The knee is the second most common site for musculoskeletal pain [1], with 11% to 17% of the cases affecting the anterior knee or the patellofemoral joint area. Muscular dysfunctions are considered an important biomechanical cause of such patella-related disorders [2-4]. Therefore, exercise therapy is widely regarded as an effective treatment strategy [4-8]. Conventional exercise therapy is typically administered on an outpatient basis under the guidance of a physiotherapist. However, the benefits of having a trained physiotherapist may be offset by drawbacks such as difficulties in scheduling appointments or a limited number of training sessions available due to regulatory limitations of German statutory health insurance. Therefore, home-based exercise interventions may present a valid alternative, as they have been reported to be at least as effective as traditional physiotherapy in reducing patellofemoral pain [9] and improving knee function [10]. Kettunen et al [11,12] found that an 8-week home exercise program alone was as effective as knee arthroscopy combined with an 8-week home exercise program in improving knee function and reducing pain. Similar findings have been reported for other knee conditions, such as meniscal tears [13-15], knee osteoarthritis [16,17], and other nonoperative knee conditions [18]. While unsupervised home-based exercise programs have demonstrated clinical effectiveness, they often lack structured feedback, personalization, and sustained engagement [19]. Frequent barriers to following physiotherapist-prescribed exercise programs also include limited social support, low self-efficacy, and a sense of helplessness [20]. These limitations have led to increased interest in digital health applications (DHAs), which aim to address these gaps by delivering exercise therapy through interactive, software-based tools, such as mobile apps or web-based platforms [21,22]. Compared to conventional physiotherapy, DHAs may also improve accessibility, especially for individuals with limited mobility or scheduling constraints [23,24], and enable more flexible and frequent engagement with exercise programs. Furthermore, they offer scalable, cost-effective solutions that can help alleviate health care resource limitations [25], as well as comprehensive possibilities to visualize exercises, educate about the condition in question, and track training sessions and progress, which may boost individual training motivation. There is evidence that DHAs may achieve better efficacy than conventional home training, for example, in treating degenerative meniscal tears [26], low back pain [27], or shoulder disorders [28]. However, only a few randomized controlled trials (RCTs) have compared DHAs directly to conventional physiotherapy covered by statutory health insurance in Germany (SHI-PT), and none of them have investigated patellar disorders. In Germany, the Digital Healthcare Act allows for DHAs to be covered by statutory health insurance once their efficacy is demonstrated. This encompasses device safety, data protection, as well as demonstrable benefits compared to standard care. Following this, DHAs have undergone rapid development in Germany since 2020. As of March 31, 2025, 40 applications have been permanently approved for prescription, while 19 applications have been preliminarily approved. This includes several DHAs for exercise therapy [29-32] that have been granted approval for coverage by German statutory health insurance.

In this context, this study aimed to evaluate the efficacy (improvement in knee function and pain) of a 12-week exercise intervention using a novel DHA (Mawendo) compared to SHI-PT in patients with disorders of the patella (International Classification of Diseases, Tenth Revision [ICD-10] code M22). The corresponding research question was as follows: is the novel DHA clinically superior to SHI-PT in improving knee function and reducing knee pain for patients with ICD-10 M22 over a 12-week intervention period?

The prespecified study protocol was approved and registered by the Institutional Ethics Committee of the Faculty of Behavioural and Social Sciences of the Chemnitz University of Technology (registration number V-439-17-CM-MAWENDO-II-18042021) in agreement with current data protection regulations. The trial was registered with the German Clinical Trials Register (World Health Organization [WHO] Primary Register) under the ID DRKS00023454 for 14 different orthopedic ICD-10 indications. All participants had to give written informed consent before enrollment in the study, and were allowed to withdraw at any time without incurring any negative consequences. The DHA Mawendo complies with all currently active data protection regulations and laws in Germany and Europe. The participants of the study did not receive any compensation, or supplemental assistance regarding the use of the DHA beyond what is generally provided. Data were deidentified before analysis.

The aforementioned 14 ICD-10 indications will be examined in 11 separate randomized controlled prospective clinical trials with 2 treatment groups. This RCT is the first among the 11 RCTs to complete the treatment of the prespecified number of patients. At the time of writing this manuscript, the 10 remaining RCTs are still in the stage of data collection. In this work, different strategies for treating diseases of the patella (ICD-10 M22) were compared between an active control group (CG) that received the current standard therapy in Germany (CG SHI-PT) and an intervention group (IG) using the DHA Mawendo (IG DHA). Participants and investigators were not blinded to the type of treatment or group allocation.

