Significant improvements in pain after a six-week physiotherapist-led exercise and education intervention, in patients with osteoarthritis awaiting arthroplasty, in South Africa: a randomised controlled trial

It is proposed that combining first line management recommendations for OA of exercise and education could provide a patient centred approach to ongoing management while awaiting arthroplasty [18, 19]. While exercise interventions alone have shown no effects on pain or function in persons awaiting total knee replacement [20] and only low to moderate effects on those awaiting total hip replacement [20, 21], education and exercise delivered by programmes such as the ESCAPE programme [22] and/or arthritis self-management programme [23] have been reported to be an effective method of not only improving knowledge, but also prolonging the belief that exercise is beneficial and achieving desired changes in exercise habits [22]. Asevidence suggests, instead of merely educating on the importance of exercise, interventions should include both educational and active participatory exercise components [23, 24]. To further expand on this it is anticipated that by combining both aspects, the optimal benefits of these two management options could be achieved by allowing the person to incorporate new knowledge from the educational component into practice and implement the advised changes in beliefs and behaviours, such as exercise, in order to experience the benefits first hand [22‐27]. Previous studies have shown benefits from various exercise and education interventions in persons with hip and knee OA; ranging from improvements in pain [24, 28‐30], function [24, 28‐31], self-efficacy [28, 29, 32, 33], and exercise participation [28, 33]. Despite literature supporting the use of combined interventions in persons with OA, studies specific to those with severe OA and awaiting surgery for extended periods such as the population in SA, are scarce.

To the authors’ knowledge, no such studies have been performed in the SA population and the above mentioned studies performed in higher income countries with more accessible resources and shorter waiting times cannot be generalised to the SA population.

Therefore, in response to the increasing burden of OA and long waiting lists for hip and knee arthroplasty in the public health system in SA, research was warranted to establish the effects of a pre-operative exercise and education intervention on pain, for patients with OA awaiting arthroplasty. Additionally, we aimed to establish whether any changes in pain were supported by individual responses when interviewed about participating in the intervention…

A randomised controlled trial was performed and is reported according to CONSORT (Consolidation of Standards of Reporting Trials) guidelines.

The University of Cape Town Human Research and Ethics Committee granted approval for this study (Ref 378-2013, 492-2013). The clinical trial was registered with the Pan African Clinical Trial Registry (PACTR201409000885765, PACTR201507001186115).

The population of interest was patients diagnosed with OA in SA. The sampling frame included patients who had been placed on a waiting list to receive a hip/knee arthroplasty, specifically in the public health system inthe Western Cape and Gauteng provinces. Principle investigators were given access to waiting lists at Tygerberg and Helen Joseph Hospital. Participants were required to be on the waiting list for a minimumof three months to allow for necessary stabilisation to any new medications that may have been prescribed or adjustments to their treatment when the patient was first put onto the list. Inclusion criteria were as follows: a willingness to commit to the study, aged between 50 – 70 years, diagnosed with OA of the hip/knee, literate in English,Afrikaans, isiXhosa or isiZulu. Exclusion criteria were any cognitive impairment, as reported in the medical records, previous trauma/surgery to the unaffected leg, deemednot eligible for exercise as per the American College of Sports Medicine (ACSM) guidelines for exercise prescription. Reasons for exclusion according to theACSM included previous cardiac conditions or surgery,uncontrolled diabetes or asthma [34]. Additionally those who had previously taken part in a six-week programme aimed at improving self-efficacy and management were excluded.

Since pain was the primary outcome measure in this study; a change in pain severity was selected to determine the required sample size using Brief Pain Inventory (BPI) data from a previous study on pain in people with OA [35]. Calculations used a smallest meaningful difference in pain scores of 3 (on a scale of 0 to 10 on the BPI), and a standard deviation of 2.49 [35]. Statistical significance was accepted as p < 0.05 with a medium to large effect size (ES > 0.5) accepted as meaningful. It was calculated that samples of 26, 36, and 45 participants in total would provide 80 %, 90 % and 95 % statistical power for change in pain respectively.

