting placebo-related or natural symptom fluctuations.

A RLMM was used to assess changes in NRS scores from baseline through 3 months post intervention, adjusting for treatment group and baseline characteristics (Table 3). There was a significant main effect of time (estimate: –8.98, 95% CI [ − 16.08, −1.88]; p = 0.014), indicating an overall reduction in knee pain across all groups (Table 3). However, none of the active interventions showed a statistically significant difference from the sham group at 3 months: active tDCS + active MBM (estimate: –3.14, 95% CI [ − 13.92, 7.65]; p = 0.569), active tDCS + sham MBM (estimate: 4.87, 95% CI [ − 5.48, 15.22]; p = 0.357), and sham tDCS + active MBM (estimate: –3.21, 95% CI [ − 13.67, 7.25]; p = 0.548). These results suggest that, while knee pain improved over time, no intervention maintained long-term effects.

Table 3.

RLMM analysis for comparing NRS measures between baseline and month 3, with demographic and baseline characteristics included^a

	Time effect	p value	Group effect (between-group difference)	p value	Time × group interaction effect	p value
	Estimate [95% CI]		Estimate [95% CI]		Estimate [95% CI]
Time: Baseline vs. Month 3	–8.98 [–16.08, –1.88]	0.014*
Study group
tDCS + MBM (vs. Sham)			–3.14 [–13.92, 7.65]	0.569	–2.13 [–12.28, 8.03]	0.682
tDCS (vs. Sham)			4.87 [–5.48, 15.22]	0.357	–7.53 [–17.42, 2.36]	0.137
MBM (vs. Sham)			–3.21 [–13.67, 7.25]	0.548	1.26 [–8.78, 11.30]	0.806

Note. tDCS + MBM refers to active tDCS combined with active MBM; tDCS refers to active tDCS with sham MBM; MBM refers to sham tDCS with active MBM; and sham refers to sham tDCS with sham MBM.

RLMM Robust Linear Mixed Model, MBM Mindfulness-Based Meditation, NRS Numeric Rating Scale, tDCS Transcranial Direct Current Stimulation.

*p <0.025 (Bonferroni-corrected significance level).

^aThe significance level after Bonferroni correction is α = 0.025.

At 3 months, group differences diminished. The active tDCS + active MBM group, despite the largest initial reduction, showed partial rebound and no longer differed significantly from sham groups. The active tDCS + sham MBM group showed continued improvement over time, yielding the lowest mean NRS score at 3 months. The active MBM + sham tDCS group demonstrated no sustained benefit beyond its initial modest effect. Among all demographic and clinical variables examined, only BMI was significantly and positively associated with NRS measures (estimate: 0.64, 95% CI [0.19, 1.09]; p = 0.005), indicating that higher BMI was linked to increased knee pain scores (Supplementary Table 2).

A RLMM model was used to accommodate NRS scores across all six time points to identify demographic and baseline predictors of pain response. BMI was the only significant predictor (estimate: 0.60, 95% CI [0.17, 1.03]; p = 0.0068). Sex-stratified effect sizes for intervention groups are presented descriptively (Supplementary Table 3).

Exploratory regression analyses examined whether BMI moderated knee pain reduction outcomes. Overall, BMI was not strongly associated with NRS knee pain reduction (Fig. 3); however, stratified analyses suggested that higher BMI was linked to greater benefit from the combined tDCS + MBM intervention for knee pain (Fig. 4).

Fig. 3. Association between body mass index and pain reduction across all participants.

Scatter plot showing the relationship between body mass index (BMI) and change in pain intensity, calculated as the difference in Numeric Rating Scale (NRS) scores between Day 10 and Day 0 (NRS reduction = Day 10 − Day 0). Each point represents an individual participant. The solid blue line indicates the linear regression fit across all participants, and the shaded area represents the 95% confidence interval.

Fig. 4. Association between body mass index and pain reduction across intervention groups.

Scatter plot showing the relationship between body mass index (BMI) and change in pain intensity, measured as the difference in Numeric Rating Scale (NRS) scores between Day 10 and Day 0 (NRS reduction = Day 10 − Day 0). Each point represents an individual participant. Colored points and corresponding regression lines indicate treatment groups: sham (red), transcranial direct current stimulation combined with meditation (tDCS+meditation; blue), tDCS alone (yellow), and meditation alone (green). Solid lines represent linear regression fits, and shaded areas indicate95% confidence intervals.

