Nonsurgical management of hallux valgus: findings of a randomised pilot and feasibility trial

The multifaceted intervention for hallux valgus (MARVELL) trial was a parallel group, participant- and assessor-blinded, randomised pilot and feasibility trial over 12 weeks [17]. The study was registered with the Australian and New Zealand Clinical Trial Registry (ACTRN12621000645853), and the protocol was developed in consultation with the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) 2013 statement [18], the CONSORT 2010 statement extension to randomised pilot and feasibility trials [19], and items from CONSORT, as recommended by Thabane and Lancaster [20]. Ethical approval was obtained from the La Trobe University Human Ethics Committee (reference number: HEC20474).

Twenty-eight participants were recruited via from the northern suburbs of Melbourne, Victoria, Australia via postal invitation using a database of patients who received podiatry treatment at the La Trobe University Health Sciences Clinic, email distribution to staff members in the School of Allied Health, Human Services and Sport at La Trobe University, Facebook advertising and posters placed in the local community. To be eligible for inclusion, participants: (i) were aged ≥ 40 years, (ii) were female, (iii) had pain in the big toe joint/s (i.e. first metatarsophalangeal or interphalangeal) for at least 12 weeks, (iv) had big toe joint pain rated at least 3 out of 10 on a numerical rating scale, (v) were able to walk household distances (more than 50 m) unaided, (vi) were capable of understanding the English language, and (vii) had at least moderate hallux valgus on one or both feet [21]. Participants were not eligible for inclusion if they self-reported: (i) surgical treatment for hallux valgus on either foot, (ii) lower limb or partial foot amputation, (iii) an inflammatory rheumatological condition or connective tissue disease, (iv) a neurological disease which interfered with walking, (v) having worn arch-contouring foot orthoses in the past 12 weeks, (vi) having performed foot exercises in the past 12 weeks, or (vii) an injury of lower limb(s) or back that may interfere with reaching their feet.

This was a pilot and feasibility trial, so was not fully powered to detect statistically significant differences between the groups. The recommended sample size for pilot studies is 12 people per group [22], however to allow for a 15% drop-out rate, we recruited 28 participants.

Participants were recruited using postal invitation using a database of patients who had recently received podiatry treatment at the La Trobe University Health Sciences Clinic, email distribution to staff members in the School of Allied Health, Human Services and Sport at La Trobe University, Facebook advertising and posters placed in the local community. Potential participants were asked to contact the chief investigator (HBM) to express their interest and were then screened for eligibility by either of two members of the research team (HBM and PQXL).

Participant characteristics were collected by structured interview at the baseline assessment and included age, height, weight, country of birth, education level, major medical conditions, and medications. The following questionnaires and clinical assessments were also conducted: the Manchester scale for hallux valgus [21], foot pain characteristics [23], shoe-wearing history [10], the Incidental and Planned Activity Questionnaire [24] and the Credibility/Expectancy Questionnaire [25].

Permuted block randomisation (with block sizes of four, six and eight) was used to randomise participants on a 1:1 ratio to the control or intervention group using an online randomisation service (www.​sealedenvelope.​com).

All face-to-face assessments were performed in the Foot and Ankle Laboratory at La Trobe University, Victoria, Australia. Postal follow-ups were conducted at 4 and 8 weeks, with the final face-to-face follow-up at 12 weeks.

Participants were blinded to group allocation by limited disclosure, in that they were told that the clinical trial was comparing two nonsurgical treatments for hallux valgus, but they were not informed about the specific characteristics of the treatments. Research staff administering the treatments could not be blinded. Outcomes were participant-reported, thus this study was also assessor-blinded (as participants were blinded). The study biostatistician performing the statistical analyses (BE) was blinded.

The primary outcome was feasibility, which was evaluated according to demand, acceptability, adherence, adverse events and retention rate [28]. Demand was determined by the recruitment rate (participants recruited per month) and the conversion rate (participants providing consent divided by those who met the selection criteria). The recruitment rate was considered acceptable if six eligible participants were recruited per month, and the conversion rate was considered acceptable if ≥ 75% of those who were eligible participated. Acceptability of the intervention was determined using questions from the Monitor Orthopaedic Shoes (MOS) questionnaire [29] which addressed issues such as appearance, comfort, weight, and ease of donning and doffing. The intervention was considered acceptable if ≥ 75% of the intervention group scored more than 5/10 for each of questions 1–6. Adherence to the footwear/orthoses intervention was documented using 4-weekly diaries and objectively assessed over 12 weeks using a small temperature sensor embedded in the orthosis (Orthotimer^®, Balingen, Germany) [30]. Adherence was considered acceptable if ≥ 75% of participants wore the footwear/orthoses for an average of ≥ 5 h per day over the 12-week follow-up period. Adherence to the exercise program was documented using 4-weekly diaries (or the PhysiTrack^® smart-phone app) and was considered acceptable if ≥ 75% of participants attempted at least 66% of the total number of exercise sessions (i.e., 24 out of 36 sessions). In both the control and intervention groups, adherence to the hot/cold packs and bunion pads were measured using 4-weekly diaries. Adverse events were assessed at 4-weekly intervals via postal diary. Serious adverse events were defined as events that were life-threatening, required hospitalisation, or resulted in persistent or significant disability or incapacity [31]. The rate of adverse events was considered acceptable if < 15% and none were considered serious. Retention rate was the proportion of recruited participants who completed the 12-week outcome assessment. A ≥ 80% retention rate in each group was considered acceptable.

One of the secondary objectives was limited efficacy testing of the outcome measures. The key secondary outcome measure was the pain subscale of the Manchester-Oxford Foot Questionnaire (MOXFQ) [32, 33]. The minimum clinically important difference for the MOXFQ pain subscale is 12 points [34].