To allow for attrition at recruitment or during the study, a higher target sample size was preferred and recruitment was aimed at a maximum of 48 participants at each site to make up 2 groups of 24 (a maximum of 96 participants across both sites).

The waiting lists at both facilities combined wereapproximately 4300 patients requiring arthroplasty for various reasons, including OA. Due to limited research personnel in the public health setting in SA and notbeing able to contact all patients, a random sample of 20 % of the total waiting list (n = 781) was drawn forlogistical reasons, using random number selection in Microsoft Excel. This was done to ensure patients had a fairer chance of being contacted, rather than contacting patients from the top of the list down. Diagnosis was made by the orthopaedic surgeon at the first clinic visit when the patient was placed on the waiting list from clinical history, physical examination and radiographic images according to American College of Rheumatology (ACR) clinical and radiological criteria. Those in the sample that met inclusion criteria were contacted telephonically by the primary investigators, to inform them of the nature of the study and to invite them to participate, if eligible. A large number of participants (n = 685) were unreachable and/or excluded from participating for various reason such as incorrect age/diagnosis other than OA according to the waiting list information, previous surgery, not interested, transport issues (Fig. 1). Thus 96 participants were invited to attend baseline testing, 21 did not arrive due to losing interest or transport issues and 1 was excluded due to incorrect diagnosis, therefore 74 arrived and were eligible to partake.

The first 10 participants’ baseline meeting was utilised to pilot the baseline procedures of providing all relevant information to the participants, obtaining informedconsent, completing demographic information and completing baseline outcome measures.

This took place at the setting intended for the study where outcome measures and the intervention took place, with the presence of the principle investigator for the information and informed consent signing and the blinded physiotherapist research assistant (RA) for completing the outcome measures, at each hospital. Piloting went smoothly as participants were able to complete outcome measures with the assistance of the RA without issues, it was noted that the time allocated to eachsession of data collection was sufficient and no changes were needed to the methods of collecting data for theprimary outcome measure. Thereafter, the remaining 64 participants gave written informed consent anddemographic information (including age, gender, educational level, employment status, joints affected, BMI, co-morbidities and waiting time for surgery) was recorded with the help of the research assistant. The participants completed the outcome measures, assisted by the research assistant with a translator where necessary.

The participants were stratified, according to OA of the hip or knee and then randomly assigned into either the experimental or control group, by means of random number allocation. This was performed by the twoprimary investigators. Those selected for the control group (n = 39) were contacted telephonically in the week that followed baseline assessment and were instructed to continue receiving their usual care, determined by their primary doctor, while awaiting surgery. The experimental group (n = 35) were contacted telephonically and were requested to begin the six-week education andexercise intervention commencing the following week (Fig. 1). Follow up measures were taken at week sixfollowing the intervention, and again at week 12 and month six, which concluded the study (Fig. 2).

It was proposed that this intervention has the potential to manage, yet probably not cure pain in this chronic condition, as it is based on the International Classification of Functioning, Disability and Health’s (ICF) understanding of health [26, 33]. This framework targets physiologicalimpairments (pain, muscle weakness, joint stiffness, and low fitness levels), personal factors (knowledge, health beliefs, self-efficacy, and stress management) and behaviours during daily activities (avoidance of movement or exercise, eating and sleeping habits, goal setting) [26, 33]. The intervention comprised of a six week physiotherapist-led education and exercise programme which was developed from previously used programmes [25, 33, 35‐38].

This multi-faceted approach addresses the participant holistically by means of education (including pain neuroscience education), self-management strategies and an active exercise component. The intervention approach and content allows for accommodation of participants with OA of both hip and knee. The intervention took place in an out-patient setting at both hospitals and wasled by the principle investigators, two qualified physiotherapists who had been trained in the respected field and who could deliver the programme in both English andAfrikaans. IsiXhosa and isiZulu translators were present for intervention sessions. The same physiotherapists led the six sessions to ensure content and delivery of the intervention remained constant. It took place for a maximum of two hours, on a weekly basis for six successive weeks, in groups of no more than 12 participants. Each two hour session included an educational component of approximately an hour, an exercise component of 20 to 30 min and a relaxation session of 10 min.