Discussion

This study represents the first fully powered, randomized, double-blind, sham-controlled trial to evaluate a digitally delivered, home-based intervention combining tDCS and MBM for chronic knee OA pain in older adults. The 2-week combined intervention produced a significant short-term reduction in knee pain immediately post-treatment; however, these effects declined over time, and group trajectories diverged across the 3-month follow-up. Throughout the section, “tDCS alone” and “MBM alone” refer to active tDCS + sham MBM and active MBM + sham tDCS, respectively.

Active tDCS produced significant short-term reductions in knee pain, both with and without concurrent MBM, whereas MBM alone did not demonstrate a statistically significant effect. Although the combined tDCS + MBM group exhibited numerically larger pain reductions, the difference between the combined intervention and tDCS alone was not statistically significant, indicating that the addition of MBM did not provide a clinically observable incremental analgesic benefit beyond tDCS in this trial^23,26. While the combined intervention showed improvement relative to double sham, the present analyses do not demonstrate an additive or synergistic interaction effect of MBM when combined with tDCS. Importantly, this study was not powered to detect additive or synergistic interaction effects between modalities, and no statistically significant difference between tDCS + MBM and tDCS-only was observed. Taken together, these findings suggest that tDCS was the primary driver of short-term analgesic effects, while the contribution of MBM remains uncertain.

The observed short-term pain reductions in both active tDCS groups are consistent with prior home-based tDCS research, including a large randomized control trial reporting Cohen’s d = 1.2³⁸, while meta-analyses of mindfulness-based interventions have generally reported smaller effects on chronic pain outcomes (Cohen’s d = 0.19–0.32)³⁹. These findings reinforce the role of neuromodulation as an effective short-term, nonpharmacological approach for knee OA pain management in older adults.

Prior studies have suggested that concurrent neuromodulation and mindfulness engagement may optimize pain regulation through complementary neurophysiological and cognitive-affective mechanisms^26,40,41. For example, Rodrigues et al.⁴⁰ demonstrated an additive effect in anxiety treatment when D-cycloserine enhanced exposure therapy, and Pimenta et al.²⁶ reported improvements in attention and headache-related disability when mindfulness was combined with anodal tDCS over the left dorsolateral prefrontal cortex in migraine patients. Additive effects have also been described in depression and migraine populations when tDCS is paired with cognitive or mindfulness-based therapies^26,41 However, despite such prior evidence, the present trial did not demonstrate an additive or synergistic interaction effect of MBM when combined with tDCS. Accordingly, while prior literature supports the plausibility of synergistic mechanisms, confirmation of such effects will require trials specifically designed and powered to evaluate interaction effects using factorial designs or prespecified contrasts.

Sustained benefits of the intervention were limited. By one month, knee OA pain scores in the combined active group increased and no longer significantly differed from those in the sham groups, consistent with the time-limited effects of brief tDCS and behavioral protocols^42–44. NRS scores increased by approximately 12–39% from Day 10 to the 1- or 2-month follow-up in both active tDCS+active MBM and active tDCS + sham MBM groups, suggesting that intervention effects lasted less than 1 month. These trajectories suggest that, while the combined intervention produced the greatest immediate knee OA pain relief, tDCS alone may yield more gradual but comparatively sustained benefits over time. At the primary endpoint, both active tDCS conditions produced clinically meaningful short-term reductions in knee OA pain⁴⁵.

These findings suggest that, without ongoing maintenance, the effects of both tDCS and MBM diminish over time. tDCS-induced cortical excitability changes are transient, and meditation-related benefits may wane without continued practice. Sustaining gains in chronic knee OA pain may therefore require continued neuromodulation or psychological reinforcement beyond the initial treatment period. Evidence from prior studies suggests that extended or repeated tDCS protocols (e.g., 15–20 sessions) and continued MBM practice may enhance durability. Home-based booster sessions or ongoing MBM practice may offer flexible and scalable options for maintaining benefits.

