A randomized two-arm, parallel group clinical trial comprised of individuals with KOA was conducted after approval by the Universidade Federal de São Carlos Research Ethics Committee (Protocol: 16767519.2.0000.5504; Approval Report Number: 3.488.466) and registration with ClinicalTrials.gov (Identifier: NCT04154059, Registered on January 2020). The study was conducted according to the guidelines of the Declaration of Helsinki and is reported based on the CONSORT and SPIRIT Extension for RCTs Revised in Extenuating Circumstances (CONSERVE).26 All participants received oral and written instructions regarding the process and aims of the study, and they gave their written informed consent before inclusion in the study.

The primary trial started in January 2020 and was conducted with the first 30 enrolled participants. Data were acquired in two blocks: the first block of individuals (n = 12) was recruited over a 3-week period from January to February 2020 (pre-COVID-19 pandemic), and the second block (n = 18) was recruited over a 4-week period from February to March 2020. Then, in March 2020, because the duration of the pandemic was unknown, we focused on mitigating the impact of the pandemic to participants already enrolled in the trial.

The COVID-19 pandemic was an extenuating circumstance due its possible effects on the population studied that involved individuals with KOA who were mostly older adults, a population that is affected more drastically and quickly from an infection.

The most impactful methodological modifications made in the trial was that the intervention for the individuals in the second block had to be adapted due to the restriction of personal contact. Prior to the pandemic, exercises were conducted in groups of a maximum of 10-12 participants at the university's facilities. In the second block, after the second week of face-to-face treatment, social distancing recommendations began. Thus, the exercise program was provided remotely to ensure continuity and conclusion of the protocol at 8 weeks.27 All individuals allocated in the first and second block performed a face-to-face baseline assessment. The evaluation after the exercise program was performed in person in an open environment following distancing and protection recommendations for both the assessor and participants with KOA. For individuals who did not feel safe going to the post-intervention assessment in person, it was performed remotely (synchronously via videoconference) following the same standards of the face-to-face assessment with regard to clinical and functional outcomes.

In addition, assessments of “foot muscle isometric strength” and “foot kinematic and kinetics during gait” were suspended. These evaluations were originally performed in a closed environment with temperature control. Due to the risk of contamination of the study population and the intent to follow social distancing recommendations issued by the World Health Organization, we decided to suspend these evaluations.

Unfortunately, we were not able to perform the complete data collection of the follow-up at 16 weeks after the end of the exercise program. Although we also proposed to perform the assessments in an open environment at week 16, following the distancing and protection recommendations for both the researcher and the study participants, most participants did not feel safe coming to the university facilities.

One of the researchers (GD) quickly planned the changes, and contacted and explained the new procedures to the participants and the supervisors of the exercises program. The changes were reviewed and approved by researchers TFS and ICNS. The data were analyzed following the decision to suspend the study.

Recruitment occurred through public announcements and lists of local and regional orthopedic and rheumatologic health services in January and February 2020. Inclusion criteria were as follows: individuals of both sexes, age range of 40 to 75 years,28 diagnosis of medial compartment KOA grade II or III based on clinical and radiological criteria of the American College of Rheumatology,29 knee pain scores ranging between 30 and 80 mm on a 0-100 visual analog scale (VAS),30 and a body mass index (BMI) less than 35 kg/m2.31 Main exclusion criteria were the diagnosis of isolated lateral compartment KOA grade II or III;27 history of physical therapy treatment or lower limb strength training in the last three months;27 history of using minimalist shoes for at least 6 h a day 5 days a week; history of intra-articular infiltration;19 and recent changes in pharmacological treatment.27

Participants randomly assigned to the intervention group, after baseline assessment, received a supervised foot-ankle exercise program focusing on strengthening of the extrinsic and intrinsic muscles and range of motion of the foot-ankle joints.27 Followed the criteria of the American College of Sports Medicine,32 the exercise program was performed three times a week for 8 consecutive weeks (24 sessions), with each session lasting around 1 h and 20 min. During the first 4 weeks, individuals performed isolated exercises to improve their foot-ankle range of motion and foot muscle strength. During the second 4 weeks, additional functional exercises were included to increase the strength of the extrinsic and intrinsic foot muscles.

