The review, conducted by Mingyu Liao of Southwest University, Yue Gu of the First Affiliated Hospital of Chongqing Medical University, and Keyin Liu of Civil Aviation Flight University of China, pooled data from thirty-six randomized controlled trials encompassing 3,811 participants. The researchers searched PubMed, Embase, CINAHL, Web of Science, and the Cochrane Library from their inception through 17 December 2025, with an updated search on 19 July 2026, and also combed clinical trial registries including ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform. Only randomized trials comparing digital health interventions against usual care or conventional rehabilitation were eligible, a design choice that places the analysis at the top of the evidence hierarchy for questions of treatment efficacy. The protocol was prospectively registered with PROSPERO, and the team assessed risk of bias with the Cochrane RoB 2 tool and graded the certainty of evidence using the GRADE framework—methodological rigor that lends weight to the findings.

The primary outcomes the investigators tracked reflect the core goals of post-arthroplasty recovery: pain intensity, the total score on the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), knee flexion and extension range of motion, and performance on the Timed Up and Go test, a widely used functional measure of mobility and fall risk. Because the constituent trials measured these outcomes on different scales, the team expressed pooled effects as standardized mean differences, or SMDs, using random-effects models that acknowledge the statistical heterogeneity expected when interventions and comparators vary across studies. Subgroup analyses and univariable meta-regression were then deployed as post hoc exploratory tools to probe whether effects differed according to the dominant delivery modality, the duration of follow-up, or the geographic region where the trial was conducted.

The headline results are nuanced rather than triumphant. Digital health interventions produced a statistically significant advantage over conventional rehabilitation on two of the five primary outcomes. The pooled SMD for the WOMAC total score was −0.33 (95% confidence interval −0.59 to −0.06), favoring the digital approach, and knee flexion range of motion showed a pooled SMD of 0.30 (95% confidence interval 0.12 to 0.48), also in favor of digital delivery. For readers unfamiliar with the metric, an SMD of roughly 0.2 is conventionally considered small, 0.5 moderate, and 0.8 large, so both of these effects sit in the small-to-modest range. No statistically significant pooled differences emerged for pain, knee extension range of motion, or Timed Up and Go performance, meaning the digital interventions did not demonstrably outperform standard care on those measures across the aggregated trials.

Perhaps the most scientifically candid element of the paper is what the authors say about the clinical importance of their own findings. Although the WOMAC and flexion results reached statistical significance, the pooled SMDs were not translated into absolute changes on familiar clinical scales, and they were not evaluated against established minimal clinically important difference thresholds—the benchmarks patients and clinicians actually use to judge whether a change is meaningful in daily life. In plain terms, a statistically detectable improvement is not automatically a noticeable one. The authors conclude that the clinical importance of these effects remains uncertain, a caveat that should temper enthusiasm among clinics considering wholesale adoption of digital rehabilitation platforms on the basis of this evidence alone.

Certainty of evidence, as graded by GRADE, ranged from very low to moderate across outcomes, and critically, the two statistically significant findings were supported only by low-certainty evidence. This matters because GRADE downgrades confidence for limitations such as risk of bias, inconsistency, imprecision, and indirectness, and the trials in this field exhibit all of the above to varying degrees. The authors also flag substantial heterogeneity across the interventions themselves—some trials tested telerehabilitation with remote therapist supervision, others mobile applications delivering home exercise programs, others extended reality systems—alongside considerable variation in what the comparator groups actually received. When both the treatment and the control arm differ so widely from study to study, the pooled estimate becomes an average of averages, useful for detecting directional signals but limited in its generalizability to any specific clinical scenario.

The post hoc exploratory analyses, examining variation by delivery modality, follow-up duration, and geographic region, reflect an attempt to untangle that heterogeneity, though the authors frame them explicitly as exploratory rather than confirmatory. This is an appropriate statistical posture: subgroup findings in meta-analyses are prone to false-positive discovery, particularly when the number of trials within each subgroup is small. What the exploratory framework does offer is a roadmap for future primary research, highlighting which delivery formats, time horizons, and care contexts appear most promising and therefore merit adequately powered, pragmatically designed randomized trials with standardized comparators and pre-specified clinically important thresholds.

The practical takeaway offered by the authors is measured and clinically sensible: digital rehabilitation should be considered an adjunct to, rather than a replacement for, conventional rehabilitation after total knee arthroplasty. This framing aligns with the broader trajectory of digital health evidence across surgical specialties, where technology-mediated care most reliably adds value when it augments—rather than substitutes for—hands-on clinical expertise. For a procedure performed hundreds of thousands of times annually worldwide, even small improvements in flexion range of motion and patient-reported function, if confirmed as clinically meaningful, could translate into substantial aggregate benefit, particularly for patients in rural or underserved areas where in-person therapy access is limited. But the review’s own limitations mean that promise remains provisional.

What makes this study notable in the crowded field of digital health meta-analyses is its discipline. Rather than declaring victory for telerehabilitation, the authors report the null findings for pain, extension, and Timed Up and Go performance with the same prominence as the positive ones, quantify their uncertainty honestly, and resist the temptation to convert small standardized effects into bold clinical claims. As health systems worldwide grapple with aging populations, rising demand for joint replacement, and workforce shortages in rehabilitation services, evidence of this caliber—registered, rigorously graded, and transparently caveated—provides exactly the foundation needed for rational policy. The next step is clear: large, well-designed trials that measure absolute clinically important change, standardize comparator care, and determine which patients benefit most from which digital tools. Until then, the smartphone in the recovery room remains a promising assistant, not yet a proven substitute for the therapist’s hands.

Subject of Research: Effects of digital health interventions on rehabilitation outcomes after total knee arthroplasty

Article Title: Effects of digital health interventions in total knee arthroplasty: a systematic review and meta-analysis

Article References: Liao, M., Gu, Y., & Liu, K. (2026). Effects of digital health interventions in total knee arthroplasty: a systematic review and meta-analysis. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15640-6

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15640-6

Keywords: total knee arthroplasty, digital health, telerehabilitation, meta-analysis, randomized controlled trials, WOMAC, range of motion, rehabilitation, mobile applications, virtual reality, physical therapy, evidence certainty