In knee OA, reinforcing central inhibition and coping mechanisms is critical for achieving durable neuromodulatory and behavioral change^46–49. Clinically, sustaining treatment benefits may require structured maintenance strategies, such as gradual tapering of stimulation sessions (e.g., three sessions in Week 3, two in Week 4, one in Week 5) or continued MBM practice supported by refresher resources or app-based tools⁵⁰. Future studies should determine optimal dosing schedules and evaluate whether flexible maintenance protocols outperform fixed protocols in sustaining long-term pain relief⁴⁸.

Prior literature has suggested that biological sex may influence responsiveness to neuromodulation and mind-body interventions targeting pain regulation^49,51. However, the present study was not powered to formally test sex-by-treatment or sex-by-time interaction effects, and no statistical comparisons of treatment response by sex were conducted. Accordingly, any sex-stratified patterns suggested in prior work, or observed descriptively in this study, should be interpreted cautiously and viewed as exploratory rather than confirmatory. Future trials specifically designed and powered to evaluate sex-related neurobiological and psychosocial moderators may help inform more personalized multimodal pain management strategies.

Higher BMI was associated with greater knee pain intensity scores, likely reflecting increased joint loading, inflammation, and altered pain modulation⁵². Group-stratified analyses (Fig. 4) indicated that participants with higher BMI in the combined tDCS + MBM intervention group experienced greater reduction in knee pain, consistent with prior evidence that individuals with obesity may respond more strongly to neuromodulation due to elevated baseline symptom burden⁵³. Higher BMI is often associated with central sensitization and impaired endogenous pain inhibition⁵⁴, which may increase responsive to interventions targeting both neural and psychological pathways²⁶. This finding supports the value of personalized approaches that considers BMI when optimizing multimodal pain interventions^55,56. Despite variability in home settings, the observed effects underscore the ecological validity of the intervention and its real-world potential⁵⁷. Future studies should explore whether tailored or booster protocols can help to sustain these benefits across BMI profiles.

Several limitations should be acknowledged. Although the study was designed to blind participants, neither treatment expectancy nor the success of participant blinding (e.g., post-intervention assessment of perceived treatment assignment) was formally assessed, which should be considered when interpreting the findings and may have implications for internal validity. With respect to intervention delivery, despite standardized protocols and real-time remote supervision, some variability in home-based tDCS administration may have occurred; however, devices were preprogrammed, electrode placement was verified, and staff were trained to minimize this risk. In terms of generalizability, participants were relatively well-educated older adults with access to technological support, and digital literacy may remain a barrier to broader implementation. Also, the MBM intervention was brief (2 weeks) and self-guided, without a live instructor component or formal fidelity assessment, which may have reduced the robustness of MBM effects compared with longer, instructor-led programs. Finally, although digital and CD options were provided, the requirement for access to playback devices and secure videoconferencing may have unintentionally excluded individuals with limited technology access.

Strengths of this trial include its large sample size, rigorous randomized double-blind sham-controlled design, and use of a fully powered analytic framework appropriate for longitudinal pain outcomes. The study is among the first to evaluate a remotely supervised, digitally delivered tDCS protocol at scale in older adults with knee OA, demonstrating high adherence, feasibility, and safety in real-world home settings. The inclusion of multiple control conditions allowed for a clear separation of tDCS-specific effects from placebo and nonspecific intervention effects. The use of robust linear mixed-effects modeling enhanced the reliability of findings in the presence of non-normal outcome distributions. Collectively, these strengths support the internal validity of the trial and underscore its contribution to advancing scalable, nonpharmacological pain management strategies for older adults with mobility limitations.

In conclusion, home-based tDCS produced clinically meaningful short-term reductions in knee OA pain, while MBM did not demonstrate independent or additive analgesic effects in this trial. Future studies should refine dosing, evaluate extended delivery and maintenance strategies, and employ trial designs powered to evaluate interaction effects in order to determine whether combined neuromodulation and mindfulness approaches can yield durable and personalized benefits.

Methods

This study was approved by the Institutional Review Board at the University of Arizona (STUDY00003164) and registered at ClinicalTrials.gov (Identifier: NCT04375072; First posted: May 5, 2020). Written informed consent was obtained from all participants in accordance with the Declaration of Helsinki. The full trial protocol and statistical analysis plan can be accessed through the ClinicalTrials.gov study record (https://clinicaltrials.gov/study/NCT04375072). Recruitment occurred from August 2020 to January 2025. Participants completed follow-up assessments through 3 months post-intervention.