The exercise program for the first block of individuals was supervised by a physical therapist in a face-to-face manner during the entire 8-week intervention. The second block of participants was also supervised in a face-to-face manner during the first 2 weeks of intervention and then remotely supervised for the remaining 6 weeks. Considering the potential for low adherence and compliance to the exercise program, a social media group was created to provide the individuals with access to videos about the exercise program, instructions on how to perform each exercise, number of sets, repetitions, rest time, and perceived effort. The physical therapist also encouraged the individuals to continue practice online and remained available throughout the session durations to support them when necessary.

Participants randomly allocated into the control group continued with usual care and treatment recommended by the healthcare team such as pharmacological treatment and/or physical activity, according to the Osteoarthritis Research Society International (OARSI) Clinical Trial Recommendation33 but performed no foot-ankle strengthening exercises.27 For compliance with ethical requirements, after the final assessment of the study, the individuals in the control group were invited to undergo 8 weeks of treatment using the foot-ankle exercise program.

Assessments at baseline (T0) and 8 weeks after the intervention (T1) were performed at the university's facilities. The primary efficacy endpoint was defined as the mean difference in the Western Ontario & McMaster Universities Osteoarthritis (WOMAC) pain subscale based on the OARSI task force between T0 and T134 (Table 1). As for the secondary endpoints, the mean difference between T0 and T1 for self-reported stiffness, function, and WOMAC total index as well as physical function performance-based tests by the 30-s chair stand test, 9-step stair climb test, and 40-m fast-paced walk test were considered (Table 1).

Offsite randomization was performed using RStudio v.1.1.463 (R Foundation for Statistical Computing, Vienna, Austria) in RStudio V.1.1.463 (RStudio, Boston, MA, USA) and processed in permuted blocks. Allocation concealment was ensured using sealed opaque envelopes inaccessible to blinded assessors.

After signing the informed consent form, eligible individuals were randomly allocated to the intervention or control group by another independent researcher 1 week after the baseline assessment. Clinical and functional outcomes were evaluated by an independent researcher blinded to group allocation, and individuals were asked not to reveal their allocation status during the study period.

A statistician blinded to the group allocation conducted all analyses before decoding the groups. The intention-to-treat approach was employed for the data analysis. To handle missing data, the data imputation method was utilized, specifically the Multiple Imputation by Chained Equations algorithm.35

Continuous variables were analyzed with the Shapiro–Wilk test of normality and are presented as means and standard deviations or medians (first and third quartile). Categorical variables are presented as counts and percentages. The Mann–Whitney U test was used to compare continuous variables between groups and Pearson's chi-squared test with Yates's continuity correction for categorical variables. ANCOVA was conducted using the intervention group as a reference and sex, BMI, and baseline measurements of each outcome as covariates. The coefficient of determination (R2) of the ANCOVA model was also reported to analyze the percentage of variation in the response to intervention explained by the model. All statistical significance was set at a two-sided p-value less than .05, also reporting the interpretation of the 95% confidence interval.36

Groups were similar at baseline (Table 2). In the ANCOVA model, after adjusting for sex, BMI, and the baseline WOMAC pain score, the intervention showed a reduction of 4.4 points (95% CI = −7.5, −1.1) on the WOMAC pain score compared to the control group, indicating a potentially clinically meaningful effect (Table 3). This model explained 49% (R2) of the variability in WOMAC pain subscale score after 8 weeks of intervention (Table 3).

The ANCOVA model showed a reduction of 7.1 (95%CI = −12.7, −1.4) and 11.9 (95%CI = −20.7, −3.1) points on the WOMAC function and total index scores, respectively, in favor of the intervention. The width of confidence intervals suggests a potentially clinically meaningful effect for both secondary outcomes (Table 3). In addition, each model explained 72% and 69% of the variability of WOMAC function and total index scores, respectively (Table 3). Regarding 30-s chair stand scores, the intervention demonstrated an improvement of 2.7 (95%CI = 1.1, 4.1) repetitions compared to the control group, with a confidence interval indicating a moderate clinically meaningful effect and the model explaining 65% of variability (Table 3).