The educational topics covered are outlined in Table 1. The model of delivery of this intervention is based on the principles of social learning and cognitive behaviouralapproaches [25, 27, 32, 39, 40]. The educational component was aimed at increasing knowledge and understanding of OA, pain neuroscience, activity and related topics affected by their condition. Important topics such as self-management skills, problem solving, goal setting, coping mechanisms, stress management, and pacing were discussed to enable the participant in self-management [27]. Each participant received a “Living with Osteoarthritis” workbook (in their preferred language) on the content of each educational session, which they were encouraged to work through each week [41].

The exercise component allowed participants to apply what was learnt into actual behaviour changes [27]. It comprised of various stretching, light aerobic exercise and different lower limb muscle group strengtheningexercises [31]. Participants were required to set exercise and activity goals on a weekly basis and to record thesein their workbooks. The exercise component commenced at low repetitions and intensity and was progressed weekly from 20 min in duration, increasing time by 10 % andintensity as appropriate, depending on each participant’s individual ability. As the group size was small, safetyduring exercise and appropriate progression was possible despite being in a group [23]. The detailed exercise programme is described in detail (see Table 1). The intervention concluded with a relaxation session led by the physiotherapist facilitating various relaxation visualisations, an example of which is included in the workbook. Although relaxation techniques are not included in the recommended management of OA, it is widely used in chronic pain conditions and is seen as a helpful self-management skill [42, 43].

Attendance for the intervention was recorded and monitored weekly. If a participant was absent for asession of the intervention, the participant was contacted telephonically and asked the reason for absenteeism and encouraged to attend the next session. The intervention took place as described above for six weeks.

The primary outcome measure was pain as measured by BPI [44]. The BPI is a short self-administered assessment tool for measuring the severity of pain and the effects thereof (out of 10) and is widely used in both clinical and research environments [45]. To address shortcomings found with other pain scales that neglect pain’s characteristic as a variable factor, the BPI evaluates pain severity on the person’s “worst”, “least”, “average” and “now” pain over the last 24 h instead of one score forintensity of pain during different tasks. A pain severity score is calculated as an average of these four scores.Pain interference is scored by evaluating the effect of pain on an activity sub-dimension (walking, generalactivity and work) and four other aspects of life (relationships with others, enjoyment of life, mood, forming the affective sub dimension, and sleep) [45] (see Additional file 1 for the BPI). These two dimensions (severity and functioning) are the first two dimensions, described by the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) panel, which should be included in all research dealing with chronic pain evaluation [46, 47].

Psychometric testing has been performed on the BPI’s use in OA [35, 48] and its’ use in the assessment of pain to distinctly depict pain intensity and its effects on function is supported [49‐51]. The BPI has high levels of testre-test reliability [35, 52‐54] and internal consistency in various languages and does not display cultural bias [55, 56]. Of particular interest to this study, the BPI has been used in studies of OA which reported good internal consistencies; particularly for the subscale of pain interference [57, 58].

The BPI has previously been translated into Afrikaans [59] as well as isiXhosa [40]. The IMMPACT panel has specifically singled out the pain interference scale from the BPI as a recommended scale to use for assessment of pain-related functional impairment [60].

Besides the BPI, secondary measures of disability [Health Assessment Questionnaire (HAQ)] [61], function (The Physical Performance Task Battery) [62], self-efficacy (The Self-Efficacy for Managing Chronic Disease 6-Item Scale) [63] and health related quality of life (EQ-5D) [64] were taken at all follow up intervals (Fig. 2).

In addition, the experimental group were asked to complete an open ended questionnaire at the six month follow up meeting. A simple questionnaire was constructed by the principle investigators to include seven open ended questions about the six-week intervention and workbook used. This questionnaire did not go through any process of validation. The questions are included in the Additional file 1. This was to gain some insight into the participants’ personal thoughts or feelings around the intervention and the individual effects thereof.