The participants in this study were recruited by orthopedists (recruiters) during their regular consultation activities at orthopedic clinics and medical practices. Patients with confirmed disorders of the patella and internet access were eligible for participation. Children (aged <12 years) were not allowed to take part in the study. Adolescents (aged <18 years) required the consent of a legal representative. The patients were fully informed about the study by the attending orthopedist and received a study information sheet outlining the study’s objectives, procedures, and treatment conditions (DHA vs physiotherapy). The pain intensity level was assessed by the recruiter before the study inclusion using a Verbal Numerical Rating Scale (VNRS, value range 0-10) [33]. Patients with very low (VNRS pain intensity ≤2) or very severe pain (VNRS pain intensity ≥8) were excluded from the study to control for disease severity. The following illnesses or physical conditions were defined as exclusion criteria:

The study data were analyzed according to the intention-to-treat principle. Therefore, the occurrence of an exclusion criterion during the study did not lead to the exclusion of participants from further data analysis. Data on additional patient characteristics, such as BMI, activity level, duration of symptoms, type of patellar disorder, and previous therapies, were not collected, as the primary aim was to evaluate the effectiveness of the DHA in real-world conditions, where broad accessibility is prioritized over detailed subgroup analyses. Collecting only essential data for the primary analysis also helped minimize participant burden. Previous experience with digital devices was not assessed, as the intervention (DHA Mawendo) was designed to be intuitive and accessible regardless of technological familiarity.

Participants were allocated to the treatment arms using simple randomization within the following strata: (1) recruiter; (2) pain medication intake of maximum WHO level 1 (yes or no); and (3) baseline VNRS pain level (3-4 or 5-7)

Simple randomization was used to impede the recruiter’s ability to predict the sequence order. Stratification by recruiter was implemented to account for potential variability in patient selection, initial assessments, and treatment initiation timelines across recruitment centers. This approach aimed to enhance internal validity and reduce confounding effects arising from differences in clinical judgment or regional health care structures. Current literature provided no evidence for potential bias between treatment arms for treatment history, stage of disease, age, or sex in previous studies and thus did not include them as strata for randomization. A browser-based web application (Rek-app; Mawendo GmbH) created specifically for recruiting purposes was provided to the recruiters. A set of 4 randomization lists (2 levels for pain medication times 2 levels for baseline VNRS pain) with 100 patients each was stored for each recruiter in the Rek-app. The statistical software R (version 3.6.2; R Foundation for Statistical Computing) was used to create these randomization lists [34]. The recruiters used the Rek-app to check exclusion criteria, include participants in the study, and randomly allocate participants to treatment arms according to the stored randomization lists (allocation sequence was concealed from the recruiters). The Rek-app provided recruiters and participants with a participant-specific ID, through which the participants gained access to the web-based survey platform. A total of 7 recruitment centers were involved with this study. Figure 1 provides a summary of the recruitment process.

The intervention for the active CG SHI-PT consisted of conventional physiotherapy (Germany’s current standard medical care), where participants selected their physiotherapists and made their appointments. IG DHA was treated using DHA Mawendo (version 1.6; certified medical device, risk class 1) and used the DHA at any suitable location.

Due to multiple testing (one hypothesis for each of the 2 primary end points), the significance level was adjusted using the Bonferroni method and set at α=0.0125 (1-tailed testing). The minimal clinically important difference was used as the effect size to be detected, and a power of 1–β=0.9 was specified to calculate the required sample size for a superiority scenario. The minimal clinically important differences for both outcomes were determined from several studies (KOOS_ADL: 10 points, SD 15 points) [57-59] and VAS: 10 points, SD 22 points [56,60,61]). This resulted in sample sizes of n=56 (KOOS_ADL) and n=121 (VAS). The more conservative (larger) sample size of n=121 was used to generate sufficient statistical power for both end points. Sample sizes were determined using the SampleSize4ClinicalTrials R package [62].

In total, 259 participants were recruited between November 10, 2021, and January 21, 2023, and randomly assigned to either IG DHA or CG SHI-PT. Among the 259 participants, IG DHA (n=136, 52.5%) received treatment using the DHA, while CG SHI-PT (n=123, 47.5%) received SHI-PT. A total of 6 participants (IG DHA: 4/136, 2.9%; CG SHI-PT: 2/123, 1.6%) provided no survey data despite several email inquiries. These participants were assumed to be missing completely at random and excluded from further analysis. Subsequently, data from 253 participants (IG DHA: n=132, 51%; CG SHI-PT: n=121, 46.7%) were analyzed. Another 6 of the remaining 253 participants provided incomplete datasets (IG DHA: 1/132, 0.8%; CG SHI-PT: 5/121, 4.1%; assumption missing at random): 0.4% (1/253) of participants did not provide PRE data, and 2% (5/253) of participants were lost to follow-up (POST data missing). These missing data were imputed for the inferential statistical analysis [74]. Figure 3 summarizes the aforementioned information in a flowchart according to CONSORT (Consolidated Standards of Reporting Trials) [75]. The demographic and clinical characteristics of IG DHA and CG SHI-PT are listed in Table 1.