Other limited efficacy outcome measures included the MOXFQ [32] walking/standing and social subscales, measured at baseline and at 4-weekly intervals until 12 weeks, foot and ankle muscle strength, measured with a hand-held dynamometer using our previously documented, reliable protocol at baseline and week 12 [35], general health-related quality of life, assessed using Short Form (SF) 12 [36] measured at 4-weekly intervals, number of participants using co-interventions, documented at 4-weekly intervals, and participants’ perception of overall treatment effect, assessed with the question “Overall, how has your foot pain changed since the start of the study?” and using a global impression of change 15-point Likert scale response (ranging from ‘a very great deal worse’ to ‘a very great deal better’), measured at 12 weeks [37].

As this was a pilot and feasibility study, it was not powered to detect changes in outcome measures, so the focus was not on inferential testing (although this was conducted) [17]. Descriptive statistics were used to report feasibility outcomes. Mean (SD) scores and mean differences (95% CI) were used to explore differences in continuous variables between the groups. Differences in the MOXFQ pain subscale between groups at 12 weeks (analysis of covariance, adjusted for baseline differences) were used to inform the sample size calculation for the main randomised trial.

Between July 8, 2021, and April 22, 2022, we recruited 89 potential participants for eligibility and then randomised and tested 28 participants (aged 44 to 80 years, mean 60.7, standard deviation 10.7). This period encompassed two COVID-related stay at home orders (July 16 to July 27 and August 5 to October 21, 2021). Forty-eight people who responded were not eligible, and 13 were eligible but chose not to participate. During the study, three participants were lost to follow-up (one in the control group and two in the intervention group) and two withdrew; one was unwell, and one had to look after a family member with COVID-19 (both were in the intervention group). This led to 13 and 10 participants completing the 12 week follow-up in the control and intervention groups, respectively. Characteristics of participants are shown in Table 1, and a flow-chart of participants throughout the trial is shown in Fig. 3.

Table 2 provides a summary of the results for the feasibility outcome measures. Demand, as determined by the recruitment rate and the conversion rate (the proportion of participants providing consent of those who met the selection criteria) was not met (2.8 participants and 68%, respectively). Acceptability of the intervention, as determined using questions from the MOS questionnaire [29] was partly met, as while only 28% considered the footwear to be attractive to others and only 50% considered the footwear to be attractive to themselves, questions 4, 5 and 6 (relating to fit, ease of use and weight) met the criteria. Adherence to the footwear/orthoses intervention, as documented using 4-weekly diaries and objectively assessed over 12 weeks using the Orthotimer^®, was not met using either method (57% and 14%, respectively) nor was adherence to the exercise program (documented using 4-weekly diaries; 34% or the PhysiTrack^® smart-phone app; 7%). Adverse events were common (64%) but not serious, so were considered partly met, and overall retention was high (82%) and met the criteria.

Table 3 provides a summary of results for the pilot and feasibility outcome measures and predetermined thresholds, Tables 4 and 5 provide MOXFQ, SF12 measures and strength measures, respectively, at baseline and follow-up, and Fig. 4 shows the improvement in the MOXFQ pain subscale. Table 4 shows that after adjusting for baseline values, a statistically significant adjusted mean difference of -9.5 (95% CI -0.8 to -18.2, effect size d = 0.55) in the MOXFQ pain score was found in favour of the intervention group. Although participants in the intervention group improved in all other MOXFQ domains at 12 weeks, none of these were statistically significant. None of the SF12 change scores were statistically significant.

Mean score values in the intervention and the control groups for strength measures are displayed in Table 5. On average, participants in the intervention group exhibited greater ankle dorsiflexion and eversion strength compared to those in the control group at 12 weeks, although both findings were borderline statistically significant. Improvements were observed in other strength measures but none of these were statistically significant.

The predetermined limited efficacy threshold, based on detecting an effect size of at least 0.20, was met for the MOXFQ pain and total subscales, partly met for muscle strength (this outcome did not meet threshold for lesser toe plantarflexion but met the threshold for all other measurements) and met for SF12 mental (but not SF12 physical). The predetermined thresholds were met for use of cointerventions (14.3% of the intervention group reported use of cointerventions compared to 35.7% of the control group) but not met for perception of overall treatment effect (41.7% of the intervention group reported feeling at least ‘somewhat better’ compared to 36.4% of the control group, a difference between groups of 5%).

The cold packs were used from 0 to 56 days (mean 12.1, SD 15.6) during the 12 weeks of the study. Reasons for not using them were “I didn’t need to” (stated on 26 occasions), “bunions don’t hurt during the day” (stated on three occasions), and “cannot tolerate cold” (three occasions), “shoes were comfortable enough” (two occasions), “did not think of using them” (one occasion), “didn’t find any difference in pain” (one occasion) and “limited time” (one occasion). The bunion pads were worn 0 to 552 h (mean 146.4, SD 192.8) during the 12 weeks of the study. Common reasons for not wearing them were “no need” (stated on 22 occasions), “uncomfortable” (10 occasions), “too loose” (four occasions), “too large” (three occasions), and “I lost them” (one occasion).

Our final secondary objective was to obtain statistical parameters to inform the main trial sample size calculation. We calculated that a fully powered parallel-group superiority trial would require 122 participants (i.e.: 61 per group) using a previously-determined minimal clinically important difference in the efficacy outcome measure (MOXFQ pain subscale) of 12 points [34], standard deviation of 20.4 (recorded in the intervention group of this pilot and feasibility trial), alpha of 5% and power of 90%.