The primary outcome measure of pain severity according to the BPI at baseline showed that data werenormally distributed [Week 0 BPI severity: N = 74, K-S p value > 0.20 (K-S p > 0.05)]. Therefore parametric statistics were used in the numerical analysis and Pearson Chi-squared (χ ²) calculations were used for categorical data. However, waiting time was not normally distributed therefore non-parametric analysis was conducted and the Mann–Whitney U test was used for thesecalculations. Data was analysed using Statistica software(StatSoft, Inc. 2004. STATISTICA, Data Analysis Software System, Version 10. www.​statsoft.​com).

Statistical significance for the two main effects of groupand time, and the interaction (group x time) for the BPI was assessed using a two-way analysis of variance (ANOVA) with repeated measures. Tukey’s post hoc comparisons were performed where necessary to determine significance between groups at different time periods. Analysis was by intention to treat. Missingdata was managed by carrying forward the last observed measurement. All numerical data was presented as the mean (M) ± standard deviation (SD). Statistical significance was accepted as p < 0.05. Effect sizes (ES) werecalculated and reported on with 95 % CI where appropriate when there were significant differences between groups. Effect size was calculated as Cohen’s d [65] using the following formula:

Finally, a descriptive thematic analysis using an inductive approach was used to identify common themes from the answers to the open-ended questions. These themes were created to foster better understanding of theresponses to the questions related to the intervention.

At baseline, the pain severity scores demonstratedmoderate to severe pain in the whole sample (M = 6.53, SD = 2.29) (Table 2). Those who participated in the intervention group (IG) showed significant improvements in pain severity when compared to the control group (CG)with moderate to large effect sizes [current effect: F(3, 216) = 8.904, p < 0.01]. This was seen at week 6 by between-group differences of: M = 2.44, SD = 2.24, p < 0.01, ES = 0.94, 95 % CI (0.45, 1.41)] and at month 6: M = 2.24, SD = 2.39, p = 0.02, ES = 0.74, 95 % CI (0.26, 1.2) (Table 2 and Fig. 3).

Similarly, at baseline the sample had moderate to severe scores for pain interference (M = 6.23, SD = 2.25) (Table 2). Likewise, those who participated in the intervention group showed significant improvements in pain interference with moderate to large effect sizes when compared to the control group [Current effect: F(3, 216) = 6.85, p < 0.01]. This was observed at all intervals by mean between-group differences: week 6 M = 2.95, SD = 2.23, p < 0.01, ES = 1.2, 95 % CI (0.7, 1.69)], week 12 M = 2.03, SD = 2.41, p = 0.04,ES = 0.68, 95 % CI (0.2, 1.14)] and month 6 M = 2.69, SD = 2.18, p < 0.01, ES = 0.98, 95 % CI (0.49, 1.45) (Table 2 and Fig. 4).

Responses from 30 participants of the experimental group were collected as five did not complete this questionnaire. Themes were identified by analysing answers to questions one, two and five. The four themes that emerged were: pain relief, increased knowledge, improvement in function/activity, and personal/social benefits. Since this article’s main focus is on the effects of the intervention on pain, the answers relating to pain were analysed. Table 3 presents the answers given by participants showing the benefits experienced related to their pain.

Adverse responses were also noted in responses. One participant stated that although the intervention had been very helpful, the effect on her pain did not last. Another participant responded that accessibility had been a problem as elevators were not always operating well and wheelchairs were not readily available. The answers to the other four questions have not been included in this article as the majority of participants agreed in their answers in that there was nothing they did not like or nothing else that needed to be added to the intervention or workbook.

Self efficacy at baseline showed a mean score of 5.88 out of 10, SD = 2.38. Those who participated in the intervention demonstrated a significant improvement in SE whencompared to the control group (Current effect: F(3, 216) = 4.37, p < 0.01). This difference was observed at week six by between group differences of M = 1.72, SD = 1.51, p = 0.03, ES = 0.76, 95 % CI (0.28, 1.22) (Table 4). No other significant differences were noted at other time intervals.

