Reproducibility of foot dimensions measured from 3-dimensional foot scans in children and adolescents with Down syndrome

Data were obtained from a previous feasibility study investigating the efficacy of custom-fitted footwear to increase physical activity levels in children and adolescents with Down syndrome [21]. The study was approved by the La Trobe University Human Ethics Committee (HEC16–027) and written informed consent was obtained from parents or guardians. Where appropriate, children and adolescents also provided written assent for participation prior to enrolment [21].

Participants were children and adolescents aged 5 to 17 years with Down syndrome. Participants were excluded if they had any health condition that may affect physical activity (e.g. inflammatory arthritis, subluxation etc.) as reported by their parents, or required the use of an ambulatory device (e.g. cane, crutches or walker). Participants were recruited through a member-based disability organisation for individuals with Down syndrome that was based in the community [21].

Two raters (NMH and AKB) performed the measurements for all foot scans, and measurements were repeated twice within 4 weeks to assess intra- and inter-rater reproducibility. Both raters were registered podiatrists with four and 14 years of clinical experience, respectively. Rater 1 had 3 months and rater 2 had 5 years of experience using 3D scans to measure foot dimensions, respectively.

Participants had their height and weight measured to calculate body mass index (BMI). Measurements were for the right foot only. Foot posture was assessed using two indices; the Foot Posture Index [22] and the Arch Index [23]. For both indices, higher positive scores indicate a flatter foot posture. The presence of lesser digital deformities (i.e. hammer, mallet and claw toes) was documented [10, 24]. The presence and severity of hallux valgus deformity was assessed using the Manchester scale [25]. The degree of deformity was graded on a scale of 0 to 3 (no deformity, mild, moderate and severe). Scores for hallux valgus were dichotomised, where scores of 0 and 1 were graded as absent and scores of 2 and 3 were graded as present [26].

Participants stood in a relaxed, full weight-bearing position and a 3D scan was taken of their right foot using the FotoScan 3D scanner (Precision 3D, Weston-super-mare, UK). The FotoScan 3D device uses a fixed system of cameras and projectors to obtain images of the foot, that are automatically converted into a 3D model [27]. According to the manufacturer, the scans obtained with this system are accurate to within less than half a millimetre. The 3D foot scans were then exported as stereolithography (STL) files (Fig. 1). The 3D-Tool© Version 13 (3D-Tool GmbH, Weinheim, Germany) was used to obtain all length, width and height measurements. For girth measurements, a cross-section of the foot at the relevant landmarks was created and exported as a drawing exchange format file. The perimeter of the cross-section was determined using Canvas© 11 software (ACD Systems International, Seattle, WA, USA).

Prior to data collection, pilot testing of the foot measurements was completed. A measurement technique protocol was developed (see Additional file 1). The protocol was piloted on the scans of five participants and the results were compared. Where there was substantial variability in results, the measurement technique was clarified between raters until consensus was reached. Each rater worked independently during data collection and compiled measurements on separate Excel spreadsheets (Microsoft® Office 365, Microsoft, Redmond, WA, USA) on two separate occasions that were two weeks apart.

We measured 13 foot dimensions that are relevant to footwear manufacturing [17, 28, 29]. These were (Figs. 2 and 3):

An a priori sample size estimation using the Power Analysis and Sample Size software (PASS 15 software, NCSS, LLC, Kaysville, UT, USA) for the ICC test, alpha of 0.05, power 0.80 and ICC of 0.60 determined that a minimum sample size of 19 participants (foot scans) was necessary. However, as we had access to foot scans of 30 children and adolescents with Down syndrome, all scans were measured. Analyses were performed using IBM SPSS Statistics version 25.0 (IBM Corp, NY, USA) and STATA SE Version 14.2 (StataCorp, College Station, TX, USA), using the kappaetc module. To satisfy the independence assumption of statistical analysis, measurements from the right foot only were analysed [30]. For continuous data, data were assessed for normality using skewness and kurtosis tests, and data were found to be normally distributed. For continuous data, reliability was calculated using ICCs with 95% confidence intervals (ICC [1, 2], consistency) [31]. Interpretation of ICCs were based on definitions provided by Portney and Watkins [32], where ICC values less than 0.5 was considered as poor, values between 0.5 to 0.75 was considered as moderate, values between 0.75 to 0.90 was considered as good and any value above 0.90 was considered as excellent reliability. For nominal data, reliability was determined by calculating Gwet’s AC1 statistic [33] and results were interpreted according to Landis and Koch cut-offs, which are: less than zero was considered poor, 0 to 0.20 slight, 0.21 to 0.40 fair, 0.40 to 0.60 moderate, 0.60 to 0.80 substantial, 0.81 to 1.00 almost perfect reliability [34]. For continuous data, agreement was determined by calculating limits of agreement (LOA) [35]. For the calculation of LOAs, the presence of heteroscedasticity was assessed, and when present, ratio LOAs were calculated by taking the antilog of the calculated LOA values [36]. For nominal data, agreement was determined by using percentage agreement.

Table 2 summarises the intra-rater reproducibility of all 13  foot measurements. The intra-rater reliability of rater 1 was good to excellent for 11 out of 13 measurements, with ICCs ranging from 0.76 to 0.99. Fifth toe height and forefoot shape measurements had moderate reliability only (ICC = 0.74, Gwet’s AC1 = 0.68 respectively). For rater 2, intra-rater reliability was good to excellent for all 13 measurements (ICCs/Gwet’s AC1 ≥ 0.83).

There were 6 measurements (foot length, diagonal foot width, Wejsflog Index, first toe height, fifth toe height and instep height) for both raters and 1 measurement for Rater 2 (horizontal foot width) that exhibited narrow LOAs ranging from − 5.7 mm to 8.1 mm. However, 5 measurements (ball of foot length, outside ball of foot length, heel width, ball girth, and instep girth) demonstrated wider LOAs ranging from − 13.5 mm to 16.3 mm. For forefoot shape, agreement ranged from 73 to 87%.

Table 3 summarises the inter-rater reproducibility of all 13 foot measurements. The inter-rater reliability was good to excellent for all 11 measurements, with ICCs ranging from 0.89 to 0.99. Fifth toe height and forefoot shape measurements had moderate reliability only (ICC = 0.73 and Gwet’s AC1 = 0.77, respectively).

Six measurements (foot length, diagonal foot width, horizontal foot width, Wejsflog Index, first toe height and fifth toe height) demonstrated narrow LOAs, ranging from − 4.4 mm to 5.8 mm. However, the remaining 5 measurements (ball of foot length, outside ball of foot length, ball girth, instep girth and instep height) displayed relatively wider LOAs ranging from − 18.8 mm to 12.7 mm. Percentage agreement for forefoot shape was 80%.