In total, 118 (89.4%) out of 132 participants in IG DHA and 103 (85.1%) out of 121 participants in CG SHI-PT reported completing “all” or “almost all” treatment sessions. Additional adherence data can be found in Table 2.

A total of 2 participants (IG DHA: 1/132, 0.8%; CG SHI-PT: 1/121, 0.8%) reported a change in therapy during the course of the study due to time restrictions (IG DHA) and supplementary medication (CG SHI-PT). Of 121 participants, 3 (2.5%) in CG SHI-PT reported the occurrence of ≥1 exclusion criteria during the treatment phase. Of 3 participants, 2 (66.7%) reported an infection during treatment (answer option “infections and fever, eg, rheumatic fever, purulent arthritis, sepsis, bacterial infections”) and 1 (33.3%) participant reported an unspecified mental disorder (answer option “mental disorders, eg, acute psychosis”).

In IG DHA, the pooled mean KOOS_ADL score improved from 66.2 points (95% CI 64.3-68.1 points) to 81.9 points (95% CI 80.3-83.5 points) from PRE to POST. This corresponds to an improvement in “knee function” of 15.7 points (95% CI 13.7-17.6 points).

In CG SHI-PT, “knee function” (pooled mean KOOS_ADL score) improved by 3.5 points (95% CI 1.5-5.5 points) from 70.4 points (95% CI 68.3-72.5 points) PRE to 73.9 points (95% CI 71.5-76.3 points) POST.

Training with DHA resulted in a 4.5 times greater improvement in “knee function” compared to SHI-PT. Figure 4 illustrates this result using the imputed dataset with the largest effect for CG SHI-PT.

The difference in KOOS_ADL score between IG DHA and CG SHI-PT was statistically significant (P<.001; Table 3) and estimated at –10.1 points (-infinity to –8.0 points; adjusted 1-sided 95% CI; Table 4) by the pooled ANCOVA (factor “Intervention”).

In IG DHA, the pooled mean VAS pain score decreased from 48.2 points (95% CI 46.1-50.3 points) to 25.6 points (95% CI 23.5-27.8 points) from PRE to POST. This corresponds to an improvement in “knee pain” of –22.5 points (95% CI –25.2 to –19.9 points).

In CG SHI-PT, pooled mean VAS pain score improved by –6.5 points (95% CI –8.7 to –4.4 points), from PRE 44.7 points (95% CI 42.4-47.1 points) to POST 38.2 points (95% CI 35.3 p-41.0 points).

Thus, training with the DHA resulted in a 3.5 times greater reduction in “knee pain” compared with SHI-PT. Figure 5 illustrates this result using the imputed dataset with the greatest effect for CG SHI-PT.

In the pooled ANCOVA for VAS pain score, the group difference between IG DHA and CG SHI-PT was also statistically significant (P<.001; Table 5), and was estimated at 14.3 points (11.7 points-infinity; adjusted 1-sided 95% CI; Table 6).

During the course of the intervention, the reported use of pain medication decreased substantially in IG DHA (PRE: 35/132, 26.5%; POST: 4/132, 3.0%), while it remained almost unchanged in CG SHI-PT (PRE: 37/121, 30.6%; POST: 33/121, 27.3%). The difference POST–PRE was -23.5% (–31/132) for IG DHA, and –3.3% (–4/121) for CG SHI-PT. Only a few participants provided no information on pain medication usage (PRE CG SHI-PT: 1/121, 0.8%; POST IG DHA: 1/132, 0.8%; POST CG SHI-PT: 4/121, 3.3%).

Table 7 enumerates the patients who received concomitant care during the course of the study. In the IG DHA group, no patient used additional physiotherapy. In contrast, of 121 patients, 1 patient (0.8%) in the CG SHI-PT group underwent additional physiotherapy sessions. The other concomitant therapies were as follows: ultrasound therapy, shock-wave therapy, cortisone injection into the knee, massage, strengthening exercises, yoga, and meditation.