Health related quality of life of the sample at baseline was low (M = 0.37, SD = 0.33). The intervention group showed a significant improvement in HRQoL when compared to the control group (current effect: F(3, 216) = 4.45, p < 0.01). This difference was seen at week 12 by between groupdifferences of M = 0.26, SD = 0.58 p = 0.03, ES = 0.71, 95 % CI (0.24, 1.18) (Table 4). No significant differences wererecorded for EQ5D health state at any time point (Current effect: F(3, 216) = 2.47, p > 0.05).

Disability according to the HAQ disability index (DI) at baseline illustrated a moderate level of disability (M = 1.12, SD = 0.69). No significant differences were observed between groups for the DI during the study (Current effect: F(3, 216) = 8.51, p < 0.01) or for the HAQ pain visual analogue scale (VAS) during the study (Current effect: F(3, 120) = 5.97, p < 0.01). Additionally, there were no significant differences between groups in any of the tests from the physical performance task battery (normal walk, fastest 15 m test, six minute walk test, forward reach, upward reach, sock test and sit to stand). Further information on secondary measures is available from the authors.

Of the entire sample, 10 % were absent at week sixfollow up; six from the control group and two from the intervention group. At week 12, a higher attrition rate was seen with 23 % of the sample being absent for follow up measures; 12 from the control and six from theexperimental group. At month six, a further 2 % of the sample was absent at follow up measures; nine from the experimental group and ten from the control group. See Table 5 for details.

The results of this study presented here support the use of a six-weekphysiotherapist-led exercise and education programme to improve pain, in patients with osteoarthritis of the hip and/or knee awaiting arthroplasty. This is seen by clinically relevant findings in the experimental group with significant improvements in pain severity and interference with moderate to large effect sizes, when compared to the control group.

A moderate sample was recruited from two different regions of the country, servicing two populations with osteoarthritis. This allows for a limited degree of generalizability to the broader population of OA sufferersawaiting a joint replacement, in the SA public health system. However, it is important to note that over 90 % of those on the waiting lists were not included in the study for different reasons (Fig. 1). This is a potential source of recruitment bias. Those not included in the study for unspecified reasons may have been either (a) too ill to participate or (b) not experiencing sufficient pain to be motivated to participate. If the majority of unspecified non-participants fell into either of these categories, the results may be biased. Another aspect is that this sample had slightly low educational levels compared to those in upperincome countries; which is representative of the large portion of the previously disadvantaged population, who consequently also have low income levels and therefore attend public health facilities [79]. Caution should be applied as these findings might not be transferable to those in the private health sector who may have higher educational levels or wait shorterperiods for surgery.

Although the use of medication and percentage painrelief obtained from medication was recorded on the BPI, specific use of medication in terms of type, dose or how often medication was used was not recorded. This could be valuable data to determine whether the interventionaffected the use of medication or whether medication use influenced how participants responded to the intervention. It is recommended that future research into this field should consider medication use and effects thereof.

As Hurley and colleagues state, loss to follow up is a common problem during research [32]. This was not too concerning at the first follow up session as seen by only 10 % loss to follow up, however higher attrition rates at week 12 (23 %) and month 6 (25 %) should be considered during interpretation of the findings. To manage missing data, data was carried forward from the lastobserved measure. This may have led to blunting ofresults and therefore it is recommended that for future studies researchers explore a different approach such as “multilevel modeling and multiple imputation to generate robust predictions of the effect of missing data” [80]. Despite the risk of further loss to follow up, it is recommended that future trials extend the follow up period to at least 12 months as long lasting benefits are especially needed in persons on a long surgical waiting list [24].

Another study is currently underway to extend follow up duration which will provide valuable comparative data whereby the current findings can be compared with prior long term research concerning persons with OA awaiting arthroplasty as well as post-operatively. In addition to this, it is suggested that future researchevaluates whether such an intervention has differenteffects on those with OA of the knee or hip, and/or any effect on the need to still be on a waiting list for surgery.