We conducted a double-blind, randomized, sham-controlled, Phase II parallel-group trial. A total of 208 participants with knee OA were randomized to one of four groups: active tDCS + active MBM (n = 53), active tDCS + sham MBM (n = 50), active MBM + sham tDCS (n = 53), or sham tDCS + sham MBM (n = 52). The participant flow diagram is shown in Fig. 1. Participants were randomized (1:1:1:1) using a pre-generated list created in R (version 4.4.3; R Foundation, Vienna, Austria) by the study statistician, balanced for age, race, and sex. Allocation was concealed in opaque envelopes.

Eligible participants were ages 50–85 and met American College of Rheumatology (ACR) clinical criteria for symptomatic knee OA⁵⁸. Additional inclusion criteria were (a) knee OA pain within the past 3 months with an average Numeric Rating Scale (NRS) pain score ≥30 (0–100 scale), (b) English proficiency, and (c) no planned changes to pain medication during the trial. ACR clinical criteria required at least three of the following: (a) age >50, (b) morning stiffness <30 min, (c) crepitus, (d) bony tenderness, (e) bony enlargement, or (f) absence of palpable warmth. Key exclusion criteria were (a) history of brain surgery, tumor, seizure, stroke, epilepsy, or intracranial metal implants; (b) systemic rheumatic disorders (e.g., rheumatoid arthritis); (c) cognitive impairment (MMSE ≤ 23); (d) prosthetic knee replacement or nonarthroscopic knee surgery; or (e) no device access for secure video conferencing.

After the Institutional Review Board of the participating institution approved the research protocol, study participants were recruited using a multipronged community-based and clinic-based strategy. Recruitment methods included advertisements placed in local institutions and community locations (e.g., retail establishments, bus stops, buses, and local print and electronic media). Additional recruitment occurred through participation in community health fairs and educational programs organized by institutional and community partners, as well as through university-affiliated hospitals and local rheumatology clinics.

Active tDCS was delivered at 2 mA for 20 min per session, 5 days per week for 2 weeks. Participants self-administered stimulation using a 1 × 1 mini-CT Stimulator (Soterix Medical Inc., NY) with 5 × 7 cm saline-soaked sponge electrodes and single-position headgear to ensure consistent placement. The anode was placed over the primary motor cortex (C3 or C4) and the cathode was placed over the supraorbital region (Fp1 or Fp2), following established protocols⁵⁹. Devices were preprogrammed with a one-time unlock code and automatically powered off after 20 min, per manufacturer protocol. Each session was digitally supervised in real time via secure videoconferencing. Study staff verified electrode placement, initiated device activation using a one-time code, and continuously monitored participants during the 20-min stimulation to ensure safety and adherence. Session attendance and completion were documented electronically. Sham tDCS used the same montage but delivered only a 30-s ramp-up and ramp-down to mimic sensations while delivering no active current beyond the initial period. This validated sham approach ensured that both participants and researchers remained blinded^25,33.

Adherence to the study protocol was promoted throughout the trial. First, participants received baseline training until they demonstrated correct use of the tDCS device. Second, participants were remotely supervised by trained research staff at each stimulation to ensure proper technique; to monitor for adverse events; and to verify that participants safely operated the device, tolerated the session without adverse effects, and followed the study schedule. Third, the research team was trained using detailed procedural manuals on all aspects of the proposed research, including treatment protocols and participant interaction, in a step-by-step fashion. All personnel were trained before study initiation, and the principal investigator conducted weekly supervision to ensure protocol adherence.

MBM was delivered concurrently for 20 min per session over 2 weeks. Participants listened to a pre-recorded meditation CD guiding them through progressively deeper mindfulness practices, including controlled breathing, body awareness, and compassion. Participants accessed the guided meditation through a digital audio file compatible with computers, tablets, or smartphones; CDs and CD players were provided upon request. The same pre-recorded track was used across all sessions in either the MBM or sham MBM conditions to ensure consistency in content, duration, and structure. Sham MBM used identical equipment and format but excluded mindfulness content, instead instructing participants to relax and take deep breaths every 3 min. Deep breathing was incorporated in both conditions to control for relaxation and respiratory effects; however, only the active MBM intervention included sustained mindfulness and compassion-based practices intended to engage cognitive–affective pain modulation mechanisms. Both versions were standardized for posture, duration, and instruction²³.