After extending the ANCOVA models with the factor “sex” and the covariate “age” (numerical), including interaction effects with factor “intervention,” no statistically significant isolated effect of either “sex” or “age” on “knee function” and “knee pain” was found (P value range=.28-.91; Tables 8 and 9). The effect sizes (partial η²) of “sex” and “age” are negligible and explain almost no additional variance compared to the statistical models without these factors or covariates (Tables 8 and 9). For the interaction effects “intervention:age” and “intervention:sex,” again no statistically significant effect was found for either end point (P value range=.48-.74; Tables 8 and 9). The effect sizes (partial η²) of the 2 interaction effects can be described as negligible and provide almost no additional explanation (Tables 8 and 9).

Figures 6 and 7 display a descriptive analysis of both statistically significant covariates (PRE score and recruiter), the relationship between PRE scores and treatment success (POST–PRE scores), as well as recruiter-specific effects (treatment success across different recruiters) for both end points.

This study provides evidence of the clinical superiority of DHA over SHI-PT for both primary end points, “knee function” (KOOS_ADL score) and “knee pain” (VAS pain score).

In both treatment arms, around 30% of patients used pain medication at PRE. Pain medication rate at POST decreased substantially to 3% for IG DHA, but remained largely unchanged in CG SHI-PT. The reduction in pain medication corresponds to the greater pain reduction achieved in IG DHA. Therefore, DHA may also have indirect beneficial health effects by limiting the risk of pain medication (WHO level 1), such as nonsteroidal anti-inflammatory drug–induced side effects and associated costs.

As expected, a statistically significant effect of the PRE scores was found for both end points [83]. Accordingly, participants with a worse PRE score tended to achieve greater improvements (Figure 6). However, the distribution of PRE scores was similar in both treatment groups (IG DHA, CG SHI-PT), indicating a limited risk of confounding. The control variable “recruiter” showed a statistically significant effect on the end point “knee pain,” while not being statistically significant on the end point “knee function.” Our statistical modeling results suggest that recruiting facilities affect treatment effectiveness. This effect is subject to a high degree of uncertainty, as recruiting performance across the 7 recruiting centers in our study varied considerably. In total, 93% (241/259) of all participants were recruited by 2 recruiting centers, while the remaining centers only acquired 1 to 8 participants, each. This suggests that this effect stems from small subsamples, which are more likely to produce extreme results. Figure 7 illustrates the distribution of treatment success (mean difference: POST–PRE) in both end points across all recruiters, for the end point “knee pain,” slightly lower recruiter variances led to statistical significance in our ANCOVA model (P=.04).

In both treatment arms, self-reporting revealed a comparably high level of adherence: 89.4% of IG DHA patients and 85.1% of CG SHI-PT patients attended “all” or “almost all” therapy sessions. Thus, treatment frequency, and not adherence, may provide a better explanation for the between-group differences found in our study. In addition, similar treatment changes occurred in both treatment arms (1 participant in each group (1/132, 0.8%; 1/121, 0.8%); no information for 2/132, 1.5% participants in DHA, and 7/121, 5.8% participants in SHI-PT). We found a slightly larger POST dropout rate in SHI-PT (4/121, 3.3% in SHI-PT vs 1/132, 0.8% in DHA). The reasons for this are speculative and may include the aforementioned scheduling and travel required for SHI-PT. In both treatment groups, the number of patients receiving concomitant care was minimal. No patient in the IG DHA group used additional physiotherapy, and the other concomitant therapies were comparable between IG DHA and CG SHI-PT. Consequently, any potential bias attributable to concomitant care is highly improbable.

While we controlled for “recruiter,” “pain medication,” and “pain level” in the ANCOVA model, there may be further confounding variables that were not considered in the model or during randomization. Since we found no evidence for possible bias of treatment effects due to age and sex of the participants, we did not include these covariates in the prespecification of the statistical analysis. The distribution of age and sex in the treatment arms is comparable, suggesting that neither of these variables had a systematic influence on the results. For confirmation purposes, the prespecified ANCOVA model was exploratively extended for both end points to include the covariates age and sex. We found no evidence of relevant effects of age or sex on either end point, including interactions with intervention. It can therefore be concluded that neither age nor sex has a statistically significant effect on the efficacy of the IG DHA and CG SHI-PT interventions [67-73].

