Study design
This was a randomised, single-blinded clinical trial with 1-year follow-up, performed in a community setting in Australia. Ethical approval was obtained from the local university Human Research Ethics Committee (PES/11/12/HREC). The trial was prospectively registered with the Australian New Zealand Clinical Trials Registry (ACTRN12612000993897).
Participants
Volunteers were included if they were aged 18-70 years and had a clinical diagnosis of LE, defined as pain over the lateral humeral epicondyle of at least six weeks’ duration provoked by palpation and resisted wrist/middle finger extension or gripping (11). In addition, participants needed to score at least 20/100 on the Patient Rated Tennis Elbow Evaluation (PRTEE) and be able to understand enough English to complete the outcome questionnaires. Exclusion criteria included any treatment for their elbow pain by a health care practitioner within the preceding three months, concomitant neck or other arm pain causing disability or requiring treatment within the last six months, clinical evidence of other primary sources of lateral elbow pain, upper limb fractures within the preceding 10 years, elbow surgery, systemic inflammatory disorder or malignancy, any contraindications to the study treatments, unresolved litigation or workers compensation claims, and pregnancy or breastfeeding.
Participants were recruited from September 2012 to June 2014, via referrals from health professionals and through local media, social and web-based advertising. Eligibility was initially assessed via a telephone screen followed by a clinical assessment by an experienced musculoskeletal physiotherapist. Volunteers meeting eligibility criteria gave informed written consent prior to enrolment by the trial administrator.
Participants were randomised to prolotherapy injections, manual therapy/exercise (physiotherapy) or a combination of both (prolotherapy+physiotherapy), using a computer-generated block randomisation schedule (N=6) generated and administered independently by the University Clinical Trials Centre. The trial administrator assigned the participants to their treatments and liaised with treating practitioners. Study personnel involved in participant screening, treatment and assessment were blind to group allocation throughout the full duration of the trial.
Sample size
An estimated 120 participants (40 per group) were required to detect a clinically important improvement of 13 points from baseline on the Patient Rated Tennis Elbow Evaluation (PRTEE; α=0.05, β=0.1) (12) and to detect a 20% difference between groups in the proportion of participants achieving ‘success’ according to the participant Global Impression of Change (GIC; α=0.05, β=0.2), assuming a success rate of 43% in the Combined group, and a 23% rate in the Physiotherapy group, and allowing for 10% loss to follow-up (5).
Treatment
All participants were provided with written educational material on their condition, with advice to use their affected arms but to avoid activities that resulted in increased pain for several minutes or more. They were encouraged to avoid use of non-trial treatments and asked to record them if they were used. In participants with bilateral LE, both elbows were treated, with the more severely affected side being the focus of outcome assessment and analysis to ensure that it met the eligibility criteria and to avoid a potential source of selection bias.
Physiotherapy: A standardised treatment protocol was implemented, based on a previously evaluated program that has demonstrated effectiveness [4, 5]. Four, 30-minute treatment sessions were provided at weekly intervals by a post-graduate trained musculoskeletal physiotherapist in a private practice setting. An evidence-based, pragmatic multimodal program, comprising education, manual therapy and therapeutic exercise, was used in conjunction with a home exercise program [4, 12]. Specific manual therapy techniques known as Mobilisation-With-Movement (MWM) were applied. In addition, three main groups of exercises were pragmatically prescribed: (a) Sensorimotor retraining of gripping and posture correction were commenced early in the physiotherapy intervention; (b) progressive resistance exercise for the wrist extensors were prescribed based on identified strength deficits; and (c) exercises geared towards general arm strengthening were also prescribed. The physiotherapist prescribed exercises based on the participant’s capabilities at each session to allow for optimal exercise volume and load setting. The overriding rule for all exercise was that pain should not be provoked during or after exercise. The physiotherapist reviewed the prescribed exercises at the commencement of each treatment session, and monitored adherence to the home program by reviewing a self-reported exercise diary completed by the participant each week.
Prolotherapy injection: The injection protocol for prolotherapy was based on one developed in the 1950s (13) and later refined (14), and which is commonly taught to practitioners in the USA (15). It was delivered at either a general practice or university-based health clinic by one of two general medical practitioners, each with more than 15 years’ experience in prolotherapy treatments. At each visit, the elbow was palpated for tenderness at points regarded as sources of pain in lateral epicondylalgia, i.e., over the lateral epicondyle, supracondylar ridge, radial head, lateral collateral and annular ligaments, and the common extensor tendon and musculotendinous junction. Each tender point was injected with 0.5 to 1.0 ml of solution containing 20% glucose and 0.4% lignocaine using a peppering technique with a 25-gauge needle. The total amount of solution injected depended on the number of tender points, but did not exceed 5 ml. Participants were advised to expect a temporary increase in pain for a few days following treatment, and to avoid anti-inflammatory medications during this period as they could theoretically reduce the effect of the injections. Non-prescription analgesics were permitted as required. The injections were repeated at 4, 8 and 12 weeks after the initial treatment session. Participants could exit this protocol early if there was either a full recovery or persistent worsening of elbow pain at any stage.
