This pilot study was an unblinded (participants and researchers), single-center, 2-arm, randomized (1:1), parallel-group study conducted in Dunedin, New Zealand. Ethical approval was granted by the New Zealand Upper South A Regional Ethics Committee (reference URA/11/05/012) and registered in the Australian New Zealand Clinical Trials Registry (#ACTRN12614000898651), registered 25th August, 2014.
Recruitment took place within the community in Dunedin, New Zealand by referral from Leading Maternity Carers (midwives) and advertisements placed in a local newspaper. Women were eligible to take part in the study if they were pregnant, at least 18 years of age, had experienced pubic symphyseal pain for at least two weeks (which was worse than any concurrent posterior pelvic pain), and had a positive response to at least two of three clinical tests: reproduction of pain from palpation , modified Trendelenburg’s test , active straight leg raise test [12,14,25] (as described in Additional file 1). Women with a known high-risk pregnancy, a history of major systemic bone disease and/or back or pelvic injury, a medical condition which contraindicated the use of a pelvic support belt (e.g. certain types of placenta previa), or those currently taking steroid medication were excluded. Figure 1 shows a flow diagram of the recruitment process.
After obtaining informed, written consent, participants completed a standardized baseline questionnaire (Additional file 2) which included pain history, a 10 cm Visual Analogue Scale (VAS) to quantify pain  intensity over the preceding 24 hours and the preceding week, and the Modified Oswestry Disability Questionnaire (MODQ) to determine the influence of symphyseal pain on activities of daily living [29,30]. The Patient Specific Functional Scale (PSFS)  was then completed in conjunction with the assessing physiotherapist (SW). Participants were tested for joint hypermobility using the nine-point Beighton Hypermobility Score (adapted from ) in which hypermobility was defined as a score of ≥4 points .
Using a computer-generated random number table and block sizes of 4 (produced by MS and concealed in sealed opaque envelopes, sequentially numbered), participants were assigned in a 1:1 allocation to either flexible or rigid belt. Following assessment and enrolment into the study, NF drew and opened the next sequentially numbered envelope and then communicated with SW about which belt each participant would be using.
Intervention: Women were randomized to three weeks wear of either a wide, flexible neoprene support belt (Smiley Belt, www.smileybelt.co.nz, Havelock North, New Zealand; NZD $58.00) or a thinner, more rigid belt made of nylon webbing with foam lining (LC symphysis pubis belt, The Orthotic Center New Zealand Limited, Greenlane, Auckland, New Zealand; NZD $21.30). Participants were shown by a physiotherapist (SW) how to wear the belt, aligned over the pubic symphysis (the so-called ‘low position’ , Figure 2) and were advised to wear it whenever possible during waking hours.
Women were sent automated standardized daily text messages and asked to respond on the number of hours the belt had been worn, whether pain had decreased (yes/no/sometimes) and if functional activities were easier to perform (yes/no/sometimes). Responses to daily text messages were recorded. Weekly phone interviews were conducted (NF) to complete the PSFS and determine adherence (frequency and duration of use) and tolerance (comfort) of the belt.
At the conclusion of the intervention (three weeks), the un-blinded researcher (NF) met with participants to reassess the PFSF, and each woman self-completed the MODQ and VAS. Participants were then fitted with the alternate belt to wear for one week. Following this (week 4), a final phone interview was conducted which included the same questions as weeks 1 and 2, respectively, and participants were asked about belt preference (the proforma for the phone interviews is shown in Additional file 3).
The primary outcome measure was the PSFS and the primary endpoint was three weeks. The PSFS has been shown to be a reliable and valid test for assessing disability and change in disability, and a 2-point difference in scores may be considered a minimal clinically important change . For each group, mean scores for all outcome measures were calculated at baseline and at the end of week 3. Data were then transformed as described by Westaway et al. . Secondary outcome measures were pain intensity rated on the 10 cm VAS for worst pain over the preceding 24 hours and the preceding week, and functional status determined by the MODQ score, interpreted using the specified guidelines .
Before a multi-center trial of pelvic support belts can be launched, it is essential to gather data on the most appropriate type of belt, symptomatic effectiveness, comfort, and adherence as well as to provide estimates of anticipated variability and correlations between repeated measures. Having 10 participants in each arm of this study was considered sufficient to obtain preliminary data to help with estimating power for a larger subsequent study.
Data were entered into a spreadsheet. Appropriate descriptive statistics were derived for all variables. Linear regression models were used to examine differences in follow-up values between the two groups (flexible and rigid belts) after adjusting for baseline values. Model assumptions were assessed using histograms of residuals, plots of residuals against fitted values, and Levene’s test for equality of variance between groups. Where model residuals were positively skewed or demonstrated heteroscedasticity, natural logarithmic transformations were investigated, after adding one in the case of measures that included zero values. Overall changes (across both groups) in measures were examined using paired t-tests where there was no evidence of differences in change between groups and regression models using follow-up values while controlling for baseline values were used to assess any association with compliance. Stata 13.1 was used for all statistical analysis and all tests were performed using two-sided p < 0.05 as indicating statistical significance.