Participants, assessors, and analysts were blinded. The tDCS devices were identical and preprogrammed with built-in double-blinding features requiring a five-digit code to initiate stimulation. MBM and sham MBM were structurally matched, which means all participants were informed they would receive a meditation recording. Further, all staff were trained to minimize bias.

At baseline, demographic characteristics, including age (continuous), gender (male vs. female), race (African American, Asian, Hispanic or Latino, or White), body mass index (BMI; kg/m²), marital status (divorced, living with partner, married, never married, or widowed), and education ( < high school, high school, 2-year college, 4-year college, master’s degree, and doctoral degree), as well as clinical characteristics, such as the average duration of OA (months), index knee (the most affected knee), and Kellgren-Lawrence score (the severity of radiographic OA)⁶⁰ were collected.

Pain intensity was primarily assessed using the NRS, where participants rated average knee OA pain over the past 24 h on a scale of 0 (no pain) to 100 (worst pain imaginable)⁶¹. The NRS is a validated, reliable tool (Cronbach’s α ≥ 0.8) widely used in knee OA studies and is suitable for older adults due to its simplicity^62,63. Ratings were collected at six time points: Day 0 (baseline), Day 5 (mid-treatment), Day 10 (end of treatment), and at 1, 2, and 3 months post intervention.

This double-blind, randomized, sham-controlled Phase II trial used a 2 × 2 factorial design, with participants assigned to one of four groups at equal probability: (a) active tDCS + active MBM, (b) active tDCS + sham MBM, (c) sham tDCS + active MBM, or (d) sham tDCS + sham MBM. The target enrollment of 200 participants (assuming 10% attrition) yielded an effective sample size of 180, providing > 80% power to detect a ≥ 2-point reduction in NRS knee pain, which is considered clinically meaningful for chronic knee pain. Based on prior home-based tDCS and MBM data (Cohen’s d = 0.8–1.8), a conservative effect size of 0.4 and a Bonferroni-adjusted α = .025 were assumed (two tests: baseline vs. Day 10, and Day 10 vs. follow-up). Power estimation used a robust linear mixed-effects model that accounted for within-subject correlations across repeated measures (baseline, Days 5 and 10, and Months 1–3). With anticipated >90% retention, the design ensured adequate power to detect intervention efficacy under the intention-to-treat framework.

Raw data were manually reviewed for quality checking; outliers and missing values were retained to preserve variability and avoid introducing bias. Descriptive statistics summarized baseline characteristics. The normality of NRS knee pain scores was assessed using the Shapiro-Wilk test. A robust linear mixed-effects model (RLMM)⁶⁴ was employed to accommodate the skewed distributions of outcomes and evaluate intervention effects at post intervention (Day 10) and across the full follow-up period (baseline, Days 5 and 10, and Months 1–3), adjusting for covariates, including BMI. The primary endpoint was NRS knee pain scores at Day 10 (post-intervention), and the secondary endpoints were NRS knee pain scores at 1-, 2-, and 3-month follow-up visits to assess sustained effects. Sustained effects and potential demographic or baseline predictors (e.g., age, sex, BMI) were examined within the same modeling framework. A Bonferroni-adjusted significance level of α = 0.025 was applied. All statistical analyses were conducted using R. This double-blind, randomized, sham-controlled, Phase II parallel-group trial focused on efficacy rather than feasibility outcomes. Adherence was tracked through session attendance logs during each remotely supervised session, and any missed sessions were recorded. Missing data were handled using an intention-to-treat approach to ensure robust conclusions.

Author contributions

J.P. designed the study, developed the research questions, and drafted the manuscript. C.L. reviewed and edited the full manuscript, with particular attention to statistical analysis and reporting of results. L.L., M.J.F., I.T., S.D., L.P., and J.H. reviewed the draft and provided constructive feedback throughout the revision process. H.M. conducted the data analyses and contributed to the statistical interpretation and reporting of findings. HA contributed to the study design, conceptual framework, and secured the R01 funding that supported the project.

Data availability

All data supporting the findings of this study are summarized in this published article. Requests for additional de-identified data will be reviewed by the corresponding author on a case-by-case basis and may require institutional and ethical approval.

Code availability

No custom code was used in this study.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41746-026-02577-7.