Trial design
This was a single centre, pragmatic randomised clinical trial (RCT).
Participants
Eligible participants were women aged 18 years or over with a neurologist-confirmed diagnosis of chronic migraine and according to the criteria of the International Classification of Headache Disorders (ICHD-III). Participants were required to have had a minimum of two cycles of tertiary care treatment and be on a consistent 'care as usual' regime to create a more homogenous group.
Exclusion criteria were: Currently having or had manual therapy for neck, shoulder or upper back in the last six weeks. A new patient without any existing management by neurologist, having a condition contraindicated for manual therapy including but not limited to inflammatory disorders, severe osteoporosis and tumours. Identification of any medical ‘red flags’ by the neurologist.
Study Setting
The study took place in a UK NHS tertiary care acute neurology unit and was given approval by the UK Heath Research Authority (IRAS 228901), Bournemouth University ethics panel, and R&D at the NHS Foundation Trust in June 2018. The Declaration of Helsinki was adhered to during this study. All data were anonymised and participants provided written informed consent.
Interventions
Participants were randomised to manual therapy group (Group M) or a care as usual group (Group C). Both groups had Botox TM (care as usual) in week 1 and again at week 12. The standard chronic migraine protocol was followed (PREEMPT). In addition, those individuals assigned to the manual therapy group were given a 12-week manual therapy treatment plan. Each session comprised 20–25 minutes manual therapy. The manual therapy session schedule was chosen to reflect a typical approach in practice and took place in weeks 1, 2, 5, 8 and 12.
The definition of manual therapy was defined as a hands-on approach utilising mobilisation (low velocity low amplitude, LVLA), manipulation (high velocity low amplitude, HVLA) and soft tissue work, singly or in combination [59][60][61]. This is distinct from physical therapy which can include manual therapy but also extends to include exercise; the use of equipment and, acupuncture.
Prior to each session participants were asked how they felt after the previous session and details of any adverse events logged.
Manual therapy techniques were performed on the upper body defined as above T12 spinal level, with a focus on the shoulder girdle, spine and head, including the temporomandibular joints. A pragmatic approach to the manual intervention was adopted whereby the PI used manual interventions deemed clinically appropriate, after an initial assessment of, and consent from, the participant at each session.
Baseline data was collected from both groups at the start of the study with final primary and secondary data collected at 12 weeks from all participants.
Outcome Measures
The primary outcome measurement instrument was the Headache Impact Test 6, recommended by the IHS as one of the most valid measures in chronic migraine studies [56]. It comprises six self-reported questions using a 5 item Likert scale to determine the level of headache-related disability. The total score ranges from 36 to 78. Anything above 60 is classed as very severe impact on life. The HIT6 has shown good reliability across studies with Cronbach’s α between 0.75–0.92 and ICC 0.76 to 0.80. The minimum clinically important difference (MCID) for between-group difference in HIT6 change scores is 2.3 units, with the within-person MCID estimated at a reduction of 3.7 to 5 from baseline [62][63][64].
The Hospital Anxiety and Depression scale (HADS) was used to measure levels of anxiety and depression with the total score reflecting the level of psychological distress on a pro-rata basis to the individual scores hence 0‐14 is considered normal and above 21 abnormal emotional distress/instability [65][66][67]. However, to date, no measures of its responsiveness to change as determined by the MCID have been calculated for headaches.
Cutaneous allodynia, a marker for central sensitisation was measured using the Allodynia Score Checklist (ASC). It is a validated easy to use, self-reported measure to measure CA in migraine. Comprising 12 statements using a 5 item Likert scale. Summation scores are categorised as none, mild, moderate and severe allodynia. Cronbach’s alpha (scale reliability) has been calculated at between 0.76 and 0.8 [68][69]. The ASC has been used in studies of central sensitisation and in particular with migraine and chronic migraine [70][71][72].
The Patient Global Impression of Change (PGIC) [73] is a validated instrument for examining the patient's perception of how they have improved after an intervention. It has been used in a wide range of chronic pain studies, is recommended by IMMPACT [74] for clinical trials in chronic pain and is recommended in the IHS chronic migraine guidelines as a secondary outcome [56]. It has seven statements, each of which uses a 7 item Likert scale to describe how the patient felt after treatment. A score of > = 5 is considered a significant change.
A weekly diary was also used to collect information on medication use, headache quantity and severity and allodynia. To reduce the likelihood of participants recognising the measurement instruments from previous experience, particularly HIT6 and HADS, and completing them based on previous experience with changes to migraine management [75], the measurement instruments were not identified individually but supplied as tables of questions in a professionally designed questionnaire booklet. Text and email reminders to complete the diary were sent each week to all participants in the study.
Sample Size Calculation
A pooled standard deviation for the Headache Impact Test (HIT6) was calculated from the analysis of eight major studies in chronic migraine involving 4400 participants and 17 measures of standard deviation which gave a pooled SD of 4.7. On this basis the study required 29 participants in each arm to ensure that a two-sided test with α = 0.05 has 80% power to detect a mean difference of 3.5 points in HIT6 before and after treatment. The project aimed to recruit a maximum of 32 participants into each group to allow for 10% dropout.
Randomisation Sequence Generation
This study employed a randomisation software tool, “Research Randomizer” [57] to create a randomised assignment sequence
Randomisation: Allocation Concealment Mechanism
Pieces of paper with the group allocation were placed in sealed opaque envelopes by the independent research assistant (nurse) and given to the Principal Investigator (PI; an experienced chiropractor with over 15 years’ experience and postgraduate education and training in headache management, and an author and teacher of soft tissue techniques [58].
Randomisation: Implementation
The allocation sequence was conducted by an independent research assistant, unattached to the project. The envelopes were opened at the first appointment, in front of the participants, who were then informed of their group allocation.
Blinding
The neurology team were blinded to the group allocations. Although it was not possible to blind the PI to the manual therapy group, the PI was blinded to the end of study survey outcomes before analysis. Another member of the research team re-coded the participant reference numbers and stored the key file on a password-protected computer.
Statistical Methods
SPSS™ v.25(IBM SPSS Inc IL USA) and Excel (™, Microsoft) were used for analysis of data.
The primary analysis used the Intention to Treat (ITT) approach. Secondary endpoint/outcomes from the baseline and final questionnaires were analysed on ITT [76].
The outcomes based on the diary data used a modified ITT (mITT) basis which dealt with missing data through the use of multiple imputation [77]. Any participant with less than 6 weeks of diary data was excluded from this part of the analysis. Baseline differences between Groups C and M were identified using means, standard deviations, medians, Student’s T-tests and chi square as appropriate. Significance testing was set at α = 0.05 with post hoc analyses using the Bonferroni correction. Measures of difference in change was by ANOVA, mixed multivariate linear regression with ordinal logistic regression to assess the secondary endpoints relationship with changes in primary end point, while adjusting for baseline parameter levels. Cohen’s d and Eta-squared when required were used to calculate effect sizes: small (0.2–0.5), medium (0.5–0.8) or large (> 0.8) [78][79]) The assumption of homoscedasticity used Levene’s test and for sphericity, Mauchly’s test. Students T, ANCOVA and Mann–Whitney U were used for non-normally distributed and ordinal data.