Combined treatment: This involved both protocols described above, but with the physiotherapy protocol timed for 1, 2, 3 and 5 weeks after the first prolotherapy treatment to minimise exacerbation of any post-injection soreness.
Outcome assessment
Baseline demographic and clinical characteristics included age, sex, body mass index, duration of current condition, affected side, hand dominance, occupation, current work status, income, physical activity status, current medications and smoking status. The primary outcomes assessed in this trial were the PRTEE and the participant's perceived Global Impression of Change (GIC). The PRTEE is a condition-specific self-reported questionnaire comprising five pain items and 10 functional disability items on 11-point numerical rating scales. The PRTEE has excellent test-retest reliability (r=0.93) and sensitivity to change (16-18), with a clinically important change of 11/100 or 37% from baseline (12). The GIC used a 6-point Likert scale ranging from ‘much worse’ to ‘completely recovered’. A dichotomous measure of success was defined as either 'much improved' or 'completely recovered' (5, 6).
Secondary outcomes included validated measures of (a) pain severity, recorded as ‘the level of pain you currently experience at rest’ and ‘the worst level of pain you have experienced in the past 7 days’, each using a 0 to10-point numerical rating scale (0 = no pain at all, 10 = worst pain imaginable) (19); (b) quality of life via the EuroQoL EQ-5D-3L scored using Australian weights (20, 21); and (c) pain-free grip strength (PFG) (22).
The use of medication and other not-per-protocol treatments related to the elbow pain were recorded at each follow-up assessment. The costs of these additional treatments, including costs for general practitioner, medical specialist, or allied health visits, aids and appliances and medications were estimated at contemporary market rates. These costs represented the costs to both the government and the participant. The cost of the trial treatments was calculated from the Australian Medicare Benefits Schedule rebates (23) and the schedule of fees from the local state workers compensation organisation (website accessed 1st September 2014).
Adverse events potentially related to treatment were recorded at all treatment visits and follow-up assessments. Recurrence of condition was defined as participants who moved from a self-reported ‘success’ on follow-up assessment up to 12 weeks, to a ‘non-success’ at 26 or 52 weeks’ follow-up (5, 6). Compliance with treatment in all groups was defined as a minimum of 75% attendance at treatment sessions; compliance with the exercise protocol was assessed by a questionnaire at each follow-up assessment and was defined as performing the exercises more than twice weekly for the first 12 weeks. All outcome measures were assessed face-to-face with a blinded assessor at baseline and at 6, 12, 26 and 52 weeks with the exception of the GIC, which was not assessed at baseline.
Data analyses
Demographic and clinical characteristics at baseline were compared between treatment groups to assess the effectiveness of the randomisation procedure. Analyses of outcome data were performed on an intention-to-treat basis by an experienced statistician who was blind to group allocation, using SPSS version 24 (IBM, Chicago, IL). The longitudinal outcome of the PRTEE was analysed using the Generalised Estimating Equation (GEE), with a first-order autoregressive relationship AR(1) working correlation structure to account for within-participant correlation for repeated measurements, and robust estimator for covariance matrix (24). The GEE is a widely used method for the analysis of longitudinal data. It considers measurements at multiple time points simultaneously and allows for testing the overall significance of the effects. With the GEE, a normal distribution with an identity link was used for scale variable outcomes, while a binomial distribution with a logit link was used for categorical binary variable outcome of success. The assumption of normality within the GEE framework was checked for scale variable outcomes. The effects of treatment, time, and treatment by time interaction were included in all models. The Wald χ2 test was used to assess between-group differences and within-group differences in outcomes over time. The GEE works well with missing data, assuming that they are missing completely at random (MCAR) so data imputation was not needed.
Subgroup analyses were performed assessing baseline demographic characteristics for their influence on treatment effects, and reporting adjusted results if they were found to significantly influence outcomes. Differences in medication use and use of other not-per-protocol treatments between groups were analysed using the Chi-squared test. Protocol and not-per-protocol treatment costs were calculated for each group and an analysis of the incremental cost-effectiveness ratio at the follow-up point of maximal differences in the proportion of responders between groups.