Blinding was not feasible in this study, as the nature of the interventions made it impossible to conceal the treatment strategy from participants or recruiters. A placebo app was not implemented, as BfArM guidelines for the evaluation of DHAs require evidence of superiority over usual care (SHI-PT), which does not include any additional DHA. A potential concern related to the lack of blinding is expectancy effects, where participants using the DHA may have rated the DHA systematically better due to its structured progression, interactive feedback, or perceived novelty compared to conventional physiotherapy. However, any resulting bias is likely limited, as all outcome measures were based on standardized, validated patient-reported outcome scores. The study did not include direct inquiries regarding the perceived quality of the intervention. While self-reported outcomes inherently carry the risk of expectancy effects, these instruments are widely used in clinical research and have established validity and reliability in similar patient populations. All outcomes in this study were assessed using validated patient-reported instruments (KOOS_ADL and VAS). Patients in both groups completed the surveys independently via a web-based platform, minimizing the potential for interviewer and social desirability bias. However, the use of self-reported measures introduces an unquantifiable risk of bias, particularly in the absence of blinding. Expectation effects and differences in perceived engagement may have influenced how participants, especially in the DHA group, reported pain, function, adherence, and medication use. Nevertheless, randomization helps ensure that any general tendency to over- or underreport is equally distributed between groups, thereby preserving internal validity. Given the significant and large clinically relevant between-group differences, we consider the risk of substantially biased results to be low.

Adherence was assessed subjectively through self-reported data collected at the end of the intervention period, introducing the potential for recall bias. Participants may not have accurately remembered or reported the number of sessions they completed, and responses could have been influenced by social desirability bias. Future studies could consider alternative methods, such as real-time adherence tracking, to improve accuracy. We did not use log-in data to verify adherence in the IG DHA group because we could not objectively record adherence in the CG SHI-PT group, making a valid comparison impossible. All physiotherapy patients chose their physiotherapist and made their appointments. Therefore, it was not economically feasible to objectively collect adherence data for IG SHI-PT. The same applies to the specific therapy content of the sessions received by participants in CG SHI-PT. While German statutory health insurance physiotherapy follows standardized guidelines, individual treatment strategies may vary depending on the treating physiotherapist. This variability limits direct comparability between DHA and SHI-PT. A comparison of physiotherapy administered under standardized ideal conditions and DHA could potentially yield a group difference that is less pronounced than this study. However, as our study aimed to evaluate the efficacy of DHA compared to standard care under real-world conditions (intention-to-treat), rather than to directly compare digital and face-to-face physiotherapy, the resulting heterogeneity reflects typical clinical practice and, therefore, supports the external validity of our findings.

Patients who did not respond to either the PRE or POST survey generated missing values. As we conducted this study according to the intention-to-treat principle, missing values were imputed. In total, only 6 (2.4%) incomplete datasets were available for 253 analyzed patients. Of these, 5 (2%) cases came from CG SHI-PT and 1 (0.4%) from IG DHA. Consequently, the likelihood of systematic bias in the study results due to missing values is minimal. Regardless of the imputation strategy used, the study’s findings are likely to be robust against the influence of missing values, given the substantial differences between the treatment arms.

Furthermore, we did not assess participants’ previous expertise, experience, or attitudes toward DHAs in general or Mawendo specifically. While participants in the IG had the necessary resources, it remains possible that some faced technological barriers. In addition, varying levels of digital literacy could influence the ease of use and engagement with the DHA. However, such difficulties would likely bias results in favor of SHI-PT rather than DHA, suggesting that any observed treatment effects were not artificially inflated by technology accessibility. Furthermore, given the increasing integration of digital health solutions in clinical practice, these barriers are expected to diminish over time.

While this study provides evidence for the short-term efficacy of DHA-based interventions, future research should investigate their long-term impact. In this regard, one potential advantage of DHAs is that they can be used beyond the prescribed therapy duration, potentially extending their benefits over time. In addition to functional improvements and pain reduction, future studies should explore other clinically relevant end points, such as quality of life and additional physical and functional measures (eg, range of motion, strength, and mobility). These may be particularly relevant for specific populations, such as athletes aiming for an early return to sport or older adults focusing on fall prevention and independence.

DHAs may also have the potential to complement traditional physiotherapy rather than replace it. Their integration into routine care could enhance accessibility, provide continuous support between in-person therapy sessions, and facilitate personalized rehabilitation strategies. Future studies should explore hybrid models that combine DHAs with supervised physiotherapy to optimize patient outcomes. Furthermore, identifying patient subgroups that benefit most from DHA interventions could improve targeted therapy approaches, ensuring more efficient allocation of health care resources. Finally, investigations into usability aspects may help to further improve the effect of DHAs.

Our findings indicate that the investigated DHA is superior to SHI-PT for treating disorders of the patella. Sex and age did not have any effect on the intervention outcomes of the study. In addition, posttreatment pain medication intake was substantially lower for DHA compared to SHI-PT. Therefore, DHA Mawendo has been approved by BfArM for persons of all sexes aged ≥12 years for the treatment of disorders of the patella.